Traffic Management System in Smart Cities to Identify the Vehicle and Detect the Speed of Vehicle to Optimize Traffic Management Strategies

Traffic Management System in Smart Cities to Identify the Vehicle and Detect the Speed of Vehicle to Optimize Traffic Management Strategies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Traffic Management System in Smart Cities to Identify the Vehicle and Detect the Speed of Vehicle to Optimize Traffic Management Strategies Venkateswara Rao Gorle, Sheik. Hidayatulla Shariff, Rajendra Kagithapu, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7476715/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The importance of intelligent vehicle recognition and counting has significantly increased in the context of contemporary highway administration. However, the precision of vehicle counts is directly challenged by the fact that cars vary in size, making it difficult to identify them accurately. This paper introduces a technique for identifying and counting vehicles using vision to tackle this problem. DeepSORT model was used, which is based on the You Only Look Once (YOLO v8) model, to recognize and track vehicles in real time within video sequences. By combining the sophisticated detection capabilities of YOLO v8 with the skilful tracking algorithms of DeepSORT, the suggested method aims to improve accuracy by taking into account the intricacies of various vehicle sizes and movement patterns. The system gives highway management real-time insights into vehicle traffic, enabling them to well-informed information to improve traffic control tactics. Furthermore, the system's capacity to smoothly adjust to intricate traffic situations is enhanced by the combination of YOLO v8 and DeepSORT. Figures Figure 1 Figure 2 Figure 3 Figure 4 1. INTRODUCTION In the past, precise vehicle counts, speed estimation, accident detection, and efficient surveillance were all difficult tasks for traditional traffic monitoring techniques. These techniques lack accuracy, efficiency and frequently involve manual intervention. The combination of cutting-edge deep learning models, such as YOLO v8 and Mobilenet SSD, which enable real-time CCTV footage processing, forms the core of the system. With a remarkable accuracy rate over 95%, these models ensure a full grasp of traffic dynamics, allowing for informed decision-making. YOLO's adoption of the Thonny Python IDE shows a dedication to reliable and effective development techniques. MobilenetSSD is implemented using Jupyter Notebook, demonstrating a commitment to transparency and cooperation in the search for innovative solutions. The field of vehicle detection and traffic analysis has witnessed significant advancements, with researchers employing state-of-the-art techniques to enhance accuracy, robustness, and efficiency [ 1 ]. A multi-level fusion network and labelling hierarchy for nighttime vehicle identification using highlight information which addresses the difficulties that arise in low-light environments [ 36 – 39 ]. An in-depth analysis of deep multi-modal object recognition and semantic segmentation, focusing on datasets, methods and problems in the field of autonomous driving [ 2 ]. [ 3 ] integration of radar and video data under adverse weather conditions is highlighted and insights into improving the resilience of autonomous vehicle systems are given. DLT-NET, a model that simultaneously identifies drivable areas, lane markings and traffic objects, demonstrating a cohesive methodology for a thorough understanding of the scene [ 4 ]. [ 5 ] improves pedestrian safety by predicting movements from a remote first-person perspective, thereby supporting intelligent driving systems. [ 6 ] presents a comprehensive analysis of the DATMO system and explains the detection and tracking of dynamic obstacles, an essential component of autonomous systems. An image processing-based framework for vehicle enumeration, speed assessment, and categorization, illustrating a comprehensive methodology for traffic analysis [ 7 ]. [ 8 ] the optimization of the YOLO v3 algorithm for effective traffic flow detection is highlighted, demonstrating the adaptability of YOLO in practical applications. [ 9 ] introduces the SSD model, a single-shot method for object identification, and explains its architecture and features. [ 10 ] presents Faster R-CNN, a model for real-time object identification that improves both speed and accuracy using Region Proposal Networks. [ 11 ] examines the implementation of an enhanced YOLO v3 network for vehicle detection in infrared aerial images is explored, highlighting the versatility of YOLO architectures. [ 12 ] presents YOLO v4, which focuses on achieving optimum speed and precision in object recognition and illustrates the continuous further development of the YOLO series. [ 13 ] introduces an automated traffic density estimation system that utilizes Single Shot Detection (SSD) and MobileNet SSD and provides insights into practical applications. [ 14 ] proposes an algorithm based on YOLO v8 for detecting vehicles in video streaming and demonstrates its effectiveness for real-time applications. [ 15 ] suggests a robust method for vehicle detection in aerial images is proposed that integrates bag-of-words and orientation-based scanning to improve accuracy. The researchers contribute to the advancement of vehicle detection, traffic analysis, and intelligent transportation systems and demonstrate progress in model architectures, optimization strategies, and real-world applications. 2. RELATED WORK This work includes a dynamic frontend interface created with HTML, CSS, and JavaScript to provide a smooth and intuitive experience. With the help of this interface's user-friendly dashboard, which is enhanced with charts, maps, and visualizations, stakeholders can easily understand traffic statistics. In order to facilitate historical analysis and reporting, a strong MySQL database is used on the backend to store and retrieve vital data, such as vehicle speeds and identifications. The development tools chosen for this work demonstrate a dedication to effectiveness and teamwork. For YOLO, the Thonny Python IDE ensures efficient development procedures, and for Mobilenet SSD, the Jupyter Notebook encourages open communication and transparency throughout the development process. The Smart City Traffic Management project is evidence of how technology may be used to solve intricate urban difficulties. The major advancement in the field of intelligent transportation systems since it emphasizes the use of state-of-the-art algorithms and approaches. 2.1 Object detection with deep learning The advancement of deep learning, particularly Convolutional Neural Networks (CNN), has significantly enhanced object identification. Ross B. Girshick et al. introduced RCNN [ 16 ] inspired by the efficacy of deep learning in categorisation. It integrated an object suggestion mechanism with a convolutional neural network classifier. Building upon RCNN, other advancements such as SPP [ 17 ], Fast RCNN [ 18 ], and Faster RCNN [ 19 ] were introduced to enhance performance in terms of both efficacy and speed. They included the proposal prediction into a network structured as a two-stage model. MSCNN [ 20 ] forecasted proposals at various stages during feature extraction and then used an object detection subnet to refine the results, hence enabling the receptive field of MSCNN to accommodate objects of diverse sizes. Another significant study is YOLO [ 21 ]. In contrast to earlier networks, they introduced a one-stage architecture that simultaneously executed proposal prediction and categorisation. This network can achieve a performance of roughly 40 frames per second. While these approaches may provide improved outcomes in general item identification, including cars, their efficacy is constrained when the visual characteristics of vehicles are diminished at night. During that period, the detection results of automobiles may be substantially enhanced by using critical information such as vehicle features. We include vehicle highlight saliency information into the detection framework, allowing the detection network to enhance vehicle feature representations, extract vehicle proposals, and identify cars with more accuracy. 3. CATALOGUING HIERARCHY Learning feature representations often encounters the issue of significant intra-class variability, which generates several ambiguities that perplex convolutional neural networks when they modify neurone weights to identify salient visual patterns. A prevalent approach involves designing networks capable of learning discriminative deep feature representations. Li et al. [22] introduced a hierarchical paradigm for the progressive learning of hidden representations. Szegedy et al. [23] included the "skip" layer to enhance feature propagation. DCE [24] integrated end-to-end learning with collaborative factor analysis to achieve optimum compatibility in representation learning and latent space discovery. The alternative approach is to use the label hierarchy. Li et al. [25] contended that pictures belonging to the same semantic category need to be embedded into the same latent representation subspace. Classifying sub-classes with enhanced visual consistency facilitates the learning process. Several studies have already investigated label hierarchies. HD-CNN [26] initially trained a coarse category classifier to segregate easy categories, and then classified the harder categories using a fine-grained category classifier. Xie et al. [27] offered two data augmentation approaches that need extra data to determine hyper-class of initial fine-grained labeled data. Lim et al. [28] introduced a feature termed “sketch token,” derived from supervised mid-level information represented by hand-drawn contours in pictures, thereafter clustering the feature to establish classes. Ohn-Bar et al. [29] and Xiang et al. [30,31] used supplementary 3D information in datasets for clustering to derive subclasses; nevertheless, the majority of datasets lack this prompt information. Kuo and Nevatia [32] clustered the HOG features after dimensionality reduction and classified the vehicle into subclasses with distinct orientations. Wang et al. [33] used semantic contexts, such as scene titles and label statistics of picture patches, to construct label hierarchies. The label subclasses automatically. The generation may include some noise; yet, [34,35] shown that labels with noise remain beneficial for the outcomes. In contrast to other studies, we establish the label hierarchy based on the unique characteristics of nocturnal vehicles. We meticulously extract concealed information from the vehicle dataset and categorise the cars in the training set into distinct subclasses based on vehicle highlight information, so establishing the label hierarchy. 3.1 YOLO v8: The most recent version of the You Only Look Once (YOLO) model, YOLO v8, created by Ultralytics, is well-known for its real-time object identification and image segmentation skills. A state-of-the-art computer vision model that blends accuracy and speed is called YOLO v8. By adding new features and enhancements, it builds on the popularity of earlier YOLO editions. Numerous visual AI tasks are supported by YOLO v8, such as object identification, segmentation, pose estimation, tracking, and classification. Key Features: Performance: YOLO v8 achieves state-of-the-art performance due to advancements in deep learning and computer vision. Adaptability: It’s suitable for various applications and can run on different hardware platforms, from edge devices to cloud APIs. Unified Framework: YOLO v8 provides a unified framework for training models across multiple tasks. GMM (Gaussian Mixture Model): Mixture of Gaussians: GMMs represent normally distributed subpopulations within an overall population. Probabilistic Clustering: Unlike hard clustering (where each point belongs to a single cluster), GMMs allow soft clustering, where data points can partially belong to multiple clusters. Parameter Estimation: GMMs estimate parameters such as means, variances, and mixing coefficients. Mathematical Formulation: A GMM is represented as a weighted sum of Gaussian component densities. For a multivariate Gaussian distribution, the probability density function is given by: Here, πl represents the mixing coefficient for the l th Gaussian component. μl is the mean vector, and Σl is the covariance matrix for the l th Gaussian. Mobile Net SSD : MobileNet SSD V2 (MobileNet Single Shot Detector) is an object detection model designed for real-time inference on devices like smartphones. It combines the power of the Mobile Net V2 base network with a Single Shot Detector (SSD) layer. Real-time Performance: Achieves good real-time results even on limited compute resources (30 frames per second). Compact Size: Once trained, MobileNet SSD V2 can be stored with just 63 MB, making it ideal for smaller devices. Architecture: Base Network: The first part consists of the MobileNetV2 network, which acts as a feature extractor. SSD Layer: The SSD layer classifies detected objects based on the features extracted by MobileNetV2. The combination of these two parts enables efficient and accurate object detection. Performance: MobileNet V2 outperforms its predecessor, MobileNet V1, with higher accuracies and lower latencies. It’s optimized for mobile devices, making it suitable for applications like image recognition, tracking, and more. Training and Deployment: You can train MobileNet SSD V2 on your custom dataset using Tensor Flow 4. APPROACH DESCRIPTION Collection of CCTV film : Compile a varied dataset of CCTV film showing various urban traffic situations. Variations in vehicle kinds, traffic density, and illumination conditions should all be included in this dataset. Definition of the Problem : Specify your goals The objectives of the traffic management system, such as precise vehicle counts, speed estimation, and vehicle type categorization, should be stated clearly. Choosing a Model : Deep Learning Model Selection: Pick suitable deep learning models for categorization, speed estimation, and object identification. Models such as YOLO, MobilenetSSD, or a mix of models designed for particular activities may be taken into consideration. Preprocessing Data : Image preprocessing: To improve the model's capacity to generalize under various circumstances, prepare the dataset by resizing, standardizing, and enhancing images. Instruction : Model Training : Use the pre-processed dataset to train the chosen deep learning models. To attain high accuracy in vehicle detection, speed estimates, and classification, adjust loss functions, hyper parameters, and other parameters. Analysis in real time : Real-time analysis implementation: Use the trained models to examine CCTV material instantly. In order to process incoming video streams continually, the models must be integrated into the system. Front-end and back-end integration : Development of Frontends : To visualize real-time traffic data, create an easy-to-use frontend interface with HTML, CSS, JavaScript, and PHP. created an intuitive dashboard with maps, charts, and visualizations to efficiently communicate information. Backend Implementation : Create a solid backend with MySQL to store and retrieve vital data, like vehicle identity and speed, allowing for reporting and historical analysis. 5. DATA EXPLORATION COCO dataset : A large-scale image recognition dataset for object detection, segmentation, and captioning applications is called COCO (Common Objects in Context). Each of the more than 330,000 photos in it has five subtitles that describe the setting and 80 object types. Numerous cutting-edge object detection and segmentation models have been trained and evaluated using the COCO dataset, which is extensively utilized in computer vision research. In machine learning, COCO dataset is frequently utilized for both research and real-world applications. The photos and their annotations make up the two primary components of the collection. A hierarchy of folders is used to arrange the photos, with subdirectories for the train, validation, and test sets located in the top-level directory. Each file corresponds to a single image, and the annotations are supplied in JSON format. Data Manipulation : Data manipulation is the process of altering or modifying data to make it easier to deal with by making it more comprehensible and structured. Only when you have the data to do so can you manipulate it. As a result, you require a database that is created from multiple data sources. Data manipulation aids in information cleaning. The database's data must be restructured and reorganized in order to complete this task. The raw video data is pre-processed and frame separated in the first step of data manipulation in order to get it ready for further analysis. In order to ensure that only pertinent frames with vehicles are kept for additional processing, preprocessing include cleaning and arranging the data to eliminate noise and unnecessary information. After that, frame separation is done to separate frames from the video stream, making it possible to analyse each frame effectively on its own. Data Preparation : One of the crucial things we must perform before analysing the data is data preparation. This procedure involves purifying the dataset's raw data prior to processing and analysis. In order to refine the data, it also entails reformatting, fixing errors, and combining the data sets. Blurry Videos : Blurry videos can significantly affect object detection accuracy. Blurriness maybe caused by camera motion, poor focus, or low frame rates, making it challenging to identify and track objects reliably. Solution : Implement image stabilization techniques to reduce the impact of camera shake. Detects and filters out excessively blurry frames to improve overall detection accuracy. Low-Light Videos : Because of the decreased visibility and increased noise in low-light (night time) recordings, it can be difficult to detect and track vehicles. Solution : To get better evening photos, use thermal or infrared (IR) cameras. To increase visibility, use sophisticated picture enhancing techniques. Make use of object identification models that have been specially trained for low light levels. 6. DATA ANALYSIS Classes in COCO Dataset A large-scale image recognition dataset for object detection, segmentation, and captioning applications is called COCO (Common Objects in Context). Each of the more than 330,000 photos in it has five subtitles that describe the setting and 80 object types. Numerous cutting-edge object detection and segmentation models have been trained and evaluated using the COCO dataset, which is extensively utilized in computer vision research. Table 1 COCO dataset to object identification evaluation. Person Fire hydrant elephant Skis Wine glass broccoli Dining table Toaster Bicycle Stop sign bear Snow board Cup Carrot toilet sink Car Parking meter zebra Sports ball Fork Hot dog tv Refrigerator Motorcycle bench giraffe Kite Knife Pizza Laptop Book Airplane bird Backpack Baseball bat Spoon Donut Mouse Clock Bus cat Umbrella Baseball glove Bowl Cake Remote Vase Train dog Handbag Skate board Banana Chain Keyboard Scissors Truck horse Tie Surf board Apple Couch Cell phone Teddy bear Boat sheep Suitcase Tennis racket Sandwich Potted plant Microwave Hair drier Traffic light cow frisbee bottle orange bed Oven toothbrush Distribution of classes in coco dataset : We must forecast the bounding boxes and the labels that go with them in order to detect objects. It's crucial to remember that the class imbalance in the COCO dataset causes inherent bias. 7. MODELLING Model Development Object Detection: • YOLO Model: To detect objects, the model loads a pre-trained YOLO model ('yolov8s.pt') using the ultralytics library. • Input: The input stream is a live video recording. • Output: It gives the output of the bounding boxes that were found around the cars. Tracking: • Object tracking: To track cars between frames, a bespoke object tracking algorithm (track.py) was implemented. • The tracking algorithm makes use of centroid tracking, which maintains object IDs by calculating and matching the centroids of bounding boxes between successive frames. • Input: Bounding boxes were found. • Output: Vehicle IDs and matching bounding boxes were tracked. Database integration: • MySQL Integration: Creates a connection to a MySQL database in order to record data about identified objects, such as lane, speed, and ID. • Input: Information about detected objects. • Result: Information was saved in the MySQL database. Model Evaluation Initialization: Establishes connections, loads the YOLO model, and initializes variables. Loop for video processing: • Resizes frames for processing; • Reads frames from the input video. • Uses the YOLO model to identify automobiles. • Follows identified cars between frames. • Determines vehicle speeds and counts those that cross specified lines. • Annotates frames with text, lane lines, and bounding boxes. • Adds identified object information to the database. Termination: When all frames have been processed or a user interrupts, resources are released and connections are closed. Real-time lane counting, speed calculation, tracking, and vehicle detection are all features of the implemented system. It interfaces with a MySQL database for data persistence, uses YOLO for object recognition, and employs unique tracking techniques to preserve object IDs. For a number of uses, including automated toll collecting, traffic monitoring, and congestion analysis, this system can be further enhanced and expanded. Table. 2: Detected data with objects and speed Traffic Information ID Speed (km/hr) 0 239 15 236 15 115 15 75 15 220 15 56 15 43 15 36 15 30 15 27 2 215 15 24 15 21 4 211 15 19 15 18 8. RESULTS AND CONCLUSIONS Detected object information, including ID, speed, and lane information, is stored in a MySQL database. This allows for further analysis and retrieval of historical data. Bounding boxes around vehicles, designated lines for speed calculation and lane counting, and text annotations displaying lane-wise vehicle counts are visualized on the video frames. The system processes the input video stream in real-time, providing immediate feedback on vehicle detection, tracking, speed calculation, and lane counting. 9. CONCLUSION The Smart City Traffic Management project represents a pioneering effort in leveraging cutting-edge technologies, notably MobilenetSSD and YOLO, to address the multifaceted challenges of urban traffic congestion. Through the tracks detected vehicles across frames. seamless integration of these advanced deep learning models, the project has achieved remarkable success in real-time vehicle detection, speed estimation, and classification. The user-friendly frontend interface, coupled with robust backend systems, ensures accessibility and efficiency for a diverse range of stakeholders, from city authorities and transportation departments to individual commuters. As we conclude this endeavor, it is evident that the Smart City Traffic Management project not only meets the immediate objectives of enhanced traffic efficiency but also aligns with the broader goals of creating sustainable, connected, and livable urban environments. The continuous improvement ethos, coupled with adaptability to emerging technologies, positions this solution as a dynamic tool for ongoing advancements in the realms of urban planning, transportation, and smart city initiatives. This project serves as a testament to the potential of technology to reshape and elevate the urban living experience. Declarations Funding Declaration: This research did not receive any specific grant from funding agencies. Author Contribution A: Data Curation, Methodology, Writing-Original, Writing-Review & Editing.B: Conceptualization, Supervision, Methodology. C: Supervision, Conceptualization, Resources, Methodology.D: Draft, Formal analysis, Software.E: Formal analysis, Investigation, SoftwareF : Writing-Review & Editing, Validation, Software. References Y. Mo, G. Han, H. Zhang, X. Xu, and W. Qu, “Highlight-assisted nighttime vehicle detection using a multi-level fusion network and label hierarchy,” Neurocomputing, vol. 355, pp. 13 23, 2019. D. Feng, C. Haase-Schuetz, L. Rosenbaum et al., “Deep multi-modal object detection and semantic segmentation for autonomous driving: datasets, methods, and challenges,” IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 3, p. 3, 2019. Z. Liu, Y. Cai, H. Wang et al., “Robust target recognition and tracking of self-driving cars with radar and camera information fusion under severe weather conditions,” IEEE Transactions on Intelligent Transportation Systems, 2021. Y. Qian, J. M. Dolan, and M. Yang, “DLT-NET: joint detection of drivable areas, lane lines, and traffic objects,” IEEE Transactions on Intelligent Transportation Systems, vol. 21, no. 11, pp. 4670–4679, 2020. Y. Cai, L. Dai, H. Wang et al., “Pedestrian motion trajectory prediction in intelligent driving from far shot first-person perspective video,” IEEE Transactions on Intelligent Transportation Systems, pp. 1–16, 2021. Á. Llamazares, E. J. Molinos, and M. Ocaña, “Detection and tracking of moving obstacles (DATMO): a review,” Robotica, vol. 38, no. 5, pp. 761–774, 2020. C. Liu, D. Q. Huynh, Y. Sun, M. Reynolds, and S. Atkinson, “A vision-based pipeline for vehicle counting, speed estimation, and classification,” IEEE Transactions on Intelligent Transportation Systems, pp. 1–14, 2020. Y.-Q. Huang, J.-C. Zheng, S.-D. Sun, C.-F. Yang, and J. Liu, “Optimized YOLOv3 algorithm and its application in traffic flow detections,” Applied Sciences, vol. 10, Article ID 3079, 2020. W. Liu, D. Anguelov, D. Erhan et al., “SSD: single shot multibox detector,” in Proceedings of the Computer Vision—ECCV 2016, pp. 21–37, Amsterdam, Netherlands, October 2016. S. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137–1149, 2017. X. Zhang and X. Zhu, “Vehicle detection in the aerial infrared images via an improved Yolov3 network,” in Proceedings of the 2019 IEEE 4th International Conference on Signal and Image Processing (ICSIP), pp. 372–376, Wuxi, China, July 2019. A. Bochkovskiy, C. Y. Wang, and H. Y. M. Liao, “YOLOv4: optimal speed and accuracy of object detection,” 2020, https://arxiv.org/abs/2004.10934v1. D. Biswas, H. Su, C. Wang, A. Stevanovic, and W. Wang, “An automatic traffic density estimation using single shot detection (SSD) and mobilenet-SSD,” Physics and Chemistry of the Earth, Parts A/B/C, vol. 110, pp. 176–184, 2019. X. Hu, Z. Wei, and W. Zhou, “A video streaming vehicle detection algorithm based on YOLOv4,” in Proceedings of the 2021 IEEE 5th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), pp. 2081–2086, Chongqing, China, March 2021. H. Zhou, L. Wei, C. P. Lim, D. Creighton, and S. Nahavandi, “Robust vehicle detection in aerial images using bag-of-words and orientation aware scanning,” IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 12, pp. 7074–7085, 2018. R. Girshick, J. Donahue, T. Darrell, J. Malik, Rich feature hierarchies for accurate object detection and semantic segmentation, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 580–587, 2014. K. He, X. Zhang, S. Ren, J. Sun, Spatial pyramid pooling in deep convolutional networks for visual recognition, IEEE TPAMI 37 (9), 1904–16, (2014). R. Girshick, Fast r-cnn, in: Proceedings of the IEEE International Conference on Computer Vision, pp. 1440–1448, 2015. S. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: towards real-time object detection with region proposal networks, in: Advances in Neural Information Processing Systems, pp. 91-99, 2015. Z. Cai, Q. Fan, R.S. Feris, N. Vasconcelos, A unified multi-scale deep convolutional neural network for fast object detection, 354–370, (2016). J. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: unified, realtime object detection,779–788, (2016). Z. Li, J. Tang, Weakly supervised deep matrix factorization for social image understanding, IEEE Trans. Image Process. 26 (1), 276–288, (2017). C. Szegedy, S. Ioffe, V. Vanhoucke, A.A. Alemi, Inception-v4, inception-resnet and the impact of residual connections on learning, in: AAAI, 4, p. 12, 2017. Z. Li, J. Tang, T. Mei, Deep collaborative embedding for social image understanding, IEEE Trans. Pattern Anal. Mach. Intell. (2018) 1. Z. Li, J. Tang, X. He, Robust structured nonnegative matrix factorization for image representation, IEEE Trans. Neural Netw. Learn. Syst. 29 (5), 1947–1960, (2018). Z. Yan, H. Zhang, R. Piramuthu, V. Jagadeesh, D. DeCoste, W. Di, Y. Yu, Hd-cnn: hierarchical deep convolutional neural networks for large scale visual recognition, in: Proceedings of the IEEE International Conference on Computer Vision, pp. 2740–2748, 2015. S. Xie, T. Yang, X. Wang, Y. Lin, Hyper-class augmented and regularized deep learning for fine-grained image classification, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2645–2654, 2015. J.J. Lim, C.L. Zitnick, P. Dollár, Sketch tokens: a learned mid-level representation for contour and object detection, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, IEEE, pp. 3158–3165, 2013. E. Ohn-Bar, M.M. Trivedi, Learning to detect vehicles by clustering appearance patterns, IEEE Trans. Intell. Transp. Syst. 16 (5), 2511–2521, (2015). Y. Xiang, W. Choi, Y. Lin, S. Savarese, Data-driven 3d voxel patterns for object category recognition, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1903–1911, 2015. Y. Xiang, W. Choi, Y. Lin, S. Savarese, Subcategory-aware convolutional neural networks for object proposals and detection, in: Workshop on Applications of Computer Vision, IEEE, pp. 924–933, 2017. C.-H. Kuo, R. Nevatia, Robust multi-view car detection using unsupervised sub- -categorization, in: Workshop on Applications of Computer Vision, IEEE, pp. 1–8, 2009. Z. Wang, H. Li, W. Ouyang, X. Wang, Learning deep representations for scene labeling with semantic context guided supervision, arXiv preprint arXiv:1706. 02493 (2017) T. Xiao, T. Xia, Y. Yang, C. Huang, X. Wang, Learning from massive noisy labeled data for image classification, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2691–2699, 2015. A. Joulin, L. van der Maaten, A. Jabri, N. Vasilache, Learning visual features from large weakly supervised data, in: Proceedings of the ECCV, Springer, pp. 67–84, 2016. Kari, J., Varaha Siva Prasad, V. & Sagari, J. Investigation of the operating characteristics of diesel engines with chromium oxide (Cr 2 O 3 ) nanoparticles dispersed in Mesua ferrea biodiesel: an experimental and predictive approach using ANNs and RSM. Journal of Non-Equilibrium Thermodynamics , 50 (1), 21-58, (2025). Kari, J., Vanthala, V. S. P., & Sagari, J. Performance and emission characteristics of a diesel engine fuelled with Mesua ferrea biodiesel with chromium oxide (Cr2O3) nanoparticles: Experimental approach and response surface methodology. International Journal of Thermofluids , 22 , 100637, (2024). Kari, J., Vanthala, V. S. P., & Sagari, J. The influence of Cr2O3 nanoparticles dispersed Mesua ferrea biodiesel on the analysis performance, combustion, and emissions of diesel engine. Environment, Development and Sustainability , 26 (2), 4551-4577, (2024). Kari, J., Vanthala, V. S. P., & Sagari, J. The effect of a surfactant and dispersant mixed Cr2O3 nanoparticles on the analysis of stability and physicochemical properties of a Mesua ferrea biodiesel blend. Petroleum Science and Technology , 41 (24), 2402-2418, (2023). Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7476715","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":509387387,"identity":"02d5afe3-1496-4dc8-9a87-8635af86c2e0","order_by":0,"name":"Venkateswara Rao Gorle","email":"","orcid":"","institution":"Andhra university Visakhapatnam","correspondingAuthor":false,"prefix":"","firstName":"Venkateswara","middleName":"Rao","lastName":"Gorle","suffix":""},{"id":509387388,"identity":"fbe1389d-9081-4f9f-81b3-be20b2f0c42a","order_by":1,"name":"Sheik. Hidayatulla Shariff","email":"","orcid":"","institution":"Avanthi Institute of Engineering and Technology","correspondingAuthor":false,"prefix":"","firstName":"Sheik.","middleName":"Hidayatulla","lastName":"Shariff","suffix":""},{"id":509387391,"identity":"fe52b3db-0ca2-468d-8e6b-b13f483b9426","order_by":2,"name":"Rajendra Kagithapu","email":"","orcid":"","institution":"Aditya University","correspondingAuthor":false,"prefix":"","firstName":"Rajendra","middleName":"","lastName":"Kagithapu","suffix":""},{"id":509387392,"identity":"e4e47e52-b446-464d-8ef6-c158409890a5","order_by":3,"name":"Pinninti Charani","email":"","orcid":"","institution":"Andhra university Visakhapatnam","correspondingAuthor":false,"prefix":"","firstName":"Pinninti","middleName":"","lastName":"Charani","suffix":""},{"id":509387394,"identity":"1419668f-1251-47ee-b243-4a9e170db2b9","order_by":4,"name":"Sirela Meena","email":"","orcid":"","institution":"Avanthi's st Theressa Institute of Engineering and Technology","correspondingAuthor":false,"prefix":"","firstName":"Sirela","middleName":"","lastName":"Meena","suffix":""},{"id":509387395,"identity":"60f449ca-ccda-4c5f-bf5a-6682e63dff80","order_by":5,"name":"Vara Lakshmi Reddy","email":"data:image/png;base64,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","orcid":"","institution":"Avanthi’s St. Theressa Institute of engineering and technology","correspondingAuthor":true,"prefix":"","firstName":"Vara","middleName":"Lakshmi","lastName":"Reddy","suffix":""}],"badges":[],"createdAt":"2025-08-28 06:08:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7476715/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7476715/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90834730,"identity":"961567d7-be7f-443e-93e5-f8b4f5880454","added_by":"auto","created_at":"2025-09-08 17:29:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":34964,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlowchart of the vehicle classification and speed estimation process\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7476715/v1/135c5647b25d7ecfbf0f36df.png"},{"id":90834731,"identity":"970a3ce3-35d6-4ead-a798-74095151ea69","added_by":"auto","created_at":"2025-09-08 17:29:18","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":637790,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDetecting objects\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7476715/v1/62cd37b2ceb279abd0566d4e.png"},{"id":90834742,"identity":"2a52526f-20c6-4c4f-b153-0378a095cdd2","added_by":"auto","created_at":"2025-09-08 17:29:19","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":728739,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSpeed estimation\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7476715/v1/a387c59a1a7e465d3b07b700.png"},{"id":90834732,"identity":"f868ce49-996f-409f-ab10-664f40634b92","added_by":"auto","created_at":"2025-09-08 17:29:18","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":718325,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHomepage\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7476715/v1/a0b399e28fb9ed6d5688f325.png"},{"id":93619711,"identity":"502573e2-f57d-458d-84de-438ced0330d1","added_by":"auto","created_at":"2025-10-15 17:38:34","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3796510,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7476715/v1/960bdbe9-e69f-45a7-9727-38c7214b01fa.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Traffic Management System in Smart Cities to Identify the Vehicle and Detect the Speed of Vehicle to Optimize Traffic Management Strategies","fulltext":[{"header":"1. INTRODUCTION","content":"\u003cp\u003eIn the past, precise vehicle counts, speed estimation, accident detection, and efficient surveillance were all difficult tasks for traditional traffic monitoring techniques. These techniques lack accuracy, efficiency and frequently involve manual intervention. The combination of cutting-edge deep learning models, such as YOLO v8 and Mobilenet SSD, which enable real-time CCTV footage processing, forms the core of the system. With a remarkable accuracy rate over 95%, these models ensure a full grasp of traffic dynamics, allowing for informed decision-making. YOLO's adoption of the Thonny Python IDE shows a dedication to reliable and effective development techniques. MobilenetSSD is implemented using Jupyter Notebook, demonstrating a commitment to transparency and cooperation in the search for innovative solutions.\u003c/p\u003e\u003cp\u003eThe field of vehicle detection and traffic analysis has witnessed significant advancements, with researchers employing state-of-the-art techniques to enhance accuracy, robustness, and efficiency [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. A multi-level fusion network and labelling hierarchy for nighttime vehicle identification using highlight information which addresses the difficulties that arise in low-light environments [\u003cspan additionalcitationids=\"CR37 CR38\" citationid=\"CR28\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. An in-depth analysis of deep multi-modal object recognition and semantic segmentation, focusing on datasets, methods and problems in the field of autonomous driving [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e] integration of radar and video data under adverse weather conditions is highlighted and insights into improving the resilience of autonomous vehicle systems are given. DLT-NET, a model that simultaneously identifies drivable areas, lane markings and traffic objects, demonstrating a cohesive methodology for a thorough understanding of the scene [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] improves pedestrian safety by predicting movements from a remote first-person perspective, thereby supporting intelligent driving systems. [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e] presents a comprehensive analysis of the DATMO system and explains the detection and tracking of dynamic obstacles, an essential component of autonomous systems. An image processing-based framework for vehicle enumeration, speed assessment, and categorization, illustrating a comprehensive methodology for traffic analysis [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e] the optimization of the YOLO v3 algorithm for effective traffic flow detection is highlighted, demonstrating the adaptability of YOLO in practical applications. [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e] introduces the SSD model, a single-shot method for object identification, and explains its architecture and features. [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e] presents Faster R-CNN, a model for real-time object identification that improves both speed and accuracy using Region Proposal Networks. [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e] examines the implementation of an enhanced YOLO v3 network for vehicle detection in infrared aerial images is explored, highlighting the versatility of YOLO architectures. [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e] presents YOLO v4, which focuses on achieving optimum speed and precision in object recognition and illustrates the continuous further development of the YOLO series. [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e] introduces an automated traffic density estimation system that utilizes Single Shot Detection (SSD) and MobileNet SSD and provides insights into practical applications. [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e] proposes an algorithm based on YOLO v8 for detecting vehicles in video streaming and demonstrates its effectiveness for real-time applications. [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e] suggests a robust method for vehicle detection in aerial images is proposed that integrates bag-of-words and orientation-based scanning to improve accuracy.\u003c/p\u003e\u003cp\u003eThe researchers contribute to the advancement of vehicle detection, traffic analysis, and intelligent transportation systems and demonstrate progress in model architectures, optimization strategies, and real-world applications.\u003c/p\u003e"},{"header":"2. RELATED WORK","content":"\u003cp\u003eThis work includes a dynamic frontend interface created with HTML, CSS, and JavaScript to provide a smooth and intuitive experience. With the help of this interface's user-friendly dashboard, which is enhanced with charts, maps, and visualizations, stakeholders can easily understand traffic statistics. In order to facilitate historical analysis and reporting, a strong MySQL database is used on the backend to store and retrieve vital data, such as vehicle speeds and identifications. The development tools chosen for this work demonstrate a dedication to effectiveness and teamwork. For YOLO, the Thonny Python IDE ensures efficient development procedures, and for Mobilenet SSD, the Jupyter Notebook encourages open communication and transparency throughout the development process. The Smart City Traffic Management project is evidence of how technology may be used to solve intricate urban difficulties. The major advancement in the field of intelligent transportation systems since it emphasizes the use of state-of-the-art algorithms and approaches.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Object detection with deep learning\u003c/h2\u003e\u003cp\u003eThe advancement of deep learning, particularly Convolutional Neural Networks (CNN), has significantly enhanced object identification. Ross B. Girshick et al. introduced RCNN [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] inspired by the efficacy of deep learning in categorisation. It integrated an object suggestion mechanism with a convolutional neural network classifier. Building upon RCNN, other advancements such as SPP [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e], Fast RCNN [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e], and Faster RCNN [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] were introduced to enhance performance in terms of both efficacy and speed. They included the proposal prediction into a network structured as a two-stage model. MSCNN [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e] forecasted proposals at various stages during feature extraction and then used an object detection subnet to refine the results, hence enabling the receptive field of MSCNN to accommodate objects of diverse sizes. Another significant study is YOLO [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. In contrast to earlier networks, they introduced a one-stage architecture that simultaneously executed proposal prediction and categorisation. This network can achieve a performance of roughly 40 frames per second. While these approaches may provide improved outcomes in general item identification, including cars, their efficacy is constrained when the visual characteristics of vehicles are diminished at night. During that period, the detection results of automobiles may be substantially enhanced by using critical information such as vehicle features. We include vehicle highlight saliency information into the detection framework, allowing the detection network to enhance vehicle feature representations, extract vehicle proposals, and identify cars with more accuracy.\u003c/p\u003e\u003c/div\u003e"},{"header":"3. CATALOGUING HIERARCHY","content":"\u003cp\u003eLearning feature representations often encounters the issue of significant intra-class variability, which generates several ambiguities that perplex convolutional neural networks when they modify neurone weights to identify salient visual patterns. A prevalent approach involves designing networks capable of learning discriminative deep feature representations. Li et al. [22] introduced a hierarchical paradigm for the progressive learning of hidden representations. Szegedy et al. [23] included the \u0026quot;skip\u0026quot; layer to enhance feature propagation. DCE [24] integrated end-to-end learning with collaborative factor analysis to achieve optimum compatibility in representation learning and latent space discovery.\u003c/p\u003e\n\u003cp\u003eThe alternative approach is to use the label hierarchy. Li et al. [25] contended that pictures belonging to the same semantic category need to be embedded into the same latent representation subspace. Classifying sub-classes with enhanced visual consistency facilitates the learning process. Several studies have already investigated label hierarchies. HD-CNN [26] initially trained a coarse category classifier to segregate easy categories, and then classified the harder categories using a fine-grained category classifier. Xie et al. [27] offered two data augmentation approaches that need extra data to determine hyper-class of initial fine-grained labeled data. Lim et al. [28] introduced a feature termed \u0026ldquo;sketch token,\u0026rdquo; derived from supervised mid-level information represented by hand-drawn contours in pictures, thereafter clustering the feature to establish classes. Ohn-Bar et al. [29] and Xiang et al. [30,31] used supplementary 3D information in datasets for clustering to derive subclasses; nevertheless, the majority of datasets lack this prompt information. Kuo and Nevatia [32] clustered the HOG features after dimensionality reduction and classified the vehicle into subclasses with distinct orientations. Wang et al. [33] used semantic contexts, such as scene titles and label statistics of picture patches, to construct label hierarchies. The label subclasses automatically.\u003c/p\u003e\n\u003cp\u003eThe generation may include some noise; yet, [34,35] shown that labels with noise remain beneficial for the outcomes. In contrast to other studies, we establish the label hierarchy based on the unique characteristics of nocturnal vehicles. We meticulously extract concealed information from the vehicle dataset and categorise the cars in the training set into distinct subclasses based on vehicle highlight information, so establishing the label hierarchy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e3.1 YOLO\u0026nbsp;v8:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe most recent version of the You Only Look Once (YOLO) model, YOLO v8, created by Ultralytics, is well-known for its real-time object identification and image segmentation skills. A state-of-the-art computer vision model that blends accuracy and speed is called YOLO v8. By adding new features and enhancements, it builds on the popularity of earlier YOLO editions. Numerous visual AI tasks are supported by YOLO v8, such as object identification, segmentation, pose estimation, tracking, and classification.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eKey\u0026nbsp;Features:\u003c/em\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cem\u003ePerformance:\u003c/em\u003e YOLO v8 achieves state-of-the-art performance due to\u0026nbsp;advancements\u0026nbsp;in deep\u0026nbsp;learning and computer vision.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eAdaptability:\u003c/em\u003e It\u0026rsquo;s\u0026nbsp;suitable\u0026nbsp;for\u0026nbsp;various\u0026nbsp;applications\u0026nbsp;and\u0026nbsp;can\u0026nbsp;run\u0026nbsp;on\u0026nbsp;different\u0026nbsp;hardware\u0026nbsp;platforms, from edge\u0026nbsp;devices to cloud APIs.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eUnified Framework:\u003c/em\u003e YOLO v8\u0026nbsp;provides\u0026nbsp;a\u0026nbsp;unified\u0026nbsp;framework\u0026nbsp;for\u0026nbsp;training\u0026nbsp;models\u0026nbsp;across\u0026nbsp;multiple\u0026nbsp;tasks.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eGMM (Gaussian Mixture Model):\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cem\u003eMixture of Gaussians:\u003c/em\u003e GMMs represent normally distributed subpopulations within an\u0026nbsp;overall\u0026nbsp;population.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eProbabilistic Clustering:\u003c/em\u003e Unlike hard clustering (where each point belongs to a single\u0026nbsp;cluster),\u0026nbsp;GMMs\u0026nbsp;allow\u0026nbsp;soft\u0026nbsp;clustering,\u0026nbsp;where\u0026nbsp;data\u0026nbsp;points\u0026nbsp;can\u0026nbsp;partially\u0026nbsp;belong\u0026nbsp;to\u0026nbsp;multiple\u0026nbsp;clusters.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eParameter Estimation:\u003c/em\u003e GMMs\u0026nbsp;estimate parameters such\u0026nbsp;as\u0026nbsp;means,\u0026nbsp;variances,\u0026nbsp;and\u0026nbsp;mixing\u0026nbsp;coefficients.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eMathematical\u0026nbsp;Formulation:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eA\u0026nbsp;GMM is represented\u0026nbsp;as a weighted\u0026nbsp;sum of Gaussian\u0026nbsp;component densities. For\u0026nbsp;a\u0026nbsp;multivariate\u0026nbsp;Gaussian\u0026nbsp;distribution,\u0026nbsp;the\u0026nbsp;probability\u0026nbsp;density\u0026nbsp;function\u0026nbsp;is\u0026nbsp;given\u0026nbsp;by:\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"249\" height=\"70\"\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eHere,\u0026nbsp;\u0026pi;l\u0026nbsp;represents\u0026nbsp;the\u0026nbsp;mixing\u0026nbsp;coefficient\u0026nbsp;for\u0026nbsp;the\u0026nbsp;l\u003csup\u003eth\u003c/sup\u003e Gaussian component.\u003c/li\u003e\n \u003cli\u003e\u0026mu;l is\u0026nbsp;the\u0026nbsp;mean\u0026nbsp;vector, and \u0026Sigma;l is the covariance matrix for the\u0026nbsp;l\u003csup\u003eth\u003c/sup\u003e Gaussian.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eMobile Net\u0026nbsp;SSD\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e\u003cem\u003e:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMobileNet SSD V2 (MobileNet Single Shot Detector) is an object detection model designed\u0026nbsp;for\u0026nbsp;real-time\u0026nbsp;inference\u0026nbsp;on\u0026nbsp;devices\u0026nbsp;like\u0026nbsp;smartphones. It\u0026nbsp;combines\u0026nbsp;the\u0026nbsp;power\u0026nbsp;of\u0026nbsp;the\u0026nbsp;Mobile Net V2 \u0026nbsp;base\u0026nbsp;network\u0026nbsp;with\u0026nbsp;a\u0026nbsp;Single\u0026nbsp;Shot Detector (SSD)\u0026nbsp;layer.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cem\u003eReal-time\u0026nbsp;Performance:\u003c/em\u003e Achieves\u0026nbsp;good\u0026nbsp;real-time\u0026nbsp;results\u0026nbsp;even\u0026nbsp;on\u0026nbsp;limited\u0026nbsp;compute\u0026nbsp;resources (30\u0026nbsp;frames per\u0026nbsp;second).\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eCompact\u0026nbsp;Size:\u003c/em\u003e Once\u0026nbsp;trained,\u0026nbsp;MobileNet\u0026nbsp;SSD\u0026nbsp;V2\u0026nbsp;can\u0026nbsp;be\u0026nbsp;stored\u0026nbsp;with\u0026nbsp;just\u0026nbsp;63\u0026nbsp;MB,\u0026nbsp;making\u0026nbsp;it ideal\u0026nbsp;for smaller devices.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eArchitecture:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cem\u003eBase\u0026nbsp;Network:\u003c/em\u003e The\u0026nbsp;first\u0026nbsp;part\u0026nbsp;consists\u0026nbsp;of\u0026nbsp;the\u0026nbsp;MobileNetV2\u0026nbsp;network,\u0026nbsp;which\u0026nbsp;acts\u0026nbsp;as a\u0026nbsp;feature extractor.\u003c/li\u003e\n \u003cli\u003e\u003cem\u003eSSD\u0026nbsp;Layer:\u003c/em\u003e The\u0026nbsp;SSD\u0026nbsp;layer\u0026nbsp;classifies\u0026nbsp;detected\u0026nbsp;objects\u0026nbsp;based\u0026nbsp;on\u0026nbsp;the\u0026nbsp;features\u0026nbsp;extracted\u0026nbsp;by MobileNetV2.\u003c/li\u003e\n \u003cli\u003eThe\u0026nbsp;combination\u0026nbsp;of\u0026nbsp;these\u0026nbsp;two\u0026nbsp;parts\u0026nbsp;enables\u0026nbsp;efficient\u0026nbsp;and\u0026nbsp;accurate\u0026nbsp;object\u0026nbsp;detection.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003ePerformance:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eMobileNet\u0026nbsp;V2\u0026nbsp;outperforms\u0026nbsp;its\u0026nbsp;predecessor,\u0026nbsp;MobileNet\u0026nbsp;V1,\u0026nbsp;with\u0026nbsp;higher\u0026nbsp;accuracies\u0026nbsp;and lower\u0026nbsp;latencies.\u003c/li\u003e\n \u003cli\u003eIt\u0026rsquo;s\u0026nbsp;optimized\u0026nbsp;for\u0026nbsp;mobile\u0026nbsp;devices,\u0026nbsp;making\u0026nbsp;it\u0026nbsp;suitable\u0026nbsp;for\u0026nbsp;applications\u0026nbsp;like\u0026nbsp;image\u0026nbsp;recognition,\u0026nbsp;tracking,\u0026nbsp;and more.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTraining\u0026nbsp;and\u0026nbsp;Deployment:\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003eYou can train MobileNet SSD V2 on your custom dataset using Tensor Flow\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"4. APPROACH DESCRIPTION","content":"\u003cp\u003e\u003cb\u003eCollection of CCTV film\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eCompile a varied dataset of CCTV film showing various urban traffic situations. Variations in vehicle kinds, traffic density, and illumination conditions should all be included in this dataset.\u003c/p\u003e\u003cp\u003e\u003cb\u003eDefinition of the Problem\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eSpecify your goals\u003c/strong\u003e\u003cp\u003eThe objectives of the traffic management system, such as precise vehicle counts, speed estimation, and vehicle type categorization, should be stated clearly.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eChoosing a Model\u003c/em\u003e: Deep Learning Model Selection: Pick suitable deep learning models for categorization, speed estimation, and object identification. Models such as YOLO, MobilenetSSD, or a mix of models designed for particular activities may be taken into consideration.\u003c/p\u003e\u003cp\u003e\u003cem\u003ePreprocessing Data\u003c/em\u003e: Image preprocessing: To improve the model's capacity to generalize under various circumstances, prepare the dataset by resizing, standardizing, and enhancing images.\u003c/p\u003e\u003cp\u003e\u003cb\u003eInstruction\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cem\u003eModel Training\u003c/em\u003e:\u003c/p\u003e\u003cp\u003eUse the pre-processed dataset to train the chosen deep learning models. To attain high accuracy in vehicle detection, speed estimates, and classification, adjust loss functions, hyper parameters, and other parameters.\u003c/p\u003e\u003cp\u003e\u003cb\u003eAnalysis in real time\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eReal-time analysis implementation: Use the trained models to examine CCTV material instantly. In order to process incoming video streams continually, the models must be integrated into the system.\u003c/p\u003e\u003cp\u003e\u003cb\u003eFront-end and back-end integration\u003c/b\u003e:\u003c/p\u003e\u003cp\u003e\u003cb\u003eDevelopment of Frontends\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eTo visualize real-time traffic data, create an easy-to-use frontend interface with HTML, CSS, JavaScript, and PHP. created an intuitive dashboard with maps, charts, and visualizations to efficiently communicate information.\u003c/p\u003e\u003cp\u003e\u003cb\u003eBackend Implementation\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eCreate a solid backend with MySQL to store and retrieve vital data, like vehicle identity and speed, allowing for reporting and historical analysis.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"5. DATA EXPLORATION","content":"\u003cp\u003e\u003cb\u003eCOCO dataset\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eA large-scale image recognition dataset for object detection, segmentation, and captioning applications is called COCO (Common Objects in Context). Each of the more than 330,000 photos in it has five subtitles that describe the setting and 80 object types. Numerous cutting-edge object detection and segmentation models have been trained and evaluated using the COCO dataset, which is extensively utilized in computer vision research. In machine learning, COCO dataset is frequently utilized for both research and real-world applications. The photos and their annotations make up the two primary components of the collection.\u003c/p\u003e\u003cp\u003eA hierarchy of folders is used to arrange the photos, with subdirectories for the train, validation, and test sets located in the top-level directory. Each file corresponds to a single image, and the annotations are supplied in JSON format.\u003c/p\u003e\u003cp\u003e\u003cb\u003eData Manipulation\u003c/b\u003e:\u003c/p\u003e\u003cp\u003eData manipulation is the process of altering or modifying data to make it easier to deal with by making it more comprehensible and structured. Only when you have the data to do so can you manipulate it. As a result, you require a database that is created from multiple data sources. Data manipulation aids in information cleaning. The database's data must be restructured and reorganized in order to complete this task. The raw video data is pre-processed and frame separated in the first step of data manipulation in order to get it ready for further analysis. In order to ensure that only pertinent frames with vehicles are kept for additional processing, preprocessing include cleaning and arranging the data to eliminate noise and unnecessary information. After that, frame separation is done to separate frames from the video stream, making it possible to analyse each frame effectively on its own.\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e\u003cem\u003eData Preparation\u003c/em\u003e:\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eOne of the crucial things we must perform before analysing the data is data preparation. This procedure involves purifying the dataset's raw data prior to processing and analysis. In order to refine the data, it also entails reformatting, fixing errors, and combining the data sets.\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e\u003cem\u003eBlurry Videos\u003c/em\u003e:\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBlurry videos can significantly affect object detection accuracy. Blurriness maybe caused by camera motion, poor focus, or low frame rates, making it challenging to identify\u003c/p\u003e\u003cp\u003eand track objects reliably.\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e\u003cem\u003eSolution\u003c/em\u003e:\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eImplement image stabilization techniques to reduce the impact of camera shake.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eDetects and filters out excessively blurry frames to improve overall detection accuracy.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e\u003cem\u003eLow-Light Videos\u003c/em\u003e:\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eBecause of the decreased visibility and increased noise in low-light (night time) recordings, it can be difficult to detect and track vehicles.\u003c/p\u003e\u003cp\u003e\u003cem\u003eSolution\u003c/em\u003e:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eTo get better evening photos, use thermal or infrared (IR) cameras.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eTo increase visibility, use sophisticated picture enhancing techniques.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eMake use of object identification models that have been specially trained for low light levels.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e"},{"header":"6. DATA ANALYSIS","content":"\u003cp\u003e\u003cstrong\u003eClasses in COCO Dataset\u003c/strong\u003e\u003cp\u003eA large-scale image recognition dataset for object detection, segmentation, and captioning applications is called COCO (Common Objects in Context). Each of the more than 330,000 photos in it has five subtitles that describe the setting and 80 object types. Numerous cutting-edge object detection and segmentation models have been trained and evaluated using the COCO dataset, which is extensively utilized in computer vision research.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eCOCO dataset to object identification evaluation.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePerson\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFire hydrant\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eelephant\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSkis\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eWine glass\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003ebroccoli\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eDining table\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eToaster\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBicycle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eStop sign\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ebear\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSnow board\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eCup\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCarrot\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003etoilet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003esink\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCar\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eParking meter\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003ezebra\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSports ball\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eFork\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eHot dog\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003etv\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eRefrigerator\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMotorcycle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ebench\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003egiraffe\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eKite\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eKnife\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePizza\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eLaptop\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eBook\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAirplane\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ebird\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBackpack\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBaseball bat\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSpoon\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eDonut\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eMouse\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eClock\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBus\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ecat\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eUmbrella\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eBaseball glove\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBowl\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCake\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eRemote\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eVase\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTrain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003edog\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eHandbag\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSkate board\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eBanana\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eChain\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eKeyboard\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eScissors\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTruck\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ehorse\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTie\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSurf board\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eApple\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eCouch\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCell phone\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eTeddy bear\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBoat\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003esheep\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSuitcase\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTennis racket\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eSandwich\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ePotted plant\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eMicrowave\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eHair drier\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTraffic light\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003ecow\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003efrisbee\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ebottle\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eorange\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003ebed\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eOven\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003etoothbrush\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003e\u003cb\u003eDistribution of classes in coco dataset\u003c/b\u003e:\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eWe must forecast the bounding boxes and the labels that go with them in order to detect objects. It's crucial to remember that the class imbalance in the COCO dataset causes inherent bias.\u003c/p\u003e"},{"header":"7. MODELLING","content":"\u003cp\u003e\u003cstrong\u003eModel\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eDevelopment Object Detection:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026bull; YOLO Model: To detect objects, the model loads a pre-trained YOLO model (\u0026apos;yolov8s.pt\u0026apos;) using the ultralytics library.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Input: The input stream is a live video recording.\u003c/p\u003e\n\u003cp\u003e\u0026bull; Output: It gives the output of the bounding boxes that were found around the cars.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTracking:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026bull; Object tracking: To track cars between frames, a bespoke object tracking algorithm (track.py) was implemented.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; The tracking algorithm makes use of centroid tracking, which maintains object IDs by calculating and matching the centroids of bounding boxes between successive frames.\u003c/p\u003e\n\u003cp\u003e\u0026bull; Input: Bounding boxes were found.\u003c/p\u003e\n\u003cp\u003e\u0026bull; Output: Vehicle IDs and matching bounding boxes were tracked.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDatabase\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eintegration:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026bull; MySQL Integration: Creates a connection to a MySQL database in order to record data about identified objects, such as lane, speed, and ID.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Input: Information about detected objects.\u003c/p\u003e\n\u003cp\u003e\u0026bull; Result: Information was saved in the MySQL database.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eModel Evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInitialization:\u0026nbsp;\u003c/strong\u003eEstablishes connections, loads the YOLO model, and initializes variables.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLoop for video processing:\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Resizes frames for processing;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Reads frames from the input video.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Uses the YOLO model to identify automobiles.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Follows identified cars between frames.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Determines vehicle speeds and counts those that cross specified lines.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Annotates frames with text, lane lines, and bounding boxes.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026bull; Adds identified object information to the database.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTermination:\u003c/strong\u003e When all frames have been processed or a user interrupts, resources are released\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eand connections are closed. Real-time lane counting, speed calculation, tracking, and vehicle detection are all features of the implemented system. It interfaces with a MySQL database for data persistence, uses YOLO for object recognition, and employs unique tracking techniques to preserve object IDs. For a number of uses, including automated toll collecting, traffic monitoring, and congestion analysis, this system can be further enhanced and expanded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable. 2: Detected data with objects and speed\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"261\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"2\" style=\"width: 261px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTraffic Information\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 111px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 150px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSpeed (km/hr)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e239\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e236\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e115\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e220\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e56\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e43\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e36\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e30\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e27\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e215\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e21\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e211\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 111px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 150px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"8.\tRESULTS AND CONCLUSIONS","content":"\u003cul\u003e\n \u003cli\u003eDetected object information, including ID, speed, and lane information, is stored in a MySQL database. This allows for further analysis and retrieval of historical data.\u003c/li\u003e\n \u003cli\u003eBounding boxes around vehicles, designated lines for speed calculation and lane counting, and text annotations displaying lane-wise vehicle counts are visualized on the video frames.\u003c/li\u003e\n \u003cli\u003eThe system processes the input video stream in real-time, providing immediate feedback on vehicle detection, tracking, speed calculation, and lane counting.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"9. CONCLUSION","content":"\u003cp\u003eThe Smart City Traffic Management project represents a pioneering effort in leveraging cutting-edge technologies, notably MobilenetSSD and YOLO, to address the multifaceted challenges of urban traffic congestion. Through the tracks detected vehicles across frames. seamless integration of these advanced deep learning models, the project has achieved remarkable success in real-time vehicle detection, speed estimation, and classification. The user-friendly frontend interface, coupled with robust backend systems, ensures accessibility and efficiency for a diverse range of stakeholders, from city authorities and transportation departments to individual commuters.\u003c/p\u003e\n\u003cp\u003eAs we conclude this endeavor, it is evident that the Smart City Traffic Management project not only meets the immediate objectives of enhanced traffic efficiency but also aligns with the broader goals of creating sustainable, connected, and livable urban environments. The continuous improvement ethos, coupled with adaptability to emerging technologies, positions this solution as a dynamic tool for ongoing advancements in the realms of urban planning, transportation, and smart city initiatives. This project serves as a testament to the potential of technology to reshape and elevate the urban living experience.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding Declaration:\u0026nbsp;\u003c/strong\u003eThis research did not receive any specific grant from funding agencies.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eA: Data Curation, Methodology, Writing-Original, Writing-Review \u0026amp; Editing.B: Conceptualization, Supervision, Methodology. C: Supervision, Conceptualization, Resources, Methodology.D: Draft, Formal analysis, Software.E: Formal analysis, Investigation, SoftwareF : Writing-Review \u0026amp; Editing, Validation, Software.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eY. Mo, G. Han, H. Zhang, X. Xu, and W. Qu, “Highlight-assisted nighttime vehicle detection using a multi-level fusion network and label hierarchy,” Neurocomputing, vol. 355, pp. 13 23, 2019.\u003c/li\u003e\n \u003cli\u003eD. Feng, C. Haase-Schuetz, L. Rosenbaum et al., “Deep multi-modal object detection and semantic segmentation for autonomous driving: datasets, methods, and challenges,” IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 3, p. 3, 2019.\u003c/li\u003e\n \u003cli\u003eZ. Liu, Y. Cai, H. Wang et al., “Robust target recognition and tracking of self-driving cars with radar and camera information fusion under severe weather conditions,” IEEE Transactions on Intelligent Transportation Systems, 2021.\u003c/li\u003e\n \u003cli\u003eY. Qian, J. M. Dolan, and M. Yang, “DLT-NET: joint detection of drivable areas, lane lines, and traffic objects,” IEEE Transactions on Intelligent Transportation Systems, vol. 21, no. 11, pp. 4670–4679, 2020.\u003c/li\u003e\n \u003cli\u003eY. Cai, L. Dai, H. Wang et al., “Pedestrian motion trajectory prediction in intelligent driving from far shot first-person perspective video,” IEEE Transactions on Intelligent Transportation Systems, pp. 1–16, 2021.\u003c/li\u003e\n \u003cli\u003eÁ. Llamazares, E. J. Molinos, and M. Ocaña, “Detection and tracking of moving obstacles (DATMO): a review,” Robotica, vol. 38, no. 5, pp. 761–774, 2020.\u003c/li\u003e\n \u003cli\u003eC. Liu, D. Q. Huynh, Y. Sun, M. Reynolds, and S. Atkinson, “A vision-based pipeline for vehicle counting, speed estimation, and classification,” IEEE Transactions on Intelligent Transportation Systems, pp. 1–14, 2020.\u003c/li\u003e\n \u003cli\u003eY.-Q. Huang, J.-C. Zheng, S.-D. Sun, C.-F. Yang, and J. Liu, “Optimized YOLOv3 algorithm and its application in traffic flow detections,” Applied Sciences, vol. 10, Article ID 3079, 2020.\u003c/li\u003e\n \u003cli\u003eW. Liu, D. Anguelov, D. Erhan et al., “SSD: single shot multibox detector,” in Proceedings of the Computer Vision—ECCV 2016, pp. 21–37, Amsterdam, Netherlands, October 2016.\u003c/li\u003e\n \u003cli\u003eS. Ren, K. He, R. Girshick, and J. Sun, “Faster r-cnn: towards real-time object detection with region proposal networks,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137–1149, 2017.\u003c/li\u003e\n \u003cli\u003eX. Zhang and X. Zhu, “Vehicle detection in the aerial infrared images via an improved Yolov3 network,” in Proceedings of the 2019 IEEE 4th International Conference on Signal and Image Processing (ICSIP), pp. 372–376, Wuxi, China, July 2019.\u003c/li\u003e\n \u003cli\u003eA. Bochkovskiy, C. Y. Wang, and H. Y. M. Liao, “YOLOv4: optimal speed and accuracy of object detection,” 2020, https://arxiv.org/abs/2004.10934v1.\u003c/li\u003e\n \u003cli\u003eD. Biswas, H. Su, C. Wang, A. Stevanovic, and W. Wang, “An automatic traffic density estimation using single shot detection (SSD) and mobilenet-SSD,” Physics and Chemistry of the Earth, Parts A/B/C, vol. 110, pp. 176–184, 2019.\u003c/li\u003e\n \u003cli\u003eX. Hu, Z. Wei, and W. Zhou, “A video streaming vehicle detection algorithm based on YOLOv4,” in Proceedings of the 2021 IEEE 5th Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), pp. 2081–2086, Chongqing, China, March 2021.\u003c/li\u003e\n \u003cli\u003eH. Zhou, L. Wei, C. P. Lim, D. Creighton, and S. Nahavandi, “Robust vehicle detection in aerial images using bag-of-words and orientation aware scanning,” IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 12, pp. 7074–7085, 2018.\u003c/li\u003e\n \u003cli\u003eR. Girshick, J. Donahue, T. Darrell, J. Malik, Rich feature hierarchies for accurate object detection and semantic segmentation, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 580–587, 2014.\u003c/li\u003e\n \u003cli\u003eK. He, X. Zhang, S. Ren, J. Sun, Spatial pyramid pooling in deep convolutional networks for visual recognition, IEEE TPAMI 37 (9), 1904–16, (2014).\u003c/li\u003e\n \u003cli\u003eR. Girshick, Fast r-cnn, in: Proceedings of the IEEE International Conference on Computer Vision, pp. 1440–1448, 2015.\u003c/li\u003e\n \u003cli\u003eS. Ren, K. He, R. Girshick, J. Sun, Faster r-cnn: towards real-time object detection with region proposal networks, in: Advances in Neural Information Processing Systems, pp. 91-99, 2015.\u003c/li\u003e\n \u003cli\u003eZ. Cai, Q. Fan, R.S. Feris, N. Vasconcelos, A unified multi-scale deep convolutional neural network for fast object detection, 354–370, (2016).\u003c/li\u003e\n \u003cli\u003eJ. Redmon, S. Divvala, R. Girshick, A. Farhadi, You only look once: unified, realtime object detection,779–788, (2016).\u003c/li\u003e\n \u003cli\u003eZ. Li, J. Tang, Weakly supervised deep matrix factorization for social image understanding, IEEE Trans. Image Process. 26 (1), 276–288, (2017).\u003c/li\u003e\n \u003cli\u003eC. Szegedy, S. Ioffe, V. Vanhoucke, A.A. Alemi, Inception-v4, inception-resnet and the impact of residual connections on learning, in: AAAI, 4, p. 12, 2017.\u003c/li\u003e\n \u003cli\u003eZ. Li, J. Tang, T. Mei, Deep collaborative embedding for social image understanding, IEEE Trans. Pattern Anal. Mach. Intell. (2018) 1.\u003c/li\u003e\n \u003cli\u003eZ. Li, J. Tang, X. He, Robust structured nonnegative matrix factorization for image representation, IEEE Trans. Neural Netw. Learn. Syst. 29 (5), 1947–1960, (2018).\u003c/li\u003e\n \u003cli\u003eZ. Yan, H. Zhang, R. Piramuthu, V. Jagadeesh, D. DeCoste, W. Di, Y. Yu, Hd-cnn:\u003cbr\u003e hierarchical deep convolutional neural networks for large scale visual recognition, in: Proceedings of the IEEE International Conference on Computer Vision, pp. 2740–2748, 2015.\u003c/li\u003e\n \u003cli\u003eS. Xie, T. Yang, X. Wang, Y. Lin, Hyper-class augmented and regularized deep\u003cbr\u003e learning for fine-grained image classification, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2645–2654, 2015.\u003c/li\u003e\n \u003cli\u003eJ.J. Lim, C.L. Zitnick, P. Dollár, Sketch tokens: a learned mid-level representation\u003cbr\u003e for contour and object detection, in: Proceedings of the IEEE Conference on\u003cbr\u003e Computer Vision and Pattern Recognition, IEEE, pp. 3158–3165, 2013.\u003c/li\u003e\n \u003cli\u003eE. Ohn-Bar, M.M. Trivedi, Learning to detect vehicles by clustering appearance\u003cbr\u003e patterns, IEEE Trans. Intell. Transp. Syst. 16 (5), 2511–2521, (2015).\u003c/li\u003e\n \u003cli\u003eY. Xiang, W. Choi, Y. Lin, S. Savarese, Data-driven 3d voxel patterns for object\u003cbr\u003e category recognition, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1903–1911, 2015.\u003c/li\u003e\n \u003cli\u003eY. Xiang, W. Choi, Y. Lin, S. Savarese, Subcategory-aware convolutional neural\u003cbr\u003e networks for object proposals and detection, in: Workshop on Applications of Computer Vision, IEEE, pp. 924–933, 2017.\u003c/li\u003e\n \u003cli\u003eC.-H. Kuo, R. Nevatia, Robust multi-view car detection using unsupervised sub-\u003cbr\u003e -categorization, in: Workshop on Applications of Computer Vision, IEEE, pp. 1–8, 2009.\u003c/li\u003e\n \u003cli\u003eZ. Wang, H. Li, W. Ouyang, X. Wang, Learning deep representations for scene\u003cbr\u003e labeling with semantic context guided supervision, arXiv preprint arXiv:1706. 02493 (2017)\u003c/li\u003e\n \u003cli\u003eT. Xiao, T. Xia, Y. Yang, C. Huang, X. Wang, Learning from massive noisy labeled data for image classification, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2691–2699, 2015.\u003c/li\u003e\n \u003cli\u003eA. Joulin, L. van der Maaten, A. Jabri, N. Vasilache, Learning visual features from large weakly supervised data, in: Proceedings of the ECCV, Springer, pp. 67–84, 2016.\u003c/li\u003e\n \u003cli\u003eKari, J., Varaha Siva Prasad, V. \u0026amp; Sagari, J. Investigation of the operating characteristics of diesel engines with chromium oxide (Cr\u003csub\u003e2\u003c/sub\u003eO\u003csub\u003e3\u003c/sub\u003e) nanoparticles dispersed in \u003cem\u003eMesua ferrea\u003c/em\u003e biodiesel: an experimental and predictive approach using ANNs and RSM. \u003cem\u003eJournal of Non-Equilibrium Thermodynamics\u003c/em\u003e, \u003cem\u003e50\u003c/em\u003e(1), 21-58, (2025).\u003c/li\u003e\n \u003cli\u003eKari, J., Vanthala, V. S. P., \u0026amp; Sagari, J. Performance and emission characteristics of a diesel engine fuelled with Mesua ferrea biodiesel with chromium oxide (Cr2O3) nanoparticles: Experimental approach and response surface methodology. \u003cem\u003eInternational Journal of Thermofluids\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e, 100637, (2024).\u003c/li\u003e\n \u003cli\u003eKari, J., Vanthala, V. S. P., \u0026amp; Sagari, J. The influence of Cr2O3 nanoparticles dispersed Mesua ferrea biodiesel on the analysis performance, combustion, and emissions of diesel engine. \u003cem\u003eEnvironment, Development and Sustainability\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(2), 4551-4577, (2024). \u003c/li\u003e\n \u003cli\u003eKari, J., Vanthala, V. S. P., \u0026amp; Sagari, J. The effect of a surfactant and dispersant mixed Cr2O3 nanoparticles on the analysis of stability and physicochemical properties of a Mesua ferrea biodiesel blend. \u003cem\u003ePetroleum Science and Technology\u003c/em\u003e, \u003cem\u003e41\u003c/em\u003e(24), 2402-2418, (2023).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7476715/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7476715/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe importance of intelligent vehicle recognition and counting has significantly increased in the context of contemporary highway administration. However, the precision of vehicle counts is directly challenged by the fact that cars vary in size, making it difficult to identify them accurately. This paper introduces a technique for identifying and counting vehicles using vision to tackle this problem. DeepSORT model was used, which is based on the You Only Look Once (YOLO v8) model, to recognize and track vehicles in real time within video sequences. By combining the sophisticated detection capabilities of YOLO v8 with the skilful tracking algorithms of DeepSORT, the suggested method aims to improve accuracy by taking into account the intricacies of various vehicle sizes and movement patterns. The system gives highway management real-time insights into vehicle traffic, enabling them to well-informed information to improve traffic control tactics. Furthermore, the system's capacity to smoothly adjust to intricate traffic situations is enhanced by the combination of YOLO v8 and DeepSORT.\u003c/p\u003e","manuscriptTitle":"Traffic Management System in Smart Cities to Identify the Vehicle and Detect the Speed of Vehicle to Optimize Traffic Management Strategies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-08 17:29:14","doi":"10.21203/rs.3.rs-7476715/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ee086f4d-46f4-4520-9f54-73387662afa4","owner":[],"postedDate":"September 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-10-15T17:38:16+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-08 17:29:14","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7476715","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7476715","identity":"rs-7476715","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}