{"paper_id":"288d0879-0a9f-4967-8da7-e5a9bc94ce2f","body_text":"Nondestructive Detection of Internal Defects in Potato by Visible/Near Infrared Spectroscopy | 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 Nondestructive Detection of Internal Defects in Potato by Visible/Near Infrared Spectroscopy Syed Mansha Rafiq, Priyanshu Priyam, Raman Mishra, Sanskar Varshney, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7176490/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 This research was carried out partly in NIFTEM-K and IIT Gandhinagar in year 2022 and investigates the application of non-destructive Visible/ Near-Infrared spectroscopy for detecting internal defects in potatoes. Conventional methods followed for defect detection are largely destructive, leading to significant material waste and increased operational costs. The study is designed to provide an innovative, efficient, and economical approach of NIR technology combined with advanced data analysis techniques to detect and segregate the quality potatoes. Specifically, it employs machine learning algorithms such as Decision Tree Classifier (DT), Logistic Regression, K Nearest Neighbor (KNN), and a Deep Learning model, the Multi-Layered Perceptron Classifier to categorize potatoes based on internal defects. The NIR sensor (NIRvascan Smart Near Infrared Spectrometer Reflective Model G1) was used for detection. The spectrum was analyzed using specialized software Python 3.7 to be run in Jupyter Notebooks. A total of 600 potatoes were selected for the experiment. After performing 4 scans per potato we had a total of 2400 scans. 400 scans were discarded due to poor signal quality. 1200 were labeled as Black and the remaining 800 as Green. Dataset was divided into Train: Test and Validation sets. A ratio of 70:20:10 was selected for the split. The research identifies the Multi-Layered Perception Classifier as the most effective model, outperforming others in terms of F1 score and accuracy. The accuracy was found to be 81%. This breakthrough offers a scalable solution for enhancing potato quality assessment, with substantial implications for the agricultural industry and food supply chains. Potato Internal Defects Machine Learning Deep Learning Non-Destructive tests Near Infrared Sensors Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 INTRODUCTION Potatoes ( Solanum tuberosum L .) play an essential role in providing food security but are susceptible to diseases and pests during the supply chain operations (Weng et al. 2024; Charkowski et al. 2020). Maintaining the quality of potato tubers throughout the production and supply chain is crucial to minimize economic losses (Imanian et al. 2021). Mostly these physiological disorders are restricted to their presence internally that gets unnoticed during grading practices. On the other end, conventional method of destructive analysis is time consuming and requires highly skilled man power to execute (Ibrahim et al. 2020). The currently used non-destructive technique for potato defect detection is restricted to the use of colorimeter, where the results are limited to surface color characterization. Recently the use of imaging and spectroscopic techniques is increasingly being used to provide a non-destructive and objective method for grading the agricultural commodities including (Sharma et al. 2024, Adesokan et al. 2023). Advanced spectroscopic methods based on Vis/NIR (400-1000 nm) gets advantage with the ease of radiation penetration without having any hazardous impact and residue generation. The number of waves getting absorbed at a particular wavelength is indicative of the type of defect present inside the potato (Semyalo et al. 2024, Sharma et al. 2023) Employing machine learning algorithms such as classifiers assist in better prediction of response variables arising from spectroscopic observations. Vis/NIR spectroscopy demonstrates excellent ability to measure quality attributes of fruits and vegetables (Pahlawan et al. 2022) like brown heart in pears and apples (Afzaal et al. 2023 ) blackheart in potatoes (Zhou et al. 2015). Imanian et al. (2021) used the reflection mode of Vis/NIR and compared with short wave to classify the potatoes with hollow heart, and brown rot and found a classification rate of 83.4%. However, the internal defects which are commonly prevalent also include the greening and internal black spots, for which we could hardly find any available research. Hence we undertook this activity to detect the specific physiological disorder at wavelength scan of (513-850 nm) and used the machine learning algorithms to predict suitable wavelength and suitable classifiers were used for predicting the defect lot. MATERIALS AND METHODS Sample acquisition & preparation: The study was conducted partly in NIFTEM-K and IIT Gandhinagar in 2022. Freshly harvested potatoes of variety (Kufri Chipsona) (400 numbers) were procured from Azadpur Mandi located in New Delhi and 200 potatoes were procured from SK Cold Storage located in Gandhinagar district of Gujarat. Samples were cleaned and washed with water to remove adhering soil matrix and dust particles. The potatoes with surface damages were sorted during the process, and the washed potatoes were labelled and immediately transferred to laboratory for the spectroscopic analysis. Preparation of samples : An extensive literature review was done to understand the environmental conditions which could induce greening inside the potatoes. Accordingly, the potatoes were exposed to sunlight for 12 hours to induce greening effect (Fig. 1a). For the case of internal black rots, the disordered potatoes were directly procured from SK Cold storage which was being sorted and certified as being black rot by the industry (Fig. 1b). Instrumentation: The NIR sensor used (NIRvascan Smart Near Infrared Spectrometer Reflective Model G1, Wavelength Range: 900-1,700 mm, 10 nm Pixel resolution 1.17nm/pixel (accuracy + or -0,7nm), Signal-to noise ratio: 6000:1 in 1 second scan, Slit Size: 1.69 x 0.025 mm) was available at IIT Gandhinagar (Fig. 2a). The source fiber was utilized to provide direct light from a DSN Xenon Arc lamp onto the surface of the potato at a distance of 15 cm. This was precisely achieved using an SP7 tip, which facilitated the focused delivery of NIR light to the targeted area on the potato (Fig. 2b). Concurrently, the detector fiber, situated adjacent to the source fiber, collected the NIR light that was reflected back from the potato. This step is critical as it involves the gathering of data that will be analyzed for potential defects. The NIR light captured by the detector fiber was then transmitted to the spectrometer. Here, the light underwent dispersion to separate it into its constituent wavelengths. Data Analysis and Modeling: The signal acquisition process took place at IIT Gandhinagar where we scanned the potatoes. Once the scan acquisition was done we used Python 3.7 and Jupyter Notebooks environment for data analysis. The software was installed using Anaconda on a desktop with windows 7 operating system. Three machine learning models, Decision Tree Classifier (DTL), Logistic Regression, K Nearest Neighbor (KNN), and a Deep Learning model, the Multi-Layered Perceptron (MLP) Classifier were used for classification and prediction of disorders. RESULTS AND DISCUSSION We had a total of 600 potatoes. The potatoes were radiated with four scans in order to ensure the rays falling to the targeted disorder spot. After performing 4 scans per potato, we had a total of 2400 scans. 400 scans were discarded due to poor signal quality. Out of 2000 scans 1200 were labeled as Black and the remaining 800 as Green. The readings were taken in the Vis/NIR range (400–1700 nm). Each scan was labeled as either internal black defect or internal green defect, depending on the truth value, which was determined by inducing the internal defects inside the potatoes. The obtained data was transferred using bluetooth to a mobile device in the form of a csv file (Fig 3a) which was then uploaded onto a server and was retrieved in Jupyter Notebook software compatible with multiple programming languages, including Python for further analysis. The retrieved data from cloud contained the values of absorption at 228 different wavelengths in the range of 600 to 1700 nm (Fig. 3a). For every 228 wavelengths there was a corresponding label, either black or green. Black was for potatoes with black internal spots (Fig. 4a) and green was for potatoes with internal greening defects (Fig. 4b). The absorbance intensity for green and black spots was different at the same wavelength (Fig. 3b). Absorbance values from the first 221 wavelengths were selected for calibration since above this threshold there was noise interference. Savgol Filter was applied on the raw data. Savgol filter is used to reduce high-frequency pitches and notes in the incoming signal using its smoothing properties and consequently help reduce low-frequency signals (e.g., due to offsets and slopes) using differentiation (Gallagher 2020). Data analysis was done using machine learning programs. DTL is an inductive algorithm derived from instances, which rely on classification metrics displaying as decision trees derived from a group of disorder and irregular instances. Logistic regression is dominantly the most widely used methods in data science in binary and general model classification specifically. KNN is a widely popular technique popularly used for the purposes of model training and classification. MLP Classifier is a widely popular neural network. Unlike other classification algorithms such as KNN or Naive Bayes Classifier, MLP Classifier relies on multiple nodes and Neural networks to perform the classification. For the input to our machine learning models we had 221 features and output as two labels either green or black. We further divided our dataset into Train: Test and Validation sets. A ratio of 70:20:10 was selected for the split. In MLP Classifier Hyper-parameter tuning was done to optimize the nodes and layers. The tuning was optimized by using precision, recall and f1 scores as shown in Fig. 5. The optimized nodes and layers were found out to be: 7 and 29. The Prediction accuracy of different models is shown in Table 1. Among all the models used, MLP Classifier resulted in higher accuracy of 80.63 % compared to others. Similar results were found by Ali et al (2022) who found that the accuracy of deep learning models is higher than other existing traditional neural network and Support Vector Machine (SVM) and Nguyen et al (2024) who found that the Multi-Layer Perceptron (MLP) neural network demonstrated superior performance, achieving an overall accuracy of 95.9 % ± 0.26 % to detect early-stage fungal infections in bok choy ( Brassica rapa subsp. chinensis ) as compared to other tree-based methods or ensemble learning techniques. CONCLUSION It was found that among all the prediction Models used the Multi-layer Perception (MLP) Classifier model was the best for classifying the potatoes according to their internal defects best based on F1 score and accuracy score. This system can be used by Multi National Companies with the likes of Pringles and other potato processing companies for speed sorting and high-quality. The same setup can be extended to other crops. This system can also be utilized by companies like Zomato and Big Basket for food classification and selling premium and safe products. Use of this system in Sorting, Grading, and classification can help countries increase their export potential by supplying high-quality and safe products. Declarations ACKNOWLEDGEMENT We acknowledge the support of IIT Gandhinagar for carrying out this research. NIFTEM-P-2024-156. CONFLICT OF INTEREST Authors declare no conflict of interest. AUTHORS` CONTRIBUTION P. Priyam performed the analysis and drafted the article. R. Mishra performed the analysis. S. Varshney analyzed the data and drafted the article. S. Satyarthi analyzed the data. V. Goyal interpreted the data. Komal Chauhanrevievwd and edited the draft. S. M. Rafiq critically revised the article and finally approved the version to be published. ORCID ID S. M. Rafiq https://orcid.org/0000-0002-9178-989X References Adesokan M, Alamu E O, Otegbayo B and Maziya-Dixon B (2023) A review of the use of near-infrared hyperspectral imaging (NIR-HSI) techniques for the non-destructive quality assessment of root and tuber crops. Applied Sciences, 13(9), 5226 Afzaal H, Farooque A, Schumann A, Hussain N, McKenzie-Gopsill A, Esau T, Abbas F, Acharya B (2021) Detection of a Potato Disease (Early Blight) Using Artificial Intelligence. Remote Sensing 13(3) 411 Ali A, Mohsen A, Leila G (2022) Potato diseases detection and classification using deep learning methods. Multimedia Tools App 82 Charkowski A, Sharma K, Parker ML, Secor GA, Elphinstone JB ( 2020) Bacterial Diseases of Potato. In The Potato Crop: Its Agricultural, Nutritional and Social Contribution to Humankind (Campos H, Ortiz O, Eds.), Springer International Publishing: Cham, Switzerland, pp. 351–388 Gallagher NB (2020) Savitzky-Golay smoothing and differentiation filter. Eigenvector Research Incorporated Ibrahim A, Grassi M, Lovati F, Parisi B, Spinelli L, Torricelli A, Rizzolo A, Vanoli M (2020) Non-destructive Detection of Potato Tubers Internal Defects: Critical Insight on the Use of Time-resolved Spectroscopy. Adv. Hort. Sci 34 43–51 Imanian K, Pourdarbani R, Sabzi S, García-Mateos G, Arribas JI, Molina-Martínez JM (2021)S Identification of Internal Defects in Potato Using Spectroscopy and Computational Intelligence Based on Majority Voting Techniques . Foods 10: 982 Nguyen DHD, Tan AJH, Lee R, Lim WF, Wong JY, Suhaimi F (2025) Monitoring of plant diseases caused by Fusarium commune and Rhizoctonia solani in bok choy using hyperspectral remote sensing and machine learning. Pest Management Science 81(1): 149-159 Pahlawan MFR, Murti BMA, Masithoh RE (2022) IOP Conf. Ser. Earth Environ. Sci. 1018 Semyalo D, Kwon O, Wakholi C, Min HJ, Cho BK ( 2024) Nondestructive Online Measurement of Pineapple Maturity and Soluble Solids Content Using Visible and Near-Infrared Spectral Analysis. Postharvest Biol. Technol. 209 112706 Sharma A, Kumar R, Kumar N, Kaur K, Saxena V, Ghosh P ( 2023). Chemometrics Driven Portable Vis-SWNIR Spectrophotometer for Non-Destructive Quality Evaluation of Raw Tomatoes. Chemom. Intell. Lab. Syst. 242: 105001 Sharma A, Kumar R, Kumar N, Saxena V (2024) Machine Learning Driven Portable Vis-SWNIR Spectrophotometer for Non Destructive Classification of Raw Tomatoes Based on Lycopene Content. Vib. Spectrosc. 130: 103628 Weng L, Tang Z, Sardar MF, Yu Y, Ai K, Liang S, Alkahtani J, Lyv D ( 2024) Unveiling the Frontiers of Potato Disease Research through Bibliometric Analysis. Front. Microbiol. 15: 1430066 Zhou Z, Zeng S, Li X, Zheng J (2015) Nondestructive Detection of Blackheart in Potato by Visible/Near Infrared Transmittance Spectroscopy. Journal of Spectroscopy 1–9 Table Table 1. Prediction accuracy of different models S. No Models Prediction Accuracy % 1 Decision Tree Classifier 68.60 2 Logistic Regression 70.48 3 K Nearest Neighbors Classifier 70.67 4 Multi-layer Perception (MLP) Classifier 80.63 Supplementary Files Supplementaryfile.docx 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-7176490\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":490114872,\"identity\":\"5e4c3f88-e15f-4f7b-be65-c0bef873fa91\",\"order_by\":0,\"name\":\"Syed Mansha Rafiq\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA5ElEQVRIiWNgGAWjYJACZhDB3szA+ABI8/ARrYXnMAOzAYhmI17LAQY2CRCDoBbd9rOHPxfU3JPnYWfeVvk1x06GjYH54aMbeLSYnclLk55xrNiwh5mt7LbstmSgw9iMjXPwaTmQY8bMw5bAuJ+Zx+y25DYgG+gdabxazr8x/szzL8G+B6ilWHJbPRFabuQYSPO2JSSCtDB+3HaYGC1vzKR5+xKSgX4plmbcdpyHjZmQX87nAB32LcG2h//wxo8/t1Xb87M3P3yMTwsyMGDmAVHMRCoHa2H8QYLqUTAKRsEoGDkAABNcP8Ha1P0sAAAAAElFTkSuQmCC\",\"orcid\":\"https://orcid.org/0000-0002-9178-989X\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Syed\",\"middleName\":\"Mansha\",\"lastName\":\"Rafiq\",\"suffix\":\"\"},{\"id\":490114873,\"identity\":\"cc2d79dd-ed25-454a-a218-e97cb9fc5664\",\"order_by\":1,\"name\":\"Priyanshu Priyam\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Priyanshu\",\"middleName\":\"\",\"lastName\":\"Priyam\",\"suffix\":\"\"},{\"id\":490114874,\"identity\":\"4566b232-cd25-4b05-a4dc-703be547300f\",\"order_by\":2,\"name\":\"Raman Mishra\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Raman\",\"middleName\":\"\",\"lastName\":\"Mishra\",\"suffix\":\"\"},{\"id\":490114875,\"identity\":\"5f908aa1-5b9b-418d-990f-6b16774e3acb\",\"order_by\":3,\"name\":\"Sanskar Varshney\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Sanskar\",\"middleName\":\"\",\"lastName\":\"Varshney\",\"suffix\":\"\"},{\"id\":490114876,\"identity\":\"c50bfbac-1ba6-4ed9-968d-1f0304325e24\",\"order_by\":4,\"name\":\"Shrinath Satyarthi\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Shrinath\",\"middleName\":\"\",\"lastName\":\"Satyarthi\",\"suffix\":\"\"},{\"id\":490114877,\"identity\":\"440d5bbb-4edf-42a2-8a3f-09be73a47de1\",\"order_by\":5,\"name\":\"Vedant Goyal\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Vedant\",\"middleName\":\"\",\"lastName\":\"Goyal\",\"suffix\":\"\"},{\"id\":490114878,\"identity\":\"46cf196c-3b46-41ce-b888-9de4c46e8c0c\",\"order_by\":6,\"name\":\"Komal Chauhan\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"NIFTEM: National Institute of Food Technology Entrepreneurship and Management\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Komal\",\"middleName\":\"\",\"lastName\":\"Chauhan\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-07-21 10:49:58\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-7176490/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-7176490/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":87823825,\"identity\":\"426a26bd-fd6a-4689-852e-fbb299b8d14d\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:29:57\",\"extension\":\"jpg\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":70364,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eImage of freshly harvested potato induced with (a) internal greening defect and (b) internal black rot\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"1.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/4898ff07a5937e56ca7c3b15.jpg\"},{\"id\":87823929,\"identity\":\"17d0b454-fa0f-4498-bc87-2565007b4f47\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:30:25\",\"extension\":\"jpg\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":45789,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eImage of (a) NIR Sensor and (b) NIR signal acquisition measurement setup\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"2.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/ec7efb7ca356ab88d7ad9860.jpg\"},{\"id\":87823827,\"identity\":\"f38c9230-0fc0-490f-ad71-74cda98bacd3\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:29:57\",\"extension\":\"jpg\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":91689,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) A plot of absorbance vs. the wavelength plotted using python and jupyter notebook and (b) Pandas Dataframe containing labels and raw input\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"3.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/0720765b119e7960feb50924.jpg\"},{\"id\":87823904,\"identity\":\"7e143212-9106-4cd4-aa71-9f98e15bd3c2\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:30:22\",\"extension\":\"jpg\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":107120,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eSpectra showing absorbance intensity for (a) Black Spots and (b) Internal Greening\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"4.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/90381890faec6a500e814e74.jpg\"},{\"id\":87823829,\"identity\":\"8f24e815-df16-4a39-814c-fc6bfb2c4db6\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:29:57\",\"extension\":\"jpg\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":38783,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eOptimized nodes and layers in MLP Classifier\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"5.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/943604beed99233f69266f65.jpg\"},{\"id\":89572201,\"identity\":\"ab2da9e5-39eb-472b-95dd-e0d4242269c9\",\"added_by\":\"auto\",\"created_at\":\"2025-08-21 12:29:16\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":715422,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/a39117cf-c7c1-46a7-852f-bb4a84bef333.pdf\"},{\"id\":87824346,\"identity\":\"46045032-7ea5-432e-b2a5-30e65e460c6e\",\"added_by\":\"auto\",\"created_at\":\"2025-07-29 11:37:57\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":592879,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Supplementaryfile.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7176490/v1/c2ee3e87f3d30b192fb69c43.docx\"}],\"financialInterests\":\"\",\"formattedTitle\":\"Nondestructive Detection of Internal Defects in Potato by Visible/Near Infrared Spectroscopy\",\"fulltext\":[{\"header\":\"INTRODUCTION\",\"content\":\"\\u003cp\\u003ePotatoes (\\u003cem\\u003eSolanum tuberosum L\\u003c/em\\u003e.) play an essential role in providing food security but are susceptible to diseases and pests during the supply chain operations\\u0026nbsp;(Weng et al. 2024; Charkowski et al. 2020).\\u0026nbsp;Maintaining the quality of potato tubers throughout the production and supply chain is crucial to minimize economic losses (Imanian et al. 2021). Mostly these physiological disorders are restricted to their presence internally that gets unnoticed during grading practices. On the other end, conventional method of destructive analysis is time consuming and requires highly skilled man power to execute (Ibrahim\\u0026nbsp;et al.\\u0026nbsp;2020). The currently used non-destructive technique for potato defect detection is restricted to the use of colorimeter, where the results are limited to surface color characterization. Recently the use of imaging and spectroscopic techniques is increasingly being used to provide a non-destructive and objective method for grading the agricultural commodities including (Sharma\\u0026nbsp;et al.\\u0026nbsp;2024, Adesokan\\u0026nbsp;et al.\\u0026nbsp;2023).\\u0026nbsp;Advanced spectroscopic methods based on Vis/NIR (400-1000 nm) gets advantage with the ease of radiation penetration without having any hazardous impact and residue generation. The number of waves getting absorbed at a particular wavelength is indicative of the type of defect present inside the potato (Semyalo\\u0026nbsp;et al.\\u0026nbsp;2024, Sharma\\u0026nbsp;et al.\\u0026nbsp;2023)\\u0026nbsp;Employing machine learning algorithms such as classifiers assist in better prediction of response variables arising from spectroscopic observations.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003eVis/NIR spectroscopy demonstrates excellent ability to measure quality attributes of fruits and vegetables (Pahlawan et al. 2022) like brown heart in pears and apples (Afzaal et al. 2023 ) blackheart in potatoes (Zhou et al. 2015). \\u0026nbsp;Imanian et al. (2021) used the reflection mode of Vis/NIR and compared with short wave to classify the potatoes with hollow heart, and brown rot and found a classification rate of 83.4%. However, the internal defects which are commonly prevalent also include the greening and internal black spots, for which we could hardly find any available research. Hence we undertook this activity to detect the specific physiological disorder at wavelength scan of (513-850 nm) and used the machine learning algorithms to predict suitable wavelength and suitable classifiers were used for predicting the defect lot.\\u0026nbsp;\\u003c/p\\u003e\"},{\"header\":\"MATERIALS AND METHODS\",\"content\":\"\\u003cp\\u003e\\u003cem\\u003eSample acquisition \\u0026amp; preparation:\\u0026nbsp;\\u003c/em\\u003eThe study was conducted partly in NIFTEM-K and IIT Gandhinagar in 2022. Freshly harvested potatoes of variety (Kufri Chipsona) (400 numbers) were procured from Azadpur Mandi located in New Delhi and 200 potatoes were procured from SK Cold Storage located in Gandhinagar district of Gujarat. Samples were cleaned and washed with water to remove adhering soil matrix and dust particles. The potatoes with surface damages were sorted during the process, and the washed potatoes were labelled and immediately transferred to laboratory for the spectroscopic analysis.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003ePreparation of samples\\u003c/em\\u003e: An extensive literature review was done to understand the environmental conditions which could induce greening inside the potatoes. Accordingly, the potatoes were exposed to sunlight for 12 hours to induce greening effect (Fig. 1a). For the case of internal black rots, the disordered potatoes were directly procured from SK Cold storage which was being sorted and certified as being black rot by the industry (Fig. 1b).\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eInstrumentation:\\u0026nbsp;\\u003c/em\\u003eThe\\u003cem\\u003e\\u0026nbsp;\\u003c/em\\u003eNIR sensor used (NIRvascan Smart Near Infrared Spectrometer Reflective Model G1, Wavelength Range: 900-1,700 mm, 10 nm Pixel resolution 1.17nm/pixel\\u0026nbsp;(accuracy + or -0,7nm), Signal-to noise ratio:\\u0026nbsp;6000:1 in 1 second scan, Slit Size:\\u0026nbsp;1.69 x 0.025 mm) was available at IIT Gandhinagar (Fig. 2a). The\\u0026nbsp;source fiber was utilized to provide direct light from a DSN Xenon Arc lamp onto the surface of the potato at a distance of 15 cm. This was precisely achieved using an SP7 tip, which facilitated the focused delivery of NIR light to the targeted area on the potato\\u0026nbsp;(Fig. 2b). Concurrently, the detector fiber, situated adjacent to the source fiber, collected the NIR light that was reflected back from the potato. This step is critical as it involves the gathering of data that will be analyzed for potential defects. The NIR light captured by the detector fiber was then transmitted to the spectrometer. Here, the light underwent dispersion to separate it into its constituent wavelengths.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eData Analysis and Modeling:\\u0026nbsp;\\u003c/em\\u003eThe signal acquisition process took place at IIT Gandhinagar where we scanned the potatoes. \\u0026nbsp; Once the scan acquisition was done we used Python 3.7 and Jupyter Notebooks environment for data analysis. The software was installed using Anaconda on a desktop with windows 7 operating system. Three machine learning models, Decision Tree Classifier (DTL), Logistic Regression, K Nearest Neighbor (KNN), and a Deep Learning model, the Multi-Layered Perceptron (MLP) Classifier were used for classification and prediction of disorders.\\u003c/p\\u003e\"},{\"header\":\"RESULTS AND DISCUSSION\",\"content\":\"\\u003cp\\u003eWe had a total of 600 potatoes. The potatoes were radiated with four scans in order to ensure the rays falling to the targeted disorder spot. After performing 4 scans per potato, we had a total of 2400 scans. 400 scans were discarded due to poor signal quality. Out of 2000 scans 1200 were labeled as Black and the remaining 800 as Green.\\u0026nbsp;The readings were taken in the Vis/NIR range (400–1700 nm).\\u0026nbsp;Each scan was labeled as either internal black defect or internal green defect, depending on the truth value, which was determined by inducing the internal defects inside the potatoes. The obtained data was transferred using bluetooth to a mobile device in the form of a csv file\\u0026nbsp;(Fig 3a)\\u0026nbsp;which was then uploaded onto a server and was\\u0026nbsp;retrieved in Jupyter Notebook software compatible with multiple programming languages, including Python for further analysis.\\u0026nbsp;The retrieved data from cloud contained the values of absorption at 228 different wavelengths in the range of 600 to 1700 nm (Fig. 3a). For every 228 wavelengths there was a corresponding label, either black or green. Black was for potatoes with black internal spots (Fig. 4a) and green was for potatoes with internal greening defects\\u0026nbsp;(Fig. 4b).\\u0026nbsp;The absorbance intensity for green and black spots was different at the same wavelength\\u0026nbsp;(Fig. 3b).\\u0026nbsp;Absorbance values from the first 221 wavelengths were selected for calibration since above this threshold there was noise interference. Savgol Filter was applied on the raw data. Savgol filter is used to reduce high-frequency pitches and notes in the incoming signal using its smoothing properties and consequently help reduce low-frequency signals (e.g., due to offsets and slopes) using differentiation (Gallagher 2020).\\u003c/p\\u003e\\n\\u003cp\\u003eData analysis was done using machine learning programs. DTL is an inductive algorithm derived from instances, which rely on classification metrics displaying as decision trees derived from a group of disorder and irregular instances. Logistic regression is dominantly the most widely used methods in data science in binary and general model classification specifically. KNN is a widely popular technique popularly used for the purposes of model training and classification. MLP Classifier is a widely popular neural network. Unlike other classification algorithms such as KNN or Naive Bayes Classifier, MLP Classifier relies on multiple nodes and Neural networks to perform the classification.\\u003c/p\\u003e\\n\\u003cp\\u003eFor the input to our machine learning models we had 221 features and output as two labels either green or black. We further divided our dataset into Train: Test and Validation sets. A ratio of 70:20:10 was selected for the split. In MLP Classifier Hyper-parameter tuning was done to optimize the nodes and layers. The tuning was optimized by using precision, recall and f1 scores as shown in\\u0026nbsp;Fig. 5. The optimized nodes and layers were found out to be: 7 and 29. The Prediction accuracy of different models is shown in\\u0026nbsp;Table 1. Among all the models used, MLP Classifier resulted in higher accuracy of 80.63 % compared to others. Similar results were found by Ali et al (2022) who found that the accuracy of deep learning models is higher than other existing\\u0026nbsp;traditional neural network and Support Vector Machine (SVM) and Nguyen et al (2024) who found that the Multi-Layer Perceptron (MLP) neural network demonstrated superior performance, achieving an overall accuracy of 95.9 % ± 0.26 % to detect early-stage fungal infections in bok choy (\\u003cem\\u003eBrassica rapa subsp. chinensis\\u003c/em\\u003e) as compared to other tree-based methods or \\u0026nbsp;ensemble learning techniques.\\u003c/p\\u003e\"},{\"header\":\"CONCLUSION\",\"content\":\"\\u003cp\\u003eIt was found that among all the prediction Models used the Multi-layer Perception (MLP) Classifier model was the best for classifying the potatoes according to their internal defects best based on F1 score and accuracy score. This system can be used by Multi National Companies with the likes of Pringles and other potato processing companies for speed sorting and high-quality. The same setup can be extended to other crops. This system can also be utilized by companies like Zomato and Big Basket for food classification and selling premium and safe products. Use of this system in Sorting, Grading, and classification can help countries increase their export potential by supplying high-quality and safe products.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eACKNOWLEDGEMENT\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eWe acknowledge the support of IIT Gandhinagar for carrying out this research.\\u0026nbsp;NIFTEM-P-2024-156.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eCONFLICT OF INTEREST\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eAuthors declare no conflict of interest.\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cstrong\\u003eAUTHORS` CONTRIBUTION\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eP. Priyam performed the analysis and drafted the article. R. Mishra performed the analysis. S. Varshney analyzed the data and drafted the article. S. Satyarthi analyzed the data. V. Goyal interpreted the data. Komal Chauhanrevievwd and edited the draft. S. M. Rafiq critically revised the article and finally approved the version to be published.\\u003c/p\\u003e\\n\\u003cp\\u003eORCID ID\\u003c/p\\u003e\\n\\u003cp\\u003eS. M. Rafiq https://orcid.org/0000-0002-9178-989X\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n\\u003cli\\u003eAdesokan M, Alamu E O, Otegbayo B and Maziya-Dixon B (2023) A review of the use of near-infrared hyperspectral imaging (NIR-HSI) techniques for the non-destructive quality assessment of root and tuber crops. Applied Sciences, 13(9), 5226\\u003c/li\\u003e\\n\\u003cli\\u003eAfzaal H, Farooque A, Schumann A, Hussain N, McKenzie-Gopsill A, Esau T, Abbas F, Acharya B (2021) Detection of a Potato Disease (Early Blight) Using Artificial Intelligence. Remote Sensing 13(3) 411\\u003c/li\\u003e\\n\\u003cli\\u003eAli A, Mohsen A, Leila G (2022) Potato diseases detection and classification using deep learning methods. Multimedia Tools App 82\\u003c/li\\u003e\\n\\u003cli\\u003eCharkowski A, Sharma K, Parker ML, Secor GA, Elphinstone JB\\u003cstrong\\u003e (\\u003c/strong\\u003e2020) Bacterial Diseases of Potato. In The Potato Crop: Its Agricultural, Nutritional and Social Contribution to Humankind (Campos H, Ortiz O, Eds.), Springer International Publishing: Cham, Switzerland, pp. 351\\u0026ndash;388\\u003c/li\\u003e\\n\\u003cli\\u003eGallagher NB (2020) Savitzky-Golay smoothing and differentiation filter. Eigenvector Research Incorporated\\u003c/li\\u003e\\n\\u003cli\\u003eIbrahim A, Grassi M, Lovati F, Parisi B, Spinelli L, Torricelli A, Rizzolo A, Vanoli M (2020) Non-destructive Detection of Potato Tubers Internal Defects: Critical Insight on the Use of Time-resolved Spectroscopy. Adv. Hort. Sci 34 43\\u0026ndash;51\\u003c/li\\u003e\\n\\u003cli\\u003eImanian K, Pourdarbani R, Sabzi S, Garc\\u0026iacute;a-Mateos G, Arribas JI, Molina-Mart\\u0026iacute;nez JM (2021)S Identification of Internal Defects in Potato Using Spectroscopy and Computational Intelligence Based on Majority Voting Techniques\\u003cem\\u003e. \\u003c/em\\u003eFoods 10: 982\\u003c/li\\u003e\\n\\u003cli\\u003eNguyen DHD, Tan AJH, Lee R, Lim WF, Wong JY, Suhaimi F (2025) Monitoring of plant diseases caused by \\u003cem\\u003eFusarium commune\\u003c/em\\u003e and \\u003cem\\u003eRhizoctonia solani\\u003c/em\\u003e in bok choy using hyperspectral remote sensing and machine learning. Pest Management Science 81(1): 149-159\\u003c/li\\u003e\\n\\u003cli\\u003ePahlawan MFR, Murti BMA, Masithoh RE (2022) IOP Conf. Ser. Earth Environ. Sci. 1018\\u003c/li\\u003e\\n\\u003cli\\u003eSemyalo D, Kwon O, Wakholi C, Min HJ, Cho BK\\u003cstrong\\u003e (\\u003c/strong\\u003e2024) Nondestructive Online Measurement of Pineapple Maturity and Soluble Solids Content Using Visible and Near-Infrared Spectral Analysis. Postharvest Biol. Technol. 209 112706\\u003c/li\\u003e\\n\\u003cli\\u003eSharma A, Kumar R, Kumar N, Kaur K, Saxena V, Ghosh P\\u003cstrong\\u003e (\\u003c/strong\\u003e2023). Chemometrics Driven Portable Vis-SWNIR Spectrophotometer for Non-Destructive Quality Evaluation of Raw Tomatoes. Chemom. Intell. Lab. Syst. 242: 105001\\u003c/li\\u003e\\n\\u003cli\\u003eSharma A, Kumar R, Kumar N, Saxena V (2024) Machine Learning Driven Portable Vis-SWNIR Spectrophotometer for Non Destructive Classification of Raw Tomatoes Based on Lycopene Content. Vib. Spectrosc. 130: 103628\\u003c/li\\u003e\\n\\u003cli\\u003eWeng L, Tang Z, Sardar MF, Yu Y, Ai K, Liang S, Alkahtani J, Lyv D\\u003cstrong\\u003e (\\u003c/strong\\u003e2024) Unveiling the Frontiers of Potato Disease Research through Bibliometric Analysis. Front. Microbiol. 15: 1430066\\u003c/li\\u003e\\n\\u003cli\\u003eZhou Z, Zeng S, Li X, Zheng J (2015) Nondestructive Detection of Blackheart in Potato by Visible/Near Infrared Transmittance Spectroscopy. Journal of Spectroscopy 1\\u0026ndash;9\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"},{\"header\":\"Table\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eTable 1. Prediction accuracy of different models\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003ctable border=\\\"1\\\" cellspacing=\\\"0\\\" cellpadding=\\\"0\\\" width=\\\"100%\\\"\\u003e\\n \\u003ctbody\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 12px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003eS. No\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 51px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003eModels\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 35px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003ePrediction Accuracy %\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 12px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003e1\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 51px;\\\"\\u003e\\n \\u003cp\\u003eDecision Tree Classifier\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 35px;\\\"\\u003e\\n \\u003cp\\u003e68.60\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 12px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003e2\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 51px;\\\"\\u003e\\n \\u003cp\\u003eLogistic Regression\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 35px;\\\"\\u003e\\n \\u003cp\\u003e70.48\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 12px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003e3\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 51px;\\\"\\u003e\\n \\u003cp\\u003eK Nearest Neighbors Classifier\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 35px;\\\"\\u003e\\n \\u003cp\\u003e70.67\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003ctr\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 12px;\\\"\\u003e\\n \\u003cp\\u003e\\u003cstrong\\u003e4\\u003c/strong\\u003e\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 51px;\\\"\\u003e\\n \\u003cp\\u003eMulti-layer Perception (MLP) Classifier\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003ctd valign=\\\"top\\\" style=\\\"width: 35px;\\\"\\u003e\\n \\u003cp\\u003e80.63\\u003c/p\\u003e\\n \\u003c/td\\u003e\\n \\u003c/tr\\u003e\\n \\u003c/tbody\\u003e\\n\\u003c/table\\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\":\"info@researchsquare.com\",\"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\":\"Potato Internal Defects, Machine Learning, Deep Learning, Non-Destructive tests, Near Infrared, Sensors\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-7176490/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-7176490/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eThis research was carried out partly in NIFTEM-K and IIT Gandhinagar in year 2022 and investigates the application of non-destructive Visible/ Near-Infrared spectroscopy for detecting internal defects in potatoes. Conventional methods followed for defect detection are largely destructive, leading to significant material waste and increased operational costs. The study is designed to provide an innovative, efficient, and economical approach of NIR technology combined with advanced data analysis techniques to detect and segregate the quality potatoes. Specifically, it employs machine learning algorithms such as Decision Tree Classifier (DT), Logistic Regression, K Nearest Neighbor (KNN), and a Deep Learning model, the Multi-Layered Perceptron Classifier to categorize potatoes based on internal defects. The\\u003cem\\u003e \\u003c/em\\u003eNIR sensor (NIRvascan Smart Near Infrared Spectrometer Reflective Model G1) was used for detection. The spectrum was analyzed using specialized software Python 3.7 to be run in Jupyter Notebooks. A total of 600 potatoes were selected for the experiment. After performing 4 scans per potato we had a total of 2400 scans. 400 scans were discarded due to poor signal quality. 1200 were labeled as Black and the remaining 800 as Green. Dataset was divided into Train: Test and Validation sets. A ratio of 70:20:10 was selected for the split. The research identifies the Multi-Layered Perception Classifier as the most effective model, outperforming others in terms of F1 score and accuracy. The accuracy was found to be 81%. This breakthrough offers a scalable solution for enhancing potato quality assessment, with substantial implications for the agricultural industry and food supply chains.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Nondestructive Detection of Internal Defects in Potato by Visible/Near Infrared Spectroscopy\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-07-29 11:29:52\",\"doi\":\"10.21203/rs.3.rs-7176490/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"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\":\"15d90adf-9d8c-42fe-b256-fcac4242bf8a\",\"owner\":[],\"postedDate\":\"July 29th, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"posted\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2025-08-21T12:21:10+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-07-29 11:29:52\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-7176490\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-7176490\",\"identity\":\"rs-7176490\",\"version\":[\"v1\"]},\"buildId\":\"8U1c8b4HqxoKbykW_rLl7\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}