A Robust Integrated Framework for Object Detection in Fisheye Traffic Surveillance Systems

Wide-area traffic monitoring benefits significantly from the use of fisheye cameras, which can capture extensive visual fields from a single observation position. Yet, the strong radial distortion and uneven resolution that are inherent to fisheye images bring notable challenges to standard object detection models-particularly near image boundaries, where the appearance of objects is severely degraded, making accurate recognition difficult. To address these issues, this study develops a detection framework engineered to maintain robust performance under the aforementioned challenging conditions. The framework integrates a simple but effective pipeline of preprocessing and postprocessing steps, which enhances the consistency of detection results across the entire image, with a focus on regions affected by severe distortion. Additionally, we train multiple state-ofthe-art detection models using fisheye traffic imagery and merge their outputs through an ensemble strategy to boost overall detection accuracy. When tested on the 2025 AI City Challenge Track 4, the proposed method achieved an F1 score of 0.6366 and secured the 8th rank among 62 participating teams. These results validate the effectiveness of the framework in resolving the inherent problems of fisheye imagery for traffic surveillance applications. Supplementary Material File (manuscript2.pdf) - Download - 1.92 MB Information & Authors Information Version history Copyright This work is licensed under a Non Exclusive No Reuse License.

Authors Metrics & Citations Metrics Article Usage 136views 102downloads Citations Download citation Yehuda Callen, Sandip Khan. A Robust Integrated Framework for Object Detection in Fisheye Traffic Surveillance Systems. Authorea. 02 December 2025. DOI: https://doi.org/10.22541/au.176463776.61525977/v1 DOI: https://doi.org/10.22541/au.176463776.61525977/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu.