Robust Person Identification and Tracking in High-Exposure Environments Using YOLOv7

The identification and detection of individuals in well-lit environments, such as university hallways, pose distinct challenges stemming from fluctuations in shadows, luminosity, and contrast. In this investigation, we devised and assessed a system for person identification and tracking utilizing the YOLOv7 model applied to a tailored dataset of images collected in brightly illuminated college corridors. The principal aim was to attain a high level of accuracy in individual detection and subsequently establish a resilient person tracking framework. Our methodological approach encompassed a comparative analysis of two techniques—contrast stretching and adaptive histogram equalization—to enhance image quality and augment detection efficacy. The YOLOv7 model, trained on the COCO dataset, constituted the foundational framework for object detection. The findings revealed a remarkable detection accuracy of 99.7%, with adaptive histogram equalization exhibiting superior performance in addressing discrepancies in shadows, brightness, and contrast.