Pedestrian detection and counting method based on YOLOv5+DeepSORT
Xiaofeng Qiu, Xiangrui Sun, chen yongchang, Wang Xinyan · 4th International Symposium on Power Electronics and Control Engineering (ISPECE 2021) · 2021
To alleviate the spread of the epidemic, most public places have begun to limit the number of trips. Therefore, this article proposes a pedestrian counting scheme based on YOLOv5 and DeepSORT for multi-target detection and tracking. Using the network weights trained by the coco data set and combining the YOLOv5 detector and the DeepSORT tracker, the pedestrians are detected and tracked, and the number of people entering and leaving is calculated, thereby realizing the control of the number of floating people. Through experiments on streets and subway stations, it is proved that this algorithm is suitable for tracking and counting high-density people, and based on ensuring the real-time performance of the system, it provides high system accuracy and robustness.