Lightweight high-precision pedestrian tracking algorithm in complex occlusion scenarios

• DOI: 10.3837/tiis.2023.03.009
• Publication Date: 2023-04-04T08:38:06Z
• Authors: Qiang Gao; Zhicheng He; Xu Jia; Yinghong Xie; Xiaowei Han
• Keywords: Tracking (education)

Aiming at the serious occlusion and slow tracking speed in pedestrian target recognition complex scenes, a method based on improved YOLO v5 combined with Deep SORT is proposed.By merging attention mechanism ECA-Net Neck part of network, using CIoU loss function non-maximum value suppression, connecting model Shuffle Net V2 as appearance feature extraction network to achieve lightweight fast purpose improving under occlusion.A large number experiments show that increases average precision by 1.3% compared other algorithms.The model, MOTA reaches 54.3% MOT17 data, accuracy 3.7% higher than related algorithms The presented this paper improves FPS nearly 5 fps indicator.