Realtime Vehicle Tracking Method Based on YOLOv5 + DeepSORT

In actual traffic scenarios, the environment is complex and constantly changing, with many vehicles that have substantial similarities, posing significant challenges to vehicle tracking research based on deep learning. To address these challenges, this article investigates application of DeepSORT (simple online realtime a association metric) multitarget algorithm in tracking. Due strong dependence target detection, YOLOv5s_DSC detection YOLOv5s proposed, which provides accurate fast data algorithm. Compared YOLOv5s, has no more than 1% difference optimal mAP0.5 (mean average precision), precision rate, recall while reducing number parameters by 23.5%, amount computation 32.3%, size weight file 20%, increasing processing speed each image 18.8%. After integrating algorithm, + reaches up 25 FPS, system exhibits better robustness occlusion.