Research on Target Detection and Tracking Algorithm in UAV Visual Perception System

In order to meet the requirements of UAV vision sensing system, an improved algorithm, named YOLO + Deep Mix SORT, is presented, which combines YOLO (You Only Look Once) and Deep SORT (Simple Online and Realtime Tracking). The algorithm utilizes YOLO's real time target detection and Deep SORT's online tracking capability, and improves the detection precision and the tracking stability. YOLO + Deep Mix SORT algorithm is analyzed in detail, and an improved feature extraction module and motion forecast model is used to detect and track multiple targets under complicated dynamic environment. Finally, the validity of the proposed algorithm is proved by the actual UAV video data. The simulation results indicate that the proposed algorithm has higher detection precision, continuous tracking and processing speed than the conventional target detection and tracking algorithm. The actual data analysis indicates that the precision of the improved algorithm is about 8.5%, and the tracking stability is 12.3%, and the real time detection is realized.