Fast Target Tracking Based on Improved Deep Sort and YOLOv3 Fusion Algorithm

Aiming at fast moving targets, such as ships, high-speed vehicles andathletes, this paper discusses a series of target detection algorithms based on neuralnetwork, YOLOv3 and background modeling. Compared KCF tracking with SSDtracking, Gaussian filter was applied to remove noise from pictures, and edgepreserving filter was used to preserve edge features. Moreover, the algorithmcombining deepsort tracking algorithm with YOLOv3 detection algorithm canimprove the accuracy of YOLOv3 target detection, solve the problem of targetloss during target tracking, adjust the frame size in real time, and improve the fitwith the target position. Experiments show that the proposed algorithm based ondetection before tracking has strong learning ability and robustness to unknownenvironment.