Target Detecting and Target Tracking Based on YOLO and Deep SORT Algorithm

The realization of the 5G/6G network can ensure high-speed data transmission, which makes it possible to realize high-speed data transmission in the monitoring video system. With the technical support of 5G/6G, the peak transmission rate can reach 10G bit/s, which solves the problems of video blur and low transmission rate in the monitoring system, and provides faster and higher resolution monitoring pictures and data, and provides a good condition for surveillance video target tracking based on 5G/6G network. In this context, based on the surveillance video in the 5G/6G network, this paper implements a two-stage processing algorithm to complete the tracking task, which solves the problem of target loss and occlusion. In the first stage, we use the Yolo V5s algorithm to detect the target and transfer the detection data to the Deep SORT algorithm in the second stage as the input of Kalman Filter, Then, the deep convolution network is used to extract the features of the detection frame, and then compared with the previously saved features to determine whether it is the same target. Due to the combination of appearance information, the algorithm can continuously track the occluded objects; The algorithm can achieve the real-time effect on the processing of surveillance video and has practical value in the future 5G/6G video surveillance network.