Research Article
Multi-object Tracking Based on YOLOX and DeepSORT Algorithm
@INPROCEEDINGS{10.1007/978-3-031-36011-4_5, author={Guangdong Zhang and Wenjing Kang and Ruofei Ma and Like Zhang}, title={Multi-object Tracking Based on YOLOX and DeepSORT Algorithm}, proceedings={6GN for Future Wireless Networks. 5th EAI International Conference, 6GN 2022, Harbin, China, December 17-18, 2022, Proceedings, Part I}, proceedings_a={6GN}, year={2023}, month={7}, keywords={Multi-object tracking YOLO Deep convolutional neural network Kalman filter}, doi={10.1007/978-3-031-36011-4_5} }- Guangdong Zhang
Wenjing Kang
Ruofei Ma
Like Zhang
Year: 2023
Multi-object Tracking Based on YOLOX and DeepSORT Algorithm
6GN
Springer
DOI: 10.1007/978-3-031-36011-4_5
Abstract
The implementation of 5G/6G network provides high-speed data transmission with a peak transmission rate of up to 10 Gbit/s, which solves the problems of blurred video and low transmission rate in monitoring systems. Faster and higher-definition surveillance images provide good conditions for tracking multiple targets in surveillance video. In this context, this paper uses a two-stage processing algorithm to complete the multi-target tracking task based on the surveillance video in the 5G/6G network, realizing the continuous tracking of multiple targets and solving the problem of target loss and occlusion well. The first stage uses YOLOX to detect the target and passes the detection data to the DeepSORT algorithm of the second stage as the input of Kalman Filtering, and then use the deep convolutional network to extract the features of the detected frames and compare them with the previously saved features. The algorithm can better continuously track multiple targets in different scenarios and achieve the real-time effect of the processing of monitoring video, which has certain significance for solving the problems of large-scale dense pedestrian detection and tracking and pedestrian multi-object tracking for pedestrians in the future 5G/6G video surveillance network.
