Moving object tracking in video surveillance using YOLOv3 and MeanShift

Video surveillance is widely used and plays a huge role in society. Due to surveillance videos are often continuously produced, using these videos to track objects is a challenge for conventional moving object tracking methods. In this paper, in order to deal with the fast moving object and the problem of target occlusion, we propose an object tracking method based on YOLOv3 and MeanShift combined with Kalman filter aiming to improve the speed and accuracy of tracking. We use YOLOv3 to realize the detection and use the MeanShift combined with Kalman filter to track the target. The results of the experiment show that our method has achieved good results.

[1]  Jianhua Li,et al.  Evolution of crowd behavior in surveillance video , 2015, 2015 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting.

[2]  Zheng Tao,et al.  Research of meanshift tracking algorithm based on feature matching estimation , 2017, 2017 Chinese Automation Congress (CAC).

[3]  Jia Song,et al.  A New Approach to Track Moving Target with Improved Mean Shift Algorithm and Kalman Filter , 2012, 2012 4th International Conference on Intelligent Human-Machine Systems and Cybernetics.

[4]  Ali Farhadi,et al.  YOLOv3: An Incremental Improvement , 2018, ArXiv.

[5]  S. R. Balaji,et al.  A survey on moving object tracking using image processing , 2017, 2017 11th International Conference on Intelligent Systems and Control (ISCO).

[6]  Qingchang Guo,et al.  Research on the detection and tracking technology of moving object in video images , 2015, 2015 IEEE International Conference on Mechatronics and Automation (ICMA).