Kalman filter-based Detection and Tracking Method for Substation Moving Targets

In the substation scene with many occlusions, the traditional target tracking algorithm is prone to lose the figure, change identity, fail to recognize the target when it is shielded, and cannot achieve accurate real-time tracking. To address this problem, we proposed a sub-station operation target tracking method that combines metric learning and Kalman filtering. Firstly, the YOLOv3 algorithm with multi-scale features is used to detect the target. The Yolov3 target detection algorithm has high detection speed and strong real-time performance. On the basis of Yolov3 algorithm, the Kalman filter is introduced to predict the trajectory, and the metric learning method is used to fit the trajectory, and combined with the appearance of the target, to realize the tracking of the moving target. Finally, the experiment was carried out on the staff in the field. The tracking of moving target can achieve high accuracy.