Multi-target tracking method and system for unmanned aerial vehicle aerial video

Abstract

A kind of multi-target tracking method and system for unmanned aerial vehicle aerial video, after target class and boundary box are extracted from video frame by multi-scale pixel-by-pixel target detection network MSPNet, further extract the apparent feature vector of target by multi-granularity fusion feature extraction network MaskMGN, calculate the apparent feature vector of trajectory by the weighted moving average method TCMWA based on time sequence and detection confidence, by apparent feature vector and Kalman motion model, data association is carried out to trajectory and detection result, and the result of multi-target tracking is obtained.The present application improves target detection and apparent feature extraction part, solves the problems of complex background, small target size, target occlusion, variable view angle in unmanned aerial vehicle aerial video, and can effectively improve the multi-target tracking precision of unmanned aerial vehicle aerial video.

Images