Real-Time Tracking of Fast Moving Weak Object Based on Siamese Network

Aiming at the problem that the existing target tracking algorithms have poor effect on fast moving weak targets, a spatio-temporal continuous multi-feature fusion siamese network algorithm is proposed. First, the full convolution siamese network is used as the basic framework; second, a robust feature combining spatial information and semantic information from coarse to fine is designed to express fast-moving weak targets, and feature attention is added; Finally, the spatio-temporal information continuity model is used to effectively update the overall information, so as to select the best tracking target. In the fast moving weak target tracking sequence, compared with five different feature selection and update algorithms, the proposed algorithm shows good real-time tracking effect; the proposed algorithm is compared with 9 different algorithms and 5 similar twin network algorithms, and the comprehensive performance of the proposed algorithm is excellent. Experimental results show that the proposed algorithm has good robustness and real-time performance, and can effectively track fast moving weak objects.