Surface Multi-target Tracking Algorithm Based on Data Association of Information Fusion

To solve the problem that the unmanned surface vehicle is vulnerable to external environmental impact in tracking multi-target, this paper proposes a water surface multi-target tracking algorithm based on information fusion data association. By transmitting the target motion information detected by the neural network to the motion measurement model, the target location is predicted. The Mahalanobis square distance between the observed and predicted values of the targets is calculated to represent the matching degree of the target motion information. In addition, the differences and similarities between the target features preserved in the trajectory and those detected in the current frame are calculated to represent the matching degree of the appearance information of the targets. The experimental results show that the proposed algorithm can track multiple targets effectively in case of wave fluctuations, illumination changes and occlusion. The tracking accuracy and precision are 66.8% and 82.5% respectively, which shows that the algorithm has good reliability.