Underwater Small Target Tracking Algorithm Based on Diver Detection Sonar Image Sequences

To resolve the problem of small targets tracking in underwater surveillance system, an area-constrained global nearest neighbour data association method is proposed based on the analysis of the features of diver detection sonar images. It models the target size with Gaussian distribution and establishes data association cost matrix constrained with it, then obtains the optical data association resolution through extended Munkres method to track multiple targets, resolves the problem of false tracks in reverberation region by the judgment of trajectory "movement direction", updates the Gaussian distribution parameters according to data association results eventually. Experiments show it overcomes the difficulties of targets being small, instability and the absence of structure information under underwater low SNR environment, and is a practical method that is robust, fast in speed and good in veracity.