An AUV Tracking Algorithm Based on the Scale-Adaptive Kernel Correlation Filter

With the widespread use of Autonomous Underwater Vehicles (AUVs), underwater object tracking has become a popular research problem in the field of underwater robotics. Due to the complexity of the underwater environment and the ambiguity of underwater visual sensors, underwater object tracking faces great challenges. To address this problem, this paper proposes an object tracking process for AUVs based on correlation filters. First, underwater images are pre-processed to recover image features before conducting object tracking. Then, the object selected in the first frame is scaled and a filter is trained to extract the object’s feature information to match objects in subsequent frames. The best scale is obtained by matching the object features with those in the scale template library. Based on the principle that objects appear larger when near and smaller when far, distance perception can be achieved by analyzing the object scale, thus enabling fixed-distance tracking using monocular vision. It is experimentally verified that the tracking process proposed in this paper demonstrates high performance in tracking accuracy, response speed, and success rate. This has strong potential for underwater object tracking applications and can provide effective technical support for the practical application of underwater robots.