Accurate Ship Tracking in Maritime Images with Kernelized Correlation Filter and Curve Fitting

At present, the single tracking algorithm is generally used to achieve the single-target ship tracking task. Although the tracking result is ideal at this time, when the target ship is blocked by other obstacles, the tracking result will be greatly reduced. In order to solve this problem, this paper proposes a ship tracking framework based on nuclear correlation filter (KCF) and curve fitting. In the first step, the KCF algorithm is used to obtain accurate position data of the target target ship before and after occlusion and position data of the occluded stage. In the second step, according to the obtained position data, a trajectory conforming to the navigation of the target ship is obtained by a curve fitting method. In the third step, the proposed ship tracking framework is verified in several offshore video databases.