Correlation Filtering Object Tracking Algorithm Based on Motion Estimation

In order to solve the bounding box deviation while the object moves rapidly, which will cause the loss of the object, an improved Kernelized Correlation Filter object tracking algorithm is proposed based on motion estimation. The block-matching algorithm is used to estimate the motion of the tracking object in the adjacent frames of the video sequence, and the estimated results are quantified to judge the severity of the object’s motion. Then the size of the search box is adaptively adjusted. Two data sets of both OTB100 and OTB50 are selected for the experiments. The experimental results show that the improved algorithm can effectively solve the problem of object loss caused by rapid object movement and achieve long-term stable tracking of objects.