A Real-Time Target Tracking Algorithm Based on Improved Kernel Correlation Filter

Abstract

At present, target tracking technology is a popular computer vision technology, which is of great significance in people's production and life. But there are still some important and difficult problems in the field of target tracking, such as background similarity, target occlusion, motion blur and so on. Although the target tracking method based on deep learning has good robustness, its off-line training process costs a lot of time and is not flexible enough to use. In this paper, starting from the background requirements of shipboard platform sea target tracking, the algorithm is improved based on kernel correlation filtering algorithm, and a set of high robust real-time target tracking algorithm based on online learning method is proposed.