Research on data association and detection algorithm in point target tracking

In the field of computer vision, point target tracking has always been an important topic and research hotspot, and it is widely used in both military and civilian fields. For the tracking of point targets under complex background, the point targets are extremely small, and their morphological characteristics are not obvious, so they are easily disturbed by background and noise. Secondly, the point targets’ maneuvering, shaking of detection equipment, etc., will change their morphology, resulting in low detection rate and high false alarm rate, which will further affect the accuracy and robustness of point target tracking. Therefore, how to effectively utilize the spatio-temporal information in sequence images to extract the target accurately is a difficult problem. This paper summarizes the existing detection and data association algorithms in point target tracking, analyzes their performance and shortcomings, and discusses the development direction of point target tracking algorithm, that is, algorithms based on multi-feature fusion with strong robustness, high accuracy and small calculation.