A moving object tracking method based on multi-feature fusion

Abstract

The invention relates to a moving object tracking method based on multi-feature fusion, belonging to the field of computer vision. At first, in that first frame image, the target area is initialized,and two position filters are respectively train by using the direction histogram and the color features; secondly, the detection samples of two features are extracted around the target in the subsequent frame, and the correlation scores between the two detection samples and the position filters trained in the previous step are calculated respectively, that is to say, the response diagrams of different features are obtained. Thirdly, according to the peak sidelobe ratios of different characteristic response diagrams, the two characteristic response values are weighted and fused, and the point with the largest response value is selected as the current center position of the target. Then the scale pyramid training scale filter is constructed by using the directional gradient histogram feature, and the maximum response point is obtained as the current scale of the target. Finally, according to the peak-to-side ratio of the final response graph of each frame, whether occlusion occurs or notis judged. In the case of occlusion, the position filter is not updated.

Description

technical field [0001] The invention discloses a moving target tracking method based on multi-feature fusion, which belongs to the field of computer vision. Background technique [0002] Object tracking is a hot spot in the field of computer vision, which is widely used in video surveillance, robot learning, industrial intelligence, etc. Its essence is to find the position and state of the target in a continuous video sequence image. Although object tracking has made great progress, it is still a challenging problem due to many factors such as occlusion, illumination change and scale change. [0003] In recent years, due to the remarkable effect of the correlation filter algorithm, many scholars have introduced the correlation filter into the target tracking framework. The feature selection in the correlation filter target tracking algorithm has a great influence on the tracking performance. Among them, the Minimun Output Sum of Square Error (MOSSE) algorithm proposed by ...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is to provide a moving target tracking method based on multi-feature fusion, which is used to solve the defects that the existing single-feature description target cannot fully express the target, and the tracking performance varies greatly in different scenarios, and Solved the problem that when the filter model is fixedly updated frame by frame, it is easy to add wrong information to the target model and cause tracking failure

Images