Real-time object tracking method based on low-rank correlation information among video frames

Abstract

The present invention discloses a real-time object tracking method based on low-rank correlation information among video frames. According to the method, a video is divided into frame segments with each frame segment comprising k frames, low rank analysis is performed on the divided video frame segments with consistent background, and therefore a trajectory of a marked object can be obtained. Themethod comprises the following steps: low-rank analysis based on bilateral random mapping is used to accurately track short-term stable video frame segment background; in order to eliminate side effects caused by the visual changes resulted from observation angle changes, prior low rank background is combined with a previous frame segment to solve of robust significant motion detection-based RPCAlow rank recovery of a current frame segment; a series of feature clues extracted from a stable frame segment are used to speed up updating of low-rank background information and overcome some stubborn problems effectively. Via the method, moving objects can be accurately detected from a video, and the same effect can be attained in a dynamic video; the method is characterized by high detection rate, good robustness, strong anti-interference and the like.

Description

technical field [0001] The present invention relates to the technical field of significant motion detection, in particular to a real-time object tracking method based on the consistency of low-rank correlation information between video frames, which is a low-rank analysis based on a block solution strategy and the use of time and space between video frames A Salient Motion Detection Approach for Semantic Connections. Background technique [0002] Salient motion detection is a very popular research direction in the field of computer vision and pattern recognition. It is of great help to many specific application packages, including object tracking, video surveillance, traffic control, intrusion detection, etc. Although the research on salient motion detection has achieved great results in recent years, some traditional limitations are still not well resolved, especially for non-static videos, which greatly limit the intelligence based on motion tracking. The application rang...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved in the present invention is: solve the problem that the background in the long video will change through the strategy of segmentation, process according to the characteristics of the background change of the frame in a short period of time is relatively small, and select continuous m frames with relatively consistent The background has the characteristics of low rank; Correction RPCA is used to correct the background to eliminate other interference factors, such as camera shake, viewing angle changes, etc.; the background information of the previous frame segment is introduced into the process of low-rank decomposition in this frame segment , use the correlation between video frames and frames to solve some noise and significant motion interference; use the difference between dynamic background and static background and the change amplitude frequency of moving objects to distinguish whether it is dynamic background, to achieve the effect of preparing to determine significant motion

Images