Robust low-rank sparse decomposition moving target detection method

Abstract

The invention discloses a robust low-rank sparse decomposition moving target detection method. A gamma norm is adopted to better approach a rank function in a traditional low-rank sparse decomposition method; a sparseness function in a traditional low-rank sparse decomposition algorithm is better approached by adopting Laplacian scale mix (LSM); a motion information vector matrix is introduced, so that whether a current pixel belongs to a motion target or not can be better judged in an auxiliary manner, and the precision of motion target detection is improved; and the introduced noise item can well express noise infection in nature, and the robustness of the robust low-rank sparse decomposition method in the aspect of moving target detection is improved.

Description

technical field [0001] The invention relates to the technical field of video processing, in particular to a robust low-rank sparse decomposition moving target detection method. Background technique [0002] Low-rank sparse decomposition has been widely used in moving object detection. The basic idea is to model and decompose videos of different scenes in reality into low-rank background and sparse foreground parts. However, this problem is an NP-hard problem. In order to solve this problem, the principal component tracking method was introduced and achieved good results, but this method is a biased estimate, which is not effective for severely corrosive video processing. Here In this case, methods such as truncated nuclear norm method and decomposition method are proposed, which further improve the defects of principal component analysis method. However, with the deepening of people's research, it is found that the previous convex function substitution effect is not good. I...

AI Technical Summary

Problems solved by technology

However, this problem is an NP-hard problem. In order to solve this problem, the principal component tracking method was introduced and achieved good results, but this method is a biased estimate, which is not effective for severely corrosive video processing. Here In this case, methods such as truncated nuclear norm method and de-decomposition method were proposed, which further improved the defects of principal component analysis method.

In this invention, the interference of noise items is not considered, and the noise matrix in the prior art is not suitable for the aforementioned target detection process. For example, the invention disclosed by the patent No. CN111429475A is based on a robust low-rank sparse decomposition. Although the background separation method can improve the robustness of the video foreground and background separation method, it cannot be directly applied to the target detection process to improve the F-measure value of the moving target and the robustness of the runtime calculation method.

Images