Moving Object Detection Based On Low-Rank Recovery

Based on image processing, moving object detection is targeted to get the moving region extracted from the image background area. It provides reliable and accurate data basis for object identification, object tracking and object behavior analysis, and thus the results of moving object detection will help achieving corresponding research tasks like target tracking and objective analysis, and so forth. This paper studies the pre-process moving object detection and a novel moving object detection technology based on Sparse and Low-Rank Matrix Decompression.The main work of this paper includes:1. This paper presents a new de-nosing technology that can be used for the pre-processing work of moving object detection; it will also provide technical support for the later-on stage of moving object detection. This algorithm can effectively avoid existing problems in image de-noising algorithm such as thresholds selection, model parameter initialization and model complexity. By imitating human eye de-noising system, it achieves fast and efficient image de-noising.2. This paper uses a novel mechanism to represent the moving target, which takes advantage of the low rank characteristics of the background and the sparse characteristics of the moving target. The moving object detection is completed by the matrix decomposition. This paper uses a method called robust principal component analysis to achieve the matrix decomposition mechanism. It projects high dimensional observed data to some low-dimensional subspaces, namely using the low-rank characteristics in the video sequence to complete the background and foreground separation.3. The paper also elaborates on some typical moving object detection methods. Starting from some basic theories and algorithm assumptions, it compares the advantages and disadvantages of each method. In the experimental session, the video sequences of multiple scenes in different scenarios are tested to achieve the comparative analysis of experiment results. The experiment results prove that the low rank matrix and sparse matrix decomposition perform significantly well.