A Moving Object Detection Method Combining Region Extraction and Improved Texture Features

Abstract

The invention discloses a moving target detection method combining region extraction and texture feature improvement, comprising the following steps: (1) collecting continuous images in a monitoring video as a sampling frame; (2) for each pixel in the sampling frame, Use its sampling information to carry out background modeling and restoration; (3) divide the image into blocks, use the statistical characteristics of the image block to extract the foreground area, and judge the illumination change of the foreground area to determine whether it is necessary to carry out binary image processing on the foreground area. (4) Perform accurate extraction of foreground pixels in the foreground area. The present invention greatly reduces the amount of calculation for subsequent accurate judgment by performing fast foreground area extraction, and eliminates both spatial displacement interference (shaking leaves, etc.) and brightness change interference (illumination change, etc.) Main interference: Accurate and efficient extraction of moving targets in image sequences.

Description

technical field [0001] The invention relates to a moving target detection method combining area extraction and texture feature improvement, belonging to the technical field of visual detection. Background technique [0002] Moving target detection technology based on image sequences is the basis of many high-level computer vision processing behaviors, such as target tracking, behavior understanding, abnormal behavior analysis, etc. The integrity and effectiveness of moving target detection results are crucial to follow-up research. At present, most moving object detection algorithms directly judge the image frames that need to be processed point by point, so as to realize the accurate detection of moving objects. This point-by-point judgment method is very susceptible to the influence of noise points (illumination, dynamic background, and imaging equipment errors, etc.), and a large number of dynamic noise points are misjudged as foreground, and in some moving scenes, the pr...

AI Technical Summary

Problems solved by technology

This point-by-point judgment method is very susceptible to the influence of noise points (illumination, dynamic background, and imaging equipment errors, etc.), and a large number of dynamic noise points are misjudged as foreground, and in some moving scenes, the proportion of moving objects The area is small, and point-by-point judgment will waste a lot of computing resources in some background areas that do not have obvious foreground features

Moreover, when performing noise interference elimination, most existing algorithms treat all interference points (illumination, swaying leaves, and noise points) as sparse noise for elimination. Methods that consider noise characteristics are limited for different types of noise cancellation

Images