A Moving Object Detection Method Based on Adaptive Parameters

Abstract

The present invention provides an adaptive parameter moving target detection method. According to the present invention, an improved mixed Gaussian background update algorithm is used for determining whether a scene is changed by a change degree of a proportion of foreground point pixel numbers in an whole image, the background is partitioned into a changed region and a non-changed region, different update rates and Gaussian distribution numbers are adopted for the changed region and non-changed region, and a new background model is obtained, thereby providing convenience to obtain a moving target in the new background model. According to the present invention, background update efficiency can be improved and the smear phenomenon and the detection missing phenomenon can be effectively inhibited.

Description

technical field [0001] The invention belongs to the field of target detection, and in particular relates to a moving target detection method with adaptive parameters. Background technique [0002] Target detection technology is the technology of segmenting and extracting the target from the image background in the sequence image. It is the key technology in the target search and tracking system, and provides initial information for subsequent target recognition, target tracking, track association and other technologies. The effect of target detection directly affects the entire recognition and tracking system. [0003] At present, moving object detection has been widely concerned and fully developed in the field of image processing. However, how to effectively detect moving objects in complex environments is still a difficult point in moving object detection. Among the current mainstream moving target detection algorithms, the hybrid Gaussian background update algorithm ha...

AI Technical Summary

Problems solved by technology

However, the literature>The traditional mixed Gaussian background update algorithm used in the

For example, a fixed number of Gaussian distributions and a fixed update rate are used in the background update, and each image must be updated, which leads to three shortcomings of the algorithm: First, after the background update is completed, it reaches After stabilization, the mixed Gaussian background model continues to update the background of the image, causing unnecessary waste of time and affecting the real-time performance of moving target detection to a certain extent; second, when the target changes from static to moving, there will be a delay Third, when the target changes from moving to static in the scene, the target will gradually blend into the background, causing the target to be lost, resulting in missed detection

Images