Gaussian mixture model foreground detection method fused with image segmentation

Abstract

The invention relates to a foreground detection method and system for separating a moving object from a background from a sequence image or a video stream, in particular to a Gaussian mixture model foreground detection method fused with image segmentation. Firstly, a to-be-detected video is received as data input; the to-be-detected video is taken as a video image frame sequence, superpixel segmentation is performed on each video image frame on the video image frame sequence, and the constituent elements of each video image frame are converted into superpixels from original pixels; gaussian mixture model processing is carried out on the video image frame sequence formed by the superpixels, so that a foreground image and a background image are separated; and finally, morphological post-processing is carried out on the foreground image, and a required foreground image sequence is output.

Description

[0001] technology neighborhood [0002] The invention relates to a foreground detection method for separating a moving object from a background in sequence images or video streams, in particular to a mixed Gaussian model foreground detection method for fused image segmentation. Background technique [0003] Foreground detection has always been one of the difficulties and hot spots in the research of visual surveillance neighborhood at home and abroad. Its purpose is to extract the changed area from the background image from the sequence image. The effective detection of the foreground object is very important for object tracking, target classification, behavior understanding, etc. Post-processing is critical. [0004] Although the foreground detection algorithm has been continuously introduced since the 1960s and 1970s, so far, there is still no algorithm that is universal. Most of the foreground detection algorithms are specially proposed to solve a specific scene, but the ac...

AI Technical Summary

Problems solved by technology

[0005] At present, in the process of directly using the mixed Gaussian model for background modeling in foreground detection, the overall data calculation amount is too large due to the large number of calculation targets, but if the calculation targets are reduced, the accuracy of the mixed Gaussian model will be reduced due to the calculation targets In addition, the foreground object may be regarded as the background object and cannot be detected correctly when it is still or the motion range is small, and in the process of image processing, the "ghost" phenomenon also affects the foreground One of the main factors of object detection accuracy

Images