Prominence and Gaussian mixture model-based method for extracting foreground of surveillance video

Abstract

The invention relates to a prominence and Gaussian mixture model-based method for extracting a foreground of a surveillance video. The method comprises the steps of firstly extracting the foreground of a video to be analyzed by utilizing a Gaussian mixture model, then adopting a frame with smaller frame number of two adjacent frames with maximal color histogram Euclidean distance in the existing frame as a reference frame, differencing the existing frame and the reference frame to acquire a frame difference image, and then acquiring a masking image by performing the binaryzation. The Gaussianfilter based on a 3*3 template is undertaken for the masking image so as to eliminate the noise of the image. The existing frame region based on the masking image is acquired through the masking image, a prominence-based mask is acquired after the prominence extraction and binaryzation is performed for the region, a union set of the prominence-based mask and the Gaussian mixture model-based mask is solved, and finally a completer foreground region is acquired.

Description

technical field [0001] The invention relates to the field of digital video analysis, in particular to a monitoring video foreground extraction method based on saliency and mixed Gaussian model. Background technique [0002] Video surveillance system has been widely used in residential security monitoring, traffic conditions, airports, banks, etc., and will have a wide range of application prospects. Since most of the existing video surveillance systems only record and save the video, and do not play a real-time and active monitoring role, the intelligent video surveillance system developed on this basis uses video analysis technology to automatically realize the detection of targets. Identify and track, and make judgments about behavior based on that. The video surveillance system integrates video collection, detection, identification, and retrieval, which represents the development trend of the future surveillance field. However, due to the variety of video types, the appl...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a monitoring video foreground extraction method based on the saliency and the mixed Gaussian model for the incomplete foreground obtained by the existing foreground extraction method based on the mixed Gaussian model, so as to extract the foreground of the monitoring video more completely area, providing strong support for target recognition, tracking and retrieval of video surveillance systems

Images