MOTION TARGET SHADOW DETECTION METHOD AND SYSTEM IN MONITORING VIDEO

摘要

A motion target shadow detection method in a monitoring video. The method comprises: in a three-primary-colour colour ratio space of a motion foreground/video background in a monitoring video, establishing a 3D Gaussian probability model of a shadow and initializing a model parameter (S1); detecting a motion foreground region and updating the video background of a current frame image in the monitoring video, and mapping each pixel in the motion foreground region to the three-primary-colour colour ratio space of the motion foreground/video background (S2); judging each pixel in the motion foreground region obtained in step (S2) using the established 3D Gaussian probability model of the shadow, and detecting and differentiating a shadow pixel and a motion target pixel (S3); obtaining data of all shadow pixel points using detection in step (S3) and updating the parameter of the 3D Gaussian probability model of the shadow (S4); and returning to step (S2) to perform motion region shadow detection on a new frame of video image (S5). The method better solves the problems of difficulties in setting parameters in shadow detection technology for a motion target, the failure to self-adapt a scene lighting change and a large volume of computation in existing monitoring video.