Static scene target detection based on VIBE algorithm

Abstract

In this paper, we analyze the foreground target extraction problem of surveillance videos and establish various mathematical models for different types of surveillance videos: two-frame difference model, Single Gaussian Model (SGM) and VIBE model, and apply these models comprehensively to achieve the extraction of foreground targets in surveillance videos with different background characteristics. On the basis of foreground target extraction, the effective features of the image sequence are extracted to determine the abnormal behavior of the crowd in the video.