A Multiscale Parametric Background Model for Stationary Foreground Object Detection

Abstract

Detection of stationary foreground objects within a dynamic scene is one of the goals of a video surveillance system. A parametric background maintenance and updating scheme, based on a multiple Gaussian mixture model that operates on multiple time scales, is proposed. Each color cluster in the proposed model is assigned a weight which measures the time duration and temporal recurrence frequency of the cluster. Sudden illumination changes are handled by using an adaptive histogram template whereas gradual illumination changes are automatically resolved with the adaptive background model. Stationary foreground objects are detected by maintaining their temporal history in the dynamic scene at multiple time scales. Experimental results show that the proposed scheme performs well in three distinct real-world settings.