Robust Hybrid Foreground Detection and Adaptive Background Model

Abstract

Identifying moving objects in video sequence is fundamental and important task in visual tracking systems and computer vision applications. Background removal algorithms are usually used to separate the foreground from the background. Despite the existence of many background removal algorithms, they didn't solve some problems such as cast shadows, highlighting and ghost effect. In this paper we propose, a robust foreground detection and background maintenance algorithm based on hybrid background removal methods in well known color space (RGB) combined with motion information. Numerous experiments were performed with different scenes showing that our system is robust to various types of background scenarios. We compared our system with different background removal algorithms and a promising performance was achieved, especially in the efficiency in detection with respect to frame rate performance.