Real-time moving object detection under complex background

Moving object detection (MOD) is a basic and important problem in video analysis and vision applications. In this paper, a novel MOD method is proposed using global motion estimation and edge information. In order to get more robust MOD results under different backgrounds and lighting conditions, a bilinear model and histogram scaling method are used respectively for spatial and illumination normalization. After normalization, edges are extracted by Canny and further filtered using morphological operators to get closed object contours. The final objects are extracted by combining the contours and moving regions from motion detection. The experimental results show the proposed approach has apparent advantages in robust and accurate detection and tracking of moving objects with changing of camera positions, lighting conditions and background for real-time applications.