Recursive-learning-based moving object detection in video with dynamic environment

Moving object detection is a fundamental and critical task in video surveillance systems. It is very challenging for complex scenes having slow-moving and paused objects. This paper proposes a moving object detection algorithm which combines Gaussian mixture model with foreground matching. This algorithm is able to detect slow-moving and paused objects very effectively. This algorithm uses adaptive learning rate to deal with different rates of change in background. The performance of the proposed algorithm is evaluated on the challenging videos containing strong dynamic background and slow-moving and paused objects using standard performance metrics. Experimental results show that the proposed method achieves 25 % average improvement in accuracy compared over existing algorithms.