Moving object detection based on an improved gaussian mixture background model
Rui Yan, Xuehua Song, Shu Jia Yan · 2009
When background subtraction method is used to detect moving objects, illumination changes can easily impact the detection. In order to deal with the problem, a novel algorithm which synthesizes the methods of background subtraction and adjacent-frame difference is proposed. This algorithm adopts Gaussian mixture model to reduce the impact of background disturbance, and uses adjacent-frame difference for reference. It deals with illumination changes by background reconstruction and the function of dynamic learning efficiency. The algorithm is simulated when background is disturbed and illumination changes. The results show that the algorithm is more efficient and more robust than traditional methods, and it can attains background model in complex conditions. The algorithm is very suitable for intelligent video systems with static cameras.