Moving object segmentation based on background subtraction and fuzzy inference

In order to improve the segmentation accuracy, reduce under-segmentation and over-segmentation, this paper proposes a new algorithm for detecting moving objects. The method is based on background subtraction algorithm and integrated with fuzzy inference for thresholding and background update. We use 7 fuzzy rules which can effectively model the membership of a pixel in a moving object during the fuzzy inference. The inference algorithm is both pixel-based and region-based. It properly segments the moving object from the stationary background. Moreover, the background model is updated by fuzzy logic with dynamic update rate over time to overcome the noise and illumination changes, which occurs frequently in complex natural environments. So the algorithm is suitable for a long run without losing accuracy. The experiment results show that our method is robust as well as fast in performance.