Moving Object Extraction in Complex Scenes

Moving object detection plays an important roll in intelligent monitor system; it may have a direct influence on the final detection result. In this paper, a self-adaptive system of moving object detection based on Gaussian mixture models (GMM) is designed, and this method can reduce some unfavorable influences, such as weather or lighting changes. Moreover, this paper modifies the algorithm of combination of background subtraction method and temporal differencing method, makes the contour of moving object more precise and removes environmental noise points effectively. Many experiments on outdoor video streams are tested and the results have shown that this method gives stable performance and good robustness.