Robust Moving Object Detection and Shadow Removing Based on Improved Gaussian Model and Gradient Information
Zhiyan Bin, Yunyi Liu · 2010
A self-adaptive moving object detection and shadow removing algorithm for video surveillance is presented in this paper. The proposed algorithm improves the classic Gaussian Mixture Model to remove some unfavorable influences, such as sudden illumination changes and gradual variations of the illumination, ghost. In order to achieve the target, the learning rate is updated according to the illumination changes factor while the mean and variance of all distributions which match the new point are updated, and not only the highest weight of distribution. Moreover, a new effective algorithm for shadow removing in all kinds of complex scene is proposed, which combines HSV color information with RGB normalized space and first-order gradient information. Finally, the corresponding experimental results show that all the detection rates are over 90%, and the improved algorithm performs more robustly and powerfully than the classical Gaussian Mixture Model in moving objects detecting. The proposed algorithm can also effectively suppress shadow.