A shadow detection method based on improved Gaussian Mixture Model

The shadows of moving objects have great influence on the accuracy and effectiveness of objects tracking and behavior recognition, this paper proposes an elimination method based on Gaussian Mixture Model (GMM). First, we improve the adaptability of GMM by making learning rate change with the speed of the moving object to eliminate ghost. Then, we come up with a shadow elimination method based on normalized RGB space and segment shadows by their characteristics of brightness, color and the spatial relationship between shadows and moving objects. At last, under different light and projecting surfaces, we take a large number of experiments of moving objects, showing the method of this paper has good adaptability and robustness.