Abstract:

To achieve effective object detection in the intelligent surveillance, a comprehensive dynamic background difference and shadow removal method is proposed. Multilevel information is adopted to update the Gaussian Mixed Model to acquire color background image with high quality. Shadow is removed by a progressive pixel classification with proper RBG, Hue and Saturation information while extracting the foreground. Based on these results, object tracking and abnormal behavior detection are realized in an intelligent surveillance system. The experimental results with different scenes show that our method could satisfy the demand of practical applications with good performance.

Key words: intelligent surveillance;background extraction;shadow removal;abnormal behavior