Automatic Target Segmentation and Tracking in Intelligent Video Surveillance

According to complex environmental identification for the problem of multi-target mobile and stationary state in a fixed scene, this paper provides a technology framework for the forthcoming coexisting multiple target detection and tracking, extracts video frame background map using the mean of multiple video sequences, makes edge detection of the background picture, judges the static target using positive and negative samples of the background image, and uses the Gaussian distribution of background difference to realize the extraction of foreground image. We utilizes the comprehension of non parametric statistical model and color model to eliminate the shadow of foreground image. After eliminating the shadow of the foreground image,we uses Mean Shift dynamic target tracking method to track the interesting objects. Experimental results show that it has strong robustness and stability for the influence of light slow change, occlusion and motion state.