Moving Target Detection And Tracking For Intelligent Video Surveillance System

The detection and tracking of moving targets in video image sequences is the key technologies to implement an intelligent video surveillance. It is also a long-standing research focuses in computer vision. The main work and contributions in the thesis includes the following two aspects:(1) On the research of the moving objects detection, we firstly improved the update strategies of background model based on the analysis of the time difference method and background subtraction method. The update region and update time are limited. So the background will be updated for special purpose. Secondly by improving and optimizing the relative parameters of the Gaussian Mixture Model, moving targets will be effectively detected in complex scenes. Lastly, a fast shadow suppression algorithm in YUV color space was proposed, and the accuracy of the moving target detection was improved.(2) On the research of the moving objects tracking, based on the moving target detection, a cost function is constructed using the centroid, area and color histogram of the target. And the tracking of multiple moving targets is realized through Kalman filter. With the motion detection results as the mask image, the color distribution of the foreground is calculated using the idea of image pyramid. The computing speed of color histogram was improved. Moreover, the PTZ(Pan/Tilt/Zoom) will be turned automatically based on a single target tracking in practical use. Thus an object can be tracked continuously.The practical result of the system shows that the proposed algorithms meet the needs of real-time and have a good detection and tracking results.