Study Of Moving Target Detection And Tracking In Video Based On Trajectory

Moving target detection and tracking is an attractive research subject of computer vision, which has broad application in many areas,such as intelligent surveillance, military control, medical diagnosis, multimedia retrieve and so on. It has important practicality values and bright future prospects. To overcome the obstacles which are caused by occlusion, illumination changes and noise in images, this paper proposes moving target detection and tracking algorithm based on trajectory.The author introduces an image segmentation way based on principal color model to get candidate areas. First of all, the images are converted from RGB space to HSV space to exclude the effect of illumination;then field region is extracted with hue and saturation components and the resultant image is binary; then through hough transformation, the mark lines are detected and erased; then after morphological filtering, we can get the image which is segmented. With contour and area features extracted from the candidate areas, the candidate objects are selected.Then this paper proposes moving target detection and tracking algorithm based on trajectory to predict and verify the candidate objects. Duing this process, each candidate is tracked by Kalman filter in successive frames, which produces many initial trajectories; through trajectory selection with Dijkstra algorithm, the true ball trajectories are verified, which indirectly get the ball position in the frames. Further more, combining Kalman filter prediction and linear interpolation, some missing balls can be made up.Experiments show that this algorithm can effectively overcome partial and full occlusion and illumination changes and achieve good performance in real-time object tracking.