Research On Moving Object Detecting And Tracking In Video Sequence Images

Moving object detecting and tracking in video sequence images is becoming a hot issue in computer vision, and it has extensively applied in video surveillance, human-machine interfaces, virtual reality and image coding. However, tracking objects accurately in various complicated environment and different conditions, such as occlusion and illumination change, is a common focus and an essential and challenge problem.A common moving object tracking system is usually composed by two major parts, moving object detection and extraction, and moving object tracking procedure. In this paper, the background and importance of moving object detecting and tracking is first described, and then the nowadays study situation and some difficult problems are illustrated, and then a detailed introduction of the whole moving object detecting and tracking is given as follow.On the research of the motion detection, three main methods of motion detection, optical flow, inter-frame difference and background subtraction, are introduced. Firstly the basic theory of them is listed, especially that of background subtraction, which is more common. Creation and update of the background in background subtraction method is a key issue for solving the whole object detection problem. Because of the complication and variety of the surrounding environment, it is difficult to obtain a correct background image. Then their own analyzed advantages and drawbacks are researched. Some experimental results show the comparison of inter-frame difference and background subtraction and some actual problems when adopt them. Finally some other approaches are introduced for getting a common object detection algorithm.On the research of the moving object tracking, two common object tracking frameworks are described first, Kalman filter and Mean Shift method. Kalman filter is a special case of Bayesian filters and is the best possible estimator in the condition of linear and Gaussian, and it is why Kalman filter is used in many object tracking systems for motion estimation. Mean Shift method is extensively applied in object tracking. Because of its simplicity and low computation, it is usually combined of other methods for better object tracking result. In this paper, some experiments are done to identify their accuracy and stability respectively. And then a novel first-detect-then-identify approach with SIFT features and discrete wavelet transform for tracking object is proposed in real surveillance scenarios. For accurate and fast moving object detection, discrete wavelet transform is adopted to eliminate the noises of the frames which may cause detection errors, and then objects are detected by applying the inter-frame difference method on the low frequency parts of two consecutive frames, and then SIFT features of an object are used for object representation and identification due to their invariant properties. Experimental results demonstrate that the proposed strategy improves the tracking performance by comparing with the classical mean shift method, and the proposed algorithm can be also applied in multiple objects tracking in real scenarios.