Moving Target Detection And Tracking

Moving object detection and tracking is one of active branches of imaging processing and computer vision, which study how to extract moving object region, describe moving feature and match the object, and estimate the concerned moving parameter in serial video image. This technology has been applied widely in military application and sociality this paper does some research in video image processing, as follow.1. Multi-motion object detection and tracking under fixed background On the condition of fixed background, mainly techniques are motion region extractions, feature description ,feature recognition and moving analysis . As for the motion regions extraction techniques, one commonly used method in most papers is called adjacent difference or optical flow field iterations. Another alternative is consists of two steps, the first step is adjacent difference, the second step is to search the motion regions in the frame being processed. The first scheme will result in overlapping phenomenon, and the second one is not available in complex background or in a scene where the processed target are similar each other and very near. In this paper, in order to reduce the errors resulting from the adjacent difference, we adopt the video sequence hierarchy segmentation theories to reconstruct background of the sequence under fixed background cases. Then the background difference is used to enhance the detection precision. According to the theory analysis and computer simulation, the method reduces the errors greatly, so do the estimation precision.Under some circumstances, we need to describe the geometry parameters of binary multi-target regions, such as the square, perimeter, minimal rectangle enveloping the target, and centroid. At the same time, dominant points is useful for their visual feature stability in feature match. In this paper, we propose a simple dominant points algorithms in the cases where the dominant points are very distinctly. Later , the corresponding matching algorithm is also proposed, which provides an excellent way to motion estimation problems concerning onthe dominant points detection.In the multi-target motion states analysis systems, most papers only deal with the tracking problems. Yet, the intelligent tracking technologies not only need the tracking algorithms, but also to detect various motion states occurring in the motion analysis procedures concurrently. In this paper, we use the motion state-discrimination equation to identify the motion states. Then various processing methods corresponding to the different states are adopted. The technique allows the existence of various motion states, track many motion states correctly, including the normal motion states, target disruption, target fusion, entering the scene, and departing the sceneBased on the paper theories, a traffic parameter detection system is setup. The system could performs many functions, includes the individual vehicle speed estimation, the queue length calculation, the average speed estimation. As for the license recognition, the paper systematically study the techniques of the orientation of license plate, the character extraction and emendation, this technique overcome many unfavorable factor in the segmentation and acquired good effects. 2 Moving object detection under moving cameraWhen object is moving, the camera is also shifting as its posture or position is changed. So the changes of moving object and changes of background will mixed together. In order to detect the moving object, the paper proposes the Moving image target detection and extraction technology based on sequence image compensation difference. This paper produces two algorithm working in 2D+T three dimension information space to detect image target in complex dynamic scene. One is four parameter method based on rotate and parallel, the other is six parameter method based on orthotropic projection. These two algorithm compensates the global motion made by camera motion in time domain. After these algorithm extract image edge in 2D space...