Research On The Technology Of Moving Object Tracking Based On Visual Image

In recent years, the technology of moving object tracking based on visual image is a hot research topic in the field of computer vision. Thus, object tracking using visual image has been widely used in security surveillance, self-driving vehicles, navigation, guidance and control, human-computer interaction. It is difficult to track object accurately and spontaneously on the influence of fast moving, occlusion, deformation and various types of background noise. Hence, the considered research has the important theory and application value.Firstly, the development of object tracking is introduced for existing research results, and the object detection and object tracking of existing methods are analyzed, and the technology difficulties of object tracking are given.Secondly, the moving object tracking algorithm which is combined with visual images and kalman filter is studied. It is necessary to preprocess the object image in order to improve the quality, then the moving object region is extracted from the background by moving object detection; On this basis, kalman filter is used to predict the location of object, and experimental results show that object tracking using centroid is effectively implemented.Subsequently, the moving object tracking algorithm which is combined with visual images and meanshift algorithm is studied. Because meanshift algorithm cann't track the fast moving object, an algorithm combining kalman filter and meanshift is studied, kalman filter is used to forecast possible position of object, then menashift search the real position near the possible position.The algorithm has a good effect on fast moving object,and can solve the occlution well. And the tracking window's self-adapting further improves the meanshift algorithm, which significantly increase adaptability and accuracy of the tracking algorithm. The adaptive updating method of tracking window further improves the meanshift algorithm, which greatly improves the accuracy and adaptability of the tracking algorithm.Following, according to the research of meanshift algorithm, the meanshift algorithm is extended to continous adaptive meanshift (CamShift) in order to overcome the lack of the essential model updating in meanshift algorithm; Similarly, CamShift algorithm is combined with kalman filter for moving object, by improving the search strategy to achieve the stability of fast tracking a moving object, and the satisfying results are obatined. Finally, the moving object tracking algorithm which is integrated with visual images and partiele filter is studied. When the object is occluded and exist various types of noise, the template update algorithm is proposed to improve the robustness of algorithm. However, in order to overcome the problem of the degradation phenomena and the computational cost of the particle filter, the meanshift algorithm is combined with the particle filter algorithm, the results show that the tracking algorithm can maintain real-time and robustness under the condition of fast moving, occlusion, deformation and various types of noise.