Research On Level Set Based Object Contour Tracking Method

Visual tracking is a challenging research topic in the field of computer vision,and,in recent years, gain a lot of attention of scholars in image processing and computer vision field. Tracking is defined as locating the interest moving target in each image of a sequence of images. This technology focuses on tracking, monitoring automation and intelligence, it combines many acvanced technology in computer vision, pattern recognition, artificial intelligence, automatic control, mathematics and other fields. Moving target tracking is widely used in human-computer interaction, vision-based control, visual guidance of military, security monitoring, traffic control, medical diagnosis, video compression, and various aspects.Tracking algorithm directly affects the stability and accuracy of the moving target tracking system, although the theory of moving target tracking has been conducted for many years, it is so far still one of the hot research issues in the field of computer vision. The current tracking technology, can achieve a reliable tracking in the simple environment over the target with easy identified characteristics. However, in more complex environment or when the target is non-rigid objects or topology changing occurs, it is easy to lose targets. Studys found that the failure of target tracking algorithm is often caused by the mismatch between the target appearance and its dynamic model. The actual reasons behind is the lack of description from the objective model for the nature target, so when the target posture changes or occlusion occurs, it will bring the unmatch between the target model and the non-linear changes in target appearance. In addition, the target usually has a complex shape, but the traditional target tracking algorithm usually represent the tracking results by simple geometric shapes, it can not provide accurate input date for the subsequent upper application. Therefore, there are many technical difficulties to be resolved if we want to extend the target tracking technology into the complex reality application. Study a robust and precise, high performance object tracking method is still a great challenge in the research field.Based on the occlusion, the appearance and topology change and the precision problems in tracking field, this paper is committed to a method of combining image segmentation and video tracking technology. We can get the outlines of objects while the tracking process and can automatically handle topology changes. Meanwhile, the outline information is used to aid tracking algorithm as well as improved it, taking the appearance information, contour information and motion information into account to effectively address the the object characteristics changing problem and occlusion, and finally get a robust and efficient tracking algorithm which can both get the accurate tracking results and the object contour.