Research Of Motion Tracking System Based On Active Contour Model

Moving target detection and tracking from image sequences are two of the main contents of computer vision and image coding. It relates to knowledge of multiple disciplines, such as image processing, pattern recognition, automatic control, applied mathematics and artificial intelligence. It is used in the fields of robotics, intelligent control systems, medical image analysis, and video compression and transmission.Active contour model is one of hot topics in the research of the image and computer vision in recent years. It combines the image features and the prior knowledge. Also it is a challenge to Marr's computational vision theory, where the vision system is divided into three bottom-up levels with information transferring in a single direction. This dissertation focuses on the application of motion tracking by active contour model.Firstly, in this thesis, a thorough study of Snake model is given. Its features and shortcomings are discussed. And the method of tracking a moving object with the GVF Snake is put forward. Secondly, the active contour model is sensitive to its initial contour, so Kalman filter is used to predict the trajectory of the moving object, in order that the initial contour would be more closely to the object's actual contour. Finally, a platform with a video capture and processing system based on DSP is set up. Motion detection program is implemented into the system.In order to make a fully automated tracking system, a quadric inter-frame differential technique combining with projection method is proposed to complete the task of motion detection and initial contour extraction. Results show that the algorithm can accurately detect moving objects.In a complex environment, the convergence of the active contour is easily disturbed by the edge of the background. For this reason, a contour constraint method is proposed. This method uses the motion information and the contour feature of the moving object to modify the target profile. Results show that this contour constraint method makes it possible to track a moving object in a very complex environment.