|
The video tracking system has many outstanding advantages, such as visual scene, strong resistance of interference and good hidden performance. With the continuously development of the computer hardware and software, the difficulty of large computation and high real-time have been gradually weakened. So it has been applied in a lot of fields such as military vision guidance, safety monitoring of civil infrastructure, video coding and decoding, and traffic control. In order to track a moving object, problems of detection, segmentation, extraction and recognition from computer video image sequence should be resolved. This requires us to accurately obtain the parameters of the target from video image sequences, such as location, size, speed, trajectory, etc.first, and then combining the appropriate object tracking algorithm to ensure the tracking system which has a better robust and higher accuracy in the process of image processing.This thesis comprehensively discusses research background, research significance and the dependent basic theory of the object tracking system, such as the object-detection theory, Mean-Shift theory, Kalman-Filtering theory and so on. This thesis focuses on algorithm implementation of object detecting, segmenting and tracking. To address the problem of object-blocking, Bhattacharyya coefficient is proposed as the basis for the object-template-upated.Core part of this thesis is to complete designing software and object-tracking algorithm, so a object tracking software system is build at fist. The system contains a CCD camera, AC6631 data acquisition card, computer, software and other components. Then we use MFC as a platform, VC++6.0 as a development tools to design a software architecture framework and work processes, a sigle document/ muti-view programs which include the video tracking window, the trajectory window and the parameter-analysising window is generated finally. Multi-threaded technique is used in this program to obtain the video image sequence from DirectShow hardware port, through processing and analysising to get the object parameters, and then provide manual mode and automatical mode in order to separate object and draw trajectory. In addition, the parameter setting, the initial state setting, mode select setting and other dialog-box-interfaces make the software and users have a good interacting.In the last of this thesis, software usage and results analysis are described. The results show that the software not only complete object tracking quickly and accurately in the tracking system, but also work out a good solution for the occlusion in the tracking process. Therefore it realizes the functions desired.
|