Study Of Target Tracking Technology Based On Robot Vision

The development of communication technology evolving rapidly,apply the wireless communication technology to the mobile robot control increasingly shows its unique advantages and great potential.On the other hand,as one of the most important research directions in the field of mobile robot,the role of robot vision in various fields has become increasingly prominent,and moving target tracking algorithm is an important as well as challenging topic in the field of robot vision.Under this premise,this paper did the research on target tracking algorithm under the complex background especially under the dynamic background,and applied it to the mobile robot system combined with the Wi-Fi wireless communication technology.Firstly set up a wireless network as the transmission medium of mobile robot target tracking system,then through the USB camera to perceive the surrounding information,applied the target tracking algorithm to the upper computer system,based on Wi-Fi wireless network for the media,realized the automatic tracking of moving object by mobile robot.Modular design is adopted in this system,and carrying out the relevant work separately from the hardware experimental platform,target tracking algorithm design,the debugging and experimental analysis,the main work of present dissertation and the specific contents are as follows:First of all,selected the USB camera equipped with the CMOS sensors to capture video information,selected STC89C52 RC as the main control chip to design the smallest single-chip system circuit,complete the design of Wi-Fi wireless transmission module,power conversion module,motor drive module;then,combined each module into a complete robot hardware platform.Secondly,studied the basic theory of the target tracking algorithms,including image feature description,optical flow method,the compressed sensing theory and three kinds of target tracking model: Boosting classifier,Decision Tree classifier and the Naive Bayes classifier.Then,according to the dynamic context of target tracking,making a comparative study of the Online-MIL tracking algorithm,TLD tracking algorithm and CT tracking algorithm.Compared the tracking performance of the three tracking algorithms through many experiments,and gave the experimental results;focused on the last two kind of tracking algorithms,analyzed the quantitative comparison of the accuracy,real-time of the two algorithms by 5 groups of video sequence with different parameters,and the experimental conclusion and analysis was presented in the end.In view of the two aspects of the accuracy and the real-time performance considerations,decided to choose the CT target tracking algorithm in the application of mobile robot platform.Finally,based on Opencv open source computer vision library and the C++ 、C# programming language in the VS2010 compiler environment to develop the robot target tracking system.According to the monocular camera can’t perceive the distance between camera and tracking object,puting forward two kinds of tracking strategy based on vision,enable the system can judge the distance in monocular camera.Debuged the overall communication system combined with the hardware platform of mobile robot and made the verification and analysis through experiments.