Moving Object For Visual Tracking

Visual Tracking for Moving Object by computer is a correlative subject with many science domains including Computer Science, Automation, Pattern Recognition and Intelligent System, and has a wide range of application perspective. Visual Tracking, which is attention-getting topic in overseas and domestic research, can be used in security surveillance, virtual reality, advanced user interfaces, etc.The thesis research issue on the whole is that image data is collected image data by single camera and input computer, and that calculation position of moving object after analytic and worked process, and that controlling camera to track moving object. The entire system was separated into four sections: Getting information of moving object, Tracing motion of moving object, Predicting position of moving object, Controlling Camera. The outline for getting information of moving object that is building gray-space at static background, and used Background Difference for getting rough information about shape and position of moving object. That information is denoised and erased shade to upgrade more exact data. The steps for tracing motion of moving object which build HSV space and split it for color character-HUE. According to exact information, calculating color histogram, meantime get data of color probability distribution; lastly track moving object at sequence frames by CamShift arithmetic. Using Kalman filter to improve efficiency and decrease iterative times. Kalman filter which is calculated to predict the position of moving object at next frame. This calculated data near measured value which can optimize initial searching value for Camshift and which decrease iterative times. Designing multithreading for this system can collect image information meanwhile control camera. Collecting thread has high priority in whole process because it needs time enough to calculate and the time which send instruction is so shortly that controlling thread can operate camera in small time slice. In detail, using VISCA protocol to control camera, trying to adjust rotational speed to keep moving object position at center of capture image.The arithmetic, which is put forward in the thesis, has sufficiently theoretical argument. The three parts which is composed of getting moving object, tracking object and controlling camera close unite into a entire system whose efficiency and robustness is advanced, can apply specially the environment that moving object's color contrast clearly with background as well as has practical enough value.