The Research Of Servo-Control System Based On Image Recognition

The paper is based on the exploited servo controller, and a set of experiment terrace is set up, which has employed the open structure of"PC+DSP"for research of image tracking system. The object tracking is realized, then control the executing agency to make sure the object in the center of view.The paper mainly focuses on the tracking algorithm research and servo-control algorithm realization.When there is single moving object in the visual field, an improved algorithm based on histogram is developed: prove the histogram statistical method, then combined with sequential similarity detection algorithm. Mode is updated using double models.When there are several moving objects in the visual field, CamShift algorithm combined with Kalman filter was used for object tracking. The using of Kalman filter estimating the object state can decrease the search scope of pattern matching ,and increase the speed of tracing object and the ability of anti-jamming.Even if there are lots of new control algorithm today, PID control is still the most widely applied in the process movement because of its simple,better robustness and high reliability, therefore, this paper adopt the PID algorithm. In the research of servo control algorithm, expert PID control algorithm is particularly researched, which can inhibit dynamic error, to make sure the system has better dynamic response. And the experimental proof showed the system has reached the request.After research the algorithm mentioned above, by use of DSP servo controller, pc machine, camera and DC servo motor, we establish the object recognition and tracking system. Visual C++ helps finish the researching of interface and realizable of imaging tracking algorithm; and embedded C helps finish the position in-phase function; the use of SCI(Serial Communication Interface) make sure PC and DSP intercoordination each other, then the object's dynamic tracking is finished. Integer performance testing has indicated that the system has better veracity and real-time.