Real-time stereovision servo system based on parallel processing technique

A stereovision servo system was proposed based on the parallel processing technique for real-time tracking and grasping the moving object. In the system, a fast template matching method was used in combination with a fast reliable stereovision matching method to achieve the image matching and identification. A Kalman filter was applied to predict the position and speed of the object. The parallel processing technique was adopted to enhance the image matching speed. The image identification results were transmitted directly to the robot and the stereovision platform controller to achieve the visual sereo of the robot and the stereovision platform to realize the tracking and grasping functions. The computer simulation and experiments on the system were performed and the results showed that both the tracking precision and matching speed were improved.