Target Tracking Research Of Autonomous Moving Robot Based On Visual Information

With the continuous improvement of computer vision and intelligent robots field,autonomous moving robot based on visual information has become a current research hot spot,which makes the combination of computer vision and moving robot more closely. More andmore computer vision algorithm is applied to the robot. The robot’s autonomous navigationhas been rapid development. Robot based on image understanding to make correspondingaction, this make the robot more intelligent. This way become more in line with the humanperception of things around. Computer vision technology appling to the robot, also can makethe robot to control their own behavior more accurately.The robot which we choose is the traveler VOYAGER-IIP-A type of the overallarchitecture and software system. Robot architecture was introduced in detail the four levels,namely the user layer, decision control layer, the underlying policy layer and executive layer.The robot’s main working parts include: power supply, several kinds of sensors and visionmodule. In the perspective of software architecture, we introduce the general structure of therobot to show that the robot motion object layer, command analysis layer, hardware controllayer on interaction of these three levels. This complete to build the robot software system.Then we introduce the design and implementation of autonomous moving robot obstacleavoidance behavior based on ultrasonic sensors, and in the indoor environment of the robot’sobstacle avoidance behavior experiment. First describes the ultrasonic sensors, the location onthe robot and the ultrasonic sensors’ number. We introduce the detailed design process of therobot obstacle avoidance algorithm. Then we do lots of obstacle avoidance experiments.Through these experiments and error we select reasonable threshold. Choose the appropriatethreshold can make the robot run more smoothly.In the visual processing part of robot we introduce the basic theory, including the RGBcolor space and HSV color space. Introduce choosing the HSV color space which can makethe robot target tracking for the position, rotation, scale changes reduced the sensitivity of theinterference factors. Such as enhance the robustness of the algorithm. This paper brieflydescribes the basic technology such as color histogram, color projection drawing.We use themto design the autonomous navigation behavior.Finally, introduce the robot target tracking based on computer vision behavior design. First introduce Meanshift algorithm. Thus Camshift target tracking algorithm based onMeanshift algorithm is derived. This paper introduce the basic principle of the two algorithmsin detail and the working process. Then expounds several basic theory of particle filter: abayesian filtering theory, bayesian importance sampling, sequential importance sampling.Camshift algorithm is proposed combined with particle filter after the implementation of thetarget tracking, and analyzes on the advantages of the two algorithms. Through theexperiments we find that Camshift combined with particle filter algorithm has high trackingaccuracy. We provide the reference for subsequent robot moving behavior.