Research And Design Of Target Tracking System Of Robot

Robots are mainly interacting with the external environment by the machine vision system,so the research of the robot vision is particularly important.The tracking of the object is an important research direction of the robot vision,so it is of great theoretical and practical value to the research and design the target tracking system of robot.In practical situations,it is difficult to achieve the accuracy of target detection,the robustness and real-time of tracking target for robot target tracking system.The purpose of this paper is to research and design a robot target tracking system in the laboratory environment to achieve the relative unity of "three characteristics".First of all,this paper studied some classical moving target detection methods,namely inter-frame difference method,background difference method and optical flow method.In many tracking algorithms,this paper studied two classic Meanshift algorithms tracking algorithm and Camshift tracking algorithm.Secondly,after studying the classical detection and tracking algorithm,this paper proposed a five-frame difference weighting method for the actual situation,and improved the classical CamShift tracking algorithm.By combining the five-frame difference weighting method with the improved CamShift,this paper proposed a new improved Camshift tracking algorithm,which is based on color histogram equalization and five-frame difference weighting method(CamShift-EW target tracking algorithm).The improved Camshift algorithm has the following characteristics: it keeps the advantages of simple,efficient and real-time of classical algorithm;Compared with the traditional frame difference method,the five-frame difference weighting method can extract more complete contour,and more sensitive to moving objects with slower moving speed,and improves the ability of detection;the color histogram equalization enhances the color information of the image and improves the ability to track the target object in the lower resolution;the area threshold and the aspect ratio threshold enhance the ability of resisting the background interference of the algorithm;saving the H weight histogram can realize effectively the continuity of tracking.The algorithm can also roughly estimate the motion state and the relative position of the target object while tracking the object,which provides the decision basis for the controlling of the four-wheel robot.Thirdly,in order to verify the reliability of the improved algorithm,the hardwareplatform of the robot target tracking system is built in a relatively stable laboratory environment.The hardware part consists of the four-wheel robot,image processing unit-Rasberry Pi,the four-wheel robot control unit-Arduino,WIFI module,USB free-drive camera and MEIZU M1 mobile phone with Android5.1(the android mobile terminal is used to control the target tracking system manually).The software part includes the core target detection and tracking CamShift-EW algorithm which is used in Raspberry Pi,the control codes in Arduino to control the four-wheel robot's motion state and the android codes in android mobile terminal to receive and display video and send manual/automatic switching and manual control instructions.Finally,the algorithm is verified on the hardware platform.According to the actual running result of the tracking system,the target tracking system of robot is better to achieve the unity of "three characters".