Research And Implementation Based On The Biped Robot For Identification And Dynamic Tracking Of Objects

With the development of computer technology, mechanical technology, electronics technology and automatic control technology, robots are being processed, manufactured and applied in various fields, which provides great convenience and security for the people and the army of each country. At present, biped robot as an intelligent robot with the appearance of the person has been widely researched.The way of humanoid biped robot action is similar to people using legs. Movement control is difficult and low stability. Machine vision works as the human eye perceived information from outside. Combining the study of machine vision and humanoid biped robots to achieve the identification, location and tracking for the moving target on humanoid biped robots.Firstly, this topic designs and analyses the system solution based on embedded Linux operating, according to humanoid biped robot hardware architecture. Secondly, to locate and track the target precisely, the system uses Zhang Zhengyou calibration algorithm based on the model of an ideal pinhole to calibrate the camera parameters, and uses Brown distortion model to calibrate the camera distortion factor, and uses the camera parameters to calibrate the image. Thirdly, depending on the characteristics of the color space, the program converts RGB image which is captured by humanoid biped robot camera to HSV image, and uses region growing algorithm to identify goal in HSV image. The system gets color probability distribution by back projection method, and gets the centroid of the target image by the hue distribution probability function for target tracking process. Next, according to the characteristics of the system and CamShift tracking algorithm processes, and because of the issues of traditional CamShift algorithms need to get the initial search window manually, this topic propose the improved CamShift algorithm that humanoid biped robot automatically obtain the search window. Finally, transplanting the program of the image distortion correction, object identification and tracking algorithm from conventional PC operating system to embedded Linux operating system. According to the hardware characteristics of humanoid biped robot used in experiment, the system designs the monocular vision distance measuring model and the track strategy for the robot, so that the robot can locate and track the target object more accurate.In the experiment, humanoid biped robot calibrates the image captured by camera first. Then, it automatically obtains the initial search window to track the target’s moving trajectory automatically. And the last, humanoid biped robot uses the tracking policy to track the target object mechanically.