Human-robot Interaction System Based On Visual Location

The application of robot is mainly through the operator's use of the teaching device.Manual control of robot joint movement,to move the robot to a predetermined position and record the location at the same time,then pass it to the robot controller,so that the robot can repeat the task automatically according to instructions.In order to solve the problem of complicated and low efficiency teaching process,a teaching device based on machine vision positioning is designed,include controller mode and trajectory generation mode.The following is the content of this paper:First of all,binocular vision system is built.The system consist of two cameras and a computer,two cameras acquire digital image.In computer doing camera calibration and correction,image preprocessing and feature extraction,object tracking and recognition,three dimensional reconstruction by stereo matching,to measure the distance of the object from the camera.Then,the method of identifying and tracking markers in the trajectory generation model is designed.Multiple markers are placed on the teaching device,and the trajectory of the teaching device is obtained by tracking the markers.This paper studies the algorithm of marker detection,image segmentation an d enhancement,contour extraction.Then,according to the location of the multiple markers in the space,the gesture of the teaching device is calculated.Finally,the method of tracking,identification and gesture calculation of hand in the control mode is studied.This paper uses method based on region template,studies a complementary method for tracker and detector.The online model is designed to train the classifier online.Fusion background difference and skin color detection algorithm,the accurate foreground area is obtained.The convex hull detection algorithm is used to obtain the landmarks of hand,and finally the pose of the hand can be obtained.