Research On Recognition Of Demonstration Action And Positioning Method Of Robot Demonstration Programming

With the large number of applications of industrial robots in industrial production,their work direction has evolved from basic mechanical processing and simple assistance to technical personnel,to the completion of complex tasks in different environments based on human-computer interaction.Robot programming technology has become a research hotspot.The thesis takes robot demo programming as the research object.To solve the problems of long programming deployment time,poor human-computer interaction ability and high usage threshold in traditional demo methods,taking the trajectory teaching task as an example,it studies the improvement and application of the trajectory generation method based on gesture recognition,combined with industry Robots and visual sensors realize the natural interaction and rapid teaching of robots and operators.In terms of gesture recognition,first of all,we conducted domestic and foreign research and research on demo programming technology and human-computer interaction technology based on gesture recognition,combined with the actual application scenarios of robots to choose equipment,and established the overall organizational framework of this article;secondly,based on skin color segmentation and The CNN combined gesture recognition method extracts gesture-related information by using the skin color model,inputs the gesture picture into the RGB and YCb C mixed skin color space to remove the non-skin color area,and then performs corrosion processing,expansion processing and Gaussian filtering to smooth the noise.Highlight the contours and edges,and finally binarize to obtain a gesture image;use Keras' self-built CNN model for feature extraction and obtain a gesture recognition model;perform verification experiments on the Marcel dataset,and for open source gesture samples,fewer cases,self Collecting images to build a gesture sample library for training and testing,and verifying the reliability of the method through experiments,which can be used for human-computer interaction.In terms of target tracking,we studied target tracking based on 3D information,applied a point cloud algorithm to robot trajectory generation,used color recognition to segment the color markers on the teach pendant,and used Euclidean algorithm to complete the target point cloud clustering.The RANSAC algorithm is used for point cloud fitting to achieve the tracking of the target color mark.The experiment verifies that the method is stable and reliable for the color mark tracking in the actual working environment and can be used for trajectory generation.Taking the UR5 robot as the research object,it performs position description and posture description to determine the position and posture information of the color mark in space,and converts the pose of the color mark in the camera coordinate system into the robot coordinate system through the homogeneous transformation matrix,Based on the target tracking method to guide the end of the robot to complete the trajectory generation;Finally,the robot teaching programming framework is established based on ROS(Robot Operation System).It is designed for gesture recognition,reading the current joint angle of the UR robot,processing the received gesture messages,and between task demonstration and execution.The switching phase and the four main nodes used to send the joint angle to the robot controller to move the robot.Apply the first three chapters of the algorithm to the framework to complete the trajectory generation method based on gesture recognition;use QT to complete the design of the human-computer interaction interface.The signal and slots mechanism of the platform was used to complete the development,and human-computer interaction experiments were conducted with the UR5 robot as the object.