A Fast Robot Teaching System Based On Binocular Vision

With the development of science and technology and the progress of manufacturing industry,industrial robots are widely used in all walks of life.The processing scale of small and medium-sized manufacturing enterprises is small,but there are many kinds of processing.When using robots to work,it is necessary to continuously conduct robot teaching according to the needs of processing products.At present,traditional robots are difficult and slow to teach,and there are few professionals in robot operation.Thus,the robots are difficultly popularized in small and medium-sized manufacturing enterprises.In order to resolve these problems,a robot teaching system with simple operation,low technical threshold and high teaching efficiency is proposed.In this study,binocular vision was combined with industrial robots to build a robot fast teaching system based on binocular vision.A hand-held teaching device is designed,which can quickly specify the position of teaching points and determine the posture of the end tool of the robot.Main research contents include:(1)Overall scheme design of visual teaching system.To compare and analyze the binocular vision structure scheme,a parallel binocular vision is selected.And then the visual system integration box and a hand-held teaching device are designed and processed by considering the selection of camera and lens.The integration module forms the overall scheme of the visual teaching system and deduces the transformation relation of the teaching point from the handheld teaching device to the base coordinate system of the robot.(2)Calibration and realization of three model parameters of binocular visual teaching system.Starting from the four coordinate systems of the vision system and the transformation relationship among different coordinate systems,the camera internal and external parameter models are derived.The camera parameters are obtained by Zhang Zhengyou’s calibration method.The "two-step method" was used to calibrate the hand-eye relation of the robot,which is the transformation relation between the tool coordinate system of the robot end and the coordinate system of binocular vision system.A method for calibrating parameters of handheld teaching device is proposed based on least square method.(3)Image feature information extraction and detection.On the one hand,the gaussian filter is selected as the image preprocessing algorithm through the filtering experiment,and the corner feature extraction of the pre-processed image is carried out to obtain the corner information and reconstruct the threedimensional space point information.On the other hand,the deep learning-based YOLOv3 detection algorithm is used to realize the target detection of image feature information.This research will be a foundation for the robot to track the movement of the hand-held teaching device.(4)Development and experimental testing of binocular visual teaching system control software.The overall software architecture and functional modules are planned.The software of C++ and Open CV are used to realize this control software development of visual teaching system.The visual teaching system of industrial robot with six degrees of freedom was built.The feasibility of the proposed binocular visual teaching system was verified by conducting welding teaching and tracking teaching experiments with the handheld teaching device.The experimental results show that the teaching efficiency and ease of use of the robot have been greatly improved through the design of the fast robot teaching system based on binocular vision.