Development Of Parallax 3D Attitude Positioning System Based On Multi-Vision

At present,with the development of science and economy,the combination of vision and industrial robots plays an important role in automobile assembly,medical treatment,aerial work and other fields.Considering the market demand,an intelligent assembly system for bolt detection based on in-depth learning is designed by using industrial robots and multi-vision stereo vision system.Among them,this paper mainly carries out the following research contents:According to the detection accuracy and economic requirements of the project,the suitable hardware such as camera,lens,light source and so on are selected.The model of the manipulator is determined according to the site environment and installation requirements as well as the range of movement of the project.The principle of hand-eye calibration algorithm and multi-eye vision calibration algorithm is explained and introduced.The calibration algorithm between camera and manipulator is realized according to the determined hardware model of manipulator and camera,and the results of calibration are analyzed and compared.Based on the analysis of the metal fittings used in this paper,according to the characteristics of the surface of the metal fittings,it is determined that the speckled light is used to increase the feature information of the surface of the precision fittings.Then,according to the values of the internal and external parameters after calibration of the two cameras,the parallax between the two cameras can be calculated.A parallax map containing three-dimensional information can be obtained by using NCC type stereo matching algorithm.Then the edge detection algorithm is used to fit the straight line to get the position of multiple points,and the least square method is used to determine the precise tool plane according to multiple feature points,so as to calculate the inclination angle of the precise tool plane.Then,the YOLO algorithm of deep learning is used to locate the four bolts of the precision tool,and the three-dimensional information of the four bolts is obtained by matching the locating results.The position and posture information data are sent to the manipulator through the network communication system,and the manipulator is guided to grasp the bolts.Finally,combined with software and hardware,the design of a parallax three-dimensional attitude positioning system based on multi-vision is completed.The positioning and grasping of four bolts by manipulator is successfully realized.The accuracy and stability of the three-dimensional position and attitude coordinates are verified by the analysis of the result data.The experimental results show that the project can complete the positioning and grasping of the bolts of the fittings,and has a high accuracy.And stability.