Research On Robot Location Technology Based On 3D Machine Vision

Machine vision,as a part of computer technology,is developing with computer technology mature gradually.Nowadays,automation industry is from the "industrial 3.0" to the "industry 4.0" which the core of manufacturing is intelligent.It plays a crucial role in the whole developing process.Binocular vision that is a momentous branch of machine vision,the usage becomes more and more popular.It is significant for improving the success rate of industrial production and production efficiency with using binocular vision to locate the feature points by three-dimensional positioning.This article mainly focuses on the robot positioning technology which based on 3D machine vision.First,in the camera calibration,calibration is the basis of positioning also the key to transform feature points from image coordinate systems to world coordinate systems.This article discusses the advantages and disadvantages of the existing calibration methods by using Zhang Zhengyou's calibration method for camera calibration.Also make the theoretical analysis and simulation experiment.Second,according to the edge detection put forward a way that ameliorate Canny operator image processing method.The two-way filtering method is used to smooth the image,edge information can be retained effectively,the image noise can be removed out.Adding gradient templates at 45 ° and 135 ° to calculate the Gradient amplitudes then it would detect edge information as much as possible.The high and low thresholds are automatically determined by the maximum between-cluster variance method On the one hand,it can obtain the rich edge information,eliminate scratches.On the other hand it can avoid other irrelevant interference,improve the detection accuracy.Besides,in the stereo matching,based on the analysis of each stereo matching algorithm,a fast matching method improved SURF is proposed.Using the line constraint and edge constraint as the matching criterion of the algorithm.The fast matching algorithm can reduce the degree to be matched,improve the timeliness and accuracy of matching.Last but not least,A three-dimensional coordinate positioning system based on binocular vision is designed,set up the system hardware platform and its software part.And test the part of hardware and the part of the software in the system,the positioning of the feature points on the workpiece is realized.Then transmits the feature point coordinates to the robot,the robot can finish the next work.