Research On Application Of Workpiece Positioning Technology Based On Machine Vision In Thermal Processing

With the continuous development of machine vision technology,monocular stereo vision has been widely used in the thermal processing industry.Compared with binocular vision,monocular stereo vision technology can obtain stereoscopic information of the target workpiece in a stable and flexible manner in a variety of situations,with obvious advantages.The key operations are as follows: workpiece identification,workpiece positioning,workpiece capture.This article uses the monocular stereo vision technology in machine vision to carry out in-depth research on workpiece recognition and positioning as well as robot grabbing artifacts.It mainly involves: identification of target artifacts,positioning of target artifacts,and grabbing of target artifacts.The details are as follows:First,an improved edge recognition based workpiece recognition algorithm is proposed.This algorithm performs spline wavelet enhancement processing on the template map and the original map,uses the Canny operator to extract the edge information,and uses the edge information as the matching feature.The similarity measure of image matching is the improved Hausdorff distance.In the computational search process,the adaptive genetic algorithm using population generation gap information makes the computational efficiency of the genetic algorithm greatly improved without affecting the solution.The show that the algorithm not only has better resistance to the change of illumination conditions,but also improves the matching speed in the image matching process under the uneven brightness.Secondly,an image matching method based on SIFT scale invariant features and wavelet transform is proposed.The wavelet transform can perform data compression and can detect local mutations in the signal,while the SIFT feature transform has invariance for scaling,translation,rotation,and partial occlusion.This method decomposes the template image and the original thing,and uses the operator to do key point detection on the workpiece image,and then uses Euclidean distance to perform feature matching on the detected key points.The last step is to mismatch elimination of feature points.Therefore,the combination of the advantages of the two not only greatly reduces the dependence on the shooting position,the shooting distance,the angle,and the lighting conditions of the image acquisition platform,but also greatly reduces the amount of computation of the workpiece image matching.Again,for the target workpiece positioning,this article details the target workpiece positioning principle and calibration method based on monocular vision,and extracts the target centroid as the positioning element.In the case of neglecting the distortion of the camera.This method reduces the operation difficulty of the camera calibration experiment.Through experiments,not only the target-oriented parameter matrix was obtained,but also the feasibility of the method was verified.Finally,a four-degree-of-freedom industrial robot was used as an implemention agency to construct a robot monocular vision system using image capture cards,PCs,and cameras.Then,the target workpiece recognition algorithm and positioning algorithm proposed above were used rationally,we successfully completed smart crawl experiments on the target artifacts.