Research On The Application Of Machine Vision In Automatic Welding

Welding plays an important role in industrial production and is called "industrial tailoring." Welding robots are widely used in automobile manufacturing,petrochemical machinery production,construction engineering,shipbuilding,power engineering and many other industries.However,at present,most of the welding robots used are teaching and reproducing robots.Due to the errors caused during the welding process,the welding quality is not high,the welding efficiency of low-volume production is low,and other problems exist.Therefore,it is of great value to apply machine vision to welding robots.However,there are many problems that need to be solved: First,there are many kinds of interference factors in the welding process.It is the first problem which is to remove the interference from the welding image and extract the coordinate information of the weld feature points to be solved.Second,real-time tracking of the weld seam is achieved after obtaining the coordinates of the weld feature points.In this paper,machine vision robot automatic welding is studied,the work results are as follows:1.An improved Zhang’s calibration algorithm is proposed for the calibration of robot vision system.Firstly,the imaging model and Zhang’s calibration principle are introduced.FAST algorithm is used instead of manual method to extract the calibration plate corner points,and then the calibration plate is imported into the MATLAB calibration toolbox to realize the subsequent calibration process.The results show that this method reduces manual operation.The error generated by the extraction of the corner points also avoids possible errors in the corner extraction process and improves the accuracy and speed of the calibration.2.Aiming at the problem of extracting the coordinates of weld feature points,a variety of image processing algorithms are proposed.First,in order to reduce the amount of calculation and interference factors of the weld image,the weld image is subjected to graying and adaptive filtering.Secondly,in order to remove the extra image information,the laser stripes that need to be processed are retained,and the weld image is binarized using the Ostu method.Then,in order to extract the weld seam feature information from the laser stripe,the edge of the laser stripe is extracted using an edge detection method,and the laser stripe is thinned into a line by the symmetry axis conversion method.Finally,the refined laser fringes are processed by Hough transform to obtain the coordinate information of the final weld feature points.The image processing process is programmed using MATLAB.The results show that the method has high reliability.3.For the problem of the weld tracking,this paper proposes a pre-detection linear interpolation method to achieve real-time tracking of welds.By detecting the feature points of the weld in advance,and then identifying the inflection points occurring in the welding process,the coordinate of the inflection point is replaced by the coordinates of the feature points of the weld detected in advance,thereby reducing the problem of larger errors in the welding trajectory at the corners.At the same time,for the assembly error caused by human errors,the use of automatic correction method to eliminate,so that the system can still work accurately in the case of external interference.4.The weld seam tracking accuracy was verified by experiments.Using ordinary linear interpolation method and this method to test the weld seam,the results show that under different welding speeds,the tracking algorithm of this paper can maintain a good tracking effect with less error.At the same time,artificially misplace the weld position and use the automatic correction method to accurately obtain the position of the new placed weld.The method is effective and feasible.