The Research On Position And Pose Error Detection Method Of Welding Parts Based On Binocular Vision

In recent years, the linear friction welding is playing an increasingly importang role in the field of processing, and has the advantage of the traditional welding cannot be compared. Weld strength of LFW is high as welding materials. In addition, LFW can realize the irregular parts to weld and the welding quality is stable. After the parts are welded, the position of the parts should be measured. According to the position, the position error is compared with the ideal model, which can get the extra margin and generate the precision machining tool path of CNC machine. Binocular vision measurement is a non-contact measurement, which has the advantages of high speed and high efficiency. Computer vision has been widely used in industrial field, the main products are defect detection, size measurement, target recognition, and workpiece positioning.The paper described the mutual transformation between visual space coordinates, and established a binocular vision system and pose solution of the mathematical model. In addition, the paper designed a method of welding parts pose errors measurement based on binocular, combing the technology of computer vision. Therefore, the paper designed a method of welding parts pose error measurement based on binocular vision and toke the test using welded blisk. Firstly, in this paper, the mathematical model of binocular vision system is established, and the algorithm was used to solve the position and pose with the SVD algorithm. What's more, the internal and external parameters of the camera were calibrated, and the actual parameters of the camera are calibrated. Then, it mark points in the welding pieces, which were fitted by the center. According to the shortcomings of the marker extraction accuracy, this paper designed an accurate marker of circle fitting method through the clustering for image segmentation, and by the improved least square method for elliptic to quasi ellipse with obtained parameters. And lastly using C++ program of simulation and demonstration, which can achieved satisfactory results. Secondly, on the basis of analyzing the characteristics of binocular vision, the feature points matching method was designed based on the gray correlation matching, and then the error matching was deleted based on the line constraint. Lastly, using C++ and OpenCv image processing library, the design of the prototype system had been designed on Visual Studio platform. Finally, the system and software were designed, and made a simulation experiment, which gives the stereo camera calibration results, feature points matching test results, and three-dimensional coordinate calculation and analyses the result.