Research On Algorithm Of Welding Seam Location Based On Machine Vision

With the arrival of industry 4.0,industrial production has been greatly developed.Welding quality and reliability deeply influences the product quality and production efficiency,therefore,how to improve the welding quality has become a hot issue in the field of industrial production.In recent years,machine vision has been widely used in industrial production as an emerging recognition technology,which has greatly improved the accuracy and efficiency of welding industry recognition.But in the current automatic welding still exist two problems:one is the most automatic welding robot without the introduction of machine vision method,still adopts the traditional contact welding method.This method is easy to be affected by welding environment and has great wear on the probe.Two is a small number of welding robots that introduce machine vision methods,which are mostly off-line programming,resulting in large errors and low efficiency.Therefore,this paper based on the machine vision "V" type welding seam recognition and localization as the research background,the laser triangulation and model construction algorithm are taken as the theoretical basis to realize the recognition and localization of "V" type welding seam.The main work of this thesis is as follows:(1)By reviewing the relevant technical documents,determine the research object of the paper,and summarizes the development of machine vision laser weld positioning technology and future trends,the welding seam was identified by combining hog and ASM.Finally,a sheet-of-light technology was used to locate the welding seam.(2)Based on the optimization of local optimal solutions,the initial positioning of the weld image based on HOG is presented.Through the industrial camera,the positive and negative samples of the welds were collected and studied and classified by SVM.This method can obtain the effective image of the weld and obtain the initial position of the weld,providing information for the establishment of the subsequent feature model.(3)Weld identification based on ASM.The combination of the aforementioned weld initial location information,proposed the establishment of weld seam feature model and search strategy,using Qt Creator programming environment platform program,and carries out the "V" font weld identification and localization experiment.Calibration feature points on the target first,and then the feature point on the characteristic model was established,finally the characteristics of the model is set up on the new test samples of model search and iteration in order to realize the purpose of the weld position.(4)Introduce the basic knowledge of the camera and the imaging principle of machine vision calibration system.Based on HALCON platform,a sheet-of-light fixing method is proposed,which solves the problem of system calibration under laser triangle method,and provides coordinate data for welding gun to locate welding position.The weld pixel coordinates are obtained by using the two dimensional image processing,and the welding coordinate transformation is carried out by using a sheet-of-light fixed system to verify the feasibility of the algorithm.(5)Summarize the research results of this paper,put forward the possible problems in the process of research,and make a prospect of the next research work.