Research On Method Of Line Structure Light Stereo Vision Weld Seam Detection

In the national pillar industries such as machining and automobile manufacturing,and in the manufacturing industries such as containers and ships,weld trajectory detection is a necessary step in the welding process and boasts as one of the most important processes in the production process of parts.Complex components have many types of welds,inconsistent weld width and large variations in base metal thickness.The process parameters of each component are also different,and due to the influence of the industrial site environment,such as arc,smoke,etc.,the weld detection is easy.There is a situation where the weld seam information cannot be obtained in real time and accurately.Therefore,improving the accuracy and efficiency of weld seam detection in intelligent welding is the key of research.The existing methods of weld seam detection at home and abroad basically extract the position information of the weld by the techniques of edge extraction and template matching in laser and stereo vision.The laser scanner has high cost,complicated structure and is susceptible to environmental noise interference;After the image filtering is denoised,the geometric feature or geometric value distribution of the gray value is used to extract the weld feature points,which lacks the robust feature representation ability and can not guarantee the automatic tracking of the weld seam.In the classification of welds,most of them are based on manual judgment,and then the process parameters are selected,and the degree of automation is low.Therefore,in view of the above-mentioned complex environment with low contrast,arc interference and low real-time performance,the author carried out research on the classification and tracking of weld seam detection,and designed a software development system for weld seam detection of line structure light stereo vision technology.According to the distortion characteristics of the line structure light projection on the surface of the weld seam,the two-dimensional information of the weld is obtained;The similarity measure is used to match the information of the left and right images,and then the RANSAC algorithm is used to match the missing points,and the mismatched points are eliminated to accurately reconstruct the three-dimensional information of the weld seam.In order to better select different processes according to weld seam type in automatic welding,the author proposed an improved neural network algorithm to classify different complex welds,and then modified the deviation of the welding process through adaptive adjustment algorithm to achieve Real-time tracking of weld seam.The simulation and experimental results show that the measurement error of the three-dimensional information of the weld is within 0.6mm,and the weld seam can be reconstructed well.The neural network algorithm proposed in this paper identifies different weld types with a recognition rate of 96%.Above,then select different process parameters for welding according to the results.However,due to the influence of arc,smoke and other welding processes,welding deviation will occur.Finally,the error tracking algorithm of the adaptive fuzzy control of the weld is proposed in this paper to correct the deviation in real time.In summary,the technology of this paper has strong applicability in actual welding operations and is of great significance for improving the automation welding technology.