Welding image identification method

Abstract

The invention discloses a welding image recognition method based on prior knowledge, which includes using a charge-coupled device (CCD) to obtain a welding seam image and using a C-V segmentation method to segment the image. The image to be welded before welding is irradiated by an ordinary indoor light source Next, the initial contour is obtained by using a CCD camera. The molten pool image in the welding process is obtained by the light reduction and filtering system. The initial contour of the image is used as the boundary, and the C-V segmentation method is used for gradual evolution to obtain the real weld before welding. The edge shape and contour information of the molten pool in the seam and welding process, the present invention realizes the image coordinates and contour information of the seam before welding as well as the contour information and size feature information of the molten pool image as a whole, and these information can be used in the automatic welding system Or the path planning of welding robots, seam tracking deviation correction and penetration control, forming quality control, etc., have applicability in the field of more intelligent welding technology of visual sensing.

Description

technical field [0001] The invention relates to the field of welding technology, in particular to a welding image recognition method based on prior knowledge, which is used to recognize seam images before welding and various molten pool images during welding. Background technique [0002] At present, visual sensing and image processing technology have been widely used in dynamic intelligent control of molten pool, weld seam tracking, prediction of welding organization, structure and robot intelligent welding. If we want to use visual technology to identify weld or extract molten pool features One of the important steps in seam tracking, penetration control, etc. is to identify welding-related images and extract the characteristic parameters of the weld or molten pool, and among these characteristic parameters, edges and contours are important characteristic quantities. For example, on a workpiece image, the gray scale of the weld seam and the workpiece forming the weld seam ...

AI Technical Summary

Problems solved by technology

[0003] After searching the existing technical documents, it was found that Xiong Zhenyu and others published "Research on Image Processing and Recognition Method of Weld Seam of Arc Welding Robot" in "Welding Technology" (2006, Vol35, No. 3: 8-11), the According to the characteristics of the weld image obtained by the charge-coupled device (CCD), the corresponding image processing method is studied, which effectively eliminates the interference of the spatter and arc light in the welding process on the weld image. The deformation at the center of the weld can be accurately identified in the image, but its shortcomings are: only the deformation of the laser band can be identified, and it cannot be applied to multi-interference image recognition without an external laser light source, and the shape and outline of the molten pool cannot be identified. For more functions such as welding quality control

However, it only recognizes local welds, and fails to recognize images containing other disturbances as a whole, which limits the application of this method

The two methods mentioned above are mainly focused on traditional image processing methods, and do not make full use of the prior knowledge of the image itself, resulting in the inability to recognize complex images and obtain good recognition results.

Images