Method for distinguishing welding bead based on colour information

Abstract

The invention relates to a method for distinguishing a welding bead based on colour information. The method comprises the following steps of in sequence: gaining a training image; acquiring training data and training through a support vector machine so as to obtain a classification hyperplane; reading a series of continuously-shot welding bead images to be distinguished; setting more than one ROI (Returner of Open Interface) on the welding bead images based on the apriori information; respectively extracting and storing a colour information vector of each pixel within each ROI on the welding bead image; identifying the colour information vector of each pixel into the testing data; classifying the testing data within each ROI range according to the classification hyperplane in previous step; defining that the testing data belongs to the welding bead pixel or the base material pixel; marking the classifying results; sequentially classifying the welding bead image of each frame; and determining the side edge of the welding bead according to the classifying result within each ROI of the welding bead image to be distinguished. The method for distinguishing the welding bead based on the colour information can be widely applied to the automatic distinguishing process of welding robots or other automatic equipment and technologies to the welding bead, such as groove detecting, wed joint tracking, post-welding nondestructive detecting tracking, weld joint formation quality detection and material surface detection.

Description

technical field [0001] The present invention relates to a welding bead identification method, in particular to a welding automation and intelligent detection method based on color information, which is suitable for robotic automatic welding, welding seam tracking, welding bead identification, automatic guidance of non-destructive testing after welding, etc. way of identification. Background technique [0002] Automatic welding bead recognition is a key technology in the development process of welding automation and intelligence, which can be widely used in automatic teaching of welding trajectory, real-time tracking of welding seam and automatic guidance of non-destructive testing equipment after welding. Visual sensing technology (visual sensor) has become the most widely used welding bead automatic identification method because of its advantages of large amount of information, no contact, and little electromagnetic interference. Process the image of the welding bead to ob...

AI Technical Summary

Problems solved by technology

[0003] The more mature seam tracking products currently used are all based on the structured light method to realize weld bead recognition, mainly using the obvious three-dimensional geometric structure of the bevel or weld bead, and the recognition object of image processing is the distortion of the structured light strip, but in welding When the bead structure is not obvious, the above method is difficult to realize the identification and tracking of the weld bead. Taking the cover welding of multi-layer multi-run welding as an example, the groove has been filled by the previous layers of weld bead before the cover welding , the structured light strip will not be significantly distorted when projected onto the surface of the weld bead, and the possible unevenness or entrained impurities in the weld bead or nearby base material may affect the shape of

Images