Measurement And Rating System For Weld Defects Based On Machine Vision

With the rapid development of computer communication and machine vision technology, the major traditional industries are facing the transformation of intelligent economy. As an important branch of the manufacturing industry, the application of welding is also gradually expanding, which has penetrated into the equipment processing, bridge, construction, defense and other industries.The quality inspection of the welding process has become more and more concerned. As we all know, computer aided X-ray nondestructive detection technology is born with the development of science and technology, combined with machine vision technology to achieve weld defect types of intelligent identification, classification and measurement ratings. By means of digital technology to obtain X-ray weld defect image, and then we complete the image processing through the way of computer aided, finally get the attention defect parameters. Compared to the traditional manual detection methods, computer aided detection has several advantages, such as accuracy, standardization, automation, scientific, efficient, resource saving and so on.This project is devoted to develop a set of intelligent defect measurement and rating system, using MATLAB as the development platform to realize the intelligent measurement and rating of weld defects. The problems of the system is mainly to solve and the innovation points as follows.In order to optimize the computer aided assessment system, we introduce the image quality analysis and judgement before the handle of defects detection of digital X-ray image. According to the characteristics of X-ray weld image and the extent of exposure to determine whether the collected images meets the requirement of detecting defect recognition system and ruled out the excessive and low exposure images in advance. And then, we will reduce redundant processing of the system and achieve the optimization of algorithm structure.Take the excessive and low exposure as two conditions to conduct the image quality judgment, meanwhile, assign different weights to the two judge conditions. Using neural network to train a large number of samples and to obtain the reasonable distribution of weight. Eventually, we can improve the accuracy of X ray image quality judgment and improve the detection accuracy of the entire system.In view of the low contrast and noise characteristics of X ray image, we conduct the simulation of images with salt and pepper noise and Gauss noise, meanwhile, contrast the simulation effect and sum up the advantages and disadvantages of each algorithm. Eventually, we adopt mixed mode of median filtering and wavelet denoising, so that we can protect the image details during the process of image denoising.This paper use the spindle method, one of the improved minimum circumscribed rectangle method, to measure the length of the weld defect. By comparing with the traditional rotation method and vertex chain code method, we found that the spindle method is more efficient.Most computer aided assessment system, stop in defect recognition after the measurement, the subject achieve intelligent rating of weld defects by studying on the pressure equipment for fusion welded butt welding joint quality rating criteria and encoding of GB.GUI MATLAB design, based on machine vision of the weld defect measurement and rating system.This paper has realized the optimization of computer aided assessment system, according to the national standard, carried out the measurements of X ray welding defect image, perfected the computer aided assessment system. Combined with the project resources, to obtain valuable prior data, to measure a large number of defect samples, and then, the feasibility of the image quality assessment module, the defect measurement and rating module were verified. The system test shows that the accuracy rate of the image quality judgment module is 93.7%. The defect measurement and rating module bar defect rating, circular defect rating and comprehensive rating were measured and compared with a prior samples, the correct rate can be 91%,92.4% and 93.2%. Eventually, this topic appeared in the form of Matlab GUI, realized the intelligent defect measurement and rating system, implemented the important reform of artificial defects detection method, made certain contribution to the intelligent defect detection of defect and the real-time meet the needs of the project.