The Design Of Weld Defect Recognition System Based On OpenCV

The traditional method of weld defect detection is that the professional workers analyze and estimate weld film by the physical tools of the sightseeing room. The obvious characteristics of this method are that the operation process is complex, slow and cumbersome, and the result is vulnerable to being affected by the external environment and the worker's experience in technology, the consumption of storage space and modifying difficultly. The fast-paced demands of social production have not been met. To settle the problems what the traditional detection methods bring to, it has been a hot research topic at home and abroad of using the method of auxiliary computer technology to realize the intelligent detection of weld defect, namely using machine vision instead of human eyes. At present, they have made great achievements on weld image preprocessing, defect feature extraction and defect identification, etc. However, there are many uncertain factors in the process of welding, which inevitably lead to the diversity of the weld films, and easily make the recognition rate of the intelligent detection mechanism decline. Therefore, how to improve the recognition rate of weld defect is also a hot professional research problem.This paper still use the auxiliary computer technology to design a weld defect recognition system based on OpenCV, which regard the project requirement as background and the development status of domestic and oversea as the basis. Having this software, the evaluated worker can achieve weld defect recognition and determine the type of defect by manipulating user interface directly. The research objects of this paper are a large number of weld defect image which are already digitized and must be qualified. Otherwise, it will affect to the difficulty of the image processing and the recognition accuracy. The application development platform is VS2012 (Visual Studio 2012). In this paper, the main work includes:(1) Read some books and papers on the X-ray detection related fields, and clear the research contents of this subject. The main content of this system includes weld image preprocessing, the extraction of weld defect, the selection and calculation of the defect features, the recognition of weld defect and database operation.(2) In the process of the weld image preprocessing, analyze the characteristics of the weld image (having many noise and low contrast), and compare the advantages and disadvantages of each algorithm to choose the appropriate preprocessing algorithm. The smooth processing of image is to use a hybrid filter of median filter and mean filter. And improving image contrast is to use histogram equalization.(3) The basic idea of splitting the weld area from the image is to determine the approximate location of the weld area, and to achieve the extraction of the region of interest based on the size of the rectangle surrounding the weld area. The extraction of weld defect is to use background subtraction. Comparing the background model established with the input image is to complete the extraction of weld defect.(4) According to the characteristics of the weld defect, select "fewer but better" characteristic parameters and calculate the parameters. Due to the weld defect is more complex, this paper use BP neural network model to identify the weld defect. It mainly includes creating neural networks, training neural network and test data.(5) This paper used the Access database to manage the characteristic parameters and the type of defect. The main function are query, save and modification.(6) The last is testing the whole system. This paper mainly studies the linear weld images. Select a large number of qualified weld images as the test objects, and adopt the method of testing and improvement to achieve the desired effect.