焊接缺陷射线DR图像自动检测与识别系统研究

发布时间：2018-06-19 02:54

本文选题：焊接缺陷 + 缺陷检测　；参考：《南昌航空大学》2017年硕士论文

【摘要】：数字射线检测是未来射线检测的主流技术,传统人工评片方法已不适合数字图像评定。随着计算机技术的发展,焊接缺陷的自动检测与识别是目前焊接缺陷无损检测的热点问题。以某型大尺寸环焊缝DR(Digital Radiography)采集图像为研究对象,进行焊接缺陷的自动检测与识别研究,以Visual Studio 2010和OpenCV 3.0开源代码为开发工具,开发焊接缺陷射线DR图像自动检测与识别软件系统。主要研究工作如下:(1)通过DR图像预处理算法提高DR图像的对比度,提出了一种焊接缺陷射线DR图像自动检测算法。首先设置平滑半径r,构造(2r+1)×(2r+1)的平滑模板。然后采用图像中值滤波创建模拟理想焊缝图像,将模拟理想焊缝图像与原始图像进行图像减影运算。最后寻找所有减影差值超过灰度连通性(给定阈值)的区域做为可疑缺陷。分析了灰度连通性和平滑半径两个检测参数对焊接缺陷自动检测结果的影响。(2)根据焊接缺陷自动检测后生成的二值图像和原始灰度图像对所有可疑缺陷进行特征参数分析与计算,提出了9个特征参数和计算算法,并得到了相应的结果。根据缺陷特征参数提出了焊接缺陷的定性分析算法,实现了焊接缺陷的自动定性分析。(3)进行了焊接缺陷射线DR图像自动识别软件系统的设计与集成。首先进行了焊接缺陷自动识别软件系统的总体框架设计。其次,设计并实现了自动识别系统的主界面、图像几何变换模块、图像预处理模块、缺陷检测模块、缺陷统计与质量评级模块。最后主要采用Visual Studio 2010开发环境自主开发完成了一套焊接缺陷射线DR图像自动识别软件系统,实现的主要功能有DR采集图像的导入、DR图像几何变换、DR图像预处理、焊接缺陷自动检测、焊接缺陷参数计算、焊接缺陷定性分析、缺陷显示列表、缺陷实时定位、缺陷特征编辑、缺陷统计、特殊缺陷描述、超标缺陷信息显示、质量自动评级、质量评级信息存储等。(4)以分散气孔、密集气孔、夹钨、未焊透等典型焊接缺陷图像进行软件的测试分析,软件测试结果表明:当检测参数选择合理时,焊接缺陷自动检测结果效果与实际缺陷信息一致,缺陷定性分析算法基本合理,软件总体运行良好。
[Abstract]:Digital ray detection is the mainstream technology in the future. Traditional manual assessment method is not suitable for digital image evaluation. With the development of computer technology, automatic detection and identification of welding defects is a hot issue in nondestructive testing of welding defects. The automatic detection and identification of welding defects are carried out with the image acquisition of a large size ring weld DRM Digital Radiography. The open source code of Visual Studio 2010 and OpenCV 3.0 is used as the development tool. A software system for automatic detection and recognition of welding defect Dr images is developed. The main research work is as follows: (1) the contrast of Dr image is improved by Dr image preprocessing algorithm, and an automatic detection algorithm of welding defect ray Dr image is proposed. First, the smoothing radius r is set and the smooth template of 2r 1 脳 2r 1) is constructed. Then the image median filter is used to create the simulated ideal weld image, and the image subtraction operation is carried out between the simulated ideal weld image and the original image. Finally, all regions whose subtraction difference exceeds gray connectivity (given threshold) are found as suspicious defects. The influence of two detection parameters of gray connectivity and smooth radius on the automatic detection results of welding defects is analyzed. (2) based on the binary image and original gray image generated by automatic detection of welding defects, all suspicious defects are characterized. Parameter analysis and calculation, Nine characteristic parameters and calculation algorithms are proposed, and the corresponding results are obtained. According to the characteristic parameters of welding defects, a qualitative analysis algorithm for welding defects is proposed, and the automatic qualitative analysis of welding defects is realized. The software system for automatic recognition of welding defects based on X-ray Dr images is designed and integrated. Firstly, the overall frame of the software system for automatic recognition of welding defects is designed. Secondly, the main interface of automatic recognition system, image geometric transformation module, image preprocessing module, defect detection module, defect statistics and quality rating module are designed and implemented. Finally, a software system for automatic recognition of welding defect X-ray Dr image is developed by using Visual Studio 2010 development environment. The main functions of this system are the import of Dr images into Dr images and the pre-processing of Dr images. Welding defect automatic detection, welding defect parameter calculation, welding defect qualitative analysis, defect display list, defect real-time location, defect feature editing, defect statistics, special defect description, defect information display, quality automatic rating, The quality rating information is stored, etc.) the software is used to test and analyze the typical welding defect images, such as dispersed pores, dense pores, intercalated tungsten, not welded thoroughly, etc. The results of the software test show that: when the detection parameters are reasonable, The results of automatic detection of welding defects are consistent with the actual defect information, the qualitative analysis algorithm of defects is basically reasonable, and the software is running well.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】