基于自适应最优化小波变换算法的焊缝缺陷检测
|
摘要
焊接件在工业领域中的应用十分广泛,目前对其质量的评定是由检验人员依据有关标准凭经验对焊缝X射线图像来进行评判的,可能会造成误判或漏判。因此,为了满足焊缝缺陷检测的客观性、科学性和规范性要求,可以采用计算机图像处理技术对X射线检测结果进行分析和识别。小波分析作为一门新兴学科,具有多分辨率分析的特点和时频局域性等优点,是检测突变信号的强有力工具,已广泛应用于图像处理领域。为此,论文将小波图像处理技术应用于X射线焊缝图像检测与识别中,通过计算机来实现焊缝缺陷的检测与识别。
提出了一种新的自适应最优化小波变换算法,它克服了传统的二进小波变换只能对长度为2的整数幂的数据进行处理,对非2的整数幂的数据则需要进行大量的边界处理这一局限性。其核心是通过解析被处理数据长度来捕获其长度的最佳逼近值,实现边界处理的最优化;通过分解最佳逼近长度来获取各层次小波变换基数,实现小波变换基数选择的自适应。其边界数据不会被忽略的特点就能够排除有缺陷在焊缝边缘处被漏查的情况,从而保证了焊缝缺陷检测的精确性与可靠性。
研究了小波奇异性检测原理,分析了噪声与图像信号在小波变换域上具有不同的Lipschitz指数的特点,并使用一种具有相同特性的阈值函数有效地去除了噪声。介绍了传统的图像多分辨率边缘检测方法优点与不足,从一个新的角度提出了一种基于小波变换的多分辨率图像重构焊缝缺陷检测方法,其检测效果比传统的多分辨率边缘检测算法更为快速更为准确。模拟了人类视觉系统由粗到细的收集和处理信息的过程,在分析焊缝图像特征的基础上建立了焊缝图像的区域模型,并且成功的提取出了焊缝区域。研究了不同焊缝缺陷的特征,详细介绍了在焊缝区内进行缺陷识别时所应依照的原则及标准。
在将小波多分辨率分析理论引入到X射线焊缝缺陷检测的研究中,根据提出的新方法研制和开发了基于自适应最优化小波变换的焊缝缺陷检测软件,并对其主要功能模块进行了详细介绍。最后将此软件应用于实际的焊缝缺陷检测中,并与几种经典的边缘检测方法做了比较。结果表明论文所提出的新算法是正确的、有效的,具有一定的实用价值。
Welded structure was widely applied to industrial fields. Nowadays, the quality of welding was assessed by the experienced inspection personnel according to relevant standards at present, which might lead to misjudgment and judgment leakage. Therefore, the computer image processing technology should be used to analyze and recognize the X-ray detection result for satisfy the objectivity, scientific and normative demand of the weld quality assessment. As a new subject, wavelet analysis which was characterized by multi-resolution and time-frequency localization, was a good tool to detect abrupt signals and was widely-used in image processing area. Making good use of this, the wavelet image processing technology was applied to X-ray weld image detection and recognition in the paper. The X-ray weld imperfection detection and recognition were implemented by computer.
Aiming at the limitation of the traditional dyadic wavelet transform, which can only dispose the data whose length is 2 of integer power, but has to do plenty of boundary treatment when the data length isn’t, a new adaptive optimization wavelet transform algorithm was proposed. The core idea are to achieving the optimization of boundary treatment by obtaining the optimal approximate length via length analysis of pending data, and actualize the adaptive selection of wavelet transform base by analyzing the optimal approximate length. The character of no neglecting boundary data of the new algorithm guaranteed the accuracy and reliability of weld imperfection detection, which eliminated the situation that some marginal weld imperfection might be neglected.
The theory of wavelet singularity detection, and the different Lipschitz exponent characteristics of noise and signal under wavelet transform domain were investigated. The noise was removed effectively by means of a kind of threshold function which had the same characteristics. Analyzing pros and cons of the traditional multi-resolution edge detection method, a new weld imperfection inspection method based on wavelet multi-resolution image reconstruction was proposed from another point of view, which could detect more rapidly and accurately than the traditional multi-resolution edge detection method. The process of the coarse-to-fine information collecting and processing of the human vision system was simulated successfully. The region model of welding image was constructed based on the characteristic analysis of weld image, and the weld region was extracted effectively. The characters of different weld imperfection are investigated, and the correlative principles and standards of weld defect recognition are introduced amply.
In the research for introducing wavelet multi-resolution analysis theory into X-ray weld imperfection detection, the weld imperfection inspection system based on the adaptive optimization wavelet transform was researched and developed according to the new methods proposed in the paper. The main function modules of the system are described in detail. Finally, compared with several traditional edge detection algorithm, the system was applied to practical weld imperfection inspection, and the detection results show that the new method proposed in the paper was correct and effective, and was of a certain practical value.
提出了一种新的自适应最优化小波变换算法,它克服了传统的二进小波变换只能对长度为2的整数幂的数据进行处理,对非2的整数幂的数据则需要进行大量的边界处理这一局限性。其核心是通过解析被处理数据长度来捕获其长度的最佳逼近值,实现边界处理的最优化;通过分解最佳逼近长度来获取各层次小波变换基数,实现小波变换基数选择的自适应。其边界数据不会被忽略的特点就能够排除有缺陷在焊缝边缘处被漏查的情况,从而保证了焊缝缺陷检测的精确性与可靠性。
研究了小波奇异性检测原理,分析了噪声与图像信号在小波变换域上具有不同的Lipschitz指数的特点,并使用一种具有相同特性的阈值函数有效地去除了噪声。介绍了传统的图像多分辨率边缘检测方法优点与不足,从一个新的角度提出了一种基于小波变换的多分辨率图像重构焊缝缺陷检测方法,其检测效果比传统的多分辨率边缘检测算法更为快速更为准确。模拟了人类视觉系统由粗到细的收集和处理信息的过程,在分析焊缝图像特征的基础上建立了焊缝图像的区域模型,并且成功的提取出了焊缝区域。研究了不同焊缝缺陷的特征,详细介绍了在焊缝区内进行缺陷识别时所应依照的原则及标准。
在将小波多分辨率分析理论引入到X射线焊缝缺陷检测的研究中,根据提出的新方法研制和开发了基于自适应最优化小波变换的焊缝缺陷检测软件,并对其主要功能模块进行了详细介绍。最后将此软件应用于实际的焊缝缺陷检测中,并与几种经典的边缘检测方法做了比较。结果表明论文所提出的新算法是正确的、有效的,具有一定的实用价值。
Welded structure was widely applied to industrial fields. Nowadays, the quality of welding was assessed by the experienced inspection personnel according to relevant standards at present, which might lead to misjudgment and judgment leakage. Therefore, the computer image processing technology should be used to analyze and recognize the X-ray detection result for satisfy the objectivity, scientific and normative demand of the weld quality assessment. As a new subject, wavelet analysis which was characterized by multi-resolution and time-frequency localization, was a good tool to detect abrupt signals and was widely-used in image processing area. Making good use of this, the wavelet image processing technology was applied to X-ray weld image detection and recognition in the paper. The X-ray weld imperfection detection and recognition were implemented by computer.
Aiming at the limitation of the traditional dyadic wavelet transform, which can only dispose the data whose length is 2 of integer power, but has to do plenty of boundary treatment when the data length isn’t, a new adaptive optimization wavelet transform algorithm was proposed. The core idea are to achieving the optimization of boundary treatment by obtaining the optimal approximate length via length analysis of pending data, and actualize the adaptive selection of wavelet transform base by analyzing the optimal approximate length. The character of no neglecting boundary data of the new algorithm guaranteed the accuracy and reliability of weld imperfection detection, which eliminated the situation that some marginal weld imperfection might be neglected.
The theory of wavelet singularity detection, and the different Lipschitz exponent characteristics of noise and signal under wavelet transform domain were investigated. The noise was removed effectively by means of a kind of threshold function which had the same characteristics. Analyzing pros and cons of the traditional multi-resolution edge detection method, a new weld imperfection inspection method based on wavelet multi-resolution image reconstruction was proposed from another point of view, which could detect more rapidly and accurately than the traditional multi-resolution edge detection method. The process of the coarse-to-fine information collecting and processing of the human vision system was simulated successfully. The region model of welding image was constructed based on the characteristic analysis of weld image, and the weld region was extracted effectively. The characters of different weld imperfection are investigated, and the correlative principles and standards of weld defect recognition are introduced amply.
In the research for introducing wavelet multi-resolution analysis theory into X-ray weld imperfection detection, the weld imperfection inspection system based on the adaptive optimization wavelet transform was researched and developed according to the new methods proposed in the paper. The main function modules of the system are described in detail. Finally, compared with several traditional edge detection algorithm, the system was applied to practical weld imperfection inspection, and the detection results show that the new method proposed in the paper was correct and effective, and was of a certain practical value.
引文
[1]周正干,滕升华,江巍等.焊缝X射线检测及其结果的评判方法综述.焊接学报,2002, 23(3): 85-88.
[2]薛迪甘.焊接概论(第三版).北京:机械工业出版社, 1995.
[3]中国机械工程学会无损检测分会.射线检测.北京:机械工业出版社, 1997.
[4]郑世才.评片技术(II).无损检测. 2000, 22(5):229~236.
[5]林克正.基于小波分析的焊缝图像处理与识别的研究.博士学位论文.哈尔滨.哈尔滨工程大学. 2001.
[6]阮秋奇.数字图像处理学.北京:电子工业出版社, 2001.
[7]李建平,唐远炎.小波分析方法的应用.重庆:重庆大学出版社, 1999.
[8]郑世才.射线检测技术20年回顾.无损检测. 1998, 20(3):61~64.
[9]苏大光.微机图像处理系统.北京:清华大学出版社, 2000.
[10]陈定华,吴林等.微型计算机自动识别焊缝缺陷的研究.无损检测. 1983, 5 (5):21-25.
[11]孙秋冬,黄思军.焊缝图象的信息压缩.无损检测,1990, 12 (10):273-275.
[12]李鹤歧,孙忠诚.数字图象处理在实时X射线成象系统中的应用.焊接,1990, (12):8-11.
[13]李衍.电脑评片技术进展.无损检测,1997, 19 (9) : 253-257.
[14] Yuhei kato等,木佳译.自动识别射线探测焊缝缺陷系统的研制.无损检测,1994, (2) :31-37
[15]藤田勉等.日立造船技报.昭和53, 39 (4) :1-6.
[16] K. Inoue et al. Trans. JWRI, 1982, 11(2):123-132.
[17]舆水大和等.电子通信学会论文志. 1983, J66-D(10):1253-1254.
[18] G. Robinson et al. Metal Construction, 1985,734-741.
[19] W. Danm, P. Rose, H. HeIdt, J. H. Boril jies. Automatic Recognition of Weld Defects in X- Ray Inspection. British Journal of NDT, 1987,29 (2):79-82.
[20]井上胜敬.神经网络技术应用于焊道外观检查.溶接学会论文集,1993, 11(2):288-293.
[21]焦李成.神经网络应用与实现.西安电子科技大学出版社,1993.
[22] V. K. Sinha.缺陷分析专家系统. Proceedings of the 14th World Conference on NDT, 1996.
[23]赵洪桥等. Internet与无损检测.无损检测,1997, 19 (8) :237-239.
[24]郁卫飞,李玉巧.射线图象数字化处理技术.无损检测,1998, 20 (7)193-195.
[25]顾本立,吕建等.X射线实时成象系统的图象退化和恢复.无损检测,1998.20(9):242-262.
[26] Y.迈耶,尤众译.小波与算子(第一卷).世界图书出版公司,1992.
[27]刘贵忠,邸双亮.小波分析及其应用.西安电子科技大学出版社,1992.
[28] S. G. Mallat. Multiresolution approximations and wavelet orthonormal bases of L2 (R). Trans.of AMS, 1989, 315 (1):68-87.
[29]唐晓初.小波分析及其应用.重庆:重庆大学出版社, 2005.
[30]李建平.小波分析与信号处理-理论、应用及软件实现.重庆:重庆出版社, 1997.
[31] Zhang Dong,Yang Yan,Qin Qianqing. Nonlinear adaptive wavelet transform for lossless image compression.Wuhan University Journal of Natural Sciences,2007,12(2):267-270.
[32]朱长青,王倩,陈虹等.基于多进制小波变换的图像放大方法.中国图象图形学报,2002,7(3):261-266.
[33]林福宗.小波与小波变换.北京:清华大学计算机科学与技术系智能技术与系统国家重点实验室,2001.
[34] K. P. William. Digital Image Processing. John Wileg and Soas. New York, 1978.
[35] M. H. Huekel. A Local Visual Edge Operator Which Recognize Edges and Lines. J. of ACM, 1973, 20 (2):181-194.
[36] R. M. Harallck. Digital Step Edge from Zero-Crossing of Second Directional Derivatives. IEEE Trans. on Pattern Analysis and Mach. intell, 1984, 6(1):75-81.
[37] R. Nevada, K. R. Rabu. Linear Feature Extraction and Description. Computer Vision Graphics Image Processing, 1980, 13 (4):731-746.
[38] D. Marr, E. Hildreth. Theory of Edge Detection Proc. Royal Soc. B207. London, 1980, 311-324.
[39] D. D. Richhard et al. Use of the Hough Transformation to Detect Lines and Curves in Pictures. ACM, 1972, 15(1):57-68.
[40] J. S. Lee et al. Morphologic Edge Detection. 8th ICPR,1986, 461-477.
[41] Mallat S. and Wang W. Singularity Detection and Processing with Wavelet. IEEE Trans. IT,1992, 38(2):617-643.
[42] Mallat S. and Zhong. S. Characterization of Signals from Multiscale Edges. IEEE Trans. PAMI, 1992, 14(7):710-732.
[43]杨慧中,钟豪,丁锋.基于多重小波变换的信号去噪及其在软测量中的应用.仪器仪表学报,2007.28(7):1245-1249.
[44] DONOHO D L.De-noising by Soft-thresholding. IEEE Trans on IT,1995,41(3):613-627.
[45]常鹏,阎平凡.一种基于小波变换的多尺度边缘检测方法.模式识别与人工智能,1996, 9 (3):251-257.
[46] Dai Qionghai,Liu Yebin,Xu Dong,etal.Data compression of light field using multiscale edges.Chinese Journal of Electronics,2007,16(1):101-106.
[47] Hu Dong,Zheng Baoyu.An improved wavelet-based deblocking algorithm for highly compressed images.Chinese Journal of Electronics,2004,13(3):432-438.
[48] Otsu. A Threshold Selection Method from Gray-level Histogram. IEEE Trans. on Syst. Man Cybern, 1979, 9 (1):62-66.
[49]郑世才.射线检测.机械工业出版社,1993.
[50] GB 3323-87,钢熔化焊对接接头射线照相和质量分级.
[2]薛迪甘.焊接概论(第三版).北京:机械工业出版社, 1995.
[3]中国机械工程学会无损检测分会.射线检测.北京:机械工业出版社, 1997.
[4]郑世才.评片技术(II).无损检测. 2000, 22(5):229~236.
[5]林克正.基于小波分析的焊缝图像处理与识别的研究.博士学位论文.哈尔滨.哈尔滨工程大学. 2001.
[6]阮秋奇.数字图像处理学.北京:电子工业出版社, 2001.
[7]李建平,唐远炎.小波分析方法的应用.重庆:重庆大学出版社, 1999.
[8]郑世才.射线检测技术20年回顾.无损检测. 1998, 20(3):61~64.
[9]苏大光.微机图像处理系统.北京:清华大学出版社, 2000.
[10]陈定华,吴林等.微型计算机自动识别焊缝缺陷的研究.无损检测. 1983, 5 (5):21-25.
[11]孙秋冬,黄思军.焊缝图象的信息压缩.无损检测,1990, 12 (10):273-275.
[12]李鹤歧,孙忠诚.数字图象处理在实时X射线成象系统中的应用.焊接,1990, (12):8-11.
[13]李衍.电脑评片技术进展.无损检测,1997, 19 (9) : 253-257.
[14] Yuhei kato等,木佳译.自动识别射线探测焊缝缺陷系统的研制.无损检测,1994, (2) :31-37
[15]藤田勉等.日立造船技报.昭和53, 39 (4) :1-6.
[16] K. Inoue et al. Trans. JWRI, 1982, 11(2):123-132.
[17]舆水大和等.电子通信学会论文志. 1983, J66-D(10):1253-1254.
[18] G. Robinson et al. Metal Construction, 1985,734-741.
[19] W. Danm, P. Rose, H. HeIdt, J. H. Boril jies. Automatic Recognition of Weld Defects in X- Ray Inspection. British Journal of NDT, 1987,29 (2):79-82.
[20]井上胜敬.神经网络技术应用于焊道外观检查.溶接学会论文集,1993, 11(2):288-293.
[21]焦李成.神经网络应用与实现.西安电子科技大学出版社,1993.
[22] V. K. Sinha.缺陷分析专家系统. Proceedings of the 14th World Conference on NDT, 1996.
[23]赵洪桥等. Internet与无损检测.无损检测,1997, 19 (8) :237-239.
[24]郁卫飞,李玉巧.射线图象数字化处理技术.无损检测,1998, 20 (7)193-195.
[25]顾本立,吕建等.X射线实时成象系统的图象退化和恢复.无损检测,1998.20(9):242-262.
[26] Y.迈耶,尤众译.小波与算子(第一卷).世界图书出版公司,1992.
[27]刘贵忠,邸双亮.小波分析及其应用.西安电子科技大学出版社,1992.
[28] S. G. Mallat. Multiresolution approximations and wavelet orthonormal bases of L2 (R). Trans.of AMS, 1989, 315 (1):68-87.
[29]唐晓初.小波分析及其应用.重庆:重庆大学出版社, 2005.
[30]李建平.小波分析与信号处理-理论、应用及软件实现.重庆:重庆出版社, 1997.
[31] Zhang Dong,Yang Yan,Qin Qianqing. Nonlinear adaptive wavelet transform for lossless image compression.Wuhan University Journal of Natural Sciences,2007,12(2):267-270.
[32]朱长青,王倩,陈虹等.基于多进制小波变换的图像放大方法.中国图象图形学报,2002,7(3):261-266.
[33]林福宗.小波与小波变换.北京:清华大学计算机科学与技术系智能技术与系统国家重点实验室,2001.
[34] K. P. William. Digital Image Processing. John Wileg and Soas. New York, 1978.
[35] M. H. Huekel. A Local Visual Edge Operator Which Recognize Edges and Lines. J. of ACM, 1973, 20 (2):181-194.
[36] R. M. Harallck. Digital Step Edge from Zero-Crossing of Second Directional Derivatives. IEEE Trans. on Pattern Analysis and Mach. intell, 1984, 6(1):75-81.
[37] R. Nevada, K. R. Rabu. Linear Feature Extraction and Description. Computer Vision Graphics Image Processing, 1980, 13 (4):731-746.
[38] D. Marr, E. Hildreth. Theory of Edge Detection Proc. Royal Soc. B207. London, 1980, 311-324.
[39] D. D. Richhard et al. Use of the Hough Transformation to Detect Lines and Curves in Pictures. ACM, 1972, 15(1):57-68.
[40] J. S. Lee et al. Morphologic Edge Detection. 8th ICPR,1986, 461-477.
[41] Mallat S. and Wang W. Singularity Detection and Processing with Wavelet. IEEE Trans. IT,1992, 38(2):617-643.
[42] Mallat S. and Zhong. S. Characterization of Signals from Multiscale Edges. IEEE Trans. PAMI, 1992, 14(7):710-732.
[43]杨慧中,钟豪,丁锋.基于多重小波变换的信号去噪及其在软测量中的应用.仪器仪表学报,2007.28(7):1245-1249.
[44] DONOHO D L.De-noising by Soft-thresholding. IEEE Trans on IT,1995,41(3):613-627.
[45]常鹏,阎平凡.一种基于小波变换的多尺度边缘检测方法.模式识别与人工智能,1996, 9 (3):251-257.
[46] Dai Qionghai,Liu Yebin,Xu Dong,etal.Data compression of light field using multiscale edges.Chinese Journal of Electronics,2007,16(1):101-106.
[47] Hu Dong,Zheng Baoyu.An improved wavelet-based deblocking algorithm for highly compressed images.Chinese Journal of Electronics,2004,13(3):432-438.
[48] Otsu. A Threshold Selection Method from Gray-level Histogram. IEEE Trans. on Syst. Man Cybern, 1979, 9 (1):62-66.
[49]郑世才.射线检测.机械工业出版社,1993.
[50] GB 3323-87,钢熔化焊对接接头射线照相和质量分级.