【作者】 安贝贝；

【导师】 曾理；

【作者基本信息】 重庆大学 ， 应用数学， 2007， 硕士

【摘要】 缺陷检测技术是提高产品质量的有力保证,对于减少或避免因缺陷引起的意外事故有积极作用。工业CT作为一种实用的无损检测技术,已广泛应用于航空、石油、钢铁、机械、汽车、采矿等领域,它可以在无损伤状态下,准确检测工件的内部结构。工业CT图像缺陷检测的目的,是从CT图像中寻找工件的缺陷所在,并获得有关缺陷的尽可能精确的信息。本文以工业CT图像为研究对象,根据目前国内外工业CT图像缺陷检测技术的发展现状,在分析小波变换提取图像边缘方法不足的基础上,将脊波变换(Ridgelet)应用于工业CT图像缺陷检测中。脊波变换是1999年美国Stanford大学的E.J.Candès和Donoho教授在小波变换的基础上提出的。脊波变换能够充分考虑图像边缘的方向性和奇异性,可以有效处理高维情况时的线状或面状奇异性。它的核心是通过Radon变换(投影)将线状奇异性转化为点状奇异性,再利用小波变换处理Radon域上的点状奇异性。所以,脊波变换适宜于检测具有线状奇异性的图像缺陷(如裂纹)。本文围绕脊波变换在工业CT图像缺陷检测中的应用开展研究,主要进行以下几个方面的工作:首先,对于工业CT图像装配缺陷问题,本文采用模板匹配的方式进行检测。为提高匹配速度,本文研究了一种基于脊波变换的图像匹配改进搜索方法。在保证匹配精度的条件下,所采用的变步长搜索策略,能有效提高图像匹配的速度,且该方法具有较好的鲁棒性。其次,由于工件本身的特点及其他物理因素影响,造成某些工业CT图像缺陷边缘不够清晰、图像噪声较多,对于这类图像,传统的边缘提取方法很难精确提取其缺陷边缘。因此本文研究了一种基于脊波变换的边缘提取方法,并将其应用在实际工业CT图像的裂纹提取中。实验表明该方法能得到定位准确、连续、独立的裂纹边缘图像。最后,在利用脊波变换提取工业CT图像小间隙裂纹边缘的基础上,本文研究了一种基于脊波变换的亚像素测量方法,采用这种方法测量小间隙裂纹的几何参数并标定裂纹宽度。实验证实该方法能够使小间隙裂纹宽度测量的精度达到1/5像素以上。更多还原

【Abstract】 Defect detection technology guarantees the improving quality of products, which can be used to reduce or avoid accidents caused by defects efficiently. Industrial Computed Tomography(ICT), as a kind of practical non-destruction detection technologies, has been widely used in many fields, such as aviation, oil, steel, engineering machine, automobile, mining and so on. It can be used to detect inner structure of work pieces availably without destruction. Defect detection technology based on ICT has been developed in recent years all over the world, and its aim is to search for the position and accurate information about defects of an object from ICT image.Wavelets are used to catch the zero-dimensional or point singularities, but not good at processing higher dimension singularities such as image edge. To overcome this weakness, ridgelet transform is applied to detecting the defects of ICT images in this thesis. Ridgelet transform, developed from wavelet transform, was put forward by Professor E.J.Candès and Donoho from Stanford University in 1999. Using ridgelet to deal with line singularities in 2-D, the orientations and singularities of image edge can be considered at the same time. The idea of ridgelet transform is first to map a line singularity into a point singularity using the Radon transform, and then using wavelets to deal with point singularities in Radon domain. Therefore, it is appropriate that ridgelet transform can detect the defects of image such as cracks, which have line singularities.The study of this thesis was focused on the application of ridgelet transform in the defect detection of ICT images, including image matching, edge extraction, and measurement as follows.Firstly, template matching is used for assembly defect detection. In order to improve its speed, a displacement-modified image matching algorithm based on ridgelet transform is proposed. The adopted alterative displacement method can not only keep precise and improve the speed, but also has strong robustness against the Gaussian noise.Secondly, some ICT images have the characters such as low contrast, blurry edges and noises owing to physical factors. It is hardly to process this kind of images by traditional edge extraction methods. Therefore, an edge extraction method based on ridgelet transform is studied and applied to extracting crack edge of ICT image. Using this algorithm, a nicety, continuous and unattached crack edge can be obtained.Finally, on the basis of crack extraction, a sub-pixel measurement method based on ridgelet transform is, furthermore, updated to get geometrical parameters such as the width of small cracks. The experiments validated that this method can get one fifth sub-pixel precision for width.更多还原