隧道结构表面病害特征快速检测研究

发布时间：2018-03-07 16:05

本文选题：畸变图像修正　切入点：图像处理　出处：《西南交通大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着目前我国隧线比高的高速铁路的大量建设,未来将会面临着大量隧道需要进行检测和维护的状况。然而传统检测方法效率低,不能满足高速铁路的检测要求。因此,急需研究一种快速检测隧道结构表面病害的方法和技术方案。采用基于图像处理的检测方法,重点对隧道结构表面裂缝病害的图像提取方法展开研究。主要开展了以下五个方面的研究：图像修正、图像去噪、图像分割、裂缝快速识别和病害检测。针对采集图像存在畸变的情况,根据畸变产生原理进行了图像修正,使得图像信息更为准确。在图像采集时,研究了引起像素当量变化的一些情况,并提出了相应的修正方法和建议。在图像处理方面,采用了自适应中值一维纳滤波法对图像进行去噪处理,该方法在去噪的同时,更多地保留了图像边缘信息。隧道图像背景灰度值杂乱,采用基于边缘信息改进的Otsu法进行分割；同时对分割后的图像进行了孤立点去除和形态学处理,减少图像中的杂质区域。在裂缝快速识别方面,由于检测系统所采集的图像数量大,且分割图像仍然含有杂质区域,不利于裂缝直接提取。基于裂缝与非裂缝间特征的区别,选取了长度、面积、占有率、平均宽度、背景连通域数目以及圆形度六个特征,采用支持向量机法建立了识别模型,从而实现了图像中裂缝区域的快速判别。在隧道病害检测中,主要检测隧道裂缝发展状况和掉块区域。通过裂缝骨架图像拼接、里程定位实现了不同时间拍摄的裂缝的对比,从而研究裂缝的发展状态。同时通过对复杂裂缝区域裂缝骨架的分析,提出了通过回路检测来确定是否含有掉块的方法。通过研究基于图像处理的隧道表面病害检测方法,试图建立一个完整的采集图像处理和病害检测的流程,为隧道结构表面病害检测系统的研发奠定基础。
[Abstract]:With the large construction of high-speed railway with high tunneling ratio in our country at present, there will be a large number of tunnels to be inspected and maintained in the future. However, the traditional detection methods are inefficient and can not meet the inspection requirements of high-speed railway. It is urgent to study a method and technical scheme for rapid detection of surface diseases of tunnel structure, which is based on image processing. This paper focuses on the image extraction method of tunnel structure surface crack disease, and mainly studies the following five aspects: image correction, image de-noising, image segmentation, rapid crack identification and disease detection. According to the distortion of the captured image, the image is corrected according to the principle of distortion generation, which makes the image information more accurate. In the process of image acquisition, some cases that cause the change of the pixel equivalent are studied. The corresponding correction methods and suggestions are put forward. In the aspect of image processing, adaptive median Wiener filter is used to de-noise the image. The method not only removes noise, but also preserves the edge information of the image, and the gray value of the tunnel image is chaotic. The improved Otsu method based on edge information is used to segment the segmented image, and the isolated points are removed and the morphology is processed to reduce the impurity area in the image. In the area of fast crack identification, because of the large number of images collected by the detection system and the fact that the segmented image still contains impurity areas, it is not conducive to the direct extraction of cracks. Based on the difference between the characteristics of cracks and non-cracks, the length and area are selected. Based on the six characteristics of occupation, average width, number of background connected domains and roundness, the recognition model is established by using support vector machine (SVM) method, which realizes the fast identification of crack regions in the image. In the detection of tunnel diseases, it mainly detects the development of tunnel cracks and the area of falling blocks. Through the splicing of crack skeleton images, mileage location realizes the comparison of cracks taken at different times. At the same time, by analyzing the fracture skeleton in the complex fracture region, a method to determine whether or not there is a falling block by loop detection is put forward. By studying the method of tunnel surface disease detection based on image processing, this paper attempts to establish a complete process of image processing and disease detection, which lays a foundation for the research and development of tunnel structure surface disease detection system.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U456.3

【参考文献】