基于数字图像处理的沥青路面裂缝识别技术研究

发布时间：2018-01-19 02:02

本文关键词： 道路工程数字图像处理路面裂缝自动识别路面管理系统　出处：《西南交通大学》2015年博士论文　论文类型：学位论文

【摘要】：作为沥青路而最重要的两种破损形式,疲劳开裂和车辙已经成为影响路面使用性能的主要因素。根据交通部《高速公路养护质量检测方法(试行)》之规定,公路养护部门需要定期对公路路面状况进行病害调查,以便指定相应的维修养护策略。而沥青路面裂缝是路面疲劳开裂与车辙的直观表现形式,因此沥青路面裂缝必然成为路面调查的一项重要指标。传统的人工裂缝调查方法越来越不能适应高速公路发展的要求,其主要缺陷是速度慢,个人主观程度大,而且花费高,危险,还影响正常交通。因此,研发有效的自动检测算法是当前急待解决的热点和难点问题。本文将针对路而破损图像裂缝检测算法中存在的一些难题,运用数字图像学原理和方法,从路面图像的增强、分割、裂缝的度量和分类、集成图像的路面管理系统等多个方而进行研究。并根据路而破损图像特点,重点研究路面裂缝分割算法。主要的研究内容和成果如下：1、在图像预处理方面,根据路面裂缝图像的特点,为了更好地突出图像裂缝线形特征的方向性,对单模板进行扩展,并构造了8个方向的模板。提出和实现了图像掩膜平滑预处理算法,并在试验中与其他方法进行对比,显示出一定得优越性。2、重点研究基于数字图像处理的路面裂缝自动识别算法。首先,对OSTU阈值分割算法进行改进,提出最大类内、类间距离阈值准则。其次,运用小波技术,分析不同的小波基,分解水平,重构策略对路面图像裂缝识别的影响。进而改进小波变换梯度和模最大值图像分割算法,提出新的小波技术图像分割算法,并在图像裂缝识别中成功的运用。最后,运用数学形态学原理,提出多尺度顶帽变换,根据结构元素的类型、结构元素的大小、膨胀操作的次数等对路面图像进行分割试验,并取得较好的效果。3、在裂缝宽度、长度、面积、密度及空洞等几何特征提取的基础上,提出了识别线性裂缝和网状裂缝的方法,识别纵向裂缝和横向裂缝的方法,以及识别块状裂缝和龟裂的方法。这些方法与复杂的人工神经网络方法相比,计算成本较低,无需进行大量的样本训练,因此可在无人为干预的情况下,对裂缝进行自动分类。4、提出了集成数字图像的路而管理信息系统。介绍了系统整体框架,各模块的划分和功能,分析公路路面管理信息系统数据库的设计和实现,讨论了公路路面使用性能的预测及评价方法,研究了公路路面维护的决策过程和维护方案。
[Abstract]:As the two most important damaged forms of asphalt road, fatigue cracking and rutting have become the main factors that affect the pavement performance. According to the regulations of Ministry of Communications < Expressway maintenance quality Test method (trial) >. The highway maintenance department needs to carry on the disease investigation regularly to the highway pavement condition, in order to assign the corresponding maintenance maintenance strategy, and the asphalt pavement crack is the road surface fatigue crack and the rut visual manifestation form. Therefore, asphalt pavement crack must become an important index of pavement investigation. The traditional artificial crack investigation method can not meet the requirements of highway development. Its main defects are slow speed and large degree of personal subjectivity. And it's expensive, dangerous, and affects normal traffic. Research and development of effective automatic detection algorithm is a hot and difficult problem to be solved. In this paper, we will use digital image theory and method to solve some difficult problems in crack detection algorithm of road damaged image. From the road image enhancement, segmentation, crack measurement and classification, integrated image pavement management system, and so on, and according to the road damage image characteristics. Focus on pavement crack segmentation algorithm. The main research content and results are as follows: 1. In image preprocessing, according to the characteristics of pavement crack image, in order to better highlight the orientation of the image crack linear features. The image mask smoothing preprocessing algorithm is proposed and implemented, and compared with other methods in the experiment, it shows certain superiority. 2. Focus on digital image processing based on automatic pavement crack recognition algorithm. Firstly, the OSTU threshold segmentation algorithm is improved, and the maximum intra-class and inter-class distance threshold criterion is proposed. Secondly, wavelet technology is used. This paper analyzes the influence of different wavelet bases, decomposition levels and reconstruction strategies on pavement image crack recognition, and then improves the wavelet transform gradient and modulus maximum image segmentation algorithm, and proposes a new wavelet image segmentation algorithm. Finally, using the principle of mathematical morphology, the multi-scale top hat transformation is proposed, according to the type of structural elements and the size of structural elements. The number of expansion operations on the road image segmentation test, and achieved good results. 3, on the basis of crack width, length, area, density and cavity and other geometric features extraction. The methods of identifying linear and network fractures, longitudinal and transverse fractures, and block fractures and cracks are presented. These methods are compared with the complex artificial neural network methods. The calculation cost is relatively low and there is no need for a large number of sample training, so the cracks can be automatically classified in the case of no one intervention. 4. This paper presents a road management information system integrating digital images, introduces the overall framework of the system, the partition and function of each module, and analyzes the design and implementation of the database of highway pavement management information system. The prediction and evaluation methods of pavement performance are discussed, and the decision process and maintenance scheme of highway pavement maintenance are studied.
【学位授予单位】：西南交通大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：U418.66;TP391.41

【参考文献】