沥青路面裂缝检测系统研究

发布时间：2018-10-30 13:01

【摘要】：随着我国高速公路网的不断扩大,公路养护工作量也日益繁重。路面病害检测作为养护工作的一项主要任务来讲,也尤为重要。就目前来讲,我国仍然停留在以人工检测为主的阶段,这种传统的检测方式效率低下、劳动强度大、危险系数高而且易造成交通拥堵现象,已经明显满足不了高速公路路面养护与管理的需求。为了适应高速公路路面养护自动化管理的发展趋势,本文主要针对沥青路面裂缝检测系统做了相关的研究。本文首先介绍了沥青路面裂缝检测的相关知识,并对裂缝类破损的类型、形成原因和破损分级及评价标准做了详细的阐述,在此基础上,概括总结了路面裂缝自动检测的原理。其次,主要探讨了路面裂缝图像的处理过程及其裂缝参数的测量算法。由于沥青路面裂缝和背景的灰度值比较相近,需要进行图像增强处理,研究表明：基于多尺度分析的图像增强算法比较适合处理路面裂缝图像。因此,本文采用非下采样Contourlet变换对路面裂缝图像进行了增强处理,使得裂缝特征凸显出来,接着,对比常见边缘检测算法得知：传统的微分边缘检测算子对噪声都比较敏感,致使在检测边缘信息时,容易出现假边缘或者漏检的现象。因而,本文提出了基于数学形态学的边缘检测,并且成功地将多尺度形态学边缘检测算子应用到裂缝图像的边缘检测中去。实验表明,该算子在有效检测裂缝边缘的同时,对椒盐噪声有很好地抑制能力,但是,遗憾的是,对高斯噪声仍然比较敏感。在此基础上又提出了抗噪型多尺度形态学边缘检测算子,此算子对椒盐噪声和高斯噪声都有很好地抑制效果,而且检测到的边缘比较平滑,实用性好。最后使用最大类间方差法对路面裂缝图像进行了分割,使得裂缝和路面背景很好地分离开来,为后续的裂缝提取做好准备工作。对分割出来的裂缝进行提取并标识,将标记出来的裂缝应用神经网络分类器划分为规则裂缝和不规则裂缝两大类,然后分别采用不同的方法对裂缝的参数进行计算。对规则裂缝的长径、短径采用常规的几何法计算,对不规则裂缝采用剔除法计算其面积。实验结果表明：本文提出的裂缝参数计算方法能够比较精确地计算出裂缝的长宽以及面积等参数,而且计算精度也基本可以满足沥青路面裂缝检测的要求。最后,本文基于图像处理技术设计了一种沥青路面裂缝检测系统,并在Visual C++6.0环境下,借助Mil-Lite8.0软件开发包对系统的主要功能模块进行了实现,实验结果表明,本文设计的系统检测精度高,能够满足路面管理要求,具有很好的实用价值。
[Abstract]:With the expansion of highway network in China, the workload of highway maintenance is becoming more and more heavy. Pavement disease detection, as a major task of maintenance, is also particularly important. At present, our country is still in the stage of manual inspection, which is inefficient, labor intensive, high risk coefficient and easy to cause traffic congestion. Has obviously not been able to meet the highway pavement maintenance and management needs. In order to adapt to the development trend of highway pavement maintenance automation management, this paper mainly focuses on the research of asphalt pavement crack detection system. This paper first introduces the relevant knowledge of asphalt pavement crack detection, and describes in detail the types, causes, classification and evaluation criteria of cracks. On this basis, it summarizes the principle of pavement crack automatic detection. Secondly, the processing process of pavement crack image and the measurement algorithm of crack parameters are discussed. Because the gray value of asphalt pavement crack and background is similar, image enhancement is needed. The research shows that the image enhancement algorithm based on multi-scale analysis is more suitable for processing pavement crack image. Therefore, the non-downsampling Contourlet transform is used to enhance the pavement crack image, which makes the crack feature prominent. Then, compared with the common edge detection algorithm, we know that the traditional differential edge detection operator is sensitive to noise. When detecting edge information, false edges or false edges are easy to appear. Therefore, in this paper, the edge detection based on mathematical morphology is proposed, and the multi-scale morphological edge detection operator is successfully applied to the edge detection of crack images. Experimental results show that the proposed operator can effectively detect crack edges and suppress salt and pepper noise, but unfortunately, it is still sensitive to Gao Si noise. On this basis, a new anti-noise multi-scale morphological edge detection operator is proposed. The operator can suppress both salt and pepper noise and Gao Si noise, and the detected edges are smooth and practical. Finally, the maximum inter-class variance method is used to segment the pavement crack image, so that the crack and the pavement background can be separated well, so as to prepare for the subsequent crack extraction. The separated cracks are extracted and marked, and the marked cracks are classified into regular cracks and irregular cracks by neural network classifier, and then the parameters of cracks are calculated by different methods. The length and short diameter of regular crack are calculated by conventional geometric method, and the area of irregular crack is calculated by eliminating method. The experimental results show that the proposed method can accurately calculate the crack parameters such as length width and area and the accuracy of calculation can basically meet the requirements of asphalt pavement crack detection. Finally, based on the image processing technology, a asphalt pavement crack detection system is designed, and the main function modules of the system are implemented with the help of Mil-Lite8.0 software development kit under the environment of Visual C 6.0. The experimental results show that, The system designed in this paper has high detection precision, can meet the requirements of pavement management, and has good practical value.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U416.217;TP391.41

【参考文献】