基于机器视觉的路面裂缝检测方法研究与实现 - 图文

河海大学本科毕业设计（论文）

Abstract

The cracks are the most common road damage, could endanger the safety of roads and highways.Be repaired in a timely manner before the cracks in the road becomes worse, can greatly reduce road maintenance funds.Important information pavement cracks assess road conditions and make the necessary road maintenance. In the past two decades, due to the image-based pavement crack detection technology to provide a safe, efficient and economical way, various image processing methods have been proposed pavement crack detection.Is usually lower than the intensity of the background based on the assumption that the strength of the edge cracks around the road surface, the intensity threshold value has been widely used in the crack detection. However, these thresholds can only produce disjoint crack debris, because the intensity along the crack below the background may not be sustainable.In addition, the road shadows tend to cause uneven illumination of the road image, which may further reduce the performance of the method of setting the threshold value. Based on edge detection method is also used for crack detection.However, cracks and background may appear low contrast between the speckle cracks might exist in the background speckle noise mistaken for crack fragment.Compared with the traditional manual inspection, visual inspection, high efficiency, the degree of automation and more secure, this project intends to study vision-based pavement crack detection method.The traditional image-based pavement crack detection method is usually assumed that the cracks in the road picture with high contrast and good continuity, but real life is often not the case.This is because the 1) road images often contain pavement from the uneven brightness caused by trees, poles and other things shadow; 2) Pavement various textures bring a large number of dot-like noise; 3) pavement crack due to the vehicle load DRUMthe role of pressure, weathering degradation resulting in decreased contrast, continuity is reduced.For these reasons, the cracks in the pavement image performance of linear low signal-to-noise ratio target has brought great difficulties to the automatic identification of cracks.In order to solve these problems, we have developed a geodesic shadow removal algorithm to remove the shadow of the road, while retaining the cracks. Then remove the shadow cracks after pretreatment, the final crack extraction.

Keyword:crack ,the shadow of pavement ,Geodesy,shadow removal,Cracks extraction

5

河海大学本科毕业设计（论文）

目 录

摘要 i Abstract iii 第一章 绪论........................................................1

1.1 引言.................................................................1 1.2 研究的目的与意义.....................................................2 1.3 国内外研究现状.......................................................2 1.3.1 线状目标的增强与提取.............................................3 1.3.2 路面裂缝的增强与提取.............................................4 1.3.3 目前存在的问题...................................................4 1.4 基于机器视觉图像处理的应用...........................................5

第二章 数字图像预处理...............................................5 2.1 引言........................................................5

2.2 图像灰度化...................................................6 2.3 图像去噪.....................................................6 2.4 图像的边缘提取................................................7 2.5.1几种算子的比较..............................................8 2.5.2边缘检测结果对比分析和评价..................................9 第三章 基于亮度高程模型的路面阴影消除.................................10 3.1 引言........................................................11 3.2 相关工作....................................................12 3.3 测地阴影去除.................................................13 3.3.1 具有纹理均衡能力的亮度补偿方法.............................14 3.3.2 基于亮度等高区域划分的阴影消除............................14

3.3.3 实验结果.................................................

3.4 裂缝地图生成.................................................15 3.4.1 检测裂缝像素..............................................16 3.4.2 张量投票裂纹增强..........................................17

6

河海大学本科毕业设计（论文）

第四章 基于灰度图像及其纹理特性的裂缝特征提取.....................18 4.1 引言....................................................18

4.2 图像预处理...............................................19 4.2.1 多种裂缝图像格式的识别.....................................19 4.2.2 灰度化...................................................20 4.2.3 灰度拉伸.................................................20 4.2.4 二值化...................................................21 4.2.5 边缘增强.................................................21 4.3 边缘提取................................................22 4.4 灰度图像形态学..............................................22 4.4.1 膨胀和腐蚀..............................................23

4.4.2 开运算和闭运算..........................................24 第五章 总结........................................................27 参考文献...........................................................28 附 录..............................................................29

一、英文原文.................................................29 二

、

英

文

翻

译..................................................33

7

河海大学本科毕业设计（论文）

第一章 绪论

1.1引言

本章介绍本论文的研究目的与意义、国内外研究现状，论文的研究内容和

组织结构。其中：

1.2节介绍本文的立题依据与研究意义。

1.3节综述了国内外关于路面裂缝检测的研究现状和进展。

1.4节提出本文的研究目标、研究内容、研究方法及拟解决的关键问题。 1.5节简述了本文的组织结构安排。

1.2 研究的背景与意义

目前，我国高等级公路的建设取得空前发展，特别是在国家《公路、水路交通“十一五”发展规划》的推动下，截至2011年底，全国高速公路通车总里程已达8.}万公里，二级以上公路达到4万公里。随着公路的建成通车，路面养护与管理已成为保障道路服务水平的关键，公路路面状况的基础数据调查迫切需要先进的技术和科学的方法作为支撑。《国家中长期科学和技术发展规划纲要》将“交通运输基础设施建设与养护技术及装备”作为重点领域及其优先主题;《公路水路交通“十一五”科技发展规划》将“道路预防性养护与快速维修技术、无损检测技术、多功能养护设备研究开发”作为重大专题攻关项目。在此背景下，公路路面养护自动化的研究也得到越来越多的重视。

在公路养护管理中，路面破损状况指数(PCI)是决策养护方案的最重要的数据。而路面裂缝类病害作为是路面破损检测的重要内容，其自动化检测一直是公路路面破损检测的热点和难点。路面裂缝的传统检测方式是定期的人工检测，由于人工检测效率低、主观性大、安全性差等，它无法满足路面破损快速检测的要求。随着科技的进步，图像检测技术得到了快速的发展，由于其具有检测速度快、测量准确、自动化程度高、获取信息丰富等特点，已经广泛应用于工业在线表面质量检测等领域。基于影像的检测手段已逐步成为路面裂缝检测的重要手段，然而由于路面裂缝具有细小不连续、对比度低、分叉且不规则等特点，使得传统的线状目标提取与识别算法不能满足要求。目前的商业化系统仍需人工交互来实现从图像中提取路面裂缝，其工作内容枯燥，工作量巨大，因此研发全自动的路面

8