基于局部纹理特征的沥青路面裂缝检测方法  

Asphalt pavement crack detection method based on local texture features

作　　者：陶健 田霖 张德津[3] 胡成雪 何莉 TAO Jian;TIAN Lin;ZHANG De-jin;HU Cheng-xue;HE Li(School of Electrical and Electronic Engineering,Hubei University of Technology,Wuhan 430068,China;Algorithm Department,Wuhan Wuda Zoyon Science and Technology Co.Ltd,Wuhan 430223,China;Guangdong Key Laboratory for Urban Informatics,Shenzhen University,Shenzhen 518060,China;College of Mechatronics and Control Engineering,Shenzhen University,Shenzhen 518060,China)

机构地区：[1]湖北工业大学电气与电子工程学院,湖北武汉430068 [2]武汉武大卓越科技有限责任公司算法部,湖北武汉430223 [3]深圳大学广东省城市空间信息工程重点实验室,广东深圳518060 [4]深圳大学机电与控制工程学院,广东深圳518060

出　　处：《计算机工程与设计》2022年第2期517-524,共8页Computer Engineering and Design

基　　金：国家重点研发计划基金项目(2019YFB2102703);深圳市科创委面上基金项目(20200125)。

摘　　要：针对不均匀光照和阴影等因素影响沥青路面图像中裂缝检测误识别问题,提出一种基于局部纹理特征的沥青路面裂缝检测方法。设计结构保持型Retinex算法将高频的纹理信号从低频光照信号和结构型纹理中分离,改进百分比阈值算法,获取高信噪比的裂缝区域二值图像,建立高置信裂缝段的特征匹配机制,利用圆形度、面积、不同置信裂缝的类间欧氏距离进行置信连通域邻域去噪,实现裂缝区域的提取。实验结果表明,该算法改善了不均匀光照和阴影情况下裂缝识别的准确性。Aiming at the problem that uneven illumination and shadow affect the recognition of cracks in asphalt pavement image,a method of asphalt pavement crack detection based on local texture features was proposed.The structure preserving Retinex algorithm was designed to separate the high-frequency texture signal from the low-frequency illumination signal and the structural texture,and the percentage threshold algorithm was improved to obtain the binary image of the fracture region with high signal-to-noise ratio.The feature matching mechanism of high confidence fracture segment was established,and the noise of confidence connected region neighborhood was removed using circularity,area and Euclidean distance between different confidence cracks to extract the fracture area.Experimental results show that the proposed method improves the accuracy of crack recognition under the condition of uneven illumination and shadow.

关 键 词：沥青路面 裂缝检测 邻域去噪 特征匹配 图像处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]