基于1mm精度路面三维图像的裂缝自动并行识别算法  

Automatic parallel cracking detection algorithm based on 1 mm resolution 3D pavement images

作　　者：彭博[1] 蒋阳升[2,3] 陈成 Kelvin C.P.Wang 

机构地区：[1]重庆交通大学交通运输学院,重庆400074 [2]西南交通大学交通运输与物流学院,成都610031 [3]西南交通大学综合运输四川省重点实验室,成都610031 [4]School of Civil and Environmental Engineering ,Oklahoma State University

出　　处：《东南大学学报（自然科学版）》2015年第6期1190-1196,共7页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(51108391);中央高校基本科研业务费专项资金科技创新项目(A0920502051208-99)

摘　　要：为了快速、准确、完整地识别裂缝,基于1 mm/像素的路面三维图像提出了具有并行结构的裂缝自动识别算法.首先,进行降维处理,分别以像素(0,0)和(4,4)为起点将源图像划分为8×8像素的子块,获得2幅部分重叠的降维图像;然后,基于降维图像进行裂缝种子识别和裂缝连接,形成10个并列的子流程,从而产生10幅初步裂缝图像;最后,对10幅图像进行裂缝融合与滑动窗口去噪处理,获得裂缝图像.测试结果表明:提出的算法具有较高的准确率(平均92.56%)和召回率(平均90.59%),并以90.59%的F值优于Otsu阈值分割及Canny边缘检测算法;该算法的并行结构有利于程序并行化,能有效提高运算速度.In order to detect pavement cracking rapidly, accurately and completely, an automatic cracking recognition algorithm with a parallel structure is proposed based on 1 mm/pixel 3D pavement images. First, image dimensional reduction is conducted. A source image is divided into blocks of 8 × 8 pixels from origin pixels （0, 0） and （4, 4）, respectively, and two partly overlapped images with lower dimensions are obtained correspondingly. Then, crack seed recognition and crack connection are conducted on the two lower-dimensional images, forming 10 parallel sub-workflows, from which 10 preliminary crack images are generated. Finally, the 10 preliminary crack images are fused and then processed via sliding-window denoising techniques, yielding final crack image. Test results show that the proposed algorithm achieves relatively high precision （ averaging 92. 56% ） and recall （averaging 90. 59% ）. It outperforms Otsu threshold segmentation and Canny edge detection with an F score of 90. 59%. Furthermore, the parallel structure of the proposed algorithm helps parallel programming, which can effectively improve computing speed.

关 键 词：道路工程 识别算法 图像处理 路面裂缝 裂缝融合 裂缝种子 

分 类 号：U416.23[交通运输工程—道路与铁道工程]