基于多特征分析的路面裂缝检测算法  

A Road Crack Detection Algorithm Based on Multi-feature Analysis

作　　者：卢紫微[1,2] 张燕[1] 常东超[1] 吴成东[2] LU Zi-wei;ZHANG Yan;CHANG Dong-chao;WU Cheng-dong(Department of Computer and Communication Engineering, Liaoning Shihua University, Fushun 113001, China;Department of Information Science and Engineering, Northeastern University, Shenyang 110819, China)

机构地区：[1]辽宁石油化工大学计算机与通信工程学院,辽宁抚顺113001 [2]东北大学信息科学与工程学院,沈阳110819

出　　处：《控制工程》2018年第4期591-595,共5页Control Engineering of China

基　　金：国家自然科学基金项目(61273078)

摘　　要：针对高速公路路面图像噪声成分复杂、路面裂缝损伤检测效率低、安全性差等问题，提出一种应用多特征分析的路面裂缝检测算法。首先将获取的高速公路图像进行分块处理，在每个分块图像上提取裂缝及其周围区域图像的灰度、局部熵和局部二进制模式（Local Binary Pattern，LBP）纹理特征构建特征向量，然后将特征向量输入支持向量机（Support Vector Machine，SVM）进行训练，最后利用得到的决策函数将图像中的每个像素划分为裂缝区域或背景区域。该方法综合利用了图像的灰度、局部熵和LBP纹理特征，最后通过实验验证了算法的有效性。In view of practical problems such as expressway pavement cracks damage detection with multi-noise, low efficiency and less security, a novel road cracks detection algorithm based on multi-feature analysis is proposed. Firstly, split the image into smaller blocks. Secondly, build feature vector using gray, local entropy and LBP texture features distribution information of each image block in cracks and the surrounding areas. And then, eigenvectors of sample images are inputed into SVM for training. At last we obtain the decision function, with which we classify each pixel into crack region or background region. This method comprehensively utilizes the gray, local entropy and LBP texture feature of crack images. Experimental results show that this method performances well in crack automatic detection of highway images.

关 键 词：裂缝检测 特征提取 LBP纹理 支持向量机 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]