几种典型的路面图像背景校正算法的比较分析-同济大学学报.doc

适于路面破损图像处理的边缘检测方法 (李 莉孙立军陈 长 （） 摘要：关键词：路面损坏；图像处理；中图分类号： TP 391.41 文献标识码：A Designed for Pavement Distress Images LI Li, SUN Lijun, CHEN Zhang （Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University, Shanghai 201804, China ） Abstract: An edge detection method designed for pavement images was presented in this study to overcome the difficulties in traditional methods of edge detection for pavement images such as nonuniform background illumination, low proportion of distress pixels and irregular directions of distress. The method offered in this study brought image preprocessing and edge enhancement to traditional edge detection methods. The image preprocessing was composed of background correction, Gaussian smoothing and histogram transformation based on the conception of effective range of gray levels which was proposed by this study for the first time, while the edge enhancement included the morphological dilation and median filtering operation. A pavement distress image was tested by Sobel detector of eight directions and Ostu method following the procedure above. The result showed that the procedure presented in this study can dramatically dampen the impact of noise on edge detection, and the background corrections as well as the histogram transformation based on effective range of gray levels are the key components of the method. Besides, Ostu method can provide good segmented edge maps if the images have been preprocessed. Key words: pavement distress; image processing; edge detection; preprocessing; effective range of gray levels 数字图像中的边缘代表像素灰度的阶跃，是图像的基本特征之一[1]。路面的开裂、坑槽等破损类病害反映在图像中都可以用边缘来描述，因此可以用边缘检测来分割路面破损图像，为路面病害识别奠定基础。 边缘检测的实质是一种模板（Mask）操作，用边缘检测算子构造模板在每个像点的邻域内进行卷积运算，输出边缘幅度后通过阈值比较判定是否为满足预设条件的边缘点。目前Sobel算子和Canny算子在路面图像检测中应用较多[1~5]，其中Sobel算子的优势为运算速度快、边缘定位精确，而Canny算子检测的边缘较为完整。经典的阈值确定方法有最大类间方差法（Ostu法）[6]、最大熵法[7]、极差法[8]、Fisher准则判别法[9]等，其中最大类间方差法由于运算简便迅速而应用最为广泛。 改进的边缘检测方法概述 尽管边缘检测方法众多，但针对路面破损图像的检测却通常