摘要
【目标】为避免沥青路面航拍图像中非路面背景及车道线对车道区域病害检测的干扰,提出了综合颜色及边缘信息的车道区域提取方法。【方法】首先,在沥青路面航拍图像中,结合车道区域颜色对应的先验信息,在HSV颜色空间内利用颜色分割法进行图像分割,并根据车道区域与图像边界相交的特征去除了颜色过分割区域;然后,利用Canny算子得到车道区域边缘信息,使用双边滤波改善高斯滤波在去噪过程中导致的边缘信息丢失问题,并引入两遍扫描法以检测与标记边缘检测图像中的连通域;最后,通过运算将颜色分割二值图像和连通域标记图像融合,在融合图像中依据面积阈值确定车道区域的标记值后,结合标记图像与航拍原图像完成车道区域的提取。利用376张多种场景下的航拍沥青路面图像对所提方法进行试验验证。【结果】本研究方法提取车道区域的平均分割精度、过分割率、欠分割率和病害丢失率分别为98.38%,1.96%,1.56%,1.60%,优于颜色分割方法、区域生长法和基于DeepLab v3+的语义分割方法。【结论】本研究方法可有效改善过分割和欠分割问题,具有较好的准确性与稳定性,为后续路面病害的准确检测奠定了良好的基础。
[Objective]To avoid the interference of non-pavement background and lane lines on lane area disease detection in asphalt pavement aerial images,the lane area extraction method,integrating color and edge information,was proposed.[Method]First,in asphalt pavement aerial images,the image segmentation was performed in HSV color space by using color segmentation,integrating the prior information corresponding to lane area color.According to the intersection characteristics of lane area and image boundary,the color over-segmentation areas were removed.Then,Canny operator was utilized to obtain the edge information of lane area.The bilateral filtering was used to improve the loss of edge information caused by using Gaussian filtering in noise reduction process.The two-pass scanning method was introduced to complete the detection and labeling of connected domain of edge detection images.Finally,the color-segmented binary images and labeled images were fused through calculation.In the fused images,the labeled values of lane area were determined according to the area threshold.The labeled images and original aerial images were combined to complete lane areas extraction.More than 376 aerial asphalt pavement images in different scenarios were used to validate the proposed method.[Result]The average segmentation accuracy,over-segmentation rate,under-segmentation rate and disease loss rate,by using the proposed method for extracting lane area,are 98.38%,1.96%,1.56%and 1.60%respectively,which are superior to color segmentation method,region growing method and semantic segmentation method based on DeepLab v3+.[Conclusion]The proposed method can effectively improve the over-segmentation and under-segmentation problems,and has better accuracy and stability.It lays a foundation for the accurate detection of subsequent road pavement diseases.
作者
夏晓华
苏建功
刘洋
李明臻
陈仕旗
XIA Xiaohua;SU Jiangong;LIU Yang;LI Mingzhen;CHEN Shiqi(Key Laboratory of Road Construction Technology and Equipment of Ministry of Education,Chang’an University,Xi’an,Shaanxi 710064,China)
出处
《公路交通科技》
北大核心
2025年第5期108-117,共10页
Journal of Highway and Transportation Research and Development
基金
陕西省重点研发计划项目(2019GY-116)。
关键词
道路工程
航拍路面图像
车道区域提取
颜色分割
边缘检测
road engineering
aerial pavement images
lane area extraction
color segmentation
edge detection
作者简介
通讯作者:夏晓华(1987-),男,山东临沂人,博士,副教授,研究方向为机器视觉与光机电一体化。(xhxia@chd.edu.cn)。
