基于无人机航拍图像的公路标线检测算法-《计算机技术与发展》

文章信息/Info

Author(s):: WANG Li-chun; LIU Ning-zhong; LI Qiang-yi; School of Computer Science and Technology,Nanjing University of Aeronautics and Astronautics,Nanjing 211100,China

Keywords:: aerial images; road markings detection; deep learning; Faster R-CNN; UAV

摘要:: 随着公路建设的迅速发展,对公路养护的任务量与日俱增,传统的人工查看方式已不再适用,利用无人机航拍对公路标线进行自动化检测无疑是一种更好的检测方式。而对于公路标线的自动化检测分析,首要的便是对图像中的公路标线进行检测。基于无人机航拍公路图像,提出一种针对公路航拍图像的公路标线检测算法。首先依据路面的颜色特征以及梯度特征进行路面分割,并提取分割图像连通区域,然后将深度学习物体检测领域的 Faster R-CNN 算法与连通区域颜色面积特征相结合进行非标线区域的过滤,最后提取未过滤的标线区域作为公路标线提取结果。实验结果表明,该算法适用性强,运行效率高,针对不同的公路均具有较高的准确率。

Abstract:: With the rapid development of highway construction,using traditional artificial way to detect breakage of road marking is no longer applicable. Because of the maturation of UAV aerial image,it is undoubtedly a better way to realize automatic detection based on aerial images. However,we need to extract the road marking first in an image for automatic detection of road breakage. Based on UAV aerial image,we propose a road marking extracting algorithm. First,we segment road surface with the color feature and the gradient feature and then extract connected regions in an image. Next,we combine faster R-CNN algorithm of object detection field with traditional features such as the color feature and the area feature for filtering the non-marking regions. Finally,we obtain road marking regions as results. The experiment shows that this algorithm has strong applicability and high efficiency with high accuracy for different roads.