| 引用本文: | 杨天麟,王卫杰,康楠,朱文清.采用改进暗通道先验算法的高速公路能见度检测[J].哈尔滨工业大学学报,2023,55(3):100.DOI:10.11918/202111066 |
| YANG Tianlin,WANG Weijie,KANG Nan,ZHU Wenqing.Highway visibility detection using improved dark channel prior algorithm[J].Journal of Harbin Institute of Technology,2023,55(3):100.DOI:10.11918/202111066 |
|
| 摘要: |
| 为提升高速公路雾天能见度检测精度,考虑大气透光强度、透射率、大气消光系数和图像中某点到摄影机的实际距离,通过能见度检测原理改进现有的暗通道先验算法。首先结合矩形区域测距和实际场景物体大小来解决高速公路二维场景到三维场景重构的问题,并利用K-means聚类的方法,并找出聚类后的视频图像中分界线的最小景深点,结合所构建的测距模型得到该点到摄像机的实际距离。其次通过发现传统暗通道先验理论在求取大气透光强度方面的不足,提出了基于图像分割的局部熵法来求取大气透光强度,再利用暗通道先验理论求出透射率,然后由能见度检测原理计算出能见度。最后,根据日兰高速公路K113+000处雾天下的视频图像,对改进暗通道先验算法与传统暗通道先验算法进行实验对比,并以能见度检测仪的检测结果为参照。结果表明:当实际能见度为100 m左右时,改进算法检测的平均相对误差（MRE）为6.25%,比传统算法减小了2.38%；当实际能见度为150 m左右时,改进算法检测的MRE为6.17%,比传统算法减小了3.06%；当实际能见度为200 m左右时,改进算法检测的MRE为5.71%,比传统算法的减小了3.41%。随着光照强度的增加,改进暗通道先验法的检测精度提升越来越明显,能更好地适应日间大雾天气下的高速公路。 |
| 关键词: 公路运输 改进的暗通道先验算法 测距模型 局部熵法 视频图像 高速公路能见度 雾天环境 |
| DOI：10.11918/202111066 |
| 分类号:U495.53 |
| 文献标识码:A |
| 基金项目:江苏省研究生科研与实践创新计划 (SJCX21_0547) |
|
| Highway visibility detection using improved dark channel prior algorithm |
|
YANG Tianlin,WANG Weijie,KANG Nan,ZHU Wenqing
|
|
(College of Transportation Engineering, Nanjing Tech University, Nanjing 211816, China)
|
| Abstract: |
| Based on the visibility detection principle, this paper aims to enhance the accuracy for highway visibility detection in foggy weather through improving the existing dark channel prior algorithm by considering the atmospheric transmittance intensity, transmittance, atmospheric extinction coefficient, and the actual distance from a point in the image to the camera. Firstly, the problem of reconstructing 2D scene to 3D scene on the highway was solved by combining the rectangular area ranging and the actual scene object size. The K-means clustering algorithm was used to obtain the point with the minimum depth of field on the dividing line of the video image. The actual distance from the point to the camera was obtained by using the constructed ranging model. Secondly, in view of the deficiency of the traditional dark channel prior theory in solving the atmospheric transmittance intensity, a local entropy method based on image segmentation was proposed to obtain the atmospheric transmittance intensity. Then, the transmittance was obtained by using the dark channel prior theory, and the visibility was calculated by the visibility detection principle. Finally, according to the video image of K113+000 of Rilan Highway in foggy weather, the improved dark channel prior algorithm was compared with the traditional dark channel prior algorithm experimentally, and the results were referenced to the detection results of visibility detector. Results showed that when the actual visibility was about 100 m, the mean relative error (MRE) detected by the improved algorithm was 6.25%, which was 2.38% less than that of the traditional algorithm. When the actual visibility was about 150 m, the MRE detected by the improved algorithm was 6.17%, which was 3.06% less than that of the traditional algorithm. When the actual visibility was about 200 m, the MRE detected by the improved algorithm was 5.71%, which was 3.41% less than that of the traditional algorithm. With the increase in light intensity, the improved dark channel prior method has higher detection accuracy and can better adapt to the highway in daytime foggy weather. |
| Key words: road transportation improved dark channel prior algorithm ranging model local entropy method video image highway visibility foggy environment |
