无需人工标记的视频对比度道路能见度检测  

Visibility Detection Based on Traffic Video Contrast Analysis without Artificial Markers

作　　者：李勃[1] 董蓉[1] 陈启美[1] 

机构地区：[1]南京大学通信技术研究所,南京210093

出　　处：《计算机辅助设计与图形学学报》2009年第11期1575-1582,共8页Journal of Computer-Aided Design & Computer Graphics

基　　金：中国交通部科技攻关项目(200435333204);江苏省交通科学研究项目(05x008)

摘　　要：为了解决传统的能见度仪价格昂贵、采样有限,及已有的一些视频测量手段需人工标记物、稳定性差等不足,提出基于路况视频对比度的能见度检测算法,进而构建无需人工标记的能见度检测系统.通过分析车道分割线,提取兴趣域,以确保所选像素的高度一致;解析各像素相对于其四邻的对比度,所取的最大值若大于给定阈值即为人眼可分辨像素;结合摄像机标定来计算距摄像机最远的可视像素,并通过Kalman滤波器滤除干扰,得到能见度值.该系统充分利用已有的路况图像,稳定性高、成本低、检测精度高,具有广阔的应用前景.Due to the fact that traditional visibility meter was expensive in addition to the limited sampling region, and the existing video measurement methods need artificial markers while the stability of output was poor, a visibility detection algorithm based on traffic video contrast analysis was brought forward, and then a visibility detection system without artificial markers was further developed. First, the region of interest was figured out by detecting the lane boundaries to ensure the height constancy of the selected pixels. Then the contrast values of each pixel to its 4-neighbors were analyzed, and the pixel whose maximum contrast value exceeds the preset threshold was considered as the distinguishable one by human operator. Finally, combined with camera calibration technology, the maximum distinguishable pixel to the camera was selected, and the distance value was further smoothed by Kalman filter, and the outcome was regarded as the current visibility distance. By making full use of existing traffic videos, with good stability, low-cost, high accuracy, this system has broad application prospects.

关 键 词：能见度 视频监控 像素对比度 兴趣域检测 摄像机标定 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]