文献类型：会议论文

英文题名：A Novel Real-time Highway Visibility Measurement System Based on Dark Channel Prior

作者：Ye, Jiongyao[1];Jin, Yue[1];Wang, Nan[1]

机构：[1]East China Univ Sci & Technol, 130 Meilong Rd, Shanghai, Peoples R China

会议论文集：International Conference on Image and Graphics Processing ICIGP)

会议日期：FEB 24-26, 2018

会议地点：Hong Kong, HONG KONG

语种：英文

外文关键词：Visibility measurement; dark channel prior; region growing; camera calibration

摘要：Low visibility is more likely to cause accidents on the highway. An accurate visibility measurement system can greatly reduce these accidents and improve driving safety. However, the existing visibility measurement systems are very expensive and cannot send the visibility information to drivers in real time. For these reasons, we propose a novel real-time highway visibility measurement system based on dark channel prior in this paper. This proposed system is running on the on-board camera, so that the cost of the measurement system can be very low, meanwhile, drivers can achieve the visibility information in real time. Firstly, the transmission, which is obtained through dark channel prior, is used to judge whether the weather is in low visibility condition or not. Secondly, the measurement bandwidth in the transmission image is determined by region growing algorithm, and the measurement line in this bandwidth is found according to the gradient variation of transmission. Thirdly, an improved real time automatic calibration method is proposed according to the highway lane line in the transmission image, and the visibility distance can be estimated finally. Experimental results show the effectiveness of proposed highway visibility measurement system, which can estimate the real-time highway visibility distance in the low visibility weather.