Abstract: 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.