Abstract: An algorithm is presented to measure atmospheric visibility by using a video camera, which is the equipment for monitoring the traffic. A series of targets with known distances are used. The image which encompasses the targets is captured from video cameras, and impulse noise is removed by a median filter. The area of first target can be divided up from the image by using the algorithm of sum of absolute differences, and the other targets can be segmented by known correlation position with the first target. The feature of the segment portion is extracted from the contrast of the edges of the targets which is measured with Kohler’s thresholding technique. The exponential curve is fitted with the data of the feature and the known distance of the targets. The visibility is decided by the contrast threshold recommended by international commission on illumination. Experimental results show that the algorithm is effective and robust.

Key words: visibility detection, image processing, pattern recognition, feature extraction

摘要： 针对现有气象能见度仪价格昂贵且无法全程分析的不足，提出一种利用路况视频图像的能见度检测算法，通过目标物预置获得距离信息，通过图像预处理得到含有目标物的去噪图像。利用目标物的相互关系和SAD图像匹配算法从图像上分割出目标物，利用Kohler的一致对比度法实现目标物图像的人眼模拟，进而对其特征进行拟合，得出能见度值，实现基于路况视频的能见度检测。实验结果表明，该算法与人眼观察一致，对噪声有较好的鲁棒性。

关键词: 能见度检测, 图像处理, 模式识别, 特征提取

CLC Number: 

• TP391.43