PTZ visibility detection based on image luminance changing tendency

Due to the fact that the existing shortage of visibility detection methods, a novel algorithm is presented to measure atmospheric visibility base on image luminance changing tendency using PTZ (Pan/Tilt/Zoom) video cameras. First, the background of image was used to remove the interference from cars and passengers, and the region of interest was figured out to ensure the height constancy of the selected pixels. Then the homogeneous road surface, without roadbed or lane marks, to ensure the luminance constancy of the selected pixels. From the selected road region, combined with camera calibration algorithm, the luminance curve which reflect that road luminance vary with the distance, was extracted to search the feature point, and then the extinction coefficient is calculated, finally, the maximum visibility distance was obtained. The experimental results show that the algorithm is consistent with observation characteristic of human eyes, and the accuracy is up to 96% while the inspection error within 10 meters. Compared with other visibility detection methods, this algorithm is more effective, robust and precise.