Visibility Detection Method Based on Camera Model Calibration
Jie Tong Zou · 2017
Low visibility is more likely to cause accidents, in order to fulfill the requirement of whole-distance surveillance as well as make full use of the resources surveillance cameras along the highway, a visibility detection Method based on the video camera calibration was proposed in this paper. At first, match the image block by the absolute difference of mismatch and SAD algorithm. After that, extract video image edge feature using the pre-coordinate information and then calculate the image contrast. Next, use the conversion formula of image pixel luminance contrast and spatial luminance contrast to calculate image eigenvectors of spatial luminance contrast. Finally the relationship between the image eigenvalues and the actual distance can be obtained by means of nonlinear least square fitting method, thus achieving visibility detection. The method is more realistic and stable than the dark channel algorithm. Then putting the algorithm into the highway, and the effect is obvious.