Visibility detection of single image in foggy days based on Fourier transform and convolutional neural network

Visibility is a very important meteorological observation factor, the reduction of which will cause inconvenience to traffic, and even cause traffic accidents. It is of great significance to estimate and monitor the visibility. In this paper, a foggy image visibility detection algorithm based on Fourier transform and convolution neural network was proposed. Firstly, Fourier transform was used to obtain the image characteristics in the frequency domain, which can reflect the general law of the image changing with visibility. Secondly, convolutional neural network was established to predict the image visibility. The experimental results show that the network convergence speed can be improved by processing the samples with Fourier transform. The detection accuracy can reach about 74.59%.