Visibility Analysis in Fog Weather Based on Convolution Neural Network and Transfer Learning

In this paper, a deep learning model based on convolutional neural network and transfer learning is adopted to realize the recognition and estimation of airport video images, and to evaluate the accuracy of the estimated visibility. In order to highlight the advantages of the transfer learning method, we have established a deep learning model and a machine learning model respectively. In the traditional image recognition method, 11-dimensional information features of the image are recognized, and BP neural network model and evaluation accuracy are established; On the deep learning model, we first match the airport video with the visibility data, and determine the parameters through the pretraining model. Then we randomly divide the image and data into training sets and testing sets at a ratio of 7:3. Fix the pre network parameter training and adjust the deep learning model. After 100 iterations, the accuracy of the training set and test set respectively reached 80.48% and 79.63%. The results show that the deep learning model based on convolutional neural network and transfer learning does have higher efficiency and recognition accuracy.