Classification of Weather Phenomenon From Images by Using Deep Convolutional Neural Network

Weather phenomenon recognition notably affects many aspects of our daily lives, for example, weather forecast, road condition monitoring, transportation, agriculture, forestry management, and the detection of the natural environment. In contrast, few studies aim to classify actual weather phenomenon images, usually relying on visual observations from humans. To the best of our knowledge, the traditional artificial visual distinction between weather phenomena takes a lot of time and is prone to errors. Although some studies improved the recognition accuracy and efficiency of weather phenomenon by using machine learning, they identified fewer types of weather phenomena. In this paper, a novel deep convolutional neural network (CNN) named MeteCNN is proposed for weather phenomena classification. Meanwhile, we establish a data set called the weather phenomenon database (WEAPD) containing 6,877 images with 11 weather phenomena, which has more categories than the previous data set. The classification accuracy of MeteCNN on the WEAPD testing set is around 92%, and the experimental result demonstrates the superiority and effectiveness of the proposed MeteCNN model. Realizing the automatic and high-quality classification of weather phenomena images can provide a reference for future research on weather image classification and weather forecasting.