BIVS: Block Image and Voting Strategy for Weather Image Classification
Run Ye, Bin Yan, Junhua Mi · 2020
Timely and accurate weather information is important for smart grid systems, autopilot systems and intelligent surveillance systems. This paper studies how to obtain weather information from a single image. The biggest challenge of weather image classification task is that there can be the same objects and features in images representing different weather conditions. To address this problem, first of all, this paper constructs the weather image dataset under outdoor transmission line scene, including images of foggy, rainy, snowy and sunny. Then, a weather image classification method based on block image and voting strategy is proposed. The method of block image and voting strategy achieves 98.74% classification accuracy in weather image dataset.