Cloud Detection Method Using Convolutional Neural Network Based on Cascaded Color and Texture Feature Attention
Cloud detection is of great significance in the sub-sequent analysis and application of remote sensing images, and it is a critical part of remote sensing image preprocessing. In recent years, the convolutional neural network is widely used in cloud detection. In this paper, we introduce the Cascaded Feature Attention Network(CFAN), a convolutional neural network for cloud detection. The CFAN uses a cascaded feature attention module(CFAM) to improve the feature extraction capabilities of the network. Our proposed CFAM contains color feature attention module and texture feature attention module, not only use color feature to assist the extraction of information but also use texture feature to help the network to learn more detailed information. The model is trained and evaluated in the GF-1 WFV data, the dataset contains complex clouds and different types of underlying surfaces. The experimental results show that the evaluation indicators of our method have reached a high level, which proves the effectiveness of our method.