Multi-task Fine-grained Feature Mining for Multi-label Remote Sensing Image Classification

Multi-label remote sensing image classification can provide comprehensive object-level semantic descriptions of remote sensing images. However, most existing methods cannot fully mine the fine-grained features of images and labels, resulting in low classification accuracy. To address this issue, we propose a novel multi-task framework for multi-label remote sensing image classification. The framework establishes the class-specific feature extraction as a binary classification auxiliary task to assist the main multi-label classification task, which can improve the model’s local and global feature extraction ability. Meanwhile, the framework updates the label correlation graph using the graph transformer layer to accurately identify label node pairs with potential correlation, which effectively mines the correlation of multiple labels to generate more accurate label co-occurrence embedding for image label prediction. Experimental results on UCM, AID and DFC15 multi-label datasets show that the proposed method outperforms existing state-of-the-art methods.