Fine-grained meta-learning with semantic augmentation for synthetic aperture radar change detection

Synthetic aperture radar (SAR) images are widely used for change detection due to their all-weather, all-day imaging capabilities. However, distinguishing between changed and unchanged areas is difficult because of complex scattering and the long-tailed distribution, where the changed areas are few. To address these challenges, we propose a fine-grained meta-learning with a semantic augmentation method for SAR image change detection. First, we propose a fine-grained classification strategy based on edge detection to construct a small, balanced dataset for training the meta-learner and construct a dataset with reduced imbalance for training the backbone classifier. This strategy improves the feature learning of hard-to-classify samples and mitigates class imbalance effects. Next, we design a lightweight yet effective multi-layer perceptron-based meta-learner and incorporate semantic data augmentation. The meta-learner automatically augments minority classes along meaningful semantic directions by learning appropriate class-wise covariance matrices, thereby improving the detection performance of the backbone classifier. We evaluated our method on four SAR datasets through cross-dataset experiments, demonstrating its superiority over six methods in effectiveness and robustness. Imbalance analysis further reveals that as the imbalance ratio increases, the performance of comparison methods degrades significantly, whereas our method exhibits the least performance deterioration, confirming its competent ability to handle imbalanced data.