Answers are grounded in this paper and cite their sources.
With the rapid development of remote sensing technology, effectively managing and retrieving vast quantities of remote sensing images has become an urgent challenge. For complex remote sensing images, multi-label annotation provides more precise scene characterization. However, existing methods inadequately model inter-label dependencies, suffer from computationally intensive feature extraction, and scale poorly to large-scale datasets, which collectively hinders their real-world deployment. To address these issues, this paper presents a novel hashing method for multi-label remote sensing image retrieval, which introduces a structured semantic dictionary to guide both feature extraction and hash code learning. The proposed framework features three dedicated components that work synergistically: a label dependency modeling module that constructs a semantic dictionary to explicitly capture inter-label correlations; a spatial-aware feature extractor based on an enhanced FasterNet, which dynamically balances local and global feature representation; and a semantic-dictionary guided joint loss function that aligns image representations with the semantic dictionary while ensuring hash code compactness and discreteness. By integrating structured semantic knowledge directly into the optimization process, our method learns highly discriminative hash codes. Extensive validation on the UCMD, AID, and MLRSNet multi-label datasets demonstrates the method's effectiveness in retrieving images with diverse and complex label combinations. Through unified feature-semantic optimization, this work provides a practical and effective solution for high-precision retrieval in large-scale remote sensing images.
No takes yet. Share an insight, caveat, or question.
He et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: