Multi-frequency PolSAR Image Fusion Classification Based on Semantic Interactive Information and Topological Structure

Compared with the rapid development of single-frequency polarimetric SAR (PolSAR) image classification technology, there is less research on land cover multi-frequency PolSAR (MF-PolSAR) images. And deep learning methods among them are mainly based convolutional neural networks (CNNs), only local spatiality considered but nonlocal relationship ignored. Therefore, this paper proposes MF semantics and topology fusion (MF-STF) model semantic interaction topological structure to improve MF-PolSAR performance. During MF-STF optimization, information-based (SIC) property-based (TPC) work collaboratively, not fully leveraging complementarity bands, also combining spatial information discrimination different categories. For SIC, designed cross-band interactive feature extraction (CIFE) module embedded explicitly correlation thereby bands make ground objects more separable. In TPC, graph sample aggregate network (GraphSAGE) employed dynamically capture representation relations between way, robustness can be further improved by information. Finally, a weighted (MFWF) strategy proposed merge inference from so as joint decisions SIC TPC. Notably, its weights adjusted total loss. The effectiveness modules proved ablation experiments three measured datasets. addition, comparative show that achieve competitive performance than some state-of-the-art methods.