HSIs contain a wealth of spatial and spectral information, but they face the following challenges in real-world applications: 1) deciphering complex spatial dependencies among geographic entities is quite difficult; 2) the phenomena of spectral ambiguity, where different objects have the same spectrum brings considerable noise into the HSI; and 3) the cost of labeling HSI is prohibitive. Although transfer learning offers a way to exploit existing labeled HSI to mitigate the annotation demands in the target domain, most methods fail to fully align the conditional distributions between the two domains. To this end, this article introduces a novel FTGN. We employ GraphSAGE to discern the spatial dependencies embedded within HSI and implement NWS to alleviate the impact of noise information. Thereafter, we adopt a pseudolabel trimming strategy, securing high-confidence pseudolabels. This ensures that samples from the target domain participating in the training process accurately mirror the...

Affected by the sensor, shooting environment, and other aspects, hyperspectral images (HSIs) in the source and target domains exhibit phenomenon of difficult feature extraction and domain shift. The above phenomena pose challenges to the cross-scene HSI classification task. Therefore, a focal transfer graph network (FTGN) for cross-scene HSI classification is proposed. First, FTGN leverages graph sample and aggregate (GraphSAGE) to capture spatial–spectral features by aggregating partial adjacency nodes, ensuring the acquisition of contextual information. The neighbor weighting strategy based on spatial–spectral information is proposed to solve the information interference caused by excessive node aggregation. Second, a pseudolabel trimming strategy based on class metrics is proposed to reduce the adverse effects of pseudolabel noise in the transfer process. Then, a specification subdomain adaptation (SSA) module is proposed, which helps to achieve effective distribution alignment by r...