GLN-LRF: Global Learning Network based on Large Receptive Fields for Hyperspectral Image Classification

mengyun dai¹tianzhe liu²youzhuang lin¹zhengyu wang³yaohai lin^1*Changcai Yang^1*RIQING CHEN¹

• ¹福建农林大学, 福州市, China
• ²Fujian Police College, Fuzhou, Fujian Province, China
• ³Newford Research Institute of Advanced Technology, wenzhou, China

Deep learning has been widely applied to high-dimensional hyperspectral image classification and has achieved significant improvements in classification accuracy. However, most current hyperspectral image classification networks follow a patch-based learning framework, which divides the entire image into multiple overlapping patches and uses each patch as input to the network. Such locality-based methods have limitations in capturing global contextual information and incur high computational costs due to patch overlap. To alleviate this problem, we propose a global learning network with a large receptive fields network (GLNet) to capture more comprehensive and accurate global contextual information and to enrich the underlying feature representation for hyperspectral image classification. The proposed GLNet is an encoder-decoder architecture with skip connections. In the encoder phase, a large receptive field context exploration (LRFC) block is proposed to extract multi-scale contextual features. LRFC block enables the network to enlarge the receptive field and capture more spectral-spatial information. In the decoder phase, to further extract rich semantic information, a multi-scale simple attention (MSA) block is proposed, which extracts deep semantic information by using multi-scale convolution kernels and fusing the obtained features with SimAM. Specifically, GLNet achieved overall accuracies (OA) of 98.72%, average accuracies (AA) of 98.63%, and Kappa coefficients of 98.3% on the IP dataset; similar improvements were observed on the PU and HOS18 datasets, confirming its superior performance compared to the baseline models.