“In response to the problem of imbalanced relationship between local and global features in hyperspectral image classification, experts have constructed a Gaussian cross feature fusion model, which dynamically weights and fuses features through specific algorithms to ensure a balance between the two. Experiments were conducted on publicly available datasets of crops, urban land features, agricultural vegetation, and heterogeneous mixed scenes. Compared with 11 mainstream methods, this method showed the best performance, demonstrating its advantages and good generalization ability in multi scene classification.”