The invention provides a local climate region classification method based on cross-
modal fusion, and relates to the technical field of geographic
information science. According to the specific implementation mode, three kinds of
modal data are input into a feature
encoder which is of the same structure but does not share parameters, and the feature
encoder is composed of a multi-scale
convolution module and channel-space attention residual modules which are cascaded in sequence; each channel-space attention residual error module obtains respective layer features of three types of
modal data by sequentially progressive the output features of the multi-scale
convolution module and serially connecting residual error
feature extraction operations of channel dimension attention and space dimension attention; sequentially progressive cross-modal explicit fusion operation is carried out on respective layer features of the three types of
modal data through a plurality of cascaded cross-modal
Transformer modules, and layer fusion features are obtained; and through a Token-level
feature fusion module, performing Token-dimension
adaptive weighting and fusion operation on the first-layer features and the first-layer fusion features of the multispectral
satellite remote sensing image data, and generating feature representation with global discrimination capability for classification. According to the technical scheme, the local climate region classification precision can be improved.