Cloud-aerosol classification method based on CALIPSO satellite-borne laser radar noise reduction technology

The invention discloses a cloud-aerosol classification method based on a CALIPSO satellite-borne laser radar noise reduction technology, and the method comprises the steps: correcting CALIPSO vertical feature mask data with an obvious COSCA error, and adjusting the CALIPSO L1 data and the vertical feature mask data to be consistent in spatial resolution through resampling; taking a signal-to-noise ratio, a root mean square error and a structural similarity index as evaluation indexes, automatically selecting an optimal noise reduction method from the noise reduction methods to be selected to perform noise reduction on the input data, and performing normalization processing on all training set data; constructing a cloud-aerosol classification model and an aerosol classification model, wherein the classification model is formed by adding a self-attention mechanism module, a residual connection module and a pyramid pooling module between an encoder part and a decoder of a U-Net neural network model; and predicti