Precipitation Intensity Recognition Based on Convolution Neural Network with Fused Encoded and Decoded Features

In order to efficiently use infrared precipitation images to determine the precipitation intensity, a precipitation intensity recognition model with fused encoded and decoded features has been proposed. The coding and decoding convolution is introduced into the deep convolution neural network classification model, which can extract the deep-seated features of rain information while reducing the loss of local information. In the coding and decoding convolution module, multi-scale receptive field convolution is considered, and local features in different ranges are fused. At the same time, coding and decoding convolution feature maps of the same scale are fused during decoding, so as to improve feature utilization. Thus, a precipitation intensity recognition model integrating coding and decoding convolution features is constructed. The proposed model has the highest classification accuracy of 91. 7% compared to state-of-the-art methods. Moreover, an ablation experiment demonstrates the effectiveness of the proposed encoded and decoded model.