Remote sensing image semantic segmentation method based on channel attention feature fusion

The invention discloses a remote sensing image semantic segmentation method based on channel attention feature fusion, and the method comprises the steps: inputting a to-be-predicted remote sensing image into a feature extraction module, extracting the initial features of four scales, inputting the extracted features into a global information module, enlarging the receptive field, and fully utilizing the global context information. Then sequentially through an attention fusion module, gradually obtaining features after multi-scale channel attention fusion; the features after attention fusion are sequentially processed through an attention decoding module, decoded features are obtained step by step, finally the last decoded feature is processed through convolution and up-sampling, and then a Softmax classifier is used to obtain a final semantic segmentation result. According to the method, dynamic and adaptive feature fusion is carried out in a context sensing mode, deviation caused by contextual information a