Knowledge distillation-based lightweight remote sensing image semantic segmentation method and device

The invention discloses a lightweight remote sensing image semantic segmentation method based on knowledge distillation. Comprising the following steps: constructing a network comprising a branch CNN for extracting local spatial features, a branch Transform for extracting global spatial features, a feature fusion module for fusing the local spatial features and the global spatial features, and a decoder for decoding the fused features; the method comprises the following steps: replacing a CNN (Convolutional Neural Network) and a Transform with ResNet50 and SegFormer-B3 to obtain a teacher network; the CNN and the Transform are replaced by the ResNet18 and the SegFormer-B0, so that a student network can be obtained; a fusion feature distillation module based on dynamic temperature is constructed to jointly train a student network and a teacher network and dynamically search temperature parameters, and a structured distillation method is used in a feature level, so that the student network pays attention to fea