Semantic Segmentation for Remote Sensing Images Based on Global And Local Features Integration

With the application of convolutional neural network (CNN) and transformer in semantic segmentation, more and more effective methods have been proposed. However, the above methods still need to be improved at the level of feature analysis and integration. Therefore, we propose a semantic segmentation method based on global and local features integration (GLFI), which firstly uses CNN to obtain local features and swin transformer to extract global features, and then combines the adaptive strategy and dilation convolution to design the feature integration way, and finally the semantic segmentation results can be obtained. Experimental results on the public dataset show that the proposed GLFI achieves better performance.