Dual decoupling semantic segmentation model for high-resolution remote sensing images

Semantic segmentation is one of the core contents of high spatial resolution remote sensing images analysis and understanding. The existing semantic segmentation network based on deep learning will lead to the loss of high-frequency information and inaccurate edge segmentation of remote sensing images. Aiming at this problem,this study designs a dual decoupling semantic segmentation network model to improve the semantic segmentation performance of high-resolution remote sensing images. The extracted two-level feature maps are decoupled into edge features with high-frequency characteristics and body features with low-frequency characteristics,and the decoupled edge and body feature maps are fused. Furthermore,a loss function considering edge and body is proposed to optimize the ground feature elements.Experiments on ISPRS Vaihingen and ISPRS Potsdam 2D high spatial resolution remote sensing image datasets. Compared with the results of the existing remote sensing images semantic segmentation network model,the dual decoupling semantic segmentation network model can effectively improve the segmentation accuracy of ground feature elements. © 2023 SinoMaps Press. All rights reserved.