计算机技术与发展2025，Vol.35Issue(3)：9-17,9.DOI:10.20165/j.cnki.ISSN1673-629X.2024.0319
基于多尺度残差增强网络的DEM超分辨率重建
DEM Super-resolution Reconstruction Based on Multi-scale Residual Enhancement Network
摘要
Abstract
Digital Elevation Models(DEMs)are considered one of the most important foundational geographic data models,with widespread applications in hydrological analysis,path planning,and modeling.However,the high cost of acquiring large-area,high-resolution DEM data with more precise sensors poses a challenge for many geographic analysis applications.Combining multi-scale features,residual learning,and multi-scale channel attention mechanisms,we propose a digital elevation model super-resolution reconstruction method based on a Multi-Scale Residual Multi-Channel Attention Enhancement Network.The Multi-Scale Residual Multi-Channel Attention Enhancement Module(MRCAEM)utilizes a combination of convolutional layers with multiple different kernel sizes,and through the multi-scale channel attention mechanism,it better captures semantic information at different scales,refines multi-scale feature extraction,and reconstructs more realistic high-resolution DEMs through feature fusion and reconstruction modules.Experimental results show that the proposed method reduces the Root Mean Square Error(RMSE)by approximately 2%～30%compared to other methods.关键词
数字高程模型/超分辨率重建/多尺度/残差融合网络/多尺度通道注意力/可变形卷积Key words
digital elevation model/super-resolution reconstruction/multi-scale/residual fusion network/multi-scale channel attention/deformable convolution分类
信息技术与安全科学引用本文复制引用
韩超,张晓滨..基于多尺度残差增强网络的DEM超分辨率重建[J].计算机技术与发展,2025,35(3):9-17,9.基金项目
陕西省自然科学基础研究计划项目(2023-JC-YB-568) （2023-JC-YB-568）
