软件导刊2025，Vol.24Issue(6)：160-167,8.DOI:10.11907/rjdk.241216
适用于飞行模拟器的城市遥感图像去雾研究
Urban Remote Sensing Image Dehazing Study for Flight Simulators
摘要
Abstract
Flight simulators often use remote sensing images with different spatial resolutions to construct 3D terrain,and remote sensing im-ages of urban areas are more susceptible to haze than other areas such as forests and mountain ranges,resulting in problems such as low con-trast and blurred details.In order to solve the above problems,a deep learning-based dehaze model is proposed,which is able to dehaze urban remote sensing images with multiple spatial resolutions and different architectural styles,so as to satisfy the needs of flight simulators for the construction of the view system.Based on the encoder-decoder model,a multi-scale attention module and a gated kernel convolution module are proposed,which are able to aggregate local and global information,and introduce fine attention to complete the detail supplementation while realizing coarse attention to present the overall information,thus realizing the effective dehazing in urban scenes.Experiments show that the proposed model achieves a PSNR of 30.48,SSIM of 0.962,and LPIPS of 0.028 under mixed spatial data,which is superior to other meth-ods compared in this paper,and has good generalization to realize effective dehaze under different spatial resolutions and different urban archi-tectural styles.关键词
城市遥感图像/遥感图像去雾/深度学习/飞行模拟器Key words
urban remote sensing images/remote sensing image dehazing/deep learning/flight simulator分类
计算机与自动化引用本文复制引用
刘琪,王博,戈文一,谭诗瀚,邹书蓉..适用于飞行模拟器的城市遥感图像去雾研究[J].软件导刊,2025,24(6):160-167,8.基金项目
民航飞行技术与飞行安全重点实验室开放基金项目(F2019KF02,FZ2022KF08) （F2019KF02,FZ2022KF08）
四川省科技计划重点研发项目(2023YFG0304,2023YFG0026) （2023YFG0304,2023YFG0026）
