摘要
卫星捕获的遥感数据容易受到成像过程中悬浮粒子的影响而造成图像雾化现象,极大地影响遥感图像的清晰度。为了弥补这一不足,遥感图像去雾(RSID)非常必要。最近兴起的状态空间模型State Space Model(SSM)在建模线性复杂性和远程依赖关系方面的性能卓越,受其启发,笔者设计了一种基于CSC-Mamba(Cross-Shaped Convolutional Mamba Model)视觉模型遥感图像去雾技术。该技术基于SSM设计了RSMamba模块,利用其线性复杂性来实现全局上下文编码,大大降低了模型的复杂度。同时,利用卷积神经网络CNN以及基于自注意力机制设计CSwin模块来聚合不同方向域上的特征,以有效地感知雾分布的空间变化特征。通过这种方式,CSC-Mamba能够更好地提取雾特征,从而有效地去除雾对遥感图像的影响。通过在SateHaze1K公共数据集上的实验,结果表明本CSC-Mamba模型遥感图像去雾技术不仅具有较好的轻量化特征的同时性,还具有较高的去雾效果。
Remote sensing data captured by satellites is prone to image haziness due to the influence of suspended particles during the imaging process,which greatly affects the clarity of remote sensing images.To address this shortcoming,remote sensing image dehazing(RSID)is essential.Inspired by the excellent performance of the recently emerged State Space Model(SSM)in modeling linear complexity and long-range dependencies,this paper designs a remote sensing image dehazing technology based on the CSCMamba visual model.This technology designs the RSMamba module based on SSM,utilizing its linear complexity to achieve global context encoding,which significantly reduces the complexity of the model.Meanwhile,Convolutional Neural Networks(CNN)and the self-attention mechanism-based CSwin module are used to aggregate features on different directional domains,effectively perceiving the spatial variation characteristics of fog distribution.In this way,CSC-Mamba can better extract fog features,thus effectively removing the impact of fog on remote sensing images.Finally,experiments on the SateHaze1K public dataset show that the CSC-Mamba model for remote sensing image dehazing technology not only has good lightweight features but also exhibits high dehazing effectiveness.
作者
王京
何建军
易善信
张俸铖
肖辉
郭洋
杨伊凡
WANG Jing;HE Jianjun;YI Shanxin;ZHANG Fengcheng;XIAO Hui;GUO Yang;YANG Yifan(College of Computer and Network Security,Chengdu University of Technology,Chengdu 610059,China;Department of Computer Engineering,Taiyuan Institute of Technology,Taiyuan 030008,China)
出处
《物探化探计算技术》
2025年第6期867-875,共9页
Computing Techniques For Geophysical and Geochemical Exploration
基金
中海油田服务股份有限公司重点研发项目(G2317C-1412T564)。
