基于生成对抗网络的建筑遥感影像细节增强技术研究

打开文本图片集

中图分类号：P237文献标识码：A

文章编号:1672-3791(2026)04-0031-03

Research on Detail Enhancement Technology of Building Remote Sensing Images Based on Generative Adversarial Network Deep Learning

SUN Yanhua LIANG Huaixiang

Geophysical prospectingand surveying team of Shandong Bureau of Coal Geology, Jinan,

Shandong Province,25010o China

Abstract: In the processof obtaining remote sensing images of urban buildings,external noise interference can affect the qualityand structural informationofthe images,resulting in a decreasein the structural similarity between thegenerated enhanced images and theoriginal images.Therefore,a research ondetail enhancement technology for urban building remote sensing images based on generative adversarial network deep learning algorithm is proposed. It builds an image detail enhancement model for generative adversarial networks,splits the input urban building remote sensing images,merges features and calibrates channel attention weights,applies multi-scale weights to feature maps,and outputs detail enhanced remote sensing images.It introduces multiple loss functions to optimize and enhance the efect.It calculates the eror between the output results and the remote sensing images of urban buildings, and uses non local mean algorithm to denoise the images.It introduces Laplacian operator to sharpen and proces images,achieves detail enhancement of urban building remote sensing images.The experimental results show that the proposed method results inclear texture of urbanbuilding remote sensing images,withSSIMvalues close to1 on various urban building scene images.The image with enhanced details is more similar to the real image.

Keywords: Generative adversarial network; Deep learning; Remote sensing images of urban buildings; Detail en - hancement; Multiple loss function

遥感技术作为一种观测手段，从空中获取地表大范围影像，能够提供城市建筑物的详细数据[1],这些数据包括城市建筑物的位置、形态、面积等多方面信息。（剩余4065字）