摘要
传统的遥感数据数字化方法效率低、容易出错,已经越来越不适应快速增长的地理信息数据需求。为了解决上述问题,提出一种基于深度学习的卫星影像自动数字化技术。利用Mask R-CNN算法技术对遥感图像的对象进行识别,对经过数据处理后的卫星影像进行训练,并提取其几何特征。能够对影像的重要对象特征进行识别,全部图像分类的结果准确率总体达到80%以上。其中识别率最高的是道路、水系、绿地,准确率可达到85%,可以作为卫星影像自动数字化工具。
Traditional methods of digitising remote sensing data arc inefficient and error-prone and have become increasingly unsuited to the rapidly growing demand for geographic information data.In order to solve the above problems*an automatic digitisation technique for satellite images based on deep learning is proposed.The Mask R-CNN algorithm technique is used to identify the objects of remote sensing images*train the satellite images after data processing*and extract their geometric features.It is able to recognise the important object features of the images*with an overall accuracy rate of over 80%for all image classification results.The highest recognition rates arc for roads*water systems*and green spaces*with an accuracy rate of 85%*which can be used as an automatic digitisation tool for satellite images.The highest recognition rates arc for roads*water systems*and green spaces*with an accuracy rate of 85%*which can be used as an automatic digitisation tool for satellite images.
作者
梁碧仪
陈颖
黄梓煌
聂佩林
LIANG Bi-yi;CHEN Ying;HUANG Zi-huang;NIE Pei-lin(School of Geography,South China Normal University,Guangzhou Guangdong 510631,China;School of Environmental and Chemical Engineering,Foshan University,Foshan Guangdong 528000,China)
出处
《现代测绘》
2023年第6期39-43,共5页
Modern Surveying and Mapping
基金
广东省自然科学基金项目(2014A030313617)
