摘要
针对国土调查实地举证中,照片数量巨大、内容丰富、语义复杂,依靠人工按照土地利用分类标准解译费时费力且主观性过强的问题,该文提出一种基于注意力的照片土地利用场景语义深度学习解析方法,自动提取照片注意力区(主场景)的土地利用语义和分布信息。实验表明,该方法能够实现自然场景照片土地利用语义包的快速提取,成功应用于江夏区国土调查工作,为全国国土调查智能化提供了技术支撑。
In the task of land survey,field photos are large in number,rich in layers,and complex in semantics,which are time-consuming and subjective to be interpreted manually according to land use classification standards.This paper proposes an attention-based deep learning semantics parsing method,which automatically extracts the land use semantics and distribution information of photo attention areas(main scenes)for the photo’s land use scenes.The experiments show that this method can realize the rapid extraction the land use semantic package from natural scene photos,and is successfully applied to land survey in Jiangxia District,providing technical support for the proof of intelligent national land survey.
作者
徐世武
李亭谕
白晓飞
吴瀚
XU Shiwu;LI Tingyu;BAI Xiaofei;WU Han(School of Geography and Information Engineering,China University of Geosciences(Wuhan),Wuhan 430074,China;China Land Surveying and Planning Institute,Beijing 100035,China;Suzhou Surveying and Mapping Institute Co.,Ltd..Suzhou,Jiangsu 215000,China)
出处
《测绘科学》
CSCD
北大核心
2021年第12期210-218,共9页
Science of Surveying and Mapping
关键词
显著式注意力
显著性语义解析
深度估计网络
语义分割
深度学习
salient attention
salient semantic parsing
depth estimation network
semantic segmentation
deep learning
