光学遥感影像土地利用分类方法综述  

Review of Land Use Classification Methods Based on Optical Remote Sensing Images

作　　者：周珂[1,2] 杨永清[1,2] 张俨娜 苗茹 杨阳[1,2] 柳乐 ZHOU Ke;YANG Yong-qing;ZHANG Yan-na;MIAO Ru;YANG Yang;LIU Le(College of Computer and Information Engineering, Henan University, Kaifeng 475004, China;Henan Key Laboratory for Big Data Analysis and Processing, Henan University, Kaifeng 475004, China;Laboratory and Equipment Management Office of Henan University, Kaifeng 475004, China)

机构地区：[1]河南大学计算机与信息工程学院,开封475004 [2]河南大学河南省大数据分析与处理重点实验室,开封475004 [3]河南大学实验室与设备管理处,开封475004

出　　处：《科学技术与工程》2021年第32期13603-13613,共11页Science Technology and Engineering

基　　金：河南省科技攻关项目(202102210381);开封市重大科技专项(18ZD007);开封市科技攻关项目(2001001)。

摘　　要：土地是实现经济社会高质量发展的基础性资源,光学遥感影像以时间、空间、光谱等特有的优势为土地利用变化的发现、识别、分析提供了很好的支持,利用光学遥感技术提取土地利用变化信息已经成为重要的研究领域。以近年来土地利用分类相关文献为基础,分别从数据源、处理流程、处理精度、计算效率等角度对土地利用分类应用较为广泛的目视解译、监督分类、非监督分类、深度学习等四类方法进行了分析和总结。最后展望了光学遥感影像土地利用分类方法的发展方向。Land is used to realize the economic and social development of high quality basic resources.Such as time,space,spectrum is used as optical remote sensing image unique advantages for the discovery of land use change,the identification,analysis provides a good support,optical remote sensing technology is utilized to extract information of land use change has become an important area of research.Based on the literature on land use classification in recent years,four widely used land use classification methods,namely visual interpretation,supervised classification,unsupervised classification and deep learning were analyzed and summarized from the perspectives of data sources,processing process,processing accuracy and computational efficiency.Finally,the development direction of land use classification methods for optical remote sensing images was prospected.

关 键 词：土地利用分类 光学遥感影像 监督分类 非监督分类 深度学习 

分 类 号：P237[天文地球—摄影测量与遥感]