黄河滩区耕地和作物遥感提取研究——以兰考段为例  

Study on Remote Sensing Extraction of Cultivated Land and Crops:Taking Lankao Section of the Yellow River Beach Area as an Example

作　　者：闰亚迪 秦耀辰[1] 樊雷[2,3,4] 刘小燕 史志方 李梦迪 李乾 崔耀平 RUN Yadi;QIN Yaochen;FAN Lei;LIU Xiaoyan;SHI Zhifang;LI Mengdi;LI Qian;CUI Yaoping(Key Laboratory of Geospatial Technology for the Middle and Lower Yelloxv River Regions,Ministry of Education/College of Geography and Environmental Science,Henan Universityt Henan Kaifeng 475004,China;Land Consolidation and Rehabilitation Center o f Henan Province,Zhengzhou 450016,China;Henan Science and Technology Innovation Center of Natural Resources(Ecological Product Value Accounting),Zhengzhou 450016,China;Henan Scientific Research Station for Land Consolidation and Ecological Restoration in Central Plains,Zhengzhou 450016,China)

机构地区：[1]河南大学黄河中下游数字地理技术教育部重点实验室/地理与环境学院,河南开封475004 [2]河南省土地整理中心,郑州450016 [3]河南省自然资源科技创新中心(生态产品价值核算研究),郑州450016 [4]中原地区国土整治与生态修复河南科研工作站,郑州450016

出　　处：《河南大学学报（自然科学版）》2022年第1期9-19,共11页Journal of Henan University:Natural Science

基　　金：中国科学院陆地表层格局与模拟院重点实验室2020年度开放基金(LB2021006);国家自然科学基金资助项目(42071415,41671425);河南省自然科学优秀青年科学基金资助项目(202300410049);国家重点研发国际合作项目(2021YFE0106700)。

摘　　要：黄河滩区耕地资源是黄河流域国土资源管控的重要领域,滩区可利用耕地资源及其对应的作物呈现出小斑块和破碎化特征,采用中高分影像和前沿算法来提炼这种特征以了解其分布和对应的农作物耕种类型,这对黄河滩区的行洪安全和国土综合整治均具有重要意义.利用水体与植被指数之间的关系识别地表水体,通过UNet深度学习网络,结合地物多种特征规则,对黄河滩区大宗农作物提取.结果表明:(1)经过实地采样、原始影像和三调数据对比,水体的提取总体精度达到91%,冬小麦提取精度达88%,夏玉米的精度达85%;(2)耕地面积在黄河滩区兰考段面积中占比较大,占滩区总面积的66.62%,冬小麦占耕地总面积的58.31%,夏玉米占耕地总面积的46.03%;(3)利用UNet深度学习网络的方法提取农作物的精度较高,可以用于冬小麦和夏玉米的提取.本研究同时探讨了持续滩涂地和耕地之间的转化,可为滩区土地利用及发展规划研究提供数据支撑.The cultivated land resources in the beach area of the Yellow River is an important field of land and resources control in the Yellow River Basin.However,the wandering Yellow River makes the available arable land resources and their corresponding crops appear with small patches and fragmentation characteristics.It is of great significance to investigate its distribution and corresponding crop types using the high-resolution images and cutting-edge algorithms to extract these characteristics for flood discharge safety and comprehensive land improvement in the Yellow River Basin.In this study,the relationship between water body and vegetation index was used to identify surface water body.We combined the deep learning method with multiple feature rules of ground objects to extract the bulk crops in the study area.The results showed that:(1)Comparing the field sampling,original image,and three surveys data,the extraction accuracy of water body,winter wheat,and summer corn were 91%,88%,and 85%,respectively.(2)The area of cultivated land in the Lankao section of the Yellow River beach area accounted for 66.62%of the total area of the tidal area.The winter wheat and summer corn accounted for 58.31%and 46.03%of the total area of cultivated land,respectively.(3)Our study proved that the UNet deep learning method has high accuracy in crop extraction and can be used for the extraction of winter wheat and summer corn.This study also discussed the continuous conversion between beach land and cultivated land,which can provide data support for the study of land use and development planning in the beach area.

关 键 词：深度学习 GEE 水土资源 大数据 云计算 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]