基于GEE平台的水稻种植范围识别研究--以黑龙江省为例  

Identification of Rice Planting Areas Based on GEE Platform:A Case Study of Heilongjiang Province

作　　者：潘建平[1] 安新永 崔伟 尚栋 谢鹏[1] PAN Jianping;AN Xinyong;CUI Wei;SHANG Dong;XIE Peng(School of Smart Cities,Chongqing Jiaotong University,Chongqing 400074,China)

机构地区：[1]重庆交通大学智慧城市学院,重庆400074

出　　处：《测绘地理信息》2024年第3期101-106,共6页Journal of Geomatics

基　　金：宁夏自治区重点研发计划(2022CMG02014)。

摘　　要：精准识别与提取水稻种植范围对于粮食估产和耕地有效利用具有重要价值。以往的研究针对于单一的关键物候期进行水稻的识别,导致水稻与某些地物在单一时相中光谱可区分性较差。以黑龙江省为研究区,在谷歌地球引擎(google earth engine,GEE)云平台的支持下,利用哨兵二号(sentinel-2 MSI)数据,使用单类支持向量机(one-class sup⁃port vector machine,OCSVM)对水稻进行了种植范围遥感识别提取,设计了一种基于像元的融合物候特征识别算法,首先对于研究区内的水稻进行时序性分析,通过建立4个关键物候期;然后对水稻和研究区内的其他地物进行物候特征的择优选择;最后将所选的物候特征融合进行水稻种植范围提取。结果表明,该方法的结果精度与实地调查数据相比总体精度(overall accuracy,OA)为0.969,Kappa为0.93。Accurate identification and extraction of rice plant⁃ing areas are of great value for grain yield estimation and effec⁃tive use of cultivated land.Previous studies focused on the identification of rice in a single key phenological phase,result⁃ing in poor spectral differentiation between rice and some land features in a single phase.Based on the sentinel-2 MSI data,this paper takes Heilongjiang Province as the research area,and uses the cloud platform of Google Earth Engine(GEE)to support the sentinel-2 MSI data.One-class Support Vec⁃tor Machine(OCSVM)is used for remote sensing recogni⁃tion and extraction of rice planting range.In this paper,a fu⁃sion phenological feature recognition algorithm based on pix⁃els is designed.Firstly,the timing analysis of rice in the study area is carried out.Then,the phenological features of rice and other land features in the study area are selected.Finally,the selected phenological features are fused to extract rice planting range.The results show that the accuracy of OA and Kappa are 0.969 and 0.93 respectively compared with the field sur⁃vey data.

关 键 词：水稻 物候特征 哨兵二号 单类支持向量机 谷歌地球引擎 

分 类 号：P237[天文地球—摄影测量与遥感]