摘要
为实现高精度、快速、高效提取长时间序列的棉花种植面积及其分布,文章基于遥感大数据计算服务(PIE Engine Studio)平台,以Landsat8 C2 SR、Landsat5 C2 SR及Sentinel-2 MSI的遥感影像作为数据源,采用中值合成法计算NDVI与EVI指数构建特征数据,并载入随机森林算法对新疆石河子市研究区近10年的棉花种植面积进行动态监测与分析,并将部分结果与GEE平台计算结果和统计数据做了对比。结果表明:1)研究区基于PIE平台的棉花种植面积提取精度良好、分类结果较理想,总体精度优于GEE平台,其中Kappa系数和总体精度OA最高可达到0.963和97.1%,与统计数据相比,精度较高;2)研究区2012–2021年棉花种植面积总体稳定,并有上升的趋势;3)研究区近10年棉花种植区域空间变化明显,种植分布变化以石河子市的西部、南部以及北部地区为主。上述结果进一步验证了PIE国产云计算平台的实用性与优越性,可为石河子市的棉花生产提供辅助信息,助力棉花产业的发展。
In order to achieve high-precision,fast and efficient extraction of long time series of cotton planting distribution and its area,the article uses remote sensing images of Landsat8 C2 SR,Landsat5 C2 SR and Sentinel-2 MSI as data sources based on the platform of remote sensing big data computing service(PIE Engine Studio),and uses the median synthesis method to calculate NDVI and EVI indices and to construct feature data,and the random forest algorithm was loaded to monitor and analyze the cotton cultivation area of the study area in Shihezi,Xinjiang for the past 10 years.The results show that:1)the cotton planted area in the study area based on PIE platform has good extraction accuracy and better classification results,and the overall accuracy is better than that of GEE platform,in which the highest Kappa coefficient and overall accuracy OA can reach 0.963 and 97.1%,and the accuracy is higher compared with the statistical data;2)the cotton planted area in the study area from 2012 to 2021 is generally stable and has an increasing trend;3)the spatial changes of cotton planting area in the study area in the past 10 years were obvious,and the planting distribution was dominated by changes in the western and southern as well as northern areas of Shihezi city.The above results further verify the practicality and superiority of PIE's domestic cloud computing platform,which can provide auxiliary information for cotton production in Shihezi and help the development of cotton industry.
作者
周磊
林志树
玉林海
窦世卿
ZHOU Lei;LIN Zhishu;YU Linhai;DOU Shiqing(College of Geomatics and Geoinformation of Guilin University of Technology,Guilin 541000,China)
出处
《航天返回与遥感》
CSCD
北大核心
2023年第3期108-118,共11页
Spacecraft Recovery & Remote Sensing
基金
国家自然科学基金(42061059)
广西八桂学者专项项目(DT2100001072)
桂林市科技局开发项目(2020010701)。
关键词
地球科学引擎
随机森林算法
棉花种植面积
遥感应用
PIE-Engine
random forest algorithm
cotton planting area
remote sensing application
