基于多时相遥感数据聚类决策树方法的白水县农作物识别  

Identification for Crops in Baishui County by Clustering Decision Tree Method Based on Multi-temporal Remote Sensing Data

作　　者：刘状 李纲 石晨烈[3] 魏永强 王建荣 杨秀策 LIU Zhuang;LI Gang;SHI Chen-lie;WEI Yong-qiang;WANG Jian-rong;YANG Xiu-ce(State Key Laboratory of Geo-information Engineering,Xi’an 710054,China;Xi’an Research Institute of Surveying and Mapping,Xi’an 710054,China;College of Urban and Environmental Sciences/Northwest University,Xi'an 710127,China)

机构地区：[1]地理信息工程国家重点实验室,陕西西安710054 [2]西安测绘研究所,陕西西安710054 [3]西北大学城市与环境学院,陕西西安710127

出　　处：《山东农业大学学报（自然科学版）》2022年第5期728-734,共7页Journal of Shandong Agricultural University：Natural Science Edition

基　　金：地理信息工程国家重点实验室开放基金(SKLGIE2021-zz-2);航天系统部综合研究项目(TJ20203C040257)。

摘　　要：利用Sentinel-2、Landsat8OLI、SRTM等多源遥感数据,根据白水县物候特征经验构建不同典型NDVI时间序列数据集进行聚类分析,同时结合影像分割和决策树支持方法首次对陕西省白水县的多种农作物进行提取分类。结果表明:1)本文提出的方法能够有效提取典型地物信息,总体分类精度达到91.22%,Kappa系数为0.8949,相比基于单一时相的光谱特征的分类有着较大优势。2)白水县的三种主要农作物苹果、玉米和小麦的物候特征显著,分步提取的地物类别极少发生类间像素重叠,且表现的物候规律与白水县农作物的实际物候规律高度吻合,侧面证明了分类结果的适用性。3)白水县苹果种植分布西多东少,中北部多南部少;玉米主要分布于白水县西部和北部较高海拔地区;小麦主要分布于白水县南部;白水县的苹果种植仍有推广空间。这些结果发展了聚类分析结合决策树分类在农作物遥感方面的用法以及为白水县的农业经济发展提供有益参考。Using Sentinel-2,landsat8OLI,SRTM and other multi-source remote sensing data,according to the experience of Baishui County’s phenological characteristics,and different typical NDVI time series datasets were constructed for cluster analysis,used image segmentation and decision tree support methods classify various land objects in BaiShui County,Shanxi Province for the first time.The results show that:1)The method proposed in this paper can effectively extract the information of typical land objects,the overall accuracy of mapping of phenological characteristics was 91.22% and the Kappa increased was 0.8949,compared with the classification of the spectral features based on a single phase,it has great advantages.2)The three main crops in Bai Shui County have significant phenological characteristics,the land object categories extracted by the step-by-step extraction rarely have pixel overlap between categories,and the phenological law of the performance is highly consistent with the actual phenological law of crops in Baishui County,which proves the applicability of the classification results.3)The distribution of apple planting in Baishui County is more in the west and less in the east,and more in the north and south than in the south;maize is mainly distributed in the west and north of Baishui County at higher altitudes;wheat is mainly distributed in the south of Baishui County;there is still room for promotion of apple planting in Baishui County.These results develop the use of cluster analysis combined with decision tree classification in remote sensing of agricultural crops as well as provide useful references for the economic development of Baishui.

关 键 词：遥感 聚类分析 农作物识别 

分 类 号：S127[农业科学—农业基础科学]