基于物候差异指数的栽培甘草遥感识别方法  

Remote sensing method for planted licorice identification based on phenological difference index

作　　者：翟富祥 史婷婷[4] 张小波[4] 蒙继华 张磊 ZHAI Fuxiang;SHI Tingting;ZHANG Xiaobo;MENG Jihua;ZHANG Lei(Key Laboratory of Digital Earth Science,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100101,China;International Research Center of Big Data for Sustainable Development Goals,Beijing 100094,China;University of Chinese Academy of Sciences,Beijing 100049,China;National Key Laboratory Breeding Base of Dao-di Herbs,National Resource Center for Chinese Materia Medica,China Academy of Chinese Medical Sciences,Beijing 100700,China)

机构地区：[1]中国科学院空天信息创新研究院数字地球重点实验室,北京100101 [2]可持续发展大数据国际研究中心,北京100094 [3]中国科学院大学,北京100049 [4]中国中医科学院中药资源中心,道地药材国家重点实验室培育基地,北京100700

出　　处：《中华中医药杂志》2023年第12期5729-5732,共4页China Journal of Traditional Chinese Medicine and Pharmacy

基　　金：第三次新疆综合科学考察项目(No.2021xjkk1201);国家重点研发计划项目(No.2020YFC1511602);国家自然科学基金项目(No.82003901);中国中医科学院科技创新工程(No.CI2021A03902)。

摘　　要：目的:及时准确地掌握栽培甘草种植面积及空间分布等信息,能够为甘草产业健康发展和精细化管理提供科学的数据支撑。方法:文章基于Sentinel-2遥感影像,以新疆库尔勒甘草主产区为研究区,根据甘草光谱的物候特征,提出一种归一化差异多年生植被指数(NDPVI),剔除具有收获性的作物对甘草分类的影响;利用归一化差异枯萎指数(NDVWI),减弱田间果树类木本植被对甘草的干扰,并通过决策树方法实现甘草的准确识别。结果:甘草的NDPVI在0.05~0.15,低于其他作物类NDPVI(0.35~0.46),而甘草的NDVWI在0.04~0.14,同样低于果树类NDVWI(0.18~0.23),NDPVI、NDVWI指数能有效地区分甘草与作物、果树,甘草识别总体精度达到83.85%。结论:研究结果证明了本方法在我国西北地区甘草主产区进行栽培甘草遥感识别的有效性。Objective:Timely and accurate grasp of licorice planting area and spatial distribution can provide scientific data support for the healthy development and delicacy management of the licorice industry.Method:Based on Sentinel-2 data in Korla,Xinjiang,we explored the phenological characteristics of licorice spectrum and proposed a normalized difference perennial vegetation index(NDPHI),which excluded harvestable crops from licorice;and we used normalized difference vegetation withered index(NDVWI)to reduce the classification interference of fruit tree-like woody plants on licorice,and finally obtained the result by decision tree algorithm.Results:It indicated that the NDPVI of licorice was between 0.05-0.15,which was lower than the NDPVI of crops(0.35-0.46),while the NDVWI of licorice was between 0.04-0.14,which was also lower than fruit trees(0.18-0.23).The proposed method can effectively distinguish licorice from crops and fruit trees.Overall accuracy reached 83.85%.Conclusion:The results prove this method's effectiveness on remote sensing method for planted licorice identification.

关 键 词：遥感 甘草 物候 植被指数 Sentinel-2 

分 类 号：S567.71[农业科学—中草药栽培] S127[农业科学—作物学]