摘要
利用35景卫星影像,探寻桃树遥感辨识的最佳时相与方法。首先对各景影像分别进行预处理,随后利用6类探试性辨识方法(即地物反射光谱比较、波段差分或比值分析、光谱指数求算与分析、光谱指数变化追踪、影像复合与辨识方法协同分析)对探试组影像进行辨识分析,并从中选出3种较佳的辨识方法,最后利用优选的3种辨识方法对验证组影像进行辨识验证。结果表明:(1)在10月初的影像中,桃树具有较高的NDVI×ρNIR值,利用其阈值可以较高精度识别桃树(桃树分类正确率可达94.8%,总体分类精度可达91.33%);(2)在4月初(桃树盛花期)的影像中,利用NDVI、1/ρGREEN-1/ρRED与ρBLUE+ρGREEN+ρRED的三重阈值也可以较高精度识别桃树(桃树分类正确率可达85??2%,总体分类精度可达80.13%);(3)在上述两期数据融合、所用辨识方法协同的情形下,可进一步提高桃树的辨识精度(桃树分类正确率可提高到96.53%,总体分类精度可提高到93.37%);(4)桃树遥感辨识的最佳时相为10月初,较佳时相为4月初。
Using 35 satellite remote sensing images the optimal phase and method for remote sensing identification of peach trees were explored. Firstly, each scene image was preprocessed separately. Secondly, the three better identification methods were selected from the six kinds of tentative identification methods (comparison of reflectance spectra of ground objects, bands difference or ratio analysis, spectral indices calculation and analysis, spectral index change tracing, image fusion and identification method collaborative analysis) which were used to identify and analyze the images of test group. Finally, the images of verification group were verified by using the three better identification methods. The results showed the peach trees had higher values of NDVI×ρ NIR in the early October, and the peach trees could be identified with higher accuracy by using threshold of the vegetation index (the correct rate of identifying peach tree species could reach 94.8%, the overall classification accuracy could reach 91.33%). In the early April (peach trees were in full bloom), the peach trees were also identified with higher accuracy by using the triple thresholds of NDVI、1/ρ GREEN -1/ρ RED and ρ BLUE +ρ GREEN +ρ RED (the correct rate of identifying peach tree species could reach 85.2%, the overall classification accuracy could reach 80.13%). The identification precision of peach trees could be further improved with the fusion of the images in the two periods and the collaboration of two identification methods (the correct rate of identifying peach tree species was increased to 96.53%, the overall classification accuracy was increased to 93.37%). The better phase for identifying the peach trees is early April, and the best phase for identifing peach trees is early October.
作者
邢东兴
王明军
陈玲侠
杨波
焦俏
张亚宁
XING Dong-xing;WANG Ming-jun;CHEN Ling-xia;YANG Bo;JIAO Qiao;ZHANG Ya-ning(College of Resources and Environment, Xianyang Normal University, Xianyang 712000, China;College of Physics and Electronic Engineering, Xianyang Normal University, Xianyang 712000, China)
出处
《江苏农业学报》
CSCD
北大核心
2019年第4期919-926,共8页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金项目(61771385)
陕西省优势学科建设项目(060103)
关键词
桃树
遥感影像
树种辨识
最佳时相
光谱指数
peach tree
remote sensing image
identification of fruit tree species
the best phase
spectral index
