西平县高标准农田Zn含量高光谱反演研究  

Hyperspectral inversion model of Zn in high standard farmland soil in Xiping County

作　　者：蔡太义[1] 王志刚 杨刘帅 王群[3] 黄会娟[1] 于海洋[1] 张传忠 张灿 刘鹏 冯玉庆 贺成龙 张合兵[1] CAI Taiyi;WANG Zhigang;YANG Liushuai;WANG Qun;HUANG Huijuan;YU Haiyang;ZHANG Chuanzhong;ZHANG Can;LIU Peng;FENG Yuqing;HE Chenglong;ZHANG Hebing(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo 454000,China;The Fourth Topographic Survey Team of the Ministry of Natural Resources,Harbin 150000,China;College of Agronomy,Henan Agricultural University/Henan Province Agro-ecosystem Field Observation and Research Station,Zhengzhou 450046,China;Henan Province Soil Conditioning and Repair Engineering Technology Research Center,Shangqiu 476000,China;View Sino Orise Technology Co.,Ltd.,Wuxi 214400,China;Mineral Resources Exploration Center of Henan Geological Bureau,Zhengzhou 450053,China)

机构地区：[1]河南理工大学测绘与国土信息工程学院,河南焦作454000 [2]自然资源部第四地形测量队,哈尔滨150000 [3]河南农业大学农学院/河南省农田生态系统野外科学观测研究站,郑州450046 [4]河南省土壤调理与修复工程技术研究中心,河南商丘476000 [5]中向旭曜科技有限公司,江苏无锡214400 [6]河南省地质局矿产资源勘查中心,郑州450053

出　　处：《农业环境科学学报》2022年第10期2223-2231,共9页Journal of Agro-Environment Science

基　　金：国家自然科学基金项目(41671225);河南省重大科技专项(181100110400)。

摘　　要：为实现高标准农田土壤重金属Zn含量的快速测定,本文以西平县土壤Zn为研究对象,通过采集168个土壤样本进行室内实验,获得土壤高光谱数据(400~2 400 nm)并进行Savitzky-Golay平滑后,利用5种光谱变换,结合连续投影算法识别最佳特征波段,采用偏最小二乘回归方法构建Zn元素最佳反演模型。结果表明:二阶微分(SD)在1 409 nm波段的相关性(-0.502)最大,一阶微分(FD)在2 323 nm波段的相关性(0.491)最大,去包络线(CR)在2 439 nm波段的相关性(0.476)最大;倒数对数(LOG)、一阶微分、二阶微分、平滑曲线(SG)、去包络线的拟合度(R2)在0.65~0.70之间,相对分析误差(RPD)处于1.71~2.29之间,其中去包络线的拟合度(R~2=0.70、RPD=2.29)最高。经过5种光谱变换后的光谱反射率可有效突出光谱反射率的变化特征,并用于构建反演模型;土壤重金属Zn的最佳模型是以去包络线光谱变换为最佳偏最小二乘模型。To rapidly determine the heavy metal Zn content in high-standard farmland soil,we collected and analyzed Zn in soil collected in Xiping County.Through indoor experiments in which 168 soil samples were collected,soil hyperspectral data(400-2400 nm)were obtained and smoothed using the Savitzky-Golay method.Five types of spectral transformations and continuous projection algorithms were used to identify the best characteristic bands,and the partial least square regression method was used to construct an optimal inversion model of Zn.The correlation of the second-order differential(-0.502)was highest at the 1409 nm band;the correlation of the first-order differential(0.491)was largest in the 2323 nm band;and the correlation of the de-envelope(0.476)was the highest in the 2439 nm band.The fitting degree of a reciprocal logarithm,first-order differential,second-order differential,smooth curve,and de-envelope was 0.65-0.70,and the residual predictive deviation(RPD)was 1.71-2.29.The de-envelope showed the highest fitting degree(R2=0.70,RPD=2.29).The five types of spectral transformation can highlight variations in spectral reflectance and can be used to construct an inversion mode.The best model for soil heavy metal Zn is the de-enveloping spectral transformation,which is a partial least square model.

关 键 词：高标准农田 高光谱反演 偏最小二乘法 连续投影算法 ZN 

分 类 号：X87[环境科学与工程—环境工程] X833