光谱学与光谱分析, 2020, 40 (8): 2571, 网络出版: 2020-12-03  

基于高光谱的草坪草叶绿素含量模拟估算

Hyperspectral-Based Estimation on the Chlorophyll Content of Turfgrass
作者单位
1 甘肃农业大学草业学院, 甘肃 兰州 730070
2 草业生态系统教育部重点实验室(甘肃农业大学), 甘肃 兰州 730070
3 中国人民武装警察部队工程大学, 陕西 西安 710086
4 西安石油大学计算机学院, 陕西 西安 710065
5 甘肃省草原技术推广总站, 甘肃 兰州 730000
6 西安航空学院, 陕西 西安 710077
摘要
草坪色泽是草坪观赏价值的最直接体现。 探索基于高光谱的草坪草叶绿素含量的估算和反演对草坪质量评定具有重要意义。 以3种常用草坪草种——“红象”高羊茅(Festuca arundinacea cv. Hongxiang)、 “百灵鸟”多年生黑麦草(Lolium perenne cv. Bailingniao)和“肯塔基”草地早熟禾(Poa pratensis cv. Kentucky)为试样, 通过盆栽实验, 在草坪草生长旺盛期, 使用SOC710VP成像光谱仪和TYS-A3500叶绿素仪分别测定了草坪草冠层光谱数据和叶绿素相对含量(SPAD), 并通过Person相关系数分析了原始SPAD, 1/SPAD和log(1/SPAD)与10个植被指数: GI(绿度植被指数)、 ARVI(大气阻抗植被指数)、 VARI(可视化气压阻抗指数)、 NDVI705(归一化植被指数705)、 MSR705(改进红边比值植被指数)、 NDVI670(归一化植被指数670)、 CI(叶绿素指数)、 PSRI(植被衰减指数)、 RGI(相对绿色指数)和EVI(增强植被指数)的相关性, 筛选与叶绿素相关性较高的高光谱波段植被指数, 构建植被指数反演叶绿素含量模型, 最后通过精度检验, 筛选最优草坪草叶绿素估算模型。 研究结果如下: (1)不同草坪草光谱曲线整体趋势相差不大, 但不同种间反射率(REF)还是有所区别。 在730~1 000 nm波段, “百灵鸟”多年生黑麦草与“红象”高羊茅REF差异不大, 但“肯塔基”草地早熟禾REF较高, 光谱特征更为明显; (2)10个植被指数中, VARI, RGI和PSRI与草坪草3个叶绿素指标极显著相关, 相关系数R2绝对值均大于0.65, 可作为首选植被指数进行草坪草叶绿素含量估算; (3)植被指数与叶绿素指标逐步回归分析发现, 单因素回归模型中, 利用VARI, RGI和PSRI估算1/SPAD的模型决定系数R2均在0.6以上, 普遍高于SPAD与log(1/SPAD)的估算模型; 而多元线性回归中, 10个植被指数中, RGI与叶绿素指标1/SPAD所构建的模型决定系数R2同样最高, 为0.817, 说明SPAD倒数形式适用于草坪草叶绿素反演; (4)选择决定系数较高(>0.7)的模型进行精度检验, 筛选的最优的草坪草叶绿素指标反演模型为: y1/SPAD=0.161xRGI+0.007xGI-0.054(R2=0.817, RMSE=0.023)。
Abstract
Lawn color is the most obvious indicator of ornamental lawn value. It is of great significance to explore the relationship between chlorophyll content of turfgrass and the hyperspectral reflectance. This relationship can be used to develop models to calculate the chlorophyll content for lawn quality evaluation purpose. In this study, three common lawn grass species—Festuca arundinacea CV. Hongxiang, Lolium perenne CV. Bailingniao and Poa pratensis CV. Kentucky was cultivated in pots. Measurements of chlorophyll content and hyperspectral reflectance were made during active growth period by tys-a3500 chlorophyll meter and SOC710VP imaging spectrometer to determine the relative chlorophyll content (SPAD) and spectral data of turf grass canopy, respectively. Person correlation analysis for each of the SPAD, 1/SPAD and log(1/SPAD) was conducted with a group of variables including vegetation index—10 G (green vegetation index), ARVI (atmospheric impedance difference vegetation index), VARI (visual pressure impedance index), NDVI705 normalized difference vegetation index (705), MSR705 red edge ratio vegetation index (improved), NDVI670 normalized difference vegetation index (670), CI (chlorophyll index), PSRI attenuation (vegetation index), RGI (relatively green index) and EVI ( Enhance the correlation of vegetation index). After screening the hyperspectral bands of vegetation index with the highest correlation with chlorophyll content, models were developed using the vegetation index based on these bands. After the best model was selecting through an accuracy test, the model was used to estimate the chlorophyll SPAD values change for turf grasses under different concentrations of heavy metals Pb2+ stress. The results are summarized as follows: (1) the overall trends of spectral curves of different turfgrass were not significantly different, but the reflectance (REF) of different species were different. At the band of 730~1 000 nm, there was no significant difference between “lark” perennial ryegrass and “red elephant” tallfestia REF, but the spectral characteristics of “Kentucky bluegrass” were unique with a higher REF. (2) among the 10 vegetation indexes, VARI, RGI and PSRI were extremely significantly correlated with 3 chlorophyll indexes of turfgrass, and the absolute value of correlation coefficient R2 was all greater than 0.65, indicating that it is feasible to estimate the chlorophyll content of turfgrass with these 3 vegetation indexes. (3) Stepwise regression analysis of vegetation index and chlorophyll index shows that in the single-factor regression model, the model determination coefficient (R2) of estimating 1/SPAD using vegetation index VARI, RGI and PSRI was above 0.626, which was generally higher than the estimation of SPAD and log(1/SPAD). In multiple linear regression, the model determination coefficient (R2) constructed by 10 vegetation indexes and chlorophyll index 1/SPAD was also the highest (0.817), showing that SPAD reciprocal form is applicable to be used the in model estimation of chlorophyll in turfgrass. (4) The best model selected from the models with a high determination coefficient (>0.7) through accuracy test was y1/SPAD=0.161xRGI+0.007xGI-0.054 (R2=0.817, RMSE=0.023).

纪童, 王波, 杨军银, 柳小妮, 王洪伟, 王彩玲, 潘冬荣, 徐君. 基于高光谱的草坪草叶绿素含量模拟估算[J]. 光谱学与光谱分析, 2020, 40(8): 2571. JI Tong, WANG Bo, YANG Jun-yin, LIU Xiao-ni, WANG Hong-wei, WANG Cai-ling, PAN Dong-rong, XU Jun. Hyperspectral-Based Estimation on the Chlorophyll Content of Turfgrass[J]. Spectroscopy and Spectral Analysis, 2020, 40(8): 2571.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!