| 武倩雯,熊黑钢,王凯龙,王莉峰,靳彦华.干旱区玉米抽雄期叶绿素含量高光谱最佳模型选择[J].干旱地区农业研究,2015,33(2):81~86 |
| 干旱区玉米抽雄期叶绿素含量高光谱最佳模型选择 |
| Selection of optimal model using hyperspectral parameters for chlorophyll content of maize during tasseling stage in arid region |
| |
| DOI：10.16302/j.cnki.1000-7601.2015.02.013 |
| 中文关键词: 叶绿素含量 高光谱参数 抽雄期 |
| 英文关键词:chlorophyll content spectral parameter maize tasselling |
| 基金项目:国家自然科学基金(41171165); 北京市属高等学校高层次人才引进与培养计划项目（IDHT20130322）；北京联合大学人才强校计划资助项目（BPHR2012E01） |
|
| 摘要点击次数: 1677 |
| 全文下载次数: 1084 |
| 中文摘要: |
| 采用相关性、线性和非线性分析法，探讨了玉米抽雄期叶片叶绿素含量与多种高光谱参数之间的关系，并建立了叶绿素含量的定量监测模型。结果表明：(1) 原始光谱反射率与叶绿素含量在713 nm处具有最大相关系数r=0.86，光谱反射率一阶微分在760 nm处与叶绿素含量具有最大相关性r=0.84。同时，最大一阶微分分别对应的波长（λr，λb，λy）、绿峰反射率（Rg）和其对应的波长λg、红边内最大一阶微分总和（SDr）、比值植被指数（SDr/SDb，SDr/SDy，(Rg-Ro)/(Rg+Ro)）以及归一化植被指数(SDr-SDb)/(SDr+SDb)等10种参数分别与叶绿素含量的相关性达到极显著相关。(2) 采用相关性达到极其显著的12种光谱参数进行建模，其中原始光谱、绿色反射峰以及光谱反射率一阶微分、基于红边面积与蓝边面积的比值植被指数和归一化植被指数所建立的10个模型R2都不小于0.72，前两者所建立的指数模型优于线性模型，而后三者所建立的线性模型则优于指数模型。(3) 所选取的五个方程中，在760 nm处的光谱反射率一阶微分值所构建的线性模型：y叶绿素=6912x760nm+44.878因其具有最大决定系数和最小的RMSE，并且其模型表达式相对简单，因此是玉米抽雄期叶绿素含量的最佳预测模型，从模型决定系数R2来看，它比其他模型至少提高了11.4%。 |
| 英文摘要: |
| A model monitoring the quantitative content of chlorophyll was established using correlation and linear and nonlinear analyses to determine the relationships between chlorophyll content of maize at the heading stage and a variety of high spectral parameters. The results showed that the raw spectral reflectance had a maximal negative correlation coefficient at 713 nm(r=0.86) with the chlorophyll content. The first derivative spectral reflectance had a maximal positive correlation coefficient at 760 nm (r=0.84). The parameters included λr, λb, λy, λg, Rg, SDr, SDr/SDb, SDr/SDy, (Rg-Ro)/(Rg+Ro) and (SDr-SDb)/(SDr+SDb), all of which reached significant correlations with the chlorophyll content. Models were built based on 12 kinds of spectral parameters that showed significant correlations. The R2 is of the models were greater than 0.72, constructed by the raw spectral reflectance, green reflection peak, the spectral reflectance of the first derivative, vegetation index and normalized difference vegetation index based on the red edge and blue edge area ratio. The first two exponential models established were better than the linear model, whereas the latter three linear models were better than the exponential models. Through precision evaluations of the estimation models, the model yChlorophyll content=6912x760nm+44.878 constructed in the first derivative spectral reflectance at 760 nm. This model seemed to be the best prediction for the chlorophyll content of maize at the heading stage due to its maximal correlation coefficient and minimal RMSE, and its relatively simple expression. The correlation coefficient R2 of the model was increased at least 11.4% more than other models. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| | |