基于光谱红边位置提取算法的番茄叶片叶绿素含量估测

Estimation of Chlorophyll Content of Tomato Leaf Using Spectrum Red Edge Position Extraction Algorithm

为了快速、准确估测番茄叶片叶绿素含量,分析了不同营养水平下的番茄叶片光谱红边参数变化规律,发现红边位置最能表征番茄叶绿素状况,统计分析了6种算法提取的光谱红边位置的差异性,并为每种算法分别建立了5种估测模型,验证结果表明每种红边位置提取算法所对应的最佳模型为线性四点内插法的指数曲线模型和其他红边位置算法的对数曲线模型。其中线性外推法模型精度最高,校正集决定系数R2c为0.618 6,验证集决定系数R2v达到0.771 1,验证集均方根误差RMSEv为8.359 6,可以有效诊断番茄叶绿素含量。线性四点内插法根据670、700、740、780 nm 4个波段的叶片反射率计算红边位置,运算简单,模型精度较高,R2c为0.621 7,R2v达到0.766 6,RMSEv为8.568 2,可以作为开发番茄叶绿素含量监测仪器的依据。

The red edge parameters of plants spectrum were used to estimate foliar chlorophyll for nitrogen content and leaf area. Among these parameters,the red edge position（ REP） is the best one for diagnosing the growth state of tomato according to statistical analysis. The REP was defined by the wavelength of the maximum first derivative of the reflectance spectrum in the region（ 660 nm to 780 nm）of the red edge. The six algorithms could be used to extract the REP,including four-point interpolation,maximum first derivative,inverted Gaussian fitting,Lagrangian,linear extrapolation,and polynomial fitting. In order to achieve a rapid and accurate application for predicting the chlorophyll content of tomato with REP, this study systematically analyzed the quantitative relationships and statistical characters between REP on various algorithms and leaf chlorophyll status,and then the linear regression,logarithmic regression,power regression,exponential regression and quadratic polynomial regression were used to develop the prediction models of the chlorophyll content for each REP extraction algorithm. The result showed that the logarithmic model of the linear extrapolation had the best accuracy and reliability.The calibration R^2 cwas 0. 618 6,the validation R^2 vwas 0. 771 1 and the root mean squared error of validation set（ RMSEv） was 8. 359 6. The exponential model of the four-point interpolation could beobtained easily according to reflectance at 670 nm,700 nm,740 nm and 780 nm,the calibration R2cwas0. 621 7,validation R^2 vwas 0. 766 6 and RMSEvwas 8. 568 2. The predictive ability was good enough to develop a monitoring instrument of tomato chlorophyll content.