基于分数阶微分的橡胶树割胶期氮素含量定量检测

Quantification detection of nitrogen content in rubber tree during tapping period based on fractional calculus

【目的】检测橡胶树割胶期叶片的实际氮素含量,建立橡胶树叶片光谱诊断模型,为实现橡胶树叶片氮素含量的快速无损检测提供参考依据。【方法】使用FieldSpec 3光谱仪采集割胶期橡胶树叶片的光谱反射率,分别以其原始光谱(R)、倒数光谱(1/R)、对数光谱(log R)和对数倒数光谱(1/log R)作为光谱信息,采用分数阶微分进行处理,获得不同分数阶阶次下的光谱数据,并通过竞争性自适应重加权算法(CARS)选择变量及偏最小二乘回归法(PLSR)建立橡胶树氮素光谱诊断模型。【结果】采用分数阶对橡胶树叶片R、1/R、log R和1/log R建立模型的最优均方根误差(RMSE)分别为0.1376、0.1175、0.1263和0.1505,且使用1/R数据建立的0.6阶模型表现最优,相关系数为0.9273,RMSE为0.1175,决定系数为0.8551。与整数阶算法相比,分数阶模型具有更强的预测能力,表明分数阶能充分挖掘光谱信息的有效信息,有效提高橡胶叶片氮含量光谱诊断模型的预测精度,实现橡胶树叶片氮素含量快速无损检测。【结论】应用近红外光谱技术并结合分数阶微分算法可快速无损检测橡胶树叶片氮素含量,为生产上橡胶树的精准可变量施肥提供技术支持。

[Objective]The actual nitrogen content of rubber leaves during tapping period was detected,and the spec-tral diagnostic model of rubber leaves was established,which could provide reference for the rapid and non-destructive de-tection of nitrogen content of rubber leaves.[Method]Fieldspec 3 spectrometer was used to collect the spectral reflectance of rubber leaves during rubber cutting period.The original spectrum(R),reciprocal spectrum(1/R),logarithmic spectrum(logR)and logarithmic reciprocal spectrum(1/logR)were used as spectral information respectively.Fractional calculus was used to process the spectral data under different fractional orders.The diagnostic model of nitrogen spectrum in rubber tree was established by selecting variables using competitive adaptive reweighted sampling(CARS)and partial least square regression(PLSR)algorithm.[Result]The optimum root mean square error(RMSE)of fractional-order for origi-nal spectrum,reciprocal spectrum,logarithmic spectrum and logarithmic reciprocal spectrum were 0.1376,0.1175,0.1263 and0.1505,respectively.Andthe0.6-order model based on reciprocal spectral data performed the best,the correlation coe-fficient index was 0.9273,RMSE was 0.1175 and determination coefficient was 0.8551.Compared with the integer-order algorithm,the fractional-order model had better prediction ability.The results showed that the fractional order was able to dig the effective information of spectral information and effectively improve the prediction accuracy of the spectral diag-nostic model of nitrogen content in rubber leaves and finally achieved the rapid and non-destructive detection of nitrogen content of rubber leaves.[Conclusion]The application of near infrared spectroscopy and fractional calculus algorithm can quickly and non-destructively detect the nitrogen content in rubber leaves,providing technical support for precise variable fertilization of rubber trees in production.