基于AIRF-CARS波段选择算法的橡胶树叶片氮含量定量研究

Quantitative Study of Nitrogen Content in Rubber Tree Leaves Based on AIRF-CARS Band Selection Algorithm

以橡胶树叶片的近红外光谱信息为分析对象,运用由粗放到细致的多分辨率特征提取思想,提出了一种融合自适应间隔随机蛙与竞争自适应重加权采样(AIRF-CARS)的算法提取橡胶树叶片的光谱特征信息,从而实现了橡胶树叶片氮含量的定量分析.实验结果表明,AIRF-CARS算法有效的压缩了光谱特征的数量,通过算法选择的特征波长为22个,使得定量分析模型的预测均方根误差(RMSEP)和决定系数(R2)分别为0.1364%和0.9596.因此,本文算法可以有效地提取信息量较大的波长特征,应用于近红外光谱检测的定量分析中,并为便携式田间多波段光谱仪的研发提供理论支撑.

In the report,the near infrared spectrum of rubber tree leaves information was selected as analysis object.The multiresolution feature from extensive to meticulous was used,and a adaptive fusion adaptive random frog and competition between heavy weighted sampling(AIRF-CARS)algorithm was used to extract the spectral feature information of rubber tree leaves,so as to perform the quantitative analysis of rubber tree leaf nitrogen content.The results showed that AIRF-CARS algorithm can effectively compress the number of spectral features,and the selected characteristic wavelength is 22,the prediction root mean square error(RMSEP)and determination coefficient(R2)of the quantitative analysis model was 0.1364%and 0.9596,respectively.Therefore,the proposed algorithm can effectively extract wavelength features with large amount of information,which can be applied to the quantitative analysis of near-infrared spectroscopy detection,and which can provide theoretical support for the development of portable field multi-band spectrometer.