基于近红外光谱的大豆叶片可溶性蛋白含量快速检测（英文）  

Rapid determination of soluble protein content for soybean leaves based on near infrared spectroscopy

作　　者：张亚坤 罗斌[2,3] 宋鹏[2,3] 潘大宇 路文超[2,3] 周亚男 王成[2,3] 赵春江 Zhang Yakun;Luo Bin;Song Peng;Pan Dayu;Lu Wenchao;Zhou Yanan;Wang Cheng;Zhao Chunjiang(School of Electrical and Information,Northeast Agricultural University,Harbin 150030,China;Beijing Research Center of Intelligent Equipment for Agriculture,Beijing 100097,China;National Research Center of Intelligent Equipment for Agriculture,Beijing 100097,China)

机构地区：[1]东北农业大学电气与信息学院,哈尔滨150030 [2]北京农业智能装备技术研究中心,北京100097 [3]国家农业智能装备工程技术研究中心,北京100097

出　　处：《农业工程学报》2018年第18期187-193,共7页Transactions of the Chinese Society of Agricultural Engineering

基　　金：National Natural Science Foundation of China(31601216);Special Construction of Innovation Ability of Beijing Academy of Agriculture and Forestry Science(KJCX20170418)

摘　　要：可溶性蛋白是植物生化及抗性生理研究的重要指标之一。快速、准确、无损测定可溶性蛋白含量对作物生长状况的动态监测及抗性作物品种的筛选具有重要意义。近红外光谱具有快速、简单方便、非破坏性的特点,已在农业、食品、化工等领域广泛应用,尤其是近年来基于光谱技术快速无损的获取作物生理生化信息的研究已成为当前农业领域研究的热点。本文采用近红外光谱技术结合化学计量学方法以实现大豆叶片可溶性蛋白含量的快速无损检测。首先,采用Savitzky-Golay平滑(SG)、一阶导数(1-Der)、二阶导数(2-Der)等7种光谱预处理方法分别建立大豆叶片可溶性蛋白含量的偏最小二乘(PLS)预测模型,经对比发现SG预处理方法为大豆叶片可溶性蛋白含量预测的最优光谱预处理方法。其次,分别采用连续投影算法(SPA)、随机蛙跳(RF)和遗传算法(GA)对SG预处理后的光谱数据进行特征波长提取。最后,基于提取的特征波长分别建立了大豆叶片可溶性蛋白含量的SPA-PLS、RF-PLS和GA-PLS预测模型,发现基于SPA提取的11个特征波长建立的大豆叶片可溶性蛋白含量SPA-PLS模型具有最佳的预测效果,其预测集相关系数(R2p)为0.864,预测均方根误差(RMSEP)为1.894 mg/g,预测偏差为2.061(RPD)。上述结果表明,应用近红外光谱技术检测大豆叶片中可溶性蛋白含量是可行的,可为大豆生长状况动态监测及抗性大豆品种的筛选提供新的方法。Soluble protein is an important indicator for the research of plant physiochemical and resistance physiology.Rapid,accurate and non-destructive detection of soluble protein content in crops is critical for dynamic monitoring of growth state and selecting varieties with strong resistance.2 soybean varieties,Qihuang35 and Zhonghuang13,were planted and treated with different copper and salt stresses.The near infrared spectra of stressed soybean leaves were obtained by the AOTF(acousto-optic tunable filter)near infrared spectrometer.The soluble protein contents of soybean leaves were measured by coomassie brilliant blue method.Chemometric methods were applied to build multivariate calibration models for the rapid and nondestructive determination of soluble protein content in soybean leaves based on near infrared spectra.Several partial least squares(PLS)models with different preprocessing methods like Savitzky-Golay smoothing(SG),first derivative(1-Der),second derivative(2-Der),standard variable normalization(SNV)and multiplicative scatter correction(MSC)were developed and compared.Then successive projections algorithm(SPA),random frog(RF)and genetic algorithm(GA)were employed to select effective wavelengths with spectral data preprocessed by SG.11,10 and 43 of effective wavelengths were selected by SPA,RF and GA respectively.These selected effective wavelengths were used as the inputs of partial least squares(PLS)to develop SPA-PLS,RF-PLS and GA-PLS models.Results showed that the best prediction results for the determination of soluble protein content were achieved by SPA-PLS model using SG spectra with prediction determination coefficient(R2 p)of 0.746,root mean squares error of prediction(RMSEP)of 1.894 mg/g and ratio of prediction to deviation(RPD)of 2.061.The overall results indicated that a strong correlation was existed between near infrared spectra and soluble protein content,and near infrared spectroscopy technology combined with SPA-PLS models was a feasible method for the rapid and nondestructive detection

关 键 词：近红外光谱 大豆叶片 可溶性蛋白 连续投影算法 偏最小二乘算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]