Bin Li, Qiu Wang, Chao-Hui Zhan, Zhao-Yang Han, Hai Yin, Jun Liao, Yan-de Liu
Institute of Optical-Electro-Mechatronics Technology and Application, East China Jiao Tong University, National and Local Joint Engineering Research Center of Fruit Intelligent Photoelectric Detection Technology and Equipment, Nanchang, 330013, China.
Plant Methods. 2022 Apr 20;18(1):52. doi: 10.1186/s13007-022-00883-1.
Anthracnose of Camellia oleifera is a very destructive disease that commonly occurs in the Camellia oleifera industry, which severely restricts the development of the Camellia oleifera industry. In the early stage of the Camellia oleifera suffering from anthracnose, only the diseased parts of the tree need to be repaired in time. With the aggravation of the disease, the diseased branches need to be eradicated, and severely diseased plants should be cut down in time. At present, aiming at the problems of complex experiments and low accuracy in detecting the degree of anthracnose of Camellia oleifera, a method is proposed to detect the degree of anthracnose of Camellia oleifera leaves by using terahertz spectroscopy (THz) combined with laser-induced breakdown spectroscopy (LIBS), so as to realize the rapid, efficient, non-destructive and high-precision determination of the degree of anthracnose of Camellia oleifera.
Mn, Ca, Ca II, Fe and other elements in the LIBS spectrum of healthy and infected Camellia oleifera leaves with different degrees of anthracnose are significantly different, and the Terahertz absorption spectra of healthy Camellia oleifera leaves, and Camellia oleifera leaves with different degrees of anthracnose there are also significant differences. Partial least squares discriminant analysis (PLS-DA), support vector machine (SVM), and linear discriminant analysis (LDA) are used to establish the fusion spectrum anthracnose classification model of Camellia oleifera. Among them, the Root mean square error of prediction (RMSEP) and the prediction determination coefficient Rp of THz-LIBS-CARS-PLS-DA of prediction set are 0.110 and 0.995 respectively, and the misjudgment rate is 1.03%; The accuracy of the modeling set of THz (CARS)-LIBS (CARS)-SVM is 100%, and the accuracy of prediction set is 100%, after preprocessing of the multivariate scattering correction (MSC), the accuracy of the THz-LIBS-MSC-CARS modeling set is 100%, and the accuracy of prediction set is 100%; The accuracy rate of THz-LIBS-MSC-CARS-LDA of modeling set is 98.98%, and the accuracy rate of the prediction set is 96.87%.
The experimental results show that: the SVM model has higher qualitative analysis accuracy and is more stable than the PLS-DA and LDA models. The results showed that: the THz spectrum combined with the LIBS spectrum could be used to separate healthy Camellia oleifera leaves from various grades of anthracnose Camellia oleifera leaves non-destructively, quickly and accurately.
油茶炭疽病是油茶产业中常见的一种极具破坏性的病害，严重制约了油茶产业的发展。在油茶遭受炭疽病的初期，只需及时修剪病树的患病部位。随着病情加重，则需要根除病枝，对于重病植株应及时砍伐。目前，针对油茶炭疽病病情检测实验复杂、检测精度低的问题，提出一种利用太赫兹光谱（THz）与激光诱导击穿光谱（LIBS）相结合的方法来检测油茶炭疽病叶片的病情程度，以实现对油茶炭疽病病情程度的快速、高效、无损、高精度测定。
健康及不同炭疽病病情程度的油茶感染叶片的LIBS光谱中的Mn、Ca、Ca II、Fe等元素存在显著差异，健康油茶叶片以及不同炭疽病病情程度的油茶叶片的太赫兹吸收光谱也存在显著差异。利用偏最小二乘判别分析（PLS-DA）、支持向量机（SVM）和线性判别分析（LDA）建立油茶融合光谱炭疽病分类模型。其中，预测集的THz-LIBS-CARS-PLS-DA的预测均方根误差（RMSEP）和预测决定系数Rp分别为0.110和0.995，误判率为1.03%；THz（CARS）-LIBS（CARS）-SVM的建模集准确率为100%，预测集准确率为100%，经多元散射校正（MSC）预处理后，THz-LIBS-MSC-CARS建模集准确率为100%，预测集准确率为100%；THz-LIBS-MSC-CARS-LDA建模集准确率为98.98%，预测集准确率为96.87%。
实验结果表明：SVM模型定性分析准确率更高，比PLS-DA和LDA模型更稳定。结果表明：太赫兹光谱结合激光诱导击穿光谱可用于无损、快速、准确地将健康油茶叶片与不同等级炭疽病油茶叶片区分开来。
