厦门大学学报（自然科学版）2025，Vol.64Issue(5)：768-774,7.DOI:10.6043/j.issn.0438-0479.202311002
基于激光诱导击穿光谱鉴别柑橘黄龙病
Identification of citrus Huanglongbing based on laser induced breakdown spectroscopy
摘要
Abstract
[Objective]Citrus Huanglongbing(HLB)is a severe threat to citrus production and quality due to its rapid spread and incurable characteristics.Laser induced breakdown spectroscopy(LIBS)offered advantages of simple sampling and fast detection.When combined with chemometric algorithms,it has the potential for rapid detection of citrus HLB.[Methods]In this study,healthy and HLB-infected navel orange(Citrus sinensis)leaves were used as research subjects.The LIBS spectra centered at 390,516,589,616 and 646 nm were collected from 222 leaves(with each leaf reconstructed as a single spectrum)using 800 nm laser pulses(about 100 fs pulse duration,100 μJ pulse energy).The spectra were then randomly divided into training set and validation set in an approximate 3∶1 ratio.Principal component analysis(PCA),partial least squares discriminant analysis(PLS-DA),and orthogonal PLS-DA(OPLS-DA)were applied to model and analyze the spectral data,respectively.The efficiency of data separation of each model and the accuracy in distinguishing HLB infection were evaluated.In addition,the ability to discriminate between different degrees of infection in navel orange leaves was assessed.[Results]The LIBS spectra showed that the intensities of the Ca(Ⅱ)and Mg(Ⅰ)characteristic spectral lines were higher in healthy leaves than in HLB-infected leaves.This is likely because HLB damages the roots and obstructs the phloem of navel oranges,which affected the absorption of mineral elements by the root system.In the discrimination of HLB infection,score plots from PCA,PLS-DA,and OPLS-DA were compared.Significant overlaps between healthy and HLB-infected leaves was observed in PCA and PLS-DA score plots,whereas the OPLS-DA model showed the best data separation ability.The OPLS-DA model reached 100.00％accuracy for both the training and validation sets.The fitting indices,Rx2 and Ry2 of the OPLS-DA model,were 0.889 and 0.988,respectively,and the prediction index Q2 was 0.917.To further verify the effectiveness of the OPLS-DA modeling,a permutation test was conducted.The Q2 of the original model was greater than that of the randomly permuted Y-observations,and the intersection of the Q2 regression line on the Y-axis was below 0.These results indicated that the model was valid and not overfitted.For discrimination of the degree of HLB infection,the score plots showed relatively small difference between mildly and severely infected navel orange leaves.The accuracies in distinguishing mild HLB of training and validation sets were 92.86％and 72.22％,respectively.[Conclusions]This study demonstrated that LIBS combined with OPLS-DA discriminant model enables fast and accurate detection of HLB infection.This approach could effectively distinguish between the healthy and HLB infected navel orange leaves,with no false positives in either training or validation sets.To a certain extent,it could also distinguish the mildly infected navel orange leaves with relatively high accuracy.Therefore,LIBS combined with OPLS-DA has the potential for early diagnosis of HLB infection.These findings are of great significance for the development of rapid identification methods for HLB.关键词
激光诱导击穿光谱/柑橘/黄龙病/正交偏最小二乘判别分析Key words
laser induced breakdown spectroscopy/citrus/Huanglongbing/orthogonal partial least squares discriminant analysis分类
化学化工引用本文复制引用
代文静,秦宏坤,徐未,何玉韩,李敏,王朝晖..基于激光诱导击穿光谱鉴别柑橘黄龙病[J].厦门大学学报（自然科学版）,2025,64(5):768-774,7.基金项目
国家自然科学基金(22072121) （22072121）
