Identification of Huanglongbing-infected navel oranges based on laser-induced breakdown spectroscopy combined with different chemometric methods
| Autor: | Liu Muhua, Xu Fanghao, Huang Lin, Tianbing Chen, Rao Gangfu, Mingyin Yao, Xue-hong Xu, Chen Jinyin, Luo Ziyi |
|---|---|
| Rok vydání: | 2018 |
| Předmět: |
0106 biological sciences
Diagnostic methods Wavelet Analysis Navel 01 natural sciences Automation Optics medicine Laser-induced breakdown spectroscopy Electrical and Electronic Engineering Engineering (miscellaneous) Scatter correction Plant Diseases Mathematics Principal Component Analysis Training set business.industry Wavelength range Lasers Spectrum Analysis 010401 analytical chemistry Models Theoretical Atomic and Molecular Physics and Optics 0104 chemical sciences medicine.anatomical_structure Principal component analysis Navel orange business Biological system Algorithms Citrus sinensis 010606 plant biology & botany |
| Zdroj: | Applied Optics. 57:8738 |
| ISSN: | 2155-3165 1559-128X |
| Popis: | In order to realize rapid identification of Gannan navel oranges infected by Huanglongbing (HLB), a full optical diagnostic method of laser-induced breakdown spectroscopy (LIBS) was proposed. All navel oranges were collected from Ganzhou, Jiangxi, China, and samples contain healthy and HLB-infected navel oranges. The LIBS spectra of the plasma plume were collected directly from the epidermis of these navel oranges. The navel orange LIBS spectra in the wavelength range of 200-1050 nm were pretreated with smoothing and multiple scatter correction; on the basis of 10×10-fold cross validation, a random forest (RF) model based on continuous wavelet transform (CWT) and principal component analysis (PCA) were analyzed to identify the navel orange of HLB. The results showed that the PCA-RF and CWT-RF models coupled with suitable methods in preprocessing data can identify HLB-infected navel oranges. The average accuracy obtained from the CWT-RF model was 96.86% in the training set and 97.45% in the test set; the average accuracy by the PCA-RF model was 97.64% in the training set and 97.89% in the test set. The overall results demonstrate that LIBS combined with CWT-RF or PCA-RF, as a valuable analytical tool, could be used for HLB-infected navel orange identification. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|