Diagnosis of Citrus Greening using Raman Spectroscopy-Based Pattern Recognition

This study verifi ed the applicability of Raman spectroscopy for the detection and classifi cation of disease in citrus leaves. The Raman spectra of citrus leaves were collected using a SENTERRA confocal microprobe Raman spectrometer and divided into fi ve types, Huanglongbing (HLB), moderate HLB, serious HLB, nutrient defi ciency, and normal. The backgrounds of the spectra were deducted by different methods, and partial least squares discrimination analysis (PLS-DA) and extreme learning machine (ELM) were used to build the mathematical model. At the same time, the data dimension was reduced using principal component analysis (PCA) and successive projection algorithm (SPA) in order to optimize and improve the classifi cation accuracy of the model. The experiments showed that the predictive ability of the PLS-DA model with 1850 input variables by 2 times polynomial fi tting deducted backgrounds was better, the recognition correct rate being 100%. The results show that Raman spectroscopy has potential for rapid diagnosis of citrus HLB.