| Citrus production is one of the most important agricultural economic activities in the world,and plays an important role in the international trade of fruits.The emergence of Citrus Huanglongbing has seriously damaged the quality of fruits and has become one of the problems that have plagued citrus growers all over the world.Moreover,at present,there is a lack of special drugs for Huanglongbing.Once the diseased tree is found,it must be removed,which affects the healthy and sustainable development of citrus industry.Therefore,timely and accurate detection of citrus leaf Huanglongbing is the key to ensure the healthy development of citrus industry.In this paper,laser induced breakdown(LIBS)technology was proposed to study the spectral intensity variation characteristics of nutrient elements in citrus leaves infected with Huanglongbing,and hyperspectrum was combined to detect yellow dragon disease in citrus leaves.It provides a rapid,nondestructive and effective detection method for citrus leaf Huanglongbing.The main content of this article is as follows:(1)With blade as the research object to study analyzed the huanglong health,moderate and severe disease that three kinds of four kinds of nutrient elements in the leaves of citrus Ca(Ⅰ),Mn(Ⅰ),Fe(Ⅰ)and Si(Ⅰ)corresponds to the changing rule of the LIBS spectrum intensity values,found in the healthy leaves of the spectral intensity were higher than in infected leaf nutrition element spectral intensity,And the spectral intensity decreases with the degree of disease.Then,the PLS-DA and SVM models of the original and pre-processed characteristic spectra were established respectively,and their comparative analysis was conducted.The results show that the SVM model has better effect.(2)The qualitative discrimination model of Huanglongbing degree by the spectral fusion of LIBS characteristic spectrum and hyperspectral three ROI regions(leaf tip,leaf and leaf tail)was studied.Compared with the PLS-DA model established by LIBS characteristic spectrum and hyperspectral three ROI regions(tip,leaf and tail)spectrum,the fused characteristic spectrum and tip spectrum fusion model have the best effect.RMSEC of the model was 0.303,R_C was 0.927,and the total misjudgment rate was 9.8%.RMSEP was 0.362,R_P was 0.896,and the total misjudgment rate was 13.3%.The original spectrum after the fusion of characteristic spectrum and blade tip spectrum is preprocessed,and the model pretreated by quantitative normalization has the best effect.Further,SPA and UVE are used to screen wavelength variables,and the discriminant analysis model is established combined with PLS-DA and MLR.Among them,UVE combined with MLR had the best modeling effect.RMSEC of modeling set was 0.101,R_C was 0.992,and the total misjudgment rate was 0.RMSEP of prediction set was 0.169,R_P was 0.991,and the total misjudgment rate was 1.3%.(3)The true contents of nutrient elements Ca(Ⅰ),Cu(Ⅰ),Mn(Ⅰ),Fe(Ⅰ)and Mg(Ⅰ)in five kinds of leaf samples were detected by flame atomic absorption spectrometry.The statistical analysis methods of one-way variance and multiple tamni black in SPSS 25.0application software were used to analyze the significant differences of the real content of each nutrient element in different disease degrees of Huanglongbing.PLS and LS-SVM quantitative analysis models were established by combining the real content of nutrient elements with the spectral information measured by LIBS.The results show that PLS quantitative analysis model has good effect and can accurately predict the concentration of nutrient elements. |
