Study On Diagnosis Method Of Citrus Greening Based On Hyperspectral Imaging

Citrus greening(Huanglongbing,HLB)is one of the most devastating diseases in citrus production,which has greatly restricted the healthy and sustainable development of the citrus industry.At present,there is no effective treatment for HLB,Only diseased trees can be identified and removed as early as possible to prevent further spread of the disease.Therefore,timely and accurate detection of HLB has important practical significance in controlling its spread and ensuring the healthy development of the citrus industry.At present,the traditional HLB detection method needs to destroy the sample,has high cost and long cycle,and it is difficult to be widely applied in actual production.Using hyperspectral imaging technology to study the spectral and image response characteristics of the main physical and chemical properties of citrus affected leaves,explore a non-destructive and accurate detection method of HLB,and provide support for the application of hyperspectral imaging technology in precision agriculture.The main research contents of this thesis are as follows:1.The non-destructive detection mechanism of HLB based on hyperspectral imaging technology was clarified,and the changes of main physical and chemical indicators in citrus leaves of three different disease levels were analyzed.There is a certain linear relationship between the change of physical and chemical index content and spectral absorption: the average chlorophyll content in normal leaves was the highest,while that in chlorophyll deficient leaves was the lowest.The average content of soluble sugar and starch in HLB leaves was higher than that in normal,which were 31.23 ± 12.81 and 9.24 ± 3.71 mg/g,respectively.2.Using modern scientific instruments and experimental analysis methods,the effects of image region selection,spectral pretreatment methods and spectral characteristic variables on the spectral discrimination model of HLB were studied.It is concluded that the region of interest with the best image is located in the middle of the left side of the leaf vein and the pixel area is 100,and the second-derivative pretreatment method combined with the successive projection algorithm(SPA)is the best choice of spectral characteristic variables.It was concluded that the partial least squares discrimination analysis(PLS-DA)model with 27 variables selected by SPA achieved the optimal discriminant effect on normal,slight HLB,moderate HLB,serious HLB and chlorophyll deficient leaves,with a misjudgment rate of 5%.3.The non-destructive detection methods of chlorophyll,soluble sugar and starch in HLB leaves based on spectral information were explored,and linear and nonlinear mathematical models were established.It was found that the spectral characteristic peak of leaves at 730 nm was lower than that of normal leaves.The results show that the calibration set composed of composite samples combined with the second-derivative spectrum can beused to predict three physicochemical indexes.Linear model has the best predictive effect.The model correlation coefficients of chlorophyll,soluble sugar and starch were 0.86,0.74 and 0.82,respectively.The root mean square error of prediction were 8.86,9.75 and 1.42,respectively.4.The multi block gray level co-occurrence matrix(MB-GLCM)algorithm for visual identification of HLB was proposed.The sensitive wavelengths of HLB were extracted and a discriminant model based on texture features was established.The results showed that the PLS-DA model of image texture features was the best when using SPA to extracted 9 spectral characteristic wavelengths,with the minimum misjudgment rate of 3.12%.The visual discriminant model based on MB-GLCM was constructed to realize the visual diagnosis of HLB.