Early Detection Of Sclerotinite Infected Oilseed Rape Plants Based On Spectrocopy Techniques

Sclerotinia sclerotiorum is one of three major diseases for rapeseed plant,which seriously restricts the yield and quality of rapeseed.Sclerotinia sclerotiorum in rapeseed plant can occur from the seedling stage to the harvest stage.Infected petals is considered to be the main source for spreading sclerotinia sclerotiorum,which can rapidly cause fungal infestation of stem and leaf tissue of rapeseed plant.The infected stem can also cause the lodging of whole plant,resulting in the most serious loss of rapeseed.Meanwhile,infected leaf tissues affect the transmission of plant nutrients,and thus affect the quality of rapeseeds.In this study,the rapeseed crops were used as the research object.Firstly,visible and near-infrared hyperspectral imaging technology was employed to finish the dynamic monitoring of rape leaves based on the spectral characteristics and chlorophyll content in the invaded tissues infected by sclerotinia sclerotiorum.Secondly,laser confocal microscopy Raman spectroscopy was used to achieve the early dynamic diagnosis of sclerotinia sclerotiorum infection using by detecting the Raman spectra of tissues around different lesion diameters in the microscopic view.Thirdly,near-infrared?NIR?hyperspectral combined with Raman spectroscopy techniques were employed to perform rapid diagnosis and detection of the rape petals infected by sclerotinia sclerotiorum,and In-situ in vivo detection of petal composition of rapeseed during the infection of sclerotinia sclerotiorum was explored by Raman imaging distribution of petal sample compounds.Fourthly,confocal laser Raman spectroscopy was selected to carry out the early diagnosis of sclerotinia sclerotiorum invading stems by using chemical imaging of the healthy tissue of the cross-section of the stems in the rapeseed plants and the compounds of the infected tissues in microscopic perspective.In all,this thesis proceed the early involvement of sclerotinia sclerotiorum on the leaf,petal and stem tissue of rapeseed plant from the macroscopic perspective?frequency multiplication and combination frequency plane of stretching vibrations containing hydrogen radicals?and microscopic perspective?fingerprint of compound/cell hierarchy?.The results of the study provides a theoretical basis for scientific spraying and precision management of field crops.And it also offers a new research approach for further investigation of host-pathogen interaction mechanisms at the cell level.The main conclusions of this research are as follows:?1?Hyperspectral imaging technology is used to rapidly detect the dynamic development of rapeseed leaves infected by sclerotinia sclerotiorum.The hyperspectral response law of the health rapeseed leaves and the sclerotinia sclerotiorum infected rapeseed leaves within different stages infection was firstly studied.Based on the analysis of variance?ANOVA?and the difference in reflectance between the four types of samples,it is found that there is a significant difference in the sample at 687 nm?p<0.05?between healthy and the infected rapeseed leaf samples.Next,principal component analysis?PCA?was executed,and result found that there were a good cluster in two-dimensionally composed of principal component-1 and principal component-4 between the healthy sample and the diseased samples at different stages.Partial least squares discriminant analysis?PLS-DA?models were established through the characteristic wavebands selected by reflectance difference,x-loadings,and Random frog?RF?algorithm.The PLS-DA models found that the accuracy of the PLS-DA model established by spectra at 630,640,650,757 and 789 nm which were selected by the random frog method was 71.3%.Then,partial least squares regression?PLSR?model was established based on spectra at the six selected characteristic wavebands by the RF and the chlorophyll content,producing in the correlation coefficient of 0.663.Based on the PLSR model,a linear equation y=4.028-2.411_x1-2.123_x2-2.771_x3-0.140_x4-2.586_x5-2.037_x6?where,y is the predicted chlorophyll content;xi is the spectral reflectance at wavelength of i nm?was obtained,which could achieve chemical imaging of chlorophyll content in rape leaves,so as to carry out rapid detection of dynamic development of rapeseed leaves infected by sclerotinia sclerotiorum.?2?Microscopic Raman spectroscopy was used to carry out the early dynamic diagnosis of rapeseed infestation by sclerotinia sclerotiorum from the microscopic view.Herein,Raman spectra in three stages of disease development?health,infected area 1 and infected area 2?were acquired,and the wavelet transform?WT?and baseline calibration were used to preprocess for Raman spectra.According to ANOVA statistical results,the Raman intensity values in the three types of samples located at 1006,1156,and 1522 cm^-1 were significantly different at the p<0.05 level;the three types of samples offered a good clustering effect in Two-dimensional space?PC-1 and PC-2?and three-dimensional space composed of 1006,1156 and 1522 cm^-1.After that,least squares support vector machine?LS-SVM?discrimination models based on PC-1,PC-2 and characteristic peaks at 1006,1156 and 1522 cm^-1 were built,and the recognition rates of the models were 100%,66.7%,70.0%,100%,and 100%,respectively.The results showed that characteristic peaks of 1156 cm^-1 and 1522 cm^-1 could be used for the early dynamic diagnosis of rapeseed leaves infected by sclerotinia sclerotiorum.?3?Near-infrared?NIR?hyperspectral imaging was employed to achieve rapid detection of rapeseed petals infected by sclerotinia sclerotiorum.First,PCA was executed and the results showed that the healthy petals and the infected petals had a clear classification on the principal component?PC?PC-3 axis.Then,ANOVA analysis found that the healthy petals and the infected petals exhibited a significant difference?p<0.05?in the 1446 nm selected by the x-loading of PC-3.After that,LS-SVM discriminant models were established by characteristic wavebands based on the x-loading coefficient and RF method.The results showed that the LS-SVM model established by six characteristic bands at 1190,1460,1463,1524,1446,and 1656 nm could reach the accuracy of 92.7%.?4?Laser Raman spectroscopy was used to rapidly detect the rapeseed petals infestation by sclerotinia sclerotiorum,and the detection of chemical composition changes of the petals was carry out by the chemical imaging of rapeseed petal components during the infection of sclerotinia sclerotiorum.Based on principal component analysis?PCA?data compression and cluster analysis,the results showed that the two types of samples have better clustering effect in the space of principal component-1?PC-1?and principal component-2?PC-2?,and then according to characteristic wavebands selected by the x-loading coefficient and regression coefficient?RC?methods,LS-SVM discriminant models were established.And the results showed that recognition rate of 95.08%was obtained using the 5 characteristic peaks(1006,1156,1188,1213,and 1522 cm^-1)identified from fingerprint peaks.However,the 7 characteristic peaks(1006,1077,1156,1188,1271,1392,and 1526 cm^-1)selected using the x-loading coefficient could achieve recognition rate of 100%.Then,the chemical imaging of based on the 5 characteristic peaks was finished,and results found that the distribution of substances on the healthy petals was uniform,and the uneven distribution of the compounds occurred due to the damage of internal cells caused by the infestation of mycelial petals.The results also showed that the chemical imaging analysis of the chemical composition of healthy petals and infected petals tissue can be performed from the microscopic point of view to achieve the in situ detection of rape petal composition during the infection of sclerotinia sclerotiorum.?5?Microscopic detection of cell cross-section of stem in rapeseed plant infected by sclerotinia sclerotiorum was carried out by surface enhancement and laser confocal microscopy Raman spectroscopy.In this study,silver nanosol was used as a surface enhancer,and median filter and adaptive iterative re-weighted penalized least squares?air-PLS?were used to perform the preprocessing of Raman spectra noise and fluorescence background.The Raman spectra of different positions of cross-section tissues of stems infected by sclerotinia sclerotiorum were obtained,and the three types of cells with different degrees of infection had good clustering effect in three-dimensional space of PC-1,PC-2 and PC-3.The recognition rate of the established LS-SVM discriminant model based on the Raman peak intensity at 1122 cm^-1 was 83.3%.Single band imaging for observation the trend of cellular compound changes was performed based on the characteristic peaks at 1005,1122,1156,1522,1307 and 1365 cm^-1.The results showed that surface-enhanced Raman microscopy imaging technology can be used to in situ detect the invasion of stem by sclerotinia sclerotiorum from the cell level.