Methods For Early Warning Diseases And Insect Pests Of Cucumber In The Greenhouse Based On The Analysis Of Chlorophyll Fluorescence Spectra

The industrial production of high yield and excellent quality is the ultimate goal ofagricultural facilities management; however, in the daily production of facility agriculture, inorder to achieve the requirement of high yield rigidly, pesticides have been widely used as acompensatory method, which makes the crop quality safety not guaranteed. Seeking newmethods to check diseases and insect pests fast, nondestructively, and accurately becomesthe key problem of facility agriculture research. This word is supported by National HighTechnology Research and Development Program―Laser-induced Plant PhysiologicalInformation Detection Sensors and Diagnosis‖(2007AA10Z203), and―Full AutomaticGrafting and Growing Seedlings Key Technology and Complete Sets of EquipmentResearch‖(2012AA10A506), and Jilin Province's Science and Technology DepartmentProgram―Key Technology Research on Major Diseases and Insect Pests Digital Monitoringand Early Warning System in Main Rice Growing Districts of Jilin Province‖(20110217),use the spectral characteristic information and plant physiological information changes toresearch greenhouse crop health status changes, with the study object of greenhousecucumber, which is the foundation to achieve plant factory manufacturing management.Chlorophyll fluorescence can reflect the plant physiological information, while this paperuses chlorophyll fluorescence spectroscopy analysis technology to reflect the physiologicalinformation changes and afterwards to reflect leaves' status changes.Starting with the spectral shape, study the distinction of healthy and verminous leaves, andextract spectral characteristic information: with the spectral second peak emissionwavelength685nm as discriminant critical value, judge that of healthy leaves to be smallerthan685nm, and that of verminous leaves greater than685nm; with the spectral first wavetrough intensity0Counts as discriminant critical value, judge that of healthy leaves to begreater than0Counts, and that of verminous leaves smaller tha n0counts. with theF512~F632wave band's spectral curve change rate k=3as discriminant critical value, judgethe leave to be healthy whose change rate k is less than3, and to be verminous whose changerate k is more than3.Use of the leaves of the physical information index, establish the leaves of the physicalinformation model, further judge the leaves of the health status. Use the fluorescence spectralindex and photosynthetic active radiation establish of leaf net photosynthetic rate model, toachieve the prediction set's correlate coefficient of0.825, being evidently interrelated in thelevel of0.01; use the fluorescence spectral index connected to leaf temperature to establishleaf moisture availability model, to achieve the predictionset's correlate coefficient of0.984, being evidently interrelated in the level of0.05; combined the spectral analysis with dataexcavation method, establish chlorophyll content SPAD model, use simple wave bandsautocorrelation method to screen the efficient, utilize the principal component analysismethod to reduce the spectral dimension, and finally use least squares support vectormachine regression to establish model, achieve the prediction set's correlate coefficient of0.946. Regard leaves' Net photosynthetic rate, moisture availability and chlorophyll contentas the auxiliary information judging leaves whether to be healthy.With the leaves of the health chlorophyll fluorescence spectra data as standard, Usechlorophyll fluorescence analysis technique combined with data excavation method toestablish the classified diagnosis model of healthy and downy mildew disease leaves,powdery mildew disease and aphid disease leaves. By contrasting the three methods of leastsquares support vector machine, discriminant analysis and BP neural network, and theprincipal component analysis method and the spectral dimension reduction method ofwavelet change, define the method of utilizing the least squares support vector machine andprincipal component analysis to reduce the model's dimension and establishing models, andrespectively establish the classified diagnosis model of gotten principal component figures,determine to use10principal components as the model input variables, at last achieve theprediction set's correlate coefficient of0.939, mean square error of0.389, accuracy rate of93.3%.Respectively analyze the spectral data of healthy and downy mildew disease leaves, bycontrasting the three method of least squares support vector machine, discriminant analysisand BP neural network, compare a derivative, two derivative, multiple scattering correction,derivative and multiple scattering correction of spectral pre-treatment methods, andprocessing the gotten spectrum by principal component analysis and wavelet noise anddimension reduction methods, finally define to use the methods of least squares supportvector machine, derivative and principal component analysis to establish downy mildewdisease prediction model, and make comparative analysis of the gotten principal componentnumber. Achieve10principal components variable number, and the effect of establishedmodels of healthy, downy mildew infection leaf, early downy mildew disease leaf, and leafwith large disease area is good, and the final RMSEP of the prediction set is0.375, accuracyrate is96.6%, correlate coefficient of the true value of prediction set leaf sample andpredictive value is0.949.Respectively analyze the spectral data of healthy, powdery mildew infection leaf, earlypowdery mildew disease leaf, and leaf with large disease area, by the methods of leastsquares support vector machine, discriminant analysis and BP neutral network, use simplewave bands autocorrelation method to screen61~70spectral wave bands as the efficient.Through the choice of principal component analysis and wavelet noise and dimensionreduction methods, and the discussion of the different influence to powdery mildew diseaseleaves prediction model from different principal components number. Define to utilize8principal components, and establish powdery mildew disease leaves prediction model. The final RMSEP of the prediction set is0.489, accuracy rate is91.3%, correlate coefficient ofthe true value of prediction set leaf sample and predictive value is0.890.Utilize chlorophyll fluorescence analysis method and establish aphid disease leavesprediction model. Respectively collect and analyze the fluorescence spectral data of healthy,a small amount of pest eggs and adult, large amount of pest eggs and adult leaves, and leafwith large disease area or filled with aphid disease, by contrasting the three method of leastsquares support vector machine, discriminant analysis and BP neural network, and define touse least squares support vector machine method. Use the wave bands with peak and troughto screen the efficient, make sure to use F632wave band as the efficient, and utilize principalcomponent analysis method to reduce spectral dimension. Discuss the influence to the modelfrom different factor number, define to use8principal components as input variables andestablish cucumber aphid disease prediction model. The final RMSEP of the prediction set is0.569, accuracy rate is97.5%, correlate coefficient of the true value of prediction set leafsample and predictive value is0.981.In addition, by the regularity of disease outbreak and actual situation, plant diseases andinsect pests in high contrast with low in hair environmental conditions, Sure the occurrenceof pest and disease conditions threshold. Analyze the maximum, minimum, difference andaverage value of temperature and humidity, and define to utilize average temperature andhumidity as the nosopoietic threshold value. The scopes of temperature and humidity are15℃≤Tmean≤25℃,40%≤RHmean≤70%; moreover, combine and then analyze theparameter of temperature and humidity in June, September and July, August; according tothe calculation of pathogen infection time, judge the continues temperature and humility for6hours, and achieve the scopes of temperature and humility are15℃≤6hTmean≤30℃,50%≤6hRHmean≤100%.Combine the survey of greenhouse environmental condition with chlorophyll fluorescencesurvey, and construct diseases and insect pests monitoring system of greenhouse cucumber.First of all, by analysing the environmental condition data and after monitoring the thresholdvalue of environmental condition, if the conditions of plant diseases and insect pests'occurrence are satisfied, monitor and analyze the chlorophyll fluorescence data. Accordingto the classified diagnosis model make sure of the occurrence type of plant diseases andinsect pests, and by the prediction model define the degree and situation, and achieve thereports of plant diseases and insect pests judging monitoring and calculating; utilize the earlywarming system to pinpoint monitoring and calculating reports, send the result to maincontrol room, greenhouse and managers, and preserve the data, then a process of earlywarming is accomplished.