Early Determination Of Rice Diseases And Pests Using Spectrum Analysis Technology

The rice plant of TN1type was used as the research object. NIR（Near Infrared spectroscopy）, Raman spectroscopy and hyperspectral imagetechnology were used in this research aimed at early detection of rice diseasesor pests. The results and conclusions are as following:1. Established the qualitative models for the insect pests rice plant based on NIR.Establish the spectrometry quantitative analysis model of rice plants damaged bybrown planthopper. After comparing different spectral pretreatment methods used inPLSDA, SVMDA, PLS-SVMDA and choosing the best method, the experimentalresults showed: the last2methods’ total accuracy rate could reach100%. Afterextracting the feature wavelength by GA （Genetic Algorithm） and comparing the bestmodel grown in each mode, the results showed that: the total accurate rate ofGA-PLSDA was92.01%, while GA-SVMDA and GA-PLS-SVMDA were97.37%;Establish the spectrometry quantitative analysis model of rice plants damaged byChilo suppressalis. The result showed: the total accuracy of all the three methodsreached100%. After extracting the feature wavelength and building models, theresult showed: the total accurate rate of GA-PLSDA was78.89%, while GA-SVMDAand GA-PLS-SVMDA was100%.2. Study on the early and non-destructive determinination of the insect pests riceplant based on NIR-Raman spectroscopy.After studying the differences of Raman spectroscopy between healthy rice andrice infected with insect pests and the possibility of early detection through theportrait research of rice plants infected with pests, the result showed: with the harmdegree becoming worse, the total vibration intensity of Raman spectrum alsoincreased; After some relative pretreatment, we can find out whether the rice planthad been damaged by rice stem borer or not via observing the Raman’s peak intensityor calculating the wavelength’s change in the point of455cm^-1,699cm^-1,1248cm^-1.3. Established disease level classification model of rice plant frost and rice pestbased on hyper-spectral imaging technology. Firstly, we designed the high-spectral image acquisition system and researchedthe optimum imaging conditions. Then, after performing the analysis of spectrum andimage processing, we found that:In the first test, at400nm⁸00nm band, spectral intensity of affected rice washigher than that of healthy rice; at460nm, the damaged rice has green mobilephenomenon; after780nm, red edge shifted to short wavelength; the best fitting wasobserved by using580nm,600nm,670nm to mix new image; using SAM methodcombined with the3rd-derivative processing can get best distinguish result.In the second test, we found that with the frozen degree getting worse, spectralintensity increased; the best fitting was obtained by using420nm,550nm,650nm tomix a new image.