Study On Hyperspectral Image Analysis Of Cucumber Leaves

The chlorophyll in the leaf of cucumber,the distribution of nutrient all kind of elements and their content are all the Important symbol for cucumber growth.General method for the determination of chlorophyll is by chemical method.In recent years,people have been trying to use hyperspectral images to detect chlorophyll,nitrogen and other nutrients in crops at home and abroad.We can get both image information and spectral information from hyperspectral images.So,this method is a new method for dynamic monitoring of greenhouse plants.In this thesis,the cucumber leaves were collected by high spectrometer.The collected hyperspectral images are imported directly into the software ENVI,and get the ROI.Then the obtained data are preprocessed,dimensionality reduction and classification.Finally,the chlorophyll estimation and analysis of cucumber leaves were carried out.This article mainly studies the following contents:(1)The average spectral data of cucumber leaves extracted were preprocessed.The pretreatment methods used in standardized processing(SNV),Multiple scattering correction(MSC)and normalization,which can well eliminate some factors that affect spectral information.(2)There are three main algorithms used: Principal component analysis(PCA),Local linear embedding(LLE)and Continuous projection(SPA)to remove redundant data in the experiment.(3)SNV and KNN were used to classify the selected characteristic bands.A total of450 samples of cucumber leaves from three different varieties were classified.(4)Using the above data to establish BP neural network model,PLS model.Therefore,the chlorophyll content of cucumber leaves in three different varieties(450samples)can be predicted.It is feasible to identify cucumber species by hyperspectral imaging.large number of data experiments MCS pretreatment method and SPA dimension reduction method have advantages in predicting chlorophyll content and classification of cucumber varieties.During the experiment of cucumber leaf classification,we compared the time and effect of these classification in detail,which proved that SVM has a good effect on accurately judging cucumber classification.Compared with PLS,BP neural network modeling is better for predicting chlorophyll effect.