Research On Nitrogen Diagnosis Model Of Rubber Tree Leaf Based On Hyperspectral Spatial-spectral Information

Nitrogen is an important nutrient element to guide the precise fertilization of rubber trees.The nitrogen content detection technology is directly related to the yield of rubber.The traditional plant detection methods are appearance feature detection method and chemical analysis method.The appearance feature detection method is subjectively affected by the operator and has the characteristics of low accuracy and strong hysteresis.The detection accuracy based on the chemical method is high,but the cost is high and the efficiency is low.The data and physical and chemical analysis labels establish a nutrition content detection model,which is an effective nitrogen detection method.However,the current research alternative on rubber trees is that the spatial information of rubber tree hyperspectral data is not highly used,and based on the hyperspectral data,the spatial spectral characteristics of the hyperspectral data extracted in this paper are used to study the accuracy of the spectral diagnosis model of the rubber tree leaf spectral collection area to the accuracy of the spectral diagnosis model.Impact,ok.A nitrogen sensitive spatial region of rubber tree leaves was established and a nitrogen diagnosis model was established.1. Propose a quantitative analysis model for nitrogen content detection of rubber trees based on a certain high spectrum.The average spectrum of random points,the average spectrum of the region of interest,and the average spectral data of the whole leaf area were used as controls.SPA and CARS were used to extract the nitrogen content sensitive wavelength information of the spectral data to establish the MLR model.The corrected Euclidean distance correction spectrum and the maximum variance projection projection spectrum are more suitable than the traditional average spectrum to establish a quantitative analysis and detection model for the nitrogen content of rubber tree leaves.The determination coefficients of the models are R~2=0.946 and R~2=0.945,respectively.2. Propose a diagnosis strategy of hypersensitive data nitrogen sensitive spatial region based on ev PCA-Kmeans.Use evPCA to amplify the characteristics of variables to differentiate the weights between points;Kmeans clusters the differentiated spatial information with K(K=2,3,4,5,6,7)to obtain hyperspectral The clustered average spectrum of the data.The incident cluster average spectrum and nitrogen physicochemical analysis results were used to establish 21 PLSR models.Finally,the main leaf veins of rubber tree leaves were initialized,and the pure mesophyll area of the lateral leaf veins was the most suitable for establishing a nitrogen quantitative analysis model of rubber trees.And it is clear that the sampling area of the homogeneous region is positively correlated with the accuracy of the model.3. Propose a quantitative detection model for nitrogen content of rubber tree leaves based on the nitrogen-sensitive spatial average spectrum.In order to further improve the accuracy of the quantitative analysis model of the nitrogen content of rubber trees,this study separately used SNV,MSC,first and second derivative pretreatment methods,combined with SPA,CARS by comparing the modeling results of different models,and finally determined the When K is 2,the SNV-SPA-PLSR model established by the second cluster average spectrum has the highest accuracy,and is most suitable for establishing a nitrogen diagnosis model of rubber tree leaves,with R~2=0.951 and RMSE=0.1497.