Study On Diagnosis Method Of Rice Nitrogen Nutrition Based On Machine Vision And Hyperspectral

Rice is the grain crop with the highest per unit yield in China.In the process of rice planting management,precision nitrogen application can increase the heading percentage,reduce the incidence of disease,and effectively increase the final yield of rice.In order to identify the nitrogen status of rice accurately,a method based on machine vision and hyperspectral techniques was developed for the diagnosis of nitrogen nutrition in rice.In order to classify and diagnose rice nitrogen status based on machine vision,four rice cultivation experiments with different nitrogen application levels were carried out with Liangyoupeijiu super hybrid rice varieties as the test objects.A scanner was used to obtain the leaf and sheath images of the first leaves and the second leaves,the third leaves from crop top in young panicle stage and full heading stage.BPNN and PNN in machine learning algorithm were used to build rice nitrogen nutrition diagnosis model.The results showed that the leaf characteristics of the third leaves in the young panicle stage were the most distinguishable,and it was easy to diagnose and identify nitrogen nutrition,which can be used as aneffective period and location for nitrogen nutrition diagnosis;regarding the recognition effect,BPNN was higher than PNN,and its overall recognition accuracy was90%;the components of six color space,RGB and HSI,can best reflect nitrogen nutrition status.In order to classify and diagnose rice nitrogen status based on hyperspectral,four rice cultivation experiments with different nitrogen application levels were carried out with Zhongjiazao 17 rice varieties as the test objects.Spectral data of the third leaves from crop top at tillering stage were obtained by means of Spectral spectrometer.Spectral pretreatment was carried out by three methods: MSC,SNV and SG.PCA and SPA were used for feature reduction and feature selection of pretreated spectrum.SVM was used to establish rice nitrogen nutrition diagnosis model.The results showed that the MSC-PCA-SVM model was used to diagnose rice nitrogen nutrition in the third leaves from crop top of tillering stage,and the accuracy was 97.5%.The results show that the machine vision and hyperspectral techniques can be applied to rice nitrogen nutrition diagnosis,which provides a new way to accurate and rapid diagnosis and identification of rice nitrogen nutrition precision.