Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

The prediction of nitrogen content of jujube leaves based on gray neural network optimized by particle swarm

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan

Citation:  2015 ASABE Annual International Meeting  152189669.(doi:10.13031/aim.20152189669)
Authors:   Wei Yang, Minzan Li, Lihua Zheng, Hong Sun
Keywords:   jujube leaves; SNV; UVE; GM-PSOBP

Abstract.A new method for building prediction model of nitrogen content of jujube leaves based on theory of spectrum was proposed. The field experiment was conducted in Shangzhuang jujube orchards of Beijing haidian. The area of jujube orchards is 2 hm2; the type of jujube is Zhanhua 1. The experiment chose 15 jujube trees, and 5 leaves from each tree. Spectra data of jujube leaves were collected during period of budding, branch leaf, flowering and coloring. Spectral instrument used in experiments is ASD, it can get the spectrum in the wavelength range of 350~2500nm, the spectrometer contains a 512 pixel detector of NMOS photodiode for 350-1000nm, and spectral resolution is 1nm. Each simple was tested 5 times and average was used as reflectivity of the sample.(FOSS)KjeltecTM2300 was used for nitrogen content testing of leaves. In order to eliminate the noise, improve the efficiency of data processing, SNV were used to process original spectral data and achieve smooth denoising. Uninformative variables elimination is presented and used to select the useful wavelength from visible near infrared spectral region. It is based on the analysis of b algorithm of PLS regression coefficients, and used to eliminate variables which do not provide information. UVE is more intuitive and easy to understand in the wavelength selection as it set noise and concentration information integrated, and also the running time of the algorithm is faster. At the same time, it uses the full spectrum data of wavelength selection, so the result has high precision. The result indicates that visible and near infrared spectra all contain useful information about jujube leaves. They are distributed in 425~445nm, 515~550nm, 568~605nm, 655~675nm, 744~770nm red light wavebands and 780~860 nm near infrared range. Finally, gray neural network optimized by particle swarm was used to establish prediction model of nitrogen content of jujube leaves. Grey GM(1,1) model has accumulated generating operation characteristics, the operation can effectively reduce the randomness of the data. The complexity of spectral data makes the prediction method must consider the non linear characteristics. The development of chaos theory and the neural network provide the effective way for prediction method. 60 samples were used as training set, 15 samples were used to test the established prediction model. Test results showed that the prediction model of nitrogen content of jujube leaves created by gray neural network optimized by particle swarm had good prediction result.

(Download PDF)    (Export to EndNotes)