Soil near infrared spectroscopy analysis method based on sparse representation and BP neural network technique

Abstract

The invention relates to a soil near infrared spectroscopy analysis method based on sparse representation and BP neural network technique, and overcomes defects that soil analysis methods of the prior art can not meet actual requirements. The method comprises the steps of: obtaining and preprocessing a training sample soil set; constructing a sparse dictionary of training sample soil data, and obtaining feature vectors of training samples; using the feature vectors of training samples as the input of a BP neural network, training network parameters, and constructing a BP neural network classification model; obtaining and preprocessing a test sample soil set; and utilizing the parse dictionary of training sample soil data to construct the feature vectors of test samples, and utilizing a trained BP neural network classification model to complete classification prediction of test sample soil components. The soil near infrared spectroscopy analysis method performs soil near infrared spectroscopy analytical prediction based on sparse representation and BP neural network technique, and improves the prediction precision of near infrared spectroscopy oil main components and model robustness.

Description

technical field [0001] The invention relates to the technical field of data analysis, in particular to a soil near-infrared spectrum analysis method based on sparse representation and BP neural network technology. Background technique [0002] For thousands of years, my country's traditional agriculture has focused on intensive farming. The acquisition and analysis of farmland information is the basis of modern digital agriculture, and soil information is the most important part of farmland information. It can be seen that both traditional agriculture and modern agriculture attach great importance to the collection of farmland soil information. In order to fertilize rationally, in recent years, the Chinese government has invested about 10 billion yuan each year to implement soil testing and formula fertilization projects nationwide. [0003] This project is based on soil testing and fertilizer field experiments, according to the law of crop fertilizer demand, soil fertilize...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the defect that the soil analysis method in the prior art cannot meet the actual needs, and to provide a soil near-infrared spectrum analysis method based on sparse representation and BP neural network technology to solve the above problems

Images