A Rapid Soil Classification Method Based on Visible-Near-Infrared Spectroscopy and Multi-Target Fusion

Abstract

The invention discloses a rapid soil classification method based on visible near infrared spectrum and multi-target fusion. The method comprises the following steps: firstly, rapidly acquiring spectral data of air-dried and ground soil samples in different soil horizons of the soil profile by a visible near infrared spectrometer, and dividing a data set into a modeling set and an independent verification set after smooth denoising; next, establishing a binary-class support vector machine model based on radial basis function for correlation between soil system classification results of the soilprofile and spectra of the soil profile in the modeling set; then, predicting an independent verification set by the binary-class support vector machine model, extracting votes of all soil type binary-class combinations in each horizon from the binary-class support vector machine model, and summing up the total votes of all the horizons in the same profile; finally, taking the soil type with thehighest total vote as the predicted soil type of the soil profile. The soil type of the profile can be predicted quickly and accurately, and a new idea is provided for large-scale and high-precision soil mapping and land resource investigation.

Description

technical field [0001] The invention relates to a rapid soil classification method, in particular to a soil classification method based on the fusion of visible and near-infrared spectra and multiple targets. [0002] technical background [0003] Accurately understanding the spatial distribution of soil resources is of great significance for sustainable soil development and management. For farmers and policy makers, a scientific understanding of the distribution of soil types helps to include precise management of soil resources (e.g. cropping planning and canal design). Traditional soil survey mainly includes field survey, laboratory analysis, expert knowledge and subsequent classification and mapping, so it is often time-consuming and labor-intensive. With the continuous development of precision agriculture, the demand for refined soil type maps in soil resource decision-making is increasing. Traditional soil surveys can no longer meet this requirement. Therefore, we need...

AI Technical Summary

Problems solved by technology

The shortcomings of the above two studies are: 1) averaging the spectral data of the soil surface and bottom layer will weaken the characteristics of the soil spectrum, ignoring the spectral classification information at different levels, resulting in a decrease in classification accuracy; 2) the method of combining fixed soil depth spectra ignores The important premise that the diagnostic layer used in the soil classification system comes from the soil genesis layer, thus reducing the scientific nature of soil classification

[0007] At present, a difficult problem in the combination of soil genetic layer and visible-near-infrared spectroscopy technology is how to deal with the phenomenon that the number of soil genetic layers in different soil profiles is not uniform

Images