KR102324230B1 - A method and system of measurement for agronomical survey using ASD-Fieldspec - Google Patents

Abstract

The purpose of the present invention is to simultaneously quantify isotope ratios of soil organic carbon(%), total nitrogen(%), stability carbon (δ^13C, ‰), and nitrogen (δ^15N, ‰) through spectral reflection information obtained by a hyperspectral sensor (ASD-Fieldspec) with respect to dried soil. According to the present invention, a difference from the conventional soil spectroscopy technique is that a spectral range used to struct a spectral library is wide (hyperspectral, 350-2500 nm) and absence of hyperspectral-based soil organic carbon-nitrogen estimation technology and the carbon-nitrogen stability isotope ratio can be quantified together.

Description

Spectral-based soil composition prediction method and system, and a program medium implementing it {A method and system of measurement for agronomical survey using ASD-Fieldspec}

The present invention relates to a technology for predicting components in soil based on a non-chemical method, a spectral sensor.

According to Environmental Business International (EBI), an environmental consulting firm in the United States, and Joint Environmental Markets Unit (JEMU), a British government agency, the global soil restoration market is estimated to be $53.4 billion in 2015 from $30 billion in 2005. (about 60 trillion won), and it is on the increasing trend so far.

Soil health can greatly affect groundwater and vegetation vitality due to cumulative damage as well as food production.

On the other hand, in the case of analyzing components such as organic carbon or nitrogen in the soil, or analyzing environmental harmful substances in the soil, it is accompanied by considerable difficulties in the analysis process because most are present in trace amounts in the environment.

As a method for analyzing the same trace amount of substances in soil, chemical analysis methods such as Inductively Coupled Plasma Mass Spectrometry (ICP-MS) or Atomic Absorption Spectrometry (AAS) were used.

The above-described methods require a lot of time, effort, and high cost through sampling, extraction, and complex purification processes representative of the analysis target region. It also requires expensive analytical instruments and skilled manpower. Therefore, there is a need for a technology that can easily and quickly monitor the component substances in the soil in real time while overcoming these shortcomings and ensuring analysis accuracy and rapid analysis speed.

In particular, among soil analysis methods, quantification of soil factors through existing chemical analysis methods requires a lot of time, manpower, and cost depending on precision, number of samples, and analysis items, so spectral reflection information is used as an alternative for efficient soil factor analysis. One predictive technique is attracting attention.

In particular, the main purpose of the prior art (Korean Patent No. 10-1674395, etc.) using the existing soil spectral reflection information is to estimate the content of heavy metals or soil organic matter, which is a soil contamination source, whereas the soil organic carbon and nitrogen content is an element representing soil health, and at the same time, it is an item that requires long-term observation of the amount of storage and change in relation to climate change. In addition, their stability isotope ratio has become a substitute for anthropogenic disturbances such as the origin of soil organic carbon-nitrogen and human activities, but the description of such component analysis has not been implemented.

Korean Patent No. 10-1674395 Korean Patent Registration No. 10-1305425

The present invention has been devised to solve the above needs, and an object of the present invention is to provide a spectroscopic-based environmental monitoring technology that simultaneously predicts the soil organic carbon-nitrogen content and the isotope ratio that can interpret it.

Specifically, the present invention predicts the organic carbon (C) and nitrogen (N) content (%) representing soil health through the reflection information (350-2500 nm) of the soil acquired by the hyperspectral sensor, and at the same time, The purpose of this study is to provide a spectroscopic-based environmental monitoring technology that predicts the stability isotope ratios (δ ¹³ C, δ ^{15 N) together.}

As a means for solving the above problems, in an embodiment of the present invention, as shown in FIGS. 1 and 2 , a first step of acquiring spectral reflection information for a group of soil samples through a spectral sensor; a second step of forming a soil component prediction model through machine learning on the spectral reflection information; a third step of performing cross-validation on the spectral reflection information for the soil sample group; It is possible to provide a spectroscopic-based soil component prediction method, including the step of quantifying the soil component in the soil sample.

In addition, as a system that can implement the above-described soil component prediction method, a soil component quantification module that calculates and verifies a soil component prediction model through machine learning based on the spectral reflection data measured by the spectral sensor module 100 ( 200); and, the soil composition quantification module 200 applies a partial least squares method based on the spectral reflection information on the soil sample to extract the wavelength having a correlation with the predictor variable, , to provide a spectral-based soil component prediction system having a soil component prediction model calculation unit 220 that forms a soil component prediction model.

Furthermore, in an embodiment of the present invention, it is also possible to provide a program medium capable of implementing the above-described soil component prediction method.

According to an embodiment of the present invention, by providing a soil composition prediction model based on spectral data for a soil sample, there is no use of a compound in the analysis process, unlike the conventional method, there is no consumption of the soil sample, and there is no consumption of the soil sample. It has the effect of enabling stable and reliable analysis as it enables repeated measurement of soil samples with non-destructive properties.

In addition, by implementing the soil composition prediction model built with the accumulated hyperspectral data through machine learning, there is no cost when analyzing a new soil sample, and the advantage of quickly deriving results through real-time analysis is realized.

Furthermore, the constructed soil hyperspectral information can be utilized as training data for predicting soil factors other than the four items implementing the prediction model in the present invention, and a hyperspectral-based soil organic carbon, nitrogen, and stability isotope prediction model also realizes the advantage of being able to quickly and accurately analyze the soil health of the target site for environmental restoration and ecosystem evaluation based on more soil samples than before.

Soil carbon-nitrogen and their stability isotope ratio, which is the subject of evaluation of the present invention, are representative factors that determine not only soil health but also anthropogenic disturbance due to human activities. The technology also realizes the advantage of being able to quickly and simultaneously analyze soil information in a wide area compared to the existing evaluation method.

1 is a configuration block diagram of a spectral-based soil component prediction system according to the present invention.
FIG. 2 is a flowchart of a system operating state according to FIG. 1 .
3 is a graph showing the results of an embodiment to which the system according to the present invention is applied.

Advantages and features of the present invention, and methods for achieving them, will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided so that the disclosed subject matter may be thorough and complete, and that the spirit of the present invention may be sufficiently conveyed to those skilled in the art.

The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present application, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It is to be understood that it does not preclude the possibility of the presence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present application. does not

1 is a block diagram of a spectral-based soil component prediction system (hereinafter, referred to as 'the present invention') according to the present invention. FIG. 2 is a flowchart of a system operating state according to FIG. 1 .

1 and 2, the present invention, the spectral sensor module 100 for measuring the spectral reflectivity to the soil sample and the spectral reflection data measured by the spectral sensor module 100 based on the soil through machine learning It may be configured to include a soil composition quantification module 200 for calculating and verifying a component prediction model.

The spectral sensor module 100 is a concept that collectively refers to a device including a spectral sensor, and in particular, as a preferred embodiment of the present invention, a hyperspectral sensor (ASD-Fieldspec) can be applied. Furthermore, the hyperspectral sensor (ASD-Fieldspec) is preferably applied so that the soil sample applied to the present invention can measure the spectral reflectivity of the dried soil sample. That is, the present invention provides soil organic carbon (%), total nitrogen (%), stability carbon (δ ¹³ The gist is to enable simultaneous quantification of C, ‰) and nitrogen (δ ^{15 N, ‰) isotope ratios.}

In addition, the soil composition quantification module 200, the spectral reflection data collection unit 210 for inputting the spectral reflectivity transmitted from the spectral sensor module 100, and the partial least squares based on the spectral reflection information on the soil sample The method (Partial least squares) is applied to extract the wavelength having a correlation with the predictor, and the soil component prediction model calculation unit 220 to form a soil component prediction model, and the soil component prediction model calculation unit 220 It may be configured to include a prediction model verification unit 230 that verifies the calculated prediction model. In addition, based on the accumulated prediction model data, it may be configured to further include a soil composition quantification determination unit 240 to perform and present a composition prediction for a new soil sample.

In particular, the soil component prediction model calculation unit 220 uses the spectral reflection information of the sample provided from the spectral reflection data collection unit 210 for inputting the spectral reflectivity transmitted from the spectral sensor module 100, a prediction model of the soil component. Machine learning is performed to form

The machine learning of the soil component prediction model calculation unit 220 uses the soil organic carbon-nitrogen content of the dry soil from which moisture is removed and the stability isotope ratio thereof as a predictor, a partial least squares method (Partial least squares) This is done by applying

Specifically, in the case of machine learning performed by the soil component prediction model calculation unit 220, the predictive variables are the soil organic carbon-nitrogen content of dry soil from which moisture is removed and their stability isotope ratio, and used to quantify it. In the spectral reflection data used for the spectral reflection data, the spectral variables are very collinear (continuous) in the range of 350-2500 nm, and since the number of predictors is overwhelmingly larger than the number of observations, the partial least squares method is used to correlate with the predictors. A process of extracting a wavelength having a relationship is performed.

In this case, an iterative partial least squares algorithm may be used in the process of performing the partial least squares method. In particular, the 'nonlinear iterative partial least squares algorithm' proceeds with dimensionality reduction that reduces predictors to extract a set of components that explain the maximum correlation between predictors and response variables.

In the soil component prediction model calculation unit 220 according to this embodiment, in order to extract the spectral region for predicting each soil factor, based on the reflection data within the range of 350-2500 nm, the axis of the data having the highest dispersion is found. Enables 'dimension reduction' to be performed. Specifically, in the process of performing this 'dimensional reduction', the first step is to summarize high-dimensional spectral data into lower-dimensional latent variables. The low-dimensional latent variable selected in this process becomes an explanatory variable to be used in the regression analysis in the second step, and ultimately predicts the estimator. Then, in the regression analysis using the partial least squares method, latent variables are extracted from the variance-covariance matrix of the explanatory variable and the response variable. In other words, the partial least squares method extracts latent variables that increase predictive power in consideration of response variables. Among the high-dimensional explanatory variables, the number of dimensions reduced for each soil factor prediction are organic carbon (dimensions: 5), total nitrogen (dimensions: 9), stable carbon and nitrogen isotopes (dimensions: 16). dog).

In this case, regression analysis using the partial least squares method can implement more stable results in predicting the results because dependent variables as well as independent variables are included in the dimensionality reduction process of variables.

In addition, in the case of the prediction model verification unit 230 of the present invention, in order to verify and improve the accuracy of the model for each factor with respect to the soil component factors calculated through the machine learning of the soil component prediction model calculation unit 220 , Allows cross validation to be performed.

Leave-One-Out Cross Validation (LOOCV) was utilized as the cross-validation method of the present invention. Predictive modeling with the spectral information of the sample was repeated as many times as the number of cases. By setting the results of existing chemical analysis methods as a reference for a total of four soil factors (organic carbon, nitrogen, carbon isotopes, nitrogen isotopes), completeness evaluation can be performed through predicted values and correlation analysis.

3 shows the results of an embodiment for illustrating the performance of the function of the present invention.

{Example}

The following experiment was performed to implement the soil composition prediction model according to the present invention.

(1) First, based on a total of 150 dry soil samples, spectral reflection information of a sample that passed through a 2 mm standard sieve was acquired through ASD-Fieldspec. Based on the spectral reflection information, the soil component prediction model was formed through the program of the soil component prediction model calculation unit.

(2) The result value of the element analyzer (Flash EA 1112; Thermo Electron, Waltham, MA, USA) for measuring the organic carbon content (%) of the existing soil was set as a reference (measured SOC).

(3) The graph of FIG. 3 shows the results of analyzing the correlation between the reference value (measured SOC) and the predicted value (predicted SOC) through the soil component prediction model through the program of the soil component prediction model calculation unit. (R ² = 0.83)

As shown in FIG. 3 , it can be confirmed that the visible and near-infrared regions in the wavelength range of 350 - 2500 nm are important regions for predicting soil organic carbon. (VIP, 0.8+) (VIP: Variable importance in projection)

In addition, the coefficient multiplied by the prediction function in each wavelength band (nm) was confirmed through the Coef value within the reflection range.

In addition, the machine learning method (partial least squares method, PLS) of the spectral reflection data and the Leave-One-Out Cross Validation (LOOCV) were performed to increase the accuracy of the development model. The accuracy of the quantified soil factor prediction values ^{was evaluated as organic carbon (R 2} =0.83), total nitrogen (R ² =0.88), δ ¹³ C (R ² =0.99), and δ ¹⁵ N (R ² =0.97).

As described above, the present invention provides soil organic carbon (%), total nitrogen (%), stability carbon ( Simultaneous quantification of δ ¹³ C, ‰) and nitrogen (δ ¹⁵ N, ‰) isotope ratios can be realized.

Furthermore, as a distinguishing feature from the existing soil spectroscopy technology, the spectral range used to construct the spectral library is wide (hyperspectral, 350-2500 nm), and the absence of hyperspectral-based soil organic carbon-nitrogen estimation technology and carbon- It is possible to quantify the nitrogen stability isotope ratio together.

In addition, the functional configuration and execution operation applied to the system for executing the above-described spectral reflection data-based soil component prediction method according to the present invention may be represented by functional block configurations and various processing steps. These functional blocks may be implemented in any number of hardware and/or software configurations that perform specific functions. For example, the present invention provides integrated circuit configurations, such as memory, processing, logic, look-up table, etc., capable of executing various functions by means of the control of one or more microprocessors or other control devices. can be hired

Similar to how the components of the present invention may be implemented as software programming or software elements, the present invention includes various algorithms implemented as data structures, processes, routines, or combinations of other programming constructs, including Python (Python). ), C, C++, Java, assembler, etc. may be implemented in a programming or scripting language. Functional aspects may be implemented in an algorithm running on one or more processors. Further, the present invention may employ prior art techniques for electronic configuration, signal processing, and/or data processing, and the like. Terms such as “module”, “part”, “mechanism”, “element”, “means”, and “configuration” may be used broadly and are not limited to mechanical and physical components. The term may include the meaning of a series of routines of software in association with a processor or the like.

As described above, the technical idea of the present invention has been specifically described in the preferred embodiment, but the preferred embodiment is for the purpose of explanation and not for limitation. As such, those skilled in the art will be able to understand that various embodiments are possible through the combination of the embodiments of the present invention within the scope of the technical spirit of the present invention.

The representative business group in which the present invention will be utilized can be applied to the existing forest ecosystem restoration project and the urban damaged area restoration project as a whole. The existing soil environmental evaluation method is a method of pre-processing samples of the minimum standard collected by field personnel, and then entrusting the analysis of the results to a specialized institution with a chemical analyzer. It has been pointed out that the existing soil evaluation method has a risk of damaging the site and has low efficiency in terms of cost, manpower, and time required. In addition, additional analysis of high cost has been required to interpret the cause of the analysis result. The spectral information-based soil carbon-nitrogen and stability isotope ratio evaluation method used in the present invention has the advantage of not using a compound, minimizing damage to the target site, and not destroying the acquired soil sample (repeat measurement is possible). In addition, the analysis time can also be significantly reduced, and due to the characteristics of the machine learning method, the prediction accuracy increases as the acquired spectral library is accumulated, and it can be used as basic data for predicting separate soil factors in the future.

100: spectral sensor module
200: Soil composition quantification module
210: spectral reflection data collection unit
220: soil composition prediction model calculation unit
230: prediction model verification unit
240: soil composition quantification determining unit

Claims (8)

Step 1 of acquiring spectral reflectivity information for a sample filtered through a standard sieve of a soil sample dried through a hyperspectral sensor;
By applying the partial least squares method using the soil organic carbon-nitrogen content, which is spectral reflection information for the dry soil sample from which moisture has been removed, and their stability isotope ratio as a predictor variable, the wavelength having a correlation with the predictor variable is extracted, and the soil Step 2 of forming a component prediction model;
a third step of performing cross-validation on the soil component prediction model for the soil sample;
Among the soil components in the soil sample, soil organic carbon (%), total nitrogen (%), stability carbon (δ ¹³ C, ‰), nitrogen (δ ¹⁵ N, ‰) isotope ratios simultaneously quantification;
containing,
Spectral-based Soil Composition Prediction Method.
The method according to claim 1,
The first step is
The hyperspectral sensor (Hyperspectral Sensor) applies spectral reflectance information in the range of spectral reflectivity 400-2200nm,
Spectral-based Soil Composition Prediction Method.
delete 3. The method according to claim 2,
Step 3 above applies Leave-One-Out Cross Validation (LOOCV) as a cross-validation method,
a) Selecting a soil sample to be verified from among n soil sample groups as a reference sample
b) performing predictive modeling with the spectral information of (n-1) soil sample groups excluding the reference sample in all cases;
c) analyzing predicted values and correlations with respect to reference result values for organic carbon, nitrogen, carbon isotopes, and nitrogen isotopes;
Containing, spectroscopy-based soil component prediction method.
a spectral sensor module 100 for obtaining spectral reflectance information for a sample filtered through a standard sieve of a dried soil sample; and
Prediction by applying the partial least squares method using the soil organic carbon-nitrogen content and their stability isotope ratio as a predictor variable based on the spectral reflection data of the dry soil sample from which moisture has been removed measured by the spectral sensor module 100 It includes; a soil composition quantification module 200 that extracts a wavelength having a correlation with a variable, forms a soil composition prediction model, and verifies it;
The soil composition quantification module 200,
Based on the spectral reflection information on the soil sample, the partial least squares method is applied to extract the wavelength having a correlation with the predictor variable, and the soil component prediction model calculation unit 220 forms a soil component prediction model. ; a spectral reflection data collection unit 210 for inputting the spectral reflectivity transmitted from the spectral sensor module 100; and a prediction model verification unit 230 for verifying the prediction model calculated by the soil component prediction model calculation unit 220;
Simultaneous quantification of isotope ratios of soil organic carbon (%), total nitrogen (%), stability carbon (δ ¹³ C, ‰), nitrogen (δ ¹⁵ N, ‰) among soil components in soil samples doing,
Spectral-based soil composition prediction system.
delete 6. The method of claim 5,
The predictive model verification unit 230,
By applying Leave-One-Out Cross Validation (LOOCV),
For organic carbon, nitrogen, carbon isotopes, and nitrogen isotopes, the reference result value is analyzed and verified with the predicted value and the correlation.
Spectral-based soil composition prediction system.
A recording medium containing a program for performing the soil component prediction method according to claim 1.

Images