CN104596957A - Estimation method for content of copper in soil on basis of visible-light near-infrared spectrum technology - Google Patents

Abstract

The invention relates to an estimation method for the content of copper in soil on the basis of visible-light near-infrared spectrum technology. The estimation method is realized by the following six steps: (1) collecting a soil sample; (2) determining a visible-light near-infrared spectrum; (3) pretreating the spectrum; (4) determining the reference value of the content of copper in soil; (5) establishing an estimation model; and (6) estimating the content of copper in an unknown soil sample. The estimation model between the visible-light neared-infrared reflectivity spectrum and the content of the copper is established by utilizing a wavelet neural network method on the basis of the visible-light near-infrared spectrum technology, so that the visible-light neared-infrared reflectivity spectrum of the unknown soil sample is substituted into the estimation model and further the content of the copper in the unknown soil sample is determined. The estimation method has the advantages that the determination can be performed without direct contact with the sample, and is complete non-destructive measurement, the operation process and the calculation method for the content of the copper in the soil are simple, the determination speed is greatly enhanced, no other chemical reagents need to be added and the like, so that the estimation method is environmental-friendly and pollution-free.

Description

Based on the soil copper content evaluation method of visible ray near-infrared spectrum technique

Technical field

The present invention relates to process and the Soil Testing of spectroscopic data, be specifically related to a kind of soil copper content evaluation method based on visible ray near-infrared spectrum technique.

Background technology

Copper (Cu) is oxidase or the indispensable ingredient of the specific construction package of cell in plant, and the rising of copper content can affect growth and the aging course of chromatin in plant (photosynthesis organ) strongly.The minimizing of Copper In The Soil content, photosynthesis and the metabolic process of plant can be affected, and the rising of Copper In The Soil content, the copper that plant absorption is excessive can be caused, make the membranous system subject to damage of plant cell membrane and various kinds of cell device, what affect plant crosses conjunction mechanism.Thus, estimate that the content in soil seems particularly important accurately and timely.

Application number a kind of method measuring metallic element mercury in soil that has been the disclosure of the invention of 201310112158.3, described determination method comprises the following steps: (1) pre-service: get unknown soil sample 1g as sample, this sample to be tested is soaked in digestion solution 1, is then heated between 80 DEG C-100 DEG C a certain thermostatic 30 minutes; This digestion solution 1 is HNO ₃with the mixed solution 100mL of HCL; In this mixed solution, add digestion solution 2 again, to be then heated between 120 DEG C-140 DEG C a certain thermostatic 30 minutes, this digestion solution 2 is HNO ₃with HClO ₄mixed solution 100mL; Learnt from else's experience after being cooled to room temperature the test solution 10ml processed, for subsequent use as mensuration test solution to 50ml with water constant volume; (2) drafting of working curve needed for mercury metal element determination: the standard soil sample pipetting one group of known mercury content, adopts the identical digestion procedure of step (1) pre-service to carry out clearing up and obtain one group of uniform standard soil sample solution; Then be incorporated in inductively coupled plasma atomic emission spectrometer and measure, draw metric works curve; Related coefficient>=0.999 of this working curve, measures the requirement of mercury metal with satisfied unknown soil sample; (3) mercury metal element determination in soil: the test solution to be measured in step (1) is introduced inductively coupled plasma atomic emission spectrometer, according to the content of mercury metal element in the standard working curve determination sample to be tested that step (2) obtains.

The assay method of whole vanadium in application number a kind of soil that has been the disclosure of the invention of 201210449109.4, described determination method comprises the following steps: (1) unknown soil sample sample pretreatment: first take pedotheque and be put in conical flask, add chloroazotic acid, on the electric furnace having ventilating kitchen, heating is smoldered near dry, perchloric acid is added after taking off pedotheque cooling, heat again and emit white cigarette, eliminate nitric acid takes off after cooling adds diluted hydrochloric acid dissolution after pedotheque is closely thick, filter constant volume in volumetric flask with distilled water, shake up to be measuredly containing the vanadium soil liquid standing; (2) in soil, whole vanadium measures: get containing the vanadium soil liquid in color comparison tube, add developer, superoxol and hydrochloric acid solution successively; Shake up placement after leaving standstill after adding screening agent successively again, measure the absorbance of colored complex with cuvette in 575nm wavelength place, and record measurement result, calculating to be analyzed; (3) result calculates.

Visible ray near-infrared spectrum technique refers to that the careful abundant spectral signature utilizing visible ray and near infrared spectrum wave band to comprise describes, and mainly differentiates material according to the spectrum of material and determines its chemical composition and the method for relative content.

The soil organism, iron and manganese oxides and clay mineral have absorption or hosting relation to heavy metal-polluted soil, these components influence spectral reflectance spectrum forms in soil and the size of reflectivity, in spectral reflectance spectrum, show its specific spectral absorption characteristics, this is just for utilizing the copper content in soil visible ray near-infrared spectral reflectance quantitative estimation soil to provide theoretical foundation simultaneously.

Said method testing procedure is many, complicated operation, reagent dosage are large, sense cycle is long, and error is large.

Therefore, need to provide a kind of method can estimating Copper In The Soil content quickly and accurately.

At present, the visible ray near infrared Fast Measurement being directed to soil copper content in mine tailing region has no report.Visible ray near-infrared spectral analysis technology has broad application prospects in heavy metal content in soil estimation.

Summary of the invention

For the deficiencies in the prior art, the present invention proposes that finding speed is fast, measurement result accurately and reliably, a kind of soil copper content evaluation method based on visible ray near-infrared spectrum technique of nondestructive measurement can be realized.

To achieve these goals, the technical solution adopted in the present invention is made up of following steps:

(1) collection of soil sample

In mine tailing Regional field, gather upper soll layer sample as with reference to soil, be divided into n part, n >=10, remove impurity;

(2) visible ray near infrared ray

Obtain every part of reflectance spectrum data with reference to soil with spectrometer, spectral range is 350 ~ 2500nm;

(3) pre-service of spectrum

Carry out envelope removal to reference to the visible ray near infrared reflectivity spectrum in the reflectance spectrum of soil, obtain the visible ray near infrared reflectivity spectrum of the reference soil after envelope is removed;

(4) mensuration of soil copper content reference value

Adopt atomic absorption spectrophotometry to measure with reference to the total copper content in soil every part, obtain every part of copper content reference value with reference to soil;

(5) foundation of appraising model

Employing wavelet neural network sets up visible ray near infrared reflectivity spectrum and with reference to the appraising model between soil copper content reference value, wherein wavelet neural network formula is as follows:

$y_j=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ji}}{\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{\mathrm{Ni}})+\stackrel{\‾}{y_j},i=\mathrm{1,2},...,M;j=\mathrm{1,2},...,N_0---\left(1\right)$

In formula, ψ _i(x ₁, x ₂, x ₃..., x _ni) be multidimensional wavelet function,

${\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{N_i})=\underset{k=1}{\overset{N_i}{\mathrm{\Π}}}\mathrm{\ψ}\left(\frac{x_k{w}_{\mathrm{ik}}-t_{\mathrm{ik}}}{{\mathrm{\λ}}_{\mathrm{ik}}}\right),k=\mathrm{1,2},...,N;i=\mathrm{1,2,3},...,M---\left(2\right)$

represent Morlet wavelet function; (3)

N _irepresent the number of input layer, w _ikrepresent network connection weight, t _ikrepresent contraction-expansion factor, λ _ikrepresent shift factor, k represents the input number of plies, and i represents the hidden layer number of plies;

W _jithe wavelet neural unit of connection i-th hidden layer and the output weight coefficient of a jth output node, a deviate is needed to process Non-zero Mean function;

(6) unknown sample soil copper content estimation

Gather the reflectance spectrum of unknown soil sample, after the pre-service of step (3), gained spectrum parameter is input in the appraising model of step (5), the copper content of unknown soil sample can be estimated.

Above-mentionedly state in step (3), preprocess method can be made up of following steps:

(3.1) the visible ray near infrared reflectivity spectrum within the scope of every part that step (2) the is gathered reservation of the reflectance spectrum with reference to soil 350 ~ 1230nm, rejects all the other spectral bands;

(3.2) do homogenization smoothing processing to the visible ray near infrared reflectivity spectrum that (3.1) retain, every part that is gathered by spectrometer 10 curves of spectrum with reference to soil are averaged as the curve of spectrum corresponding to this reference soil;

(3.2) calculate envelope according to formula (1), envelope removal is carried out to the spectrum with reference to soil;

$\mathrm{CR}=\frac{{\mathrm{\ρ}}_{\mathrm{\λ}}}{{\mathrm{\ρ}}_{\mathrm{c\λ}}}$

Wherein: λ represents specific wavelength, ρ _λrepresent the spectral reflectivity of specific wavelength, ρ _{c λ}represent the reflectivity envelope be associated.

The concrete grammar of above-mentioned steps (5) is made up of following steps:

(5.1) netinit

The contraction-expansion factor t of random initializtion wavelet neural network _ik, shift factor λ _ikwith network connection weight w _ji, e-learning speed η is set;

(5.2) sample classification

Modeling is carried out with reference to the visible ray near infrared reflectivity spectrum of pedotheque and the copper content reference value data set of step (4) gained with step (3) gained, adopt cross validation method optimization wavelet neural network model, the scope n of setting cross validation, input amendment data are divided into n group, choose arbitrarily wherein n-1 group subset data and, as training set, remain 1 group as checking collection;

(5.3) estimation exports

Training set is inputted wavelet neural network, wherein input layer is the visible ray near infrared reflectivity spectrum with reference to pedotheque, output layer is the copper content reference value with reference to soil, utilizes the error between the network estimated value of wavelet neural network computing reference soil copper content and expectation value;

(5.4) modified weight

According to the adjustment of step (5.3) the errors number of plies of hidden layer, the interstitial content of each layer and wavelet function parameter, what make wavelet neural network estimated value approaches expectation value, the training error of wavelet-neural network model reaches minimum, obtains optimum appraising model.

Soil copper content evaluation method of the present invention is based on visible ray near-infrared spectrum technique, wavelet neural network method is utilized to set up appraising model between the visible ray near infrared reflectivity spectrum of soil and copper content, thus the visible ray near infrared reflectivity spectrum of unknown soil sample is substituted in appraising model, thus determine the copper content of unknown soil sample, compared with prior art, the present invention has following beneficial effect:

(1) simple, the inventive method utilizes spectroscopic data to realize the quantitative estimation of mine tailing regional soil copper content, and this mensuration, without the need to directly contacting with sample, be nondestructive measurement completely, and the computing method of operating process and soil copper content is simple.

(2) quick, the method that the present invention proposes is based on wavelet neural network, the computer program of assay method is less than 10 minutes working time, and compared with the loaded down with trivial details detection operating process needing at least 1 hour with conventional leaf chlorophyll contents detection method, finding speed is accelerated greatly.

(3) environmental protection, the soil copper content evaluation method that the present invention proposes is pollution-free.

Accompanying drawing explanation

Fig. 1 is the visible ray near infrared reflectivity curve of spectrum with reference to soil sample in embodiment.

Fig. 2 is the structure of wavelet neural network.

Fig. 3 is the comparison diagram between the copper content estimated value of unknown soil sample and soil copper content reference value.

Embodiment

Below in conjunction with accompanying drawing, case study on implementation of the present invention is elaborated: the implementation case is implemented under premised on technical solution of the present invention; give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

Embodiment 1

Now using Jin Dui city, Hua County, the Shaanxi molybdenum tailing pedotheque that is study area as detected object, carry out soil copper content estimation, specific implementation step is as follows:

(1) collection of pedotheque

With Jin Dui city, Hua County, Shaanxi molybdenum tailing for study area, according to mining area exploitation situation and ground flora coverage condition, whole mining area is divided into 10 communities, to layout method by plum blossom in each community, random acquisition 6 pedotheques, gather soil sample 60 altogether, gather upper soll layer sample (0 ~ 20cm).The soil of each sampling point loads in different sample boxes, and outside sample sack, fill in sample label.

After pedotheque takes back laboratory, foreign matter such as removing fragment, cobble etc., 2mm nylon mesh is crossed after air-dry, being divided equally by soil sample after screening is 2 parts, wherein 42 parts as with reference to soil be used for soil spectrum measure and copper content analysis, other 18 parts as test sample book for estimating the reliability of rear checking evaluation method of the present invention.

(2) visible ray near-infrared spectral measurement

Utilize the reflectance spectrum data of ASD FieldSpec3 field spectroradiometer witness mark soil, measurement spectral range is 350 ~ 2500nm, Measuring Time is 10:30 ~ 12:00, at outdoor natural light according under condition, select 8 ° of field angle probes, probe is 1.35m to the surface distance with reference to soil sample, measures 1m ²soil spectrum in scope; This spectrometer needs with blank school zero before use, and before each sample of measurement, measurement dark current and spectrometer obtain 10 measure spectrum with reference to blank is each continuously with reference to soil sample.

(3) pre-service of soil spectrum

For eliminating the system noise that ASD FieldSpec3 field spectroradiometer produces when witness mark soil spectrum, the reflectance spectrum of the reference soil of step (2) being surveyed is rejected, retain the visible ray near infrared reflectivity spectrum within the scope of 350 ~ 1230nm, reject all the other spectral bands; 10 visible ray near infrared reflectivity spectrum of the reference soil measured by each sampled point more afterwards do homogenization smoothing processing, get the curve of spectrum of its mean value as this reference soil sample, as shown in Figure 1; Again by calculating the reflectivity (ρ of specific wavelength (λ) _λ) with the reflectivity envelope (ρ be associated _{c λ}) and obtain Envelope Equations, the envelope completed with reference to the visible ray near infrared reflectivity spectrum of soil is removed; Envelope Equations is:

$\mathrm{CR}=\frac{{\mathrm{\ρ}}_{\mathrm{\λ}}}{{\mathrm{\ρ}}_{\mathrm{c\λ}}}$

In formula: λ represents the specific wavelength with reference to soil, ρ _λrepresent the spectral reflectivity of specific wavelength, ρ _{c λ}represent the reflectivity envelope be associated.

(4) mensuration of soil copper content

With muller, the reference soil sample of screening is pulverized, adopt atomic absorption spectrophotometry (GB/T17138-1997) to measure its total copper content, obtain the copper content reference value with reference to soil.

(5) foundation of appraising model

Employing wavelet neural network sets up the visible ray near infrared spectrum after envelope removal and with reference to the appraising model between soil copper content reference value, concrete grammar is:

(5.1) netinit

The contraction-expansion factor t of random initializtion wavelet neural network _ik, shift factor λ _ikwith network connection weight w _ji, e-learning speed η is set; Wavelet neural network is a new important branch in neural network research, and the thought in conjunction with wavelet theory and artificial neural network is formed.Wavelet neural network is a kind of based on BP neural network topology mechanism, using the transport function of small echo odd function as hidden layer node, and the neural network of error back propagation while signal propagated forward.The structure of wavelet neural network as shown in Figure 2.

Wavelet neural network formula is as follows:

$y_j=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ji}}{\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{\mathrm{Ni}})+\stackrel{\‾}{y_j},i=\mathrm{1,2},...,M;j=\mathrm{1,2},...,N_0---\left(1\right)$

In formula, ψ _i(x ₁, x ₂, x ₃..., x _ni) be multidimensional wavelet function,

${\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{N_i})=\underset{k=1}{\overset{N_i}{\mathrm{\Π}}}\mathrm{\ψ}\left(\frac{x_k{w}_{\mathrm{ik}}-t_{\mathrm{ik}}}{{\mathrm{\λ}}_{\mathrm{ik}}}\right),k=\mathrm{1,2},...,N;i=\mathrm{1,2,3},...,M---\left(2\right)$

represent Morlet wavelet function; (3)

N _irepresent the number of input layer, w _ikrepresent network connection weight, t _ikrepresent contraction-expansion factor, λ _ikrepresent shift factor, k represents the input number of plies, and i represents the hidden layer number of plies;

W _jithe wavelet neural unit of connection i-th hidden layer and the output weight coefficient of a jth output node, a deviate is needed to process Non-zero Mean function.

(5.2) sample classification

Based on the reference soil visible ray near infrared reflectivity spectrum after 42 pre-service and 42 reference soil sample copper content reference values, adopt cross validation method optimization wavelet neural network model, the scope of setting cross validation, input amendment data are divided into 42 groups, choose arbitrarily wherein 41 groups of subset data as training set, remain 1 group as checking collection, set up 41 training datasets (train data set) and 1 verification msg (validation data set).

(5.3) estimation exports

Training set is inputted wavelet neural network, wherein input layer is the visible ray near infrared reflectivity spectrum with reference to pedotheque, output layer is the copper content reference value with reference to soil, utilizes the error between the network estimated value of wavelet neural network computing reference soil copper content and expectation value.

(5.4) modified weight

According to the number of plies of the error transfer factor hidden layer of step (5.3) gained, the interstitial content of each layer and wavelet function parameter, wavelet neural network estimated value is made to approach expectation value, thus the training error of wavelet-neural network model reaches minimum, then obtain optimum wavelet-neural network model.

(6) unknown sample soil copper content estimation

Gather unknown soil sample, its reflectance spectrum is measured according to the method for step (2), and the visible ray near infrared reflectivity spectrum of its correspondence is obtained according to the preprocess method of step (3), gained visible ray near infrared reflectivity spectrum parameter is input in the appraising model of step (5), the copper content corresponding to unknown soil sample can be estimated.

In order to this verifies the beneficial effect of evaluation method of the present invention, the method of above-described embodiment is utilized to carry out analyses and prediction to 18 of gained in the same area testing soil sample spectrum, obtain the copper content estimated value of soil sample, contrast with the soil copper content utilizing atomic absorption spectrophotometry to measure, evaluation index is coefficient R ²with square error MSE, formula is as follows:

$R^2=1-\frac{\underset{i=1}{\overset{T}{\mathrm{\Σ}}}{(y_i-{\hat{y}}_i)}^2}{\underset{i=1}{\overset{T}{\mathrm{\Σ}}}{(y_i-\stackrel{\‾}{y})}^2}$

$\mathrm{MSE}=\frac{\underset{i=1}{\overset{T}{\mathrm{\Σ}}}{(y_i-y)}^2}{T}$

Wherein, T is the quantity of test sample book, i-th sample actual measured value, the mean value of test sample book;

Result as shown in Figure 3, related coefficient between the estimated value of method of the present invention and atomic absorption spectrophotometry measured value is 0.7482, square error is 84.2724, this shows, the estimation result of the evaluation method that the present invention proposes is high correlation with the soil copper content value utilizing atomic absorption spectrophotometry to measure, and may be used for the quantitative estimation carrying out mine tailing regional soil copper content fast.

Claims (3)

1. based on a soil copper content evaluation method for visible ray near-infrared spectrum technique, by following steps:

(1) collection of soil sample

In mine tailing Regional field, gather upper soll layer sample as with reference to soil, be divided into n part, n >=10, remove impurity;

(2) visible ray near infrared ray

Obtain every part of reflectance spectrum data with reference to soil with spectrometer, spectral range is 350 ~ 2500nm;

(3) pre-service of spectrum

Carry out envelope removal to reference to the visible ray near infrared reflectivity spectrum in the reflectance spectrum of soil, obtain the visible ray near infrared reflectivity spectrum of the reference soil after envelope is removed;

(4) mensuration of soil copper content reference value

Adopt atomic absorption spectrophotometry to measure with reference to the total copper content in soil every part, obtain every part of copper content reference value with reference to soil;

(5) foundation of appraising model

Employing wavelet neural network sets up the visible ray near infrared reflectivity spectrum after envelope removal and with reference to the appraising model between soil copper content reference value, wherein wavelet neural network formula is as follows:

$y_j=\underset{i=1}{\overset{M}{\mathrm{\Σ}}}{w}_{\mathrm{ji}}{\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{\mathrm{Ni}})+\stackrel{\‾}{y_j}i=\mathrm{1,2},...,M;j=\mathrm{1,2},...,N_0---\left(1\right)$

In formula, ψ _i(x ₁, x ₂, x ₃..., x _ni) be multidimensional wavelet function,

${\mathrm{\ψ}}_i(x_1,x_2,x_3,...,x_{N_i})=\underset{k=1}{\overset{N_i}{\mathrm{\Π}}}\mathrm{\ψ}\left(\frac{x_k{w}_{\mathrm{ik}}-t_{\mathrm{ik}}}{{\mathrm{\λ}}_{\mathrm{ik}}}\right)k=\mathrm{1,2},...,N;i=\mathrm{1,2,3},...,M---\left(2\right)$

represent Morlet wavelet function; (3)

N _irepresent the number of input layer, w _ikrepresent network connection weight, t _ikrepresent contraction-expansion factor, λ _ikrepresent shift factor, k represents the input number of plies, and i represents the hidden layer number of plies;

W _jithe wavelet neural unit of connection i-th hidden layer and the output weight coefficient of a jth output node, a deviate is needed to process Non-zero Mean function;

(6) the copper content estimation of unknown sample soil

Measure the reflectance spectrum of unknown sample soil, after the pre-service of step (3), gained spectrum parameter is input in the appraising model of step (5), the copper content of unknown soil sample can be estimated.

2. the soil copper content evaluation method based on visible ray near-infrared spectrum technique according to claim 1, is characterized in that: in described step (3), preprocess method is made up of following steps:

(3.1) the visible ray near infrared reflectivity spectrum within the scope of every part that step (2) the is gathered reservation of the reflectance spectrum with reference to soil 350 ~ 1230nm, rejects all the other spectral bands;

(3.2) do homogenization smoothing processing to the visible ray near infrared reflectivity spectrum that (3.1) retain, every part that is gathered by spectrometer 10 visible ray near infrared reflectivity curves of spectrum with reference to soil are averaged as the curve of spectrum corresponding to this reference soil;

(3.2) calculate envelope according to formula (1), envelope removal is carried out to the visible ray near infrared reflectivity spectrum with reference to soil;

$\mathrm{CR}=\frac{{\mathrm{\ρ}}_{\mathrm{\λ}}}{{\mathrm{\ρ}}_{\mathrm{c\λ}}}$

Wherein: λ represents the specific wavelength with reference to soil, ρ _λrepresent the spectral reflectivity of specific wavelength, ρ _{c λ}represent the reflectivity envelope be associated.

3. the soil copper content evaluation method based on visible ray near-infrared spectrum technique according to claim 1, is characterized in that: the concrete grammar of described step (5) is made up of following steps:

(5.1) netinit

The contraction-expansion factor t of random initializtion wavelet neural network _ik, shift factor λ _ikwith network connection weight w _ji, e-learning speed η is set;

(5.2) sample classification

Modeling is carried out with reference to the visible ray near infrared reflectivity spectrum of pedotheque and the copper content reference value data set of step (4) gained with step (3) gained, adopt cross validation method optimization wavelet neural network model, the scope n of setting cross validation, input amendment data are divided into n group, choose arbitrarily wherein n-1 group subset data and, as training set, remain 1 group as checking collection;

(5.3) estimation exports

Training set is inputted wavelet neural network, wherein input layer is the visible ray near infrared reflectivity spectrum with reference to pedotheque, output layer is the copper content reference value with reference to soil, utilizes the error between the network estimated value of wavelet neural network computing reference soil copper content and expectation value;

(5.4) modified weight

According to the adjustment of step (5.3) the errors number of plies of hidden layer, the interstitial content of each layer and wavelet function parameter, what make wavelet neural network estimated value approaches expectation value, the training error of wavelet-neural network model reaches minimum, obtains optimum appraising model.