CN114357891A - Hyperspectral remote sensing quantitative inversion method for soil cadmium element content - Google Patents
Hyperspectral remote sensing quantitative inversion method for soil cadmium element content Download PDFInfo
- Publication number
- CN114357891A CN114357891A CN202210024419.5A CN202210024419A CN114357891A CN 114357891 A CN114357891 A CN 114357891A CN 202210024419 A CN202210024419 A CN 202210024419A CN 114357891 A CN114357891 A CN 114357891A
- Authority
- CN
- China
- Prior art keywords
- soil
- sample
- cadmium element
- individual
- remote sensing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 229910052793 cadmium Inorganic materials 0.000 title claims abstract description 108
- BDOSMKKIYDKNTQ-UHFFFAOYSA-N cadmium atom Chemical compound [Cd] BDOSMKKIYDKNTQ-UHFFFAOYSA-N 0.000 title claims abstract description 108
- 239000002689 soil Substances 0.000 title claims abstract description 78
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000001228 spectrum Methods 0.000 claims abstract description 51
- 238000011160 research Methods 0.000 claims abstract description 40
- 241000282461 Canis lupus Species 0.000 claims abstract description 29
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 23
- 238000013528 artificial neural network Methods 0.000 claims abstract description 22
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 12
- 238000005516 engineering process Methods 0.000 claims abstract description 10
- 230000009466 transformation Effects 0.000 claims abstract description 10
- 238000010987 Kennard-Stone algorithm Methods 0.000 claims abstract description 5
- 238000005070 sampling Methods 0.000 claims description 22
- 238000005259 measurement Methods 0.000 claims description 16
- 238000010521 absorption reaction Methods 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000000694 effects Effects 0.000 claims description 8
- 239000013307 optical fiber Substances 0.000 claims description 6
- 238000002310 reflectometry Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 5
- 238000009616 inductively coupled plasma Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- GNFTZDOKVXKIBK-UHFFFAOYSA-N 3-(2-methoxyethoxy)benzohydrazide Chemical compound COCCOC1=CC=CC(C(=O)NN)=C1 GNFTZDOKVXKIBK-UHFFFAOYSA-N 0.000 claims description 3
- NYQDCVLCJXRDSK-UHFFFAOYSA-N Bromofos Chemical compound COP(=S)(OC)OC1=CC(Cl)=C(Br)C=C1Cl NYQDCVLCJXRDSK-UHFFFAOYSA-N 0.000 claims description 3
- FGUUSXIOTUKUDN-IBGZPJMESA-N C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 Chemical compound C1(=CC=CC=C1)N1C2=C(NC([C@H](C1)NC=1OC(=NN=1)C1=CC=CC=C1)=O)C=CC=C2 FGUUSXIOTUKUDN-IBGZPJMESA-N 0.000 claims description 3
- 241000196324 Embryophyta Species 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 230000002925 chemical effect Effects 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 229910052736 halogen Inorganic materials 0.000 claims description 3
- 150000002367 halogens Chemical class 0.000 claims description 3
- 238000007790 scraping Methods 0.000 claims description 3
- 239000004575 stone Substances 0.000 claims description 3
- 238000007605 air drying Methods 0.000 claims description 2
- 238000000227 grinding Methods 0.000 claims description 2
- 239000012535 impurity Substances 0.000 claims description 2
- 238000007873 sieving Methods 0.000 claims description 2
- 238000013098 chemical test method Methods 0.000 abstract 1
- 230000004069 differentiation Effects 0.000 abstract 1
- 230000003595 spectral effect Effects 0.000 description 23
- 238000009826 distribution Methods 0.000 description 13
- 239000000126 substance Substances 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000012795 verification Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000000985 reflectance spectrum Methods 0.000 description 3
- 239000002699 waste material Substances 0.000 description 3
- 230000002265 prevention Effects 0.000 description 2
- 239000011573 trace mineral Substances 0.000 description 2
- 235000013619 trace mineral Nutrition 0.000 description 2
- 239000002351 wastewater Substances 0.000 description 2
- 241000717544 Aconitum lycoctonum subsp. vulparia Species 0.000 description 1
- 238000003723 Smelting Methods 0.000 description 1
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229940065285 cadmium compound Drugs 0.000 description 1
- 150000001662 cadmium compounds Chemical class 0.000 description 1
- 239000000919 ceramic Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000009713 electroplating Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000010419 fine particle Substances 0.000 description 1
- 239000003673 groundwater Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 231100000003 human carcinogen Toxicity 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000002440 industrial waste Substances 0.000 description 1
- 239000010842 industrial wastewater Substances 0.000 description 1
- 230000002262 irrigation Effects 0.000 description 1
- 238000003973 irrigation Methods 0.000 description 1
- JQJCSZOEVBFDKO-UHFFFAOYSA-N lead zinc Chemical compound [Zn].[Pb] JQJCSZOEVBFDKO-UHFFFAOYSA-N 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005272 metallurgy Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000012466 permeate Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 229910052761 rare earth metal Inorganic materials 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 239000003403 water pollutant Substances 0.000 description 1
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention relates to application of a reflection spectrum in the field of ecological environment, in particular to a hyperspectral remote sensing quantitative inversion method of soil cadmium element content. The method comprises the steps of collecting soil samples in a research area, measuring the content of cadmium elements, obtaining reflection spectrums of the cadmium elements, carrying out spectrum transformation by adopting first-order differentiation, using the wavelength range of an image spectrum, selecting training and testing samples by random variation and a Kennard-Stone algorithm, optimally obtaining an optimal waveband combination by using a Grey wolf optimization algorithm, establishing a cadmium element content quantitative inversion model by combining a neural network technology, verifying based on a chemical testing result, and quantitatively inverting the content of the cadmium elements in the soil in the research area by using a CASI/SASI aviation hyperspectral remote sensing image. The method organically fuses the ground and the image spectrum, quickly and nondestructively obtains the content of the cadmium element in the soil in a large-scale research area, and has higher prediction precision.
Description
Technical Field
The invention relates to application of a reflection spectrum in the field of ecological environment, in particular to a hyperspectral remote sensing quantitative inversion method of soil cadmium element content.
Background
In 2019, cadmium and cadmium compounds were listed as toxic and harmful water pollutants. The cadmium element in the environment is mainly from crust and industrial pollution, and the plant food is easy to absorb three wastes discharged from metallurgy, smelting, ceramics, electroplating industry, chemical industry and the like due to the absorption of the cadmium element. Research shows that cadmium element is discharged into the environment through waste water and waste, cadmium-containing industrial waste diffuses and naturally settles, and is accumulated in soil around a factory, and beneficiation waste water of lead-zinc ore and related industrial waste water are discharged into ground water or permeate into underground water, and then pollute crops through irrigation, planting and other ways. With the proposal of the green industrial revolution concept, the problem that cadmium element diffuses into the environment to pollute the environment becomes one of the problems which need to be solved urgently. In addition, cadmium has been classified as a human carcinogen by the international cancer research institute and will also cause serious health damage to humans. Research has shown that over the past 30 years, the number of relevant studies on the source, level and distribution of cadmium elements in the environment has been increasing. Therefore, how to quantitatively analyze the content of cadmium in soil in a research area has important significance for further research of related technical methods, and meanwhile, the method can also provide technical support for researching the distribution of cadmium.
The traditional soil cadmium pollution monitoring method mainly adopts manual means with strong accuracy and high measurement precision, such as field fixed-point sampling, indoor chemical analysis and the like, but the method wastes time and labor and has certain destructiveness to the environment, and large-area and spatially continuous pollution distribution information is difficult to obtain by monitoring partial sampling points and profiles. The ground spectrometer can collect the spectral characteristics of the cadmium element in soil on site, but the operation area is relatively limited. The hyperspectral remote sensing has the characteristics of high spectral resolution, multiple and continuous wave bands and the like, can acquire fine ground object spectral information, and provides a new way for monitoring cadmium elements. In recent years, the trace element semi-quantitative inversion and regional content distribution inversion evaluation have shown good application value. With the emission of a series of hyperspectral remote sensing satellites at home and abroad, a series of quantitative inversion methods for the content of cadmium elements in soil emerge. However, the satellite image spectral information is easy to distort, the accuracy of information extraction is relatively low, and the information is difficult to be directly matched with the characteristic distribution of the actual soil cadmium element content. In addition, the image contains certain redundant information due to the high spectral resolution, and the dimension reduction of the sample is necessary, so that the inversion accuracy is further improved while the redundant features are removed.
In conclusion, the research on the relationship between the content of cadmium and the spectral reflectance at home and abroad is still in the beginning stage, and particularly, the research on the cadmium and the reflectance spectrum of soil in a large-scale research area is still under search. In view of the strategic importance of the soil cadmium element on environmental pollution prevention and control, the production value of the soil cadmium element and the like, the existing soil cadmium element geochemical analysis means and detection equipment have the limitations of precision and cost, and the distribution condition of the cadmium element in a large-scale research area is difficult to analyze. It is very necessary to develop a hyperspectral remote sensing quantitative inversion method for rapidly and nondestructively acquiring the content of cadmium elements in soil in a large-scale research area.
In view of this, the invention is particularly proposed.
Disclosure of Invention
The invention aims to provide a hyperspectral remote sensing quantitative inversion method of soil cadmium element content, which organically fuses ground and image spectrums, is a hyperspectral remote sensing quantitative inversion method for quickly and nondestructively acquiring the soil cadmium element content in a large-range research area and has higher prediction precision.
A hyperspectral remote sensing quantitative inversion method for soil cadmium element content comprises the following steps:
(1) collecting a soil sample of a research area;
(2) testing the content of the cadmium element in the soil of the sample by adopting an inductively coupled plasma mass spectrometer;
(3) acquiring a sample reflection spectrum by using a surface feature spectrometer;
(4) smoothing the reflection spectrum, and performing spectrum transformation by using a first-order differential technology to control a baseline effect;
(5) selecting training and testing samples by random variation and Kennard-Stone algorithm by using the wavelength range of the image spectrum;
(6) selecting a wave band by adopting a wolf optimization algorithm to obtain an optimal wavelength combination, and constructing a soil cadmium element quantitative inversion model by combining a neural network;
(7) and (5) carrying out quantitative inversion on the content of cadmium elements in the soil in the research area by carrying out operation on the CASI/SASI aviation hyperspectral remote sensing image.
Preferably, the soil samples of the research area collected in step (1) are: 5 sub-sampling points are distributed in a quincunx manner within the range of 50m of each sampling point in each research area, stone, weed and tree root impurities in the sample are removed, then a mixed sample is synthesized by a quartering method, and the weight of each sample is more than 1 kg; and (3) air-drying, grinding and sieving the sample by a 200-mesh sieve, and dividing the sample into two sub-samples, wherein one sub-sample is used for the reflection spectrum measurement in the step (3), and the other sub-sample is used for the soil cadmium element content analysis in the step (2).
Preferably, the inductively coupled plasma mass spectrometer in the step (2) adopts NexIoN 350x type/CSY-066. And (2) accurately measuring the content of the soil cadmium element of the sample by a chemical method, and establishing a connection with the sample spectrum so as to establish a quantitative inversion model.
Preferably, the surface feature spectrometer in the step (3) is an SVC HR1024 surface feature spectrometer manufactured by Spectra Vista company of America.
Preferably, step (3) is: placing the sample in a sample container, and scraping the surface; the halogen lamp is used as a spectrum measurement light source, a white reference plate is used for radiometric calibration before measurement, a ground object spectrometer is used for obtaining a sample spectrum, the wavelength range is 350 nm-2500 nm, the distance between an optical fiber of the ground object spectrometer and a sample is 5-10 cm, and the angle of field of the optical fiber is 25 degrees.
Preferably, in step (4), a first order differential technique is used for correcting the baseline effect in the spectrum to eliminate non-chemical effects and establish a robust correction model; the first order differential equation is expressed as follows:
wherein,finger-shapedThe first order differential of (a) is,for the reflectivity of the next sampling band,is the reflectivity of the last sampling band,is the sampling interval.
Preferably, step (6) is:
step a: initializing parameters needed by the population and the algorithm, and setting the parameters:
including the number of wolf individualsMaximum number of iterationsConvergence factorSum coefficient vector、(ii) a WhereinLinearly decreasing from an initial value of 2 to 0 during the iteration,andis a random number between 0 and 1,andpositively correlating, wherein each wolf individual in the population represents a wave band combination, and randomly initializing each individual position;
step b: and (3) predicting the content of cadmium in the test sample:
taking each individual grey wolf position as an input, taking each individual grey wolf position as a candidate wavelength combination, and continuously adjusting a weight and a threshold value minimization target function through back propagation to finally output a predicted value of the content of the cadmium element; continuously iterating through a neural network to obtain a cadmium element predicted value;
step c: calculating an individual fitness value:
obtaining the initial position of an individual in a wolf optimization algorithm, calculating the fitness of the individual, and determining the optimal value of the fitnessSub-optimum valueAnd the next order of meritProtection ofStoring corresponding individual positions;
step e: decoding the individual positions, and weighting the absorption wavelength positions of the cadmium elements;
sorting all the individual fitness values in a descending order, and reserving an optimal valueSub-optimum valueAnd the next order of meritThe corresponding individual position;
step g: judging a termination condition;
judging whether the current iteration number reaches the maximum iteration numberIf it has, it willAnd (4) constructing a soil cadmium element quantitative inversion model by combining the optimal waveband combination obtained by decoding and a neural network, and otherwise, repeating the steps c-f.
Preferably, in step b, the objective function is expressed as follows:
wherein,in order to train the sample size,as output variable dimension, which corresponds to the total number of bands,andrespectively obtaining a true value and a predicted value of the cadmium element content of the sample;
in the step b, continuously iterating through a neural network to obtain a cadmium element predicted value, which is defined as follows:
wherein,is the function of the activation of the function,is the firstThe output of each of the hidden layers is,is connected with a hidden layerAndthe weight of (a) is determined,is the firstThe layer outputs the bias of the layer.
In the step c, the individual fitness calculating method comprises the following steps: and calculating by combining a neural network technology and utilizing an objective function to obtain the fitness value of the band combination corresponding to each group of codes, wherein a fitness calculation formula is as follows:
wherein,representing an individualThe value of the fitness value of (a) is,the prediction accuracy of the content of cadmium element corresponding to the individually selected wave band combination is shown,which indicates the number of the selected bands,weighting parameters to balance accuracy and number of bands.
Taking the individual position codes as candidate band combinations, calculating the fitness value of the band combination corresponding to each group of codes through an objective function, and keeping the optimal value of the wolf pack individualsSub-optimum valueAnd the next order of meritWhen the algorithm satisfies the termination conditionAnd decoding and outputting the optimal band combination.
Preferably, the inversion model input item is the corresponding optimal band combination after decoding by the wolf optimization algorithm.
Compared with the prior art, the invention has the following beneficial effects:
(1) the regional inversion of the soil components is realized based on the aviation hyperspectral remote sensing image, and the possibility is provided for solving the distribution condition of the soil trace elements. The invention provides a hyperspectral remote sensing quantitative inversion method of soil cadmium element content, which is characterized by collecting soil samples of a plurality of research areas to test the cadmium element content, acquiring a reflection spectrum, smoothing the spectrum, representing the absorption characteristic of the cadmium element through first-order differential transformation, selecting training and testing samples by using the wavelength range of the image spectrum and combining random variation and a Kennard-Stone algorithm, selecting an optimal waveband combination through a wolf optimization algorithm, and simultaneously carrying out quantitative inversion modeling by combining a neural network, applying the method to CASI/SASI aviation hyperspectral images and analyzing the distribution condition of the soil cadmium element content of the research areas. This process organically fuses the ground with the image spectra so that both maintain similar trends. Therefore, on the basis of ensuring certain ground sampling, the invention trains a relatively accurate quantitative inversion model of the content of the cadmium element in the soil. The model can quickly and accurately carry out quantitative inversion on the content of the cadmium element in the soil in a large-range research area, and provides a valuable reference for further soil environment evaluation.
(2) The hyperspectral remote sensing sensor can monitor a large-range research area in real time, and the meta-heuristic algorithm has the characteristics of high convergence rate, strong optimization capability and the like, and can efficiently select a band combination with small correlation from all spectral characteristics. In addition, the soil cadmium element and the reflection spectrum generally present a nonlinear relation, and the neural network has good generalization capability when fitting the nonlinear relation, thereby bringing new ideas and precision improvement for quantitative inversion of the soil cadmium element. The invention organically fuses the ground and the image spectrum, is a hyperspectral remote sensing quantitative inversion method for quickly and nondestructively acquiring the content of the cadmium element in the soil in a large-range research area, realizes the prediction and evaluation of the content of the cadmium element in the soil in the research area, and searches the distribution condition of the cadmium element. The adopted spectrum transformation technology effectively eliminates the sample baseline effect and enhances the spectrum absorption characteristic, wherein the first order differential transformation has better enhanced spectrum characteristics; the adopted wolf optimization algorithm has the characteristics of simple structure, few parameters, easy realization and the like, and can quickly and efficiently select the optimal wave band combination in the spectrum sampling process; the adopted neural network has better prediction capability, stable performance and higher prediction precision. The method can rapidly and nondestructively carry out quantitative inversion on the content of the cadmium element in the soil, and meanwhile provides technical support for regional evaluation.
Drawings
In order to explain the specific technical solution of the present invention, the drawings required in the description of the related art will be briefly introduced.
FIG. 1 is a schematic overall flow diagram of the present invention;
FIG. 2 is a schematic flow chart of the gray wolf optimization algorithm of the present invention;
FIG. 3 is an aerial hyperspectral remote sensing image of a research area;
FIG. 4 is a diagram of the distribution of cadmium element content in soil in a research area obtained by quantitative inversion.
Detailed Description
It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the disclosure. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms also include the plural forms unless the context clearly dictates otherwise, and further, it is understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of the stated features, steps, operations, devices, components, and/or combinations thereof.
The technical solutions of the present invention will be described clearly and completely with reference to the following embodiments, and it should be understood that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Examples
A hyperspectral remote sensing quantitative inversion method for soil cadmium element content is shown in figures 1-2 and comprises the following steps:
(1) sample collection
The sampling place is mainly a farmland in a research area, and a few sampling places are in a building land and a tailing pile, 264 soil samples of 0-20 cm plough layers are collected in total for chemical analysis and spectral measurement. The sampling time of the sample is 9-10 months in 2020, according to a sampling method specified in soil environment monitoring technical Specification (implemented by HJ/T166-. The sample testing and analysis is completed by a test center of a geophysical survey institute of the China metallurgical geological institute, after the sample is brought back to the test center, the sample is dried, ground and sieved (200 meshes), each obtained fine particle is divided into two sub-samples, one sub-sample is used for the reflection spectrum measurement in the step (3), and the other sub-sample is used for the cadmium element content analysis in the step (2).
(2) And (3) measuring the content of the cadmium element in the soil by adopting an inductively coupled plasma mass spectrometer NexIoN 350x type/CSY-066. The content of the soil cadmium element of the sample is accurately measured by a chemical method and is used for establishing connection with the sample spectrum so as to establish a quantitative inversion model.
(3) Reflectance spectrum acquisition
An SVC HR1024 terrain spectrometer manufactured by the American Spectra Vista company is used for measuring the spectrum of a soil sample with the wavelength range of 350-2500 nm, wherein the wavelength range covers a near infrared region and a short wave infrared region, and the total wavelength range is 1024 bands. Placing the sample in a sample container, and scraping the surface; a halogen lamp is used as a spectrum measurement light source, a white reference plate is used for radiation calibration before measurement, the distance between a spectrometer optical fiber and a sample is 5-10 cm, and the angle of field of the optical fiber is 25 degrees. The samples were averaged after 5 measurements to minimize measurement error. In order to eliminate the influence of the environment on spectral measurement, a part of ultraviolet spectral bands (with the wavelength less than 400 nm) with large noise are removed, and finally, spectral characteristics in the range of 400 nm-2500 nm are selected, and the total number of the spectral characteristics is 924 bands.
(4) Spectral transformation
The reflection spectrum preprocessing step and method play a crucial role in predicting the quality of results. However, since the content of cadmium element in the soil sample is too low to complete the spectral reflection, a series of spectral processing is required to enhance the spectral characteristics.
Firstly, the spectral curve in the reflection spectrum measurement stage usually has noise, and in order to effectively eliminate the spectral noise and simultaneously store the chemical information expressed by the spectrum, the spectrum smoothing processing is performed.
In addition, in order to eliminate the possibility of baseline effect brought by the particle size and enhance the spectral absorption characteristic of cadmium element in soil, first-order differential transformation is carried out. A first order differential technique is used to correct for baseline effects in the spectrum to eliminate non-chemical effects and to build a robust correction model. Specifically, the first derivative of the spectrum represents a measure of the slope of the spectral curve at each point, and the slope of the curve is not imaged by a purely additive baseline shift in the spectrum, and the first derivative minimizes background interference and baseline drift in the reflectance spectrum and effectively eliminates the baseline shift. The first order differential equation is expressed as follows:
wherein,finger-shapedThe first order differential of (a) is,for the reflectivity of the next sampling band,is the reflectivity of the last sampling band,is the sampling interval.
(5) Sample selection
For training a learner with good fitting effect, 264 samples subjected to spectral transformation are divided into training samples and verification samples before model construction, wherein the proportion of the training samples to the verification samples is 7: 3. the uniformity of the sample property distribution is significantly improved by using the wavelength range of the image spectrum, by using random variation and Kennard-Stone techniques, so that the training and testing samples cover the entire sample space. The technology is calculated based on Euclidean distance, the random variation factor is used for 10%, the positions of samples in a training set and a testing set are continuously updated in an iterative mode, and balance is kept between randomness and convergence. And selecting a sample with the largest decision coefficient, the smallest root mean square error and the large decision coefficient of the verification set for testing through iteration and cross verification.
(6) Band selection
Because the spectrum contains a large amount of redundant information, a large amount of time is consumed for carrying out quantitative inversion of the cadmium element in the soil in a large-scale research area. Therefore, band selection is necessary in the modeling process. The optimal band combination is obtained by adopting a wolf optimization algorithm, as shown in fig. 2, the specific process is as follows:
step a: initializing the population and the parameters needed by the algorithm.
Specifically, parameters required by a population and an algorithm are initialized, and various parameters are set: including the number of wolf individualsMaximum number of iterationsConvergence factorSum coefficient vector、(ii) a WhereinLinearly decreasing from an initial value of 2 to 0 during the iteration,andis a random number between 0 and 1,andpositively correlating, wherein each wolf individual in the population represents a wave band combination, and randomly initializing each individual position;
step b: and predicting the content of cadmium element in the test sample.
The neural network is based on the gradient descent principle and mainly comprises two processes: information forward propagation and error reverse propagation. The technology takes the individual positions (namely candidate wavelength combinations) of the wolfsbane as input, and finally outputs the predicted value of the content of the cadmium element by continuously adjusting the weight and the threshold value minimization target function through back propagation. The objective function is expressed as follows:
wherein,in order to train the sample size,as output variable dimension, which corresponds to the total number of bands,andrespectively obtaining a true value and a predicted value of the cadmium element content of the sample;
continuously iterating through a neural network to obtain a cadmium element predicted value, which is defined as follows:
wherein,is the function of the activation of the function,is the firstThe output of each of the hidden layers is,is connected with a hidden layerAndthe weight of (a) is determined,is the firstThe layer outputs the bias of the layer.
And the neural network updates the weight and the bias of each layer of neuron through iteration and finally outputs a predicted value of the content of the cadmium element.
Step c: an individual fitness value is calculated.
Obtaining the initial position of an individual in a wolf optimization algorithm, calculating the fitness of the individual, and determining the optimal value of the fitnessSub-optimum valueAnd the next order of meritAnd saving the corresponding individual position.
Specifically, the fitness value of the band combination corresponding to each group of codes is calculated by using an objective function in combination with a neural network technology, and higher prediction accuracy is obtained by selecting as little wavelength information as possible.
The fitness calculation formula is as follows:
wherein,representing an individualThe value of the fitness value of (a) is,the prediction accuracy of the content of cadmium element corresponding to the individually selected wave band combination is shown,which indicates the number of the selected bands,weighting parameters to balance accuracy and number of bands.
Step d: updating individual locations and parameters、、Specifically, the gray wolf optimization algorithm considers、Andthe potential position of the prey is closer, the position of the prey is judged by utilizing the information contained in the three parts, and other individuals gradually approach the prey according to the positions of the three parts. Updating individual locations and parameters、、The specific updating formula is as follows:
wherein,、、respectively in step c、Andthe corresponding individual positions, namely the positions of the three wolfs.
Step e: decoding the individual positions and weighting the absorption wavelength positions of the cadmium elements.
Sorting all the individual fitness values in a descending order, and reserving an optimal valueSub-optimum valueAnd the next order of meritThe corresponding individual position. The initial position vector of an individual in the wolf optimization algorithm is coded into a corresponding band combination by adopting a binary coding space, the binary coding space is designed to code a code '0' to indicate that a band corresponding to a component is not selected, and a code '1' to indicate that the band corresponding to the component is selected. Calculating through an objective function to obtain the fitness value of the corresponding band combination of each group of codes, and reserving the optimal value of the wolf pack individualSub-optimum valueAnd the next order of meritWhen the algorithm satisfies the termination conditionDecoding is carried out, and the optimal band combination is output. According to the geochemical principle, the characteristics of the cadmium element in the soil at the absorption wavelength are extracted, the wave band combination is weighted, and the adaptability value of the wave band combination is evaluated. In principle, the spectral feature has an extreme value in the vicinity of the absorption peak, and no significant absorption feature appears for the absorption wavelength corresponding to the non-cadmium element.
specifically, the content of cadmium element is predicted by combining the weight determined in step 4 and the wavelength combination determined in step 3 through a neural network. Calculating individual fitness value by using the objective function, recalculating fitness values of all wolfs, and updating the sorted individual fitness values、Andthe corresponding individual position.
Step g: and (5) judging the termination condition.
Specifically, the algorithm determines whether the current iteration count reaches a maximum iteration countIf not, returning to the step c to continue execution, repeating the steps c-f, and searching the optimal wave band combination; if it has, it willAnd (5) constructing a soil cadmium element quantitative inversion model by combining the optimal waveband combination obtained by decoding and a neural network.
(7) Quantitative inversion of soil cadmium element content
The image acquisition time of the research area is 10-14 points per day in 6-8 months in 2018, the acquired data are subjected to radiometric calibration, geometric correction and orthorectification through a system provided by ITRES company, FLAASH is adopted for atmospheric correction, and the SASI file is resampled to CASI, the spectral range is 400-2500 nm, and the total wave band number is 173. The method comprises the steps of performing first-order differential transformation on a reflection spectrum, selecting a training and testing sample by using a random variation and Kennard-Stone algorithm, obtaining an optimal wave band combination based on a Grey wolf optimization algorithm, directly applying the optimal wave band combination to an aviation hyperspectral remote sensing image, and quantitatively inverting the content of cadmium elements in soil in a research area through a neural network, wherein the result is shown in table 1, fig. 3 is the aviation hyperspectral remote sensing image in the research area, and fig. 4 is a distribution situation diagram of the content of cadmium elements in soil in the research area obtained through quantitative inversion.
TABLE 1 neural network training and testing results
For the training samples, the decision coefficient of the ground samples reached 0.9286, indicating that the predicted values have a high correlation with the measured values. For the test samples, the decision coefficient for the ground samples was further improved to 0.9948, indicating that the predicted values are very close to the measured value distribution. Experimental results show that the quantitative inversion method for the cadmium element in the soil has good adaptability to ground samples in a research area and can meet the requirements of practical application. And further testing is carried out through an aviation hyperspectral remote sensing image, and the cadmium element content in most regions of the research area is in the range of a risk screening value and a background value, so that earlier-stage related treatment measures are shown to effectively avoid further pollution of the whole soil environment. However, since small smelters are becoming the major economic source of the area and are difficult to eradicate completely in a short time, the area is still controlled by the risk management of part of the area cadmium element content, and subsequently more strict treatment measures are required to ensure that the rare earth element content of the area reaches the normal level. The invention provides a hyperspectral remote sensing quantitative inversion model of soil cadmium element content by taking a new male security area as a research area. Through sampling, measurement, data processing, modeling analysis and experimental verification, the reflection spectrum is proved to be applicable to rapid and nondestructive evaluation of the cadmium element pollution condition of the soil, and scientific and technical and theoretical bases are provided for pollution prevention and control work in research areas.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (10)
1. A hyperspectral remote sensing quantitative inversion method for soil cadmium element content is characterized by comprising the following steps:
(1) collecting a soil sample of a research area;
(2) testing the content of the cadmium element in the soil of the sample by adopting an inductively coupled plasma mass spectrometer;
(3) acquiring a sample reflection spectrum by using a surface feature spectrometer;
(4) smoothing the reflection spectrum, and performing spectrum transformation by using a first-order differential technology to control a baseline effect;
(5) selecting training and testing samples by random variation and Kennard-Stone algorithm by using the wavelength range of the image spectrum;
(6) selecting a wave band by adopting a wolf optimization algorithm to obtain an optimal wavelength combination, and constructing a soil cadmium element quantitative inversion model by combining a neural network;
(7) and (5) carrying out quantitative inversion on the content of cadmium elements in the soil in the research area by carrying out operation on the CASI/SASI aviation hyperspectral remote sensing image.
2. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1, characterized in that the collection of soil samples in the research area in step (1) is: 5 sub-sampling points are distributed in a quincunx manner within the range of 50m of each sampling point in each research area, stone, weed and tree root impurities in the sample are removed, then a mixed sample is synthesized by a quartering method, and the weight of each sample is more than 1 kg; and (3) air-drying, grinding and sieving the sample by a 200-mesh sieve, and dividing the sample into two sub-samples, wherein one sub-sample is used for the reflection spectrum measurement in the step (3), and the other sub-sample is used for the soil cadmium element content analysis in the step (2).
3. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1 is characterized in that the inductively coupled plasma mass spectrometer in the step (2) adopts NexIoN 350x type/CSY-066.
4. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1 is characterized in that the surface feature spectrometer in the step (3) adopts an SVC HR1024 surface feature spectrometer.
5. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 4, characterized in that the step (3) is as follows: placing the sample in a sample container, and scraping the surface; the halogen lamp is used as a spectrum measurement light source, a white reference plate is used for radiometric calibration before measurement, a ground object spectrometer is used for obtaining a sample spectrum, the wavelength range is 350 nm-2500 nm, the distance between an optical fiber of the ground object spectrometer and a sample is 5-10 cm, and the angle of field of the optical fiber is 25 degrees.
6. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1 is characterized in that in the step (4), a first order differential technology is used for correcting the baseline effect in the spectrum so as to eliminate non-chemical effects and establish a robust correction model; the first order differential equation is expressed as follows:
7. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1, characterized in that the step (6) is as follows:
step a: initializing parameters needed by the population and the algorithm, and setting the parameters:
including the number of wolf individualsMaximum number of iterationsConvergence factorSum coefficient vector、(ii) a WhereinLinearly decreasing from an initial value of 2 to 0 during the iteration,andis a random number between 0 and 1,andpositively correlating, wherein each wolf individual in the population represents a wave band combination, and randomly initializing each individual position;
step b: and (3) predicting the content of cadmium in the test sample:
taking each individual grey wolf position as an input, taking each individual grey wolf position as a candidate wavelength combination, and continuously adjusting a weight and a threshold value minimization target function through back propagation to finally output a predicted value of the content of the cadmium element; continuously iterating through a neural network to obtain a cadmium element predicted value;
step c: calculating an individual fitness value:
obtaining the initial position of an individual in a wolf optimization algorithm, calculating the fitness of the individual, and determining the optimal value of the fitnessSub-optimum valueAnd the next order of meritStoring the corresponding individual positions;
step e: decoding the individual positions, and weighting the absorption wavelength positions of the cadmium elements;
sorting all the individual fitness values in a descending order, and reserving an optimal valueSub-optimum valueAnd the next order of meritThe corresponding individual position;
step g: judging a termination condition;
judging whether the current iteration number reaches the maximum iteration numberIf it has, it willAnd (4) constructing a soil cadmium element quantitative inversion model by combining the optimal waveband combination obtained by decoding and a neural network, and otherwise, repeating the steps c-f.
8. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 7 is characterized in that in the step b, the objective function is expressed as follows:
wherein,in order to train the sample size,as output variable dimension, which corresponds to the total number of bands,andrespectively obtaining a true value and a predicted value of the cadmium element content of the sample;
in the step b, continuously iterating through a neural network to obtain a cadmium element predicted value, which is defined as follows:
9. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 7 is characterized in that in the step c, the calculation method of individual fitness comprises the following steps: and calculating by combining a neural network technology and utilizing an objective function to obtain the fitness value of the band combination corresponding to each group of codes, wherein a fitness calculation formula is as follows:
wherein,representing an individualThe value of the fitness value of (a) is,the prediction accuracy of the content of cadmium element corresponding to the individually selected wave band combination is shown,which indicates the number of the selected bands,weighting parameters to balance accuracy and number of bands.
10. The hyperspectral remote sensing quantitative inversion method of soil cadmium element content according to claim 1, characterized in that the inversion model input item is the corresponding optimal band combination after decoding by the wolf optimization algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210024419.5A CN114357891A (en) | 2022-01-11 | 2022-01-11 | Hyperspectral remote sensing quantitative inversion method for soil cadmium element content |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210024419.5A CN114357891A (en) | 2022-01-11 | 2022-01-11 | Hyperspectral remote sensing quantitative inversion method for soil cadmium element content |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114357891A true CN114357891A (en) | 2022-04-15 |
Family
ID=81108375
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210024419.5A Pending CN114357891A (en) | 2022-01-11 | 2022-01-11 | Hyperspectral remote sensing quantitative inversion method for soil cadmium element content |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114357891A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117269078A (en) * | 2023-11-21 | 2023-12-22 | 核工业北京地质研究院 | Method for determining cation exchange capacity of soil |
CN117688835A (en) * | 2023-12-11 | 2024-03-12 | 哈尔滨航天恒星数据***科技有限公司 | Soil nutrient inversion method, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111144308A (en) * | 2019-12-25 | 2020-05-12 | 中国冶金地质总局矿产资源研究院 | Kaolin mineral information extraction model and building method and application thereof |
CN112699596A (en) * | 2020-12-04 | 2021-04-23 | 湖南工商大学 | Wide-area electromagnetic method induced polarization information nonlinear extraction method based on learning |
CN112766075A (en) * | 2020-12-31 | 2021-05-07 | 中国冶金地质总局矿产资源研究院 | Hyperspectral remote sensing black and odorous water body grading method based on semi-supervised learning strategy |
WO2021184550A1 (en) * | 2020-03-19 | 2021-09-23 | 中国科学院南京土壤研究所 | Target soil property content prediction method based on soil transfer function |
CN113866102A (en) * | 2021-09-24 | 2021-12-31 | 中国地质调查局西安地质调查中心(西北地质科技创新中心) | Soil health investigation monitoring method based on spectrum |
-
2022
- 2022-01-11 CN CN202210024419.5A patent/CN114357891A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111144308A (en) * | 2019-12-25 | 2020-05-12 | 中国冶金地质总局矿产资源研究院 | Kaolin mineral information extraction model and building method and application thereof |
WO2021184550A1 (en) * | 2020-03-19 | 2021-09-23 | 中国科学院南京土壤研究所 | Target soil property content prediction method based on soil transfer function |
CN112699596A (en) * | 2020-12-04 | 2021-04-23 | 湖南工商大学 | Wide-area electromagnetic method induced polarization information nonlinear extraction method based on learning |
CN112766075A (en) * | 2020-12-31 | 2021-05-07 | 中国冶金地质总局矿产资源研究院 | Hyperspectral remote sensing black and odorous water body grading method based on semi-supervised learning strategy |
CN113866102A (en) * | 2021-09-24 | 2021-12-31 | 中国地质调查局西安地质调查中心(西北地质科技创新中心) | Soil health investigation monitoring method based on spectrum |
Non-Patent Citations (4)
Title |
---|
HUANG,ZQ 等: "Inversion Evaluation of Rare Eath Elements in Soil by Visible-Shortwave infrared Spectroscopy", 《REMOTE SENSING》 * |
王轩慧等: "基于SGA-RF算法的农业土壤镉浓度反演研究", 《农业机械学报》 * |
田泽众 等: "基于 IGWO 算法的冬小麦作物土壤全氮含量一体化监测", 《农业机械学报》 * |
黄照强 等: "基于随机变异-Kennard-Stone和偏最小二乘法的土壤重金属镉含量反演---以雄安新区西南部为例", 《地质论评》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117269078A (en) * | 2023-11-21 | 2023-12-22 | 核工业北京地质研究院 | Method for determining cation exchange capacity of soil |
CN117269078B (en) * | 2023-11-21 | 2024-03-22 | 核工业北京地质研究院 | Method for determining cation exchange capacity of soil |
CN117688835A (en) * | 2023-12-11 | 2024-03-12 | 哈尔滨航天恒星数据***科技有限公司 | Soil nutrient inversion method, electronic equipment and storage medium |
CN117688835B (en) * | 2023-12-11 | 2024-06-04 | 哈尔滨航天恒星数据***科技有限公司 | Soil nutrient inversion method, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhou et al. | Hyperspectral inversion of soil heavy metals in Three-River Source Region based on random forest model | |
Vaudour et al. | Sentinel-2 image capacities to predict common topsoil properties of temperate and Mediterranean agroecosystems | |
Wang et al. | New methods for improving the remote sensing estimation of soil organic matter content (SOMC) in the Ebinur Lake Wetland National Nature Reserve (ELWNNR) in northwest China | |
Malenovský et al. | Retrieval of spruce leaf chlorophyll content from airborne image data using continuum removal and radiative transfer | |
Bartholomeus et al. | Soil organic carbon mapping of partially vegetated agricultural fields with imaging spectroscopy | |
CN110174359B (en) | Aviation hyperspectral image soil heavy metal concentration assessment method based on Gaussian process regression | |
WO2023087630A1 (en) | Method for estimating soil salinity of straw residue farmland by using remote sensing construction index | |
Mashimbye et al. | Model-based integrated methods for quantitative estimation of soil salinity from hyperspectral remote sensing data: A case study of selected South African soils | |
CN104266982B (en) | A kind of large area insect pest quantifies monitoring system | |
CN114357891A (en) | Hyperspectral remote sensing quantitative inversion method for soil cadmium element content | |
CN107421894A (en) | Based on unmanned plane EO-1 hyperion inverting heavy metal in soil pollution monitoring method | |
CN114018833A (en) | Method for estimating heavy metal content of soil based on hyperspectral remote sensing technology | |
CN103792246A (en) | X-ray fluorescence spectrum quantitative analysis method based on LM-BP neural network | |
CN114460013B (en) | Coastal wetland vegetation overground biomass GAN model self-learning remote sensing inversion method | |
Lin et al. | Estimating the heavy metal contents in farmland soil from hyperspectral images based on Stacked AdaBoost ensemble learning | |
Wang et al. | Effects of hyperspectral data with different spectral resolutions on the estimation of soil heavy metal content: From ground-based and airborne data to satellite-simulated data | |
Diek et al. | Minimizing soil moisture variations in multi-temporal airborne imaging spectrometer data for digital soil mapping | |
CN110836870B (en) | GEE-based large-area lake transparency rapid drawing method | |
CN113466143B (en) | Soil nutrient inversion method, device, equipment and medium | |
CN114813651A (en) | Remote sensing water quality inversion method combining difference learning rate and spectrum geometric characteristics | |
CN115128013A (en) | Soil organic matter content space prediction evaluation method based on partition algorithm | |
CN114241331A (en) | Wetland reed aboveground biomass remote sensing modeling method taking UAV as ground and Sentinel-2 intermediary | |
Hang et al. | Estimation of chlorophyll-a concentration in Lake Taihu from Gaofen-1 wide-field-of-view data through a machine learning trained algorithm | |
CN114814167B (en) | Soil heavy metal content inversion method fusing multi-source environment variables and spectral information | |
CN117409334A (en) | Aerosol type identification method based on stationary satellite remote sensing data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220415 |
|
RJ01 | Rejection of invention patent application after publication |