CN114878508A - Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site - Google Patents
Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site Download PDFInfo
- Publication number
- CN114878508A CN114878508A CN202210490117.7A CN202210490117A CN114878508A CN 114878508 A CN114878508 A CN 114878508A CN 202210490117 A CN202210490117 A CN 202210490117A CN 114878508 A CN114878508 A CN 114878508A
- Authority
- CN
- China
- Prior art keywords
- hyperspectral imaging
- data
- sample
- visible light
- water content
- 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
- 238000000701 chemical imaging Methods 0.000 title claims abstract description 93
- 238000000034 method Methods 0.000 title claims abstract description 48
- 239000002352 surface water Substances 0.000 title claims abstract description 25
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 59
- 238000001228 spectrum Methods 0.000 claims abstract description 33
- 238000002156 mixing Methods 0.000 claims abstract description 4
- 230000003595 spectral effect Effects 0.000 claims description 27
- 238000009499 grossing Methods 0.000 claims description 23
- 238000002310 reflectometry Methods 0.000 claims description 21
- 238000007781 pre-processing Methods 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 16
- 238000010238 partial least squares regression Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 6
- 239000004927 clay Substances 0.000 claims description 3
- 238000012847 principal component analysis method Methods 0.000 claims description 3
- 239000002689 soil Substances 0.000 description 23
- 239000011159 matrix material Substances 0.000 description 10
- 238000012795 verification Methods 0.000 description 7
- 230000006870 function Effects 0.000 description 6
- 230000009466 transformation Effects 0.000 description 5
- 238000010200 validation analysis Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 239000008367 deionised water Substances 0.000 description 2
- 229910021641 deionized water Inorganic materials 0.000 description 2
- 238000012886 linear function Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 150000003839 salts Chemical class 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- BPQQTUXANYXVAA-UHFFFAOYSA-N Orthosilicate Chemical compound [O-][Si]([O-])([O-])[O-] BPQQTUXANYXVAA-UHFFFAOYSA-N 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000002329 infrared spectrum Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/3563—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor
- G01N2021/3568—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light for analysing solids; Preparation of samples therefor applied to semiconductors, e.g. Silicon
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/30—Assessment of water resources
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention provides a method for detecting the surface water content of cultural relics in an earthen site based on hyperspectral imaging. The method comprises the following steps: collecting samples around the cultural relics in the earthen site, adding water to the samples, uniformly mixing the samples, and preparing the samples into gradient standard samples with different external water contents; collecting visible light-near infrared hyperspectral imaging data of the gradient standard sample; extracting characteristic wave bands of the visible light-near infrared hyperspectral imaging data to obtain characteristic spectrum data; establishing a relation between characteristic spectrum data of a gradient standard sample and an external water content, establishing an earthen site cultural relic surface water content prediction model, and inputting visible light-near infrared hyperspectral imaging data of a sample to be detected of the earthen site cultural relic into the prediction model to obtain the predicted water content of the sample to be detected of the earthen site cultural relic.
Description
Technical Field
The invention belongs to the technical field of cultural relic detection, particularly relates to nondestructive detection for protecting the water content of the cultural relic, and particularly relates to a method for detecting the water content of the surface of the cultural relic on the basis of hyperspectral imaging.
Background
China possesses a large number of clay sites and heritages, and is an important component in excellent cultural heritage of China. The earthen site is a cultural heritage which is built by taking soil as a main material historically and has historical, artistic, scientific, social and cultural values. These cultural heritages are important carriers of long history and splendid culture in China, and are real signs of Chinese civilization. Many earthen relics are subjected to factors such as natural erosion and environmental change, and the protection status is not optimistic. Among many diseases, the damage of salt and moisture is the most difficult and serious to control and is also a difficult point and a hot point of cultural relic protection and research works at home and abroad.
Under the interaction of underground water and atmospheric water, the surface moisture of the earthen site becomes a potential threat to damage of the earthen site, so that the acquisition of the surface moisture content of the earthen site plays an important role in protecting the earthen site in a drought or humid area. The traditional soil water content testing method can damage the surface information of the earthen site, causes environmental pollution, wastes time and labor, has low resolution and is not suitable for the field of earthen site cultural relic protection. At present, a hyperspectral soil moisture inversion model mainly comprises partial least square regression, a neural network method, a stepwise regression method and a multiple linear regression method, and due to the fact that the variables of the model are numerous, the model has low operation efficiency and poor stability, and the application of quickly and accurately measuring the surface water content of an earthen site is difficult. Meanwhile, the reflectivity of the visible light-near infrared spectrum is closely related to various properties of the sample such as moisture, salt, organic matters, color, oxide content and the like, and a soil water content prediction model is difficult to be applied to determination of the surface water content of the earthen site, so detailed research needs to be carried out on materials on the surface of the specific earthen site.
The existing data show that the research and related reports for establishing a reasonable prediction model of the surface water content of the earthen site by adopting a method combining hyperspectral imaging with chemometrics do not exist in the field of cultural relic protection.
Disclosure of Invention
Aiming at the short plate existing in the determination of the surface water content of the cultural relic in the earthen site in the prior art, the invention provides the detection method of the surface water content of the cultural relic in the earthen site based on hyperspectral imaging, and the method is used as an effective technique for determining the surface water content of the cultural relic in the earthen site, thereby improving the accuracy of the determination of the surface water content of the earthen site.
Therefore, the invention provides a method for detecting the surface water content of the cultural relic in the earthen site based on hyperspectral imaging. The method comprises the following steps:
step (i): collecting samples around the cultural relics in the earthen site, adding water to the samples, uniformly mixing the samples, and preparing the samples into gradient standard samples with different external water contents;
step (ii): collecting visible light-near infrared hyperspectral imaging data of the gradient standard sample;
step (iii): extracting characteristic wave bands of the visible light-near infrared hyperspectral imaging data to obtain characteristic spectrum data;
step (iv): establishing a relation between characteristic spectrum data of a gradient standard sample and an external water content, establishing an earthen site cultural relic surface water content prediction model, and inputting visible light-near infrared hyperspectral imaging data of a sample to be detected of the earthen site cultural relic into the prediction model to obtain the predicted water content of the sample to be detected of the earthen site cultural relic.
Preferably, the characteristic spectrum data of the visible light hyperspectral imaging is spectrum data of 400-1000nm wave band; the characteristic spectral data of the near-infrared hyperspectral imaging is the spectral data of a band of 1000-2500 nm.
Preferably, the method further comprises: and preprocessing the visible light-near infrared hyperspectral imaging spectral data before extracting the characteristic wave bands of the visible light-near infrared hyperspectral imaging data.
Preferably, one or more of Savitzky-Golay smoothing, standard normal variation, first-order derivative and second-order derivative are combined to preprocess the visible light-near infrared hyperspectral imaging spectral data; preferably, the visible light hyperspectral imaging data, especially the reflectivity data, of the standard sample are preprocessed by adopting a combination of Savitzky-Golay smoothing and standard normal variation; the near-infrared hyperspectral imaging data, especially the reflectivity data, of the standard sample are preprocessed by adopting a combination of Savitzky-Golay smoothing and second-order derivative.
Preferably, in each standard sample, the spectral curves of the visible light-near infrared hyperspectral are arithmetically averaged in a selected area, and actual reflectivity spectral data of each standard sample for preprocessing are obtained.
Preferably, one or more of a continuous projection algorithm and a principal component analysis method are combined to extract the characteristic wave band of the visible light-near infrared hyperspectral imaging data.
Preferably, a prediction model of the surface water content of the cultural relics in the earthen ruins is established by adopting a partial least squares regression and/or support vector regression method; preferably, a prediction model is established for the characteristic spectrum data based on visible light hyperspectral imaging by adopting partial least squares regression, and a prediction model is established for the characteristic spectrum data based on near-infrared hyperspectral imaging by adopting support vector regression.
Preferably, the earthen site cultural relics are clay-based cultural relics. Including but not limited to sandstone, stony, etc., some silicate cultural relics that facilitate the selection of a perimeter simulation sample and have potential water damage.
Preferably, the amount of water added is the mass of water added per oven dry mass of the sample.
Preferably, the applied moisture content of the gradient standard sample is distributed between 0 and 0.45.
The method for detecting the surface water content of the relic cultural relic on the basis of hyperspectral imaging can establish a specific model aiming at the relic samples with different properties and in different areas, and has strong popularization. In addition, the method adopts the sample basically consistent with the composition of the cultural relics of the earthen ruins, and the adaptability of the prediction result is further improved. Through verification, the method is stable in model, strong in prediction pertinence and high in prediction precision.
Drawings
FIG. 1 is the visible light hyperspectral imaging (400-1000nm) spectral data of Dunhuang soil sample;
FIG. 2 is the near infrared hyperspectral imaging (1000-;
FIG. 3 is the visible light hyperspectral imaging (400-1000nm) spectrum preprocessing data of Dunhuang soil sample obtained by the method of combining Savitzky-Golay smoothing and standard normal transformation;
FIG. 4 is the near infrared hyperspectral imaging (1000-;
FIG. 5 is a characteristic wave band of the visible light hyperspectral imaging (400-;
FIG. 6 is a characteristic band of near-infrared hyperspectral imaging (1000-;
FIG. 7 is a schematic diagram of the operation of a continuous projection algorithm;
FIG. 8 is a schematic diagram of an algorithm for establishing a prediction model for characteristic spectrum data based on visible light hyperspectral imaging (400-1000nm) by adopting partial least squares regression;
FIG. 9 is an algorithm diagram of a prediction model established by using support vector regression on characteristic spectrum data based on near-infrared hyperspectral imaging (1000-;
FIG. 10 is the result of predicting the water content of Dunhuang soil sample based on visible light hyperspectral imaging (400-;
FIG. 11 shows the result of predicting the water content of Dunhuang soil sample based on near-infrared hyperspectral imaging (1000-.
Detailed Description
The present invention is further illustrated by the following examples, which are to be understood as merely illustrative of, and not restrictive on, the present invention. The following exemplarily illustrates the method for detecting the surface water content of the cultural relic on the basis of hyperspectral imaging.
A sample is collected. Samples which are (basically) consistent with the composition of the cultural relics of the earthen archaeological site are collected at the periphery of the cultural relics of the earthen archaeological site. Before sampling, impurities such as plant branches and stems, gravels and the like which have large interference on sample components are preferably removed from the periphery of the cultural relic in the earthen site. Meanwhile, the flatness of the sample is kept basically consistent with the surface of the earthen archaeological site cultural relic as much as possible, so that the later data can be conveniently collected. And mixing additional water into the sample, thereby configuring a plurality of samples with different external water contents, namely gradient standard samples. The gradient here means that the applied water content of different gradient samples shows a gradient distribution. The water content applied is the mass of water applied (g)/the oven dry mass of the sample (g). The oven dry mass of the sample may be the oven dry mass of the sample at 105 ℃. In some embodiments, the applied moisture content of the gradient standard sample is between 0 and 0.45. The number of standard samples can be selected according to actual needs. For example, the number of the standard samples may be 100. In order to ensure that the sample is fully mixed with water, the dried earthen site cultural relic sample is placed in a sealed sample bag, stirred and sprayed with deionized water with certain mass.
And collecting visible light-near infrared hyperspectral imaging data of the gradient standard sample. In the test, the wet sample added with water is used for visible light-near infrared hyperspectral imaging. Preferably, the visible light-near infrared hyperspectral imaging data (spectral data) of the standard sample are arithmetically averaged to obtain actual reflectance spectral data of the standard sample. Specifically, the method comprises the steps of collecting visible light-near infrared hyperspectral imaging data of the gradient standard samples, selecting an interested area at the middle position of each standard sample, and carrying out arithmetic mean on a spectral curve to obtain actual reflectivity spectral data of each standard sample. For example, the hyper-spectral imaging system data analysis software HSI Analyzer is used to derive an arithmetic mean of selected spectral data.
And preprocessing the visible light-near infrared hyperspectral imaging spectral data by adopting one or more combinations of Savitzky-Golay smoothing (convolution balance algorithm), standard normal variation, first-order derivative and second-order derivative. The purpose of preprocessing is to reduce interference from external factors and to improve reliability and accuracy of the model. Preferably, the standard sample reflectivity data based on visible light hyperspectral imaging (400-; standard sample reflectance data based on near-infrared hyperspectral imaging (1000-.
The operation of preprocessing the reflectivity data of the standard sample based on visible light hyperspectral imaging (400-1000nm) by adopting the combination of Savitzky-Golay smoothing and standard normal variation is as follows: firstly, two groups of data acquired by hyperspectral imaging are imported into The Unscramble X10.4 software, wherein The column value is The spectral reflectivity of each wave band, and The row value is The number of samples. In the parameters of the Savitzky-Golay smoothing method, because the data is linear, the polynomial degree is selected to be 1, the number of smoothing points is 9, and two groups of Savitzky-Golay smoothed data are obtained. And then, respectively carrying out standard normal transformation and second derivative preprocessing on the two groups of pre-data.
The standard normal transformation principle is as follows:
wherein x i For the smoothed spectral reflectance of the ith sample Savitzky-Golay,
is the average spectral reflectivity of the ith sample after Savitzky-Golay smoothing, m is the number of wavelength points, x i，snv Namely the data preprocessed by combining Savitzky-Golay smoothing and standard normal transformation.
Similarly, the operation of preprocessing the standard sample reflectivity data based on the near-infrared hyperspectral imaging (1000-: performing second derivative operation on the data after Savitzky-Golay smoothing based on near-infrared hyperspectral imaging, wherein the operation formula is as follows:
wherein the original spectrum is x n The wavelength point is n, and the wavelength point is n,the difference width is G, the invention takes 2 to obtain data x after Savitzky-Golay smoothing and second derivative combination preprocessing n,2nd 。
The pretreatment was carried out using The Unscamblebler X10.4 software. In the specific implementation mode, different preprocessing methods are adopted to establish a model, and a better preprocessing method is preferred according to the decision coefficient and the root mean square error.
And extracting the characteristic wave band of the preprocessed visible light-near infrared hyperspectral imaging data. The characteristic wave band extraction can be carried out on the preprocessed visible light-near infrared hyperspectral imaging data by adopting methods such as a continuous projection algorithm, principal component analysis and the like. By establishing a continuous projection algorithm, a principal component analysis method and other feature extraction methods, visible light and near infrared feature bands are extracted, the feature bands cover effective information in hyperspectral data, the variable number of spectra is reduced, the complexity of a prediction model of the water content of a sample on the surface of an earthen site is reduced, and the operation efficiency and the prediction precision are improved. And (3) establishing a continuous projection algorithm for extracting characteristic wave bands by adopting Matlab software, wherein the principle is that the number of the characteristic wave bands is optimized to obtain the minimum Root Mean Square Error (RMSE), and the fitting effect of the model is optimal at the moment. The continuous projection algorithm is a forward-circulating characteristic wavelength extraction method, and a matrix X is set n×p Where n is the number of samples and p is the spectral wavelength. N is the number of extracted characteristic wavelengths, X k(0 ) For the initial iteration vector, the specific operation of the algorithm is as follows:
step 1: before iteration begins, one column j of the spectrum matrix is selected randomly, and the j-th column of the modeling set is assigned to x j Is marked as X k(0) ；
Step 2: the set of remaining column vector positions is denoted as S,
and step 3: separately calculate x j Projection of the remaining column vectors:
and 4, step 4: note the book
And 5: memory
Step 6: let n equal n +1, if n<N, returning to the first step; the resulting wavelength is { x k(n) (ii) a N is 0, …, N-1 }. And (3) performing multiple linear regression after one cycle corresponding to each k (0) and N to obtain a Root Mean Square Error (RMSE), wherein the k (0) and the N corresponding to the minimum root mean square error are optimal values.
And establishing a prediction model of the surface water content of the cultural relics in the earthen archaeological site based on the gradient standard sample. Establishing a relation between characteristic spectrum data of a gradient standard sample and an external water content, establishing an earthen site cultural relic surface water content prediction model, and inputting visible light-near infrared hyperspectral imaging data of a sample to be detected of the earthen site cultural relic into the prediction model to obtain the predicted water content of the sample to be detected of the earthen site cultural relic. Preferably, the reflectivity spectrum data of the sample to be tested of the earthen ruins cultural relics is input into the prediction model.
The prediction model of the surface water content of the earthen site can be established by adopting methods such as partial least squares regression, support vector regression and the like. Preferably, a prediction model is established on the characteristic spectrum data based on the visible light hyperspectral imaging (400-. The establishment of the prediction model is realized in Matlab software.
The method comprises the following specific operation steps of:
step 1: the spectral reflectivity matrix X and the water content matrix Y are decomposed, and the model is as follows: x ═ TP + E, Y ═ UQ + F, where T and U are the scoring matrices of the X and Y matrices, respectively; p and Q are respectively the load matrix of X and Y matrix; e and F are partial least squares regression fitting residual matrixes of the X matrix and the Y matrix respectively;
step 2: linear regression analysis was performed on T and U, which is expressed as: u ═ UTB，B＝(T T T) -1 T T Y；
And step 3: obtaining a predicted moisture content value according to the formula of step 1: y is Is unknown ＝T Is unknown BQ, wherein, T Is unknown Spectral matrix X for unknown sample Is unknown The scoring matrix of (2).
The support vector regression algorithm comprises the following specific operation steps:
step 1: support vector regression is through a non-linear function
The input space X is projected to a higher dimensional space F, so the main task of this method is to model the sample set D { (X) according to the spectrum 1 ,y 1 ),…,(x n ,y n )}，x n Is the spectral reflectance, y n For water content, find the linear function equation f (x) w T x + b, w and b are model parameters such that f (x) is as close as possible to y.
Step 2: xi is introduced in the model as a sensitive factor, assuming that a maximum bias of xi between f (x) and y is tolerated. From this model a convex optimization problem is transformed, i.e.
Where C is the error penalty factor and L (-) is the loss function. Let the relaxation variable xi i And
and introducing the formula, adopting a re-weighted quadratic programming optimization method, and introducing a Lagrangian multiplier to construct a Lagrangian function. The solution to support vector regression is finally obtained as:
and step 3: the kernel function of the invention adopts a Gaussian Radial Basis Function (RBF), and the specific form is as follows: k (x, x) i )＝exp(-γ||x-x i || 2 )。
The performance of support vector regression at this time is determined by the error penalty factor C and the gaussian radial basis function width γ. The generalization capability of the model is reduced due to the fact that the error penalty factor C is too large or too small, the Gaussian radial basis function width gamma reflects the degree of directional correlation between the support vectors, and the accuracy of the model is also affected due to the fact that the error penalty factor C is too large or too small. The invention adopts a particle swarm optimization algorithm to obtain optimal parameters: c is 11 and γ is 0.0039.
In some technical schemes, a standard sample for visible light-near infrared hyperspectral imaging acquisition is divided into a modeling sample and a verification sample. The quantity proportion of the modeling samples and the verification samples can be adjusted according to needs. For example, the number ratio of modeling samples to validation samples is 7: 3. The modeling samples are used to build a water cut prediction model. The validation samples are used to validate the prediction model. Preferably, the external water content of the modeling sample and the external water content of the verification sample are distributed in an equal numerical value in the external water content distribution interval of the standard sample.
Determining the coefficient R according to the following formula 2 And the root mean square error RMSE determines the degree of fit of the model. R 2 The closer to 1, the better the correlation of the prediction model is, the higher the fitting degree is, the more reliable the model is; the smaller the RMSE, the higher the stability of the model. The calculation formula is as follows:
in the above formula, y i Is a predicted value of the water content of the sample, y is a true value of the water content of the sample,
the average value of the true water content of the sample is shown, and n is the number of the samples. The true sample moisture content value is (sample mass-sample oven dry mass)/sample oven dry mass.
The invention provides a method for detecting the surface water content of an earthen site based on visible light-near infrared hyperspectral imaging, which is efficient, convenient and fast and has similar guiding significance for the surface water content of other earthen sites based on visible light-near infrared hyperspectral imaging.
The present invention will be described in detail by way of examples. It is also to be understood that the following examples are illustrative of the present invention and are not to be construed as limiting the scope of the invention, and that certain insubstantial modifications and adaptations of the invention by those skilled in the art may be made in light of the above teachings. The specific process parameters and the like of the following examples are also only one example of suitable ranges, i.e., those skilled in the art can select the appropriate ranges through the description herein, and are not limited to the specific values exemplified below.
Typical Dunhuang soil is taken as an earthen relic sample, and a method for detecting the water content of the earthen relic sample based on visible light-near infrared hyperspectral imaging is explained in detail by combining with figures 1 to 11.
Step 1: the Dunhuang soil sample is prepared into (0-0.45) standard samples with different external water contents according to a gradient. In order to ensure that the Dunhuang soil sample is fully mixed with water, the Dunhuang soil dried under the condition of 105 ℃ is placed in a sealed sample bag, stirred and sprayed with deionized water with certain mass. The water content applied is the mass of water applied/mass of dunhuang soil sample dried at 105 ℃. The number of standard samples was 100. The prepared standard sample is weighed and recorded, and is placed for 24h for standby.
Step 2: the hyperspectral imaging is divided into two wave bands, namely visible light hyperspectral imaging (400 + 1000nm) and near-infrared hyperspectral imaging (1000 + 2500 nm). And respectively adopting visible light hyperspectral imaging (400-1000nm) and near-infrared hyperspectral imaging (1000-2500nm) to collect data of the prepared Dunhuang soil standard sample. And (3) after the collection is finished, drying the Dunhuang soil sample in a 105 ℃ oven for 12 hours, calculating the true value of the water content of the sample, and recording. True moisture content value (sample mass-sample oven dry mass)/sample oven dry mass.
And analyzing the data of each Dunhuang soil sample based on visible light-near infrared hyperspectral imaging, selecting an interested area in the middle position of each sample, and arithmetically averaging a spectral curve to obtain the actual reflectivity spectrum data of each Dunhuang soil sample. The reflectivity data of the dunghuang soil sample based on visible light hyperspectral imaging is shown in figure 1. The reflectivity data of the Dunhuang soil sample based on near-infrared hyperspectral imaging is shown in figure 2.
And step 3: the reflectivity data of the Dunhuang soil sample obtained in the step 2 based on visible light hyperspectral imaging (400-1000nm) is preprocessed by adopting a method of combining Savitzky-Golay smoothing and standard normal variation, and the result is shown in figure 3. Meanwhile, data preprocessing is carried out on the reflectivity data of the Dunhuang soil sample based on near-infrared hyperspectral imaging (1000-.
And 4, step 4: and (3) performing characteristic length extraction on the spectrum preprocessing data based on visible light hyperspectral imaging (400-1000nm) obtained in the step (3) by adopting a continuous projection algorithm, and when the maximum characteristic wavelength is set to be 18, obtaining the minimum root mean square error (RMSE ═ 0.0163) at the moment, wherein the characteristic wavelengths are 414.384nm, 523.410nm, 629.895nm, 788.908nm and 932.025nm respectively, and the specific result is shown in FIG. 5. And (3) performing characteristic length extraction on the spectrum preprocessing data based on the near-infrared hyperspectral imaging (1000-2500nm) obtained in the step (3) by adopting a continuous projection algorithm, and when the maximum characteristic wavelength is set to be 8, obtaining the minimum root mean square error (RMSE ═ 0.0133) at the moment, wherein the characteristic wavelengths are 1149.251nm, 1233.7nm, 1464.197nm, 2214.801nm and 2262.161nm respectively, and the specific result is shown in fig. 6. The continuous projection algorithm was performed in Matlab software, and the running algorithm is shown in fig. 7.
And 5: dividing 100 Dunhuang soil standard samples for visible light-near infrared hyperspectral imaging acquisition into a modeling sample and a verification sample. 70 Dunhuang soil samples were selected as modeling samples and the remaining 30 samples were selected as validation samples. A prediction model is established for characteristic spectrum data based on visible light hyperspectral imaging (400-1000nm) by adopting partial least squares regression, and a model algorithm is shown in FIG. 8. A support vector regression is adopted to establish a prediction model for the characteristic spectrum data based on the near-infrared hyperspectral imaging (1000-. Both models were run using Matlab software.
Step 6: and analyzing and evaluating the result of the prediction model. A prediction model established based on characteristic spectrum data of visible light hyperspectral imaging (400-1000nm) and known external water content, wherein the model is used for determining a coefficient R 2 0.905 rms error EMSE 0.0359, and verification determination factor R 2 The root mean square error RMSE is 0.0438, and the specific results are shown in fig. 10. A prediction model established based on the characteristic spectrum data of the near-infrared hyperspectral imaging (1000- 2 0.917, 0.0327, and the determination coefficient R 2 The root mean square error RMSE is 0.0354, and the specific results are shown in fig. 11.
The model accuracy is determined by a coefficient of determination R 2 And performing comprehensive evaluation on the RMSE, wherein the larger the decision coefficient of the model is, the more stable the model is, and the smaller the RMSE is, the better the prediction capability of the model is. According to the conclusion obtained by experiments, the model established by the partial least squares regression method for the characteristic spectrum data of visible light hyperspectral imaging (400 + 1000nm) and the model established by the support vector regression method for the characteristic spectrum data of near infrared hyperspectral imaging (1000 + 2500nm) both have ideal water content prediction effects. Therefore, a prediction model of the surface water content of the earthen site can be established by adopting visible light-near infrared hyperspectral imaging, the prediction model is used for quickly detecting the water content of the sample on the surface of the earthen site in a nondestructive and quick manner, and the determination coefficient and the root mean square error both meet the requirements.
And other data preprocessing methods are combined to establish different water content prediction models. RAW is RAW data without any processing; SG is Savitzky-Golay smoothing; SG + FD is a Savitzky-Golay smoothing and first derivative combination; SG + SD is a Savitzky-Golay smoothing and second derivative combination; SG + MSC is a combination of Savitzky-Golay smoothing and multivariate scattering correction; SG + SNV is a combination of Savitzky-Golay smoothing and standard normal transformation.
Table 1 results of water content prediction models using different preprocessing methods based on visible light hyperspectral imaging
TABLE 2 results of water content prediction models based on near-infrared hyperspectral imaging using different pre-processing methods
Wherein R is C 2 To model the absolute coefficients of a sample, RMSE C /% is the root mean square error, R, of the modeled samples p 2 To verify the absolute coefficients of a sample, RMSE p /% is the root mean square error of the validation samples. As can be seen from tables 1 and 2, when the standard sample reflectivity data based on visible light hyperspectral imaging is preprocessed by adopting the combination of Savitzky-Golay smoothing and standard normal variation, and the standard sample reflectivity data based on near-infrared hyperspectral imaging is preprocessed by adopting the combination of Savitzky-Golay smoothing and second-order derivative, the verification set absolute coefficient and root-mean-square error data show that the water content prediction model has the best effect at the moment.
Claims (10)
1. A method for detecting the surface water content of cultural relics in an earthen site based on hyperspectral imaging is characterized by comprising the following steps:
step (i): collecting samples around the cultural relics in the earthen site, adding water to the samples, uniformly mixing the samples, and preparing the samples into gradient standard samples with different external water contents;
step (ii): collecting visible light-near infrared hyperspectral imaging data of the gradient standard sample;
step (iii): extracting characteristic wave bands of the visible light-near infrared hyperspectral imaging data to obtain characteristic spectrum data;
step (iv): establishing a relation between characteristic spectrum data of a gradient standard sample and an external water content, establishing an earthen site cultural relic surface water content prediction model, and inputting visible light-near infrared hyperspectral imaging data of a sample to be detected of the earthen site cultural relic into the prediction model to obtain the predicted water content of the sample to be detected of the earthen site cultural relic.
2. The method as claimed in claim 1, wherein the characteristic spectrum data of the visible light hyperspectral imaging is spectrum data of 400-1000nm band; the characteristic spectral data of the near-infrared hyperspectral imaging is the spectral data of a band of 1000-2500 nm.
3. The method according to claim 1 or 2, characterized in that the method further comprises: and preprocessing the visible light-near infrared hyperspectral imaging spectral data before extracting the characteristic wave bands of the visible light-near infrared hyperspectral imaging data.
4. The method of claim 3, wherein the visible-near infrared hyperspectral imaging spectral data is preprocessed with a combination of one or more of Savitzky-Golay smoothing, standard normal variation, first derivative, and second derivative; preferably, the visible light hyperspectral imaging data, especially the reflectivity data, of the standard sample are preprocessed by adopting a combination of Savitzky-Golay smoothing and standard normal variation; the near-infrared hyperspectral imaging data, especially the reflectivity data, of the standard sample are preprocessed by adopting a combination of Savitzky-Golay smoothing and second-order derivative.
5. The method according to any one of claims 1 to 4, wherein the spectral curves of visible light-near infrared hyperspectral are arithmetically averaged at selected areas in each standard sample to obtain actual reflectance spectral data for each standard sample for preprocessing.
6. The method of claim 5, wherein the visible-near infrared hyperspectral imaging data is subjected to characteristic band extraction using a combination of one or more of a continuous projection algorithm and a principal component analysis method.
7. The method according to any one of claims 1 to 6, characterized in that a prediction model of the surface water content of the cultural relics of the earthen site is established by adopting a partial least squares regression method and/or a support vector regression method; preferably, a prediction model is established for the characteristic spectrum data based on visible light hyperspectral imaging by adopting partial least squares regression, and a prediction model is established for the characteristic spectrum data based on near-infrared hyperspectral imaging by adopting support vector regression.
8. The method of any one of claims 1 to 7, wherein the earthen archaeological relic is a clay-based relic.
9. Method according to any one of claims 1 to 8, characterized in that the applied water content is the applied water mass/oven dry mass of the sample.
10. The method of any one of claims 1 to 9, wherein the applied moisture content of the gradient standard sample is distributed between 0 and 0.45.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490117.7A CN114878508A (en) | 2022-05-07 | 2022-05-07 | Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490117.7A CN114878508A (en) | 2022-05-07 | 2022-05-07 | Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114878508A true CN114878508A (en) | 2022-08-09 |
Family
ID=82674648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210490117.7A Pending CN114878508A (en) | 2022-05-07 | 2022-05-07 | Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114878508A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252875A (en) * | 2023-11-17 | 2023-12-19 | 山东大学 | Medical image processing method, system, medium and equipment based on hyperspectral image |
-
2022
- 2022-05-07 CN CN202210490117.7A patent/CN114878508A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117252875A (en) * | 2023-11-17 | 2023-12-19 | 山东大学 | Medical image processing method, system, medium and equipment based on hyperspectral image |
| CN117252875B (en) * | 2023-11-17 | 2024-02-09 | 山东大学 | Medical image processing method, system, medium and equipment based on hyperspectral image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shi et al. | Development of a national VNIR soil-spectral library for soil classification and prediction of organic matter concentrations | |
| Jin et al. | Comparison of different satellite bands and vegetation indices for estimation of soil organic matter based on simulated spectral configuration | |
| CN106918565B (en) | Heavy metal-polluted soil Cd content Inverse modeling and its spectral response characteristics wave band recognition methods based on indoor standard specimen bloom spectrum signature | |
| CN108152235B (en) | Heavy metal content inversion method combining soil indoor and outdoor spectra | |
| CN114018833B (en) | Method for estimating heavy metal content of soil based on hyperspectral remote sensing technology | |
| CN101231274B (en) | Method for rapid measuring allantoin content in yam using near infrared spectrum | |
| CN110118742A (en) | Land table Vegetation canopy ecology water content remote sensing inversion method based on spectrum analysis | |
| CN103712923A (en) | Method for eliminating moisture influence factor in field in-situ soil measurement spectrums | |
| CN110687053A (en) | Regional organic matter content estimation method and device based on hyperspectral image | |
| Xuemei et al. | Using short wave visible–near infrared reflectance spectroscopy to predict soil properties and content | |
| CN114676636B (en) | Grassland area soil moisture rapid inversion method integrating vegetation and habitat characteristics | |
| CN114878508A (en) | Hyperspectral imaging-based method for detecting surface water content of cultural relic in earthen site | |
| Yu et al. | Prediction of soil properties based on characteristic wavelengths with optimal spectral resolution by using Vis-NIR spectroscopy | |
| CN107748144A (en) | The middle infrared spectrum detecting system of quick measure SOIL CARBON AND NITROGEN and its stable isotope | |
| CN113670913A (en) | Construction method for inverting hyperspectral vegetation index by using nitrogen content of rice | |
| CN111595806A (en) | Method for monitoring soil carbon component by using mid-infrared diffuse reflection spectrum | |
| CN113049526B (en) | Corn seed moisture content determination method based on terahertz attenuated total reflection | |
| CN115392120A (en) | LSTM-based soil organic matter content high-spectrum modeling method | |
| Yan et al. | Hyperspectral response and monitoring study of soil moisture content based on the optimized spectral index | |
| Yu et al. | Improving SPAD spectral estimation accuracy of rice leaves by considering the effect of leaf water content | |
| CN111521568A (en) | Soil water content prediction method based on spectrum angle | |
| Zhang et al. | Design and planning of urban ecological landscape using machine learning | |
| Jing et al. | Machine learning for estimating leaf dust retention based on hyperspectral measurements | |
| CN114235715B (en) | Full phosphorus content nondestructive monitoring method for karst plant leaves | |
| CN115392305A (en) | Soil organic matter content high spectrum modeling method based on improved time convolution network |
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 |