Method for detecting phosphate content of mural ground layer based on three-band spectrum
Through the three-band spectral index and fractional differential treatment, the sampling damage and double-band index saturation problems in mural salt content detection are solved, and the lossless and accurate phosphate content detection is achieved, which improves the adaptability and accuracy of the model. Effective protection of cultural heritage.
Patent Information
- Application Number
- CN202411129753.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-16
- Publication Date
- 2024-11-12
AI Technical Summary
The existing mural salt content detection technology relies on sampling analysis, which has the risk of physical damage to precious murals. The dual-band spectral index has saturation phenomena and accuracy limitations. The third characteristic band selection lacks innovation and data-driven analysis, resulting in the model in complex applications. Inadequate adaptability and accuracy.
Using the three-band spectral index, through fractional differential processing and partial least squares regression model, the optimal spectral index and band are selected, and the three-band spectral index PTSR is established, which is used to detect the phosphate content of murals, avoid saturation, and improve detection. Accuracy and anti-interference ability.
The non-destructive and accurate detection of phosphate content in murals has been achieved, which enhances the adaptability and accuracy of the model, and can identify salt damage problems in early stages, reduce emergency restoration needs, reduce protection costs, and improve the efficiency of cultural heritage protection.
Smart Images

Figure CN118937249A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of cultural relics protection, and in particular to a method for detecting phosphate content in mural ground based on a three-band spectrum. Background Art
[0002] As a national treasure-level cultural heritage, my country's Dunhuang murals are famous for their rich themes, exquisite craftsmanship and unique historical value. Unfortunately, due to their age, the murals are affected by the environment, resulting in fading, alkali, wind and sand erosion, mildew and other diseases. The separation of the pigment layer and the base layer of the murals, and the separation of the base layer and the cliff support layer, have seriously affected the beauty and historical value of the murals. In the process of protecting these irreplaceable cultural heritage, it is crucial to understand the phosphate content in the mural base for the prevention and control of salt damage. Phosphates can form dodecahydrates under specific hydrothermal conditions, which pose a greater threat to the cultural heritage of murals than common sulfates.
[0003] At present, the quantitative detection of the salt content of murals is mainly based on sampling analysis. In related studies, Du Hongying and others used capillary electrophoresis (CE), also known as high-performance capillary electrophoresis (HPCE), but this method requires field sampling, which will cause certain damage to the murals. When studying the salt content of murals, hyperspectrometers can reveal important information on chemical composition, spatial distribution, content estimation, crystallization monitoring, environmental impact, and non-destructive testing in the form of spectral reflection and absorption.
[0004] Related research shows that the use of spectral index is a simple and effective method to measure the surface properties of murals, which can effectively avoid the physical damage to cultural heritage that may be caused by traditional sampling and analysis, and the band optimization algorithm is widely used in the development of hyperspectral technology. Compared with one-dimensional spectral data, this method can provide more spectral features and enhance the relationship between soil properties and spectral features. For example, in the study Jia, P.; Zhang, J.; He, W.; Hu, Y.; Zeng, R.; Zamanian, K.; Jia, K.; Zhao, X., Combination of hyperspectral and machine learning to invert soil electrical conductivity. Remote Sensing. 2022, 14, 2602, Lihan Chen et al. discussed the application effect of visible-near infrared spectroscopy technology in the estimation of soil heavy metal content. By collecting 120 soil samples in Xuzhou City, Jiangsu Province, the heavy metal content and spectral characteristics of the samples were measured. The fractional derivative spectral preprocessing method (FOD) and a new three-band spectral index were used to enhance the spectral information of soil heavy metals, while solving the problems of collinearity and redundancy in hyperspectral data. In another study, Dong, T.; Wu, B. In Estimate faction of photosynthetically active radiation with three-band vegetation indices based on HJ-CDD satellite in wheat, 2012 First International Conference on Agro-Geoinformatics (Agro-Geoinformatics), Ieee: 2012; pp 1-5, Jing Yuan et al. proposed a soil organic matter content estimation method based on the improved Hapke model. This method is based on the radiation transmission process of soil reflectance spectrum, and constructs spectral index by converting reflectance and single scattering reflectance, so as to estimate soil organic matter content in remote sensing. According to the above analysis results, spectral index can effectively capture the high spectral response of components in soil. By expressing them in two-dimensional or multi-dimensional spectral space, they reduce the interference of other soil components on the estimation results. In addition, spectral index can enhance the weak correlation between bands, simplify the model structure, and eliminate redundant information variables. Therefore, the application of spectral index to quantitatively analyze the interaction between characteristic bands of phosphate in mural ground can significantly improve the accuracy of phosphate content estimation model.Existing studies have used hyperspectral data to study the relationship between soil reflectance and soil composition through different spectral index combination methods (such as normalization, difference and ratio, etc.). Compared with the two-band spectral index, the three-band spectral index adds a third band in a specific sensitive area. This processing method can often improve the estimation accuracy of the index, enhance the anti-interference ability, eliminate the saturation phenomenon of the commonly used two-band index, and make the estimation of soil composition more robust and accurate.
[0005] In summary, the prior art has the following problems:
[0006] (1) At present, the quantitative detection technology for the salt content of murals mainly relies on sampling analysis, including the use of Merckoquant qualitative test paper method, capillary electrophoresis (CE) and ion chromatography. Although these technologies can analyze the salt damage and environmental impact of murals in detail, they all require direct sampling from the murals. This sampling process may cause certain physical damage to the precious murals. Therefore, these methods have potential risks to the protection of cultural heritage in practical application.
[0007] (2) The current spectral inversion model used for mural salt content analysis is mainly based on dual-band spectral index. However, the dual-band method has limitations such as saturation. Moreover, the model that only relies on the dual-band spectral index cannot fully realize the potential of spectral analysis in terms of accuracy and anti-interference.
[0008] (3) In previous studies, the selection of the third characteristic band relied on previous research results and lacked sufficient innovation and data-driven analysis. Traditional methods failed to fully utilize the potential of fractional differentials, resulting in a lack of systematicity and scientificity in the selection of band positions and orders, which limited the improvement of the model on a theoretical and empirical basis. In addition, due to the lack of in-depth analysis of the differential characteristics and band sensitivity of spectral data, the adaptability and accuracy of the model in practical application scenarios have not been effectively enhanced. Therefore, the existing technology has failed to provide a model optimization path that is both accurate and practical, making it impossible to guarantee effectiveness and reliability when dealing with complex practical application problems.
[0009] Therefore, there is still no solution to accurately and non-destructively quantitatively detect the phosphate content in murals without the drawbacks of the dual-band method. Summary of the invention
[0010] The purpose of the present invention is to provide a method for detecting the phosphate content of mural floor coating based on a three-band spectral index. Based on the establishment of the third-band spectral index, the saturation problem of the traditional two-band index in analyzing the salt content of mural floor coating is effectively avoided, thereby realizing accurate and non-destructive detection of the phosphate content of mural floor coating.
[0011] To achieve the above object, the present invention provides the following solutions:
[0012] A method for detecting phosphate content in mural floor coverings based on a three-band spectral index comprises the following steps:
[0013] S1. Make mural ground sample;
[0014] S2, measuring the spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, and preprocessing the spectral reflectance data;
[0015] S3, performing fractional differential processing on the spectral reflectance data, and calculating the Pearson correlation coefficient between the PNDI and PSR spectral index of the spectral reflectance data and the phosphate content of the mural ground, and obtaining the optimal spectral index with the highest correlation with the phosphate content;
[0016] S4. Using the optimal spectral index and the corresponding conductivity to establish partial least squares regression models PSR-PLSR and PNDI-PLSR, by comparing the accuracy of the two models, the optimal model and its corresponding specific band are selected, and then the basic index R for constructing the three-band spectral index PTSR is determined. c , which is used as the basis for establishing a three-band spectral index;
[0017] S5. Based on the GL algorithm, a differential expression of v-order fractional differential is obtained, and the differential expression and the expression of the PTSR index are applied to calculate the three-band spectral index under different fractional differentials;
[0018] S6. Using the three-band spectral index as the explanatory variable and the corresponding mural floor conductivity as the response variable, a mural floor phosphate content detection model is established, and it is verified using the validation set. The verified mural floor phosphate content detection model is obtained and used for the mural floor phosphate content measurement.
[0019] Further, in S3, the Pearson correlation coefficients between the PNDI and PSR spectral indexes of the spectral reflectance data and the phosphate content of the mural ground are calculated respectively to obtain the optimal spectral index with the highest correlation with the phosphate content, specifically including:
[0020] S3.1. Construct PSR spectral index:
[0021] Construct the PSR spectral index:
[0022] Among them, R a and R b Both represent the spectral reflectance at a certain wavelength;
[0023] S3.2. Use the formula of the PSR spectral index and the PNDI spectral index to calculate the spectral index of the spectral reflectance data in the verification set, perform univariate regression analysis on the spectral index with the largest absolute value of the Pearson correlation coefficient at each order and the phosphate content of the mural floor, and screen out the spectral index with the optimal order at each order.
[0024] Furthermore, in S4, the partial least squares regression models PSR-PLSR and PNDI-PLSR are established using the optimal spectral index and the corresponding conductivity, specifically including:
[0025] The spectral index of the best order among the first six orders was taken as the explanatory variable, and the corresponding conductivity was taken as the response variable to establish the PNDI-PLSR model and the PSR-PLSR model.
[0026] Furthermore, the three-band spectrum index is:
[0027] In the formula, R a and R b Respectively represent the spectral reflectance of specific band a and specific band b in the range of 400-2400nm, R c It is a specific band selected based on the optimal model.
[0028] Further, the S2, measuring the original spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, specifically includes:
[0029] S2.1. After subjecting the mural flooring samples to phosphate corrosion degradation treatment of different concentrations and drying in the shade, the surface spectral reflectance thereof is measured to obtain original spectral reflectance data under different corrosion conditions;
[0030] S2.2. For mural samples that have been degraded by phosphate erosion, the conductivity is measured using a soil salt conductivity tester to obtain conductivity data under different erosion conditions.
[0031] Furthermore, the method also includes, after measuring the conductivity of the mural floor sample under different erosion conditions, performing a statistical description of the density distribution, mean, standard deviation, minimum value, quartile and coefficient of variation of the conductivity, wherein the coefficient of variation calculation formula is: In the formula, σ represents the sample standard deviation and u represents the sample mean.
[0032] Furthermore, the S1, preparing the mural ground sample, specifically comprises:
[0033] S1.1. Desalination treatment of the clay, sand and wheat straw used to make the mural ground sample;
[0034] S1.2. Evenly mix the desalted clay, sand and wheat straw in a mass ratio of 64:36:3, then add 20% of the solid mass of deionized water, and make a mural floor sample by molding the sample blocks.
[0035] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the method for detecting phosphate content in mural flooring based on three-band spectral index provided by the present invention has the following advantages:
[0036] (1) Non-invasive detection: Non-contact spectroscopy technology is used to avoid physical damage to precious murals while providing accurate chemical composition analysis.
[0037] (2) On-site simulation of salt damage process: Different from the traditional laboratory preset salt content method, the present invention can more realistically reflect the salt damage of murals by simulating the salt damage process in the actual environment of murals - capillary salt ion adsorption and crystallization erosion degradation process, making the research results closer to the actual situation and improving the practical value of the present invention. This method not only enables the research results to more effectively guide the protection and restoration of murals on site, but also increases the applicability of the research and the effectiveness of field operations.
[0038] (2) Ability to eliminate saturation phenomenon: By adding a third band, the three-band salt spectral index of the present invention can capture spectral information more comprehensively than the traditional two-band index. This improvement reduces information loss and the saturation problem of the traditional two-band index in analyzing the salt content of murals, improves the accuracy of phosphate content detection and the robustness of the model, thereby enhancing the model's resistance to environmental interference.
[0039] (3) Optimizing the selection of bands and differential orders: Based on in-depth data-driven analysis, the selection of the third characteristic band is improved, and fractional differential techniques are systematically applied. This optimization not only makes the band selection more scientific and systematic, but also significantly improves the performance of the model on both theoretical and empirical basis, and further enhances the adaptability and accuracy of the model in complex environments.
[0040] (4) Through proactive and regular technical testing, potential degradation problems of cultural heritage can be identified early and timely intervention can be made. This preventive measure not only prolongs the life of cultural heritage, but also reduces the need for emergency repairs, thereby reducing the cost and complexity of protection and maintenance. The proactive and preventive nature of the present invention greatly improves the efficiency and effectiveness of cultural heritage protection. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0042] Figure 1 This is a flow chart of a method for detecting phosphate content in mural floor coverings based on a three-band spectral index according to an embodiment of the present invention;
[0043] Figure 2 A flowchart of spectral data collection and processing and establishment of a phosphate content detection model for murals and floor coverings in an embodiment of the present invention;
[0044] Figure 3 It is a graph of the partial least squares regression model of the dual-band spectral index according to an embodiment of the present invention;
[0045] Figure 4 Schematic diagram of the correlation coefficient between the three-band spectral index of different FOD orders and the phosphate content of the mural flooring in the embodiment of the present invention, wherein (a)-(f) respectively represent the correlation coefficient between the three-band spectral index and the phosphate content of the mural flooring when the differential order is 0.1, 0.2, 0.3, 0.4, 0.7, and 0.8;
[0046] Figure 5 A diagram showing the prediction results of the phosphate content detection model for mural flooring established in an embodiment of the present invention;
[0047] Figure 6 Schematic diagram of statistical analysis of EC value of mural floor coating in an embodiment of the present invention, wherein (a)-(e) represent the statistical description of data under five working conditions, namely, density distribution, mean, standard deviation (SD), minimum value (Min), first quartile (Q1), third quartile (Q3), and coefficient of variation (CV), and (f) represents the statistical analysis of all the total data of (a)-(e). DETAILED DESCRIPTION
[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0049] The purpose of the present invention is to provide a method for detecting the phosphate content of mural floor coating based on a three-band spectral index. Based on the establishment of the third-band spectral index, the saturation problem of the traditional two-band index in analyzing the salt content of mural floor coating is effectively avoided, thereby realizing accurate and non-destructive detection of the phosphate content of mural floor coating.
[0050] Terminology explanation:
[0051] Univariate regression refers to a method in which only one independent variable is related to one dependent variable. For variable selection, the Pearson correlation coefficient is used to measure the strength of the linear relationship between the independent variable and the dependent variable. In the present invention, the correlation coefficients of PNDI, PSR, PTSR spectral index and phosphate content of mural flooring are calculated. The absolute value of the correlation coefficient describes the linear correlation degree between two variables.
[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0053] like Figure 1-Figure 2 As shown, the method for detecting phosphate content in mural flooring based on three-band spectral index provided by an embodiment of the present invention comprises the following steps:
[0054] S1. Make mural ground sample;
[0055] S2, measuring the spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, and preprocessing the spectral reflectance data;
[0056] S3, performing fractional differential processing on the spectral reflectance data, and calculating the Pearson correlation coefficient between the PNDI and PSR spectral index of the spectral reflectance data and the phosphate content of the mural ground, and obtaining the optimal spectral index with the highest correlation with the phosphate content;
[0057] S4. Using the optimal spectral index and the corresponding conductivity to establish partial least squares regression models PSR-PLSR and PNDI-PLSR, by comparing the accuracy of the two models, the optimal model and its corresponding specific band are selected, and then the basic index R for constructing the three-band spectral index is determined. c , which is used as the basis for establishing a three-band spectral index;
[0058] S5. Based on the GL algorithm, a differential expression of v-order fractional differential is obtained (i.e., Formula 6), and the differential expression and the expression of the PTSR index (i.e., Formula 4) are applied to calculate the three-band spectral index under different fractional differentials;
[0059] S6. Using the three-band spectral index as the explanatory variable and the corresponding mural floor conductivity as the response variable, a mural floor phosphate content detection model is established, and it is verified using the validation set. The verified mural floor phosphate content detection model is obtained and used for the mural floor phosphate content measurement.
[0060] In order to improve the accuracy of the model, the data was preprocessed by standardization and mean centering. The model architecture is based on the Kernel PLS algorithm, and the upper limit of the number of principal components is set to seven. The optimal number of components was determined by random cross-validation, and specific model warning parameters were set (the thresholds of the calibration and validation residual variance ratios were set to 0.5 and 0.75, respectively, and the upper limit of the residual variance increase was set to 6%) to ensure the stability and reliability of the model.
[0061] In this embodiment, the S2, measuring the original spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, specifically includes:
[0062] S2.1. After subjecting the mural flooring samples to phosphate corrosion degradation treatment of different concentrations and drying in the shade, the surface spectral reflectance thereof is measured to obtain original spectral reflectance data under different corrosion conditions;
[0063] S2.2. For mural samples that have been degraded by phosphate erosion, the conductivity is measured using a soil salt conductivity tester to obtain conductivity data under different erosion conditions.
[0064] In this embodiment, the specific process of measuring the spectral reflectance data includes:
[0065] In order to study the hyperspectral characteristics of the mural flooring's response to phosphate concentration, the present invention designed five different concentrations of disodium hydrogen phosphate dodecahydrate solutions, which are 0.608, 0.808, 1.008, 1.208, and 1.408 mol / L, corresponding to the lowest to highest erosion conditions. A high-precision low-temperature constant temperature water tank is used to control the temperature of the sample, and the mural flooring sample is eroded at a constant temperature of 32.5°C by simulating the capillary absorption effect. After the sample is dried in the shade, the ASD-FieldSpec4HI-RES spectrometer is used to measure its spectral reflectance and collect hyperspectral data. The wavelength range is 350-2500nm, and the sampling interval is 1.4nm (350-1000nm) and 2nm (1001-2500nm). In order to avoid the influence of other light, the measurement is carried out under indoor closed conditions at night without other light sources. After the instrument is preheated for 30 minutes, the standard reflectance calibration is first performed. The 57 optical fibers that come with the instrument are embedded in the ASD Analytical (Model: A122317) high-brightness contact probe. The light source is a 70W quartz-tungsten-halogen lamp, and the aperture of the light outlet is 25mm. The light outlet is vertically inverted on the center surface of the sample, the light source switch is turned on, and the curve is collected after it stabilizes. The probe is rotated 90° parallel to the sample surface and the collection is repeated. The four directions are measured, and then the center point of the sample is measured again. The average value is taken as the sample spectrum data, and the whole process is completed in a dark room. In this embodiment, the five-point sampling method is also used for average processing, which helps to improve the stability and representativeness of the data, and ensures that the results are more reliable by smoothing accidental errors and outliers. This method not only reduces the impact of measurement errors, but also simplifies the subsequent data analysis process, making data processing more efficient. In this embodiment, in order to eliminate the interference that the salt in the raw materials may cause to the results, the raw materials are desalted. This step ensures that the conductivity measurement can accurately reflect the changes caused by phosphate, rather than the influence of other salts.
[0066] In this embodiment, the specific process of measuring the conductivity includes:
[0067] The determination of electrical conductivity is a key indicator used to assess the phosphate content in soil samples. Based on this, we proposed a hypothesis that changes in phosphate content in the sample will be closely related to changes in electrical conductivity. This is supported in the literature. Specifically, Xiaoguang Zhang (2019) and Xiangyu Ge et al. (2022) pointed out in their studies that the electrical conductivity in the soil solution can effectively reflect the total salt content in the soil. Although these studies did not directly link phosphate content to electrical conductivity, they provided strong evidence that electrical conductivity can reflect the salt content in the soil, thus supporting our hypothesis. The specific steps for measuring electrical conductivity are as follows: In the laboratory, the mural samples after phosphate erosion were sieved through a 2mm soil sieve. Soil leachate was prepared at a soil to water ratio of 1:5 and filtered at a room temperature of 25°C. Soil electrical conductivity (EC) in the soil leachate was measured using a soil salt EC conductivity tester.
[0068] In this embodiment, in order to reduce the impact of noise on spectral data analysis, the collected spectral reflectance data of the mural floor is preprocessed to eliminate high-frequency noise and outliers. This step includes removing the 350-399nm and 2401-2500nm bands with low signal-to-noise ratio, and fitting the signal with a local polynomial using a Savitzky-Golay filter as a sliding window, and then using the center point value of the polynomial as the smoothed output value, the formula is:
[0069]
[0070] Among them, y' i is the smoothed data point, y i+j is a point in the original data sequence, c j are the coefficients in the convolution kernel (determined by the fitting polynomial coefficients and derivative information), which are calculated by fitting a second-order polynomial least squares fit to the data in a local window; m is half the window size (in the case of a 21-point window, m=10); j represents an index from -m to m, used to traverse the data points around point i; Δ is usually 1 unless the intervals between data points are not uniform.
[0071] In recent years, many studies have focused on estimating soil components by spectral index, especially using dual-band spectral index. These indices help to reveal the external response and intrinsic meaning of spectral data, and can amplify the weak correlation between bands. At the same time, the phosphate ratio index (Phosphate Simple Ratio, PSR) and the phosphate normalized difference index (Phosphate Normalized Difference Index, PNDI) place the strongest reflection band and the weakest reflection band in the numerator and denominator respectively, and further expand the gap between the reflectivity values by ratio and normalized ratio calculation, so as to maximize the sensitivity of the spectral characteristics of the object to be measured. In the present embodiment, in the S3, the Pearson correlation coefficient between the PNDI, PSR spectral index of the spectral reflectance data and the phosphate content of the mural ground is calculated respectively, and the optimal spectral index with the highest correlation with the phosphate content is obtained:
[0072] S3.1. Construct PSR spectral index:
[0073] Construct the PSR spectral index:
[0074] Among them, Ra and Rb both represent the spectral reflectance at a certain wavelength;
[0075] S3.2. Use the formula of the PSR spectral index and the PNDI spectral index to calculate the spectral index of the spectral reflectance data in the verification set, and for the spectral index with the largest absolute value of the Pearson correlation coefficient at each order and the phosphate content of the mural floor, that is: select the spectral index corresponding to the largest absolute value of the correlation coefficient as the independent variable, and the phosphate content of the mural floor as the dependent variable, and perform univariate regression analysis in the ratio of 7:3 between the test set and the verification set, and screen out the spectral index of the optimal order at each order, to provide a basis for the subsequent establishment of a multivariate regression model.
[0076] In this embodiment, we proposed a new three-band spectral index, PTSR (Phosphate ThreeSimple Ratio, PTSR), which aims to improve the monitoring of phosphate content in murals by improving the estimation accuracy of the spectral index and enhancing its resistance to interference. The design process of this spectral index includes multiple steps. The first is to select the spectral index with the highest correlation with phosphate content by calculating the correlation coefficient between the phosphate content of the mural ground and the PSR spectral index and the PNDI spectral index. This process includes univariate regression analysis of the spectral index with the largest correlation coefficient at each order, so as to determine the optimal spectral index at each order, as well as the optimal order sorting.
[0077] In this embodiment, in S4, the partial least squares regression models PSR-PLSR and PNDI-PLSR are established using the optimal spectral index and the corresponding conductivity, specifically including:
[0078] By analyzing the spectral indices in the top six univariate regressions with the best performance, these indices were used as explanatory variables, and the corresponding conductivity was used as the response variable to conduct a partial least squares regression (PLSR) analysis of the ground phosphate content, that is, PSR-PLSR and PNDI-PLSR models were established respectively. The accuracy of the PSR-PLSR and PNDI-PLSR models was compared, and the best performing model was selected as the basis for the three-band spectral index study. c (indicates the band corresponding to the spectral index with the largest correlation coefficient with phosphate content) is included as a specific sensitive band to further optimize the design of the spectral index. The final three-band spectral index PTSR is defined as:
[0079]
[0080] In the formula, R a and R b Respectively represent the spectral reflectance of specific band a and specific band b in the range of 400-2400nm, R c It is a specific band selected based on the optimal model.
[0081] Therefore, the selection of specific three bands no longer simply relies on previous research results, but adopts a data-based analysis method. The core of this method is to go deep into the field of fractional differentials and systematically screen out the most suitable order and band position. This strategy not only improves the scientific basis of model construction based on theoretical and empirical data, but also significantly enhances the adaptability and accuracy of the model in practical application scenarios. By comprehensively considering the differential characteristics and band sensitivity of spectral data, the present invention establishes a model optimization path that is both accurate and practical, thereby enhancing the scientific nature of the model while ensuring its effectiveness and reliability when facing complex practical application problems.
[0082] In this embodiment, the method further includes: after measuring the conductivity of the mural floor sample under different erosion conditions, performing a statistical description of the density distribution, mean, standard deviation, minimum value, quartile and coefficient of variation of the conductivity under five groups of different erosion conditions, such as Figure 6As shown, the red dashed line in the figure is used to represent the first and third quartiles, and the blue dashed line represents the mean. Among them, the coefficient of variation (CV), as a standardized measure of dispersion, is calculated by the ratio of the standard deviation to the mean, and is used to evaluate the variability of the data. The evaluation criteria for data variability are: CV ≤ 15% indicates low variability, 15% < CV ≤ 35% indicates moderate variability, and CV > 35% indicates strong variability. The formula for calculating the coefficient of variation is: In the formula, σ represents the sample mean, and u represents the sample standard deviation.
[0083] The conductivity value of the mural plaster sample is a direct reflection of the difference in salt content in the same material, showing a certain degree of variability. As Figure 6 shown, the range of the EC value of the sample is 2.06 ms·m -1 ~6.43 ms·m -1 . The mean and standard deviation of the EC value of the sample are 3.62 ± 0.88 ms·m -1 . The first quartile (Q1) and the third quartile (Q3) of the data are 2.97 ms·m -1 and 4.27 ms·m -1 respectively, meaning that 50% of the EC values of the samples are between these two values. In addition, the coefficient of variation (CV) is 24%, meaning that the relative variability of the EC value is at a moderate level.
[0084] The fractional-order differentiation adopted in this embodiment:
[0085] Fractional-order differentiation (FOD) is an extension of traditional integer-order differentiation, and it has various forms in the fields of mathematics and engineering, such as Riemann-Liouville (R-L), Riesz, Weyl, Caputo, and Grünwald-Letnikov (G-L). The Grünwald-Letnikov method has been widely used in the fields of engineering and science because it is suitable for processing discrete data. One of the main advantages of this method is that it is directly based on the difference definition of differentiation, which makes its application in numerical calculations more direct and convenient. Especially when dealing with discrete data such as hyperspectral signals, the Grünwald-Letnikov method shows its unique applicability. Therefore, according to the G-L algorithm, the fractional order (v) of f(x) in the wavelength domain [a, t] can be defined as:
[0086]
[0087] where v represents the order of differentiation, h represents the differentiation step size, n represents the number of iterations, t represents the upper limit of differentiation,, x represents the position where the derivative is calculated, a represents the lower limit of differentiation, and Γ represents the Gamma function (Γ(v) = (v - 1)!).
[0088] When the function f(x) is a one-dimensional spectrum, the band range is [a, t], x∈[a, t] divided by the differential step size h. The retention interval of the spectrometer is 1 nm, so the differential step size can be set to h = 1. Therefore, the differential expression of the v-th order fractional differential of the function f(x) can be derived from equation (5) as follows:
[0089]
[0090] In formula (6), v = 0 refers to the 0.0 order differential, that is, no fractional order differential is performed on the hyperspectrum. And v = 1 and v = 2 correspond to integer order differentials of 1.0 and 2.0, respectively. In this embodiment, v ranges from 0 to 2 with an interval of 0.1; when v = 0, the spectrum corresponds to the original spectrum without any derivative, and when v = 1 and v = 2, the first and second order differential spectra are calculated.
[0091] About Pearson correlation coefficient:
[0092] In statistics, the Pearson correlation coefficient is one of the methods used for the extensive relationship between variables. It evaluates the strength of the relationship between two vectors based on the covariance matrix of the data. The Pearson correlation coefficient calculation formula is:
[0093] Among them, cov(α i ,α j ) is the covariance, var(α i ) is α i The variance of var(α j ) is α j The variance of .
[0094] The Pearson correlation coefficient can be applied to a sample or a population. The absolute values of the Pearson correlation coefficients of the sample and the population are both less than or equal to 1. In this embodiment, the maximum absolute correlation coefficient (MACC) between the phosphate content of the mural floor and the spectral index is calculated to determine the best performing spectral index.
[0095] In this embodiment, the S1, preparing a mural ground sample, specifically includes:
[0096] S1.1. Desalination treatment of the clay, sand and wheat straw used to make the mural ground sample;
[0097] S1.2. Evenly mix the desalted clay, sand and wheat straw in a mass ratio of 64:36:3, then add 20% of the solid mass of deionized water, and make a mural floor sample by molding the sample blocks.
[0098] The specific preparation process is as follows:
[0099] First, the Chengban soil, sand and wheat straw used to make the samples were desalted. The purpose of desalination is to prevent the inherent salt in the raw materials from interfering with the experiment. The specific operation method of desalination is: weigh a certain mass of dried and crushed soil samples through a sieve with an aperture of 0.56mm, mix the soil sample and deionized water in a mass ratio of 1:5, and stir with a stirrer in a clockwise direction for 0.5h to fully dissolve the salt in the soil. Let it stand for 24h. After the solution is clarified, use a syringe to extract and filter out the clear water on the top to leave the soil sample, and inject deionized water again for stirring and standing. Repeat the salt washing 5 times, measure the conductivity of the soil sample, convert the ion content with the conductivity, and desalination is considered to be completed when the total salt content in the soil does not exceed 0.1%. Then the desalted soil sample is dried, rolled and sieved and stored for later use. After the desalinated chengban soil, sand and wheat straw are mixed evenly in a mass ratio of 64:36:3, 20% of the solid mass of distilled water is added. Then the sample block is molded. The mixed raw materials are filled into the mold by smearing, and the surface is smoothed, and the excess air in the sample is expelled by vibration. The sample is then placed in a drying oven at 90°C for 2 hours for drying. The reason for setting this temperature is to obtain the dry weight of each sample under completely dry conditions. It is believed that the moisture content of the dried sample is 0. In addition, the initial moisture content is controlled at 20% during the sample preparation process. The mixture of chengban soil and sand in this moisture content state is most suitable for smearing on the surface of the cave cliff. Therefore, the moisture content of the ground layer in the actual cave may also be around this value during the production process. Among them, the desalination process follows the GB / T50123-2019 standard.
[0100] In this embodiment, five accuracy evaluation indicators are used to evaluate the effect of the prediction model for the phosphate content of murals. The performance indicators of the model include the determination coefficient of the calibration data set expressed as The root mean square error of the calibration dataset is denoted as RMSE c The coefficient of determination of the validation data set is expressed as The root mean square error of the validation dataset is denoted as RMSE v And the mean absolute error is expressed as MAE. Among them, R 2 It is used to evaluate the model fit. The closer the value is to 1, the higher the model accuracy is. RMSE and MAE are used to evaluate the model stability. The closer the value is to 0, the better the RMSE and MAE are. 2The calculation formulas for , RMSE and MAE are as follows:
[0101]
[0102] Where n represents the number of mural ground samples; y i represents the measurement value of the i-th mural ground sample; It represents the average value of all the measured values of the mural floor samples; represents the predicted value of the i-th mural ground sample; z i represents the predicted value of the i-th mural ground sample; Represents the average value of the predicted values of all mural floor samples.
[0103] Dual-band spectral index analysis: In this paper, we explored the application of fractional order derivative (FOD) in hyperspectral remote sensing technology, specifically optimizing the spectral estimation of phosphate content in mural ground. Through experimental analysis, we calculated the dual-band spectral index (Phosphorus Normalized Difference Index, PNDI) of mural ground using different FOD orders, aiming to explore the optimal FOD order to maximize the correlation between PNDI and phosphate content.
[0104] Table 1 The characteristic bands and dual-band spectral index estimation results of phosphate content in murals obtained by linear regression method (PNDI)
[0105]
[0106]
[0107] It can be observed from Table 1 that the maximum absolute correlation coefficient (MACC) between PNDI and phosphate content first increases and then decreases with the increase of FOD order.
[0108] Through the univariate regression analysis on the validation data set shown in Table 1, we can observe the performance of the linear regression model based on the two-band spectral index (PNDI) at different FOD orders. The FOD orders involved in this analysis include 0.3, 0.8, 0.2, 0.4, 0.9 and 1.2, and the model accuracy of these orders is in the top six of all tested orders.
[0109] At the same time, in the experiment, different FOD orders were used to calculate the dual-band spectral index (PSR), aiming to explore the optimal FOD order to maximize the correlation between PSR and phosphate content and achieve the maximization of the correlation between PSR and phosphate content.
[0110] Table 2 The characteristic bands and dual-band spectral index estimation results of phosphate content in murals obtained by linear regression method (PSR)
[0111]
[0112]
[0113] From Table 2, it can be observed that the maximum absolute correlation coefficient (MACC) between PSR and phosphate content also shows a trend of first increasing and then decreasing with the increase of FOD order, which is the same as the trend of PNDI. In particular, at the 0.6 order, MACC reaches the highest value of 0.653.
[0114] In the process of further analysis, through the data shown in Table 2, we found that when using a single dual-band spectral index (PSR) to build the model, the 0.3-order fractional order derivative (FOD) processing made the model achieve the highest coefficient of determination (R 2 =0.728), significantly improving the accuracy of phosphate content estimation. In addition, through comparative analysis of 0.1 to 0.8 order and 1 to 2 order FOD treatments, we found that low-order (0.1 to 0.8 order) fractional derivatives are more excellent in estimating phosphate content.
[0115] Through the univariate regression analysis on the validation data set shown in Table 2, we can observe the performance of the linear regression model based on the dual-band spectral index (PSR) at different FOD orders. The FOD orders involved in this analysis include 0.3, 0.7, 0.8, 0.2, 0.1 and 0.4, and the model accuracy of these orders is in the top six of all the test orders with excellent performance.
[0116] Partial least squares regression modeling of dual-band spectral index: Within the framework of the present invention, the selection of specific three bands no longer simply relies on previous research results, but adopts a data-based analysis method. First, the PLSR regression model was used to conduct a high-precision comprehensive analysis of the dual-band spectral index modeling. By analyzing the best performing spectral indices in the top six univariate regressions, these indices were used as explanatory variables to conduct partial least squares regression (PLSR) analysis of ground phosphate content. Finally, as Figure 3 As shown in Figure 2, the PSR-PLSR modeling accuracy is higher than that of the PNDI-PLSR model, and its determination coefficient R 2 =0.759.
[0117] Modeling using the three-band spectral index (PTSR): The correlation coefficient between the three-band spectral index of different FOD orders and the phosphate content of the mural ground is shown in Figure 4 and Table 3. The horizontal and vertical axes represent spectral bands. The maximum absolute correlation coefficients (i.e., the maximum values between the maximum positive correlation coefficient and the maximum negative correlation coefficient) are listed in Table 3. By cumulative comparison method, the maximum absolute correlation coefficient of the three-band spectral index is 5.32% higher than that of PSR, indicating that the correlation between the spectral index and the phosphate content is significantly enhanced after the three-band combination is processed by fractional derivative.
[0118] Table 3 The characteristic bands and three-band spectral index estimation results of phosphate content in murals obtained by linear regression method (PTSR)
[0119]
[0120] In addition, it can be seen in Table 3 that the model accuracy of the three-band spectral index has also been significantly improved compared with the univariate regression model of the two-band spectral index (PSR and PNDI) in the validation dataset. This finding not only proves the superiority of the three-band spectral index in modeling phosphate content, but also emphasizes that in hyperspectral remote sensing analysis, the model prediction accuracy can be significantly improved through careful spectral feature selection and appropriate mathematical processing techniques. The experimental results further emphasize that through in-depth analysis and precise processing of spectral data, more representative spectral indices can be extracted, thus providing a new theoretical basis and methodological guidance for the application and development of hyperspectral remote sensing technology.
[0121] Phosphate content detection model for mural floor: The phosphate content detection model for mural floor established by the present invention is actually a new three-band spectral index phosphate content estimation model for mural floor established based on a data-driven decision-making method. In the present invention, we use the fractional order derivative (FOD) technology and the three-band spectral index (PTSR) to construct a high-precision model for phosphate monitoring of mural floor. Figure 5 It is the optimal phosphate monitoring model for murals and ground based on FOD spectrum and three-band spectral index (PTSR) (based on validation set data). Through in-depth analysis, the PTSR-PLSR model showed significant performance improvement compared with the previous PNDI-PLSR model and PSR-PLSR model, which is specifically reflected in the determination coefficient (R 2) reached 0.815, and the root mean square error (RMSE) dropped to 0.327, which was 10.4% and 7.38% higher than the PNDI-PLSR and PSR-PLSR models, respectively. This result not only confirms the high accuracy and reliability of the PTSR-PLSR model in monitoring the phosphate content of murals, but also highlights the application potential of FOD technology and the new three-band spectral index in the field of hyperspectral remote sensing monitoring.
[0122] In summary, the present invention provides a method for detecting phosphate content in mural floor coverings based on a three-band spectral index, which uses fractional order derivative (FOD) technology and a three-band spectral index (PTSR) to construct a high-precision model for monitoring phosphate content in mural floor coverings, thereby achieving high-precision monitoring of phosphate content in mural floor coverings. This method is different from the previous method of prefabricating whole samples with a specific salt content in laboratories by simulating the capillary salt ion adsorption and crystallization erosion degradation process in the in-situ environment of the mural. This method is closer to the actual situation of murals suffering from salt damage, and can deeply explore the specific impact mechanism of salt on murals at the microscopic level, providing a more accurate and practical scientific basis for mural protection. Furthermore, a new three-band spectral index is developed based on the absorption mechanism of phosphate and its characteristic bands and using a band optimization algorithm to more accurately reflect the phosphate content and fill the gap in the accurate inversion of phosphate content in different phases in existing research; by designing a three-band spectral index, the accuracy, anti-interference ability and elimination of saturation of the traditional two-band spectral index in salt estimation are avoided. The three-band index adds an additional characteristic wavelength through a data-driven decision-making method, which can more effectively evaluate the salt content in the gypsum of the mural. This is a relatively new attempt in hyperspectral remote sensing technology. The fractional differential method provides a more flexible processing framework that can capture more subtle spectral changes, especially suitable for processing nonlinear and complex physical processes. At the same time, the construction of a three-band spectral index can often improve the estimation accuracy of the index, enhance the anti-interference ability, and eliminate the saturation phenomenon of the commonly used two-band index. Through this method, we hope to accurately identify and quantify the types and distribution of salt in the murals without causing physical damage to the murals, thereby providing a more scientific basis for the protection and restoration of the murals. This invention emphasizes the development and application of advanced diagnostic technology as a regular "health check" of cultural heritage to ensure its long-term protection and inheritance. The proactive and preventive nature of this technology greatly improves the efficiency and effectiveness of cultural heritage protection.
[0123] The remaining technical features in this embodiment can be flexibly selected by those skilled in the art according to actual conditions to meet different specific practical needs. The modifications and changes made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the protection scope of the claims attached to the present invention. In the above description, in order to provide a thorough understanding of the present invention, a large number of specific details are explained. However, it is obvious to those of ordinary skill in the art that these specific details do not have to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, well-known technologies, such as specific construction details, operating conditions and other technical conditions, are not described in detail.
[0124] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of the present invention. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. A method for detecting phosphate content in murals based on a three-band spectral index, characterized in that: The following steps are involved: S1. Make mural ground sample; S2, measuring the spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, and preprocessing the spectral reflectance data; S3, performing fractional differential processing on the spectral reflectance data, and calculating the Pearson correlation coefficient between the PNDI and PSR spectral indexes of the spectral reflectance data and the phosphate content of the mural ground, and obtaining the optimal spectral index with the highest correlation with the phosphate content; S4. Using the optimal spectral index and the corresponding conductivity to establish partial least squares regression models PSR-PLSR and PNDI-PLSR, by comparing the accuracy of the two models, the optimal model and its corresponding specific band are selected, and then the basic index R for constructing the three-band spectral index is determined. c , which is used as the basis for establishing a three-band spectral index; S5. Based on the GL algorithm, a differential expression of v-order fractional differential is obtained, and the differential expression and the expression of the PTSR index are applied to calculate the three-band spectral index under different fractional differentials; S6. Using the three-band spectral index as the explanatory variable and the corresponding mural floor conductivity as the response variable, a mural floor phosphate content detection model is established, and it is verified using the validation set. The verified mural floor phosphate content detection model is obtained and used for the mural floor phosphate content measurement.
2. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: In S3, the Pearson correlation coefficients between the PNDI and PSR spectral indexes of the spectral reflectance data and the phosphate content of the mural ground are calculated respectively to obtain the optimal spectral index with the highest correlation with the phosphate content, specifically including: S3.
1. Construct PSR spectral index: Construct the PSR spectral index: Among them, R a and R b Both represent the spectral reflectance at a certain wavelength; S3.
2. Use the formula of the PSR spectral index and the PNDI spectral index to calculate the spectral index of the spectral reflectance data in the verification set, perform univariate regression analysis on the spectral index with the largest absolute value of the Pearson correlation coefficient at each order and the phosphate content of the mural floor, and screen out the spectral index with the optimal order at each order.
3. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: In S4, the use of the optimal spectral index and the corresponding conductivity to establish the partial least squares regression models PSR-PLSR and PNDI-PLSR specifically includes: The spectral index of the best order among the first six orders was taken as the explanatory variable, and the corresponding conductivity was taken as the response variable to establish the PNDI-PLSR model and the PSR-PLSR model.
4. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: The three-band spectral index of the spectrum is: In the formula, R a and R b Respectively represent the spectral reflectance of specific band a and specific band b in the range of 400-2400nm, R c It is a specific band selected based on the optimal model.
5. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: The S2, measuring the spectral reflectance data and electrical conductivity of the mural floor sample under different erosion conditions, specifically includes: S2.
1. After subjecting the mural flooring samples to phosphate corrosion degradation treatment of different concentrations and drying in the shade, the surface spectral reflectance thereof is measured to obtain spectral reflectance data under different corrosion conditions; S2.
2. For mural samples that have been degraded by phosphate erosion, the conductivity is measured using a soil salt conductivity tester to obtain conductivity data under different erosion conditions.
6. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: The method further comprises: after measuring the conductivity of the mural floor sample under different erosion conditions, performing a statistical description of the density distribution, mean, standard deviation, minimum value, quartile and coefficient of variation of the conductivity, wherein the coefficient of variation calculation formula is: In the formula, σ represents the sample standard deviation and u represents the sample mean.
7. The method for detecting phosphate content in murals based on three-band spectroscopy according to claim 1, characterized in that: The S1, preparing a mural ground sample, specifically comprises: S1.
1. Desalination treatment of the clay, sand and wheat straw used to make the mural ground sample; S1.
2. Evenly mix the desalted clay, sand and wheat straw in a mass ratio of 64:36:3, then add 20% of the solid mass of deionized water, and make a mural floor sample by molding the sample blocks.
Citation Information
Cited By
Hyperspectral unmixing detection method for content of soluble salt in mural
CN120609755A
Method for hyperspectral inversion of sodium sulfate concentration in mural by considering temperature and humidity
CN121253517A
