Rapid quantitative method of sulfates in saline soil based on mid-infrared spectroscopy combined with multiple linear regression
By combining mid-infrared spectroscopy with multiple linear regression, the problems of time-consuming and inaccurate sulfate detection in saline soils have been solved, enabling rapid, non-destructive, and high-precision detection of sulfate in saline soils, which is suitable for field applications.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XINJIANG UNIVERSITY
- Filing Date
- 2026-03-30
- Publication Date
- 2026-07-24
AI Technical Summary
Existing technologies for sulfate detection in saline soils are time-consuming, cumbersome, and lack precision. Furthermore, existing mid-infrared spectroscopy applications suffer from noise interference and model complexity, failing to meet the needs for rapid and accurate field detection.
A method combining mid-infrared spectroscopy and multiple linear regression was adopted. The spectrum was preprocessed by Savitzky-Golay smoothing, multiple scattering correction and second derivative transformation to screen out the sensitive characteristic bands of sulfate, and a multiple linear regression model was constructed for quantitative detection.
It achieves rapid, non-destructive, and high-precision detection of sulfate in saline soil without the need for chemical reagents, with a detection time of less than 5 minutes. It is suitable for field applications and has a simple model structure with strong interpretability.
Smart Images
Figure CN122448780A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of saline soil detection technology, specifically a rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression. Background Technology
[0002] Saline soil is a special soil type widely distributed in the arid and semi-arid regions of Northwest my country. Excessive salt content severely inhibits crop growth and causes engineering problems such as road salt swelling and subsidence, seriously threatening regional agricultural production and engineering construction safety. Sulfate is one of the most significant harmful salts in saline soil. Sodium sulfate, calcium sulfate, and other sulfates not only lead to soil compaction and decreased fertility but also undergo crystal swelling reactions under temperature changes, causing roadbed deformation and building corrosion. Therefore, rapid and accurate quantitative detection of sulfate content in saline soil is crucial for the monitoring, remediation, and engineering geological investigation of saline soils.
[0003] Currently, traditional standard methods for detecting sulfate content in saline soils mainly include gravimetric methods, EDTA titration, and spectrophotometry. Gravimetric methods involve reacting sulfate ions in the soil leachate with barium ions to form barium sulfate precipitate. After filtration, washing, drying, and weighing, the sulfate content is calculated. While this method offers high accuracy, the process is extremely cumbersome, requiring multiple steps such as extraction, precipitation, aging, filtration, and drying. The entire testing process takes 2-3 days, failing to meet the needs of rapid on-site testing. Furthermore, this method requires large amounts of chemical reagents, potentially causing environmental pollution, and the testing process is easily influenced by the operator's experience, resulting in significant errors. EDTA titration indirectly calculates sulfate content through titration, also requiring complex pretreatment processes, is time-consuming, and necessitates professional laboratory personnel, making it difficult to promote and apply in the field. While spectrophotometry offers improved detection speed, it still requires sample extraction and chemical colorimetric reactions, failing to achieve non-destructive testing. It is also susceptible to interference from other ions, leading to unstable detection accuracy.
[0004] In recent years, with the development of spectroscopic technology, hyperspectral and near-infrared spectroscopy have been gradually applied to the rapid detection of soil salinity. These technologies do not require complex chemical pretreatment and can achieve non-destructive and rapid detection, attracting widespread attention. For example, in existing technologies, Sun Yanan et al. (2019) used saline soil in the Hetao Irrigation Area as the research object and constructed a sulfate inversion model based on near-infrared hyperspectroscopy. They used the support vector machine method to achieve quantitative prediction of sulfate, and the validation set determination coefficient R² reached 0.74. Chen Ruihua et al. (2021) used near-infrared spectroscopy combined with an SVM model to achieve sulfate inversion for saline soil in the Yinbei area of Ningxia, and the R² reached 0.84. However, these near-infrared spectroscopy-based methods have obvious defects: the absorption peaks of near-infrared spectra are mainly the overtones and combination frequencies of molecular vibrations, with weak absorption intensity, serious overlap of characteristic peaks, and easy interference from matrix components such as soil moisture, organic matter, and clay minerals, resulting in low detection accuracy and difficulty in meeting the needs of high-precision quantitative detection.
[0005] Mid-infrared spectroscopy (MIR) operates in the 4000–400 cm⁻¹ band, which corresponds to the fundamental frequency absorption of molecular vibrations. This band exhibits high absorption intensity and highly specific characteristic peaks; each functional group has its own unique characteristic absorption peak. Therefore, it enables precise identification of specific substances and has significantly stronger anti-interference capabilities than near-infrared spectroscopy. For example, the S=O asymmetric stretching vibration of sulfate ions exhibits a very strong characteristic absorption peak near 1100 cm⁻¹, and the O=S=O bending vibration also shows a significant characteristic absorption near 600 cm⁻¹. These two characteristic peaks are highly specific and almost unaffected by other common soil components, making them ideal for the quantitative detection of sulfates.
[0006] However, existing technologies for the application of mid-infrared spectroscopy in sulfate detection in saline soil still have many shortcomings: First, the matrix of saline soil is extremely complex, and the particle size of different samples varies greatly, which leads to severe scattering effects. At the same time, the spectrum has obvious baseline drift. These noises and interferences can mask the characteristic absorption peaks of sulfate, resulting in a very low signal-to-noise ratio of the original spectrum. Most of the preprocessing methods in existing technologies use single smoothing or derivative processing, which cannot simultaneously eliminate the interference of scattering, baseline drift and random noise, resulting in poor preprocessing effects. Secondly, existing technologies mostly use full-band spectral data as input when constructing quantitative models, combined with complex machine learning models such as partial least squares regression (PLSR) and support vector machines (SVM). Although these models can handle high-dimensional data, their complex structure and large computational load require specialized software and hardware support, making them unsuitable for rapid calculation in the field. Furthermore, full-band data suffers from severe collinearity, rendering traditional multiple linear regression (MLR) models unusable. While MLR models offer advantages such as simple structure, fast computation speed, and strong interpretability, making them ideal for rapid field detection, current technologies fail to address the issues of collinearity and feature selection, resulting in extremely low accuracy in sulfate detection and failing to meet the requirements.
[0007] In summary, existing technologies suffer from problems such as long detection time, cumbersome operation, complex models, and insufficient accuracy. To address these issues, this invention proposes a rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression. Summary of the Invention
[0008] The purpose of this invention is to provide a rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression, so as to solve the problems mentioned in the background art.
[0009] To achieve the above objectives, the present invention provides the following technical solution: A rapid quantitative method for sulfate in saline soils based on mid-infrared spectroscopy combined with multiple linear regression includes the following steps: S1. Collection and pretreatment of saline soil samples: Saline soil samples with different degrees of salinization were collected, impurities were removed, and the samples were air-dried, ground, and passed through a 150-250 mesh standard sieve to obtain soil powder samples to be tested. S2. Mid-infrared spectral data acquisition: A Fourier transform infrared spectrometer was used to acquire diffuse reflectance spectra of the soil powder sample to obtain raw mid-infrared absorbance spectral data in the 4000-400cm⁻¹ band. S3. Joint preprocessing of spectral data: The original spectral data are sequentially subjected to Savitzky-Golay smoothing, multivariate scattering correction, and second derivative transformation to obtain preprocessed spectral data. S4. Screening of sulfate-sensitive characteristic bands: Based on the correlation analysis between the preprocessed spectral data and the measured sulfate content of the sample, combined with the stepwise regression algorithm, the sensitive characteristic bands of sulfate are screened out. The sensitive characteristic bands include the 1050-1150 cm⁻¹ band and the 550-650 cm⁻¹ band. S5. Construction of a multiple linear regression quantitative model: Using the spectral absorbance data within the sensitive characteristic band as the input variable and the measured sulfate content of the sample as the output variable, a multiple linear regression quantitative prediction model is constructed. S6. Rapid quantitative detection of the sample to be tested: For the saline soil sample to be tested, perform the sample and spectrum preprocessing steps S1 to S3 in sequence, input the preprocessed characteristic band absorbance data into the multiple linear regression quantitative prediction model, and output the sulfate content of the sample to be tested.
[0010] Preferably, in step S1, the natural air-drying temperature is 20-25°C, the air-drying time is 7-10 days, direct sunlight is avoided during the air-drying process, and the sample particles are passed through a 200-mesh standard sieve after grinding to ensure uniform particle size.
[0011] Preferably, in step S2, the scanning resolution of the Fourier transform infrared spectrometer is 4 cm⁻¹, the number of scans is 32, a standard KBr plate is used for background correction before acquisition to eliminate environmental background interference, and the spectrum of each sample is collected three times and the average value is taken as the original spectral data.
[0012] Preferably, in step S3, the window size of the Savitzky-Golay smoothing process is 15 data points, and the polynomial fitting order is 2, which is used to eliminate random noise in the spectral acquisition process.
[0013] Preferably, in step S3, the multivariate scattering correction processing is used to eliminate the scattering effect and baseline drift caused by differences in particle size in the saline soil samples, specifically including: calculating the average spectrum of all samples as a reference spectrum:
[0014] in, For the sample size, For wavelength points, For the first One sample in Absorbance at a wavelength point; Linear regression was performed on the spectrum of each sample and the reference spectrum to obtain the intercept for each sample. With slope ; Correct the original spectrum using correction parameters:
[0015] Preferably, in step S3, the second derivative transformation process is used to eliminate background baseline shift in the spectrum, separate overlapping absorption peaks, and enhance the resolution of sulfate characteristic absorption peaks.
[0016] Preferably, in step S4, the correlation analysis is Pearson correlation analysis, and the calculation formula is:
[0017] in, Absorbance in the specified wavelength range To measure sulfate content, bands with an absolute correlation coefficient greater than 0.8 with sulfate content were selected as candidate sensitive bands. Then, the optimal combination of characteristic bands was selected from the candidate bands using a stepwise regression algorithm to eliminate the problem of collinearity between variables.
[0018] Preferably, in step S5, the regression equation of the multiple linear regression quantitative prediction model is:
[0019] in, The predicted sulfate content is expressed in g / kg. The average absorbance is in the 1050–1150 cm⁻¹ wavelength range; The average absorbance is in the 550–650 cm⁻¹ band.
[0020] Compared with the prior art, the beneficial effects of the present invention are: This invention uses mid-infrared spectroscopy for quantitative detection of sulfate. It utilizes the fundamental frequency characteristic absorption peak of sulfate ions in the mid-infrared band. Compared with the overtone absorption of near-infrared band, the characteristic peak has stronger specificity, higher absorption intensity, and stronger anti-interference ability. It can effectively resist the interference of soil organic matter, clay minerals, and other salts, and significantly improve the accuracy of detection.
[0021] This invention proposes a combined preprocessing method of Savitzky-Golay smoothing, multivariate scattering correction, and second derivative transformation, which can simultaneously eliminate random noise, scattering effects caused by sample particle differences, baseline drift, and other interferences during the spectral acquisition process. This significantly improves the signal-to-noise ratio of the spectrum, allowing the characteristic absorption peaks of sulfate to be clearly highlighted, and solves the problem of spectral interference caused by the complex matrix of saline soil.
[0022] This invention uses correlation analysis combined with stepwise regression algorithm to accurately screen out two specific sensitive characteristic bands of sulfate. This not only significantly reduces the input variables of the model and the amount of computation, but also effectively eliminates the collinearity problem between spectral variables. It solves the defect that traditional multiple linear regression models cannot be applied to quantitative spectral detection, enabling multiple linear regression models to achieve extremely high prediction accuracy.
[0023] The multiple linear regression model constructed in this invention has an extremely simple structure, requiring only two input variables to achieve high-precision prediction. It has an extremely fast calculation speed, requiring no complex machine learning algorithms or professional software support. In field testing, only simple calculations are needed to obtain results, making it very suitable for rapid field testing applications. At the same time, the model has extremely strong interpretability, clearly explaining the quantitative relationship between characteristic bands and sulfate content, and has higher reliability compared to black-box machine learning models.
[0024] The detection method of this invention requires no chemical reagents or complex pretreatment throughout the entire process, and the detection time for a single sample is no more than 5 minutes. Compared with the traditional method of 2 to 3 days, the detection speed is increased by hundreds of times. At the same time, it realizes non-destructive testing, which can provide efficient technical support for large-scale rapid monitoring of saline soil and has extremely high practical value. Attached Figure Description
[0025] Figure 1 This is a flowchart of the rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy as described in this invention. Detailed Implementation
[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0027] Example 1 This embodiment uses saline soil in area A as the research object and employs the method of the present invention for rapid quantitative detection of sulfate content. The specific process is as follows: S1. Sample Collection and Preprocessing: First, 80 saline soil samples were collected from the top 0-20cm layer in different salinized areas of the Yongji irrigation area in the Hetao Irrigation District. The samples covered different types of soil, including non-saline soil, slightly saline soil, moderately saline soil, and heavily saline soil. The sulfate content ranged from 0.12 to 9.78 g / kg, covering all sulfate content ranges in the area.
[0028] After collection, the samples were first cleaned of plant debris, stones, and other impurities. They were then placed in a well-ventilated, shady place to air dry naturally at 23°C for 8 days. Direct sunlight was avoided during the drying process to prevent salt loss and organic matter decomposition. Once dried, the samples were ground using an agate grinder and then passed through a 200-mesh standard sieve to obtain a uniform soil powder sample, which was used for subsequent spectral acquisition and chemical analysis.
[0029] S2. Actual measured sulfate content of the sample: Following the standard method in the "Standard for Geotechnical Testing Methods" (GB / T50123-2019), the sulfate content of each sample was measured and used as the model's annotation data. The specific procedure was as follows: 10g of sieved soil sample was weighed, 50ml of deionized water was added, and the sample was extracted by shaking at 25℃ for 30 minutes. Then, the sample was centrifuged and filtered to obtain the soil extract. The sulfate ion content in the extract was determined using EDTA titration. Finally, the sulfate content of each sample was calculated, and the measured results were used as the true output value of the model.
[0030] S3, Mid-infrared spectral data acquisition: A Bruker VERTEX 70 Fourier transform infrared spectrometer, equipped with a diffuse reflectance accessory, was used to acquire spectra for each soil powder sample. Before acquisition, background correction was performed using a standard KBr plate. The soil powder was then placed in a sample cup, the surface was leveled, and spectral acquisition was performed. The parameters for spectral acquisition were: scanning range 4000–400 cm⁻¹, resolution 4 cm⁻¹, 32 scans, ambient temperature 24℃, and relative humidity 35% to avoid interference from moisture. Each sample was acquired three times, and the average value was used as the raw spectral data for that sample, ultimately yielding raw mid-infrared absorbance spectra for 80 samples.
[0031] S4. Joint preprocessing of spectral data: The obtained raw spectral data are subjected to joint preprocessing sequentially, as follows: First, Savitzky-Golay smoothing is performed, with a window size of 15 data points and a polynomial fitting order of 2. This smooths the raw spectrum, eliminating random noise during spectral acquisition and improving the signal-to-noise ratio. This method involves performing polynomial least-squares fitting on the spectral data within a sliding window, as shown in the formula:
[0032] in, The fitting coefficients are obtained by polynomial least squares fitting, with a window size corresponding to 15 data points. This effectively eliminates the interference of random noise.
[0033] Then, multivariate scattering correction is performed. First, the average spectrum of all 80 samples is calculated as the reference spectrum. Then, a univariate linear regression is performed on the smoothed spectrum of each sample and the reference spectrum to obtain the intercept (translation parameter) and slope (scaling parameter) of each sample. Then, these two parameters are used to correct the spectrum of each sample to eliminate the scattering effect and baseline drift caused by the difference in particle size of different samples, and to unify the spectral baseline of all samples.
[0034] Finally, a second-order derivative transformation is performed on the corrected spectrum to calculate the second derivative, eliminating the remaining background baseline shift and separating overlapping absorption peaks, thus enhancing the resolution of sulfate characteristic peaks. After these three combined preprocessing steps, preprocessed spectral data are obtained.
[0035] The original spectrum showed significant baseline drift, with large differences in baseline between different samples, and the characteristic peaks were obscured by the background and difficult to distinguish. However, after preprocessing, the baseline drift was completely eliminated, and the characteristic peaks of sulfate were clearly highlighted. The S=O stretching vibration peak near 1100 cm⁻¹ and the O=S=O bending vibration peak near 600 cm⁻¹ were particularly prominent, and the intensity of these two peaks showed a significant positive correlation with the sulfate content, indicating that the preprocessing effect was extremely significant.
[0036] S5. Screening of sulfate-sensitive characteristic bands: First, for the preprocessed spectral data, the Pearson correlation coefficient between the absorbance of each band and the measured sulfate content of the sample was calculated to analyze the correlation between each band and the sulfate content. The results showed that the absolute value of the correlation coefficient was the highest in the 1050 / 1150 cm⁻¹ band, reaching 0.92, and the absolute value of the correlation coefficient in the 550 / 650 cm⁻¹ band also reached 0.88, while the correlation coefficients of other bands were all below 0.7, indicating that these two bands are sensitive characteristic bands for sulfate.
[0037] Then, these two bands were selected as candidate bands, and a stepwise regression algorithm was used for variable screening. The results of the stepwise regression showed that the VIF (variance inflation factor) of these two bands were 1.21 and 1.18, respectively, which are much less than 5, indicating that there is no collinearity problem between the two variables, and they can be used as input variables for multiple linear regression. Therefore, these two bands were finally determined as the sensitive feature bands of this invention, namely 1050 / 1150 cm⁻¹ and 550 / 650 cm⁻¹.
[0038] S6. Construction of a Multiple Linear Regression Model: Eighty samples were randomly divided into a training set and a validation set at a ratio of 3:1, with 60 samples in the training set and 20 samples in the validation set. A multiple linear regression model was constructed using the average absorbance of two characteristic bands of the training set samples as the input variable and the measured sulfate content as the output variable.
[0039] After fitting, the resulting regression equation is:
[0040] in, This refers to the sulfate content (g / kg). The average absorbance in the 1050 / 1150 cm⁻¹ band. The average absorbance is measured in the 550 / 650 cm⁻¹ band.
[0041] Then, the model's accuracy was validated using 20 samples from the validation set. The formula for calculating the model evaluation metric is as follows: Coefficient of determination :
[0042] Root mean square error :
[0043] Relative analysis error :
[0044] in, These are measured values. For predicted values, The average of the measured values. To determine the standard deviation of the measured values in the validation set.
[0045] The validation results show that the determination coefficients between the predicted and measured values on the training set are... The validation set reached 0.97. The root mean square error reached 0.96. The relative analytical error is 0.31 g / kg. The RPD reaches 3.82. According to the evaluation criteria for spectral models, an RPD greater than 3.0 indicates that the model has excellent predictive ability. The RPD of the model in this invention reaches 3.82, which means that the model has extremely high prediction accuracy and can fully meet the needs of quantitative detection.
[0046] S7. Sample detection: For the saline soil sample to be tested, sample preprocessing is performed according to step S1, and then the infrared spectrum is collected. Spectral joint preprocessing is performed according to step S3 to extract the average absorbance of the 1050 / 1150 cm⁻¹ and 550 / 650 cm⁻¹ bands. Substituting these values into the regression equation above, the sulfate content of the sample can be quickly calculated. The entire process takes only 4 minutes, requires no chemical reagents, and achieves rapid and non-destructive detection.
[0047] Example 2 This embodiment uses saline soil from Region B in my country as the research object to verify the applicability of the method of the present invention in saline soil samples from different regions. The specific process is as follows: S1. Sample Collection and Preprocessing: In Pingluo County, Yinbei Irrigation District, Ningxia, a total of 75 saline soil samples were collected from the top 0 to 20 cm layer. This area is arid and semi-arid, and the soil sulfate content is generally high. The samples cover different types of salinization, including mild, moderate, severe, and saline soils. The sulfate content ranges from 0.21 to 12.35 g / kg, covering the typical sulfate content range of this area.
[0048] The sample pretreatment process was the same as in Example 1: after removing impurities, the sample was air-dried at 22°C for 9 days, avoiding direct sunlight. After air-drying, the sample was ground through a 200-mesh standard sieve to obtain a uniform soil powder sample.
[0049] S2. Actual measured sulfate content of the sample: Similarly, following the EDTA titration method in the "Standard for Geotechnical Testing Methods" (GB / T50123-2019), the sulfate content of each sample was measured to serve as the true data for the model.
[0050] S3. Spectral Acquisition and Preprocessing: Using the same Bruker VERTEX 70 Fourier transform infrared spectrometer and with the same parameters, raw mid-infrared spectral data of 75 samples were obtained. Then, Savitzky-Golay smoothing (window 15, order 2), multivariate scattering correction, and second-order derivative transform were performed in sequence for joint preprocessing to obtain preprocessed spectral data.
[0051] S4. Feature band selection and model construction: Similarly, two sensitive feature bands, 1050–1150 cm⁻¹ and 550–650 cm⁻¹, were selected. The 75 samples were divided into a training set of 56 and a validation set of 19 in a 3:1 ratio to construct a multiple linear regression model.
[0052] The validation results show that the coefficient of determination R² for the training set reaches 0.96, the R² for the validation set reaches 0.95, the root mean square error RMSE is 0.42 g / kg, and the relative analysis error RPD reaches 3.51, still meeting the standard of excellent predictive ability. This indicates that the method of the present invention also has extremely high detection accuracy in saline soil samples in Yinbei area of Ningxia, and the method has good regional adaptability.
[0053] S5. Sample detection: The sample to be tested is preprocessed and spectrally acquired according to the same procedure. The sulfate content can be quickly obtained by substituting it into the regression equation. The entire detection process takes no more than 5 minutes, realizing the rapid detection of sulfate in saline soil in this area.
[0054] Example 3 This embodiment uses saline soil from region C in my country as the research object to verify the detection capability of the method of the present invention in saline soil with high sulfate content. The specific process is as follows: S1. Sample Collection and Preprocessing: Ninety saline soil samples from the 0-20cm surface layer were collected in different salinization areas of the Wei-Ku oasis in Xinjiang. This area is a typical extremely arid region in Xinjiang with extremely high salinity. Some areas are typical sulfate-type saline soils. The sulfate content of the samples ranged from 0.35 to 18.62 g / kg, which is much higher than the content range of the previous two examples. This was used to verify the detection accuracy of this method in the high content range.
[0055] Sample pretreatment process: After removing plant residues and stone impurities, the sample is air-dried at 24℃ for 10 days, ground, and then passed through a 200-mesh standard sieve to obtain uniform soil powder, ensuring the consistency of sample particles.
[0056] S2. Actual measured sulfate content of the sample: The sulfate content of the samples was measured using the standard gravimetric method to obtain high-precision true data for model calibration and validation.
[0057] S3. Spectral Acquisition and Preprocessing: The same Fourier transform infrared spectrometer was used for spectral acquisition under the same environmental conditions (temperature 24℃, humidity 32%), with a scanning range of 4000–400 cm⁻¹ and a resolution of 4 cm⁻¹. Each sample was acquired three times and the average was taken. The raw spectra were then subjected to joint preprocessing to eliminate interference from noise, scattering, and baseline drift.
[0058] S4. Feature band selection and model validation: For samples from this region, two sensitive characteristic bands, 1050–1150 cm⁻¹ and 550–650 cm⁻¹, were also selected. The correlation coefficients between these two bands and sulfate content reached 0.91 and 0.87, respectively, still showing a very strong correlation. The 90 samples were divided into a training set of 67 samples and a validation set of 23 samples in a 3:1 ratio to construct a multiple linear regression model.
[0059] The validation results show that even in the extreme case where the sulfate content reaches as high as 18.62 g / kg, the training set R² still reaches 0.97, the validation set R² reaches 0.96, the root mean square error RMSE is 0.58 g / kg, and the relative analysis error RPD reaches 3.67. The model still maintains extremely high prediction accuracy, indicating that the method of the present invention can cover a very wide range of sulfate content detection, from low-content non-saline soil to high-content saline soil, and can achieve accurate detection, demonstrating strong applicability.
[0060] S5. Sample detection: For the samples to be tested in this area, only simple air drying, grinding and spectral acquisition are required to complete the quantitative detection of sulfate content within 5 minutes. No complicated chemical pretreatment is required, which can meet the needs of large-scale saline soil monitoring in this area.
[0061] Comparative Example 1 To verify the advantages of the method of this invention compared with the prior art, this case selected the same saline soil sample from the Hetao Irrigation District and simultaneously detected it using the traditional gravimetric method, EDTA titration method, near-infrared spectroscopy combined with SVM model method, and the method of this invention, respectively, and compared the detection performance of different methods.
[0062] Traditional gravimetric method: The test is carried out according to the gravimetric method in standard GB / T50123-2019. The sample is extracted, precipitated, aged, filtered, dried and weighed. The whole process takes 48 hours and requires the use of chemical reagents such as barium chloride and hydrochloric acid. It is a destructive test and the test process requires professional laboratory personnel to operate.
[0063] EDTA titration method: The test is carried out according to the standard EDTA titration method. After sample extraction, the sulfate content is calculated by titration. The whole process takes 3 hours and requires the use of chemical reagents such as EDTA standard solution and buffer solution. It is also a destructive test.
[0064] Near-infrared spectroscopy combined with SVM model method: Using the existing near-infrared spectroscopy detection method, the near-infrared spectrum (10000~4000cm⁻¹) of the sample is collected, and an SVM model is used to construct a sulfate quantification model. This method does not require chemical reagents, is a non-destructive detection method, and the detection time is about 10 minutes.
[0065] This application uses the mid-infrared spectroscopy combined with multiple linear regression method proposed in this application for detection, and the process is as described in Example 1.
[0066]
[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A rapid quantitative method for sulfate in saline soils based on mid-infrared spectroscopy combined with multiple linear regression, characterized in that, Includes the following steps: S1. Collection and pretreatment of saline soil samples: Saline soil samples with different degrees of salinization were collected, impurities were removed, and the samples were air-dried, ground, and passed through a 150-250 mesh standard sieve to obtain soil powder samples to be tested. S2. Mid-infrared spectral data acquisition: A Fourier transform infrared spectrometer was used to acquire diffuse reflectance spectra of the soil powder sample to obtain raw mid-infrared absorbance spectral data in the 4000-400cm⁻¹ band. S3. Joint preprocessing of spectral data: The original spectral data are sequentially subjected to Savitzky-Golay smoothing, multivariate scattering correction, and second derivative transformation to obtain preprocessed spectral data. S4. Screening of sulfate-sensitive characteristic bands: Based on the correlation analysis between the preprocessed spectral data and the measured sulfate content of the sample, combined with the stepwise regression algorithm, the sensitive characteristic bands of sulfate are screened out. The sensitive characteristic bands include the 1050-1150 cm⁻¹ band and the 550-650 cm⁻¹ band. S5. Construction of a multiple linear regression quantitative model: Using the spectral absorbance data within the sensitive characteristic band as the input variable and the measured sulfate content of the sample as the output variable, a multiple linear regression quantitative prediction model is constructed. S6. Rapid quantitative detection of the sample to be tested: For the saline soil sample to be tested, perform the sample and spectrum preprocessing steps S1 to S3 in sequence, input the preprocessed characteristic band absorbance data into the multiple linear regression quantitative prediction model, and output the sulfate content of the sample to be tested.
2. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S1, the natural air-drying temperature is 20-25℃, the air-drying time is 7-10 days, direct sunlight is avoided during the air-drying process, and after grinding, the sample is passed through a 200-mesh standard sieve to ensure that the sample particles are uniform.
3. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S2, the scanning resolution of the Fourier transform infrared spectrometer is 4 cm⁻¹, the number of scans is 32, and a standard KBr plate is used for background correction before acquisition to eliminate environmental background interference. The spectrum of each sample is collected three times and the average value is taken as the original spectral data.
4. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S3, the window size of the Savitzky-Golay smoothing process is 15 data points, and the polynomial fitting order is 2, which is used to eliminate random noise in the spectral acquisition process.
5. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S3, the multivariate scattering correction process is used to eliminate the scattering effect and baseline drift caused by differences in particle size in saline soil samples. Specifically, it includes: calculating the average spectrum of all samples as a reference spectrum. in, For the sample size, For wavelength points, For the first One sample in Absorbance at a wavelength point; Linear regression was performed on the spectrum of each sample and the reference spectrum to obtain the intercept for each sample. With slope ; Correct the original spectrum using correction parameters: 。 6. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S3, the second derivative transformation process is used to eliminate the background baseline shift of the spectrum, separate overlapping absorption peaks, and enhance the resolution of sulfate characteristic absorption peaks.
7. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S4, the correlation analysis is a Pearson correlation analysis, and the calculation formula is: in, Absorbance in the specified wavelength range To measure sulfate content, bands with an absolute correlation coefficient greater than 0.8 with sulfate content were selected as candidate sensitive bands. Then, the optimal combination of characteristic bands was selected from the candidate bands using a stepwise regression algorithm to eliminate the problem of collinearity between variables.
8. The rapid quantitative method for sulfate in saline soil based on mid-infrared spectroscopy combined with multiple linear regression according to claim 1, characterized in that, In step S5, the regression equation of the multiple linear regression quantitative prediction model is: in, The predicted sulfate content is expressed in g / kg. The average absorbance is in the 1050–1150 cm⁻¹ wavelength range; The average absorbance is in the 550–650 cm⁻¹ band.