Polarization spectrum-based unfrozen water content prediction method and system, medium and terminal

Abstract

The application discloses a kind of based on polarization spectrum unfrozen water content prediction method and system, medium, terminal, it is related to soil detection technical field, main purpose is in to improve the problem of the prediction accuracy reduction of unfrozen water content prediction model based on reflectivity construction caused by the change of soil surface reflection characteristic in the ice crystal structure formation process.It includes: collecting the multi-angle polarization spectrum of the soil sample to be measured at the target temperature, and carrying out decomposition processing to obtain a set of polarization component spectrum;Determine the target freeze-thaw stage based on the target temperature;Based on the stage input feature subset corresponding to the target freeze-thaw stage, generate target input feature combination;Based on the stage unfrozen water content prediction model matched with the target freeze-thaw stage, generate the predicted unfrozen water content according to the target input feature combination.

Description

Technical Field

[0001] This application relates to the field of soil testing technology, and in particular to a method, system, medium, and terminal for predicting unfrozen water content based on polarization spectroscopy. Background Technology

[0002] Permafrost is widely distributed in high-latitude, high-altitude regions of my country, with seasonally frozen soil accounting for more than half of the country's land area. Soil freeze-thaw cycles are one of the most common physical processes in cold regions, essentially involving the periodic phase transition of soil pore water between liquid and solid states driven by temperature changes. During freeze-thaw cycles, not all water in the soil pores freezes into ice; some remains in liquid form on the surface of mineral particles and in fine pores—this portion is called unfrozen water. Unfrozen water content is a key physical parameter characterizing the hydrothermal processes of permafrost, directly affecting soil thermal conductivity, water migration patterns, solute transport characteristics, and frost heave and thaw settlement behavior. Therefore, accurately predicting unfrozen water content is crucial for ecological restoration, agricultural production, and engineering stability assessment in permafrost regions. Traditional methods for determining unfrozen water content, such as nuclear magnetic resonance (NMR) and thermal analysis, while highly accurate, suffer from problems such as complex equipment, long testing cycles, and difficulty in achieving rapid in-situ monitoring.

[0003] In recent years, visible-near-infrared spectroscopy has been widely used in soil moisture retrieval research due to its advantages such as being non-destructive and rapid. It mainly uses reflectance spectra to construct empirical models between unfrozen water content and the unfrozen water content in order to predict the unfrozen water content.

[0004] However, during the freezing process of soil, the formation of ice crystals significantly alters the surface reflection characteristics, resulting in the inclusion of specular reflection components in the reflection signal. Empirical models based on reflectivity are unable to effectively eliminate the interference caused by this specular reflection, leading to the introduction of noise unrelated to the unfrozen water content into the model, ultimately reducing the accuracy of unfrozen water content prediction. Summary of the Invention

[0005] In view of this, this application provides a method, system, medium, and terminal for predicting unfrozen water content based on polarization spectroscopy. The main purpose is to improve the problem of reduced prediction accuracy of unfrozen water content prediction models based on reflectivity caused by changes in the reflectivity of the soil surface during the formation of ice crystals.

[0006] According to one aspect of this application, a method for predicting unfrozen water content based on polarization spectroscopy is provided, comprising:

[0007] The multi-angle polarization spectrum of the soil sample to be tested is collected at the target temperature, and the multi-angle polarization spectrum is decomposed to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0008] Based on the target temperature and the preset freeze-thaw stage division rules, the target freeze-thaw stage of the soil sample to be tested is determined.

[0009] The experimental polarization component spectra of multiple experimental soil samples were obtained during each freeze-thaw stage. By performing correlation analysis between the spectra of the multiple experimental polarization components and the measured unfrozen water content, the dominant polarization component of each freeze-thaw stage that responds most significantly to the change in unfrozen water content and the corresponding preferred band were selected to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0010] Based on the stage input feature subset corresponding to the target freeze-thaw stage, the polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectrum set to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0011] Based on a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample to be tested at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

[0012] Preferably, after generating the predicted unfrozen water content of the soil sample to be tested at the target temperature, the method further includes:

[0013] Obtain the polarization component spectral set of the soil sample to be tested at different temperatures and predict the unfrozen water content;

[0014] For each temperature point, the polarization component spectrum of the dominant polarization component corresponding to the current temperature point is obtained from the set of polarization component spectra corresponding to the current temperature point. Based on the polarization component spectrum of the dominant polarization component corresponding to the current temperature point and the polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point, the band spectral change rate of each band is calculated respectively. The mean of all the band spectral change rates is determined as the polarization change intensity index of the current temperature point. Based on the polarization change intensity index, the physical monotonicity constraint modulation weight of the current temperature point is constructed. Based on the physical monotonicity constraint modulation weight, the physical monotonicity constraint optimization objective function of the current temperature point is constructed. The physical monotonicity constraint optimization objective function is solved to perform monotonicity correction processing on the predicted unfrozen water content of the current temperature point, and the predicted unfrozen water content after monotonicity correction is obtained. The current temperature point and the next adjacent temperature point are obtained by sorting all temperature points according to the temperature decrease rule or the temperature increase rule.

[0015] Preferably, after obtaining the predicted unfrozen water content after monotonicity correction, the method further includes:

[0016] Based on the pre-trained residual prediction model, residual estimation is performed according to the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage to obtain the corresponding residual estimation compensation amount.

[0017] The predicted unfrozen water content after monotonicity correction is summed with the residual estimate compensation to obtain the final predicted unfrozen water content of the soil sample under test at the target temperature.

[0018] Preferably, before determining the target freeze-thaw stage of the soil sample to be tested based on the target temperature and a preset freeze-thaw stage division rule, the method further includes:

[0019] Freeze-thaw cycle experiments were conducted on multiple experimental soil samples, and experimental multi-angle polarization spectra of each experimental soil sample were collected during the complete freeze-thaw process. The initial water content of each experimental soil sample was different, and the complete freeze-thaw process included multiple experimental temperature points. The experimental multi-angle polarization spectra included the experimental multi-angle polarization spectra of the experimental soil samples at each experimental temperature point.

[0020] The multi-angle polarization spectra of each experiment are decomposed to obtain multiple sets of experimental polarization component spectra, and a mapping relationship is established between the sets of polarization component spectra and the corresponding experimental temperature points and initial water content.

[0021] For each polarization component, the spectrum of all polarization components of the current polarization component is obtained. The spectrum of all polarization components is sorted according to the temperature decrease rule or the temperature increase rule to obtain the polarization component spectrum sequence of the current polarization component during the complete freeze-thaw process.

[0022] For each pair of adjacent experimental temperature points, based on the polarization component spectral sequence, the experimental band spectral change rate of the current polarization component between the current adjacent experimental temperature points is calculated, and the average value of all experimental band spectral change rates is determined as the experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points.

[0023] Obtain the experimental polarization change intensity index of all polarization components between the current adjacent experimental temperature points, determine the maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices as the experimental polarization change intensity index between the current adjacent experimental temperature points, and integrate the experimental polarization change intensity indices between all adjacent experimental temperature points to obtain the experimental polarization change intensity index sequence of the complete freeze-thaw process.

[0024] According to a preset screening method, continuous temperature ranges with polarization change intensity higher than the background fluctuation level are selected from the experimental polarization change intensity index sequence, and the continuous temperature ranges are determined as phase transition stages. Continuous temperature ranges located on the low-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as stable freezing stages. Continuous temperature ranges located on the high-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as melting and stabilizing stages. The preset screening method is any one of the statistical threshold method, peak ratio method, and extreme value identification method. The continuous temperature range contains at least two pairs of adjacent experimental temperature points.

[0025] Based on the temperature ranges corresponding to the stable freezing stage, the phase transition stage, and the thawing stabilization stage, a preset freeze-thaw stage division rule is constructed.

[0026] Preferably, the step involves acquiring the experimental polarization component spectra of multiple experimental soil samples during each freeze-thaw stage, and by performing correlation analysis between the spectra of the multiple experimental polarization components and the measured unfrozen water content, selecting the dominant polarization component and its corresponding preferred band that most significantly responds to changes in unfrozen water content during each freeze-thaw stage, to generate a stage input feature subset corresponding to each freeze-thaw stage, including:

[0027] Based on the experimental temperature points corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules, the freeze-thaw stage corresponding to each experimental polarization component spectral set is determined respectively.

[0028] Within each freeze-thaw stage, for each polarization component, based on the experimental polarization component spectrum of the current polarization component, the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content is calculated, and the correlation distribution data of the current polarization component within the current freeze-thaw stage is generated.

[0029] From the correlation distribution data, bands with absolute values ​​of correlation coefficients higher than a preset correlation coefficient threshold are selected to obtain a set of candidate sensitive bands for the current polarization component in the current freeze-thaw stage.

[0030] The stability of each candidate sensitive band in the candidate sensitive band set is evaluated to screen out candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error threshold or whose correlation dispersion coefficient exceeds a preset correlation dispersion threshold, so as to obtain the preferred band of the current polarization component in the current freeze-thaw stage.

[0031] The polarization components are comprehensively scored from the dimensions of the number of preferred bands, the correlation coefficient of preferred bands, and the stability evaluation results of preferred bands. The polarization component with the highest comprehensive score is determined as the dominant polarization component of the current freeze-thaw stage. Based on the dominant polarization component and the corresponding preferred band, a subset of stage input features corresponding to the current freeze-thaw stage is generated.

[0032] Preferably, the step of collecting the multi-angle polarization spectrum of the soil sample to be tested at the target temperature and decomposing the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature includes:

[0033] Using a polarization spectrometer, multi-angle polarization spectra of the soil sample to be tested are obtained from multiple preset polarization directions, wherein each preset polarization direction corresponds to a polarization spectrum, and the multi-angle polarization spectrum contains multiple polarization spectra;

[0034] By synthesizing the polarization spectra corresponding to all the preset polarization directions, the total reflectance component spectrum of the soil sample under test at the target temperature is obtained;

[0035] Two mutually orthogonal preset polarization directions are selected from all the preset polarization directions, and the polarization spectra corresponding to the two preset polarization directions are determined as the parallel polarization component spectrum and the vertical polarization component spectrum of the soil sample to be tested at the target temperature.

[0036] Stokes parameters are calculated based on the multi-angle polarization spectrum, and the linear polarization degree component spectrum and polarization angle component spectrum of the soil sample under test at the target temperature are calculated based on the Stokes parameters.

[0037] Preferably, before generating the predicted unfrozen water content of the soil sample at the target temperature based on the target input feature combination using a pre-trained model that matches the target freeze-thaw stage, the method further includes:

[0038] For each freeze-thaw stage, construct a prediction model for the initial unfrozen water content corresponding to the current freeze-thaw stage;

[0039] For each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, the experimental polarization component data of the dominant polarization component at the preferred band is obtained from the current experimental polarization component spectral set, and the experimental polarization component data, the experimental temperature point corresponding to the current experimental polarization component spectral set, and the initial water content are combined to generate the training sample of the current freeze-thaw stage.

[0040] Based on multiple training samples of the current freeze-thaw stage, the model for predicting the unfrozen water content of the initial stage is trained to obtain the stage unfrozen water content prediction model corresponding to the current freeze-thaw stage that has completed model training.

[0041] According to another aspect of this application, a system for predicting unfrozen water content based on polarization spectroscopy is provided, comprising:

[0042] A multi-angle polarization spectral acquisition module is used to acquire the multi-angle polarization spectrum of the soil sample to be tested at the target temperature, and to decompose the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0043] The freeze-thaw stage determination module is used to determine the target freeze-thaw stage of the soil sample to be tested based on the target temperature and the preset freeze-thaw stage division rules.

[0044] The stage input feature subset generation module is used to obtain the experimental polarization component spectra of multiple experimental soil samples in each freeze-thaw stage, and to screen out the dominant polarization component and its corresponding preferred band that responds most significantly to the change in unfrozen water content in each freeze-thaw stage by performing correlation analysis on the spectra of multiple experimental polarization components and the measured unfrozen water content, so as to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0045] The input feature combination generation module is used to obtain the polarization component data of the dominant polarization component at the preferred band from the polarization component spectrum set based on the stage input feature subset corresponding to the target freeze-thaw stage, so as to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0046] The unfrozen water content prediction module is used to generate the predicted unfrozen water content of the soil sample to be tested at the target temperature based on the target input feature combination, using a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage. Each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

[0047] Preferably, after the unfrozen water content prediction module, the system further includes a monotonicity correction module, used for:

[0048] Obtain the polarization component spectral set of the soil sample to be tested at different temperatures and predict the unfrozen water content;

[0049] For each temperature point, the polarization component spectrum of the dominant polarization component corresponding to the current temperature point is obtained from the set of polarization component spectra corresponding to the current temperature point. Based on the polarization component spectrum of the dominant polarization component corresponding to the current temperature point and the polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point, the band spectral change rate of each band is calculated respectively. The mean of all the band spectral change rates is determined as the polarization change intensity index of the current temperature point. Based on the polarization change intensity index, the physical monotonicity constraint modulation weight of the current temperature point is constructed. Based on the physical monotonicity constraint modulation weight, the physical monotonicity constraint optimization objective function of the current temperature point is constructed. The physical monotonicity constraint optimization objective function is solved to perform monotonicity correction processing on the predicted unfrozen water content of the current temperature point, and the predicted unfrozen water content after monotonicity correction is obtained. The current temperature point and the next adjacent temperature point are obtained by sorting all temperature points according to the temperature decrease rule or the temperature increase rule.

[0050] Preferably, after the monotonicity correction module, the system further includes a residual compensation module, used for:

[0051] Based on the pre-trained residual prediction model, residual estimation is performed according to the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage to obtain the corresponding residual estimation compensation amount.

[0052] The predicted unfrozen water content after monotonicity correction is summed with the residual estimate compensation to obtain the final predicted unfrozen water content of the soil sample under test at the target temperature.

[0053] Preferably, before the freeze-thaw stage determination module, the system further includes a freeze-thaw stage division rule construction module, used for:

[0054] Freeze-thaw cycle experiments were conducted on multiple experimental soil samples, and experimental multi-angle polarization spectra of each experimental soil sample were collected during the complete freeze-thaw process. The initial water content of each experimental soil sample was different, and the complete freeze-thaw process included multiple experimental temperature points. The experimental multi-angle polarization spectra included the experimental multi-angle polarization spectra of the experimental soil samples at each experimental temperature point.

[0055] The multi-angle polarization spectra of each experiment are decomposed to obtain multiple sets of experimental polarization component spectra, and a mapping relationship is established between the sets of polarization component spectra and the corresponding experimental temperature points and initial water content.

[0056] For each polarization component, the spectrum of all polarization components of the current polarization component is obtained. The spectrum of all polarization components is sorted according to the temperature decrease rule or the temperature increase rule to obtain the polarization component spectrum sequence of the current polarization component during the complete freeze-thaw process.

[0057] For each pair of adjacent experimental temperature points, based on the polarization component spectral sequence, the experimental band spectral change rate of the current polarization component between the current adjacent experimental temperature points is calculated, and the average value of all experimental band spectral change rates is determined as the experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points.

[0058] Obtain the experimental polarization change intensity index of all polarization components between the current adjacent experimental temperature points, determine the maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices as the experimental polarization change intensity index between the current adjacent experimental temperature points, and integrate the experimental polarization change intensity indices between all adjacent experimental temperature points to obtain the experimental polarization change intensity index sequence of the complete freeze-thaw process.

[0059] According to a preset screening method, continuous temperature ranges with polarization change intensity higher than the background fluctuation level are selected from the experimental polarization change intensity index sequence, and the continuous temperature ranges are determined as phase transition stages. Continuous temperature ranges located on the low-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as stable freezing stages. Continuous temperature ranges located on the high-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as melting and stabilizing stages. The preset screening method is any one of the statistical threshold method, peak ratio method, and extreme value identification method. The continuous temperature range contains at least two pairs of adjacent experimental temperature points.

[0060] Based on the temperature ranges corresponding to the stable freezing stage, the phase transition stage, and the thawing stabilization stage, a preset freeze-thaw stage division rule is constructed.

[0061] Preferably, the stage input feature subset generation module is used for:

[0062] Based on the experimental temperature points corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules, the freeze-thaw stage corresponding to each experimental polarization component spectral set is determined respectively.

[0063] Within each freeze-thaw stage, for each polarization component, based on the experimental polarization component spectrum of the current polarization component, the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content is calculated, and the correlation distribution data of the current polarization component within the current freeze-thaw stage is generated.

[0064] From the correlation distribution data, bands with absolute values ​​of correlation coefficients higher than a preset correlation coefficient threshold are selected to obtain a set of candidate sensitive bands for the current polarization component in the current freeze-thaw stage.

[0065] The stability of each candidate sensitive band in the candidate sensitive band set is evaluated to screen out candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error threshold or whose correlation dispersion coefficient exceeds a preset correlation dispersion threshold, so as to obtain the preferred band of the current polarization component in the current freeze-thaw stage.

[0066] The polarization components are comprehensively scored from the dimensions of the number of preferred bands, the correlation coefficient of preferred bands, and the stability evaluation results of preferred bands. The polarization component with the highest comprehensive score is determined as the dominant polarization component of the current freeze-thaw stage. Based on the dominant polarization component and the corresponding preferred band, a subset of stage input features corresponding to the current freeze-thaw stage is generated.

[0067] Preferably, the multi-angle polarization spectrum acquisition module is used for:

[0068] Using a polarization spectrometer, multi-angle polarization spectra of the soil sample to be tested are obtained from multiple preset polarization directions, wherein each preset polarization direction corresponds to a polarization spectrum, and the multi-angle polarization spectrum contains multiple polarization spectra;

[0069] By synthesizing the polarization spectra corresponding to all the preset polarization directions, the total reflectance component spectrum of the soil sample under test at the target temperature is obtained;

[0070] Two mutually orthogonal preset polarization directions are selected from all the preset polarization directions, and the polarization spectra corresponding to the two preset polarization directions are determined as the parallel polarization component spectrum and the vertical polarization component spectrum of the soil sample to be tested at the target temperature.

[0071] Stokes parameters are calculated based on the multi-angle polarization spectrum, and the linear polarization degree component spectrum and polarization angle component spectrum of the soil sample under test at the target temperature are calculated based on the Stokes parameters.

[0072] Preferably, before the unfrozen water content prediction module, the system further includes a staged unfrozen water content prediction model construction module, used for:

[0073] For each freeze-thaw stage, construct a prediction model for the initial unfrozen water content corresponding to the current freeze-thaw stage;

[0074] For each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, the experimental polarization component data of the dominant polarization component at the preferred band is obtained from the current experimental polarization component spectral set, and the experimental polarization component data, the experimental temperature point corresponding to the current experimental polarization component spectral set, and the initial water content are combined to generate the training sample of the current freeze-thaw stage.

[0075] Based on multiple training samples of the current freeze-thaw stage, the model for predicting the unfrozen water content of the initial stage is trained to obtain the stage unfrozen water content prediction model corresponding to the current freeze-thaw stage that has completed model training.

[0076] According to another aspect of this application, a storage medium is provided that stores at least one executable instruction, which causes a processor to perform an operation corresponding to the above-described method for predicting unfrozen water content based on polarization spectroscopy.

[0077] According to another aspect of this application, a terminal is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0078] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described method for predicting the content of unfrozen water based on polarization spectroscopy.

[0079] By employing the above technical solutions, the technical solutions provided in the embodiments of this application have at least the following advantages:

[0080] This application provides a method, system, medium, and terminal for predicting unfrozen water content based on polarization spectroscopy. First, multi-angle polarization spectra of the soil sample to be tested are collected at a target temperature, and the multi-angle polarization spectra are decomposed to obtain a set of polarization component spectra of the soil sample at the target temperature. Further, based on the target temperature and a preset freeze-thaw stage division rule, the target freeze-thaw stage of the soil sample is determined. Further, experimental polarization component spectra of multiple experimental soil samples are obtained within each freeze-thaw stage, and by performing correlation analysis between the multiple experimental polarization component spectra and the measured unfrozen water content, the most significant responses to changes in unfrozen water content within each freeze-thaw stage are selected. The dominant polarization component of the stage and its corresponding preferred band are used to generate a stage input feature subset corresponding to each freeze-thaw stage. Further, based on the stage input feature subset corresponding to the target freeze-thaw stage, polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectrum set to generate a target input feature combination for the soil sample at the target temperature. Finally, based on a stage unfrozen water content prediction model that has been trained and matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model. Compared with existing technologies, this application's embodiments effectively separate specular reflection components by decomposing multi-angle polarization spectra into multiple polarization component spectra, suppressing the interference of ice crystal formation on reflection signals, and eliminating noise unrelated to unfrozen water content, thereby improving the prediction accuracy of unfrozen water content. Furthermore, by adaptively dividing the freeze-thaw stages based on the full-band change rate of the polarization spectrum, it overcomes the limitations of traditional methods that rely on fixed temperature thresholds or empirical critical values, achieving accurate identification of the stable freezing stage, phase transition stage, and thawing stabilization stage. At the same time, by selecting the dominant polarization component and preferred band that are most sensitive to changes in unfrozen water within each stage, redundant features are eliminated, avoiding the information redundancy problem introduced by directly splicing multiple polarization components. Moreover, by independently modeling and predicting in stages, the prediction model for each stage can focus on the physical laws within the corresponding stage, avoiding interference caused by differences in mechanisms between different freeze-thaw stages on the unified model, further improving the prediction accuracy of unfrozen water content.

[0081] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0082] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0083] Figure 1 A flowchart of a method for predicting unfrozen water content based on polarization spectroscopy, provided in an embodiment of this application, is shown.

[0084] Figure 2 A flowchart of another method for predicting unfrozen water content based on polarization spectroscopy provided in an embodiment of this application is shown;

[0085] Figure 3 This paper illustrates a block diagram of an unfrozen water content prediction system based on polarization spectroscopy, provided in an embodiment of this application.

[0086] Figure 4 A schematic diagram of the structure of a terminal provided in an embodiment of this application is shown. Detailed Implementation

[0087] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0088] This application provides a method for predicting the content of unfrozen water based on polarization spectroscopy, such as... Figure 1 As shown, the method includes:

[0089] 101. Collect the multi-angle polarization spectrum of the soil sample to be tested at the target temperature, and decompose the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0090] The target temperature is used to characterize the temperature of the soil sample being tested. The multi-angle polarization spectrum contains polarization spectra with multiple polarization directions, each corresponding to a polarization spectrum. The polarization directions can take values ​​of 0°, 45°, 90°, and 135°. The set of polarization component spectra includes the total reflectance component spectrum, parallel polarization component spectrum, perpendicular polarization component spectrum, linear polarization degree component spectrum, and polarization angle component spectrum. The total reflectance component spectrum is synthesized from polarization spectra of all polarization directions. The parallel polarization component spectrum and perpendicular polarization component spectrum are used to characterize the polarization spectra of two mutually orthogonal polarization directions; that is, the polarization spectra at 0° and 90° are defined as the parallel polarization component spectrum and perpendicular polarization component spectrum, respectively. The linear polarization degree component spectrum measures the proportion of polarized light in the reflected light and is sensitive to specular reflection caused by ice crystals. The polarization angle component spectrum characterizes the principal vibration direction of polarized light and can reflect the orientation information of the surface microstructure. In this embodiment, the current execution terminal can be the unfrozen water content prediction module in a freeze-thaw soil analysis system.

[0091] Optionally, the spectra of each polarization component in the polarization component spectrum set can be preprocessed, and subsequent steps can be performed based on the preprocessed spectra of each polarization component. Preprocessing includes, but is not limited to, smoothing and denoising, band selection, resampling to a uniform wavelength range, and normalization or standardization.

[0092] 102. Based on the target temperature and the preset freeze-thaw stage division rules, determine the target freeze-thaw stage of the soil sample to be tested.

[0093] The freeze-thaw stage includes a stable freezing stage, a phase transition stage, and a thawing-stable stage. The phase transition stage characterizes a continuous temperature range where the polarization change intensity is higher than the background fluctuation level. The stable freezing stage characterizes a continuous temperature range located at the low temperature side of the phase transition stage and where the polarization change intensity is lower than the background fluctuation level. The thawing-stable stage characterizes a continuous temperature range located at the high temperature side of the phase transition stage and where the polarization change intensity is lower than the background fluctuation level. The preset freeze-thaw stage division rule is constructed by collecting experimental multi-angle polarization spectra of multiple experimental soil samples during the freeze-thaw process. Based on the full-band spectral change rate of the experimental multi-angle polarization spectra under different temperature conditions, a polarization change intensity index is constructed. The freeze-thaw stage division operation based on the polarization change intensity index is then used to construct the temperature range corresponding to each freeze-thaw stage. For example, below -5℃ is the stable freezing stage, [-5℃, 0℃] is the phase transition stage, and above 0℃ is the thawing-stable stage.

[0094] It is understandable that since the temperature range has been clearly defined when dividing the freeze-thaw stage, in this embodiment of the application, the freeze-thaw stage of the soil sample to be tested can be directly determined based on the target temperature and denoted as the target freeze-thaw stage.

[0095] 103. Obtain the experimental polarization component spectra of multiple experimental soil samples during each freeze-thaw stage, and by performing correlation analysis between the spectra of multiple experimental polarization components and the measured unfrozen water content, screen out the dominant polarization component of each freeze-thaw stage that responds most significantly to the change in unfrozen water content and the corresponding preferred band, so as to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0096] The dominant polarization component of a stage is used to characterize the polarization component that responds most significantly to changes in unfrozen water content within each freeze-thaw stage. The preferred band is used to characterize the band that responds significantly and stably to changes in unfrozen water content within the current freeze-thaw stage for the current polarization component. The stage input feature subset contains the name of the dominant polarization component of that freeze-thaw stage and its corresponding preferred band.

[0097] 104. Based on the stage input feature subset corresponding to the target freeze-thaw stage, obtain the polarization component data of the dominant polarization component at the preferred band from the polarization component spectrum set to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0098] For example, assuming the target freeze-thaw stage is a stable freezing stage, according to step 103 of the embodiment, the stage input feature subset corresponding to the stable freezing stage is [stage dominant polarization component: vertical polarization component spectrum, preferred band (a, b)]. Then, the vertical polarization component spectrum is extracted from the polarization component spectrum set obtained in step 101 of the embodiment, and the vertical polarization component data corresponding to band (a, b) is obtained from the vertical polarization component spectrum. This data is then combined with the initial moisture content of the soil sample to be tested and the target temperature to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0099] 105. Based on the pre-trained model for predicting the unfrozen water content of the target soil sample at the target temperature, and according to the combination of target input features, generate the predicted unfrozen water content of the soil sample to be tested at the target temperature.

[0100] Each freeze-thaw stage corresponds to a prediction model for the unfrozen water content of that stage.

[0101] In this embodiment, the target input feature combination generated in step 104 of the embodiment is input into the unfrozen water content prediction model that has been trained and matches the target freeze-thaw stage, so as to obtain the predicted unfrozen water content of the soil sample to be tested at the target temperature.

[0102] Compared with existing technologies, this application's embodiments effectively separate specular reflection components by decomposing multi-angle polarization spectra into multiple polarization component spectra, suppressing the interference of ice crystal formation on reflection signals, and eliminating noise unrelated to unfrozen water content, thereby improving the prediction accuracy of unfrozen water content. Furthermore, by adaptively dividing the freeze-thaw stages based on the full-band change rate of the polarization spectrum, it overcomes the limitations of traditional methods that rely on fixed temperature thresholds or empirical critical values, achieving accurate identification of the stable freezing stage, phase transition stage, and thawing stabilization stage. At the same time, by selecting the dominant polarization component and preferred band that are most sensitive to changes in unfrozen water within each stage, redundant features are eliminated, avoiding the information redundancy problem introduced by directly splicing multiple polarization components. Moreover, by independently modeling and predicting in stages, the prediction model for each stage can focus on the physical laws within the corresponding stage, avoiding interference caused by differences in mechanisms between different freeze-thaw stages on the unified model, further improving the prediction accuracy of unfrozen water content.

[0103] This application provides another method for predicting unfrozen water content based on polarization spectroscopy, such as... Figure 2 As shown, the method includes:

[0104] 201. Construct a pre-defined rule for dividing the freeze-thaw stage.

[0105] Accordingly, step 201 of the embodiment specifically includes: conducting freeze-thaw cycle experiments on multiple experimental soil samples, and collecting experimental multi-angle polarization spectral sets of each experimental soil sample during the complete freeze-thaw process. The initial water content of each experimental soil sample is different, and the complete freeze-thaw process includes multiple experimental temperature points. The experimental multi-angle polarization spectral sets contain the experimental multi-angle polarization spectra of the experimental soil samples at each experimental temperature point. Each experimental multi-angle polarization spectrum is decomposed to obtain multiple experimental polarization component spectral sets, and a mapping relationship is established between the polarization component spectral sets and the corresponding experimental temperature points and initial water content. For each polarization component, all polarization component spectra of the current polarization component are obtained, and all polarization component spectra are sorted according to a temperature decrease rule or a temperature increase rule to obtain the polarization component spectral sequence of the current polarization component during the complete freeze-thaw process. For each pair of adjacent experimental temperature points, based on the polarization component spectral sequence, the experimental band spectral change rate of the current polarization component between the current adjacent experimental temperature points is calculated, and the average value of the experimental band spectral change rate is determined as the experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points. The experimental polarization change intensity index of all polarization components between current adjacent experimental temperature points is determined by taking the maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices. This is then integrated to obtain a complete sequence of experimental polarization change intensity indices for the freeze-thaw process. Following a pre-defined screening method, continuous temperature intervals with polarization change intensity higher than the background fluctuation level are selected from the sequence and designated as the phase transition stage. Furthermore, continuous temperature intervals located on the low-temperature side of the phase transition stage with polarization change intensity lower than the background fluctuation level are designated as the stable freezing stage, and continuous temperature intervals located on the high-temperature side of the phase transition stage with polarization change intensity lower than the background fluctuation level are designated as the thawing stable stage. The pre-defined screening method can be any one of the statistical threshold method, peak ratio method, or extreme value identification method. Each continuous temperature interval must contain at least two pairs of adjacent experimental temperature points. Based on the temperature intervals corresponding to the stable freezing stage, phase transition stage, and thawing stable stage, a pre-defined freeze-thaw stage division rule is constructed.

[0106] In this embodiment, firstly, freeze-thaw cycle experiments are conducted on multiple experimental soil samples under laboratory or field conditions, with each sample having a different initial moisture content. For each soil sample, experimental multi-angle polarization spectra are collected at different experimental temperature points during a complete freeze-thaw process. Further, each experimental multi-angle polarization spectrum is decomposed to obtain the corresponding experimental polarization component spectrum set, and a mapping relationship is established between this polarization component spectrum set and the corresponding experimental temperature point and initial moisture content. Further, for each polarization component, it is considered the current polarization component, and all polarization component spectra of the current polarization component are obtained, with each polarization component spectrum corresponding to an experimental temperature point. Subsequently, according to a temperature decrease rule or a temperature increase rule, all the above polarization component spectra are sorted to obtain the polarization component spectrum sequence of the current polarization component during the complete freeze-thaw process. Further, for each pair of adjacent experimental temperature points... and Based on the polarization component spectral sequence of the current polarization component, calculate the rate of change of the experimental band spectrum of the current polarization component at each wavelength between the current adjacent experimental temperature points. The calculation formula is as follows:

[0107] ,

[0108] in, Indicates the current polarization component At adjacent experimental temperature points and Between wavelengths The rate of spectral change at the experimental band, Indicates the current polarization component At the experimental temperature point Time wavelength Polarization component data at that location, Indicates the current polarization component At the experimental temperature point Time wavelength Polarization component data at the location;

[0109] The mean of the spectral change rate across all experimental bands is determined as the index of the experimental polarization change intensity of the current polarization component between adjacent experimental temperature points.

[0110] ,

[0111] in, Indicates the current polarization component At adjacent experimental temperature points and The intensity index of experimental polarization change between them Indicates the number of effective bands. Indicates the shortest wavelength. Indicates the longest wavelength.

[0112] Furthermore, the experimental polarization change intensity index of all polarization components between the current adjacent experimental temperature points is obtained. The maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices is determined as the experimental polarization change intensity index between the current adjacent experimental temperature points. The experimental polarization change intensity indices between all adjacent experimental temperature points are then integrated to obtain the complete experimental polarization change intensity index sequence during the freeze-thaw process. Furthermore, continuous temperature ranges with polarization change intensity higher than the background fluctuation level are selected from the experimental polarization change intensity index sequence, and these continuous temperature ranges are identified as phase transition stages. The preset selection method can be any one of the statistical threshold method, peak ratio method, or extreme value identification method. Specifically, the statistical threshold method selects continuous temperature ranges exceeding a preset change intensity threshold based on the statistical characteristics of the experimental polarization change intensity index sequence. The preset change intensity threshold can be the mean of the experimental polarization change intensity index sequence plus a certain multiple of the standard deviation. The peak ratio method selects continuous temperature ranges exceeding the maximum change intensity by a certain proportion, such as 50%. The extreme value identification method automatically detects continuous extreme value ranges. After determining the phase transition stages, continuous temperature ranges located on the low-temperature side of the phase transition stage with polarization change intensity lower than the background fluctuation level are identified as the stable freezing stage, and continuous temperature ranges located on the high-temperature side of the phase transition stage with polarization change intensity lower than the background fluctuation level are identified as the melting and stabilizing stage. Finally, based on the temperature ranges corresponding to the stable freezing stage, the phase transition stage, and the thawing stable stage, a preset freeze-thaw stage division rule is constructed. For example, less than -5℃ is the stable freezing stage, [-5℃, 0℃] is the phase transition stage, and greater than 0℃ is the thawing stable stage.

[0113] 202. Obtain the experimental polarization component spectra of multiple experimental soil samples during each freeze-thaw stage, and by performing correlation analysis between the spectra of multiple experimental polarization components and the measured unfrozen water content, screen out the dominant polarization component of each freeze-thaw stage that responds most significantly to the change in unfrozen water content and the corresponding preferred band, so as to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0114] Accordingly, step 202 of the embodiment specifically includes: determining the freeze-thaw stage corresponding to each experimental polarization component spectral set based on the experimental temperature point corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules; within each freeze-thaw stage, for each polarization component, based on the experimental polarization component spectrum of the current polarization component, calculating the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content, and generating the correlation distribution data of the current polarization component within the current freeze-thaw stage; from the correlation distribution data, selecting the bands with an absolute value of the correlation coefficient higher than the preset correlation coefficient threshold, to obtain the correlation distribution data of the current polarization component within the current freeze-thaw stage. A candidate sensitive band set is generated; the stability of each candidate sensitive band in the candidate sensitive band set is evaluated to filter out candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error or whose correlation dispersion coefficient exceeds a preset correlation dispersion coefficient threshold, so as to obtain the preferred bands of the current polarization component in the current freeze-thaw stage; each polarization component is comprehensively scored from the dimensions of the number of preferred bands, the correlation coefficient of preferred bands, and the stability evaluation results of preferred bands, and the polarization component with the highest comprehensive score is determined as the stage dominant polarization component of the current freeze-thaw stage, and the stage input feature subset corresponding to the current freeze-thaw stage is generated based on the stage dominant polarization component and the corresponding preferred bands.

[0115] In this embodiment, firstly, based on the experimental temperature points corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules constructed in step 201 of the embodiment, the freeze-thaw stage corresponding to each experimental polarization component spectral set is determined. Further, within each freeze-thaw stage, for each polarization component, it is considered as the current polarization component. Based on the experimental polarization component spectrum of the current polarization component, the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content is calculated. Any one of the following methods can be used: Pearson correlation coefficient, Spearman rank correlation coefficient, or sensitivity measurement method based on regression residuals. Correlation distribution data of the current polarization component within the current freeze-thaw stage is generated, i.e., a sensitivity curve, with wavelength on the horizontal axis and correlation coefficient on the vertical axis. Further, from this correlation distribution data, bands with absolute correlation coefficient values ​​higher than a preset correlation coefficient threshold are selected to obtain a candidate sensitive band set for the current polarization component within the current freeze-thaw stage. These bands are considered to respond significantly to changes in unfrozen water content. Furthermore, to avoid noise interference or the influence of local abnormal bands, a stability assessment is performed on each candidate sensitive band in the candidate sensitive band set. This eliminates candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error threshold or whose correlation dispersion coefficient exceeds a preset correlation dispersion coefficient threshold. This yields the preferred bands for the current polarization component within the current freeze-thaw stage. The stability assessment can be achieved by calculating the correlation dispersion between bands or by evaluating the prediction error stability of these bands when modeled individually using cross-validation. An individual modeling prediction error exceeding the preset prediction error threshold indicates poor generalization ability for that band; a correlation dispersion coefficient exceeding the preset correlation dispersion coefficient threshold indicates significant fluctuations in that band. Finally, each polarization component is comprehensively scored based on the number of preferred bands, the correlation coefficient of preferred bands, and the stability assessment results of preferred bands. The polarization component with a large number of preferred bands, a concentrated and stable correlation distribution, and a low prediction error within the freeze-thaw stage is identified as the dominant polarization component for the current freeze-thaw stage. Based on the dominant polarization component and its corresponding preferred bands, a subset of stage input features corresponding to the current freeze-thaw stage is generated.

[0116] It should be noted that if there are multiple polarization components with the highest overall scores, they can be regarded as the dominant polarization components in the cooperative stage.

[0117] 203. Construct a prediction model for the unfrozen water content of each freeze-thaw stage that has been trained.

[0118] Accordingly, step 203 of the embodiment specifically includes: for each freeze-thaw stage, constructing an initial stage unfrozen water content prediction model corresponding to the current freeze-thaw stage; for each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, obtaining experimental polarization component data of the dominant polarization component at the preferred band from the current experimental polarization component spectral set, and combining the experimental polarization component data, the experimental temperature point corresponding to the current experimental polarization component spectral set, and the initial water content to generate training samples for the current freeze-thaw stage; and training the initial stage unfrozen water content prediction model based on multiple training samples of the current freeze-thaw stage to obtain a stage unfrozen water content prediction model that has completed model training for the current freeze-thaw stage.

[0119] In this embodiment, firstly, for each freeze-thaw stage, a prediction model for the initial unfrozen water content corresponding to the current freeze-thaw stage is constructed. This model can employ linear regression, partial least squares, support vector machine, random forest, gradient boosting, or neural network models, etc., and is not specifically limited in this embodiment. Further, for each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, experimental polarization component data of the dominant polarization component at the preferred band is obtained from the current experimental polarization component spectral set. For details, please refer to the detailed explanation in step 104 of the embodiment, which will not be repeated here. The obtained experimental polarization component data, along with the experimental temperature point and initial water content corresponding to the current experimental polarization component spectral set, are combined to generate training samples for the current freeze-thaw stage. Finally, based on multiple training samples from the current freeze-thaw stage, the model for predicting the unfrozen water content in the initial stage is trained. The model parameters can be optimized through cross-validation, and after performance evaluation, the unfrozen water content prediction model for the current freeze-thaw stage with completed model training is obtained. The performance evaluation indicators include, but are not limited to, the coefficient of determination, root mean square error, and mean absolute error.

[0120] It should be noted that if there are multiple candidate models for the same freeze-thaw stage, the model with the smallest prediction error or the best generalization ability will be selected as the final prediction model for that freeze-thaw stage.

[0121] 204. Collect the multi-angle polarization spectrum of the soil sample to be tested at the target temperature, and decompose the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0122] Accordingly, step 204 of the embodiment specifically includes: using a polarization spectrometer to acquire multi-angle polarization spectra of the soil sample to be tested from multiple preset polarization directions, wherein each preset polarization direction corresponds to a polarization spectrum, and the multi-angle polarization spectrum contains multiple polarization spectra; synthesizing the polarization spectra corresponding to all preset polarization directions to obtain the total reflectance component spectrum of the soil sample to be tested at the target temperature; selecting two mutually orthogonal preset polarization directions from all preset polarization directions, and determining the polarization spectra corresponding to the two preset polarization directions as the parallel polarization component spectrum and the perpendicular polarization component spectrum of the soil sample to be tested at the target temperature; calculating the Stokes parameters based on the multi-angle polarization spectra, and calculating the linear polarization degree component spectrum and the polarization angle component spectrum of the soil sample to be tested at the target temperature based on the Stokes parameters.

[0123] In this embodiment, polarization spectra of the soil sample to be tested are first acquired using a polarization spectrometer from multiple preset polarization directions, such as 0°, 45°, 90°, and 135°, to obtain multi-angle polarization spectra. Further, all polarization spectra from these multi-angle polarization spectra are synthesized to obtain the total reflectance component spectrum. Simultaneously, two mutually orthogonal preset polarization directions are selected from all preset polarization directions, and the polarization spectra corresponding to these two preset polarization directions are determined as the parallel polarization component spectrum and the perpendicular polarization component spectrum of the soil sample to be tested at the target temperature; that is, the polarization spectra at 0° and 90° are determined as the parallel polarization component spectrum and the perpendicular polarization component spectrum. Further, Stokes parameters are calculated. and ,in, This represents the difference between the horizontal polarization component spectrum and the vertical polarization component spectrum, that is, the difference between the 0° polarization component spectrum and the 90° polarization component spectrum; This represents the difference between the 45° polarization component spectrum and the 135° polarization component spectrum, i.e., the difference between the 45° polarization component spectrum and the 135° polarization component spectrum; subsequently, based on the Stokes parameters, the linear polarization component spectrum of the soil sample under test at the target temperature is calculated.

[0124] ,

[0125] in, Represents the linear polarization degree component spectrum. Represents the total reflectance component spectrum;

[0126] Polarization angular component spectrum

[0127] ,

[0128] in, Represents the polarization angular component spectrum;

[0129] The polarization degree component spectrum can also be calculated based on the parallel polarization component spectrum and the perpendicular polarization component spectrum, using the following formula:

[0130] ,

[0131] in, Represents the parallel polarization component spectrum. This represents the vertical polarization component spectrum.

[0132] In addition, the parallel polarization component spectrum, the vertical polarization component spectrum, and the total reflection component spectrum can be obtained directly by a polarization spectrometer or calculated based on Stokes parameters.

[0133] 205. Based on the target temperature and the preset freeze-thaw stage division rules, determine the target freeze-thaw stage of the soil sample to be tested.

[0134] 206. Based on the stage input feature subset corresponding to the target freeze-thaw stage, obtain the polarization component data of the dominant polarization component at the preferred band from the polarization component spectrum set to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0135] 207. Based on the unfrozen water content prediction model of the stage that has been trained and matched with the target freeze-thaw stage, the predicted unfrozen water content of the soil sample to be tested at the target temperature is generated according to the combination of target input features.

[0136] Steps 205-207 of the embodiment can be referred to in the detailed description of steps 102, 104 and 105 of the aforementioned embodiment, and will not be repeated here.

[0137] 208. The predicted unfrozen water content of the soil sample to be tested at the target temperature is subjected to monotonicity correction to obtain the predicted unfrozen water content after monotonicity correction.

[0138] Accordingly, step 208 of the embodiment specifically includes: acquiring the set of polarization component spectra of the soil sample to be tested at different temperatures and predicting the unfrozen water content; for each temperature point, acquiring the polarization component spectrum of the dominant polarization component corresponding to the current temperature point from the set of polarization component spectra corresponding to the current temperature point; calculating the band spectral change rate of each band based on the polarization component spectrum of the dominant polarization component corresponding to the current temperature point and the polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point; determining the average value of the spectral change rate of all bands as the polarization change intensity index of the current temperature point; constructing the physical monotonicity constraint modulation weight of the current temperature point based on the polarization change intensity index; constructing the physical monotonicity constraint optimization objective function of the current temperature point based on the physical monotonicity constraint modulation weight; solving the physical monotonicity constraint optimization objective function to perform monotonicity correction processing on the predicted unfrozen water content of the current temperature point, and obtaining the predicted unfrozen water content after monotonicity correction. The current temperature point and the next adjacent temperature point are obtained by sorting all temperature points according to the temperature decrease rule or the temperature increase rule.

[0139] Regarding the predicted unfrozen water content of the soil sample obtained in step 207 of the embodiment at the target temperature, in order to ensure that it conforms to the basic physical law that the unfrozen water content does not decrease with increasing temperature, and to improve the rationality and reliability of the prediction results, in this embodiment, the polarization component spectrum set of the soil sample at different temperatures and the predicted unfrozen water content are first obtained. Further, for each temperature point, it is considered the current temperature point. From the corresponding polarization component spectrum set, the polarization component spectrum of the dominant polarization component corresponding to the current temperature point is obtained, denoted as polarization component spectrum 1. The polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point is denoted as polarization component spectrum 2. Based on polarization component spectrum 1 and polarization component spectrum 2, the band spectral change rate of each band is calculated, and the average value of the spectral change rate of all bands is determined as the polarization change intensity index of the current temperature point. Further, based on this polarization change intensity index, a physical monotonicity constraint modulation weight for the current temperature point is constructed.

[0140] ,

[0141] in, Indicates temperature point The physical monotonicity constraint modulates the weights. Indicates temperature point The intensity index of polarization change, This indicates the maximum value of the polarization change intensity index;

[0142] Based on this physical monotonicity constraint, the modulated weights are used to construct the physical monotonicity constraint optimization objective function for the current temperature point.

[0143] ,

[0144] ,

[0145] in, This represents the objective function for optimizing the physical monotonicity constraint at the current temperature point. This indicates the error in the model for predicting the unfrozen water content at a given stage. Indicates the adjustment parameter. Indicates a monotonicity penalty term. Representing temperature point Predicted unfrozen water content, Representing temperature point The predicted unfrozen water content.

[0146] Finally, the objective function of the physical monotonicity constraint is solved to perform monotonic correction on the predicted unfrozen water content at the current temperature point, thus obtaining the predicted unfrozen water content after monotonic correction.

[0147] 209. Perform residual compensation on the predicted unfrozen water content after completing monotonicity correction to obtain the final predicted unfrozen water content of the soil sample to be tested at the target temperature.

[0148] Accordingly, step 209 of the embodiment specifically includes: based on the pre-trained residual prediction model, performing residual estimation operation according to the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage to obtain the corresponding residual estimation compensation amount; summing the predicted unfrozen water content after monotonicity correction with the residual estimation compensation amount to obtain the final predicted unfrozen water content of the soil sample to be tested at the target temperature.

[0149] In this embodiment, for each training sample, the residual between the predicted unfrozen water content after monotonicity correction and the measured unfrozen water content is calculated in advance. The polarization change intensity index and freeze-thaw stage label of the training sample are used as model inputs, and the residual of the training sample is used as the model output to construct a residual prediction model, which can employ linear regression or a simple nonlinear model. During residual compensation, the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage are directly input into the residual prediction model to obtain the corresponding residual estimate compensation amount. The predicted unfrozen water content after monotonicity correction is summed with the residual estimate compensation amount to obtain the final predicted unfrozen water content of the soil sample at the target temperature.

[0150] This application provides a method for predicting unfrozen water content based on polarization spectroscopy. First, multi-angle polarization spectra of the soil sample to be tested are collected at a target temperature, and the multi-angle polarization spectra are decomposed to obtain the set of polarization component spectra of the soil sample at the target temperature. Further, based on the target temperature and a preset freeze-thaw stage division rule, the target freeze-thaw stage of the soil sample is determined. Further, experimental polarization component spectra of multiple experimental soil samples are obtained within each freeze-thaw stage, and by performing correlation analysis between the multiple experimental polarization component spectra and the measured unfrozen water content, the stage with the most significant response to changes in unfrozen water content within each freeze-thaw stage is selected. The polarization components and their corresponding preferred bands are used to generate a subset of stage input features corresponding to each freeze-thaw stage. Further, based on the subset of stage input features corresponding to the target freeze-thaw stage, polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectral set to generate a target input feature combination for the soil sample at the target temperature. Finally, based on a stage unfrozen water content prediction model that has been trained and matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model. Compared with existing technologies, this application's embodiments effectively separate specular reflection components by decomposing multi-angle polarization spectra into multiple polarization component spectra, suppressing the interference of ice crystal formation on reflection signals, and eliminating noise unrelated to unfrozen water content, thereby improving the prediction accuracy of unfrozen water content. Furthermore, by adaptively dividing the freeze-thaw stages based on the full-band change rate of the polarization spectrum, it overcomes the limitations of traditional methods that rely on fixed temperature thresholds or empirical critical values, achieving accurate identification of the stable freezing stage, phase transition stage, and thawing stabilization stage. At the same time, by selecting the dominant polarization component and preferred band that are most sensitive to changes in unfrozen water within each stage, redundant features are eliminated, avoiding the information redundancy problem introduced by directly splicing multiple polarization components. Moreover, by independently modeling and predicting in stages, the prediction model for each stage can focus on the physical laws within the corresponding stage, avoiding interference caused by differences in mechanisms between different freeze-thaw stages on the unified model, further improving the prediction accuracy of unfrozen water content.

[0151] Furthermore, as a response to the above Figure 1 The implementation of the method shown in this application provides a system for predicting the content of unfrozen water based on polarization spectroscopy, such as... Figure 3 As shown, the system includes:

[0152] Multi-angle polarization spectrum acquisition module 31, freeze-thaw stage determination module 32, stage input feature subset generation module 33, input feature combination generation module 34, and unfrozen water content prediction module 35.

[0153] The multi-angle polarization spectrum acquisition module 31 is used to acquire the multi-angle polarization spectrum of the soil sample to be tested at the target temperature, and to decompose the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0154] The freeze-thaw stage determination module 32 is used to determine the target freeze-thaw stage of the soil sample to be tested based on the target temperature and the preset freeze-thaw stage division rules.

[0155] The stage input feature subset generation module 33 is used to obtain the experimental polarization component spectra of multiple experimental soil samples in each freeze-thaw stage, and to screen out the dominant polarization component and its corresponding preferred band in each freeze-thaw stage by performing correlation analysis between the multiple experimental polarization component spectra and the measured unfrozen water content, so as to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0156] The input feature combination generation module 34 is used to obtain the polarization component data of the dominant polarization component at the preferred band from the polarization component spectrum set based on the stage input feature subset corresponding to the target freeze-thaw stage, so as to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0157] The unfrozen water content prediction module 35 is used to generate the predicted unfrozen water content of the soil sample to be tested at the target temperature based on the target input feature combination and a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage. Each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

[0158] In specific application scenarios, after the unfrozen water content prediction module, the system further includes a monotonicity correction module, used for:

[0159] Obtain the polarization component spectral set of the soil sample to be tested at different temperatures and predict the unfrozen water content;

[0160] For each temperature point, the polarization component spectrum of the dominant polarization component corresponding to the current temperature point is obtained from the set of polarization component spectra corresponding to the current temperature point. Based on the polarization component spectrum of the dominant polarization component corresponding to the current temperature point and the polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point, the band spectral change rate of each band is calculated respectively. The mean of all the band spectral change rates is determined as the polarization change intensity index of the current temperature point. Based on the polarization change intensity index, the physical monotonicity constraint modulation weight of the current temperature point is constructed. Based on the physical monotonicity constraint modulation weight, the physical monotonicity constraint optimization objective function of the current temperature point is constructed. The physical monotonicity constraint optimization objective function is solved to perform monotonicity correction processing on the predicted unfrozen water content of the current temperature point, and the predicted unfrozen water content after monotonicity correction is obtained. The current temperature point and the next adjacent temperature point are obtained by sorting all temperature points according to the temperature decrease rule or the temperature increase rule.

[0161] In specific application scenarios, after the monotonicity correction module, the system further includes a residual compensation module, used for:

[0162] Based on the pre-trained residual prediction model, residual estimation is performed according to the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage to obtain the corresponding residual estimation compensation amount.

[0163] The predicted unfrozen water content after monotonicity correction is summed with the residual estimate compensation to obtain the final predicted unfrozen water content of the soil sample under test at the target temperature.

[0164] In specific application scenarios, before the freeze-thaw stage determination module, the system further includes a freeze-thaw stage division rule construction module, used for:

[0165] Freeze-thaw cycle experiments were conducted on multiple experimental soil samples, and experimental multi-angle polarization spectra of each experimental soil sample were collected during the complete freeze-thaw process. The initial water content of each experimental soil sample was different, and the complete freeze-thaw process included multiple experimental temperature points. The experimental multi-angle polarization spectra included the experimental multi-angle polarization spectra of the experimental soil samples at each experimental temperature point.

[0166] The multi-angle polarization spectra of each experiment are decomposed to obtain multiple sets of experimental polarization component spectra, and a mapping relationship is established between the sets of polarization component spectra and the corresponding experimental temperature points and initial water content.

[0167] For each polarization component, the spectrum of all polarization components of the current polarization component is obtained. The spectrum of all polarization components is sorted according to the temperature decrease rule or the temperature increase rule to obtain the polarization component spectrum sequence of the current polarization component during the complete freeze-thaw process.

[0168] For each pair of adjacent experimental temperature points, based on the polarization component spectral sequence, the experimental band spectral change rate of the current polarization component between the current adjacent experimental temperature points is calculated, and the average value of all experimental band spectral change rates is determined as the experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points.

[0169] Obtain the experimental polarization change intensity index of all polarization components between the current adjacent experimental temperature points, determine the maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices as the experimental polarization change intensity index between the current adjacent experimental temperature points, and integrate the experimental polarization change intensity indices between all adjacent experimental temperature points to obtain the experimental polarization change intensity index sequence of the complete freeze-thaw process.

[0170] According to a preset screening method, continuous temperature ranges with polarization change intensity higher than the background fluctuation level are selected from the experimental polarization change intensity index sequence, and the continuous temperature ranges are determined as phase transition stages. Continuous temperature ranges located on the low-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as stable freezing stages. Continuous temperature ranges located on the high-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as melting and stabilizing stages. The preset screening method is any one of the statistical threshold method, peak ratio method, and extreme value identification method. The continuous temperature range contains at least two pairs of adjacent experimental temperature points.

[0171] Based on the temperature ranges corresponding to the stable freezing stage, the phase transition stage, and the thawing stabilization stage, a preset freeze-thaw stage division rule is constructed.

[0172] In specific application scenarios, the stage input feature subset generation module is used for:

[0173] Based on the experimental temperature points corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules, the freeze-thaw stage corresponding to each experimental polarization component spectral set is determined respectively.

[0174] Within each freeze-thaw stage, for each polarization component, based on the experimental polarization component spectrum of the current polarization component, the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content is calculated, and the correlation distribution data of the current polarization component within the current freeze-thaw stage is generated.

[0175] From the correlation distribution data, bands with absolute values ​​of correlation coefficients higher than a preset correlation coefficient threshold are selected to obtain a set of candidate sensitive bands for the current polarization component in the current freeze-thaw stage.

[0176] The stability of each candidate sensitive band in the candidate sensitive band set is evaluated to screen out candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error threshold or whose correlation dispersion coefficient exceeds a preset correlation dispersion threshold, so as to obtain the preferred band of the current polarization component in the current freeze-thaw stage.

[0177] The polarization components are comprehensively scored from the dimensions of the number of preferred bands, the correlation coefficient of preferred bands, and the stability evaluation results of preferred bands. The polarization component with the highest comprehensive score is determined as the dominant polarization component of the current freeze-thaw stage. Based on the dominant polarization component and the corresponding preferred band, a subset of stage input features corresponding to the current freeze-thaw stage is generated.

[0178] In specific application scenarios, the multi-angle polarization spectrum acquisition module is used for:

[0179] Using a polarization spectrometer, multi-angle polarization spectra of the soil sample to be tested are obtained from multiple preset polarization directions, wherein each preset polarization direction corresponds to a polarization spectrum, and the multi-angle polarization spectrum contains multiple polarization spectra;

[0180] By synthesizing the polarization spectra corresponding to all the preset polarization directions, the total reflectance component spectrum of the soil sample under test at the target temperature is obtained;

[0181] Two mutually orthogonal preset polarization directions are selected from all the preset polarization directions, and the polarization spectra corresponding to the two preset polarization directions are determined as the parallel polarization component spectrum and the vertical polarization component spectrum of the soil sample to be tested at the target temperature.

[0182] Stokes parameters are calculated based on the multi-angle polarization spectrum, and the linear polarization degree component spectrum and polarization angle component spectrum of the soil sample under test at the target temperature are calculated based on the Stokes parameters.

[0183] In specific application scenarios, prior to the unfrozen water content prediction module, the system further includes a staged unfrozen water content prediction model construction module, used for:

[0184] For each freeze-thaw stage, construct a prediction model for the initial unfrozen water content corresponding to the current freeze-thaw stage;

[0185] For each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, the experimental polarization component data of the dominant polarization component at the preferred band is obtained from the current experimental polarization component spectral set, and the experimental polarization component data, the experimental temperature point corresponding to the current experimental polarization component spectral set, and the initial water content are combined to generate the training sample of the current freeze-thaw stage.

[0186] Based on multiple training samples of the current freeze-thaw stage, the model for predicting the unfrozen water content of the initial stage is trained to obtain the stage unfrozen water content prediction model corresponding to the current freeze-thaw stage that has completed model training.

[0187] This application provides a system for predicting unfrozen water content based on polarization spectroscopy. First, multi-angle polarization spectra of the soil sample to be tested are collected at a target temperature, and these spectra are decomposed to obtain a set of polarization component spectra of the soil sample at the target temperature. Further, based on the target temperature and a preset freeze-thaw stage division rule, the target freeze-thaw stage of the soil sample is determined. Further, experimental polarization component spectra of multiple experimental soil samples are obtained within each freeze-thaw stage, and by performing correlation analysis between the multiple experimental polarization component spectra and the measured unfrozen water content, the stage with the most significant response to changes in unfrozen water content within each freeze-thaw stage is selected. The polarization components and their corresponding preferred bands are used to generate a subset of stage input features corresponding to each freeze-thaw stage. Further, based on the subset of stage input features corresponding to the target freeze-thaw stage, polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectral set to generate a target input feature combination for the soil sample at the target temperature. Finally, based on a stage unfrozen water content prediction model that has been trained and matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model. Compared with existing technologies, this application's embodiments effectively separate specular reflection components by decomposing multi-angle polarization spectra into multiple polarization component spectra, suppressing the interference of ice crystal formation on reflection signals, and eliminating noise unrelated to unfrozen water content, thereby improving the prediction accuracy of unfrozen water content. Furthermore, by adaptively dividing the freeze-thaw stages based on the full-band change rate of the polarization spectrum, it overcomes the limitations of traditional methods that rely on fixed temperature thresholds or empirical critical values, achieving accurate identification of the stable freezing stage, phase transition stage, and thawing stabilization stage. At the same time, by selecting the dominant polarization component and preferred band that are most sensitive to changes in unfrozen water within each stage, redundant features are eliminated, avoiding the information redundancy problem introduced by directly splicing multiple polarization components. Moreover, by independently modeling and predicting in stages, the prediction model for each stage can focus on the physical laws within the corresponding stage, avoiding interference caused by differences in mechanisms between different freeze-thaw stages on the unified model, further improving the prediction accuracy of unfrozen water content.

[0188] According to one embodiment of this application, a storage medium is provided that stores at least one executable instruction, which can execute the polarization spectroscopy-based method for predicting unfrozen water content in any of the above method embodiments.

[0189] Based on this understanding, the technical solution of this application can be embodied in the form of a software product. The software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, or portable hard drive), and includes several instructions to cause a computer device (such as a personal computer, server, or network device) to execute the methods described in the various implementation scenarios of this application.

[0190] Figure 4 The diagram shows a structural schematic of a terminal according to one embodiment of the present application. The specific embodiments of the present application do not limit the specific implementation of the terminal.

[0191] like Figure 4 As shown, the terminal may include: a processor 402, a communications interface 404, a memory 406, and a communications bus 408.

[0192] The processor 402, communication interface 404, and memory 406 communicate with each other via communication bus 408.

[0193] Communication interface 404 is used to communicate with other network elements such as clients or other servers.

[0194] The processor 402 is used to execute program 410, specifically the relevant steps in the above-described embodiment of the method for predicting unfrozen water content based on polarization spectroscopy.

[0195] Specifically, program 410 may include program code that includes computer operation instructions.

[0196] Processor 402 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The computer device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or they may be processors of different types, such as one or more CPUs and one or more ASICs.

[0197] Memory 406 is used to store program 410. Memory 406 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0198] Specifically, program 410 can be used to cause processor 402 to perform the following operations:

[0199] The multi-angle polarization spectrum of the soil sample to be tested is collected at the target temperature, and the multi-angle polarization spectrum is decomposed to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature.

[0200] Based on the target temperature and the preset freeze-thaw stage division rules, the target freeze-thaw stage of the soil sample to be tested is determined.

[0201] The experimental polarization component spectra of multiple experimental soil samples were obtained during each freeze-thaw stage. By performing correlation analysis between the spectra of the multiple experimental polarization components and the measured unfrozen water content, the dominant polarization component of each freeze-thaw stage that responds most significantly to the change in unfrozen water content and the corresponding preferred band were selected to generate the stage input feature subset corresponding to each freeze-thaw stage.

[0202] Based on the stage input feature subset corresponding to the target freeze-thaw stage, the polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectrum set to generate the target input feature combination of the soil sample to be tested at the target temperature.

[0203] Based on a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample to be tested at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

[0204] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the physical device for predicting unfrozen water content based on polarization spectroscopy, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0205] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for predicting unfrozen water content based on polarization spectroscopy, characterized in that, include: The multi-angle polarization spectrum of the soil sample to be tested is collected at the target temperature, and the multi-angle polarization spectrum is decomposed to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature. Based on the target temperature and the preset freeze-thaw stage division rules, the target freeze-thaw stage of the soil sample to be tested is determined. The experimental polarization component spectra of multiple experimental soil samples were obtained during each freeze-thaw stage. By performing correlation analysis between the spectra of the multiple experimental polarization components and the measured unfrozen water content, the dominant polarization component of each freeze-thaw stage that responds most significantly to the change in unfrozen water content and the corresponding preferred band were selected to generate the stage input feature subset corresponding to each freeze-thaw stage. Based on the stage input feature subset corresponding to the target freeze-thaw stage, the polarization component data of the dominant polarization component at the preferred band is obtained from the polarization component spectrum set to generate the target input feature combination of the soil sample to be tested at the target temperature. Based on a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage, the predicted unfrozen water content of the soil sample to be tested at the target temperature is generated according to the target input feature combination, wherein each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

2. The method according to claim 1, characterized in that, After generating the predicted unfrozen water content of the soil sample to be tested at the target temperature, the method further includes: Obtain the polarization component spectral set of the soil sample to be tested at different temperatures and predict the unfrozen water content; For each temperature point, the polarization component spectrum of the dominant polarization component corresponding to the current temperature point is obtained from the set of polarization component spectra corresponding to the current temperature point. Based on the polarization component spectrum of the dominant polarization component corresponding to the current temperature point and the polarization component spectrum of the dominant polarization component corresponding to the next adjacent temperature point, the band spectral change rate of each band is calculated respectively. The mean of all the band spectral change rates is determined as the polarization change intensity index of the current temperature point. Based on the polarization change intensity index, the physical monotonicity constraint modulation weight of the current temperature point is constructed. Based on the physical monotonicity constraint modulation weight, the physical monotonicity constraint optimization objective function of the current temperature point is constructed. The physical monotonicity constraint optimization objective function is solved to perform monotonicity correction processing on the predicted unfrozen water content of the current temperature point, and the predicted unfrozen water content after monotonicity correction is obtained. The current temperature point and the next adjacent temperature point are obtained by sorting all temperature points according to the temperature decrease rule or the temperature increase rule.

3. The method according to claim 2, characterized in that, After obtaining the predicted unfrozen water content after monotonicity correction, the method further includes: Based on the pre-trained residual prediction model, residual estimation is performed according to the polarization change intensity index of the target temperature point and the label of the target freeze-thaw stage to obtain the corresponding residual estimation compensation amount. The predicted unfrozen water content after monotonicity correction is summed with the residual estimate compensation to obtain the final predicted unfrozen water content of the soil sample under test at the target temperature.

4. The method according to claim 1, characterized in that, Before determining the target freeze-thaw stage of the soil sample to be tested based on the target temperature and preset freeze-thaw stage division rules, the method further includes: Freeze-thaw cycle experiments were conducted on multiple experimental soil samples, and experimental multi-angle polarization spectra of each experimental soil sample were collected during the complete freeze-thaw process. The initial water content of each experimental soil sample was different, and the complete freeze-thaw process included multiple experimental temperature points. The experimental multi-angle polarization spectra included the experimental multi-angle polarization spectra of the experimental soil samples at each experimental temperature point. The multi-angle polarization spectra of each experiment are decomposed to obtain multiple sets of experimental polarization component spectra, and a mapping relationship is established between the sets of polarization component spectra and the corresponding experimental temperature points and initial water content. For each polarization component, the spectrum of all polarization components of the current polarization component is obtained. The spectrum of all polarization components is sorted according to the temperature decrease rule or the temperature increase rule to obtain the polarization component spectrum sequence of the current polarization component during the complete freeze-thaw process. For each pair of adjacent experimental temperature points, based on the polarization component spectral sequence, the experimental band spectral change rate of the current polarization component between the current adjacent experimental temperature points is calculated, and the average value of all experimental band spectral change rates is determined as the experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points. Obtain the experimental polarization change intensity index of all polarization components between the current adjacent experimental temperature points, determine the maximum experimental polarization change intensity index or the weighted sum of all experimental polarization change intensity indices as the experimental polarization change intensity index between the current adjacent experimental temperature points, and integrate the experimental polarization change intensity indices between all adjacent experimental temperature points to obtain the experimental polarization change intensity index sequence of the complete freeze-thaw process. According to a preset screening method, continuous temperature ranges with polarization change intensity higher than the background fluctuation level are selected from the experimental polarization change intensity index sequence, and the continuous temperature ranges are determined as phase transition stages. Continuous temperature ranges located on the low-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as stable freezing stages. Continuous temperature ranges located on the high-temperature side of the phase transition stage and with polarization change intensity lower than the background fluctuation level are determined as melting and stabilizing stages. The preset screening method is any one of the statistical threshold method, peak ratio method, and extreme value identification method. The continuous temperature range contains at least two pairs of adjacent experimental temperature points. Based on the temperature ranges corresponding to the stable freezing stage, the phase transition stage, and the thawing stabilization stage, a preset freeze-thaw stage division rule is constructed.

5. The method according to claim 1, characterized in that, The process involves acquiring the experimental polarization component spectra of multiple experimental soil samples during various freeze-thaw stages, and performing correlation analysis between these spectra and the measured unfrozen water content to identify the dominant polarization component and its corresponding preferred band for each freeze-thaw stage that exhibits the most significant response to changes in unfrozen water content. This process generates a subset of stage input features for each freeze-thaw stage, including: Based on the experimental temperature points corresponding to each experimental polarization component spectral set and the preset freeze-thaw stage division rules, the freeze-thaw stage corresponding to each experimental polarization component spectral set is determined respectively. Within each freeze-thaw stage, for each polarization component, based on the experimental polarization component spectrum of the current polarization component, the correlation coefficient between the experimental polarization component data of the current polarization component at each band and the measured unfrozen water content is calculated, and the correlation distribution data of the current polarization component within the current freeze-thaw stage is generated. From the correlation distribution data, bands with absolute values ​​of correlation coefficients higher than a preset correlation coefficient threshold are selected to obtain a set of candidate sensitive bands for the current polarization component in the current freeze-thaw stage. The stability of each candidate sensitive band in the candidate sensitive band set is evaluated to screen out candidate sensitive bands whose individual modeling prediction error exceeds a preset prediction error threshold or whose correlation dispersion coefficient exceeds a preset correlation dispersion threshold, so as to obtain the preferred band of the current polarization component in the current freeze-thaw stage. The polarization components are comprehensively scored from the dimensions of the number of preferred bands, the correlation coefficient of preferred bands, and the stability evaluation results of preferred bands. The polarization component with the highest comprehensive score is determined as the dominant polarization component of the current freeze-thaw stage. Based on the dominant polarization component and the corresponding preferred band, a subset of stage input features corresponding to the current freeze-thaw stage is generated.

6. The method according to claim 1, characterized in that, The process involves collecting multi-angle polarization spectra of the soil sample under test at the target temperature, and decomposing the multi-angle polarization spectra to obtain a set of polarization component spectra of the soil sample under test at the target temperature, including: Using a polarization spectrometer, multi-angle polarization spectra of the soil sample to be tested are obtained from multiple preset polarization directions, wherein each preset polarization direction corresponds to a polarization spectrum, and the multi-angle polarization spectrum contains multiple polarization spectra; By synthesizing the polarization spectra corresponding to all the preset polarization directions, the total reflectance component spectrum of the soil sample under test at the target temperature is obtained; Two mutually orthogonal preset polarization directions are selected from all the preset polarization directions, and the polarization spectra corresponding to the two preset polarization directions are determined as the parallel polarization component spectrum and the vertical polarization component spectrum of the soil sample to be tested at the target temperature. Stokes parameters are calculated based on the multi-angle polarization spectrum, and the linear polarization degree component spectrum and polarization angle component spectrum of the soil sample under test at the target temperature are calculated based on the Stokes parameters.

7. The method according to claim 1, characterized in that, Before generating the predicted unfrozen water content of the soil sample at the target temperature based on the target input feature combination, the method further includes: For each freeze-thaw stage, construct a prediction model for the initial unfrozen water content corresponding to the current freeze-thaw stage; For each experimental polarization component spectral set included in the current freeze-thaw stage, based on the stage input feature subset corresponding to the current freeze-thaw stage, the experimental polarization component data of the dominant polarization component at the preferred band is obtained from the current experimental polarization component spectral set, and the experimental polarization component data, the experimental temperature point corresponding to the current experimental polarization component spectral set, and the initial water content are combined to generate the training sample of the current freeze-thaw stage. Based on multiple training samples of the current freeze-thaw stage, the model for predicting the unfrozen water content of the initial stage is trained to obtain the stage unfrozen water content prediction model corresponding to the current freeze-thaw stage that has completed model training.

8. A system for predicting unfrozen water content based on polarization spectroscopy, characterized in that, include: A multi-angle polarization spectral acquisition module is used to acquire the multi-angle polarization spectrum of the soil sample to be tested at the target temperature, and to decompose the multi-angle polarization spectrum to obtain the set of polarization component spectra of the soil sample to be tested at the target temperature. The freeze-thaw stage determination module is used to determine the target freeze-thaw stage of the soil sample to be tested based on the target temperature and the preset freeze-thaw stage division rules. The stage input feature subset generation module is used to obtain the experimental polarization component spectra of multiple experimental soil samples in each freeze-thaw stage, and to screen out the dominant polarization component and its corresponding preferred band that responds most significantly to the change in unfrozen water content in each freeze-thaw stage by performing correlation analysis on the spectra of multiple experimental polarization components and the measured unfrozen water content, so as to generate the stage input feature subset corresponding to each freeze-thaw stage. The input feature combination generation module is used to obtain the polarization component data of the dominant polarization component at the preferred band from the polarization component spectrum set based on the stage input feature subset corresponding to the target freeze-thaw stage, so as to generate the target input feature combination of the soil sample to be tested at the target temperature. The unfrozen water content prediction module is used to generate the predicted unfrozen water content of the soil sample to be tested at the target temperature based on the target input feature combination, using a pre-trained unfrozen water content prediction model that matches the target freeze-thaw stage. Each freeze-thaw stage corresponds to a stage unfrozen water content prediction model.

9. A storage medium storing at least one executable instruction, characterized in that, The executable instructions cause the processor to perform the operations corresponding to the polarization spectroscopy-based unfrozen water content prediction method as described in any one of claims 1-7.

10. A terminal, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, characterized in that the executable instruction causes the processor to perform the operation corresponding to the polarization spectroscopy-based unfrozen water content prediction method as described in any one of claims 1-7.

Images