CN-121994717-A - Unfrozen water content prediction method and system based on polarization spectrum, medium and terminal

CN121994717ACN 121994717 ACN121994717 ACN 121994717ACN-121994717-A

Abstract

The application discloses a method, a system, a medium and a terminal for predicting unfrozen water content based on polarization spectrum, relates to the technical field of soil detection, and mainly aims to solve the problem that prediction accuracy of an unfrozen water content prediction model constructed based on reflectivity is reduced due to change of reflection characteristics of a soil surface in the formation process of an ice crystal structure. The method comprises the steps of collecting multi-angle polarization spectrums of a soil sample to be detected at a target temperature, performing decomposition treatment to obtain a polarization component spectrum set, determining a target freeze-thawing stage based on the target temperature, generating a target input characteristic combination based on a stage input characteristic subset corresponding to the target freeze-thawing stage, and generating predicted unfrozen water content according to the target input characteristic combination based on a stage unfrozen water content prediction model matched with the target freeze-thawing stage.

Inventors

Bao Nisha
PENG SIHAN
LIU BOYAN
LUO JIAYIN
ZHANG FAN
HE LIMING

Assignees

东北大学

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. A method for predicting unfrozen water content based on polarization spectrum, comprising: Collecting multi-angle polarization spectrums of a soil sample to be detected at a target temperature, and decomposing the multi-angle polarization spectrums to obtain a polarization component spectrum set of the soil sample to be detected at the target temperature; Determining a target freeze-thawing stage in which the soil sample to be tested is positioned based on the target temperature and a preset freeze-thawing stage division rule; Obtaining experimental polarization component spectrums of a plurality of experimental soil samples in each freeze thawing stage, and screening out a stage dominant polarization component with the most obvious response to the unfrozen water content change in each freeze thawing stage and a corresponding preferred wave band through correlation analysis of the experimental polarization component spectrums and the actually measured unfrozen water content so as to generate a stage input feature subset corresponding to each freeze thawing stage; Based on a stage input feature subset corresponding to the target freeze thawing stage, acquiring polarization component data of a stage dominant polarization component at a preferred wave band from the polarization component spectrum set to generate a target input feature combination of the soil sample to be detected at the target temperature; and generating the predicted unfrozen water content of the soil sample to be detected at the target temperature according to the target input characteristic combination based on a stage unfrozen water content prediction model which is matched with the target freeze thawing stage and is subjected to model training, wherein each freeze thawing stage corresponds to one stage unfrozen water content prediction model.
2. The method of claim 1, wherein after generating the predicted unfrozen water content of the soil sample under test at the target temperature, the method further comprises: acquiring a polarization component spectrum set of the soil sample to be detected at different temperatures and predicting unfrozen water content; For each temperature point, acquiring a polarization component spectrum of a phase-dominant polarization component corresponding to the current temperature point from a polarization component spectrum set corresponding to the current temperature point, calculating band spectrum change rates of each band respectively based on the polarization component spectrum of the phase-dominant polarization component corresponding to the current temperature point and the polarization component spectrum of a phase-dominant polarization component corresponding to the next adjacent temperature point, determining the average value of all band spectrum change rates as a polarization change intensity index of the current temperature point, constructing a physical monotonicity constraint modulation weight of the current temperature point based on the polarization change intensity index, constructing a 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, and performing monotonicity correction processing on the predicted unfrozen water content of the current temperature point to obtain the predicted unfrozen water content after monotonicity correction, wherein the current temperature point and the next adjacent temperature point are obtained by sequencing all temperature points according to a temperature descending rule or a temperature ascending rule.
3. The method of claim 2, wherein after said obtaining a predicted unfrozen water content that completes the monotonicity correction, the method further comprises: performing residual estimation operation according to the polarization change intensity index of the target temperature point and the label of the target freeze thawing stage based on a pre-trained residual prediction model to obtain a corresponding residual estimation compensation quantity; and adding and calculating the predicted unfrozen water content subjected to monotonicity correction and the residual error estimation compensation quantity to obtain the final predicted unfrozen water content of the soil sample to be detected at the target temperature.
4. The method of claim 1, wherein the determining the target freeze-thaw stage in which the soil sample to be tested is located is preceded by determining the target freeze-thaw stage based on the target temperature and a preset freeze-thaw stage partitioning rule, the method further comprising: Respectively carrying out freeze thawing cycle experiments on a plurality of experimental soil samples, and respectively collecting experimental multi-angle polarization spectrum sets of each experimental soil sample in a complete freeze thawing process, wherein the initial water content of each experimental soil sample is different, the complete freeze thawing process comprises a plurality of experimental temperature points, and the experimental multi-angle polarization spectrum sets comprise experimental multi-angle polarization spectrums of each experimental soil sample in each experimental temperature point; decomposing each experimental multi-angle polarization spectrum to obtain a plurality of experimental polarization component spectrum sets, and establishing a mapping relation between the polarization component spectrum sets and corresponding experimental temperature points and initial water content; for each polarized component, acquiring all polarized component spectrums of the current polarized component, and sequencing all polarized component spectrums according to a temperature drop rule or a temperature rise rule to obtain a polarized component spectrum sequence of the current polarized component in a complete freezing and thawing process; For each pair of adjacent experimental temperature points, calculating the experimental band spectrum change rate of each wavelength of the current polarization component between the current adjacent experimental temperature points based on the polarization component spectrum sequence, and determining the average value of all the experimental band spectrum change rates as an experimental polarization change intensity index of the current polarization component between the current adjacent experimental temperature points; Acquiring experimental polarization change intensity indexes of all the polarization components between the current adjacent experimental temperature points, determining the maximum experimental polarization change intensity index or the weighted sum of all the experimental polarization change intensity indexes in all the experimental polarization change intensity indexes as the experimental polarization change intensity indexes between the current adjacent experimental temperature points, and integrating the experimental polarization change intensity indexes between all the adjacent experimental temperature points to obtain an experimental polarization change intensity index sequence in the complete freeze thawing process; According to a preset screening method, screening a continuous temperature interval with the polarization change intensity higher than the background fluctuation level from the experimental polarization change intensity index sequence, determining the continuous temperature interval as a phase change transition stage, determining a continuous temperature interval which is positioned at the low temperature side of the phase change transition stage and has the polarization change intensity lower than the background fluctuation level as a stable freezing stage, and determining a continuous temperature interval which is positioned at the high temperature side of the phase change transition stage and has the polarization change intensity lower than the background fluctuation level as a melting stable stage, wherein the preset screening method is any one of a statistical threshold method, a peak value proportion method and an extremum recognition method, and the continuous temperature interval at least comprises two pairs of adjacent experimental temperature points; And constructing a preset freezing and thawing stage division rule according to the stable freezing stage, the phase change transition stage and the temperature interval corresponding to the thawing stable stage.
5. The method of claim 1, wherein obtaining experimental polarization component spectra of a plurality of experimental soil samples in each freeze-thawing stage, and screening out a stage dominant polarization component and a corresponding preferred band, which have the most significant response to changes in unfrozen water content in each freeze-thawing stage, by performing correlation analysis on a plurality of experimental polarization component spectra and measured unfrozen water content, so as to generate a stage input feature subset corresponding to each freeze-thawing stage, wherein the method comprises: According to experimental temperature points corresponding to each experimental polarization component spectrum set and a preset freezing and thawing stage division rule, respectively determining freezing and thawing stages corresponding to each experimental polarization component spectrum set; In each freeze thawing stage, calculating a correlation coefficient between experimental polarization component data of the current polarization component at each wave band and the actually measured unfrozen water content based on an experimental polarization component spectrum of the current polarization component for each polarization component, and generating correlation distribution data of the current polarization component in the current freeze thawing stage; Screening out wave bands with absolute values of correlation coefficients higher than a preset correlation coefficient threshold value from the correlation distribution data to obtain a candidate sensitive wave band set of the current polarization component in the current freeze thawing stage; Performing stability evaluation on each candidate sensitive wave band contained in the candidate sensitive wave band set to screen out candidate sensitive wave bands of which the independent modeling prediction errors exceed a preset prediction error threshold value or the correlation discrete coefficients exceed a preset correlation discrete coefficient threshold value, so as to obtain a preferred wave band of the current polarization component in the current freeze thawing stage; And comprehensively scoring each polarization component from the preferred band quantity dimension, the preferred band correlation coefficient dimension and the preferred band stability evaluation result dimension, determining the polarization component with the highest comprehensive score as a phase dominant polarization component of the current freeze thawing phase, and generating a phase input feature subset corresponding to the current freeze thawing phase based on the phase dominant polarization component and the corresponding preferred band.
6. The method according to claim 1, wherein the collecting the multi-angle polarization spectrum of the soil sample to be measured at the target temperature and decomposing the multi-angle polarization spectrum to obtain the set of polarization component spectrums of the soil sample to be measured at the target temperature comprises: Obtaining multi-angle polarization spectrums of the soil sample to be detected from a plurality of preset polarization directions by using a polarization spectrometer, wherein each preset polarization direction corresponds to one polarization spectrum, and the multi-angle polarization spectrums comprise a plurality of polarization spectrums; synthesizing polarization spectrums corresponding to all the preset polarization directions to obtain a total reflection component spectrum of the soil sample to be detected at the target temperature; Selecting two mutually orthogonal preset polarization directions from all the preset polarization directions, and determining polarization spectrums corresponding to the two preset polarization directions as a parallel polarization component spectrum and a perpendicular polarization component spectrum of the soil sample to be detected at the target temperature; And calculating a Stokes parameter based on the multi-angle polarization spectrum, and calculating a linear polarization degree component spectrum and a polarization angle component spectrum of the soil sample to be measured at the target temperature based on the Stokes parameter.
7. The method of claim 1, wherein the generating the predicted unfrozen water content of the soil sample to be tested at the target temperature is preceded by generating a model of the predicted unfrozen water content of the soil sample to be tested at the target temperature based on a model-trained stage unfrozen water content prediction model that matches the target freeze-thaw stage, the method further comprising: aiming at each freeze-thawing stage, constructing an unfrozen water content prediction model of an initial stage corresponding to the current freeze-thawing stage; Aiming at each experimental polarization component spectrum set contained in the current freeze-thawing stage, acquiring experimental polarization component data of a stage dominant polarization component at a preferred wave band from the current experimental polarization component spectrum set based on a stage input feature subset corresponding to the current freeze-thawing stage, and combining the experimental polarization component data, an experimental temperature point corresponding to the current experimental polarization component spectrum set and an initial water content to generate a training sample of the current freeze-thawing stage; And carrying out model training on the unfrozen water content prediction model in the initial stage based on a plurality of training samples in the current freeze-thawing stage to obtain a stage unfrozen water content prediction model which corresponds to the current freeze-thawing stage and is subjected to model training.
8. A polarization spectrum-based unfrozen water content prediction system, comprising: The multi-angle polarization spectrum acquisition module is used for acquiring multi-angle polarization spectrums of a soil sample to be detected at a target temperature, and decomposing the multi-angle polarization spectrums to obtain a polarization component spectrum set of the soil sample to be detected at the target temperature; The freeze-thawing stage determining module is used for determining a target freeze-thawing stage of the soil sample to be detected based on the target temperature and a preset freeze-thawing stage dividing rule; The stage input feature subset generating module is used for acquiring experimental polarization component spectrums of a plurality of experimental soil samples in each freeze thawing stage, and screening out a stage dominant polarization component with the most obvious response to the unfrozen water content change in each freeze thawing stage and a corresponding preferred wave band through correlation analysis of the experimental polarization component spectrums and the actually measured unfrozen water content so as to generate a stage input feature subset corresponding to each freeze thawing stage; the input feature combination generation module is used for acquiring polarization component data of a phase dominant polarization component at a preferred wave band from the polarization component spectrum set based on a phase input feature subset corresponding to the target freeze thawing phase so as to generate a target input feature combination of the soil sample to be detected at the target temperature; the unfrozen water content prediction module is used for generating the predicted unfrozen water content of the soil sample to be detected at the target temperature according to the target input characteristic combination based on a stage unfrozen water content prediction model which is matched with the target freezing and thawing stage and is subjected to model training, wherein each freezing and thawing stage corresponds to one stage unfrozen water content prediction model.
9. A storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the polarization spectrum based unfrozen water content prediction method of any one of claims 1-7.
10. A terminal comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus; the memory is configured to store at least one executable instruction, wherein the executable instruction causes the processor to perform operations corresponding to the method for predicting unfrozen water content based on polarization spectrum of any one of claims 1 to 7.

Description

Unfrozen water content prediction method and system based on polarization spectrum, medium and terminal Technical Field The application relates to the technical field of soil detection, in particular to a method and a system for predicting unfrozen water content based on polarization spectrum, a medium and a terminal. Background The frozen soil is widely distributed in high latitude and high altitude areas of China, wherein the seasonal frozen soil occupies more than half of the territorial area. The soil freeze-thawing cycle is one of the most common physical processes in cold regions, and the essence of the soil freeze-thawing cycle is that soil pore water periodically changes phase between liquid state and solid state under the drive of temperature change. During the freeze-thaw cycle, not all of the moisture in the soil pores freezes into ice, but some remains in liquid form on the surface of the mineral particles and in the fine pores, and this part of the moisture is referred to as unfrozen water. Unfrozen water content is a key physical parameter for representing the hydrothermal process of frozen soil, and directly influences the heat conducting property, the water migration rule, the solute transport characteristics and the frost heaving and thawing and sinking behaviors of the soil. Therefore, the accurate prediction of the unfrozen water content has important significance for ecological restoration in frozen soil areas, agricultural production and engineering stability evaluation. Although the traditional method for measuring the unfrozen water content such as a nuclear magnetic resonance method, a thermal analysis method and the like has higher precision, the problems of complex equipment, long test period, difficulty in realizing in-situ rapid monitoring and the like exist. In recent years, the visible light-near infrared spectrum technology is widely applied to soil moisture inversion research due to the advantages of no damage, rapidness and the like, and mainly utilizes an empirical model between reflectance spectrum construction and unfrozen water content to predict the unfrozen water content. However, since the formation of ice crystal structure during the soil freezing process can significantly change the surface reflection characteristics, so that the reflected signal contains components from specular reflection, and the empirical model constructed based on reflectivity is difficult to effectively eliminate the interference caused by such specular reflection, noise irrelevant to the unfrozen water content is mixed in the model, and finally the precision of the unfrozen water content prediction is reduced. Disclosure of Invention In view of the above, the application provides a method and a system for predicting unfrozen water content based on polarization spectrum, a medium and a terminal, and aims to solve the problem that the prediction precision of an unfrozen water content prediction model constructed based on reflectivity is reduced due to the change of the reflection characteristic of the soil surface in the formation process of an ice crystal structure. According to one aspect of the present application, there is provided a method for predicting unfrozen water content based on polarization spectrum, comprising: Collecting multi-angle polarization spectrums of a soil sample to be detected at a target temperature, and decomposing the multi-angle polarization spectrums to obtain a polarization component spectrum set of the soil sample to be detected at the target temperature; Determining a target freeze-thawing stage in which the soil sample to be tested is positioned based on the target temperature and a preset freeze-thawing stage division rule; Obtaining experimental polarization component spectrums of a plurality of experimental soil samples in each freeze thawing stage, and screening out a stage dominant polarization component with the most obvious response to the unfrozen water content change in each freeze thawing stage and a corresponding preferred wave band through correlation analysis of the experimental polarization component spectrums and the actually measured unfrozen water content so as to generate a stage input feature subset corresponding to each freeze thawing stage; Based on a stage input feature subset corresponding to the target freeze thawing stage, acquiring polarization component data of a stage dominant polarization component at a preferred wave band from the polarization component spectrum set to generate a target input feature combination of the soil sample to be detected at the target temperature; and generating the predicted unfrozen water content of the soil sample to be detected at the target temperature according to the target input characteristic combination based on a stage unfrozen water content prediction model which is matched with the target freeze thawing stage and is subjected to model training, wherein each freeze thawing stage corresponds to o