Nuclear magnetic resonance and image intelligent dual-mode dynamic recognition method and system for unfrozen water

Abstract

The application provides a nuclear magnetic resonance and image intelligent dual-mode unfrozen water dynamic identification method and system, and belongs to the technical field of frozen soil engineering and geophysical exploration, and comprises the following steps: sealing a frozen soil sample in a nuclear magnetic sample tube, and realizing gradient cooling of the sample in a preset low temperature range through temperature control; scanning and acquiring a transverse relaxation time T2 spectrum and a nuclear magnetic image of the frozen soil sample at different temperature nodes by using a nuclear magnetic resonance technology; automatically extracting a temperature-signal intensity curve in the nuclear magnetic image by using an image recognition technology after image preprocessing of the nuclear magnetic image; performing integral calculation on liquid water signals in the T2 spectrum, and dynamically calculating the content of unfrozen water at each temperature based on a nuclear magnetic signal intensity formula; and constructing a pore-fracture collaborative evolution matrix, so that accurate analysis of the dynamic unfrozen water and pore evolution in a freezing and thawing process is realized, the water-heat coupling characteristics of Qinghai-Tibet soil under a complex freezing and thawing environment are analyzed, and reliable technical support is provided for stability evaluation of frozen soil engineering.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of frozen soil engineering and geophysical exploration, and particularly relates to a nuclear magnetic resonance and image intelligent dual-mode frozen water dynamic identification method and system. BACKGROUND

[0002] As the largest frozen soil distribution area in the global mid-low latitude, the stability of frozen soil engineering in the Qinghai-Tibet Plateau directly affects the safe operation of major infrastructure such as railways, highways and energy pipelines. In the freezing and thawing cycle of the unique high-salinity and multi-fractured Qinghai-Tibet soil in this region, the dynamic change of unfrozen water content significantly affects the formation mechanism of frost heaving deformation and thawing settlement disasters through the effects of latent heat of phase change and osmotic pressure. Traditional detection methods have significant limitations: differential scanning calorimetry (DSC) requires destruction of the sample structure and cannot achieve in-situ monitoring; time domain reflectometry (TDR) is limited by the single-point measurement mode and cannot capture the three-dimensional water distribution characteristics of the soil; and the resistivity method has a detection error rate as high as 15%-20% due to high-salinity ion interference. Although nuclear magnetic resonance (NMR) technology has gradually been applied in this field due to its non-destructive detection and high-resolution advantages, there are still key bottlenecks: the temperature-signal relationship curve in the nuclear magnetic image relies on manual extraction, the data processing efficiency is low and the subjective error is significant, and the signal-to-noise ratio is poor; existing methods cannot fuse T2 spectrum analysis and image feature recognition technology, and cannot quantify the coordinated evolution law of the pore-fracture system during the freezing and thawing process. In addition, the current image processing technology has insufficient analytical accuracy for complex curves and coordinate scales, which seriously restricts the construction accuracy of the temperature-unfrozen water correlation model. SUMMARY

[0003] The purpose of the present application is to provide a multi-fractured frozen soil unfrozen water dynamic identification and detection method to solve the core problems of low data analysis efficiency, lack of pore evolution quantification, and insufficient adaptability to complex environments in multi-fractured frozen soil unfrozen water monitoring.

[0004] To achieve the above purpose, the present application proposes a nuclear magnetic resonance (LF-NMR) and image intelligent dual-mode driven unfrozen water dynamic precise identification method, comprising the following steps:

[0005] S1: sealing the frozen soil sample in a nuclear magnetic sample tube and achieving gradient cooling of the sample in a preset low temperature range through temperature control;

[0006] S2: scanning and obtaining the transverse relaxation time T2 spectrum and nuclear magnetic image of the frozen soil sample at different temperature nodes using nuclear magnetic resonance technology;

[0007] S3: after image preprocessing of the nuclear magnetic image, automatically extracting the temperature-nuclear magnetic signal intensity curve in the nuclear magnetic image using image recognition technology;

[0008] S4: Integrate the liquid water signal in the T2 spectrum to calculate the content of unfrozen water at each temperature based on the following nuclear magnetic signal intensity formula:

[0009]

[0010] In the formula, M T is the nuclear magnetic signal intensity at temperature T, obtained by integrating the T2 spectrum peak area; M T0 is the nuclear magnetic signal intensity corresponding to the initial liquid water; w0 is the initial total water content.

[0011] Further, steps S5 and S6 are further included:

[0012] S5: Use the transverse relaxation time T2 spectrum to inverse the pore radius, extract the pore distribution characteristics, calculate the pore volume proportion of each pore size interval, and establish the linear relationship between the nuclear magnetic image gray value (0-255) and the pore radius;

[0013] S6: Construct a pore-fracture co-evolution matrix, define the evolution matrix M to describe the coupling relationship between pore structure and unfrozen water and temperature during the freezing and thawing process, and realize the accurate analysis of the dynamic of unfrozen water and the evolution of pores during the freezing and thawing process:

[0014]

[0015] Where, ΔR 微孔隙 is the radius change amount of micropores with a pore size less than 0.2 μm in the freezing and thawing process due to the stress action of ice crystal growth, which quantifies the mesoscopic frost heaving characteristics; ΔR 裂隙 represents the width change amount of fractures with a width greater than 1 μm in the freezing and thawing process due to the action of ice wedge, which represents the macroscopic expansion rate; is the unfrozen water gradient, is the temperature gradient.

[0016] Further, step S5 is specifically:

[0017] S5.1: Use the linear relationship between the transverse relaxation time T2 spectrum and the pore radius R to inverse the pore radius:

[0018] R = 3ρ2·T2;

[0019] Wherein, the surface relaxation rate ρ2 takes the empirical value of 10 nm / ms, and the T2 distribution curve is converted into the pore radius distribution;

[0020] S5.2: Pore classification and quantification:

[0021] According to the pore size, the pores are divided into ultra-micropores (<0.01 μm), micropores (0.01 μm~0.2 μm), small pores (0.2 μm~1 μm), medium pores (1 μm~3 μm) and large pores (>3 μm), and the pore volume proportion of each pore size interval is calculated by integrating the T2 spectrum peak area:

[0022]

[0023] Wherein, V s is the pore volume of a pore size interval, V a is the total pore volume;

[0024] When the micropore volume proportion is more than 60%, the micropore radius change ΔR 微孔隙 is the main evolution, reflecting the meso-frost heave mechanism; when the large pore volume proportion is more than 10%, the crack width change ΔR 裂隙 is the main variable, corresponding to the macroscopic failure process;

[0025] S5.3: The pore radius distribution of the soil body is measured by the mercury injection method, the gray value (0-255) of the pore region of the nuclear magnetic image of the same soil sample is extracted, the gray value-pore radius mapping relationship of the nuclear magnetic image is established, and the linear relationship (linear regression model) between the gray value of the nuclear magnetic image and the pore radius is constructed.

[0026] The gray value G=k·R+b;

[0027] The parameters k and b are calibrated by the mercury injection method (MIP), the quantitative conversion of the gray value of the nuclear magnetic image to the pore radius is realized, and the pore-crack co-evolution matrix is constructed.

[0028] Further, in step S1, the preset low temperature range is -30℃ to 20℃;

[0029] In step S2, the 0.5T permanent magnet low-field nuclear magnetic resonance instrument is used to perform nuclear magnetic scanning on the frozen soil sample, the nuclear magnetic scanning is triggered once every 5℃, so as to obtain the transverse relaxation time T2 spectrum and the nuclear magnetic image of the frozen soil sample at 20℃, 10℃, 0℃, -2℃, -5℃, -7℃, -10℃, -15℃ and -30℃, and the unfrozen water phase change characteristics in the whole freezing and thawing process are comprehensively captured.

[0030] In the nuclear magnetic data acquisition process, the CPMG sequence is used, the echo time TE=0.2ms, the number of accumulations is 32 times, and the system can ensure that the signal-to-noise ratio is greater than or equal to 90dB in the temperature range of -30℃ to 20℃.

[0031] Further, in step S3, the image preprocessing process includes:

[0032] Gray processing: convert the nuclear magnetic image from the RGB color space to the gray image;

[0033] Binaryzation: using adaptive threshold method, threshold set to 120, converting gray image to binary image;

[0034] Median filter processing: using 3*3 median filter to remove the image of the shot noise, to improve the image quality.

[0035] Further, in step S3, the image recognition technology comprises:

[0036] Using improved Hough transform to identify temperature axis and signal intensity axis, combining Canny edge detection algorithm and least square fitting, realizing coordinate axis tilt correction (error ≤0.5°), false detection rate <5%;

[0037] Integrating Tesseract OCR engine and regular expression (supporting scientific notation, fraction and exponential format) to extract scale value and parse, and automatically calibrating scale offset error, accurately converting scale information in the image into actual physical quantity value, scale recognition rate ≥95%;The OCR analysis supports scientific notation and fraction format, and the value is matched by regular expression;

[0038] Based on HSV color space segmentation (red curve threshold: (H=0-10, S≥120, V≥70) and morphological closing operation, realizing pixel-level extraction of temperature-nuclear magnetic signal intensity curve.

[0039] Using multi-thread parallel architecture, single frame image processing time <0.5 seconds, supporting 10 frames per second of continuous monitoring, meeting the real-time demand of engineering.

[0040] Further, the parameter setting of the improved Hough transform is: ρ=1, θ=π / 180, threshold=100, minimum line segment length=100 pixels;

[0041] Based on HSV color space segmentation and morphological closing operation, the specific method for realizing pixel-level extraction of temperature-nuclear magnetic signal intensity curve is: first, color segmentation is carried out on the red curve through HSV color space, and the HSV threshold range is set as lower=[0, 120, 70] and upper=[10, 255, 255]; second, the curve contour is extracted by using OpenCV library, and noise points with an area less than 100 pixels are filtered out; finally, the pixel coordinates are converted into actual values, and the linear interpolation method is used to calculate the temperature and nuclear magnetic signal intensity values corresponding to the curve points. The extraction method effectively establishes the relationship between temperature and nuclear magnetic signal intensity, and provides a basis for subsequent data analysis and processing.

[0042] Further, the method of the present application has the implementation of closed loop, which specifically shows that:

[0043] Data flow: from the start of the nuclear magnetic signal acquisition, through the imaging processing, feature extraction, model calculation, and finally providing the basis for engineering decision-making, engineering practice feedback further optimizes the entire technical process.

[0044] Verification closed loop: first, parameter calibration in the laboratory, then in the field to verify the actual environment, according to the verification result, parameter iteration, and finally realize large-scale application, form a complete technology implementation closed loop, and continuously improve the reliability and practicality of the technology.

[0045] The application also provides a nuclear magnetic resonance and image intelligent dual-mode unfrozen water dynamic identification system for realizing the identification method, the system comprising:

[0046] Multi-modal data acquisition subsystem: including a nuclear magnetic acquisition module and a temperature control module, for obtaining the transverse relaxation time T2 spectrum and nuclear magnetic image of the frozen soil sample at different temperature nodes;

[0047] Intelligent image processing unit: equipped with OpenCV and TesseractOCR library, for detecting the coordinate axis of the nuclear magnetic image, extracting the curve fitting, scale analysis and data conversion;

[0048] Data analysis module: based on the Python platform, with built-in unfrozen water content calculation algorithm (unfrozen water content dynamic calculation model) and pore evolution model (pore-fracture collaborative evolution model).

[0049] Further, the data analysis module includes a Qinghai-Tibet soil special analysis model and a freeze-thaw coupling analysis module:

[0050] The Qinghai-Tibet soil special analysis model includes an unfrozen water content dynamic calculation model and a pore-fracture collaborative evolution model.

[0051] Wherein, the pore-fracture collaborative evolution model introduces a pore characteristic factor F s (micropore <0.2 μm, fracture >1 μm):

[0052]

[0053] In the formula, V 微孔 is the micropore volume with a pore diameter less than 0.2 μm, which is calculated by integrating the peak area in the T2 spectrum corresponding to the relaxation time interval, reflecting the mesoscopic frost heave mechanism. V 裂隙 is the fracture volume with a width greater than 1 μm, which is determined by mapping the nuclear magnetic image gray scale and the pore radius calibrated by the mercury injection method, representing the macroscopic fracture expansion.

[0054] And the surface relaxation rate ρ2=10 nm / ms is calibrated, and a multi-scale evolution equation set is established:

[0055]

[0056] wherein the evolution coefficients k1, k2 are measured by mercury intrusion method on the mercury intrusion radius of the micro-pores and fissures and fitted.

[0057] In the temperature interval ΔT, the gradient is approximately constant, and the time integration gives the state change

[0058]

[0059] After normalizing the time variable (Δt = 1), the pore-fissure co-evolution model has the following relationship with the pore-fissure co-evolution matrix:

[0060]

[0061] For the calibration process, first, data acquisition is performed, the pore radius is measured by mercury intrusion method (MIP), the micro-pore volume V 微孔 is calculated using the T2 spectrum measured by nuclear magnetic resonance, and the fissure volume V 裂隙 is obtained, and F s is obtained; then parameter fitting is performed, the data is substituted into the evolution equation, and k1, k2 are adjusted so that |R MIP -R 模型 |<=0.05 μm; finally, verification is performed, in the freeze-thaw experiment, the fissure expansion calculated by the model is ensured to be within a deviation of ≤0.05 μm from the scanning.

[0062] The freeze-thaw coupling analysis module is used to construct a pore-fissure co-evolution matrix, and through eigenvalue analysis of the matrix, a frost-heave risk level (low / medium / high risk) is divided, and the prediction error is better than that of a traditional empirical model. In linkage with a cloud platform, a frost-heave risk thermal map is generated, supporting real-time early warning and decision optimization for projects such as the Qinghai-Tibet Railway and cold region tunnels.

[0063] Further, in the multi-modal data acquisition subsystem,

[0064] The nuclear magnetic acquisition module: a low-field nuclear magnetic resonance instrument configured with a 0.5T permanent magnet is adopted, the minimum echo time TEmin = 0.1ms is supported, CPMG sequence acquisition is supported, the low-field nuclear magnetic resonance instrument is equipped with an anti-interference double-layer shielding probe coil, the magnetic field uniformity is ≤10ppm, the signal-to-noise ratio is ≥90dB through dynamic shimming technology and frequency adaptive filtering algorithm, the reliability of the signal under the condition of extremely low temperature (-30℃) and complex pore structure is ensured, and the low-field nuclear magnetic resonance instrument is suitable for frozen soil detection;

[0065] Temperature control module: integrated liquid nitrogen circulation pipeline and PID temperature controller, the working temperature range is-30 DEG C to 20 DEG C, the temperature control precision is ± 0.01 DEG C, the temperature closed loop control is realized through the PID temperature controller, the temperature sensor adopted is PT100 platinum resistance, and the temperature measurement precision is ± 0.01 DEG C;Through gradient cooling design (rate 1 DEG C / min) and 9 temperature node scanning strategy, the whole process of soil sample freeze-thaw is comprehensively covered.

[0066] The intelligent image processing unit supports multi-thread parallel computing, and the intelligent image processing unit comprises an adaptive color segmentation module, supports red and blue double-color curve identification, and the HSV threshold range is as follows:

[0067] Red: lower=[0, 120, 70], upper=[10, 255, 255],

[0068] Blue: lower=[100, 120, 70], upper=[130, 255, 255];

[0069] The data analysis module has a built-in data verification function, which automatically corrects calculation errors by comparing experimental data and theoretical models.

[0070] Further, the data analysis module integrates a cubic spline interpolation algorithm, the interpolation error is less than or equal to 2%, and supports pore distribution visualization output and frost heave risk level division.

[0071] The application also provides a computer readable storage medium, which stores program codes for executing the above-mentioned identification method, and the codes comprise an image preprocessing unit, a Hough transform detection unit, an OCR analysis unit and a non-frozen water content calculation unit, wherein the image preprocessing unit supports batch processing and can process multiple nuclear magnetic images simultaneously, and the processing speed is significantly improved; the non-frozen water content calculation unit supports multi-core parallel computing, and the calculation speed is significantly improved compared with single core.

[0072] The technical upgrading direction of the application includes multi-physical field coupling, edge computing and 5G transmission and artificial intelligence optimization. By integrating temperature, seepage and stress data, a full-dimensional model of frost heaving mechanism is constructed. Through millisecond-level data real-time analysis, rapid decision-making in engineering site is realized. Deep learning algorithms (such as CNN) are introduced to optimize curve segmentation and feature extraction, and the contour integrity is significantly improved.

[0073] Compared with the prior art, the application has the following advantages:

[0074] 1. The present application first combines low-field nuclear magnetic resonance (LF-NMR) with image intelligent processing technology in depth, builds a temperature-nuclear magnetic signal dynamic correlation model, breaks through the limitations of traditional single-mode detection, realizes the synchronous monitoring of unfrozen water content and pore evolution through the dual driving of nuclear magnetic signal dynamic analysis and image feature extraction, significantly improves the data dimension and analysis accuracy, fills the technical gap of multi-modal non-destructive monitoring, avoids the problems of strong destructiveness, single-point measurement limitation and high salt interference of traditional methods (such as DSC, TDR), and solves the core problems of low data analysis efficiency, lack of pore evolution quantization and insufficient adaptability in complex environment in the monitoring of unfrozen water in multi-fractured frozen soil, and solves the technical bottleneck of dynamic monitoring of frozen water and pore structure analysis, with high precision, real-time and engineering applicability, providing a revolutionary tool for stability evaluation of frozen soil engineering, disaster warning and optimization of artificial freezing technology.

[0075] 2. Through the unfrozen water dynamic identification method of the present application, the test accuracy is effectively improved, the unfrozen water content detection error is significantly reduced, the pore radius inversion precision error is ≤0.05 μm, the resolution can reach 0.1 μm, which meets the microstructure analysis demand, at the same time, the unfrozen water dynamic identification system of the present application can stably operate in low temperature environment, the temperature change and data fluctuation are kept in a small range, and has good environmental adaptability; through anti-interference design, the stability of nuclear magnetic signal of the system under complex conditions is ensured, the signal-to-noise ratio is ≥90 dB, and the salt ion interference suppression rate is >90%.

[0076] 3. The present application realizes non-destructive and accurate monitoring of the unfrozen water content of multi-fractured Qinghai-Tibet soil by integrating nuclear magnetic signal dynamic analysis, intelligent image processing and freeze-thaw coupling modeling, providing technical support for engineering safety in alpine regions, especially suitable for stability evaluation of frozen soil engineering, frost heaving disaster warning and quality control of artificial freezing construction in alpine regions.

[0077] 4. The present application has wide application scene expandability, can be applied to polar scientific research station foundation monitoring, frozen soil area ecological restoration and underground energy storage engineering, first expands the nuclear magnetic-image dual-mode technology to the field of frozen soil thermal stability evaluation and clean energy optimization, which is adapted to-30℃ low temperature environment, can monitor the unfrozen water content and pore structure change of foundation in real time, and ensures the safety of structure under extreme conditions. Through water-pore collaborative analysis, the vegetation restoration strategy is optimized, the ecological restoration efficiency is improved, the frozen soil thermal stability is evaluated, the ground source heat pump system design is optimized, and the efficient utilization of clean energy is facilitated.

[0078] 5、The application realizes the coordinate axis tilt correction (error ≤0.5°) by the improved Hough transform (parameters: ρ=1, θ=π / 180) combined with Canny edge detection and least square fitting, solves the subjective error problem of traditional manual extraction; through intelligent OCR analysis, the data conversion efficiency is significantly improved; and through the HSV color space segmentation technology and morphological closing operation, the pixel-level accurate extraction of the temperature-nuclear magnetic signal intensity curve is realized, and the noise points with an area of <100 pixels are effectively filtered.

[0079] 6、The application establishes a temperature-nuclear magnetic signal intensity dynamic correlation model of temperature and nuclear magnetic signal intensity dynamic mapping by gradient cooling and multi-temperature node nuclear magnetic scanning, successfully realizes the frozen water phase change feature capture of the whole freeze-thaw process. By combining the nuclear magnetic image gray value and the mercury injection method verification data to form a pore-fracture co-evolution model, an evolution matrix is constructed to realize the quantitative frost heaving risk grade. BRIEF DESCRIPTION OF DRAWINGS

[0080] Figure 1 The nuclear magnetic resonance and image intelligent dual-mode frozen water dynamic identification method flowchart for the embodiment 1 of the application is shown. DETAILED DESCRIPTION

[0081] In order to make the purpose, technical scheme and advantages of the application clearer, the technical scheme of the application will be further described below.

[0082] Taking the typical frozen soil area of the Qinghai-Tibet Plateau as an example, the embodiment 1 proposes a nuclear magnetic resonance and image intelligent dual-mode frozen water dynamic identification method. As shown in Figure 1 The method comprises the following steps:

[0083] 1、Sample preparation and pretreatment:

[0084] The undisturbed soil sample is collected from the typical frozen soil area of the Qinghai-Tibet Plateau by the hollow drill method, a cylindrical sample with a diameter of 39.1mm and a height of 80mm is obtained, and then the pretreatment of screening and centrifugation is cooperated to remove gravel or impurity particles with a diameter of >2mm, and the pretreated soil sample is sealed in a nuclear magnetic sample tube.

[0085] In this embodiment, the hollow drill method is used to collect the undisturbed soil samples from the typical permafrost region of the Qinghai-Tibet Plateau, which can effectively ensure that the internal structure of the soil is not damaged. The cylindrical sample size obtained is 39.1 mm in diameter and 80 mm in height, which can maximize the preservation of the characteristics of the soil sample in its natural state while facilitating experimental operation. In addition, the gravel segments with a diameter greater than 2 mm are removed, and the screening and centrifugation operations are coordinated. By using a 2 mm aperture screen, larger particle impurities can be initially removed, and then by centrifugation at a speed of 3000 rpm for 5 minutes, the soil sample can be further homogenized, so that the coefficient of variation (CV) of the soil sample moisture content is less than 5%, ensuring the stability and reliability of the subsequent experimental data. The sample is sealed in a polytetrafluoroethylene (PTFE) sample tube, which has stable chemical properties and a tube wall thickness of ≤1 mm, minimizing interference with the nuclear magnetic signal. Ceramic sealing caps are installed at both ends, with a porosity of <0.1%, which can effectively prevent material exchange between the sample and the external environment and maintain the consistency of the sample state.

[0086] After the Qinghai-Tibet permafrost samples are pretreated, initial parameter calibration is required, which is specifically as follows: the initial total moisture content w0 is determined by the drying method, which is continuously maintained at 105°C for 24 hours, which can ensure that the water in the soil sample is completely evaporated, providing accurate basic data for subsequent unfrozen water content calculation.

[0087] 2. Nuclear magnetic data acquisition and temperature control settings:

[0088] After the Qinghai-Tibet permafrost samples are sealed in the nuclear magnetic sample tube, the temperature control module is used to achieve a gradient temperature drop of the sample in the low temperature range of -30°C to 20°C; the low-field nuclear magnetic resonance instrument is used to scan and obtain the transverse relaxation time T2 spectrum and nuclear magnetic image of the permafrost sample at different temperature nodes.

[0089] In this embodiment, the specific operation and configuration are as follows:

[0090] 1) Nuclear magnetic resonance instrument configuration

[0091] In this embodiment, a low-field nuclear magnetic resonance instrument with a 0.5T permanent magnet is selected, with a magnetic field uniformity of ≤10ppm, which ensures the stability of the magnetic field environment during the experiment. The minimum echo time TEmin is 0.1ms, which supports CPMG sequence acquisition and meets the demand for accurate acquisition of sample signals. The probe coil adopts a double-layer copper shielding design, combined with a frequency adaptive filtering algorithm, with a signal-to-noise ratio (SNR) of ≥90dB, which greatly suppresses high-frequency noise interference. In terms of pulse sequence parameter settings, the echo time TE is 0.2ms, the echo number N is 1024, and the number of accumulations is 32, which, through dynamic shimming technology (using gradient coil adjustment), ensures further improvement of the magnetic field uniformity, making the T2 spectrum resolution reach ±0.01ms, so that the sample transverse relaxation time distribution information can be more accurately obtained.

[0092] 2) Temperature control module operation

[0093] In this embodiment, the temperature control module integrates the liquid nitrogen circulation pipeline and the PID temperature controller. After starting the liquid nitrogen circulation pipeline, the temperature can be accurately set from 20°C to -30°C at a rate of 1°C / min linearly by means of the PID temperature controller, with a temperature control accuracy of up to ±0.5%. Nuclear magnetic scanning is triggered once every 5°C, covering 20°C, 10°C, 0°C, -2°C, -5°C, -7°C, -10°C, -15°C, and -30°C, a total of 9 temperature nodes, fully covering the whole process of soil freezing and thawing. The temperature sensor uses a PT100 platinum resistance, with a temperature measurement accuracy of ±0.01°C. The real-time collected temperature data is transmitted to the main control system, realizing the synchronous recording of temperature and nuclear magnetic signal, and providing reliable data correlation for subsequent analysis of the changes of soil characteristics at different temperatures.

[0094] 3) Data acquisition and storage

[0095] At each temperature point, the system synchronously collects T2 spectrum data (i.e. transverse relaxation time distribution) and nuclear magnetic images (resolution: 512x512 pixels, grayscale range: 0-255). The data storage adopts a hierarchical architecture, in which the T2 spectrum data is saved in CSV format, with detailed records of relaxation time (ms), signal amplitude (a.u.), and corresponding temperature labels, facilitating subsequent data retrieval and analysis. The nuclear magnetic images are stored in DICOM format, retaining not only the spatial resolution but also metadata such as magnetic field strength and sampling time, providing necessary information for accurate image interpretation. The database is backed up to the cloud in real time, supporting multi-terminal access. Researchers can retrieve historical data for retrospective analysis anytime and anywhere, greatly improving the efficiency and security of data utilization.

[0096] 3. Image processing and feature extraction:

[0097] After image preprocessing of the obtained nuclear magnetic images, the temperature-nuclear magnetic signal intensity curve in the nuclear magnetic image is automatically extracted using image recognition technology through coordinate axis detection and correction, scale analysis and numerical conversion, curve extraction and coordinate conversion. The specific process is as follows:

[0098] 1) Image preprocessing

[0099] Gray scale processing is performed using the OpenCV library to convert the RGB nuclear magnetic image to an 8-bit grayscale image, removing color information from the image and reducing interference factors. Adaptive thresholding is used for binary processing, with a threshold value of 120, effectively enhancing the contrast of the coordinate axes and curves in the image. A median filter is applied to remove shot noise, and an edge-preserving filter is used to remove noise while maximizing the protection of curve details, laying a good foundation for subsequent coordinate axis detection and curve extraction operations.

[0100] 2) Coordinate axis detection and correction

[0101] The image edge is extracted by Canny algorithm (low threshold 50, high threshold 150) to generate a clear edge intensity map. The improved Hough transform (parameters p = 1, q = p / 180, threshold = 100) is used to detect the horizontal and vertical lines in the image, and the length of the interference line segment is filtered out <100 pixels to improve the detection accuracy. Based on the least square method, the axis equation is fitted, the coordinate axis inclination angle is accurately calculated (tolerance <0.5°), and the image is geometrically corrected by affine transformation to ensure that the coordinate axis in the image is in the standard position, which is convenient for subsequent scale analysis and numerical conversion.

[0102] 3) Scale analysis and numerical conversion

[0103] The image bottom 50 pixel height range is intercepted as the x-axis scale area, and the left 50 pixel width range is intercepted as the y-axis scale area to extract the scale information. The Tesseract OCR engine parameters are configured to accurately extract the scale text. Regular expressions are used to analyze the scientific notation (such as "2.5e-3" to 0.0025) and fraction format (such as "1 / 293.15" to 0.00341), and automatically calibrate the scale offset error to accurately convert the scale information in the image to the actual physical quantity value.

[0104] 4) Curve extraction and coordinate conversion

[0105] The target curve is segmented based on the HSV color space. For red curves, H = 0-10, S >= 120, V >= 70; for blue curves, H = 100-130, S >= 120, V >= 70. Morphological closing operation (3x3 rectangular kernel, dilated 2 times and then eroded 1 time) is applied to fill the contour breakage and connect the discrete curve segments. The continuous contours with an area >= 100 pixels are selected, sorted in ascending order of x coordinate, and the curve key points are extracted. Through bilinear interpolation, the pixel coordinates are mapped to the actual physical quantity, realizing accurate conversion from image pixel coordinates to actual physical quantity coordinates, and the calculation formula is:

[0106]

[0107] In the calculation formula, x real , y real are the converted actual physical horizontal and vertical coordinate values (the unit is determined by the scale unit analyzed by OCR);

[0108] x pixel , y pixel are the horizontal and vertical pixel coordinates of the curve points in the image (the origin is the upper left corner of the image, x pixel ∈[0,W], y pixel∈ [0, H]);

[0109] W, H are image width and height (pixel number, such as img_shape[::-1][:2] to get the width and height value);

[0110] x max , x min represent the x-axis scale range (i.e. the actual physical quantity maximum / minimum value);

[0111] y max , y min represent the y-axis scale range (i.e. the actual physical quantity maximum / minimum value).

[0112] 4. Data analysis and model calculation:

[0113] Numerical integration is performed on the liquid water signal in the T2 spectrum, and the unfrozen water content is dynamically calculated and freeze-thaw coupling analysis is performed. The specific calculation and analysis method is:

[0114] 1) Unfrozen water content calculation

[0115] In this embodiment, numerical integration (Simpson's rule, step size 0.01 ms) is performed on the T2 spectrum of each temperature point, and the integration range is set to 10 ms-1000 ms. Since the ice signal T2<10 μs will be filtered by hardware, this integration range can effectively obtain the liquid water signal.

[0116] The integral value of the liquid water signal is:

[0117]

[0118] The integral calculation of the liquid water signal in the T2 spectrum is based on the following nuclear magnetic signal intensity formula, and the content of the unfrozen water at each temperature is dynamically calculated;

[0119]

[0120] In the formula, M T is the nuclear magnetic signal intensity at temperature T; M T0 is the nuclear magnetic signal intensity corresponding to the initial liquid water; w0 is the initial total water content.

[0121] According to the unfrozen water content at each temperature point, the system automatically generates a temperature-unfrozen water content curve, and accurately labels the freezing point, stable section and other key inflection points, and intuitively presents the trend of the unfrozen water content with temperature change.

[0122] 2) Freeze-thaw coupling analysis

[0123] In this embodiment, by establishing a linear mapping relationship between the nuclear magnetic image gray value (0-255) and the pore radius (such as gray=200 corresponding to R=0.1 μm), a pore-fracture co-evolution matrix is constructed to realize the accurate analysis of the dynamic of unfrozen water and pore evolution in the freezing and thawing process. The specific method is as follows:

[0124] 2.1: Extracting pore distribution characteristics and establishing mapping relationship

[0125] (1) T2 spectrum inversion pore radius:

[0126] Using the linear relationship between transverse relaxation time T2 and pore radius R:

[0127] R=3ρ2·T2;

[0128] Wherein, the surface relaxation rate ρ2 takes the empirical value of 10 nm / ms, and the T2 distribution curve is converted into the pore radius distribution.

[0129] (2) Pore classification and quantification

[0130] The pores are divided into ultramicro-pores (<0.01 μm), micro-pores (0.01 μm-0.2 μm), small pores (0.2 μm-1 μm), medium pores (1 μm-3 μm) and large pores (>3 μm). By integrating the peak area of T2 spectrum, the pore volume proportion of each pore diameter interval is calculated:

[0131]

[0132] Wherein, V s is the pore volume of a certain pore diameter interval, V a is the total pore volume.

[0133] When the micro-pore volume proportion is more than 60%, the micro-pore radius change amount (ΔR 微孔隙 ) is the main variable, reflecting the mesoscopic frost heave mechanism; when the large pore volume proportion is more than 10%, the fracture width change amount (ΔR 裂隙 ) is the main variable, corresponding to the macroscopic damage process.

[0134] Nuclear magnetic image gray-pore radius mapping establishes a linear relationship between the nuclear magnetic image gray value (0-255) and the pore radius:

[0135] Gray value G=k·R+b;

[0136] The parameters k and b are calibrated by mercury injection method (MIP) to realize the quantitative conversion of image gray value to pore radius.

[0137] 2.2: Analysis of pore evolution law

[0138] Pore-fracture co-evolution matrix construction, define the evolution matrix M to describe the coupling relationship between pore structure and unfrozen water, temperature during freeze-thaw process:

[0139]

[0140] wherein, ΔR is the pore radius change amount (μm / ℃), is the unfrozen water gradient (% / mm), is the temperature gradient (℃ / mm).

[0141] ΔR 微孔隙 is the radius change amount of micropore with a pore size less than 0.2 μm in the freeze-thaw process due to the stress effect of ice crystal growth, and its physical meaning is to quantify the frost heave characteristics of micropore at the mesoscale

[0142] ΔR 裂隙 is the width change amount of fracture with a width greater than 1 μm in the freeze-thaw process due to the effect of ice wedge, and its physical meaning is to represent the expansion rate of fracture at the macro scale.

[0143] Both of them provide key parameters for the analysis of pore-water-thermal coupling mechanism in the multi-scale freeze-thaw process through the difference in pore structure evolution between mesoscale and macro scale.

[0144] 5. Verification and optimization:

[0145] Through the precision verification experiment, the measurement precision of unfrozen water identification method and the pore radius measurement precision of the embodiment are verified; through the robustness test, the salt ion interference resistance ability of the unfrozen water identification method of the embodiment is verified.

[0146] 1) Precision verification experiment

[0147] Based on the theoretical model and simulation analysis of the method of the embodiment, the related precision and ability of the Qinghai-Tibet permafrost sample are predicted: through the nuclear magnetic signal integral model and the temperature-water coupling algorithm, the average error of unfrozen water content can be controlled within 5%, which is significantly lower than the 15%-20% error of traditional differential scanning calorimetry (DSC), and the precision is improved due to the non-destructive quantification ability of nuclear magnetic technology for liquid water signal, avoiding the sample structure interference of traditional destructive detection;

[0148] Based on the theoretical mapping relationship between T2 spectrum and pore radius, combined with surface relaxation rate calibration, when the pore radius R = 0.5 μm, the measurement error is ≤0.05 μm, and the resolution can reach 0.1 μm. This precision can effectively capture the micro features of soil pore structure and provide quantitative basis for freeze-thaw damage analysis; through the pore-fracture co-evolution matrix and the freeze-thaw coupling model, the error of frost heaving risk grade division is better than that of the traditional empirical model (the error of traditional model is ≥30%), the prediction accuracy is improved by more than 50%, and the dynamic classification of low risk (|λ1|<0.3), medium risk (0.3≤|λ1|<0.7) and high risk (|λ1|≥0.7) can be realized, which provides risk warning support based on eigenvalue threshold for Qinghai-Tibet Railway, cold region tunnel and other projects.

[0149] 2) Robustness test

[0150] Within the range of 0-5% salt content, the system signal-to-noise ratio is stable ≥90dB, and the salt ion interference suppression rate is >90%, indicating that this technology has strong resistance to salt ion interference. The multi-thread parallel computing architecture supports processing 10 frames of images per second, meeting the efficiency requirements of real-time monitoring.

[0151] In addition, in the implementation process of the recognition method of the embodiment, the following method is used for parameter optimization:

[0152] The optimal parameters (p = 1, q = n / 180) of Hough transform are determined by grid search (p = 0.5-2, q = n / 360-n / 90), which improves the accuracy of coordinate axis detection. The OCR regular expression is extended to support formats such as "3.2x10-3" and "5.6x10 2 " to improve the recognition rate. The deep learning model (CNN) is introduced to optimize the curve segmentation, which further optimizes the image processing and feature extraction effect.

[0153] Embodiment 2

[0154] Embodiment 2 proposes a nuclear magnetic resonance and intelligent image dual-modal unfrozen water dynamic recognition system to realize the unfrozen water dynamic precise recognition method of embodiment 1. The system includes a multi-modal data acquisition subsystem, an intelligent image processing unit, and a data analysis module.

[0155] In embodiment 2, the multi-modal data acquisition subsystem is used to obtain the transverse relaxation time T2 spectrum and nuclear magnetic image of the frozen soil sample at different temperature nodes. The acquisition subsystem includes:

[0156] 1) NMR acquisition module: a low-field NMR instrument with a 0.5T permanent magnet is used, with a minimum echo time TEmin = 0.1ms, supporting CPMG sequence acquisition. The low-field NMR instrument is equipped with an anti-interference double-layer shielding probe coil, with a magnetic field uniformity ≤ 10ppm. Through dynamic shimming technology and frequency adaptive filtering algorithm, the signal-to-noise ratio is ≥ 90dB, ensuring the reliability of the signal under extreme low temperature (-30℃) and complex pore structure conditions, and it is suitable for frozen soil detection.

[0157] 2) Temperature control module: integrated with liquid nitrogen circulation pipeline and PID temperature controller, the working temperature range is -30℃ to 20℃, the temperature control accuracy is ±0.01℃, the temperature closed-loop control is realized through the PID temperature controller, the temperature sensor used is PT100 platinum resistance, and the temperature measurement accuracy is ±0.01℃; through gradient cooling design (rate 1℃ / min) and 9 temperature node scanning strategy, the whole process of soil sample freezing and thawing is fully covered.

[0158] In this embodiment 2, the intelligent image processing unit is equipped with OpenCV and TesseractOCR library, which is used for coordinate axis detection, curve fitting extraction, scale analysis and data conversion of NMR images;

[0159] The intelligent image processing unit adopts a multi-thread parallel architecture, supports multi-thread parallel computing, and the single-frame image processing time is <0.5s, supporting continuous monitoring of 10 frames per second, meeting the real-time demand of engineering.

[0160] The intelligent image processing unit includes an adaptive color segmentation module, which supports red and blue double-color curve recognition, and the HSV threshold range is as follows:

[0161] Red: lower = [0, 120, 70], upper = [10, 255, 255],

[0162] Blue: lower = [100, 120, 70], upper = [130, 255, 255];

[0163] In this embodiment 2, the data analysis module is based on Python platform, with built-in Qingzang soil special analysis model, freeze-thaw coupling analysis module and data verification function.

[0164] 1) Qingzang soil special analysis model: including unfrozen water content dynamic calculation model and pore-fracture co-evolution model.

[0165] 1.1) The unfrozen water content dynamic calculation model calculates the unfrozen water content ω by integrating the liquid water peak area in the T2 spectrum and combining the initial total water content w0.

[0166] 1.2) The pore-crack co-evolution model introduces the pore characteristic factor Fs and the surface relaxation rate ρ for calibration, and combines the validation data of the mercury intrusion method to achieve a pore radius resolution of 0.1μm and can identify microcrack extension (accuracy ±0.05μm). The specific method is as follows:

[0167] The pore-fracture coevolution model introduces the pore characteristic factor F s (Micropores <0.2μm, cracks >1μm):

[0168]

[0169] Where V 微孔 V is the volume of micropores with a pore size less than 0.2 μm, which is calculated by integrating the peak area of ​​the corresponding relaxation time interval in the T2 spectrum, reflecting the microscopic frost heave mechanism. 裂隙 The volume of cracks with a width greater than 1 μm is determined by mapping the grayscale of the NMR image with the pore radius calibrated by mercury intrusion porosimetry, which characterizes the macroscopic crack expansion.

[0170] Calibrated with the surface relaxation rate ρ2 = 10 nm / ms, the multi-scale evolution equations are established:

[0171]

[0172] Among them, the evolution coefficients k1 and k2 are the mercury injection radius of micropores and cracks measured by mercury injection method. And fitted.

[0173] In the temperature range ΔT, the gradient Approximately constant, the state change can be obtained by integrating the time:

[0174]

[0175] After normalizing the time variable (Δt=1), the pore-fracture coevolution model and the pore-fracture coevolution matrix have the following relationship:

[0176]

[0177] The specific calibration process is as follows: first, data acquisition is performed, the pore radius is measured by mercury intrusion porosimetry (MIP), and the micropore volume V is calculated using the nuclear magnetic resonance T2 spectrum. 微孔 , fracture volume V 裂隙 and get F s Then perform parameter fitting, substitute the data into the evolution equation, adjust k1 and k2, and make |R MIP -R 模型 |<=0.05μm; Finally, verification was carried out to ensure that the crack extension calculated by the model and the scanning deviation were ≤0.05μm in the freeze-thaw experiment.

[0178] Freeze-thaw coupling analysis module: used to construct the pore-fracture co-evolution matrix (such as matrix M), and analyze the frost heaving risk level (low / medium / high risk) by the matrix eigenvalue (λ1, λ2, sorted by the modulus |λ|, |λ1|≥|λ2|). The specific risk classification is as follows:

[0179] Low risk: |λ1|<0.3, corresponding to gentle pore-fracture evolution, such as stable section in permafrost region, frost heaving deformation rate <0.5% / year, unfrozen water gradient Temperature gradient Weak water-heat-force coupling, high soil stability.

[0180] Medium risk: 0.3≤|λ1|<0.7, corresponding to transition section in seasonal frozen soil region, frost heaving deformation rate 0.5%-1.5% / year, and Moderate coupling Pore-fracture evolution is significant but controllable.

[0181] High risk: |λ1|≥0.7, corresponding to water-rich fracture development area, such as Qinghai-Tibet Railway subgrade thawing section, frost heaving deformation rate >1.5% / year, Fracture rapid expansion (ΔR 裂隙 ≥0.1 μm / ℃), high disaster risk.

[0182] Data verification function: automatically correct calculation errors by comparing experimental data with theoretical model.

[0183] Data analysis module: integrated cubic spline interpolation algorithm, as a piecewise cubic polynomial interpolation method, by constructing a cubic polynomial between adjacent data points, ensuring the continuity of function value, first derivative and second derivative, thus forming a smooth interpolation curve. This algorithm can accurately fit discrete data such as pore distribution, meeting the requirement of interpolation error ≤2%, and can generate continuous pore structure curve, which on the one hand provides support for visual analysis of frost heaving risk, including realizing pore size distribution and frost heaving potential energy correlation modeling and frost heaving risk level classification, on the other hand provides pore connectivity interpolation quantification and other continuous medium parameters for unfrozen water migration simulation, thereby improving the accuracy of freeze-thaw coupling analysis.

[0184] System expansion and technology upgrade:

[0185] In terms of system expansion, the unfrozen water dynamic precise identification system of embodiment 2 also has the following expansion directions:

[0186] Technical upgrade direction: In the future, higher field strength (1.0T) nuclear magnetic resonance instrument can be introduced, which is expected to further improve the T2 spectrum resolution to ±0.005ms and obtain more detailed sample information. Develop multi-physical field coupling model, integrate temperature, seepage, stress and other data, realize full-dimensional analysis of frost heaving mechanism, and provide more comprehensive theoretical support for engineering application. Combined with edge computing technology, realize millisecond-level real-time data analysis and early warning to meet the needs of rapid decision-making in engineering site.

[0187] Application expansion scenario: In the ecological restoration field of permafrost regions, by monitoring the coordinated changes of soil moisture-salt-pore, scientific basis is provided for formulating reasonable vegetation restoration strategies. In underground energy storage engineering, the thermal stability of permafrost is evaluated, the design of ground source heat pump system is optimized, the energy utilization efficiency is improved, and the development and utilization of clean energy are promoted.

[0188] In addition, the frozen water dynamic precise identification method and system of the present application has the implementation closed loop, which is specifically manifested as:

[0189] Data flow: From the beginning of nuclear magnetic signal acquisition, through image processing, feature extraction, model calculation, finally providing basis for engineering decision-making, engineering practice feedback further optimizes the whole technical process.

[0190] Verification closed loop: First, parameter calibration is carried out in the laboratory, then verification is carried out in the actual environment, parameter iteration is carried out according to the verification result, and finally large-scale application is realized, forming a complete technical implementation closed loop, and continuously improving the reliability and practicality of the technology.

[0191] Embodiment 3

[0192] Embodiment 3 provides a computer readable storage medium which stores program codes for executing the above identification method, and the codes include an image preprocessing unit, a Hough transform detection unit, an OCR analysis unit and a unfrozen water content calculation unit. The image preprocessing unit supports batch processing and can process multiple nuclear magnetic images at the same time, and the processing speed is significantly improved. The unfrozen water content calculation unit supports multi-core parallel computing, and the calculation speed is significantly improved compared with single core.

[0193] The above is only the preferred embodiment of the present application, and does not limit the present application in any way. Any person skilled in the art, without departing from the scope of the technical solution of the present application, makes any form of equivalent replacement or modification of the technical solution and technical content disclosed by the present application, and still belongs to the protection scope of the present application.

Claims

1. A method for dynamic identification of unfrozen water using nuclear magnetic resonance and image intelligence dual modalities, characterized in that: The steps include: S1: Seal the frozen soil sample in a nuclear magnetic resonance sample tube and use temperature control to achieve gradient cooling of the sample within a preset low temperature range; S2: Use nuclear magnetic resonance technology to scan and obtain the transverse relaxation time T2 spectrum and nuclear magnetic images of frozen soil samples at different temperature nodes; S3: After performing image preprocessing on the nuclear magnetic image, automatically extracting a temperature-nuclear magnetic signal intensity curve from the nuclear magnetic image using image recognition technology; S4: Integrate the liquid water signal in the T2 spectrum and dynamically calculate the content of unfrozen water at various temperatures based on the following NMR signal intensity formula; Where M T is the nuclear magnetic signal intensity at temperature T; M T0 is the nuclear magnetic signal intensity corresponding to the initial liquid water; w0 is the initial total water content; S5: using the transverse relaxation time T2 spectrum to invert the pore radius, extract the pore distribution characteristics, calculate the pore volume proportion of each pore size interval, and establish a linear relationship between the grayscale value of the nuclear magnetic resonance image and the pore radius; Step S5 is specifically as follows: S5.1: Using the linear relationship between the transverse relaxation time T2 spectrum and the pore radius R, invert the pore radius and convert the T2 distribution curve into a pore radius distribution: R=3ρ2·T2; Among them, the surface relaxation rate ρ2 takes an empirical value of 10nm / ms; S5.2: Pore classification and quantification: According to the pore size, the pores are divided into ultramicropores, micropores, small pores, mesopores and macropores. The pore volume proportion of each pore size interval is calculated by integrating the T2 spectrum peak area: Among them, V s is the pore volume in a certain pore size range, V a is the total pore volume; When the micropore volume accounts for more than 60%, the change in micropore radius ΔR 微孔隙 The main evolution reflects the microscopic frost heave mechanism; when the volume of macropores exceeds 10%, the change in crack width ΔR 裂隙 Become the main variable, corresponding to the macro-destruction process; S5.3: Measure the soil pore radius distribution using mercury intrusion porosimetry, extract the grayscale value of the pore area in the NMR image of the same soil sample, establish a mapping relationship between the NMR image grayscale and the pore radius, and thus construct a linear relationship between the NMR image grayscale value and the pore radius; S6: Construct the pore-fracture co-evolution matrix and define the evolution matrix M to describe the coupling relationship between pore structure, unfrozen water and temperature during the freeze-thaw process: Where, ΔR 微孔隙 ΔR is the radius change of micropores with a pore size less than 0.2 μm caused by the stress of ice crystal growth during the freeze-thaw process, 裂隙 It represents the width change of cracks with a width greater than 1 μm caused by ice wedge action during the freezing and thawing process. is the unfrozen water gradient, is the temperature gradient.

2. The method for dynamic identification of unfrozen water by nuclear magnetic resonance and image intelligence dual modality according to claim 1, characterized in that: In step S1, the preset low temperature range is -30°C to 20°C; In step S2, a 0.5T permanent magnet low-field nuclear magnetic resonance instrument is used to perform nuclear magnetic resonance scanning on the frozen soil sample, triggering a nuclear magnetic resonance scan every 5°C to obtain the transverse relaxation time T2 spectrum and nuclear magnetic resonance image of the frozen soil sample at 9 temperature nodes: 20°C, 10°C, 0°C, -2°C, -5°C, -7°C, -10°C, -15°C, and -30°C, fully covering the entire process of soil sample freezing and thawing; During the nuclear magnetic resonance data acquisition process, the CPMG sequence is used, the echo time TE = 0.2ms, the accumulation number is 32 times, and the system can ensure a signal-to-noise ratio ≥ 90dB within the temperature range of -30℃ to 20℃.

3. The method for dynamic identification of unfrozen water by nuclear magnetic resonance and image intelligence dual modality according to claim 1, characterized in that: In step S3, The image preprocessing process includes: Grayscale processing: convert the MRI image from RGB color space to grayscale image; Binarization: Adaptive threshold method is used with the threshold set to 120 to convert the grayscale image into a binary image; Median filtering: Use a 3×3 median filter to remove shot noise from the image; The image recognition technology includes: Improved Hough transform is used to identify the temperature axis and signal intensity axis, and the Canny edge detection algorithm and least squares fitting are combined to achieve coordinate axis tilt correction; Integrate the Tesseract OCR engine with regular expressions to extract and parse scale values, automatically calibrate scale offset errors, and accurately convert scale information in images into actual physical quantity values; Based on HSV color space segmentation and morphological closing operation, pixel-level extraction of temperature-nuclear magnetic signal intensity curve is achieved.

4. The method for dynamic identification of unfrozen water by nuclear magnetic resonance and image intelligence dual modality according to claim 3, characterized in that: The parameters of the improved Hough transform are set as follows: ρ = 1, θ = π / 180, threshold = 100, minimum line segment length = 100 pixels; The specific method for pixel-level extraction of the temperature-NMR signal intensity curve based on HSV color space segmentation and morphological closing operation is as follows: first, the red curve is segmented using the HSV color space, and the HSV threshold range is set to lower = [0, 120, 70] and upper = [10, 255, 255]. Second, the curve contour is extracted using the OpenCV library, and noise points with an area of ​​less than 100 pixels are filtered out. Finally, the pixel coordinates are converted into actual numerical values, and the temperature and NMR signal intensity values ​​corresponding to the curve points are calculated using linear interpolation.

5. A nuclear magnetic resonance and image intelligent dual-modality unfrozen water dynamic identification system, used to implement the identification method according to any one of claims 1 to 4, characterized in that: include: Multimodal data acquisition subsystem: includes a nuclear magnetic resonance acquisition module and a temperature control module, used to obtain transverse relaxation time T2 spectra and nuclear magnetic resonance images of frozen soil samples at different temperature nodes; Intelligent image processing unit: equipped with OpenCV and TesseractOCR libraries, used for coordinate axis detection, curve fitting extraction and scale analysis of nuclear magnetic resonance images; Data analysis module: Based on the Python platform, it has built-in unfrozen water content calculation algorithm and pore evolution model.

6. The nuclear magnetic resonance and image intelligent dual-modality unfrozen water dynamic identification system according to claim 5, characterized in that: In the multimodal data acquisition subsystem, NMR acquisition module: uses a low-field NMR instrument equipped with a 0.5T permanent magnet, with a minimum echo time TEmin = 0.1ms, supports CPMG sequence acquisition, is equipped with an anti-interference double-layer shielded probe coil, has a magnetic field uniformity of ≤10ppm, and a signal-to-noise ratio of ≥90dB; Temperature control module: Integrates liquid nitrogen circulation piping and PID temperature controller, with an operating temperature range of -30°C to 20°C and a temperature control accuracy of ±0.01°C. The PID temperature controller achieves closed-loop temperature control. The temperature sensor is a PT100 platinum resistance with a temperature measurement accuracy of ±0.01°C. The intelligent image processing unit supports multi-threaded parallel computing. The intelligent image processing unit includes an adaptive color segmentation module and supports red and blue color curve recognition. The HSV threshold range is as follows: Red: lower = [0, 120, 70], upper = [10, 255, 255], Blue: lower = [100, 120, 70], upper = [130, 255, 255]; The data analysis module has a built-in data verification function, which automatically corrects calculation errors by comparing experimental data with theoretical models.

7. The nuclear magnetic resonance and image intelligent dual-modality unfrozen water dynamic identification system according to claim 5, characterized in that: The data analysis module integrates a cubic spline interpolation algorithm with an interpolation error of ≤2%, and supports visualization output of pore distribution and frost heave risk level classification.

8. A computer-readable storage medium storing a program code for executing the identification method according to any one of claims 1 to 4, characterized in that: The code includes an image preprocessing unit, a Hough transform detection unit, an OCR parsing unit and an unfrozen water content calculation unit, wherein the image preprocessing unit supports batch processing and the unfrozen water content calculation unit supports multi-core parallel computing.

Images