Soil pore structure monitoring method and system based on low-field nuclear magnetic resonance imaging

Abstract

The invention relates to the technical field of soil pore structure monitoring, and discloses a soil pore structure monitoring method, system and device based on low-field nuclear magnetic resonance imaging and a storage medium, and the method comprises the steps: obtaining original data, and preprocessing the original data to obtain first correction data; calculating second correction data based on the first correction data; performing feature extraction and enhancement on the second correction data to generate third correction data; reconstructing a pore three-dimensional model, and extracting a first pore feature data set from the pore three-dimensional model; establishing a first quantitative correlation model by using a support vector machine algorithm in combination with a preset biological model; and optimizing the pore characteristic data set based on a simulated annealing algorithm to obtain a final quantitative correlation model. According to the method, through multi-step data processing and modeling, the quantitative relation between the soil pore structure and the plant root system function can be disclosed efficiently and accurately, and the problem of low accuracy of a feature extraction result in a traditional microscopic image analysis method is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of soil physical property monitoring, and in particular to a soil pore structure monitoring method, system, electronic equipment and storage medium based on low-field nuclear magnetic resonance imaging. Background Art

[0002] At present, the monitoring and analysis of soil pore structure is of great significance in the fields of agricultural production, environmental protection and geological research. Soil pore structure directly affects the water permeability, aeration and nutrient transport capacity of the soil, and plays a vital role in crop root growth, water resource management and ecosystem stability. However, existing technologies still have many problems and challenges in monitoring soil pore structure.

[0003] In one prior art, the traditional microscopic image analysis method obtains soil pore information through microscopic observation of two-dimensional slices. However, since the sample preparation process requires operations such as cutting, polishing or staining, the original structure of the soil is often destroyed, causing changes in the actual pore morphology and size. In addition, two-dimensional images can only provide limited pore information and lack a description of the three-dimensional distribution and connectivity of pores. This limitation leads to insufficient data integrity, affecting the accuracy and repeatability of the results.

[0004] In summary, the pore information obtained by traditional microscopic image analysis methods is limited and the data is incomplete, resulting in low accuracy of pore feature extraction results. Summary of the invention

[0005] The present invention provides a soil pore structure monitoring method, system, electronic equipment and storage medium based on low-field nuclear magnetic resonance imaging to achieve higher-precision soil pore characteristic analysis.

[0006] In a first aspect, in order to solve the above technical problems, the present invention provides a soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging, comprising: Acquiring low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and performing preprocessing to obtain first correction data; Calculate magnetic field inhomogeneity based on the first correction data, and obtain second correction data according to the magnetic field inhomogeneity; Extracting and enhancing features of the second corrected data to generate third corrected data; reconstructing a high-resolution three-dimensional pore model from the third corrected data, and extracting a first pore characteristic data set from the three-dimensional pore model; Using a support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and combining with a preset biological model to establish a first quantitative association model between pore structure and root function; The imaging parameters were adjusted based on the simulated annealing algorithm to obtain the optimized imaging parameters, the pore characteristic data set under the optimized parameters was re-obtained, and the final quantitative correlation model between pore structure and root function was output.

[0007] In one embodiment, the step of obtaining low-field nuclear magnetic resonance imaging data, temperature data, and moisture data of a soil sample and preprocessing the data to obtain first correction data includes: Obtain low-field nuclear magnetic resonance imaging data, temperature data, and moisture data of soil samples; Ensure the alignment of data from different sources in the time dimension through interpolation algorithms; The original data is adjusted using the preset temperature compensation coefficient and moisture correction coefficient to obtain first correction data.

[0008] In one embodiment, the calculating the magnetic field inhomogeneity based on the first correction data and obtaining the second correction data according to the magnetic field inhomogeneity includes: constructing a magnetic field gradient map based on the first correction data, and calculating the magnetic field inhomogeneity according to the magnetic field gradient map; When the magnetic field inhomogeneity exceeds a preset inhomogeneity threshold, a uniformity optimization algorithm based on magnetic field gradient mapping is applied to process the first correction data to obtain second correction data; when the magnetic field inhomogeneity does not exceed the preset inhomogeneity threshold, the first correction data is directly used as the second correction data.

[0009] In one embodiment, the step of extracting and enhancing features of the second corrected data to generate third corrected data includes: Extracting and enhancing the second corrected data to obtain an enhanced feature set; The enhanced feature set is subjected to wavelet transformation to remove high-frequency noise, while the multi-scale low-frequency features of the pore structure are retained and enhanced to generate third correction data.

[0010] In one embodiment, reconstructing a high-resolution three-dimensional pore model from the third corrected data and extracting a first pore characteristic data set from the three-dimensional pore model comprises: Using voxelization and isosurface extraction techniques, a high-resolution pore 3D model is reconstructed from the third-corrected data; Multidimensional features are extracted from the three-dimensional pore model to establish a first pore feature data set; wherein the multidimensional features reflect the pore size, shape, orientation and connectivity.

[0011] In one embodiment, the method of using a support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and combining a preset biological model to establish a first quantitative association model between pore structure and root function includes: The preset biological model includes root function data of plants in the soil sample; Using the support vector machine algorithm, the pores in the first pore characteristic data set are classified into large, medium and small pores to obtain the pore structure category; The pore structure class is fused with the root function data to obtain a first quantitative association model between pore structure and root function.

[0012] In one embodiment, the method of adjusting imaging parameters based on a simulated annealing algorithm to obtain optimized imaging parameters, reacquiring a second pore characteristic data set under the optimized parameters, and outputting a final quantitative correlation model between pore structure and root function includes: Initializing imaging parameters, wherein the imaging parameters include echo time and sampling frequency; The imaging parameters are optimized using a simulated annealing algorithm to minimize the objective function value; When the objective function value is lower than a preset target threshold, the imaging parameters at this time are used as optimized imaging parameters; Reacquiring optimized nuclear magnetic resonance imaging data using the optimized imaging parameters; generating a second pore characteristic data set based on the optimized nuclear magnetic resonance imaging data; The second pore characteristic data set is integrated with the first quantitative correlation model to output a final quantitative correlation model.

[0013] In a second aspect, the present invention provides a soil pore structure monitoring system based on low-field nuclear magnetic resonance imaging, comprising: A data acquisition and preprocessing module, used to acquire low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and perform preprocessing to obtain first correction data; A magnetic field correction module, configured to calculate magnetic field inhomogeneity based on the first correction data, and obtain second correction data according to the magnetic field inhomogeneity; A feature enhancement module, used for performing wavelet transformation and nonlinear enhancement on the second corrected data to generate third corrected data; A pore feature extraction module, used to reconstruct a high-resolution pore three-dimensional model from the third corrected data, and extract a first pore feature data set from the pore three-dimensional model; A model generation module, for classifying the pores in the first pore characteristic data set into large, medium and small pores using a support vector machine algorithm, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model; The imaging parameter optimization module is used to adjust the imaging parameters based on the simulated annealing algorithm to obtain the optimized imaging parameters, re-acquire the pore characteristic data set under the optimized parameters, and output the final quantitative correlation model between pore structure and root function.

[0014] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging as described above is implemented.

[0015] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the soil pore structure monitoring methods based on low-field nuclear magnetic resonance imaging described above.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging, comprising obtaining low-field nuclear magnetic resonance imaging data, temperature data and moisture data of a soil sample and performing preprocessing to obtain first correction data; calculating magnetic field inhomogeneity based on the first correction data, and obtaining second correction data according to the magnetic field inhomogeneity; extracting and enhancing the second correction data to generate third correction data; reconstructing a high-resolution pore three-dimensional model from the third correction data, and extracting a first pore characteristic data set from the pore three-dimensional model; classifying the pores in the first pore characteristic data set into large, medium and small pores by using a support vector machine algorithm, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model; adjusting imaging parameters based on a simulated annealing algorithm to obtain optimized imaging parameters, reacquiring the pore characteristic data set under the optimized parameters, and outputting a final quantitative association model between pore structure and root function.

[0017] The present invention combines support vector machines and simulated annealing optimization algorithms to effectively preprocess and enhance the features of limited data, and can maintain good performance even when the amount of data is relatively small, ensuring accurate reflection of the true situation of soil pores while improving the accuracy of monitoring results. The final output of the quantitative correlation model between pore structure and root function helps to understand the impact of soil pore structure on plant growth and guide precision agriculture practices. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic flow chart of a soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging provided by the first embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a soil pore structure monitoring system based on low-field nuclear magnetic resonance imaging provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Reference Figure 1 The first embodiment of the present invention provides a soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging, comprising the following steps: S100, acquiring low-field nuclear magnetic resonance imaging data, temperature data and moisture data of a soil sample and performing preprocessing to obtain first correction data; S200, calculating magnetic field inhomogeneity based on the first correction data, and obtaining second correction data according to the magnetic field inhomogeneity; S300, extracting and enhancing features of the second corrected data to generate third corrected data; S400, reconstructing a high-resolution three-dimensional pore model from the third corrected data, and extracting a first pore characteristic data set from the three-dimensional pore model; S500, using a support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model; S600, adjusting the imaging parameters based on the simulated annealing algorithm to obtain the optimized imaging parameters, reacquiring the pore characteristic data set under the optimized parameters, and outputting the final quantitative correlation model between the pore structure and the root function.

[0021] In step S100, it is necessary to obtain low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample.

[0022] Low-field nuclear magnetic resonance imaging data refers to the internal microstructure information of soil obtained by LF-NMR technology. It can provide detailed images of the spatial distribution, size and connectivity of soil pores, and is the basis for building a high-resolution pore three-dimensional model. Temperature data reflects information about the environment in which the soil sample is located or the temperature change of the soil sample itself. Moisture data indicates the water content in the soil.

[0023] In step S200, it is necessary to calculate the magnetic field inhomogeneity based on the first correction data.

[0024] In one implementation, first, it is necessary to construct a magnetic field gradient mapping based on the first correction data. This step evaluates the spatial distribution characteristics of the magnetic field through frequency domain analysis techniques, such as Fourier transform. Specifically, by converting the spatial domain signal into the frequency domain, frequency domain characteristic parameters reflecting the inhomogeneity of the magnetic field can be obtained. For an ideal uniform magnetic field, its spectrum should have only one main frequency component; however, the actual magnetic field often has inhomogeneities, which will appear as additional frequency components in the spectrum. The amplitude and frequency of these additional components can quantify the degree of inhomogeneity of the magnetic field. Based on the frequency domain characteristic parameters, a magnetic field gradient mapping model can be constructed, which can intuitively show the changes in magnetic field intensity in space. By analyzing the gradient mapping, the position distribution of the gradient peaks can be determined. These peaks often correspond to areas where the magnetic field changes most dramatically, and are related to the distribution of paramagnetic substances in the soil.

[0025] Next, the calculation of the magnetic field inhomogeneity index is a key parameter for evaluating the quality of the magnetic field. It can be obtained by calculating the ratio of the standard deviation of the magnetic field strength to the average value. If this ratio exceeds the preset inhomogeneity threshold, it means that the magnetic field inhomogeneity has reached a level that affects the measurement accuracy, and the magnetic field uniformity optimization algorithm needs to be started. The evaluation of magnetic field inhomogeneity is not limited to overall statistical indicators, but also needs to consider local magnetic field changes. Therefore, in some cases, more sophisticated methods, such as the sliding window method or the local mean method, are used to calculate the magnetic field inhomogeneity in different regions. In addition, in order to ensure the reliability of the evaluation results, multiple independent measurements are performed and the results are averaged to reduce the impact of random errors.

[0026] If it is confirmed that the magnetic field inhomogeneity exceeds the preset inhomogeneity threshold, the uniformity optimization algorithm based on magnetic field gradient mapping is started. This algorithm is an iterative optimization method that improves the magnetic field uniformity by adjusting the magnetic field distribution to make the gradient flat. This process involves adjusting the position or strength of the magnetic field source, or adding a compensation coil. After each iteration, the magnetic field gradient mapping is recalculated until the expected uniformity level is reached. Assuming that the initial magnetic field distribution causes abnormally high gradients in certain areas, the optimization algorithm will gradually adjust the magnetic field sources near these areas to make the magnetic field more consistent throughout the measurement volume. During the optimization process, experimental verification is also combined to ensure that the final magnetic field distribution meets the design requirements through repeated testing and adjustment. Preferably, a machine learning algorithm can also be introduced to assist the optimization process to automatically find the best magnetic field configuration solution.

[0027] Next, the optimized magnetic field distribution is used as a constraint condition to perform a secondary correction on the first correction data to obtain the second correction data with improved magnetic field uniformity. For example, if the magnetic field strength in a certain area is 5% higher than the average value, the signal strength in that area can be reduced by 5% accordingly to ensure that the measurement results at different locations are comparable. This correction ensures that the data at all locations can be reliably used to describe the soil pore structure. In addition, after completing the secondary correction, a series of quality checks are required, including but not limited to verifying the improvement effect of the magnetic field uniformity, confirming that no new error sources have been introduced, and ensuring the overall consistency of the data. Only data that has been strictly reviewed can be used for the next step - feature extraction and enhancement.

[0028] In another implementation, if it is confirmed that the magnetic field inhomogeneity does not exceed the preset inhomogeneity threshold, the first correction data is directly used as the second correction data without optimization.

[0029] In step S300, the second corrected data needs to be subjected to feature extraction and enhancement to generate third corrected data.

[0030] In one implementation, feature extraction is first performed on the second corrected data. This step aims to identify and quantify key properties of soil pores, such as pore size distribution, shape characteristics (including circularity, aspect ratio, etc.), and connectivity between pores. By analyzing these characteristics, a more comprehensive understanding of the complexity and diversity of soil pore structure can be achieved. To achieve this, a variety of image processing techniques are used, including morphological operations to define pore boundaries, gray-level co-occurrence matrices (GLCMs) to obtain texture information, and statistical methods to describe the distribution of pore size and shape.

[0031] Next, feature enhancement is performed to improve the quality of the data and the discernibility of the features. The methods used here include applying filters to remove noise, enhancing the image contrast using methods such as histogram equalization or contrast stretching, and edge detection algorithms to highlight pore boundaries.

[0032] Preferably, multi-feature fusion is also performed, that is, different types of features (such as pore size, shape, connectivity) are combined to form a comprehensive feature representation, which can not only capture the information of individual features but also reflect the relationship between them.

[0033] Finally, all the extracted and enhanced features are integrated to generate the third correction data. This process ensures that each pore is described in detail and all features are standardized on the same scale to facilitate subsequent quantitative analysis. At the same time, in order to ensure the quality of the data, necessary quality checks are also performed at this stage to eliminate the introduction of new error sources.

[0034] In step S400, a high-resolution three-dimensional pore model needs to be reconstructed from the third corrected data, and a first pore characteristic data set needs to be extracted from the three-dimensional pore model.

[0035] In one implementation, a voxelization algorithm is first applied to the third corrected data to discretize the continuous pore space into regular voxel units. This process can convert the complex pore structure into a digital representation that is easy to calculate and analyze. After that, a triangular mesh can be constructed between adjacent voxels through an isosurface extraction algorithm to form a three-dimensional surface model of the pores. This method can not only restore the geometric shape of the pores, but also retain the complexity of their internal structure.

[0036] Then, from the reconstructed three-dimensional model, a variety of features reflecting the pore size distribution can be extracted, including pore volume, surface area, equivalent diameter, etc., as well as features describing the pore geometry, including pore aspect ratio, sphericity, roundness, etc. In addition, the angle between the main direction of the pore and the coordinate axis is calculated based on the pore three-dimensional model to obtain the orientation characteristic data characterizing the three-dimensional angular distribution of the pore. Finally, by analyzing the connection relationship between the pores, the number and size of the pore connected clusters can be calculated to obtain the characteristic data measuring the pore connectivity.

[0037] All these extracted features - multi-dimensional heterogeneous features such as pore size, shape, orientation, connectivity, etc., are integrated together to comprehensively construct the first pore characteristic dataset that comprehensively depicts the diversity and complexity of soil pore structure. The first pore characteristic dataset reflects the spatial distribution of soil pores and also contains information about the potential functions of soil. This distribution map can be used to guide precision agriculture practices, such as optimizing irrigation strategies or evaluating the effectiveness of soil improvement measures.

[0038] In step S500, a support vector machine (SVM) algorithm is used to classify the pores in the first pore characteristic data set into large, medium and small pores, and a first quantitative association model between pore structure and root function is established in combination with a preset biological model.

[0039] First, a pre-set biological model is used, which contains the root function data of plants in soil samples. These data come from laboratory cultivation experiments or field observations, and provide information on functional parameters such as plant root growth preferences and nutrient absorption capacity. Such biological models can help understand how different types of pore structures affect root growth and function.

[0040] Next, based on the first pore characteristic data set mentioned above, the support vector machine algorithm is applied to classify the pores. The SVM algorithm can be trained to classify pores into heterogeneous categories such as large pores, medium pores, and small pores. This classification not only relies on the size characteristics of the pores, but also considers multidimensional characteristics such as shape, orientation, and connectivity, thereby more accurately describing the spatial distribution pattern of soil pores. By carefully classifying the pores, the influence of soil structure on water movement, gas exchange, and root activity can be better revealed.

[0041] Finally, the classified pore structure categories are fused with the root function data in the preset biological model. This step aims to establish a quantitative association model that can quantify the relationship between pore structure (such as pore size distribution, connectivity index) and root function (such as root extension and nutrient absorption efficiency). For example, large pores (diameter greater than 200 microns) mainly affect the drainage performance of the soil, while medium pores (50-200 microns) are essential for plant root growth, and small pores (less than 50 microns) determine the water retention capacity of the soil. In this way, it is possible not only to determine which pore characteristics have a significant impact on root growth and nutrient absorption, but also to predict the performance of plants under different pore conditions.

[0042] In step S600, it is necessary to adjust the imaging parameters based on the simulated annealing algorithm to optimize the acquired pore characteristic data, and finally output the final quantitative correlation model between the pore structure and the root function.

[0043] First of all, initializing the imaging parameters is the first step of the entire optimization process. At this stage, it is necessary to set the initial imaging parameters, mainly including the echo time and sampling frequency. These parameters directly affect the quality and information content of the image. For example, in practical applications, researchers use a 0.5 Tesla low-field nuclear magnetic resonance instrument equipped with a probe suitable for soil samples. The initial settings are an echo time of 10 milliseconds and a sampling frequency of 1000 Hz. This initial setting provides a starting point for the subsequent optimization and is also the spatial basis for the simulated annealing algorithm to start searching for the optimal solution.

[0044] Next, the simulated annealing algorithm was used to optimize the imaging parameters in order to minimize the objective function value. The simulated annealing algorithm is a global optimization method inspired by the annealing process in metallurgy. In this study, it was used to optimize the key parameters of low-field NMR imaging, namely the echo time and sampling frequency. These parameters not only affect the image quality and information content, but also determine important characteristics such as signal intensity, signal-to-noise ratio, and spatial resolution. The simulated annealing algorithm iteratively adjusts the echo time and sampling frequency, and re-performs low-field NMR imaging based on the parameters obtained in each iteration to obtain the corresponding pore structure and root function data. Then, the residuals of the acquired pore structure and root function data with the first quantitative association model are calculated, and the residual minimization is used as the objective function of the simulated annealing optimization. A key feature of this optimization method is that it can jump out of the local optimal solution. In practice, this means that even if the result obtained in a certain iteration is slightly worse, the algorithm still has a certain probability of accepting this result to explore a wider solution space. For example, during the optimization process, even if a certain iteration increases the residual from 0.05 to 0.06, the algorithm still accepts this change because it leads to a better global optimal solution. The optimization process will set a convergence condition or a maximum number of iterations. For example, it is set to converge when the residual improvement is less than 0.1% for 50 consecutive iterations, or a maximum of 1000 iterations. This can strike a balance between obtaining a good enough result and controlling the computational cost.

[0045] When the objective function value is lower than the preset target threshold, it means that a set of optimized imaging parameters has been found. The imaging parameter combination at this time is the optimal solution, which can significantly improve the image quality and the predictive ability of the associated model. Once the optimized imaging parameters are determined, it is time to reacquire high-quality imaging data. Using the optimized parameters for nuclear magnetic resonance imaging, the root structure in the image will be clearer and the pore boundaries will be sharper. This improvement allows researchers to more accurately quantify soil structural characteristics and root morphology at the micron level, improving the accuracy of the analysis.

[0046] Finally, a new pore characterization dataset was generated based on the optimized NMR imaging data reacquired with the optimized imaging parameters. This step ensures that the latest and most accurate soil pore structure information is available for further analysis and modeling. Based on these high-quality data, a second pore characterization dataset was constructed, which not only has higher spatial resolution and better signal-to-noise ratio, but also more accurately reflects the true situation of soil pores. Subsequently, the second pore characterization dataset was combined with the first quantitative association model established previously for integrated analysis. In this way, the original association model can be further verified and improved to ensure that it can more accurately describe the relationship between pore structure and root function. The final output is an optimized and verified quantitative association model, which not only reveals the impact of soil pore structure on plant growth, but also provides a reliable scientific basis for precision agriculture practices, such as guiding farmers to optimize farming methods to improve crop yields and resource utilization efficiency. The entire process embodies an integrated solution from data acquisition to model building to optimization and verification, aiming to provide the most accurate soil-plant system characterization.

[0047] In order to facilitate the understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0048] In another implementation, the first pore characteristic data may also be classified based on a CNN convolutional neural network.

[0049] First, the first pore feature data is preprocessed, including denoising, normalization and other operations to improve the quality and consistency of the data. The preprocessing step ensures that the data input to the CNN model has a high signal-to-noise ratio and all features are compared at the same scale, thus providing a solid foundation for subsequent feature extraction and classification.

[0050] Next, a CNN model specially designed for identifying and classifying pore structures is constructed. The model contains multiple convolutional layers, pooling layers, and fully connected layers. The convolutional layer scans the input image by sliding windows to extract local features such as edges and textures; the pooling layer is used to reduce the spatial size of the feature map while retaining important information, reducing the amount of calculation and preventing overfitting; the fully connected layer globally integrates the features extracted by the convolutional layer and the pooling layer to form a higher-level feature representation, which can better capture the essential properties of the pore structure.

[0051] In the training phase, the CNN model is trained using the labeled pore structure dataset. During the training process, the model parameters are adjusted through the back propagation algorithm so that it can accurately identify and classify different pore types. In order to ensure the robustness and generalization ability of the model, the training dataset should be as diverse as possible, covering the pore structure characteristics under various soil conditions. In addition, data enhancement techniques such as rotation and flipping can be introduced to further expand the number and variability of training samples.

[0052] After training, test the trained CNN model on an independent dataset to evaluate its classification accuracy and generalization ability. If the model's performance is not satisfactory, you can improve the model's performance by adjusting the network structure, adding training data, or optimizing hyperparameters. You can try adding more convolutional layers to extract more complex features, or adjusting hyperparameters such as the learning rate and regularization coefficient to improve the model's learning effect.

[0053] Finally, the trained CNN model was applied to new low-field MRI data to automatically identify and classify pore structures and generate detailed pore feature data sets. These data sets not only contain geometric features such as pore size, shape, and distribution density, but also multidimensional heterogeneous features such as connectivity and orientation, providing a rich information basis for subsequent analysis. Finally, the pore feature data sets identified and classified by the CNN model were integrated and analyzed with the first quantitative association model established previously. This method can not only further verify and improve the original association model, but also ensure that it more accurately describes the relationship between pore structure and root function. The final output is an optimized and verified quantitative association model, which not only reveals the impact of soil pore structure on plant growth, but also provides a reliable scientific basis for precision agriculture practice.

[0054] The working process of the present invention is described below using a relatively common scenario as an example.

[0055] First, multiple soil samples were collected from farmland and placed in a low-field nuclear magnetic resonance imager in the laboratory. LF-NMR imaging data, temperature data, and moisture content data were obtained for each soil sample under a freeze-thaw cycle. These data are time-stamped to ensure alignment of data from different sources in the time dimension. In a 100-second experimental cycle, LF-NMR data is recorded every 0.1 seconds, for a total of 1,000 data points; the temperature sensor is sampled every 5 seconds, obtaining 20 points; and the moisture sensor is sampled every 10 seconds, obtaining 10 points.

[0056] The interpolation algorithm is used to adjust the time-misaligned data, and then the preset temperature compensation coefficient (for sandy soil, the signal strength decreases by 0.5% for every 1°C increase) and moisture correction coefficient (the signal strength increases by 30% when the moisture content increases from 10% to 20%) are used to correct the original data to eliminate the impact of temperature and moisture changes and obtain the first corrected data. After this step, the data originally affected by temperature and moisture fluctuations becomes more stable and reliable, making the subsequent analysis results more accurate.

[0057] Next, the magnetic field inhomogeneity in the first correction data is analyzed, and the magnetic field gradient mapping is constructed. Based on this, a secondary correction is performed to improve the magnetic field uniformity and generate the second correction data. When the magnetic field inhomogeneity index exceeds 5%, the magnetic field uniformity optimization algorithm is activated, and the magnetic field gradient is finally flattened. After the correction, the signal intensity in the area is reduced by 5%, which improves the consistency and comparability of the data.

[0058] Wavelet transform is used to remove high-frequency noise, while retaining and enhancing the multi-scale low-frequency characteristics of the pore structure to generate the third correction data. The preset inhomogeneity threshold is 0.8, so any high-frequency coefficient with an absolute value greater than 0.8 will be regarded as noise and set to zero. Through this treatment, high-frequency noise is significantly reduced, while the main features of the pore structure are retained, and the accuracy of pore identification is increased from the original 85% to 95%, while the noise level is reduced by 70%.

[0059] On this basis, the gradient direction information of the magnetic field gradient mapping is extracted, and the pore size, shape, orientation and other features are combined to form an enhanced feature set that fuses the magnetic field information. The fused feature set is [0.2, 0.1, 0.7, 5, 0.8, 30], where the first three numbers represent the gradient direction, and the following three numbers represent the pore size, roundness and inclination, respectively. This fusion allows the consideration of both magnetic field changes and pore morphology, which helps to more accurately describe and classify pore areas.

[0060] Based on the third correction data, voxelization and isosurface extraction techniques are used to reconstruct a high-resolution three-dimensional model of pores. A 1 cubic centimeter soil sample is divided into 1 million cubic voxels with a side length of 100 microns, and each voxel is marked as a pore or soil particle according to its location. By setting an appropriate threshold, such as considering voxels with a grayscale value greater than 128 as pores, the algorithm can construct a triangular mesh between adjacent voxels to form a three-dimensional surface model of the pores. From the reconstructed three-dimensional model, a variety of features reflecting the pore size, shape, orientation and connectivity can be extracted, including pore volume, surface area, equivalent diameter, aspect ratio, sphericity, roundness, etc. These features can comprehensively characterize the diversity and complexity of soil pore structure, laying the foundation for further analysis.

[0061] The support vector machine algorithm is used to classify the pores in the first pore characteristic data set into large, medium and small pores, and combined with the preset biological model, the first quantitative association model between pore structure and root function is established. This model includes the root function data of plants in soil samples, such as root growth preference and nutrient absorption capacity. The rationality of the classification result is verified by comparing the large pore enrichment area in the classification result with the gradient peak position of the magnetic field gradient mapping. If the overlap exceeds 80%, the classification result is considered to be highly credible. The final output pore heterogeneous category distribution map not only reflects the spatial distribution of soil pores, but also contains information on soil formation history and potential functions.

[0062] In order to further improve the prediction performance of the model, the parameters such as the echo time and sampling frequency of LF-NMR imaging were adjusted based on the simulated annealing algorithm to minimize the residual error of the association between pore structure and root function. The initial settings were an echo time of 10 milliseconds and a sampling frequency of 1000 Hz. Through the simulated annealing process, these parameters were optimized to an echo time of 8.5 milliseconds and a sampling frequency of 1200 Hz, thereby improving image resolution and signal-to-noise ratio. The LF-NMR imaging data of the soil samples were reacquired using the optimized parameters to generate a three-dimensional pore structure image with higher spatial resolution and better signal-to-noise ratio. Before optimization, only pores with a diameter greater than 150 microns could be distinguished, while after optimization, pores with a diameter greater than 80 microns could be clearly identified, greatly improving the accuracy of the analysis.

[0063] For the final pore characteristic data set, a cross-validated nonlinear regression algorithm was used to establish the nonlinear relationship between pore surface area, volume, root growth rate, and nutrient absorption flux. An independent validation sample was used to evaluate the generalization ability of the association model to ensure its reliability. The R² of the model reached 0.85 when predicting root growth rate, but only 0.70 when predicting nutrient absorption flux, which means that the effect of pore structure on root growth is more direct or easier to model than the effect on nutrient absorption. Based on the validation results, the nonlinear regression model was fine-tuned and optimized to improve the robustness and adaptability of the model and ensure that the model can accurately describe the quantitative association between pore structure and root function.

[0064] In summary, the present invention provides a soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging, the method comprising: obtaining low-field nuclear magnetic resonance imaging data, temperature data and moisture data of a soil sample and preprocessing to obtain first correction data; calculating magnetic field inhomogeneity based on the first correction data, and obtaining second correction data according to the magnetic field inhomogeneity; extracting and enhancing the second correction data to generate third correction data; reconstructing a high-resolution pore three-dimensional model from the third correction data, and extracting a first pore characteristic data set from the pore three-dimensional model; using a support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and combining a preset biological model to establish a first quantitative association model between pore structure and root function; adjusting imaging parameters based on a simulated annealing algorithm to obtain optimized imaging parameters, reacquiring the pore characteristic data set under the optimized parameters, and outputting the final quantitative association model between pore structure and root function. The method achieves a detailed description of soil pore structure and an in-depth understanding of its quantitative relationship with root function by performing a series of complex preprocessing, modeling and optimization steps on soil pore structure data. This method not only improves the accuracy and reliability of soil structure analysis, but also provides a scientific basis for precision agriculture practices, helping to optimize farming methods and increase crop yields and resource utilization efficiency.

[0065] Reference Figure 2 The second embodiment of the present invention provides a soil pore structure monitoring system based on low-field nuclear magnetic resonance imaging, comprising: A data acquisition and preprocessing module, used to acquire low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and perform preprocessing to obtain first correction data; A magnetic field correction module, configured to calculate magnetic field inhomogeneity based on the first correction data, and obtain second correction data according to the magnetic field inhomogeneity; A feature enhancement module, used for performing wavelet transformation and nonlinear enhancement on the second corrected data to generate third corrected data; A pore feature extraction module, used to reconstruct a high-resolution pore three-dimensional model from the third corrected data, and extract a first pore feature data set from the pore three-dimensional model; A model generation module, for classifying the pores in the first pore characteristic data set into large, medium and small pores using a support vector machine algorithm, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model; The imaging parameter optimization module is used to adjust the imaging parameters based on the simulated annealing algorithm to obtain the optimized imaging parameters, re-acquire the pore characteristic data set under the optimized parameters, and output the final quantitative correlation model between pore structure and root function.

[0066] It should be noted that the soil pore structure monitoring system based on low-field nuclear magnetic resonance imaging provided in an embodiment of the present invention is used to execute all the process steps of the soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0067] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an imaging parameter optimization program. When the processor executes the computer program, the steps in the above-mentioned soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the imaging parameter optimization module.

[0068] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0069] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0070] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0071] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0072] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0073] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0074] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging, characterized in that: Executed by a computer, including: Acquiring low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and performing preprocessing to obtain first correction data; Calculate magnetic field inhomogeneity based on the first correction data, and obtain second correction data according to the magnetic field inhomogeneity; Extracting and enhancing features of the second corrected data to generate third corrected data; reconstructing a high-resolution three-dimensional pore model from the third corrected data, and extracting a first pore characteristic data set from the three-dimensional pore model; Using a support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and combining with a preset biological model to establish a first quantitative association model between pore structure and root function; The imaging parameters were adjusted based on the simulated annealing algorithm to obtain the optimized imaging parameters, the pore characteristic data set under the optimized parameters was re-obtained, and the final quantitative correlation model between pore structure and root function was output.

2. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1 is characterized in that: The step of obtaining low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and performing preprocessing to obtain first correction data includes: Obtain low-field nuclear magnetic resonance imaging data, temperature data, and moisture data of soil samples; Ensure the alignment of data from different sources in the time dimension through interpolation algorithms; The original data is adjusted using the preset temperature compensation coefficient and moisture correction coefficient to obtain first correction data.

3. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1 is characterized in that: The step of calculating the magnetic field inhomogeneity based on the first correction data and obtaining the second correction data according to the magnetic field inhomogeneity includes: constructing a magnetic field gradient map based on the first correction data, and calculating the magnetic field inhomogeneity according to the magnetic field gradient map; When the magnetic field inhomogeneity exceeds a preset inhomogeneity threshold, a uniformity optimization algorithm based on magnetic field gradient mapping is applied to process the first correction data to obtain second correction data; when the magnetic field inhomogeneity does not exceed the preset inhomogeneity threshold, the first correction data is directly used as the second correction data.

4. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1 is characterized in that: The step of extracting and enhancing the features of the second corrected data to generate third corrected data comprises: Extracting and enhancing the second corrected data to obtain an enhanced feature set; The enhanced feature set is subjected to wavelet transformation to remove high-frequency noise, while the multi-scale low-frequency features of the pore structure are retained and enhanced to generate third correction data.

5. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1 is characterized in that: The step of reconstructing a high-resolution three-dimensional pore model from the third corrected data and extracting a first pore characteristic data set from the three-dimensional pore model comprises: Using voxelization and isosurface extraction techniques, a high-resolution pore 3D model is reconstructed from the third-corrected data; Multidimensional features are extracted from the three-dimensional pore model to establish a first pore feature data set; wherein the multidimensional features reflect the pore size, shape, orientation and connectivity.

6. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1, characterized in that: The method of using the support vector machine algorithm to classify the pores in the first pore characteristic data set into large, medium and small pores, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model, includes: The preset biological model includes root function data of plants in the soil sample; Using the support vector machine algorithm, the pores in the first pore characteristic data set are classified into large, medium and small pores to obtain the pore structure category; The pore structure category is fused with the root function data to obtain a first quantitative association model between the pore structure and the root function.

7. The soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging according to claim 1 is characterized in that: The method adjusts the imaging parameters based on the simulated annealing algorithm to obtain optimized imaging parameters, reacquires the second pore characteristic data set under the optimized parameters, and outputs the final quantitative correlation model between pore structure and root function, including: Initializing imaging parameters, wherein the imaging parameters include echo time and sampling frequency; The imaging parameters are optimized using a simulated annealing algorithm to minimize the objective function value; When the objective function value is lower than a preset target threshold, the imaging parameters at this time are used as optimized imaging parameters; Reacquiring optimized nuclear magnetic resonance imaging data using the optimized imaging parameters; generating a second pore characteristic data set based on the optimized nuclear magnetic resonance imaging data; The second pore characteristic data set is integrated with the first quantitative correlation model to output a final quantitative correlation model.

8. A soil pore structure monitoring system based on low-field nuclear magnetic resonance imaging, characterized in that: include: A data acquisition and preprocessing module, used to acquire low-field nuclear magnetic resonance imaging data, temperature data and moisture data of the soil sample and perform preprocessing to obtain first correction data; A magnetic field correction module, configured to calculate magnetic field inhomogeneity based on the first correction data, and obtain second correction data according to the magnetic field inhomogeneity; A feature enhancement module, used for performing wavelet transformation and nonlinear enhancement on the second corrected data to generate third corrected data; A pore feature extraction module, used to reconstruct a high-resolution pore three-dimensional model from the third corrected data, and extract a first pore feature data set from the pore three-dimensional model; A model generation module, for classifying the pores in the first pore characteristic data set into large, medium and small pores using a support vector machine algorithm, and establishing a first quantitative association model between pore structure and root function in combination with a preset biological model; The imaging parameter optimization module is used to adjust the imaging parameters based on the simulated annealing algorithm to obtain the optimized imaging parameters, re-acquire the pore characteristic data set under the optimized parameters, and output the final quantitative correlation model between pore structure and root function.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the soil pore structure monitoring method based on low-field nuclear magnetic resonance imaging as described in any one of claims 1 to 7.

Images