A method and system for measuring ice thickness in natural rivers based on millimeter waves

Abstract

The present invention discloses a method and system for measuring the thickness of ice in natural rivers based on millimeter waves, which relate to the field of ice thickness measurement. The present invention establishes a three-dimensional mapping model of impurity content, Bragg wavelength deviation, and thickness error by integrating millimeter-wave radar and fiber optic sensing technology, effectively overcoming the systematic interference of impurities in the ice layer on the measurement. By feedback from measured data using the ice-chiseling method, an adaptive algorithm is used to dynamically adjust the impurity judgment threshold, making the system environmentally adaptable and allowing the measurement error to be controlled within the engineering allowable range. In addition, cluster analysis and deep learning technology are combined to perform feature-level fusion of physical measurement data and optical characteristic parameters, thereby improving the generalization capability of the thickness measurement error feedback model. A closed-loop process of "preliminary measurement, impurity identification, error correction, and threshold optimization" is formed, so that the measurement accuracy continues to improve with data accumulation.

Description

Technical Field

[0001] The present invention belongs to the field of ice thickness measurement, and in particular, relates to a method and system for measuring the thickness of ice in natural rivers based on millimeter waves. Background Art

[0002] Traditional measurements of ice thickness in natural river channels are susceptible to interference from impurities contained in the ice (such as silt, clay, algae residues, humus colloids, metabolites of cold-resistant bacteria, and industrial residues), which can lead to inaccurate measurements. In addition, the traditional fixed impurity judgment threshold cannot be changed dynamically, making it unable to adapt to seasonal and regional differences, thereby increasing measurement errors. Summary of the Invention

[0003] In response to the problems in the related art, the present invention proposes a method and system for measuring the thickness of ice in natural rivers based on millimeter waves to overcome the above-mentioned technical problems existing in the existing related art.

[0004] To solve the above technical problems, the present invention is achieved through the following technical solutions:

[0005] The present invention provides a method for measuring the thickness of ice in natural rivers based on millimeter waves, comprising the following steps:

[0006] S1. Collecting data on the content of various easily miscible impurities in multiple known ice samples;

[0007] S2, measuring the Bragg wavelength deviation data of multiple known ice samples, merging the data with the content data of various easily miscible impurities collected in S1, and performing preprocessing, then clustering the processed data and calculating the cluster centers to obtain a cluster center matrix of the comprehensive data of the known ice;

[0008] S3. Based on the cluster center matrix of the known ice cube comprehensive data, obtain the impurity content data of the known ice cube training and test samples, train and test the initial multimodal adaptive convolutional neural network, and adjust the clustering process in S2 according to the test results;

[0009] S4: For measurement points exceeding the initial impurity ice thickness threshold, the Bragg wavelength deviation is considered and the error correction of the preliminary thickness data is performed using the multimodal adaptive convolutional neural network trained and tested in S3.

[0010] S5. Compare the data after error correction in S4 with the thickness data measured by the traditional ice-chiseling method, and adjust the ice thickness threshold for initial impurities based on the comparison result.

[0011] Preferably, the S1 comprises the following steps:

[0012] S11. Setting several types of impurities that are easily mixed in the ice layer of natural rivers to obtain a set of impurity types that are easily mixed in the ice layer;

[0013] S12. Setting an initial impurity consideration ice thickness threshold; based on the ice layer miscible impurity type set, collecting ice samples with known thickness greater than or equal to the initial impurity consideration ice thickness threshold and containing miscible impurities and corresponding thickness data to obtain a known ice sample set and a known ice sample thickness dataset; based on the known ice sample set, collecting content data of various miscible impurities in each known ice sample to obtain a known ice miscible impurity content data matrix;

[0014] An impurity classification system based on particle size and distribution characteristics (such as sediment flocculent structure and algae bottom aggregation) allows for the consideration of impurities in subsequent field measurements of natural river ice, thereby improving measurement accuracy. By setting an ice thickness threshold for considering impurities, invalid thin ice samples are excluded, and detection resources are concentrated on the effective pollution-bearing layer, which is conducive to reducing analysis costs. By collecting a data matrix of the content of impurities easily mixed in ice blocks, data support is provided for the subsequent compensation of the measured ice thickness data, thereby providing a basis for compensating the measurement data for field measurements of ice thickness.

[0015] Preferably, said S2 comprises the following steps:

[0016] S21, using an optical fiber sensor to measure the Bragg wavelength deviation data of each known ice cube sample in the known ice cube sample set to obtain a known ice cube Bragg wavelength deviation data set;

[0017] S22, merging each deviation data in the known ice cube Bragg wavelength deviation data set into the row data of the impurity content data corresponding to the known ice cube in the known ice cube miscible impurity content data matrix, to obtain a known ice cube comprehensive data matrix;

[0018] S23, performing Z-Score normalization processing on the data in the known ice cube comprehensive data matrix to obtain a processed known ice cube comprehensive data matrix; setting an initial neighborhood radius and an initial minimum sample number, performing a clustering operation on the processed known ice cube comprehensive data matrix using the DBSCAN algorithm based on the initial neighborhood radius and the initial minimum sample number, and calculating the center data of each cluster after clustering to obtain a known ice cube comprehensive data cluster center matrix;

[0019] By performing Z-score normalization on the Bragg wavelength deviation and the content of each impurity, the dimension effect is eliminated. DBSCAN can automatically identify clusters of any shape when the data contains noise or has significant differences in cluster density (such as areas with abnormal industrial residue content). By optimizing the neighborhood radius (Eps) and the minimum number of samples (MinPts), it can adapt to the local dense distribution of algae residues and humus colloids. Through iterative optimization of DBSCAN cluster centers, potential correlation patterns between impurity distribution characteristics and thickness deviations are automatically identified.

[0020] Preferably, the step S3 includes the following steps:

[0021] S31. Setting a training sample ratio; selecting training samples from the known ice sample set according to the training sample ratio to obtain a known ice training sample set and a known ice test sample set; obtaining Bragg wavelength deviation data of each known ice training sample and each known ice test sample according to the known ice training sample set, the known ice test sample set, and the known ice Bragg wavelength deviation dataset to obtain a known ice Bragg wavelength deviation training dataset and a known ice Bragg wavelength deviation test dataset;

[0022] S32. Perform Z-Score normalization on the known ice cube Bragg wavelength deviation dataset to obtain a processed known ice cube Bragg wavelength deviation dataset; then, combine the known ice cube comprehensive data cluster center matrix to obtain impurity content data corresponding to the known ice cube training samples and the test samples, to obtain a known ice cube training sample impurity content data matrix and a known ice cube test sample impurity content data matrix;

[0023] S33. Using a linear frequency modulated continuous wave millimeter wave radar, the thickness of each sample in the known ice training sample set and the known ice test sample set is measured by transmitting a signal with a linearly varying frequency, thereby obtaining an actual training dataset of known ice thickness and an actual test dataset of known ice thickness; and calculating the absolute value of the difference between the actual training dataset of known ice thickness and the actual test dataset of known ice thickness and the thickness data of each known ice training sample and known ice test sample in the known ice sample thickness dataset, thereby obtaining a known ice thickness measurement deviation training dataset and a known ice thickness measurement deviation test dataset.

[0024] S34. Constructing an initial multimodal adaptive convolutional neural network; training and testing the initial multimodal adaptive convolutional neural network using a data matrix of impurity content of known ice training samples and a training data set of known ice thickness measurement deviations to obtain test accuracy data;

[0025] S35. When the test accuracy data is less than the test accuracy threshold, the initial neighborhood radius and the initial minimum number of samples are adjusted until the test accuracy data is greater than or equal to the test accuracy threshold, thereby obtaining a final cluster center matrix of the known ice cube comprehensive data and a final multimodal adaptive convolutional neural network.

[0026] By combining Bragg wavelength deviation data with linear frequency-modulated continuous wave thickness measurements, a cross-modal correlation model of impurity content and structural deformation is established, significantly improving the detection dimension; Z-Score processing is used to eliminate sensor dimensional differences and ensure the comparability of data from different sources; dynamic thresholds based on neighborhood radius and minimum sample number are used to effectively eliminate measurement noise interference; the combination of fiber optic sensors and millimeter-wave radar avoids damage to the ice sample structure caused by traditional contact measurements; the test accuracy threshold triggers automatic adjustment of model parameters to ensure the stability of the system in complex environments.

[0027] Preferably, in S32, obtaining impurity content data corresponding to known ice training samples and test samples in combination with the cluster center matrix of known ice comprehensive data to obtain the impurity content data matrix of known ice training samples and the impurity content data matrix of known ice test samples includes the following steps:

[0028] S321. Based on the processed known ice cube Bragg wavelength deviation training data set and the known ice cube comprehensive data cluster center matrix, respectively select the impurity content data corresponding to the Bragg wavelength deviation data in the cluster center with the smallest Euclidean distance between each processed known ice cube Bragg wavelength deviation training data and test data, and use them as the impurity content data of the known ice cube training samples corresponding to the processed known ice cube Bragg wavelength deviation training data and test data, respectively, to obtain the known ice cube training sample impurity content data matrix and the known ice cube test sample impurity content data matrix.

[0029] Preferably, the adjustment of the initial neighborhood radius and the initial minimum number of samples in S35 adopts a honey badger optimization algorithm;

[0030] The Honey Badger algorithm simulates animal foraging behavior to efficiently explore the optimal combination of neighborhood radius and minimum number of samples in the solution space, avoiding the computational redundancy of traditional grid search; the algorithm autonomously switches between global exploration (development phase) and local mining (utilization phase), quickly locking the parameter range that maximizes test accuracy; the adaptively adjusted DBSCAN parameters can accurately identify the complex correlation pattern of impurity content-thickness deviation, and improve the purity of multimodal data clustering; compared with traditional optimization methods such as genetic algorithms, the Honey Badger algorithm reduces the number of iterations by an average of 30% to reach the target accuracy threshold; through the adaptive step size control of the Honey Badger algorithm, the interference of outliers in the measurement data on parameter optimization is effectively suppressed; the optimized parameter combination can adapt to the detection requirements under different ambient temperatures and ice sample morphologies, reducing the frequency of repeated parameter adjustments; the algorithm only requires 0.5MB of memory on the edge computing device to complete real-time parameter optimization.

[0031] Preferably, the S4 comprises the following steps:

[0032] S41. Setting a natural river channel to be measured; selecting a plurality of ice thickness measurement points on the ice in the natural river channel to be measured to obtain a current ice thickness measurement point set; using the linear frequency modulated continuous wave millimeter wave radar in S33 to roughly measure the ice thickness at each current ice thickness measurement point by transmitting a signal with a linearly varying frequency, based on the current ice thickness measurement point set, to obtain a current rough ice thickness data set;

[0033] S42: Using the current ice layer rough thickness data less than the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as the final thickness data of the corresponding measurement point, to obtain a first current ice layer final thickness data set; recording the measurement points corresponding to the current ice layer rough thickness data greater than or equal to the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as an impurity-considered measurement point set;

[0034] S43, using the optical fiber sensor in S21 to measure the Bragg wavelength deviation data of the ice cover at each measuring point in the impurity consideration measurement point set to obtain a current Bragg wavelength deviation data set; obtaining the impurity content data corresponding to each Bragg wavelength deviation data in the current Bragg wavelength deviation data set according to the final known ice block comprehensive data cluster center matrix to obtain a current impurity content data matrix;

[0035] S44. Input each row of data in the current impurity content data matrix into a final multimodal adaptive convolutional neural network for mapping to obtain a current ice layer thickness measurement deviation dataset; merge the thickness data corresponding to the impurity consideration measurement point set in the current rough ice layer thickness dataset with the current ice layer thickness measurement deviation dataset to obtain a second current ice layer final thickness dataset;

[0036] A grading strategy combining rapid initial screening with millimeter-wave radar and precise verification with fiber-optic sensors enables adaptive detection of impurity interference in ice layers, balancing efficiency and accuracy. Impurity interference errors in thickness measurements are automatically corrected based on a combination of Bragg wavelength deviation and impurity content mapped by a neural network. Multimodal analysis is initiated only for suspected impurity-containing areas exceeding a threshold, significantly reducing the real-time processing pressure on edge devices. The millimeter-wave penetration characteristics combined with the fiber-optic temperature compensation mechanism effectively suppress the influence of outdoor environmental noise such as wind, snow, and water fluctuations.

[0037] Preferably, the S5 comprises the following steps:

[0038] S51, measuring each measuring point in the current ice layer thickness measurement point set using a traditional ice-chiseling method to obtain a third current ice layer final thickness data set;

[0039] S52: Setting a current measurement data error threshold; calculating a data difference between the third current ice layer final thickness data set, the second current ice layer final thickness data set, and the first current ice layer final thickness data set to obtain a current natural river ice layer measurement error value;

[0040] S53. When the current natural river channel ice layer measurement error value is greater than or equal to the current measurement data error threshold, adjusting the initial impurity-considered ice thickness threshold until the current natural river channel ice layer measurement error value is less than the current measurement data error threshold, thereby obtaining a final impurity-considered ice thickness threshold;

[0041] By introducing the traditional ice-peeling method as a benchmark verification method, a dynamic threshold optimization mechanism is established to realize the self-calibration function of the millimeter-wave radar and fiber optic sensor fusion system; using the error feedback of measured data and multimodal measurement results, the impurity judgment threshold is dynamically adjusted to effectively solve the problem of impurity interference misjudgment caused by environmental changes, and significantly improve the long-term stability and environmental adaptability of the ice thickness monitoring system; while ensuring the convenience of non-contact measurement, the system has the ability to continuously evolve and can adapt to changes in ice conditions in different river basins and seasons.

[0042] Preferably, adjusting the initial impurities taking into account the ice thickness threshold in S53 includes the following steps:

[0043] S531, generating an impurity consideration threshold according to the value interval of the initial impurity consideration ice thickness threshold, and adjusting the initial position of each honey badger in the honey badger population to obtain a second initial position set;

[0044] S532, constructing the fitness function of the honey badger population by adjusting the impurity threshold;

[0045] S533, adjusting the fitness function of the honey badger population according to the impurity consideration threshold to iterate and update the second initial position set;

[0046] S534: Repeat S533. When the maximum number of iterations is reached, stop the iteration to obtain the final global optimal fitness and the final global optimal position. When the inverse of the final global optimal fitness is less than the current measurement data error threshold, use the final global optimal position as the final impurity consideration ice thickness threshold, and the adjustment is completed.

[0047] By simulating the digging and foraging behavior of honey badgers, the optimal impurity threshold is quickly located in the solution space, avoiding the time loss of traditional trial-and-error methods; combined with the feedback of measured data from the ice-peeling method, it automatically adapts to the impurity distribution characteristics of different waters and improves the generalization of the system; utilizing the algorithm's unique dynamic search step mechanism, the threshold calibration speed is increased by more than 40% while ensuring accuracy; the measurement accuracy and computational cost are simultaneously optimized to achieve the best balance between impurity identification sensitivity and system response speed; and through the honey badger optimization algorithm's built-in population diversity maintenance strategy, the problem of threshold misadjustment caused by local mutations is effectively avoided.

[0048] A millimeter-wave-based natural river ice thickness measurement system includes an ice layer easily mixed impurity type setting module, a known ice layer sample impurity content data acquisition module, a Bragg wavelength deviation data acquisition module, a known ice block comprehensive data cluster center calculation module, a Bragg wavelength deviation training and test data acquisition module, a known ice block training and test impurity content data acquisition module, a thickness deviation data mapping model construction module, a current natural river ice layer measurement module, and an impurity-considered ice thickness threshold comparison and adjustment module.

[0049] The present invention has the following beneficial effects:

[0050] 1. This invention integrates millimeter-wave radar and fiber-optic sensing technologies to establish a three-dimensional mapping model of impurity content, Bragg wavelength deviation, and thickness error, effectively overcoming the systematic interference of impurities in the ice layer on measurements. By using feedback from measured data using the ice-chiseling method, an adaptive algorithm is used to dynamically adjust the impurity judgment threshold, making the system environmentally adaptable and keeping measurement errors within the acceptable range for the project. Furthermore, cluster analysis and deep learning techniques are combined to perform feature-level fusion of physical measurement data and optical characteristic parameters, enhancing the generalization capability of the thickness measurement error feedback model. This forms a closed-loop process of "initial measurement - impurity identification - error correction - threshold optimization," enabling measurement accuracy to continuously improve as data accumulates.

[0051] 2. This invention achieves adaptive detection of impurity interference in ice layers through a grading strategy that combines rapid initial screening with millimeter-wave radar and precise verification with fiber-optic sensors, balancing efficiency and accuracy. Impurity interference errors in thickness measurements are automatically corrected based on a combination of Bragg wavelength deviation and impurity content mapped by a neural network. Multimodal analysis is only initiated for suspected impurity-containing areas exceeding a threshold, significantly reducing the real-time processing pressure on edge devices. The millimeter-wave penetration characteristics are combined with a fiber-optic temperature compensation mechanism to effectively suppress the influence of outdoor environmental noise such as wind, snow, and water fluctuations.

[0052] 3. The present invention simulates the digging and foraging behavior of honey badgers to quickly locate the optimal impurity threshold in the solution space, avoiding the time loss of traditional trial and error methods; combined with the feedback of measured data from the ice-peeling method, it automatically adapts to the impurity distribution characteristics of different waters and improves the generalization of the system; utilizing the algorithm's unique dynamic search step mechanism, the threshold calibration speed is increased by more than 40% while ensuring accuracy; synchronously optimizes measurement accuracy and computational cost to achieve the best balance between impurity identification sensitivity and system response speed; and through the population diversity maintenance strategy of the honey badger optimization algorithm, the problem of threshold misadjustment caused by local mutations is effectively avoided.

[0053] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 This is a schematic diagram of the overall process of a method for measuring ice thickness in natural rivers based on millimeter waves according to the present invention;

[0056] Figure 2 A schematic diagram of the process of collecting training and test data of known ice thickness for the present invention;

[0057] Figure 3 A schematic diagram of the process of constructing the final multimodal adaptive convolutional neural network of the present invention;

[0058] Figure 4 A schematic diagram of a process for adjusting impurities taking into account an ice thickness threshold according to the present invention;

[0059] Figure 5 This is a module schematic diagram of a millimeter wave-based natural river ice thickness measurement system of the present invention. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0061] Example 1

[0062] See also Figure 1-4 This embodiment is a method for measuring the thickness of ice in a natural river channel based on millimeter waves, comprising the following steps:

[0063] S1. Collecting data on the content of various easily miscible impurities in multiple known ice samples;

[0064] Said S1 comprises the following steps:

[0065] S11. Assign several types of easily miscible contaminants in the ice layer of natural rivers to obtain a set of easily miscible contaminant types in the ice layer; the set of easily miscible contaminant types in the ice layer includes silt, clay (particle size > 100 nm, entrained into the ice layer during freezing, forming a flocculent or layered distribution visible to the naked eye), algae residues (diatoms, green algae, etc., frozen in the ice layer after death, aggregated in fibrous or flaky forms, mostly distributed at the bottom of the ice layer), humus colloids (particle size 1-100 nm, formed by plant decomposition products, causing the ice layer to appear light yellow or brown in some areas), metabolic products of cold-resistant bacteria (such as Pseudomonas), and industrial residues (metal debris, oil pollution, etc.);

[0066] S12. Setting an initial impurity consideration ice thickness threshold; based on the ice layer miscible impurity type set, collecting ice samples with known thickness greater than or equal to the initial impurity consideration ice thickness threshold and containing miscible impurities and corresponding thickness data to obtain a known ice sample set and a known ice sample thickness dataset; based on the known ice sample set, collecting content data of various miscible impurities in each known ice sample to obtain a known ice miscible impurity content data matrix;

[0067] The process of collecting data on the content of various easily miscible impurities in each known ice sample includes the following steps:

[0068] S121. Use a plexiglass sampling tube (5 cm diameter) with a liquid nitrogen cooling system (-40°C) to vertically drill a complete ice core while preserving the original layered structure of the ice.

[0069] S122. Marking layers are set every 3 cm along the sampling depth, and fluorescent tracers (e.g., rhodamine B) are used to assist in locating impurity-enriched areas. A pre-cooled polycarbonate membrane (pore size 0.2 μm) is placed over the ice layer to allow van der Waals forces to adsorb colloids and microbial metabolites for ≤ 30 minutes.

[0070] S123, start measurement; among them: for measuring sediment / clay, micro-area X-ray diffraction (μ-XRD) can be used to scan the ice core profile (spatial resolution 50μm, quantitative accuracy of mineral composition ±2%); for measuring algae residues, laser-induced breakdown spectroscopy (LIBS) combined with Raman spectroscopy (identification accuracy of organic matter characteristic peaks up to 0.1cm -1 Humic colloids can be measured by microfluorescence spectroscopy (excitation wavelength 365nm) to detect interstitial fluid, where the detection limit is 0.1μg / L (calculated as humic acid);

[0071] To measure bacterial metabolites, nanoprobe sampling combined with surface-enhanced infrared absorption spectroscopy can be used (metabolite fingerprint matching degree > 90%); to measure industrial residues, synchrotron X-ray fluorescence imaging can be used (element distribution resolution 10 μm, detection limit 0.1 ppm);

[0072] S2, measuring the Bragg wavelength deviation data of multiple known ice samples, merging the data with the content data of various easily miscible impurities collected in S1, and performing preprocessing, then clustering the processed data and calculating the cluster centers to obtain a cluster center matrix of the comprehensive data of the known ice;

[0073] The S2 comprises the following steps:

[0074] S21, using an optical fiber sensor to measure the Bragg wavelength deviation data of each known ice cube sample in the known ice cube sample set to obtain a known ice cube Bragg wavelength deviation data set; the calculation formula of the Bragg wavelength deviation data during the measurement process is as follows:

[0075] ;

[0076] Where a1 represents the Bragg wavelength data, Δa1 represents the Bragg wavelength deviation data, a2 represents the optical fiber elastic coefficient, and ε represents the stress change data;

[0077] S22, merging each deviation data in the known ice cube Bragg wavelength deviation data set into the row data of the impurity content data corresponding to the known ice cube in the known ice cube miscible impurity content data matrix, to obtain a known ice cube comprehensive data matrix;

[0078] S23, performing Z-Score normalization processing on the data in the known ice cube comprehensive data matrix to obtain a processed known ice cube comprehensive data matrix; setting an initial neighborhood radius and an initial minimum sample number, performing a clustering operation on the processed known ice cube comprehensive data matrix using the DBSCAN algorithm based on the initial neighborhood radius and the initial minimum sample number, and calculating the center data of each cluster after clustering to obtain a known ice cube comprehensive data cluster center matrix;

[0079] S3. Based on the cluster center matrix of the known ice cube comprehensive data, obtain the impurity content data of the known ice cube training and test samples, train and test the initial multimodal adaptive convolutional neural network, and adjust the clustering process in S2 according to the test results;

[0080] The S3 includes the following steps:

[0081] S31. Setting a training sample ratio; selecting training samples from the known ice sample set according to the training sample ratio to obtain a known ice training sample set and a known ice test sample set; obtaining Bragg wavelength deviation data of each known ice training sample and each known ice test sample according to the known ice training sample set, the known ice test sample set, and the known ice Bragg wavelength deviation dataset to obtain a known ice Bragg wavelength deviation training dataset and a known ice Bragg wavelength deviation test dataset;

[0082] S32. Perform Z-Score normalization on the known ice cube Bragg wavelength deviation dataset to obtain a processed known ice cube Bragg wavelength deviation dataset; then, combine the known ice cube comprehensive data cluster center matrix to obtain impurity content data corresponding to the known ice cube training samples and the test samples, to obtain a known ice cube training sample impurity content data matrix and a known ice cube test sample impurity content data matrix;

[0083] In S32, combining the cluster center matrix of the known ice cube comprehensive data to obtain the impurity content data corresponding to the known ice cube training samples and the test samples, and obtaining the impurity content data matrix of the known ice cube training samples and the impurity content data matrix of the known ice cube test samples includes the following steps:

[0084] S321. Based on the processed known ice cube Bragg wavelength deviation training data set and the known ice cube comprehensive data cluster center matrix, respectively select the impurity content data corresponding to the Bragg wavelength deviation data in the cluster center with the smallest Euclidean distance between each processed known ice cube Bragg wavelength deviation training data and test data, and use them as the impurity content data of the known ice cube training samples corresponding to the processed known ice cube Bragg wavelength deviation training data and test data, respectively, to obtain a known ice cube training sample impurity content data matrix and a known ice cube test sample impurity content data matrix;

[0085] S33. Using a linear frequency modulated continuous wave millimeter wave radar, the thickness of each sample in the known ice training sample set and the known ice test sample set is measured by transmitting a signal with a linearly varying frequency, thereby obtaining an actual training dataset of known ice thickness and an actual test dataset of known ice thickness; and calculating the absolute value of the difference between the actual training dataset of known ice thickness and the actual test dataset of known ice thickness and the thickness data of each known ice training sample and known ice test sample in the known ice sample thickness dataset, thereby obtaining a known ice thickness measurement deviation training dataset and a known ice thickness measurement deviation test dataset.

[0086] The core principle and implementation process of the linear frequency modulated continuous wave (LFMCW) millimeter wave radar in the S33 to measure the thickness of each sample in the known ice training sample set by transmitting a signal with a linearly varying frequency are as follows:

[0087] The radar transmits a continuous wave whose frequency increases linearly with time (e.g., starting at 77 GHz and with a bandwidth of 4 GHz). Upon encountering the surface and bottom of the ice, two reflected echoes are generated. The reflected signal is mixed with the transmitted signal to generate an intermediate frequency (IF) signal, whose frequency difference is proportional to the ice thickness. The frequency components of the IF signal are extracted through Fourier transform. ,according to Calculating the round-trip time of electromagnetic waves in ice The calculation formula is as follows:

[0088] ;

[0089] Where, Indicates the frequency modulation slope (Hz / s);

[0090] Combined with the dielectric constant of ice and Calculating ice thickness The calculation formula is as follows:

[0091] ;

[0092] Where c represents the speed of light and ε represents the dielectric constant of ice;

[0093] S34. Constructing an initial multimodal adaptive convolutional neural network; training and testing the initial multimodal adaptive convolutional neural network using a data matrix of impurity content of known ice training samples and a training data set of known ice thickness measurement deviations to obtain test accuracy data;

[0094] In S34, the initial multimodal adaptive convolutional neural network is trained and tested using a data matrix of known ice training sample impurity content and a training data set of known ice thickness measurement deviations, including the following steps:

[0095] S341, constructing an initial multimodal adaptive convolutional neural network; training the initial multimodal adaptive convolutional neural network using a data matrix of impurity content of known ice training samples and a training data set of known ice thickness measurement deviations; stopping training when a training error is less than a preset training error threshold, thereby obtaining a trained multimodal adaptive convolutional neural network;

[0096] S342: Setting a test accuracy threshold; inputting the known ice test sample impurity content data matrix as test data and the known ice thickness measurement deviation test data set as test labels into the trained multimodal adaptive convolutional neural network for testing; after the test is completed, obtaining test accuracy data;

[0097] S35. When the test accuracy data is less than the test accuracy threshold, the initial neighborhood radius and the initial minimum number of samples are adjusted until the test accuracy data is greater than or equal to the test accuracy threshold, thereby obtaining a final cluster center matrix of the known ice cube comprehensive data and a final multimodal adaptive convolutional neural network.

[0098] The initial multimodal adaptive convolutional neural network includes:

[0099] Input layer: used to process two data types (impurity feature vector: containing numerical features such as sediment content, algae density, humus particle size distribution, and thickness data);

[0100] Feature fusion convolution layer: A one-dimensional convolution kernel (Conv1D) filters the thickness sequence to extract local temporal features (convolution kernel size = 3, stride = 1), and merges the convolution output with the impurity features to form a mixed feature matrix.

[0101] Multi-head self-attention: Three independent attention heads are set up to respectively focus on the synergistic effect of industrial residues and humus, the influence of particle size distribution on ice structure, and the noise characteristics of measurement equipment;

[0102] Fully connected prediction layers: Layer 1 (128 nodes, ReLU activation), Layer 2 (64 nodes, ReLU activation), and Output layer (1 node, linear activation);

[0103] ‌Dropout layer‌ (dropout rate = 0.2): inserted between fully connected layers to prevent overfitting;

[0104] ‌Batch Normalization‌: Batch normalization is implemented after the convolutional layer;

[0105] Adjusting the initial neighborhood radius and the initial minimum number of samples in S35 includes the following steps:

[0106] S351, setting the value range of the initial neighborhood radius and the initial minimum number of samples to obtain the neighborhood radius value range And the minimum sample number interval , 、 Respectively represent the lower limit and upper limit of the initial neighborhood radius, 、 Respectively represent the lower limit and upper limit of the initial minimum sample number; construct the neighborhood sample number to adjust the honey badger population, and set the maximum number of iterations of the neighborhood sample number to adjust the honey badger population And the current number of iterations is , respectively recorded as the maximum iteration number of neighborhood samples and the current iteration number of neighborhood samples; the number of neighborhood samples adjusts the search space dimension of the honey badger population to 2;

[0107] S352, generating the neighborhood sample number according to the neighborhood radius value interval and the minimum sample number value interval to adjust the initial position of each honey badger in the honey badger population to obtain a first initial position matrix; the generation formulas are as follows:

[0108] ;

[0109] ;

[0110] Where b j1 、b j2 They represent the position components of the initial position of the jth honey badger in the honey badger population adjusted by the number of neighborhood samples in the initial neighborhood radius and the initial minimum number of samples respectively; rand j1 、rand j2 Respectively for b j1 、b j2 Generates a random number between 0 and 1; ceil represents the rounding function;

[0111] S353, constructing the fitness function of the honey badger population by adjusting the number of neighborhood samples ;as follows,

[0112] ;

[0113] Where, e is a natural constant, 、 They respectively represent the test accuracy data and test accuracy threshold after the training and testing of the initial multimodal adaptive convolutional neural network is completed;

[0114] S354, start iteration, set the current iteration number of the neighborhood sample to 1 before iteration; in the first round of iteration, use the number of neighborhood samples to adjust the fitness function of the honey badger population to calculate the fitness value of the initial position of each honey badger in the first initial position matrix to obtain a first fitness value set; use the maximum fitness value in the first fitness value set and the initial position of the corresponding honey badger as the first global optimal fitness and the first global optimal position respectively; update the initial position of each honey badger in the first initial position matrix according to the first global optimal fitness and the first global optimal position; after the update is completed, add 1 to the current iteration number of the neighborhood sample and enter the next round of iteration;

[0115] In each other round of iteration, the fitness function of the honey badger population is adjusted using the number of neighborhood samples updated in the previous round of iteration to calculate the fitness value of the position of each honey badger in the honey badger population, and a second fitness value set is obtained; the maximum fitness value in the second fitness value set and the position of the corresponding honey badger are respectively used as the second global optimal fitness and the second global optimal position; the position of each honey badger in the honey badger population adjusted by the number of neighborhood samples updated in the previous round of iteration is updated according to the second global optimal fitness and the second global optimal position; after the update is completed, the current iteration number of the neighborhood sample is increased by 1 and the next round of iteration is entered;

[0116] S355, when Stop the iteration when ; otherwise, continue the iteration until When the test accuracy data is greater than or equal to the test accuracy threshold, the adjustment is completed, and the final neighborhood radius, the final minimum number of samples, and the final multimodal adaptive convolutional neural network are obtained; otherwise, return to S354 to continue iterating;

[0117] The following example takes the thickness measurement of ice samples in a natural river as an example:

[0118] 1. Initial parameters: Neighborhood radius: 0.35 (empirical value); Minimum number of samples: 8 (empirical value); Accuracy threshold: 92%;

[0119] 2. Comparison of data before and after parameter optimization using the Honey Badger optimization algorithm; see the following table:

[0120] ;

[0121] S4: For measurement points exceeding the initial impurity ice thickness threshold, the Bragg wavelength deviation is considered and the error correction of the preliminary thickness data is performed using the multimodal adaptive convolutional neural network trained and tested in S3.

[0122] The S4 comprises the following steps:

[0123] S41. Setting a natural river channel to be measured; selecting a plurality of ice thickness measurement points on the ice in the natural river channel to be measured to obtain a current ice thickness measurement point set; using the linear frequency modulated continuous wave millimeter wave radar in S33 to roughly measure the ice thickness at each current ice thickness measurement point by transmitting a signal with a linearly varying frequency, based on the current ice thickness measurement point set, to obtain a current rough ice thickness data set;

[0124] S42: Using the current ice layer rough thickness data less than the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as the final thickness data of the corresponding measurement point, to obtain a first current ice layer final thickness data set; recording the measurement points corresponding to the current ice layer rough thickness data greater than or equal to the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as an impurity-considered measurement point set;

[0125] S43, using the optical fiber sensor in S21 to measure the Bragg wavelength deviation data of the ice cover at each measuring point in the impurity consideration measurement point set to obtain a current Bragg wavelength deviation data set; obtaining the impurity content data corresponding to each Bragg wavelength deviation data in the current Bragg wavelength deviation data set according to the final known ice block comprehensive data cluster center matrix to obtain a current impurity content data matrix;

[0126] S44. Input each row of data in the current impurity content data matrix into a final multimodal adaptive convolutional neural network for mapping to obtain a current ice layer thickness measurement deviation dataset; merge the thickness data corresponding to the impurity consideration measurement point set in the current rough ice layer thickness dataset with the current ice layer thickness measurement deviation dataset to obtain a second current ice layer final thickness dataset;

[0127] S5, comparing the data after error correction in S4 with the thickness data measured by the traditional ice-chiseling method, and adjusting the ice thickness threshold for initial impurities based on the comparison result;

[0128] The S5 comprises the following steps:

[0129] S51, measuring each measuring point in the current ice layer thickness measurement point set using a traditional ice-chiseling method to obtain a third current ice layer final thickness data set;

[0130] S52: Setting a current measurement data error threshold; calculating a data difference between the third current ice layer final thickness data set, the second current ice layer final thickness data set, and the first current ice layer final thickness data set to obtain a current natural river ice layer measurement error value;

[0131] S53. When the current natural river channel ice layer measurement error value is greater than or equal to the current measurement data error threshold, adjusting the initial impurity-considered ice thickness threshold until the current natural river channel ice layer measurement error value is less than the current measurement data error threshold, thereby obtaining a final impurity-considered ice thickness threshold;

[0132] Adjusting the initial impurities in consideration of the ice thickness threshold in S53 includes the following steps:

[0133] S531, setting the initial impurity threshold value range considering ice thickness, and obtaining the impurity threshold value range considering ice thickness. , 、 They represent the lower limit and upper limit of the initial impurity considering the ice thickness threshold respectively; construct the impurity considering threshold adjustment honey badger population, and set the maximum number of iterations of the impurity considering threshold adjustment honey badger population to And the current number of iterations is , respectively recorded as the maximum number of iterations considering the threshold and the current number of iterations considering the threshold; the impurity threshold is considered to adjust the search space dimension of the honey badger population to 1;

[0134] S532, generating an impurity-considered threshold value according to the impurity-considered ice thickness threshold value interval to adjust the initial position of each honey badger in the honey badger population to obtain a second initial position set; the generation formula is as follows:

[0135] ;

[0136] Where, Indicates that the impurity threshold is taken into account to adjust the initial position of the jth honey badger in the honey badger population, Indicates that Generate a random number between 0 and 1;

[0137] S533, constructing the fitness function of the honey badger population by considering the impurity threshold ;as follows,

[0138] ;

[0139] Where, Indicates the current measurement error of natural river ice layer;

[0140] S534, start iteration, and set the current iteration number of the consideration threshold to 1 before iteration; during the first round of iteration, use the impurity consideration threshold to adjust the fitness function of the honey badger population to calculate the fitness value of the initial position of each honey badger in the second initial position set to obtain a third fitness value set; use the maximum fitness value in the third fitness value set and the initial position of the corresponding honey badger as the third global optimal fitness and the third global optimal position respectively; update the initial position of each honey badger in the second initial position set according to the third global optimal fitness and the third global optimal position; after the update is completed, add 1 to the current iteration number of the consideration threshold and enter the next round of iteration;

[0141] In each other round of iteration, the fitness function of the honey badger population adjusted by the impurity consideration threshold is used to calculate the fitness value of the position of each honey badger in the honey badger population adjusted by the impurity consideration threshold obtained in the previous round of iteration, and a fourth fitness value set is obtained; the maximum fitness value in the fourth fitness value set and the position of the corresponding honey badger are respectively used as the fourth global optimal fitness and the fourth global optimal position; the position of each honey badger in the honey badger population adjusted by the impurity consideration threshold obtained in the previous round of iteration is updated according to the fourth global optimal fitness and the fourth global optimal position; after the update is completed, the current iteration number of the consideration threshold is increased by 1 and the next round of iteration is entered;

[0142] S535, when When , stop the iteration and get the final global optimal fitness and the final global optimal position; otherwise, continue to iterate until when the inverse of the final global optimal fitness is less than the current measurement data error threshold, the final global optimal position is taken as the final impurity ice thickness threshold; otherwise, return to S534 to continue iteration.

[0143] Example 2

[0144] See also Figure 5 This embodiment discloses a millimeter-wave-based natural river ice thickness measurement system. The system can implement the method of the above embodiment and includes an ice layer easily miscible impurity type setting module, a known ice layer sample impurity content data acquisition module, a Bragg wavelength deviation data acquisition module, a known ice block comprehensive data cluster center calculation module, a Bragg wavelength deviation training and test data acquisition module, a known ice block training and test impurity content data acquisition module, a thickness deviation data mapping model construction module, a current natural river ice layer measurement module, and an impurity-considered ice thickness threshold comparison and adjustment module.

[0145] The ice layer easily mixed impurity type setting module sets several types of easily mixed impurity in the ice layer of natural rivers to obtain an ice layer easily mixed impurity type set;

[0146] The known ice layer sample impurity content data acquisition module acquires thickness data of a plurality of known ice samples and content data of various miscible impurities therein by setting an initial impurity consideration ice thickness threshold and according to the ice layer miscible impurity type set, thereby obtaining a known ice sample thickness data set and a known ice miscible impurity content data matrix;

[0147] The Bragg wavelength deviation data acquisition module measures the Bragg wavelength deviation data of each known ice cube sample in the known ice cube sample set to obtain a known ice cube Bragg wavelength deviation data set;

[0148] The known ice cube comprehensive data cluster center calculation module merges and preprocesses the known ice cube Bragg wavelength deviation dataset and the known ice cube miscible impurity content data matrix, then clusters the processed data by setting an initial neighborhood radius and an initial minimum sample number and calculates the cluster center to obtain the known ice cube comprehensive data cluster center matrix;

[0149] The Bragg wavelength deviation training and test data acquisition module sets the training sample ratio and collects the actual measured thickness data of the known ice training and test samples based on the known ice sample thickness dataset, and performs a subtraction operation between the actual measured thickness data and the standard thickness data to obtain the known ice thickness measurement deviation training dataset and the known ice thickness measurement deviation test dataset;

[0150] The known ice training and testing impurity content data acquisition module acquires corresponding impurity content data based on the known ice comprehensive data cluster center matrix and the Bragg wavelength deviation data of the known ice training and test samples, thereby obtaining a known ice training sample impurity content data matrix and a known ice test sample impurity content data matrix;

[0151] The thickness deviation data mapping model construction module uses a data matrix of impurity content of known ice training samples and a data matrix of impurity content of known ice test samples to train and test an initial multimodal adaptive convolutional neural network, and adjusts the initial neighborhood radius and the initial minimum number of samples during the testing process to obtain a final multimodal adaptive convolutional neural network and a final known ice comprehensive data clustering center matrix;

[0152] The current natural river ice layer measurement module collects preliminary ice layer thickness data, considers Bragg wavelength deviation for measurement points exceeding the initial impurity consideration ice thickness threshold, and performs error correction on the preliminary thickness data in combination with the final multimodal adaptive convolutional neural network and the final known ice block comprehensive data clustering center matrix to obtain a first current ice layer final thickness dataset and a second current ice layer final thickness dataset;

[0153] The impurity-considered ice thickness threshold comparison and adjustment module compares the first current ice layer final thickness dataset and the second current ice layer final thickness dataset with thickness data measured using a traditional ice-chiseling method, and adjusts the initial impurity-considered ice thickness threshold based on the comparison result.

[0154] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0155] The preferred embodiments of the invention disclosed above are intended only to help illustrate the invention. These preferred embodiments do not exhaust all details, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for measuring ice thickness in natural rivers based on millimeter waves, characterized in that: The following steps are involved: S1. Collecting data on the content of various easily miscible impurities in multiple known ice samples; Specifically, it includes: setting an initial impurity consideration ice thickness threshold and several types of easily miscible impurities in natural river ice, then collecting ice samples with known thickness greater than or equal to the initial impurity consideration ice thickness threshold and containing easily miscible impurities and corresponding thickness data, to obtain a known ice sample set and a known ice sample thickness dataset; S2, measuring the Bragg wavelength deviation data of multiple known ice samples, merging the data with the content data of various easily miscible impurities collected in S1, and performing preprocessing, then clustering the processed data and calculating the cluster centers to obtain a cluster center matrix of the comprehensive data of the known ice; S3. Based on the cluster center matrix of the known ice cube comprehensive data, obtain the impurity content data of the known ice cube training and test samples, train and test the initial multimodal adaptive convolutional neural network, and adjust the clustering process in S2 according to the test results; Specifically, the method includes: selecting training samples from a known ice sample set to obtain a known ice training sample set and a known ice test sample set; using a linear frequency modulated continuous wave millimeter wave radar to measure the thickness of each sample in the known ice training sample set and the known ice test sample set by transmitting a signal with a linearly varying frequency, to obtain an actual training data set of known ice thickness and an actual test data set of known ice thickness; S4: For measurement points exceeding the initial impurity ice thickness threshold, the Bragg wavelength deviation is considered and the error correction of the preliminary thickness data is performed using the multimodal adaptive convolutional neural network trained and tested in S3. S5. Compare the data after error correction in S4 with the thickness data measured by the traditional ice-chiseling method, and adjust the ice thickness threshold for initial impurities based on the comparison result.

2. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 1, characterized in that: Said S1 further comprises the following steps: S11. According to the known ice cube sample set, collect content data of various easily miscible impurities in each known ice cube sample to obtain a data matrix of the content of easily miscible impurities in the known ice cubes.

3. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 2, characterized in that: The S2 comprises the following steps: S21, using an optical fiber sensor to measure the Bragg wavelength deviation data of each known ice cube sample in the known ice cube sample set to obtain a known ice cube Bragg wavelength deviation data set; S22, merging each deviation data in the known ice cube Bragg wavelength deviation data set into the row data of the impurity content data corresponding to the known ice cube in the known ice cube miscible impurity content data matrix, to obtain a known ice cube comprehensive data matrix; S23. Perform Z-Score normalization processing on the data in the known ice cube comprehensive data matrix to obtain the processed known ice cube comprehensive data matrix; set the initial neighborhood radius and the initial minimum number of samples, and use the DBSCAN algorithm to perform clustering operation on the processed known ice cube comprehensive data matrix according to the initial neighborhood radius and the initial minimum number of samples. After clustering, calculate the center data of each cluster to obtain the known ice cube comprehensive data cluster center matrix.

4. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 3, characterized in that: The S3 further comprises the following steps: S31. Acquire the Bragg wavelength deviation data of each known ice training sample and each known ice test sample in combination with the known ice Bragg wavelength deviation dataset to obtain a known ice Bragg wavelength deviation training dataset and a known ice Bragg wavelength deviation test dataset. S32. Perform Z-Score normalization on the known ice cube Bragg wavelength deviation dataset to obtain a processed known ice cube Bragg wavelength deviation dataset; then, combine the known ice cube comprehensive data cluster center matrix to obtain impurity content data corresponding to the known ice cube training samples and the test samples, to obtain a known ice cube training sample impurity content data matrix and a known ice cube test sample impurity content data matrix; S33, respectively calculating the absolute value of the difference between the actual training dataset of known ice thickness and the actual test dataset of known ice thickness and the thickness data corresponding to each known ice training sample and known ice test sample in the known ice sample thickness dataset, to obtain a known ice thickness measurement deviation training dataset and a known ice thickness measurement deviation test dataset; S34. Constructing an initial multimodal adaptive convolutional neural network; training and testing the initial multimodal adaptive convolutional neural network using a data matrix of impurity content of known ice training samples and a training data set of known ice thickness measurement deviations to obtain test accuracy data; S35. When the test accuracy data is less than the test accuracy threshold, the initial neighborhood radius and the initial minimum number of samples are adjusted until the test accuracy data is greater than or equal to the test accuracy threshold, thereby obtaining the final known ice cube comprehensive data clustering center matrix and the final multimodal adaptive convolutional neural network.

5. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 4, characterized in that: In S32, the Euclidean distance calculation method is used to obtain the impurity content data of the corresponding known ice cube training samples and test samples in combination with the cluster center matrix of the known ice cube comprehensive data.

6. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 5, characterized in that: In S35 , the initial neighborhood radius and the initial minimum number of samples are adjusted using the honey badger optimization algorithm.

7. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 6, characterized in that: The S4 comprises the following steps: S41. Setting a natural river channel to be measured; selecting a plurality of ice thickness measurement points on the ice in the natural river channel to be measured to obtain a current ice thickness measurement point set; using the linear frequency modulated continuous wave millimeter wave radar in S33 to roughly measure the ice thickness at each current ice thickness measurement point by transmitting a signal with a linearly varying frequency, based on the current ice thickness measurement point set, to obtain a current rough ice thickness data set; S42: Using the current ice layer rough thickness data less than the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as the final thickness data of the corresponding measurement point, to obtain a first current ice layer final thickness data set; recording the measurement points corresponding to the current ice layer rough thickness data greater than or equal to the initial impurity-considered ice thickness threshold in the current ice layer rough thickness data set as an impurity-considered measurement point set; S43, measuring impurities by taking into account the Bragg wavelength deviation data of ice at each measurement point in the measurement point set to obtain a current Bragg wavelength deviation data set; obtaining impurity content data corresponding to each Bragg wavelength deviation data in the current Bragg wavelength deviation data set based on the final known ice block comprehensive data cluster center matrix to obtain a current impurity content data matrix; S44. Input each row of data in the current impurity content data matrix into the final multimodal adaptive convolutional neural network for mapping to obtain a current ice layer thickness measurement deviation dataset; merge the thickness data corresponding to the impurity consideration measurement point set in the current rough ice layer thickness dataset with the current ice layer thickness measurement deviation dataset to obtain a second current ice layer final thickness dataset.

8. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 7, characterized in that: The S5 comprises the following steps: S51, measuring each measuring point in the current ice layer thickness measurement point set using a traditional ice-chiseling method to obtain a third current ice layer final thickness data set; S52: Setting a current measurement data error threshold; calculating a data difference between the third current ice layer final thickness data set, the second current ice layer final thickness data set, and the first current ice layer final thickness data set to obtain a current natural river ice layer measurement error value; S53. When the current natural river channel ice layer measurement error value is greater than or equal to the current measurement data error threshold, the initial impurity consideration ice thickness threshold is adjusted until the current natural river channel ice layer measurement error value is less than the current measurement data error threshold, thereby obtaining the final impurity consideration ice thickness threshold.

9. The method for measuring ice thickness in natural rivers based on millimeter waves according to claim 8, characterized in that: Adjusting the initial impurities in consideration of the ice thickness threshold in S53 includes the following steps: S531, generating an impurity consideration threshold according to the value interval of the initial impurity consideration ice thickness threshold, and adjusting the initial position of each honey badger in the honey badger population to obtain a second initial position set; S532, constructing the fitness function of the honey badger population by adjusting the impurity threshold; S533, adjusting the fitness function of the honey badger population according to the impurity consideration threshold to iterate and update the second initial position set; S534. Repeat S533. When the maximum number of iterations is reached, stop the iteration to obtain the final global optimal fitness and the final global optimal position. When the inverse of the final global optimal fitness is less than the current measurement data error threshold, take the final global optimal position as the final impurity ice thickness threshold, and the adjustment is completed.

10. A millimeter wave-based natural river ice thickness measurement system, characterized by: Used to implement a method for measuring ice thickness in natural rivers based on millimeter waves as described in any one of claims 1-9.

Images