Evaluation method and device for embankment cracks

Abstract

The present invention provides a method and device for evaluating embankment cracks, which can be applied to the field of geotechnical engineering. The evaluation method includes: obtaining initial information to be evaluated corresponding to cracks on the embankment surface; processing the initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment; the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to threshold quantize the decomposed initial information to be evaluated; and inputting the soil information, geometric information, and environmental information into an evaluation model to obtain size and rate evaluation results of the cracks, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering, and in particular to a method and device for evaluating embankment cracks. Background Art

[0002] Embankment cracks can cause numerous problems. For example, in terms of road structure, cracks can weaken the overall strength of the roadbed and pavement, leading to a decrease in the pavement's bearing capacity and accelerated damage to the pavement structure. Regarding traffic safety, cracks can reduce the smoothness of the pavement, affecting vehicle comfort and stability. Cracks also provide infiltration channels for rain and snow, exacerbating the softening and erosion of the roadbed soil, increasing road maintenance costs and difficulty. Related technologies for predicting and simulating embankment cracks include: using finite element methods to simulate the stress distribution and deformation of embankments under different conditions to predict the location and expansion trends of cracks; using deep learning-based networks (such as Transformers) to analyze and segment crack images and predict the location and morphology of cracks in future time periods; and installing sensors (such as strain gauges and displacement meters) on the embankment to monitor the road surface in real time.

[0003] However, numerical simulation methods require the construction of relatively accurate mathematical models, and the accuracy of the simulation results depends on the accuracy of the input parameters; traditional deep learning models lack appropriate preprocessing for multi-source measurement data during prediction, and the measurement data are mostly non-stationary signals, with the possibility of losing key features, resulting in low accuracy of the prediction results; the real-time monitoring method of the equipment is greatly affected by external environmental factors (such as temperature and humidity) and the equipment cost is relatively high. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method, device, equipment, medium and program product for evaluating embankment cracks.

[0005] According to a first aspect of the present invention, a method for evaluating embankment cracks is provided, comprising: obtaining initial information to be evaluated corresponding to cracks on the embankment surface; processing the initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; inputting the soil information, geometric information, and environmental information into an evaluation model to obtain size evaluation results and rate evaluation results of the cracks, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel.

[0006] The second aspect of the present invention provides an evaluation device for embankment cracks, comprising: an information acquisition module for acquiring initial information to be evaluated corresponding to cracks on the embankment surface; an information processing module for processing the initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information and environmental information corresponding to the embankment based on different types, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to threshold quantize the decomposed initial information to be evaluated; an information input module for inputting soil information, geometric information and environmental information into an evaluation model to obtain size evaluation results and rate evaluation results of the cracks, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel based on different functions.

[0007] A third aspect of the present invention provides an electronic device, comprising: one or more processors; a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the above method.

[0008] The fourth aspect of the present invention further provides a computer-readable storage medium having a computer program or instructions stored thereon, which implements the steps of the above method when the computer program or instructions are executed by a processor.

[0009] The fifth aspect of the present invention further provides a computer program product, comprising a computer program or instructions, which implement the steps of the above method when executed by a processor.

[0010] According to embodiments of the present invention, the information to be evaluated, processed through a decomposition strategy and preset thresholds, retains the key features of the original information, improving the signal-to-noise ratio and data quality. Because the evaluation model integrates features and state evaluation functions for multi-source and time series information, it extracts key features from these multi-source information to capture long-term trends, enabling the processing of nonlinear, time-varying characteristics of embankment cracks. This allows for future crack size and rate assessments, improving the rationality and reliability of crack assessments. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0012] Figure 1 A diagram showing an application scenario of the method, device, equipment, medium, and program product for evaluating embankment cracks according to an embodiment of the present invention is shown;

[0013] Figure 2 A flow chart showing a method for evaluating embankment cracks according to an embodiment of the present invention is shown;

[0014] Figure 3A An example schematic diagram of an initial evaluation model training process according to an embodiment of the present invention is shown;

[0015] Figure 3B A schematic diagram of the structure of a CNN layer according to an embodiment of the present invention is shown;

[0016] Figure 3C A schematic diagram of the structure of a BiLSTM layer according to an embodiment of the present invention is shown;

[0017] Figure 4A A three-dimensional schematic diagram of widening a longitudinal crack in an embankment according to an embodiment of the present invention is shown;

[0018] Figure 4B A comparison chart of the measured and predicted widths of cracks in a widened embankment according to an embodiment of the present invention;

[0019] Figure 5 It shows a structural block diagram of an embankment crack assessment device according to an embodiment of the present invention;

[0020] Figure 6 A block diagram of an electronic device suitable for implementing a method for evaluating embankment cracks according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0021] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present invention. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of embodiments of the present invention. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concept of the present invention.

[0022] The terms used herein are only for describing specific embodiments and are not intended to limit the present invention. The terms "comprise", "include", etc. used herein indicate the presence of the features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0023] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0024] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0025] In related technologies, numerical simulation methods require the construction of relatively accurate mathematical models, and the accuracy of the simulation results depends on the accuracy of the input parameters; traditional deep learning models lack appropriate preprocessing for multi-source measurement data during prediction, and the measurement data are mostly non-stationary signals, with the possibility of losing key features, resulting in low accuracy of prediction results; the real-time monitoring method of equipment is greatly affected by external environmental factors (such as temperature and humidity) and the equipment cost is relatively high.

[0026] In light of this, the present invention, through a decomposition strategy and pre-set thresholds, preserves the key features of the original information after processing, improving the signal-to-noise ratio and data quality. Because the assessment model integrates the evaluation of the characteristics and states of multi-source and time series information, it extracts key features from these multi-source information to capture long-term trends, enabling the processing of the nonlinear, time-varying characteristics of embankment cracks. This allows for the generation of future crack size and rate estimates, enhancing the rationality and reliability of crack assessment.

[0027] An embodiment of the present invention provides a method for evaluating embankment cracks, comprising: obtaining initial information to be evaluated corresponding to cracks on the embankment surface; processing the initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; inputting the soil information, geometric information, and environmental information into an evaluation model to obtain size evaluation results and rate evaluation results of the cracks, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel.

[0028] Figure 1 A diagram illustrating an application scenario of a method, apparatus, device, medium, and program product for evaluating embankment cracks according to an embodiment of the present invention is shown.

[0029] like Figure 1As shown, the application scenario 100 according to this embodiment may include a terminal device 101, a network 102, a server 103, and a data acquisition device 104. The network 102 is used as a medium for providing a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0030] Data acquisition equipment 104 can be used to collect embankment geometry, soil, and environmental information. This includes, but is not limited to, surveying drones, temperature sensors, humidity sensors, and earth pressure sensors. For example, a drone equipped with a high-precision camera can be used to take aerial photos of the embankment, generating its geometry and a three-dimensional model using photogrammetry. For example, temperature and humidity sensors can be used to monitor soil temperature and humidity in real time, including their temporal variations. For example, earth pressure sensors can be used to monitor the total pressure at the interface between the embankment and the soil, enabling assessment of the soil's stress state and stability.

[0031] The data collection device 104 can send the collected information about the embankment to the server 103. The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages.

[0032] The terminal device 101 may be any electronic device having a display screen and supporting web browsing, including but not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, and the like.

[0033] The server 103 may be a server that provides various services. The server 103 may process the received relevant information of the embankment (such as soil information, geometric information, and environmental information), obtain the processing results of the cracks on the embankment surface, and send the processing results to the terminal device 101.

[0034] It should be noted that the embankment crack assessment method provided in the embodiments of the present invention can generally be executed by the server 103. Accordingly, the embankment crack assessment device provided in the embodiments of the present invention can generally be located in the server 103. The embankment crack assessment method provided in the embodiments of the present invention can also be executed by a server or server cluster that is different from the server 103 and that is capable of communicating with the terminal device 101 and / or the server 103. Accordingly, the embankment crack assessment device provided in the embodiments of the present invention can also be located in a server or server cluster that is different from the server 103 and that is capable of communicating with the terminal device 101 and / or the server 103.

[0035] It should be understood that Figure 1The number of terminal devices, networks, data collection devices and servers in the embodiment is merely illustrative. Any number of terminal devices, data collection devices, networks and servers may be provided as required.

[0036] Figure 2 A flow chart of a method for evaluating embankment cracks according to an embodiment of the present invention is shown.

[0037] like Figure 2 As shown, the method for evaluating embankment cracks in this embodiment includes operations S210 to S230.

[0038] In operation S210 , initial information to be evaluated corresponding to cracks on the embankment surface is obtained.

[0039] In embodiments of the present invention, the embankment may be a completed and in-service embankment, including at least one of an existing embankment and a newly constructed embankment. The initial information to be evaluated can be used to assess the development trend and status of existing cracks on the embankment surface, using multi-source information, or to assess the probability of cracks appearing on the embankment surface in the future. The method and equipment for obtaining the initial information to be evaluated are based on the comprehensiveness and accuracy of the information obtained and are not limited herein.

[0040] In operation S220, the initial information to be evaluated is processed based on the decomposition strategy and the preset threshold to obtain the information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment. The decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated.

[0041] In embodiments of the present invention, a decomposition strategy can be used to decompose the acquired initial information to be evaluated to obtain decomposition coefficients in multiple dimensions. A preset threshold can represent a target threshold selected for quantizing the decomposed initial information to be evaluated to remove signal noise. Geometric information can include width and height information of the embankment in different dimensions. Soil information can include state and mechanical information of the embankment soil. Environmental information can include external environmental information and settlement and deformation information corresponding to the embankment.

[0042] For example, the geometric information, soil information and environmental information of the embankment are decomposed and processed to obtain the multi-dimensional width and height information of the embankment, the state information and mechanical information of the land, the external environment information corresponding to the embankment, and the settlement deformation information.

[0043] In operation S230 , soil information, geometric information, and environmental information are input into an evaluation model to obtain a size evaluation result and a rate evaluation result of the crack, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel.

[0044] In an embodiment of the present invention, the feature assessment sub-model can be used to extract key spatial relationship information from multi-source information, namely time series information. The state assessment sub-model can be used to process the extracted time series information to obtain long-term dependencies between the time series information, thereby obtaining an assessment of the size of the crack in the future time period and the assessment of the crack growth rate. The size assessment results can include information on the width, length, and depth of the crack.

[0045] For example, soil information, geometric information, and environmental information are input into the feature evaluation sub-model, and the key spatial relationship information of the time series is output. The key spatial relationship information is then input into the state assessment sub-model as input information to obtain the size assessment results and rate assessment results of the cracks.

[0046] According to embodiments of the present invention, the information to be evaluated, processed through a decomposition strategy and preset thresholds, retains the key features of the original information, improving the signal-to-noise ratio and data quality. Because the evaluation model integrates features and state evaluation functions for multi-source and time series information, it extracts key features from these multi-source information to capture long-term trends, enabling the processing of nonlinear, time-varying characteristics of embankment cracks. This allows for future crack size and rate assessments, improving the rationality and reliability of crack assessments.

[0047] It can be understood that the above has described how to determine the size evaluation results and rate evaluation results of the cracks. The following will describe how to obtain the information to be evaluated.

[0048] According to an embodiment of the present invention, the decomposition strategy includes a decomposition function and a decomposition dimension;

[0049] The initial information to be evaluated is processed based on a decomposition strategy and a preset threshold to obtain the information to be evaluated, including: decomposing the initial information to be evaluated using a decomposition function and a decomposition dimension to obtain decomposition coefficients of multiple frequencies corresponding to the initial information to be evaluated; threshold quantizing the decomposition coefficients of the multiple frequencies based on a preset threshold to obtain processing coefficients of the multiple frequencies; and fusing the processing coefficients of the multiple frequencies based on an inverse transformation function to obtain the information to be evaluated.

[0050] In embodiments of the present invention, a decomposition function can be used to decompose multi-source information to be evaluated to obtain functions with different scale characteristics. The decomposition dimension is determined based on the decomposition function and actual requirements, and is proportional to the frequency resolution of the signal. The decomposition coefficients can be high-frequency wavelet coefficients and / or low-frequency wavelet coefficients obtained by the decomposition function.

[0051] In an embodiment of the present invention, the preset threshold may be a threshold set during the quantization operation of high-frequency wavelet coefficients or low-frequency wavelet coefficients. Coefficients less than the preset threshold may be reset to zero or subjected to other forms of quantization operations to remove or reduce noise components while retaining important features in the signal. The processed coefficient may be the result obtained after threshold quantization processing. The inverse transform function may be a stepwise combination of low-frequency coefficients and high-frequency coefficients through a reconstruction process of the coefficients subjected to multi-level decomposition, thereby gradually recovering approximate information corresponding to the original information to obtain complete information to be evaluated.

[0052] Related technologies lack comprehensive analysis of basic information on embankment cracks at different scales, both at the macro and micro levels. This results in poor adaptability and low accuracy in the analysis process. This application decomposes and reconstructs the acquired information to be evaluated using a decomposition function, removing or suppressing noise components.

[0053] For example, to address the symmetrical cracks that often occur during embankment widening, the Symlets wavelet algorithm and a decomposition scale (determined based on actual needs, typically 3-5 layers) can be used to perform wavelet decomposition on the preprocessed soil information, breaking the signal into subband signals of different frequency bands. This decomposition process better preserves the symmetry of the cracks, facilitating a more accurate analysis of their shape and distribution.

[0054] For example, after performing a multi-level decomposition on the initial information to be evaluated and obtaining processing coefficients of multiple frequencies, the processing coefficients can be reconstructed in a corresponding multi-level manner through an inverse transformation function, and the processed low-frequency coefficients and high-frequency coefficients can be gradually combined to restore an approximate version of the original signal step by step until a complete reconstructed signal is obtained.

[0055] According to the embodiments of the present invention, the decomposition function removes noise while retaining effective crack information, making crack features more prominent, improving the detection algorithm's robustness to noise, and enabling stable crack detection in complex environments. By comprehensively analyzing crack characteristics and their changing patterns at different scales, a richer information base is provided for crack assessment and prediction. For example, by analyzing crack geometric characteristics at different scales, crack development trends and potential risks can be more accurately determined.

[0056] According to an embodiment of the present invention, the method further includes: determining a decomposition function and a decomposition dimension according to at least one of a length feature and a frequency feature of the initial information to be evaluated.

[0057] In an embodiment of the present invention, the decomposition dimension may be determined by the signal length and high and low frequencies corresponding to the acquired information to be evaluated. The specific dimension may be determined according to actual conditions or requirements and is not limited here.

[0058] For example, when processing embankment crack information, if the focus is on the overall crack trend, such as the crack direction and general distribution, a lower resolution scale (e.g., 1-2 layers) can be selected. Conversely, if the focus is on analyzing detailed crack features, such as crack microtexture and fine branching, a higher resolution scale (e.g., 4-6 layers) can be selected.

[0059] According to an embodiment of the present invention, soil information, geometric information and environmental information are input into an evaluation model to obtain size evaluation results and rate evaluation results of cracks, including: inputting soil information, geometric information and environmental information into a feature evaluation sub-model to output feature evaluation results; inputting feature evaluation results into a state evaluation sub-model to output size evaluation results and rate evaluation results.

[0060] In an embodiment of the present invention, the feature evaluation sub-model can be a deep learning model comprising multiple network layers for extracting multi-source initial information to be evaluated. The state evaluation sub-model can be an evaluation model based on a gating algorithm and a fusion strategy, comprising long short-term memory network modules in different directions.

[0061] For example, soil information, geometric information, and environmental information are input into a trained convolutional neural network model to extract the key spatial relationship information of multi-element data and obtain the feature evaluation results of the time series. The feature evaluation results are then input into a trained bidirectional long short-term memory network. By simultaneously processing the forward and backward information in the time series information, the contextual information from front to back and from back to front in the sequence are determined respectively. The contextual information from front to back and the contextual information from back to front are then fused at each time step to obtain the final output representation, namely the size evaluation results and rate evaluation results of the cracks.

[0062] According to an embodiment of the present invention, by processing the information to be evaluated through an evaluation model that integrates a feature evaluation sub-model and a state evaluation sub-model, it is possible to extract multi-dimensional key feature information from multi-source data using the feature evaluation sub-model, capture the spatial distribution, morphology, location and other characteristics of cracks, as well as the mutual influence between different locations, and provide rich spatial details for prediction. For processing time series data, the evolution patterns and trends of embankment cracks in the time dimension can be captured, and the dependencies between the front and back in the time series can be integrated. The fused evaluation model can organically integrate spatial features and temporal features, and can more comprehensively and accurately model the development process of embankment cracks, thereby improving prediction accuracy, more accurately predicting the expansion direction, speed and extent of cracks, and providing a reliable basis for timely and effective maintenance measures.

[0063] According to an embodiment of the present invention, a feature evaluation sub-model includes an extraction layer, a conversion layer and an update layer; soil information, geometric information and environmental information are input into the feature evaluation sub-model, and a feature evaluation result is output, including: inputting soil information, geometric information and environmental information into the extraction layer for feature extraction to obtain initial feature information; inputting the initial feature information into the conversion layer for nonlinear conversion to obtain intermediate feature information; inputting the intermediate feature information into the update layer for dimensionality reduction and feature enhancement processing to obtain multidimensional feature information; and performing conversion processing on the multidimensional feature information to obtain a feature evaluation result.

[0064] In an embodiment of the present invention, the extraction layer can be used to extract features from multidimensional information to be evaluated to obtain initial feature information; the conversion layer can be used to transform the extracted initial feature information by introducing nonlinear factors to fit more complex relationships in the data to obtain intermediate feature information; and then the intermediate feature information obtained is reduced in dimension, compressed, or enhanced using the updated layer height to retain key feature information and obtain key multidimensional feature information.

[0065] In an embodiment of the present invention, considering that the information format of the multidimensional feature information obtained by the update layer cannot be processed by the state assessment sub-model, the multidimensional feature information can be subjected to dimensionality reduction transformation to obtain the reduced-dimensional feature assessment result.

[0066] For example, consider a convolutional neural network (CNN) as the feature evaluation submodel and a bidirectional long short-term memory (LSTM) as the state evaluation submodel. The feature information output by the feature evaluation submodel is multidimensional and has a spatial structure, such as height, width, and number of channels. However, the input information received by the state evaluation submodel is in sequence form, namely, a one-dimensional vector sequence. To input the features extracted by the CNN into the bidirectional LSTM for subsequent time series modeling, a flattening layer is used to convert the high-dimensional features into a sequence format suitable for the LSTM input.

[0067] According to an embodiment of the present invention, by converting the multidimensional feature information, the input vector of each time step contains the comprehensive features of the information to be evaluated at that position, so that the state assessment sub-model can better capture the front-end dependency relationship in the time series information, thereby improving the accuracy of the model in predicting the development of embankment cracks.

[0068] According to an embodiment of the present invention, multidimensional feature information is converted and processed to obtain a feature evaluation result, including: reducing the dimension of the multidimensional feature information based on the time step or feature dimension to obtain multiple reduced-dimensional features; and fusing the multiple reduced-dimensional features to obtain a feature evaluation result.

[0069] In the embodiment of the present invention, dimensionality reduction processing may refer to flattening multi-dimensional feature information to obtain reduced-dimensional features. Feature fusion may refer to further integrating and transforming the reduced-dimensional features after the flattening operation to extract higher-level feature representations.

[0070] For example, consider receiving multidimensional feature information output by a feature evaluation submodel, which has a certain height, width, and number of channels. For example, assuming the feature information output by the feature evaluation submodel has a size of [height, width, number of channels] = [h, w, c], the feature vector at each position can be flattened, converting the three-dimensional feature information into a two-dimensional matrix. For example, the feature information can be flattened along the height and width dimensions to obtain a matrix of shape [h × w, c], where each row represents a feature vector at a position. Alternatively, the feature information can be flattened along the number of channels dimension to obtain a vector sequence of shape [h × w × c, 1].

[0071] After flattening, the feature vectors can be further integrated and transformed using fully connected layers. Fully connected layers map each feature vector to a new feature space by learning weight matrices and bias terms to extract higher-level feature representations. The output dimension of the fully connected layer can be set based on the input requirements of the bidirectional long short-term memory network. For example, it can be set to the input dimension of the bidirectional long short-term memory network at each time step.

[0072] An embodiment of the present invention inputs feature evaluation results into a state evaluation sub-model and outputs size evaluation results and rate evaluation results, including: processing the feature evaluation results based on a forward propagation strategy to obtain first state information corresponding to multiple moments, and processing the feature evaluation results based on a backward propagation strategy to obtain second state information corresponding to multiple moments; splicing the first state information and the second state information to obtain target state information; and performing pooling processing on the target state information to obtain a size evaluation result and a rate evaluation result.

[0073] In an embodiment of the present invention, the first state information may represent state information obtained by sequentially processing the feature evaluation results backward from the first time step of a plurality of time steps using a forward propagation layer in the state evaluation submodel. The second state information may represent state information obtained by sequentially processing the feature evaluation results forward from the last time step of a plurality of time steps using a backward propagation layer in the state evaluation submodel. The target state information may be fused state information obtained by concatenating or adding the first state information and the second state information. Pooling processing may represent global pooling processing performed on the fused target state information.

[0074] For example, the forward LSTM network in the bidirectional LSTM network can start from the first time step of the sequence and process the features of each time step in sequence backwards; the backward LSTM network can start from the last time step of the sequence and process the features of each time step in sequence forwards. Through bidirectional processing, the state assessment sub-model can capture the dependencies and evolution trends of cracks in the temporal dimension. The bidirectional LSTM network then outputs a hidden state for each time step, which incorporates the temporal information of the crack in both the preceding and following directions at that time step.

[0075] Furthermore, the temporal features output by the bidirectional long-short-term memory network can be fused with the spatial features extracted by the convolutional neural network. These two features can be combined through methods such as concatenation and addition to form the final target state information. This target state information is then globally pooled to obtain the scale and rate assessment results of the crack.

[0076] In an embodiment of the present invention, the embankment includes an existing embankment and a newly built embankment, the geometric information includes first geometric information of the existing embankment and second geometric information of the newly built embankment; the soil information includes density information, water content information, compression modulus information, internal friction angle and cohesion information; and the environmental information includes temperature information, humidity information, rainfall information and settlement deformation information.

[0077] In an embodiment of the present invention, the first geometric information may include the width, height, and roadbed depth of an existing embankment, and the second geometric information may include the width, height, and roadbed depth of a newly constructed embankment. The environmental information may also include traffic flow information on the embankment.

[0078] For example, within a specific time frame, measuring equipment is used to accurately measure the settlement and deformation data of the embankment. Targeted measurements can be made at key locations such as the junction of the existing and newly built embankments, the embankment slopes, and the bottom of the embankment. Crack observation instruments are used to record key information such as the width and length of cracks in detail, and ultrasonic methods are used to measure the crack depth. Temperature and humidity sensors, rain gauges, and other equipment are deployed to collect real-time data on environmental factors such as temperature, humidity, and rainfall at the site and surrounding environment, and obtain traffic flow information from relevant databases. Detailed construction drawings can also be collected to obtain construction process-related parameters and the physical and mechanical parameters of the roadbed soil.

[0079] In an embodiment of the present invention, an evaluation model is trained by the following operations: obtaining initial sample information to be evaluated corresponding to cracks on the surface of a sample embankment; processing the initial sample information to be evaluated based on a decomposition strategy and a preset threshold to obtain sample information to be evaluated, wherein the sample information to be evaluated includes sample geometric information of the sample embankment, sample soil information, sample crack information, and sample environmental information corresponding to the sample embankment based on different types, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; using the sample crack information, sample geometric information, sample soil information, and sample environmental information to train the initial evaluation model to obtain an evaluation model that meets the preset evaluation conditions.

[0080] In an embodiment of the present invention, the sample initial information to be evaluated is processed based on the decomposition strategy and the preset threshold to obtain the sample information to be evaluated. This is the same as the process of processing the initial information to be evaluated based on the decomposition strategy and the preset threshold to obtain the information to be evaluated in the application stage, and will not be repeated here.

[0081] In an embodiment of the present invention, considering the influence of different data dimensions of multi-source information, the sample information to be evaluated can be normalized to eliminate the influence of multi-source data dimensions and improve model training efficiency. The Min-Max normalization method is used to scale the eigenvalues ​​to between 0 and 1, as shown in the following formula (1):

[0082] (1);

[0083] Among them, X' can be the normalized data; X can be the original data; X min Can be the minimum value in the data; X max It can be the maximum value in the data.

[0084] For example, the initial evaluation model can include an initial feature evaluation submodel and an initial state evaluation submodel. For example, if the initial feature evaluation submodel and the initial state evaluation submodel are hybrid models of a convolutional neural network (CNN) and a bidirectional long short-term memory (BiLSTM), the initial feature evaluation submodel and the initial state evaluation submodel can be considered the CNN layer and the BiLSTM layer of the initial evaluation model.

[0085] Figure 3A An example schematic diagram of an initial evaluation model training process according to an embodiment of the present invention is shown.

[0086] like Figure 3AAs shown, the sample information to be evaluated can be divided into a training set 31 and a test set 32. The initial evaluation model can include an input layer 301, a CNN layer 302, a flattening layer 303, a BiLSTM layer 304, and an output layer 305. The CNN layer 302 can be composed of a stack of convolutional layers 3021, activation function layers 3022, and max pooling layers 3023. The BiLSTM layer 304 can be composed of a forward LSTM layer 3041, a dropout layer 3042, a backward LSTM 3043, and a fully connected layer 3044. The CNN layer 302 can be used to extract key spatial relationship information from multi-source data, and the BiLSTM layer 304 can be used to obtain long-term dependency information related to time series.

[0087] The data in the training set 31 are serialized by time step and input into the input layer 301. Each time step contains all the eigenvalue information. The convolution layer 3021 in the CNN layer 302 is then used to perform a convolution operation on the input data to extract the spatial relationship between the widened embankment parameters and the generated cracks to generate spatial features.

[0088] Figure 3B A structural diagram of a CNN layer according to an embodiment of the present invention is shown.

[0089] like Figure 3B As shown, the convolution layer 3021 performs a convolution operation on the input sample information to be evaluated. The convolution kernel can be set to 2×2. The data is processed by sliding the window to obtain the spatial relationship between the widened embankment parameters and the generated cracks, and a feature matrix (feature map) is generated. Setting two convolution layers can further combine and refine the key features of the processed data, as shown in the following formula (2):

[0090] (2);

[0091] in,

[0092] A i+m,j+n An element in the matrix can be input, whose position is determined by the current output position (i, j) and the position of the convolution kernel (n, m); f cov Can be an activation function; w m,n Can be the weight of the convolution kernel in the mth row and nth column; b m,n It can be the bias parameter of the convolution kernel in the mth row and nth column; i, j can be the output feature map position, It can be an element in the output matrix or feature map, located in the i-th row and j-th column, and Σ can be a summation symbol, indicating the weighted summation of the elements in the area covered by the convolution kernel.

[0093] The activation function layer can use the ReLu activation function to activate the extracted spatial features and add nonlinearity; the maximum pooling layer 3023 can perform maximum pooling dimensionality reduction operation on the feature map generated by the convolution layer 3021, and the pooling window is set to 2×2 with a step size of 2. The maximum value is taken in the local area of ​​the data to reduce the dimension of the data and enhance the robustness of the model, while retaining the key features in the data and generating several feature maps; the flattening layer 303 operates in the form of a fully connected layer to integrate the feature vectors generated by the CNN layer 302 to adapt to the input form of the BiLSTM layer 304.

[0094] Figure 3C A schematic structural diagram of a BiLSTM layer according to an embodiment of the present invention is shown.

[0095] like Figure 3C As shown, the results of the flattening layer 303 can be arranged in chronological order as input 3031, and the forward LSTM layer 3041 and the backward LSTM 3043 are operated to simultaneously learn the forward and reverse information of the crack features in the time dimension, and the forward and reverse results are combined to obtain the output result 3032; it is set as two layers, and a Dropout layer 3042 is added between the two layers, and the dropout rate is set to 0.2 to improve the generalization ability of the model, improve the model performance, prevent overfitting problems, and strengthen robustness; the calculation formula of the BiLSTM layer can be shown as follows (3) to (8):

[0096] (3);

[0097] (4);

[0098] (5);

[0099] (6);

[0100] (7);

[0101] (8);

[0102] Among them, x t is the input data; i t 、f t 、o t are the values ​​of input gate, forget gate, and output gate respectively; c t is the unit state; h t is the hidden state; σ is the Sigmoid function; ⊙ is the element-by-element multiplication; w is the weight; and b is the bias parameter.

[0103] like Figure 3AAs shown, the fully connected layer 3044 can be connected to the output layer 305 to set as a multi-output result, outputting the width, length, depth and development rate of the crack respectively.

[0104] The training set can be used to perform iterative training and hyperparameter optimization on the initial evaluation model; the hyperparameter optimization can adopt the Adam algorithm 308, which combines the advantages of the Adagrad algorithm and the RMSProp algorithm, and dynamically adjusts the learning rate of each parameter based on the first-order moment estimation and second-order moment estimation of the gradient. The learning rate can be dynamically adjusted according to the gradient history information of each parameter, giving a larger learning rate for parameter updates with lower frequency and a smaller learning rate for parameter updates with higher frequency, so as to converge to the optimal solution faster during the training process.

[0105] For example, the learning rate of the Adam algorithm 308 can be initialized to 0.001, the first-order moment estimate β1=0.8, the second-order moment estimate β2=0.999, and the constant term ε=e- 8 ; Forward propagation, for a given input data, it can be input into the constructed initial evaluation model, and calculation is performed according to the structure and parameters of the neural network, and the initial prediction result 306 of the network is obtained by sequentially passing through the input layer 301, CNN layer 302, flattening layer 303, BiLSTM layer 304 and output layer 305; the loss function 307 is calculated based on the prediction result and the true value, and the mean square error is used as the loss function to evaluate the neural network; then back propagation is performed, and the gradient of the loss function with respect to each parameter is calculated using the chain rule, as shown in the following formulas (9) to (11):

[0106] (9);

[0107] (10);

[0108] (11);

[0109] Among them, MSE(θ) can be the loss function; n can be the number of samples; y i Can be the true value of the i-th sample; Can be the corresponding predicted value; the gradient vector of the loss function with respect to the parameter θ is the partial derivative of the loss function with respect to the parameter Indicates; m can be the total number of parameters.

[0110] The Adam algorithm 308 can be used to update parameters according to the rules based on the calculated gradient to minimize the loss function; the above operation steps are repeated until all samples are processed, and the prediction accuracy of the model is determined 33. If the prediction accuracy 33 meets the requirements, the model is determined as the evaluation model 30, otherwise the parameters are adjusted to continue optimization; the performance and accuracy of the model are evaluated using the test set 32, and the evaluation parameter is the determination coefficient (R 2 ), the closer the coefficient of determination is to 1, the better the model prediction performance is, and the closer it is to 0, the worse the model prediction effect is. 2 The calculation method is shown in the following formula (12):

[0111] (12);

[0112] Among them, y i It can be the true value of the i-th sample. It can be the corresponding predicted value, that is, the value obtained by predicting each sample point according to the regression model. It can be the average of the observations, which is used to measure the central location of the data. The residual sum of squares can be expressed as the sum of squares of the differences between the model predicted values ​​and the actual observed values, reflecting the error of the model fitting data. The total sum of squared deviations can be expressed as the sum of squares of the differences between the actual observed values ​​and the mean, reflecting the variability of the data itself.

[0113] Figure 4A A three-dimensional schematic diagram of widening a longitudinal crack in an embankment according to an embodiment of the present invention is shown.

[0114] like Figure 4A As shown, a case study of a road embankment widening in City A was selected for verification. The widened embankment includes an existing embankment 401 and a newly built embankment 402. Multiple longitudinal cracks 404 are generated along the junction 403 of the refueling embankment and the newly built embankment.

[0115] Figure 4B This is a comparison chart of the measured and predicted values ​​of the crack width of the widened embankment according to an embodiment of the present invention.

[0116] The soil information, geometric information, and environmental information obtained by chronologically arranging a section of the widened embankment in City A are used as input to the constructed CNN-BiLSTM model. After a series of iterative training and hyperparameter optimization, the evaluation value obtained using the intermediate evaluation model is obtained. Figure 4B As shown in the figure, the predicted value is compared with the measured value obtained by the actual detection equipment to obtain the comparison result, and the determination coefficient (R 2) is 0.98833, indicating that the prediction results of the trained model can meet the actual prediction accuracy requirements, and at the same time verify the rationality and feasibility of the crack assessment method.

[0117] Based on the above-mentioned evaluation method for embankment cracks, the present invention also provides an evaluation device for embankment cracks. Figure 5 The device is described in detail.

[0118] Figure 5 A structural block diagram of an embankment crack assessment device according to an embodiment of the present invention is shown.

[0119] like Figure 5 As shown, the embankment crack assessment device of this embodiment includes an information acquisition module 510 , an information processing module 520 and an information input module 530 .

[0120] The information acquisition module 510 is used to acquire initial information to be evaluated corresponding to cracks on the embankment surface. In one embodiment, the information acquisition module 510 can be used to perform the operation S210 described above, which will not be described in detail here.

[0121] Information processing module 520 is configured to process the initial information to be evaluated based on a decomposition strategy and a preset threshold value to obtain information to be evaluated. The information to be evaluated may include embankment geometry information, soil information, and environmental information corresponding to the embankment, depending on its type. The decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold value is used to perform threshold quantization on the decomposed initial information to be evaluated. In one embodiment, information processing module 520 may be configured to perform operation S220 described above, which will not be further described here.

[0122] Information input module 530 is used to input soil information, geometric information, and environmental information into the assessment model to obtain crack size and rate assessment results. The assessment model includes a feature assessment submodel and a state assessment submodel based on different functions. In one embodiment, information input module 530 can be used to perform operation S230 described above and will not be further described here.

[0123] According to an embodiment of the present invention, the information acquisition module 510, information processing module 520, and information input module 530 in the embankment crack assessment device retain the key features of the original information after processing using a decomposition strategy and preset thresholds, thereby improving the signal-to-noise ratio and data quality. Because the assessment model integrates features and state evaluation functions for multi-source information and time series information, it extracts key features from multi-source information to capture long-term trends, enabling the processing of nonlinear, time-varying characteristics of embankment cracks. This allows for the generation of future crack size and rate assessment results, thereby enhancing the rationality and reliability of crack assessment.

[0124] According to an embodiment of the present invention, the decomposition strategy includes a decomposition function and a decomposition dimension; the information processing module 520 includes: an information processing module 520, a quantization submodule, and a coefficient fusion submodule. The information decomposition submodule is configured to decompose the initial information to be evaluated using the decomposition function and the decomposition dimension to obtain decomposition coefficients of multiple frequencies corresponding to the initial information to be evaluated; the quantization submodule is configured to perform threshold quantization on the decomposition coefficients of the multiple frequencies based on a preset threshold to obtain processing coefficients for the multiple frequencies; and the coefficient fusion submodule is configured to fuse the processing coefficients of the multiple frequencies based on an inverse transformation function to obtain the information to be evaluated.

[0125] According to an embodiment of the present invention, the apparatus further comprises: a dimension determination module, configured to determine a decomposition function and a decomposition dimension according to at least one of a length feature and a frequency feature of the initial information to be evaluated.

[0126] According to an embodiment of the present invention, the information input module 530 includes a result output submodule and a result input submodule. The result output submodule is used to input soil information, geometric information, and environmental information into the feature assessment submodel and output feature assessment results. The result input submodule is used to input the feature assessment results into the state assessment submodel and output size assessment results and velocity assessment results.

[0127] According to an embodiment of the present invention, the feature evaluation submodel includes an extraction layer, a conversion layer, and an update layer; the result output submodule includes an extraction unit, a conversion unit, an enhancement unit, and an information conversion unit. The extraction unit is used to input soil information, geometric information, and environmental information into the extraction layer for feature extraction to obtain initial feature information; the conversion unit is used to input the initial feature information into the conversion layer for nonlinear conversion to obtain intermediate feature information; the enhancement unit is used to input the intermediate feature information into the update layer for dimensionality reduction and feature enhancement to obtain multidimensional feature information; and the information conversion unit is used to convert the multidimensional feature information to obtain feature evaluation results.

[0128] According to an embodiment of the present invention, the information conversion unit includes: a dimensionality reduction subunit and a fusion subunit. The dimensionality reduction subunit is used to reduce the dimensionality of multidimensional feature information based on time steps or feature dimensions to obtain multiple reduced-dimensional features; the fusion subunit is used to fuse the multiple reduced-dimensional features to obtain a feature evaluation result.

[0129] According to an embodiment of the present invention, the result input submodule includes: a result processing unit, a splicing unit, and a pooling processing unit. The result processing unit is used to process the feature evaluation results based on the forward propagation strategy to obtain first state information corresponding to multiple moments, and to process the feature evaluation results based on the backward propagation strategy to obtain second state information corresponding to multiple moments; the splicing unit is used to splice the first state information and the second state information to obtain target state information; and the pooling processing unit is used to pool the target state information to obtain size evaluation results and rate evaluation results.

[0130] According to an embodiment of the present invention, the embankment includes an existing embankment and a newly built embankment, the geometric information includes first geometric information of the existing embankment and second geometric information of the newly built embankment; the soil information includes density information, water content information, compression modulus information, internal friction angle and cohesion information; and the environmental information includes temperature information, humidity information, rainfall information and settlement deformation information.

[0131] According to an embodiment of the present invention, an evaluation model is trained by the following operations: obtaining sample initial information to be evaluated corresponding to cracks on the surface of a sample embankment; processing the sample initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain sample information to be evaluated, wherein the sample information to be evaluated includes sample geometric information, sample soil information, sample crack information and sample environmental information corresponding to the sample embankment based on different types, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; using the sample crack information, sample geometric information, sample soil information and sample environmental information to train the initial evaluation model to obtain an evaluation model that meets the preset evaluation conditions.

[0132] According to embodiments of the present invention, any multiple modules among the information acquisition module 510, information processing module 520, and information input module 530 may be combined into a single module, or any one of these modules may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in a single module. According to embodiments of the present invention, at least one of the information acquisition module 510, information processing module 520, and information input module 530 may be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or may be implemented in hardware or firmware through any other reasonable means of circuit integration or packaging, or may be implemented in any one of the three implementation methods of software, hardware, and firmware, or any appropriate combination of these. Alternatively, at least one of the information acquisition module 510, information processing module 520, and information input module 530 may be at least partially implemented as a computer program module that, when executed, performs the corresponding functionality.

[0133] Figure 6 A block diagram of an electronic device suitable for implementing a method for evaluating embankment cracks according to an embodiment of the present invention is shown.

[0134] like Figure 6 As shown, an electronic device according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 602 or programs loaded from a storage unit 608 into a random access memory (RAM) 603. Processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or related chipsets and / or a dedicated microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. Processor 601 may also include onboard memory for caching purposes. Processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0135] Various programs and data required for the operation of the electronic device are stored in RAM 603. Processor 601, ROM 602, and RAM 603 are connected to each other via bus 604. Processor 601 executes the programs in ROM 602 and / or RAM 603 to perform various operations according to the method flow of the embodiment of the present invention. It should be noted that the programs may also be stored in one or more memories other than ROM 602 and RAM 603. Processor 601 may also execute the programs stored in the one or more memories to perform various operations according to the method flow of the embodiment of the present invention.

[0136] According to an embodiment of the present invention, the electronic device may further include an input / output (I / O) interface 605, which is also connected to the bus 604. The electronic device may further include one or more of the following components connected to the I / O interface 605: an input portion 606 including a keyboard, a mouse, etc.; an output portion 607 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and speakers; a storage portion 608 including a hard disk; and a communication portion 609 including a network interface card such as a LAN card or a modem. The communication portion 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. Removable media 611, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 610 as needed, so that computer programs read from the removable media can be installed in the storage portion 608 as needed.

[0137] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0138] According to an embodiment of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present invention, a computer-readable storage medium may include the ROM 602 and / or RAM 603 described above, and / or one or more memories other than ROM 602 and RAM 603.

[0139] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for executing the method shown in the flowchart. When the computer program product is executed in a computer system, the program code causes the computer system to implement the embankment crack assessment method provided in an embodiment of the present invention.

[0140] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when the computer program is executed by the processor 601. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0141] In one embodiment, the computer program may be stored on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may be transmitted and distributed in the form of a signal on a network medium, downloaded and installed via the communication portion 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0142] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 609 and / or installed from a removable medium 611. When the computer program is executed by the processor 601, the above-described functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the systems, devices, means, modules, units, etc. described above can be implemented by computer program modules.

[0143] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect via the Internet).

[0144] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0145] It will be understood by those skilled in the art that the features described in the various embodiments of the present invention may be combined and / or coupled in various ways, even if such combinations or couplings are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention may be combined and / or coupled in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or couplings fall within the scope of the present invention.

[0146] The above describes embodiments of the present invention. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which should all fall within the scope of the present invention.

Claims

1. A method for evaluating embankment cracks, characterized in that: The method comprises: obtaining initial information to be assessed corresponding to cracks on the embankment surface; Processing the initial information to be evaluated based on a decomposition strategy and a preset threshold to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; Inputting the soil information, the geometric information, and the environmental information into an evaluation model to obtain a size evaluation result and a rate evaluation result of the crack, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel, and the feature evaluation submodel includes an extraction layer, a conversion layer, and an update layer; The step of inputting the soil information, the geometric information, and the environmental information into an evaluation model to obtain a size evaluation result and a rate evaluation result of the crack includes: Inputting the soil information, the geometric information and the environmental information into the extraction layer to perform feature extraction to obtain initial feature information; Inputting the initial feature information into the conversion layer for nonlinear conversion to obtain intermediate feature information; Inputting the intermediate feature information into the update layer for dimensionality reduction and feature enhancement processing to obtain multidimensional feature information; Performing conversion processing on the multi-dimensional feature information to obtain a feature evaluation result; The feature evaluation result is input into the state evaluation sub-model, and the size evaluation result and the speed evaluation result are output.

2. The method according to claim 1, characterized in that The decomposition strategy includes a decomposition function and a decomposition dimension; The initial information to be evaluated is processed based on the decomposition strategy and the preset threshold to obtain information to be evaluated, including: Decomposing the initial information to be evaluated by using the decomposition function and the decomposition dimension to obtain decomposition coefficients of multiple frequencies corresponding to the initial information to be evaluated; Performing threshold quantization on the decomposition coefficients of the multiple frequencies based on the preset threshold to obtain processing coefficients of the multiple frequencies; The processing coefficients of the multiple frequencies are fused based on an inverse transformation function to obtain the information to be evaluated.

3. The method according to claim 2, characterized in that The method further comprises: The decomposition function and the decomposition dimension are determined according to at least one of a length feature and a frequency feature of the initial information to be evaluated.

4. The method according to claim 1, wherein Converting the multidimensional feature information to obtain the feature evaluation result includes: Performing dimensionality reduction processing on the multidimensional feature information based on the time step or feature dimension to obtain multiple dimensionality reduction features; Perform feature fusion on the multiple dimensionality reduction features to obtain the feature evaluation result.

5. The method according to claim 1, wherein Inputting the feature evaluation result into the state evaluation sub-model and outputting the size evaluation result and the rate evaluation result includes: Processing the feature evaluation results based on a forward propagation strategy to obtain first state information corresponding to a plurality of moments, and processing the feature evaluation results based on a backward propagation strategy to obtain second state information corresponding to the plurality of moments; concatenating the first state information and the second state information to obtain target state information; Pooling is performed on the target state information to obtain the size evaluation result and the rate evaluation result.

6. The method according to claim 1, characterized in that The embankment includes an existing embankment and a newly built embankment, the geometric information includes the first geometric information of the existing embankment and the second geometric information of the newly built embankment; the soil information includes density information, water content information, compression modulus information, internal friction angle and cohesion information; the environmental information includes temperature information, humidity information, rainfall information and settlement deformation information.

7. The method according to any one of claims 1 to 6, characterized in that The evaluation model is trained by the following operations: Obtaining initial sample information to be evaluated corresponding to cracks on the surface of the sample embankment; The sample initial information to be evaluated is processed based on a decomposition strategy and a preset threshold to obtain sample information to be evaluated, wherein the sample information to be evaluated includes sample geometry information of the sample embankment, sample soil information, sample crack information, and sample environmental information corresponding to the sample embankment based on different types, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold is used to perform threshold quantization on the decomposed initial information to be evaluated; An initial evaluation model is trained using the sample crack information, the sample geometry information, the sample soil information, and the sample environment information to obtain an evaluation model that meets preset evaluation conditions.

8. An evaluation device for embankment cracks, characterized in that: The device comprises: An information acquisition module, configured to acquire initial information to be evaluated corresponding to cracks on the embankment surface; an information processing module, configured to process the initial information to be evaluated based on a decomposition strategy and a preset threshold value to obtain information to be evaluated, wherein the information to be evaluated includes geometric information of the embankment, soil information, and environmental information corresponding to the embankment based on different types, the decomposition strategy is used to decompose the initial information to be evaluated, and the preset threshold value is used to perform threshold quantization on the decomposed initial information to be evaluated; an information input module, configured to input the soil information, the geometric information, and the environmental information into an evaluation model to obtain a size evaluation result and a rate evaluation result of the crack, wherein the evaluation model includes a feature evaluation submodel and a state evaluation submodel based on different functions, and the feature evaluation submodel includes an extraction layer, a conversion layer, and an update layer; Among them, the information input module includes: a result output submodule and a result input submodule. The result output submodule includes an extraction unit, which is used to input the soil information, the geometric information and the environmental information into the extraction layer for feature extraction to obtain initial feature information; a conversion unit, which is used to input the initial feature information into the conversion layer for nonlinear conversion to obtain intermediate feature information; an enhancement unit, which is used to input the intermediate feature information into the update layer for dimensionality reduction and feature enhancement processing to obtain multidimensional feature information; an information conversion unit, which is used to convert the multidimensional feature information to obtain a feature evaluation result; and a result input submodule, which is used to input the feature evaluation result into the state evaluation submodel and output the size evaluation result and the rate evaluation result.

Images