A bridge array radar signal processing system and method

Abstract

This invention relates to the field of bridge inspection technology and discloses a bridge array radar signal processing system and method. The method involves periodically acquiring reflected electromagnetic wave signals through a bridge array radar system and preprocessing them to improve signal quality. Utilizing 3D modeling and the spatial resolution capability of the array radar, the signals are precisely matched to the bridge model, forming an electromagnetic wave reflection image of the internal structure. Based on the electromagnetic properties of the bridge materials, intensity thresholds and morphological recognition rules are set to distinguish between normal and defective structures and analyze defect development trends. Furthermore, deep learning algorithms are used to extract and classify features from the reflection images, automatically identifying the location, type, and size of defects and generating an inspection report. Finally, combined with Geographic Information System (GIS), the spatial distribution of defects is displayed, and targeted maintenance plans are formulated. This invention improves the accuracy and efficiency of detecting internal bridge defects.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge detection, and in particular to a bridge array radar signal processing system and method. Background Art

[0002] As an important transportation infrastructure, the structural safety of bridges is directly related to the smoothness of transportation and the safety of life and property of the public. However, during the long-term use of bridges, due to the influence of various factors such as environmental erosion, vehicle loads, natural disasters, etc., their internal structures often suffer various forms of damage and defects, such as cracks, corrosion, and spalling. If these potential defects are not discovered and repaired in time, they will seriously threaten the bearing capacity and service life of the bridge, and even lead to serious accidents such as bridge collapse.

[0003] Existing bridge inspection methods, such as manual inspection, static load test, dynamic load test, etc., can detect bridge defects to a certain extent, but they have limitations such as long inspection cycle, low efficiency, difficulty in fully covering all structural parts, and great impact on traffic. In particular, for some internal defects that are more concealed, such as cracks and cavities inside the bridge, traditional methods are often difficult to effectively detect.

[0004] With the continuous development of radar technology, radar detection technology has gradually been applied to the field of bridge structure detection. Radar detection has the advantages of non-contact, long-distance, high precision, and high efficiency. It can penetrate the surface of the bridge and perform non-destructive detection of the internal structure. However, when processing electromagnetic wave signals reflected from bridges, existing radar detection methods often face problems such as large signal interference, complex data processing, and poor imaging effects, which affect the accuracy and reliability of the detection results to a certain extent. Summary of the invention

[0005] The object of the present invention is to provide a bridge array radar signal processing system and method to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a bridge array radar signal processing method, the method comprising:

[0007] Step 1: The reflected electromagnetic wave signal of the bridge structure is collected regularly by the bridge array radar system, wherein the reflected electromagnetic wave signal contains the echo information of the internal structure of the bridge, and the timestamp of the signal collection and the transmission parameters of the radar are recorded at the same time;

[0008] Step 2: Preprocess the collected reflected electromagnetic wave signal, including signal denoising, gain adjustment, time-frequency analysis and phase correction;

[0009] Step 3: Build a three-dimensional model of the bridge structure, and use the spatial resolution capability of the array radar to accurately match the pre-processed reflected electromagnetic wave signal with the spatial coordinates of the bridge model. Through the signal imaging algorithm, map the reflected signal to the bridge model to form an electromagnetic wave reflection image of the internal structure of the bridge;

[0010] Step 4: According to the electromagnetic properties and structural characteristics of the bridge material, the intensity threshold and morphological recognition rules of the reflected signal are set to distinguish normal structures from potential defects. Combined with the timestamp information, the changes of defects over time are analyzed to evaluate their development trends.

[0011] Step 5: Analyze the reflected image and use deep learning algorithms to extract and classify features in the image, automatically identify and mark the location, type and size of defects inside the bridge, and generate a defect detection report;

[0012] Step 6: Combine the defect detection report with the bridge’s geographic information GIS, display the spatial distribution of bridge defects through the GIS platform, and formulate targeted maintenance plans based on the severity and location of the defects.

[0013] Preferably, the radar transmission parameters recorded in step 1 include the radar's transmission frequency, beam width and scanning angle.

[0014] Preferably, the step of mapping the reflected signal onto the bridge model by a signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge includes:

[0015] Step A: Perform spatial domain processing on the preprocessed reflected electromagnetic wave signal, and use the signal data obtained by multiple receiving channels of the array radar to construct a signal matrix S, where each row of S corresponds to a signal of a receiving channel and each column corresponds to a time sampling point;

[0016] Step B: According to the geometric layout of the array radar and the beamforming principle of signal acquisition, the spatial spectrum of the signal matrix S is calculated to obtain the spatial spectrum matrix P, where the elements of P represent the signal strength in different spatial directions;

[0017] Step C: Use the spatial spectrum matrix P and the three-dimensional model of the bridge structure to perform spatial coordinate matching, map the signal intensity in the spatial spectrum to the corresponding spatial position on the bridge model, and obtain the initial electromagnetic wave reflection image I_0;

[0018] Step D: performing image enhancement processing on the initial reflection image I_0, including contrast adjustment, edge detection and texture enhancement, to improve the recognizability of defects in the image, and obtaining an enhanced reflection image I_e;

[0019] Step E: According to the electromagnetic properties and structural characteristics of the bridge material, the intensity threshold T_h and the morphological recognition rule R of the reflected signal are set, and the enhanced reflected image I_e is analyzed pixel by pixel. The pixels whose signal intensity exceeds the threshold T_h and meets the morphological recognition rule R are marked as potential defect areas to obtain the marked reflected image I_m;

[0020] The spatial spectrum is calculated using the following formula:

[0021] ,

[0022] Among them, W is the weight coefficient matrix, which is calculated according to the beamforming algorithm of the array radar. represents the conjugate transpose of W, and S is the preprocessed reflected electromagnetic wave signal matrix.

[0023] Preferably, in step 4, the method for identifying bridge defects is:

[0024] Step 4a: Based on the electromagnetic properties of bridge materials, a material electromagnetic properties database is established, which includes electromagnetic parameters of different materials at specific frequencies;

[0025] Step 4b: In combination with the geometric characteristics of the bridge structure and the expected normal state, a strength reference value I_b of the reflection signal is set, which represents the typical reflection intensity of the bridge structure in the absence of defects;

[0026] Step 4c: According to the material electromagnetic property database and the intensity reference value I_b, the intensity threshold T_d of the reflected signal is set, where T_d = I_b + ΔI, ΔI is the intensity offset determined according to the material properties and structural tolerance, and is used to distinguish normal structures from potential defects;

[0027] Step 4d: According to the geometric morphology of the bridge structure and the typical manifestations of defects, formulate the morphological recognition rules R_m, including the shape, size and edge characteristics of the defects, to further confirm the potential defects;

[0028] Step 4e: Perform pixel-by-pixel analysis on the electromagnetic wave reflection image of the internal structure of the bridge obtained by the signal imaging algorithm, and mark the pixels whose signal intensity exceeds the intensity threshold T_d and meets the morphological recognition rule R_m as potential defective pixels.

[0029] Preferably, the potential defects identified in step 4 include cracks, cavities, corrosion, misplaced or missing steel bars, and concrete quality defects inside the bridge.

[0030] Preferably, in step 5, the reflected image is analyzed, and features in the image are extracted and classified using a deep learning algorithm to automatically identify and annotate the location, type and size of defects inside the bridge. The specific steps include:

[0031] Step 5a: Build a deep learning model and select a convolutional neural network (CNN) as the feature extractor. The model contains multiple convolutional layers, pooling layers, fully connected layers, and a classification output layer to automatically learn and extract defect features in the reflection image.

[0032] Step 5b: Prepare a training dataset, which includes a large number of labeled bridge array radar reflection images with the location, type and size information of defects annotated in the images for training the deep learning model;

[0033] Step 5c: Input the training data set into the deep learning model, calculate the loss between the model output and the true annotation through the forward propagation algorithm, and iteratively update the model parameters using the backpropagation algorithm and gradient descent optimization method until the model loss converges below the preset threshold;

[0034] Step 5d: Validate the trained deep learning model and use an independent validation dataset to evaluate the recognition accuracy and generalization ability of the model to ensure that the model can accurately identify defects in the reflection image.

[0035] Step 5e: Input the radar reflection image of the bridge array to be inspected into the trained deep learning model. The model automatically extracts and classifies the features of the image and outputs the location, type and size information of the defects.

[0036] Preferably, in step 5c, the calculation formula of the gradient descent algorithm is:

[0037] ,

[0038] in, Represents the model parameters, including all weights and biases that the model needs to learn; Represents the learning rate, which is used to control the step size of parameter update; Represents the loss function About parameters The gradient of is a vector pointing to the direction where the loss function grows fastest; Indicates an assignment operation, that is, updating parameters The value of .

[0039] Preferably, in each iteration of model training, the cross entropy loss function is used to evaluate the performance of the model, and its calculation formula is:

[0040] ,

[0041] in, is the loss value, is the sample size, For true marking, is the model's predicted value.

[0042] Preferably, the method further includes: establishing a bridge health monitoring database to store all collected reflected electromagnetic wave signals, preprocessing results, reflected images, defect detection reports and maintenance records to provide data support for long-term monitoring and maintenance of the bridge.

[0043] Preferably, a bridge array radar signal processing system comprises:

[0044] A signal acquisition module is used to regularly acquire the reflected electromagnetic wave signal of the bridge structure through the bridge array radar system, wherein the reflected electromagnetic wave signal contains the echo information of the internal structure of the bridge, and records the timestamp of the signal acquisition and the transmission parameters of the radar;

[0045] The preprocessing module is connected to the signal acquisition module and is used to preprocess the collected reflected electromagnetic wave signals, including signal denoising, gain adjustment, time-frequency analysis and phase correction;

[0046] The three-dimensional modeling and matching module is connected to the preprocessing module and is used to construct a three-dimensional model of the bridge structure and use the spatial resolution capability of the array radar to accurately match the preprocessed reflected electromagnetic wave signal with the spatial coordinates of the bridge model;

[0047] The signal imaging module is connected to the 3D modeling and matching module and is used to map the reflected signal onto the bridge model through the signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge;

[0048] The defect recognition and analysis module is connected to the signal imaging module and is used to set the intensity threshold and morphological recognition rules of the reflected signal according to the electromagnetic properties and structural characteristics of the bridge material, distinguish between normal structures and potential defects, and analyze the changes of defects over time and evaluate their development trends in combination with timestamp information;

[0049] The deep learning analysis module is connected to the defect recognition and analysis module and is used to analyze the reflected image. It uses the deep learning algorithm to extract and classify the features in the image, automatically identify and annotate the location, type and size of the defects inside the bridge, and generate a defect detection report.

[0050] The geographic information integration module, connected to the deep learning analysis module, is used to combine the defect detection report with the geographic information GIS of the bridge, display the spatial distribution of bridge defects through the GIS platform, and formulate targeted maintenance plans based on the severity and location of the defects.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] By adopting the bridge array radar system, the present invention can regularly collect the reflected electromagnetic wave signals of the bridge structure and perform refined preprocessing on them, including signal denoising, gain adjustment, time-frequency analysis and phase correction, effectively reducing signal interference and improving signal quality. Combined with three-dimensional modeling and precise matching technology, the reflected signal is accurately mapped to the bridge model to form a clear electromagnetic wave reflection image, thereby greatly improving the accuracy and reliability of defect detection.

[0053] Existing bridge inspection methods often fail to fully cover all structural parts, especially the internal defects that are more hidden. However, the present invention uses the non-contact and long-distance characteristics of radar detection to penetrate the bridge surface and conduct all-round and no-dead-angle inspection of the internal structure, effectively discovering hidden defects that are difficult to reach with traditional methods, such as cracks and cavities inside the bridge.

[0054] The present invention introduces a deep learning algorithm to extract and classify the features of the reflection image, which can automatically identify and mark the location, type and size of the defects inside the bridge, and generate a detailed defect detection report. Combined with the geographic information GIS platform, the spatial distribution of bridge defects can be intuitively displayed, providing a scientific basis for the formulation of targeted maintenance plans, and realizing the intelligent and precise maintenance of bridges. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 A step diagram of a bridge array radar signal processing method according to the present invention;

[0056] Figure 2 A diagram showing the steps for generating an electromagnetic wave reflection image;

[0057] Figure 3 This is a flowchart for building the deep learning model in step 5. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0059] See also Figure 1-3 The present invention provides a technical solution: a bridge array radar signal processing method, the method comprising the following steps:

[0060] Step 1: The reflected electromagnetic wave signals of the bridge structure are collected regularly through the bridge array radar system. Specifically, the array radar system is configured to emit electromagnetic wave pulses at predetermined time intervals and receive electromagnetic wave signals reflected by the internal structure of the bridge. These reflected signals contain echo information of the internal structure of the bridge, which is crucial for subsequent analysis. At the same time, the system also records the timestamp of each signal acquisition and the radar's transmission parameters, which include the radar's transmission frequency, beam width, and scanning angle.

[0061] Step 2: Preprocess the collected reflected electromagnetic wave signal. The preprocessing process includes signal denoising to remove noise and interference in the signal; gain adjustment to optimize the signal strength; time-frequency analysis to obtain the characteristics of the signal in the time and frequency domains; and phase correction to correct the phase offset that may occur during the transmission process.

[0062] Step 3: Construct a three-dimensional model of the bridge structure and use the spatial resolution of the array radar to accurately match the signal with the model. First, construct a three-dimensional model of the bridge structure based on the design drawings or actual measurement data of the bridge. Then, use the spatial resolution provided by the array radar to accurately match the pre-processed reflected electromagnetic wave signal with the spatial coordinates of the bridge model. Through the signal imaging algorithm, the reflected signal is mapped to the bridge model, thereby forming an electromagnetic wave reflection image of the internal structure of the bridge.

[0063] Step 4: According to the electromagnetic properties and structural characteristics of the bridge material, set the intensity threshold and morphological recognition rules of the reflected signal. By analyzing the electromagnetic properties of the bridge material, such as dielectric constant, conductivity, etc., and the characteristics of the bridge structure, such as thickness, shape, etc., set a reasonable reflection signal intensity threshold. At the same time, according to the signal morphology that the defects may present, such as signal interruption caused by cracks, signal intensity changes caused by corrosion, etc., formulate morphological recognition rules. These thresholds and rules are used to distinguish normal structures from potential defects, and analyze the changes of defects over time in combination with timestamp information to evaluate their development trends. The potential defects include cracks, cavities, corrosion, dislocation or missing steel bars, and concrete quality defects inside the bridge.

[0064] Step 5: Analyze the reflected image and use the deep learning algorithm to extract and classify the features in the image. Specifically, the deep learning algorithm is used to extract and classify the features of the reflected image. By training the deep learning model, it can automatically identify and annotate the location, type and size of the defects inside the bridge. This step realizes the automatic identification and annotation of defects, greatly improving the detection efficiency and accuracy. Finally, a defect detection report is generated based on the recognition results.

[0065] Step 6: Combine the defect detection report with the bridge's geographic information GIS. Specifically, the defect location, type, and size in the defect detection report are associated with the bridge's geographic information (such as longitude, latitude, altitude, etc.). Then, the spatial distribution of bridge defects is displayed through the GIS platform, allowing maintenance personnel to intuitively understand the location and distribution of defects. Finally, according to the severity and location information of the defects, a targeted maintenance plan is formulated to ensure the safe operation of the bridge.

[0066] A bridge health monitoring database is established to store all collected reflected electromagnetic wave signals, preprocessing results, reflected images, defect detection reports and maintenance records, providing data support for long-term monitoring and maintenance of bridges.

[0067] The present invention will be further described below in conjunction with Examples 1 to 3: Embodiment 1:

[0068] The step of mapping the reflected signal onto the bridge model through a signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge includes:

[0069] Step A: Perform spatial domain processing on the preprocessed reflected electromagnetic wave signal. Specifically, a two-dimensional signal matrix S is constructed using the signal data simultaneously acquired by multiple receiving channels of the array radar. In this matrix, each row of S corresponds to a signal received by a receiving channel, and each column corresponds to a specific time sampling point.

[0070] Step B: Calculate the spatial spectrum of the signal matrix S. According to the geometric layout of the array radar (including the position and orientation of the receiving channel) and the beamforming principle of signal acquisition, the spatial spectrum of the signal matrix S is calculated to obtain a spatial spectrum matrix P. The elements of the spatial spectrum matrix P represent the signal strength in different spatial directions, which reflects the propagation and reflection of electromagnetic waves in the internal structure of the bridge. The spatial spectrum is calculated using the following formula:

[0071]

[0072] Where W is the weight coefficient matrix, which is pre-calculated according to the beamforming algorithm of the array radar and is used to perform spatial filtering on the signal; represents the conjugate transposed matrix of W; S is the reflected electromagnetic wave signal matrix after preprocessing.

[0073] Step C: Map the spatial spectrum to the bridge model. Use the calculated spatial spectrum matrix P and the three-dimensional model of the bridge structure to accurately match the spatial coordinates. Specifically, map the signal intensity values ​​in the spatial spectrum to the corresponding spatial positions on the bridge model according to the corresponding spatial directions, thereby obtaining an initial electromagnetic wave reflection image I_0. This image intuitively shows the reflection of electromagnetic waves by the internal structure of the bridge.

[0074] Step D: Perform image enhancement processing on the initial reflection image. In order to improve the recognizability of defects in the image, a series of image enhancement processing is performed on the initial reflection image I_0, including contrast adjustment, edge detection and texture enhancement. Contrast adjustment can make the difference between light and dark in the image more obvious; edge detection can highlight the edge information in the image, such as the outline of defects such as cracks and corrosion; texture enhancement can enhance the detail information in the image, making the defects more clearly visible. After these processing, the enhanced reflection image I_e is obtained.

[0075] Step E: Marking potential defect areas. According to the electromagnetic properties and structural characteristics of the bridge material, a reflection signal intensity threshold T_h and a set of morphological recognition rules R are set. Then, the enhanced reflection image I_e is analyzed pixel by pixel. For pixels whose signal intensity exceeds the threshold T_h and meets the morphological recognition rule R, they are marked as potential defect areas. In this way, a marked reflection image I_m is obtained, in which the potential defect areas are clearly marked, providing an important basis for subsequent analysis and maintenance. Embodiment 2:

[0076] In the process of bridge defect identification, the present invention adopts the following detailed steps to achieve accurate detection of the internal structure of the bridge:

[0077] Step 4a: Establish a database of electromagnetic properties of materials. First, collect the electromagnetic parameters of the main materials used in bridges (such as concrete, steel bars, steel, etc.) at specific frequencies, including but not limited to dielectric constant, conductivity, magnetic permeability, etc. These parameters are the basis for the propagation and reflection of electromagnetic waves in materials. Then, organize these parameters into a database for use in subsequent analysis and calculations. For example, for concrete materials, it may include its dielectric constant and conductivity at different frequencies.

[0078] Step 4b: Set the intensity reference value of the reflected signal. Combined with the geometric characteristics of the bridge structure (such as thickness, shape, number of layers, etc.) and the expected normal state, a reflection signal intensity reference value I_b is set through simulation or actual measurement. This reference value represents the typical reflection intensity of the bridge structure in the absence of defects and is an important basis for subsequent defect judgment. For example, for a section of uniform concrete beam, its reflection intensity reference value at a specific frequency can be obtained through simulation calculation or actual measurement.

[0079] Step 4c: Set the intensity threshold of the reflected signal. According to the parameters in the material electromagnetic property database and the intensity reference value I_b, set a reflection signal intensity threshold T_d. This threshold is used to distinguish normal structures from potential defects. Its calculation formula is T_d = I_b + ΔI, where ΔI is the intensity offset determined according to material properties and structural tolerances. The setting of ΔI needs to take into account factors such as material heterogeneity, measurement errors, and changes in reflection intensity that may be caused by defects. For example, for concrete materials, ΔI may be set based on the fluctuation range of its experimentally measured reflection intensity.

[0080] Step 4d: Formulate morphological recognition rules. According to the geometric morphology of the bridge structure and the typical manifestations of defects (such as the shape of cracks, corroded areas, peeling edges, etc.), formulate a set of morphological recognition rules R_m. These rules include the shape of the defect (such as linear, circular, irregular, etc.), size (such as length, width, depth, etc.) and edge features (such as sharp, blurred, etc.), which are used to further confirm potential defects. The formulation of morphological recognition rules needs to be based on a large number of actual cases and professional knowledge to ensure its accuracy and reliability.

[0081] Step 4e: Analyze and mark potential defects pixel by pixel. Perform pixel-by-pixel analysis on the electromagnetic wave reflection image of the internal structure of the bridge obtained by the signal imaging algorithm. For each pixel, compare its signal strength with the intensity threshold T_d, and check whether it meets the morphological recognition rule R_m. If the signal strength of a pixel exceeds the intensity threshold T_d and its morphology meets the morphological recognition rule R_m, the pixel is marked as a potential defective pixel. By traversing the entire image, a set of all potential defective pixels can be obtained, providing important information for subsequent analysis and maintenance.

[0082] Through the implementation of the above steps, the present invention can achieve accurate detection of the internal structure of the bridge and effectively identify potential defects. Embodiment 3:

[0083] In the process of automatic identification and labeling of internal defects of bridges, the present invention adopts deep learning algorithms, especially convolutional neural networks (CNNs), to achieve efficient extraction and accurate classification of defect features in reflection images. The following are the detailed implementation steps of the process:

[0084] Step 5a: Build a deep learning model. A convolutional neural network (CNN) is selected as the feature extractor to build a deep learning model. The model design includes multiple convolutional layers to capture local features in the image; pooling layers to reduce the dimension of the feature map and retain important information; fully connected layers to integrate features and output classification results; and a classification output layer to finally determine the location, type, and size of the defect. The specific architecture of the model (such as the number of layers, convolution kernel size, activation function, etc.) is optimized according to the task requirements and data characteristics. For example, for crack defects inside the bridge, deeper convolutional layers are designed to capture subtle changes in the cracks.

[0085] Step 5b: Prepare training datasets. Collect and organize a large number of annotated bridge array radar reflection images as training datasets. These images should contain defects of various types, locations, and sizes, with detailed annotation information. The annotation process can be completed manually by experts or with the assistance of existing defect detection algorithms. For example, for each image, annotate the location (represented by a bounding box or segmentation mask), type (such as transverse cracks, longitudinal cracks, etc.) and size (such as length, width) of the cracks.

[0086] Step 5c: Train the deep learning model. Input the prepared training data set into the deep learning model, and calculate the loss between the model output and the true annotation through the forward propagation algorithm. Use the backpropagation algorithm and gradient descent optimization method (such as stochastic gradient descent, Adam, etc.) to iteratively update the model parameters to minimize the loss function. The calculation formula of the gradient descent algorithm is:

[0087] ,

[0088] in, Represents the model parameters, including all weights and biases that the model needs to learn; Represents the learning rate, which is used to control the step size of parameter update; Represents the loss function About parameters The gradient of is a vector pointing to the direction where the loss function grows fastest; Indicates an assignment operation, that is, updating parameters During the training process, data enhancement techniques (such as rotation, scaling, flipping, etc.) are introduced to increase the diversity of data and improve the generalization ability of the model. At the same time, by monitoring the training loss and validation loss, hyperparameters such as learning rate and batch size are adjusted until the model loss converges below the preset threshold. The cross entropy loss function is used to evaluate the performance of the model, and its calculation formula is:

[0089] ,

[0090] in, is the loss value, is the sample size, For true marking, is the model's predicted value.

[0091] Step 5d: Validate the deep learning model. Validate the trained deep learning model using an independent validation dataset. The validation dataset should contain radar reflection images of bridge arrays that are different from the training dataset but have similar characteristics. Evaluate the performance of the model by calculating the recognition accuracy (such as accuracy, recall, F1 score, etc.) and generalization ability (such as performance on unseen defect types) of the model on the validation dataset. If the model performance does not meet the requirements, return to step 5a to adjust the model architecture or training strategy and retrain the model.

[0092] Step 5e: Apply deep learning model for defect identification. Input the radar reflection image of the bridge array to be inspected into the trained deep learning model. The model automatically extracts and classifies the image features and outputs the location (represented by bounding box or segmentation mask), type (such as crack, corrosion, spalling, etc.) and size (such as length, width, area, etc.) of the defect. This information can be directly used for maintenance decision-making and repair plan formulation of the bridge. For example, for the identified crack defect, a targeted repair plan can be formulated based on its location and size information.

[0093] The present invention also includes a bridge array radar signal processing system, the system comprising:

[0094] A signal acquisition module is used to regularly acquire the reflected electromagnetic wave signal of the bridge structure through the bridge array radar system, wherein the reflected electromagnetic wave signal contains the echo information of the internal structure of the bridge, and records the timestamp of the signal acquisition and the transmission parameters of the radar;

[0095] The preprocessing module is connected to the signal acquisition module and is used to preprocess the collected reflected electromagnetic wave signals, including signal denoising, gain adjustment, time-frequency analysis and phase correction;

[0096] The three-dimensional modeling and matching module is connected to the preprocessing module and is used to construct a three-dimensional model of the bridge structure and use the spatial resolution capability of the array radar to accurately match the preprocessed reflected electromagnetic wave signal with the spatial coordinates of the bridge model;

[0097] The signal imaging module is connected to the 3D modeling and matching module and is used to map the reflected signal onto the bridge model through the signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge;

[0098] The defect recognition and analysis module is connected to the signal imaging module and is used to set the intensity threshold and morphological recognition rules of the reflected signal according to the electromagnetic properties and structural characteristics of the bridge material, distinguish between normal structures and potential defects, and analyze the changes of defects over time and evaluate their development trends in combination with timestamp information;

[0099] The deep learning analysis module is connected to the defect recognition and analysis module and is used to analyze the reflected image. It uses the deep learning algorithm to extract and classify the features in the image, automatically identify and annotate the location, type and size of the defects inside the bridge, and generate a defect detection report.

[0100] The geographic information integration module, connected to the deep learning analysis module, is used to combine the defect detection report with the geographic information GIS of the bridge, display the spatial distribution of bridge defects through the GIS platform, and formulate targeted maintenance plans based on the severity and location of the defects.

[0101] The implementation of the system refers to the above embodiment and will not be described in detail in the specification.

[0102] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0103] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A bridge array radar signal processing method, characterized in that: The method comprises: Step 1: The reflected electromagnetic wave signal of the bridge structure is collected regularly by the bridge array radar system, wherein the reflected electromagnetic wave signal contains the echo information of the internal structure of the bridge, and the timestamp of the signal collection and the transmission parameters of the radar are recorded at the same time; Step 2: Preprocess the collected reflected electromagnetic wave signal, including signal denoising, gain adjustment, time-frequency analysis and phase correction; Step 3: Build a three-dimensional model of the bridge structure, and use the spatial resolution capability of the array radar to accurately match the pre-processed reflected electromagnetic wave signal with the spatial coordinates of the bridge model. Through the signal imaging algorithm, map the reflected signal to the bridge model to form an electromagnetic wave reflection image of the internal structure of the bridge; Step 4: According to the electromagnetic properties and structural characteristics of the bridge material, the intensity threshold and morphological recognition rules of the reflected signal are set to distinguish normal structures from potential defects. Combined with the timestamp information, the changes of defects over time are analyzed to evaluate their development trends. Step 5: Analyze the reflected image and use deep learning algorithms to extract and classify features in the image, automatically identify and mark the location, type and size of defects inside the bridge, and generate a defect detection report; Step 6: Combine the defect detection report with the bridge's geographic information GIS, display the spatial distribution of bridge defects through the GIS platform, and formulate targeted maintenance plans based on the severity and location of the defects; In step 4, the method for identifying bridge defects is: Step 4a: Based on the electromagnetic properties of bridge materials, a material electromagnetic properties database is established, which includes electromagnetic parameters of different materials at specific frequencies; Step 4b: In combination with the geometric characteristics of the bridge structure and the expected normal state, a strength reference value I_b of the reflection signal is set, which represents the typical reflection intensity of the bridge structure in the absence of defects; Step 4c: According to the material electromagnetic property database and the intensity reference value I_b, the intensity threshold T_d of the reflected signal is set, where T_d = I_b + ΔI, ΔI is the intensity offset determined according to the material properties and structural tolerance, and is used to distinguish normal structures from potential defects; Step 4d: According to the geometric morphology of the bridge structure and the typical manifestations of defects, formulate the morphological recognition rules R_m, including the shape, size and edge characteristics of the defects, to further confirm the potential defects; Step 4e: Analyze the electromagnetic wave reflection image of the internal structure of the bridge obtained by the signal imaging algorithm pixel by pixel, and mark the pixels whose signal intensity exceeds the intensity threshold T_d and meets the morphological recognition rule R_m as potential defect pixels; Potential defects identified in step 4 include cracks, voids, corrosion, misplaced or missing steel bars, and concrete quality defects inside the bridge; In step 5, the reflected image is analyzed, and the features in the image are extracted and classified using a deep learning algorithm to automatically identify and mark the location, type and size of defects inside the bridge. The specific steps include: Step 5a: Build a deep learning model and select a convolutional neural network (CNN) as the feature extractor. The model contains multiple convolutional layers, pooling layers, fully connected layers, and a classification output layer to automatically learn and extract defect features in the reflection image. Step 5b: Prepare a training dataset, which includes a large number of labeled bridge array radar reflection images with the location, type and size information of defects annotated in the images for training the deep learning model; Step 5c: Input the training data set into the deep learning model, calculate the loss between the model output and the true annotation through the forward propagation algorithm, and iteratively update the model parameters using the backpropagation algorithm and gradient descent optimization method until the model loss converges below the preset threshold; Step 5d: Validate the trained deep learning model and use an independent validation dataset to evaluate the recognition accuracy and generalization ability of the model to ensure that the model can accurately identify defects in the reflection image. Step 5e: Input the radar reflection image of the bridge array to be inspected into the trained deep learning model. The model automatically extracts and classifies the features of the image and outputs the location, type and size information of the defects.

2. The bridge array radar signal processing method according to claim 1 is characterized in that: The radar transmission parameters recorded in step 1 include the radar's transmission frequency, beam width and scanning angle.

3. The bridge array radar signal processing method according to claim 2 is characterized in that: The step of mapping the reflected signal onto the bridge model through a signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge includes: Step A: Perform spatial domain processing on the preprocessed reflected electromagnetic wave signal, and use the signal data obtained by multiple receiving channels of the array radar to construct a signal matrix S, where each row of S corresponds to a signal of a receiving channel and each column corresponds to a time sampling point; Step B: According to the geometric layout of the array radar and the beamforming principle of signal acquisition, the spatial spectrum of the signal matrix S is calculated to obtain the spatial spectrum matrix P, where the elements of P represent the signal strength in different spatial directions; Step C: Use the spatial spectrum matrix P and the three-dimensional model of the bridge structure to perform spatial coordinate matching, map the signal intensity in the spatial spectrum to the corresponding spatial position on the bridge model, and obtain the initial electromagnetic wave reflection image I_0; Step D: performing image enhancement processing on the initial reflection image I_0, including contrast adjustment, edge detection and texture enhancement, to improve the recognizability of defects in the image, and obtaining an enhanced reflection image I_e; Step E: According to the electromagnetic properties and structural characteristics of the bridge material, the intensity threshold T_h and the morphological recognition rule R of the reflected signal are set, and the enhanced reflected image I_e is analyzed pixel by pixel. The pixels whose signal intensity exceeds the threshold T_h and meets the morphological recognition rule R are marked as potential defect areas to obtain the marked reflected image I_m; The spatial spectrum is calculated using the following formula: ， Where W is the weight coefficient matrix, which is calculated according to the beamforming algorithm of the array radar. represents the conjugate transpose of W, and S is the preprocessed reflected electromagnetic wave signal matrix.

4. The bridge array radar signal processing method according to claim 1 is characterized in that: In step 5c, the calculation formula of the gradient descent algorithm is: ， in, Represents the model parameters, including all weights and biases that the model needs to learn; Represents the learning rate, which is used to control the step size of parameter update; Represents the loss function About parameters The gradient of is a vector pointing to the direction where the loss function grows fastest; Indicates an assignment operation, that is, updating parameters The value of .

5. The bridge array radar signal processing method according to claim 1, characterized in that: In each iteration of model training, the cross entropy loss function is used to evaluate the performance of the model, and its calculation formula is: ， in, is the loss value, is the sample size, For true marking, is the model's predicted value.

6. The bridge array radar signal processing method according to claim 1 is characterized in that: The method also includes: establishing a bridge health monitoring database to store all collected reflected electromagnetic wave signals, preprocessing results, reflected images, defect detection reports and maintenance records to provide data support for long-term monitoring and maintenance of the bridge.

7. A bridge array radar signal processing system, characterized in that: The system comprises: A signal acquisition module is used to regularly acquire the reflected electromagnetic wave signal of the bridge structure through the bridge array radar system, wherein the reflected electromagnetic wave signal contains the echo information of the internal structure of the bridge, and records the timestamp of the signal acquisition and the transmission parameters of the radar; The preprocessing module is connected to the signal acquisition module and is used to preprocess the collected reflected electromagnetic wave signals, including signal denoising, gain adjustment, time-frequency analysis and phase correction; The three-dimensional modeling and matching module is connected to the preprocessing module and is used to construct a three-dimensional model of the bridge structure and use the spatial resolution capability of the array radar to accurately match the preprocessed reflected electromagnetic wave signal with the spatial coordinates of the bridge model; The signal imaging module is connected to the 3D modeling and matching module and is used to map the reflected signal onto the bridge model through the signal imaging algorithm to form an electromagnetic wave reflection image of the internal structure of the bridge; The defect recognition and analysis module is connected to the signal imaging module and is used to set the intensity threshold and morphological recognition rules of the reflected signal according to the electromagnetic characteristics and structural characteristics of the bridge material, distinguish normal structures from potential defects, and analyze the changes of defects over time in combination with timestamp information to evaluate their development trends; the specific method of bridge defect recognition is as follows: i) According to the electromagnetic characteristics of the bridge material, a material electromagnetic characteristic database is established, which includes the electromagnetic parameters of different materials at specific frequencies; ii) In combination with the geometric characteristics of the bridge structure and the expected normal state, the intensity reference value I_b of the reflected signal is set, which represents the typical reflection intensity of the bridge structure in the absence of defects; iii) According to the material electromagnetic characteristic database and the intensity reference value I_b, the intensity threshold T_d of the reflected signal is set, where T_d = I_b + ΔI, ΔI is the intensity offset determined according to the material characteristics and structural tolerance, which is used to distinguish normal structures from potential defects; iv) According to the geometric morphology of the bridge structure and the typical manifestations of defects, a morphological recognition rule R_m is formulated, including the shape, size and edge characteristics of the defect, which is used to further confirm the potential defect; v) The electromagnetic wave reflection image of the internal structure of the bridge obtained by the signal imaging algorithm is analyzed pixel by pixel, and the pixels whose signal intensity exceeds the intensity threshold T_d and meets the morphological recognition rule R_m are marked as potential defect pixels; the potential defects identified include cracks, cavities, corrosion, dislocation or missing steel bars, and concrete quality defects inside the bridge; The deep learning analysis module is connected to the defect recognition and analysis module and is used to analyze the reflected image. It uses the deep learning algorithm to extract and classify the features in the image, automatically identify and annotate the location, type and size of the defects inside the bridge, and generate a defect detection report. The specific implementation steps include: i) constructing a deep learning model, selecting a convolutional neural network CNN as a feature extractor, and the model contains multiple convolutional layers, pooling layers, fully connected layers, and a classification output layer, which is used to automatically learn and extract the defect features in the reflected image; ii) preparing a training data set, which includes a large number of annotated bridge array radar reflection images, in which the location, type and size information of the defects are annotated for training the deep learning model; iii) inputting the training data set into the deep learning model, calculating the loss between the model output and the true annotation through the forward propagation algorithm, and iteratively updating the model parameters using the back propagation algorithm and the gradient descent optimization method until the model loss converges to below the preset threshold; iv) verifying the trained deep learning model, using an independent verification data set to evaluate the recognition accuracy and generalization ability of the model to ensure that the model can accurately identify defects in the reflected image; v) The radar reflection image of the bridge array to be inspected is input into the trained deep learning model, which automatically extracts and classifies the image features and outputs the location, type and size information of the defects. The geographic information integration module, connected to the deep learning analysis module, is used to combine the defect detection report with the geographic information GIS of the bridge, display the spatial distribution of bridge defects through the GIS platform, and formulate targeted maintenance plans based on the severity and location of the defects.

Images