Bridge crack detection method and system
By combining multispectral imaging and deep learning algorithms, precise and automated detection of bridge cracks has been achieved, solving the problems of low efficiency and insufficient accuracy of traditional detection methods and providing an efficient solution for bridge health monitoring and maintenance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-13
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional methods for detecting cracks in bridges are inefficient, lack precision, and are highly subjective in their assessments, making it difficult to meet the needs for rapid and accurate detection of large-scale bridge groups.
Multispectral imaging equipment is used to collect image data of the bridge surface. A crack identification model is constructed by combining deep learning algorithms. Through multispectral data fusion and image preprocessing, the geometric and texture feature parameters of the cracks are extracted, and automated evaluation is carried out in combination with preset evaluation standards.
It has achieved precision, automation and standardization in bridge crack detection, reduced subjective errors in manual assessment, and provided an efficient and reliable technical solution for structural health monitoring and maintenance.
Smart Images
Figure CN121830658A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of bridge crack detection technology, specifically to a bridge crack detection method and system. Background Technology
[0002] As a core component of transportation infrastructure, the structural safety of bridges is directly related to public safety and economic operation. However, due to factors such as environmental erosion, load-bearing effects, and material aging, bridge surfaces are prone to cracks and defects. If these are not detected and repaired in a timely manner, they may lead to structural failure or even collapse. Traditional crack detection methods mainly rely on manual visual inspection or single-spectrum imaging technology, which have the following limitations: 1. Low inspection efficiency: Manual inspection requires the erection of scaffolding or the use of inspection vehicles, which is time-consuming and has a limited coverage area, making it difficult to meet the rapid inspection needs of large-scale bridge groups.
[0003] 2. Insufficient accuracy: Single spectrum (such as visible light) is easily affected by light, stains or coatings, leading to missed detection of micro or hidden cracks; although near-infrared / mid-infrared bands can penetrate surface coatings, they lack multi-band data fusion, making it difficult to comprehensively characterize crack features.
[0004] 3. High degree of subjectivity in assessment: Traditional methods rely on human experience to judge the severity of cracks, lack quantitative assessment standards, and are prone to errors in maintenance decisions due to subjective differences. Summary of the Invention
[0005] To address the aforementioned technical problems, a method and system for detecting cracks in bridges are provided. This technical solution resolves the issues raised in the background section.
[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect of the present invention, a method for detecting cracks in bridges is provided, comprising: Bridge surface image data was collected, and multispectral imaging equipment was used to continuously scan and photograph the bridge surface under different spectral bands to obtain composite image data including visible light, near infrared and mid-infrared bands; The acquired composite image data is preprocessed, including image denoising, image enhancement, and image registration operations, to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands; Based on the preprocessed image data, a crack recognition model constructed using deep learning algorithms is used to perform preliminary crack detection and identify possible crack areas in the image. Feature extraction and analysis are performed on the initially detected crack areas to obtain the geometric and texture feature parameters of the cracks; The severity of the crack is assessed based on its geometric and textural characteristics, combined with a pre-defined crack assessment standard. Based on the assessment results of crack severity, a bridge crack detection report is generated, which includes the location, size, severity of cracks, and recommended repair measures.
[0007] Preferably, image denoising of the composite image data specifically includes: Wavelet decomposition was performed on the composite image data to obtain wavelet coefficients at different scales; Based on the statistical characteristics of wavelet coefficients, a pre-set threshold is retrieved from the database; The wavelet coefficients are compared with a pre-set threshold. Wavelet coefficients smaller than the threshold are set to zero, while wavelet coefficients larger than the threshold are retained or shrunk. Wavelet reconstruction is performed on the processed wavelet coefficients to obtain the denoised image data.
[0008] Preferably, image enhancement of the composite image data specifically includes: Adaptive histogram equalization is performed on the denoised composite image data. The cumulative distribution function of the local region is determined by calculating the gray-level histogram of the denoised image; Based on the cumulative distribution function, the gray-level mapping relationship of local regions is adaptively adjusted to obtain an enhanced image.
[0009] Preferably, the image registration operation employs a feature point-based registration method, with the following specific steps: Feature points, such as corner points and edge points, are extracted from images of different spectral bands using the Harris corner detection algorithm, whose response function is: ,in , and These are the gradients of the image in the x and y directions, respectively. This is an empirical constant, with a value of 0.05. Points with response function values greater than a set threshold are considered corner points, while points with smaller or negative response function values are considered edge points. represent The determinant value of the matrix, and represent The sum of the elements on the main diagonal of a matrix; The extracted feature points are matched to obtain feature point pairs. The nearest neighbor ratio matching method is used, and the nearest neighbor feature point p in the first image is... The second nearest neighbor feature point is ,like , The matching threshold is set to 0.6, which means p and They are matched feature point pairs; Calculate the transformation parameters and translation parameters between images based on feature point pairs. Rotation parameters Given the scaling parameter s, the transformation equation is solved using the least squares method: ; in and These are the coordinates of the matched feature point pairs in the two images, respectively; The image is transformed according to the transformation parameters to achieve accurate registration of images in different spectral bands.
[0010] Preferably, the specific steps for constructing a crack recognition model using deep learning algorithms are as follows: Collect a large amount of bridge crack image data and non-crack image data to construct a training dataset. and test dataset The number of crack images in the training dataset is The number of non-crack images is The number of crack images in the test dataset is The number of non-crack images is ; Choose a convolutional neural network structure and initialize the network parameters. Let the kernel size of the convolutional layer be k×k, the stride be s, the padding be p, and the number of neurons in the fully connected layer be n. The deep learning network is trained using the training dataset, and the network parameters are updated using the backpropagation algorithm. The cross-entropy loss function is used as follows: ; in, For the sample size, For the true label of the sample, The predicted label for the sample; Until the network's recognition accuracy on the test dataset Reaching the preset threshold, among which For a real example, For a true negative example, This is a false positive example. This is a false negative example; The trained crack recognition model is optimized using a model pruning method to obtain a new crack recognition model.
[0011] Preferably, feature extraction and analysis are performed on the initially detected crack areas to obtain the geometric feature parameters of the cracks, specifically including: An image edge detection-based method is employed. First, the crack edges are detected. Then, the sum of distances between adjacent pixels on the edge is calculated. and Calculate the distance between them: ; The crack length is ,in This represents the number of pixels on the edge. In the direction perpendicular to the crack, multiple measurement points are selected, and the distance between the two edges of the crack at each measurement point is calculated. Let the pixel coordinates of the two edges of the crack at the j-th measurement point be respectively... and The crack width at that measurement point is... ; Take the average value as the crack width ,in This represents the number of measurement points.
[0012] Preferably, feature extraction and analysis are performed on the initially detected crack areas to obtain the texture feature parameters of the cracks, specifically including: The gray-level co-occurrence matrix is calculated using the gray-level co-occurrence matrix method. ,in and Grayscale For pixel spacing, For the direction angle, calculate the entropy, contrast, and correlation of the gray-level co-occurrence matrix: ; ; ; in For gray levels, and They are respectively The mean of the rows and columns, and They are respectively The standard deviation of the rows and columns; The roughness of the cracked surface is assessed based on these statistical measures as follows: ; in, , and These are the weights for entropy, contrast, and correlation, respectively.
[0013] Preferably, the preset crack evaluation criteria are set based on the geometric and textural characteristic parameters of the crack, specifically as follows: When the crack length > ,width > At that time, it was determined to be a severe crack, among which , The threshold for determining severe cracks is determined based on bridge design requirements and usage experience, and is preset in the system. When the crack length ≤ ≤ ,width ≤ ≤ Meanwhile, crack roughness > At that time, it was determined to be a moderate crack, of which , , The threshold for determining moderate cracks is also preset in the system; All other cases were determined to be minor cracks.
[0014] Preferably, the generated bridge crack detection report includes the following: The specific location of the crack in the bridge, such as the pier, bridge deck, or beam, is determined by mapping the image coordinates to the actual structure of the bridge. Historical inspection records and trend analysis of cracks, by comparing the results of multiple inspections, let the first... The crack length detected was and width are Calculate the rate of change of crack length and the rate of change of crack width: ; ; Based on the rate of change of crack length and the rate of change of crack width, determine whether the crack is propagating or worsening; We provide repair solutions and cost estimates for cracks of varying severity.
[0015] In a second aspect of the invention, a bridge crack detection system is also provided, comprising: The acquisition module is used to acquire image data of the bridge surface. It uses a multispectral imaging device to continuously scan and photograph the bridge surface under different spectral bands to obtain composite image data including visible light, near-infrared and mid-infrared bands. The preprocessing module is used to preprocess the acquired composite image data. The preprocessing includes image denoising, image enhancement, and image registration operations to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands. The detection module is used to perform preliminary crack detection based on preprocessed image data and a crack recognition model constructed using a deep learning algorithm, and to identify possible crack areas in the image. The extraction module is used to extract and analyze features from the initially detected crack area to obtain the geometric and texture feature parameters of the crack. An evaluation module is used to evaluate the severity of a crack based on its geometric and textural features, combined with a preset crack evaluation standard. The report generation module is used to generate a bridge crack detection report based on the crack severity assessment results. The report includes the crack location, size, severity, and recommended repair measures.
[0016] Compared with the prior art, the present invention provides a method and system for detecting cracks in bridges, which has the following beneficial effects: This invention simultaneously acquires data in the visible, near-infrared, and mid-infrared bands, capturing crack characteristics under different spectra (such as morphology under visible light, material damage under near-infrared, and thermal radiation anomalies under mid-infrared), avoiding the limitations of single-band detection. Furthermore, it constructs a convolutional neural network (CNN) crack recognition model to achieve end-to-end crack localization and segmentation, adapting to crack recognition in complex backgrounds (such as corrosion and joints). It extracts crack geometric (length, width) and texture (roughness) feature parameters, automatically classifying the severity (mild / medium / severe) based on preset thresholds, providing a quantitative basis for repair solutions (such as grouting repair and carbon fiber reinforcement), reducing subjective errors in manual assessment. Thus, through multi-technology collaborative innovation, this invention achieves "precision, automation, and standardization" in bridge crack detection, providing an efficient and reliable technical solution for structural health monitoring and maintenance. Attached Figure Description
[0017] Figure 1 This is a schematic diagram of the method flow of S101-S106 in this invention; Figure 2 This is a schematic diagram of the method flow for S201-S204 in this invention; Figure 3 This is a schematic diagram of the method flow of S301-S303 in this invention; Figure 4 This is a schematic diagram of the method flow for S401-S404 in this invention; Figure 5 This is a schematic diagram of the method flow for S501-S505 in this invention. Detailed Implementation
[0018] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0019] Example 1 Please refer to Figure 1 As shown, in a first aspect of the present invention, a method for detecting cracks in bridges is provided, comprising: S101. Collect bridge surface image data. Use a multispectral imaging device to continuously scan and photograph the bridge surface under different spectral bands to obtain composite image data containing visible light, near-infrared and mid-infrared bands. S102. Preprocess the acquired composite image data. The preprocessing includes image denoising, image enhancement and image registration operations to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands. S103. Based on the preprocessed image data, a crack recognition model constructed using a deep learning algorithm is used to perform preliminary crack detection and identify possible crack areas in the image. S104. Perform feature extraction and analysis on the initially detected crack areas to obtain the geometric and texture feature parameters of the cracks. S105. Based on the geometric and textural characteristic parameters of the crack, and in conjunction with the preset crack evaluation criteria, assess the severity of the crack. S106. Based on the assessment results of crack severity, generate a bridge crack detection report. The report includes the location, size, severity of the crack, and recommended repair measures.
[0020] Those skilled in the art will understand that this invention, by simultaneously acquiring visible, near-infrared, and mid-infrared data, can capture the characteristics of cracks under different spectra (such as morphology under visible light, material damage under near-infrared, and thermal radiation anomalies under mid-infrared), avoiding the limitations of single-band detection. Furthermore, it constructs a convolutional neural network (CNN) crack recognition model to achieve end-to-end crack localization and segmentation, adapting to crack recognition in complex backgrounds (such as corrosion and joints). It extracts crack geometric (length, width) and texture (roughness) feature parameters, and automatically grades the severity (mild / medium / severe) based on preset thresholds, providing a quantitative basis for repair solutions (such as grouting repair and carbon fiber reinforcement), reducing subjective errors in manual assessment. Thus, through multi-technology collaborative innovation, this invention achieves "precision, automation, and standardization" in bridge crack detection, providing an efficient and reliable technical solution for structural health monitoring and maintenance.
[0021] Please refer to Figure 2 As shown, image denoising is performed on the acquired composite image data, specifically including: S201. Perform wavelet decomposition on the composite image data to obtain wavelet coefficients at different scales; S202. Based on the statistical characteristics of wavelet coefficients, retrieve the preset threshold from the database; S203. Compare the wavelet coefficients with a pre-set threshold, set the wavelet coefficients smaller than the threshold to zero, and retain or shrink the wavelet coefficients larger than the threshold. S204. Perform wavelet reconstruction on the processed wavelet coefficients to obtain the denoised image data.
[0022] Please refer to Figure 3 As shown, image enhancement is performed on the acquired composite image data, specifically including: S301. Perform adaptive histogram equalization on the denoised composite image data. S302. By calculating the gray-level histogram of the denoised image, the cumulative distribution function of the local region is determined; S303. Based on the cumulative distribution function, adaptively adjust the gray-level mapping relationship of the local region to obtain an enhanced image.
[0023] Please refer to Figure 4 As shown, the image registration operation uses a feature point-based registration method, and the specific steps are as follows: S401. Extract feature points from images of different spectral bands, such as corner points and edge points, using the Harris corner detection algorithm. Its response function is: ,in , and These are the gradients of the image in the x and y directions, respectively. This is an empirical constant, with a value of 0.05. Points with response function values greater than a set threshold are considered corner points, while points with smaller or negative response function values are considered edge points. represent The determinant value of the matrix, and represent The sum of the elements on the main diagonal of a matrix; S402. Match the extracted feature points to obtain feature point pairs. Use the nearest neighbor ratio matching method to find the nearest neighbor feature point p in the first image. The second nearest neighbor feature point is ,like , The matching threshold is set to 0.6, which means p and They are matched feature point pairs; S403. Calculate the transformation parameters and translation parameters between images based on feature point pairs. Rotation parameters Given the scaling parameter s, the transformation equation is solved using the least squares method: ; in and These are the coordinates of the matched feature point pairs in the two images, respectively; S404. Transform the image according to the transformation parameters to achieve accurate registration of images in different spectral bands.
[0024] Please refer to Figure 5 As shown, the specific steps for constructing a crack recognition model using deep learning algorithms are as follows: S501. Collect a large amount of bridge crack image data and non-crack image data to construct a training dataset. and test dataset The number of crack images in the training dataset is The number of non-crack images is The number of crack images in the test dataset is The number of non-crack images is ; S502. Select a convolutional neural network structure and initialize the network structure parameters. Let the kernel size of the convolutional layer be k×k, the stride be s, the padding be p, and the number of neurons in the fully connected layer be n. S503. Train the deep learning network using the training dataset, update the network parameters using the backpropagation algorithm, and use the cross-entropy loss function as the loss function: ; in, For the sample size, For the true label of the sample, The predicted label for the sample; S504, until the network's recognition accuracy on the test dataset. Reaching the preset threshold, among which For a real example, For a true negative example, This is a false positive example. This is a false negative example; S505. Optimize the trained crack recognition model using the model pruning method to obtain a crack recognition model.
[0025] Feature extraction and analysis are performed on the initially detected crack areas to obtain the geometric feature parameters of the cracks, specifically including: An image edge detection-based method is employed. First, the crack edges are detected. Then, the sum of distances between adjacent pixels on the edge is calculated. and Calculate the distance between them: ; The crack length is ,in This represents the number of pixels on the edge. In the direction perpendicular to the crack, multiple measurement points are selected, and the distance between the two edges of the crack at each measurement point is calculated. Let the pixel coordinates of the two edges of the crack at the j-th measurement point be respectively... and The crack width at that measurement point is... ; Take the average value as the crack width ,in This represents the number of measurement points.
[0026] Feature extraction and analysis are performed on the initially detected crack areas to obtain the texture feature parameters of the cracks, specifically including: The gray-level co-occurrence matrix is calculated using the gray-level co-occurrence matrix method. ,in and Grayscale For pixel spacing, For the direction angle, calculate the entropy, contrast, and correlation of the gray-level co-occurrence matrix: ; ; ; in For gray levels, and They are respectively The mean of the rows and columns, and They are respectively The standard deviation of the rows and columns; The roughness of the cracked surface is assessed based on these statistical measures as follows: ; in, , and These are the weights for entropy, contrast, and correlation, respectively.
[0027] The preset crack evaluation criteria are set based on the geometric and textural feature parameters of the crack, specifically as follows: When the crack length > ,width > At that time, it was determined to be a severe crack, among which , The threshold for determining severe cracks is determined based on bridge design requirements and usage experience, and is preset in the system. When the crack length ≤ ≤ ,width ≤ ≤ Meanwhile, crack roughness > At that time, it was determined to be a moderate crack, of which , , The threshold for determining moderate cracks is also preset in the system; All other cases were determined to be minor cracks.
[0028] The generated bridge crack detection report includes the following: The specific location of the crack in the bridge, such as the pier, bridge deck, or beam, is determined by mapping the image coordinates to the actual structure of the bridge. Historical inspection records and trend analysis of cracks, by comparing the results of multiple inspections, let the first... The crack length detected was and width are Calculate the rate of change of crack length and the rate of change of crack width: ; ; Based on the rate of change of crack length and the rate of change of crack width, determine whether the crack is propagating or worsening; We provide repair solutions and cost estimates for cracks of varying severity.
[0029] In a second aspect of the invention, a bridge crack detection system is also provided, comprising: The acquisition module is used to acquire image data of the bridge surface. It uses a multispectral imaging device to continuously scan and capture images of the bridge surface under different spectral bands to obtain composite image data containing visible light, near-infrared and mid-infrared bands. The preprocessing module is used to preprocess the acquired composite image data. The preprocessing includes image denoising, image enhancement and image registration operations to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands. The detection module is used to perform preliminary crack detection based on the preprocessed image data and a crack recognition model built using deep learning algorithms, and to identify possible crack areas in the image. The extraction module is used to extract and analyze features from the initially detected crack areas to obtain the geometric and texture feature parameters of the cracks. The evaluation module is used to assess the severity of cracks based on their geometric and textural characteristics, combined with preset crack evaluation criteria. The report generation module is used to generate a bridge crack detection report based on the crack severity assessment results. The report includes the crack location, size, severity, and recommended repair measures.
[0030] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. A method for detecting cracks in bridges, characterized in that, include: Bridge surface image data was collected, and multispectral imaging equipment was used to continuously scan and photograph the bridge surface under different spectral bands to obtain composite image data including visible light, near infrared and mid-infrared bands; The acquired composite image data is preprocessed, including image denoising, image enhancement, and image registration operations, to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands; Based on the preprocessed image data, a crack recognition model constructed using deep learning algorithms is used to perform preliminary crack detection and identify possible crack areas in the image. Feature extraction and analysis are performed on the initially detected crack areas to obtain the geometric and texture feature parameters of the cracks; The severity of the crack is assessed based on its geometric and textural characteristics, combined with a pre-defined crack assessment standard. Based on the assessment results of crack severity, a bridge crack detection report is generated, which includes the location, size, severity of cracks, and recommended repair measures.
2. The method for detecting bridge cracks according to claim 1, characterized in that, Image denoising of the composite image data specifically includes: Wavelet decomposition was performed on the composite image data to obtain wavelet coefficients at different scales; Based on the statistical characteristics of wavelet coefficients, a pre-set threshold is retrieved from the database; The wavelet coefficients are compared with a pre-set threshold. Wavelet coefficients smaller than the threshold are set to zero, while wavelet coefficients larger than the threshold are retained or shrunk. Wavelet reconstruction is performed on the processed wavelet coefficients to obtain the denoised image data.
3. The method for detecting bridge cracks according to claim 1, characterized in that, Image enhancement of the composite image data specifically includes: Adaptive histogram equalization is performed on the denoised composite image data. The cumulative distribution function of the local region is determined by calculating the gray-level histogram of the denoised image; Based on the cumulative distribution function, the gray-level mapping relationship of local regions is adaptively adjusted to obtain an enhanced image.
4. The method for detecting bridge cracks according to claim 1, characterized in that, The image registration operation employs a feature point-based registration method, with the following specific steps: Feature points, such as corner points and edge points, are extracted from images of different spectral bands using the Harris corner detection algorithm, whose response function is: ,in , and These are the gradients of the image in the x and y directions, respectively. This is an empirical constant, with a value of 0.
05. Points with response function values greater than a set threshold are considered corner points, while points with smaller or negative response function values are considered edge points. represent The determinant value of the matrix, and represent The sum of the elements on the main diagonal of a matrix; The extracted feature points are matched to obtain feature point pairs. The nearest neighbor ratio matching method is used, and the nearest neighbor feature point p in the first image is... The second nearest neighbor feature point is ,like , The matching threshold is set to 0.6, which means p and They are matched feature point pairs; Calculate the transformation parameters and translation parameters between images based on feature point pairs. Rotation parameters Given the scaling parameter s, the transformation equation is solved using the least squares method: ； in and These are the coordinates of the matched feature point pairs in the two images, respectively; The image is transformed according to the transformation parameters to achieve accurate registration of images in different spectral bands.
5. The method for detecting bridge cracks according to claim 1, characterized in that, The specific steps for constructing a crack recognition model using deep learning algorithms are as follows: Collect a large amount of bridge crack image data and non-crack image data to construct a training dataset. and test dataset The number of crack images in the training dataset is The number of non-crack images is The number of crack images in the test dataset is The number of non-crack images is ; Choose a convolutional neural network structure and initialize the network parameters. Let the kernel size of the convolutional layer be k×k, the stride be s, the padding be p, and the number of neurons in the fully connected layer be n. The deep learning network is trained using the training dataset, and the network parameters are updated using the backpropagation algorithm. The cross-entropy loss function is used as follows: ； in, For the sample size, For the true label of the sample, The predicted label for the sample; Until the network's recognition accuracy on the test dataset Reaching the preset threshold, among which For a real example, For a true negative example, This is a false positive example. This is a false negative example; The trained crack recognition model is optimized using a model pruning method to obtain a new crack recognition model.
6. The method for detecting bridge cracks according to claim 1, characterized in that, Feature extraction and analysis are performed on the initially detected crack areas to obtain the geometric feature parameters of the cracks, specifically including: An image edge detection-based method is employed. First, the crack edges are detected. Then, the sum of distances between adjacent pixels on the edge is calculated. and Calculate the distance between them: ； The crack length is ,in This represents the number of pixels on the edge. In the direction perpendicular to the crack, multiple measurement points are selected, and the distance between the two edges of the crack at each measurement point is calculated. Let the pixel coordinates of the two edges of the crack at the j-th measurement point be respectively... and The crack width at that measurement point is... ; Take the average value as the crack width ,in This represents the number of measurement points.
7. The method for detecting bridge cracks according to claim 6, characterized in that, Feature extraction and analysis are performed on the initially detected crack areas to obtain the texture feature parameters of the cracks, specifically including: The gray-level co-occurrence matrix is calculated using the gray-level co-occurrence matrix method. ,in and Grayscale For pixel spacing, For the direction angle, calculate the entropy, contrast, and correlation of the gray-level co-occurrence matrix: ； ； ； in For gray levels, and They are respectively The mean of the rows and columns, and They are respectively The standard deviation of the rows and columns; The roughness of the cracked surface is assessed based on these statistical measures as follows: ； in, , and These are the weights for entropy, contrast, and correlation, respectively.
8. The method for detecting bridge cracks according to claim 1, characterized in that, The preset crack evaluation criteria are set based on the geometric and textural feature parameters of the crack, specifically as follows: When the crack length > ,width > At that time, it was determined to be a severe crack, among which , The threshold for determining severe cracks is determined based on bridge design requirements and usage experience, and is preset in the system. When the crack length ≤ ≤ ,width ≤ ≤ Meanwhile, crack roughness > At that time, it was determined to be a moderate crack, of which , , The threshold for determining moderate cracks is also preset in the system; All other cases were determined to be minor cracks.
9. The method for detecting bridge cracks according to claim 1, characterized in that, The generated bridge crack detection report includes the following: The specific location of the crack in the bridge, such as the pier, bridge deck, or beam, is determined by mapping the image coordinates to the actual structure of the bridge. Historical inspection records and trend analysis of cracks, by comparing the results of multiple inspections, let the first... The crack length detected was and width are Calculate the rate of change of crack length and the rate of change of crack width: ； ； Based on the rate of change of crack length and the rate of change of crack width, determine whether the crack is propagating or worsening; We provide repair solutions and cost estimates for cracks of varying severity.
10. A bridge crack detection system, used to implement the bridge crack detection method as described in any one of claims 1-9, characterized in that, include: The acquisition module is used to acquire image data of the bridge surface. It uses a multispectral imaging device to continuously scan and photograph the bridge surface under different spectral bands to obtain composite image data including visible light, near-infrared and mid-infrared bands. The preprocessing module is used to preprocess the acquired composite image data. The preprocessing includes image denoising, image enhancement, and image registration operations to eliminate noise interference in the image, enhance crack feature information, and accurately register images of different spectral bands. The detection module is used to perform preliminary crack detection based on preprocessed image data and a crack recognition model constructed using a deep learning algorithm, and to identify possible crack areas in the image. The extraction module is used to extract and analyze features from the initially detected crack area to obtain the geometric and texture feature parameters of the crack. An evaluation module is used to evaluate the severity of a crack based on its geometric and textural features, combined with a preset crack evaluation standard. The report generation module is used to generate a bridge crack detection report based on the crack severity assessment results. The report includes the crack location, size, severity, and recommended repair measures.