A method for concrete bridge damage detection and residual life prediction based on machine vision

Abstract

The application provides a concrete bridge damage detection and residual life prediction method based on machine vision, comprising: obtaining a concrete bridge surface image and corresponding structure coordinates; embedding the structure coordinates into the concrete bridge surface image to obtain a coordinate surface image; inputting the coordinate surface image into an image recognition model to recognize the image type and corresponding structure coordinates, wherein the image type comprises an explicit damage image and a marked image, and the marked image is a marked image generated by human testing; determining each first life influence parameter according to the explicit damage image and the corresponding structure coordinates; determining each second life influence parameter according to each marked image and the corresponding structure coordinates; and estimating the residual life of the concrete bridge according to each first life influence parameter and each second life influence parameter. Through the implementation of the application, the comprehensiveness of the concrete bridge damage detection and the accuracy of the residual life prediction can be improved.

Description

Technical Field

[0001] This invention belongs to the field of concrete bridge life prediction technology, specifically relating to a method for damage detection and remaining life prediction of concrete bridges based on machine vision. Background Technology

[0002] The remaining service life assessment of concrete bridges is a crucial step in ensuring their safe operation. Traditional bridge inspection and life assessment methods have many shortcomings. Traditional manual inspection methods suffer from strong subjectivity, numerous blind spots, and data discretization, making it difficult to meet the refined inspection needs of long concrete bridges. While existing computer vision-based inspection technologies have made some progress in damage identification, in the life prediction stage, due to the limitations of computer vision, current methods mostly rely on single, visible damage images, such as concrete cracking, steel corrosion, fissures, spalling, and exposed rebar. These damage images only reflect the external damage, and using only a single type of image information to predict the remaining service life can lead to significant biases in the remaining service life assessment. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide a machine vision-based method for damage detection and remaining life prediction of concrete bridges, so as to meet the need to improve the accuracy of the remaining life of concrete bridges.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] According to a first aspect, the present invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges, comprising: acquiring a surface image of a concrete bridge and its corresponding structural coordinates; explicitly embedding the structural coordinates into the surface image of the concrete bridge to obtain a coordinateized surface image; inputting the coordinateized surface image into an image recognition model to identify the image type and the corresponding structural coordinates, wherein the image type includes visible damage images and marked images, wherein the marked images are marked images generated by human testing; determining each first life-affecting parameter based on each visible damage image and its corresponding structural coordinates; determining each second life-affecting parameter based on each marked image and its corresponding structural coordinates; and estimating the remaining life of the concrete bridge based on each first life-affecting parameter and each second life-affecting parameter.

[0006] According to a second aspect, the present invention provides a machine vision-based device for damage detection and remaining life prediction of concrete bridges, comprising: an acquisition module for acquiring surface images of concrete bridges and their corresponding structural coordinates; a coordinateization module for embedding the structural coordinates into the surface images of concrete bridges to obtain coordinateized surface images; a recognition module for inputting the coordinateized surface images into an image recognition model to recognize image types and corresponding structural coordinates, wherein the image types include visible damage images and marked images, and the marked images are marked images generated by human testing; a first life influence parameter determination module for determining each first life influence parameter based on each visible damage image and its corresponding structural coordinate; a second life influence parameter determination module for determining each second life influence parameter based on each marked image and its corresponding structural coordinate; and a remaining life determination module for estimating the remaining life of the concrete bridge based on each first life influence parameter and each second life influence parameter.

[0007] According to a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor performing the steps of the machine vision-based method for detecting damage and predicting the remaining life of concrete bridges as described in the first aspect or any embodiment of the first aspect.

[0008] According to a fourth aspect, embodiments of the present invention provide a computer storage medium storing computer instructions that, when executed by a processor, implement the steps of the machine vision-based method for damage detection and remaining life prediction of concrete bridges as described in the first aspect or any embodiment of the first aspect.

[0009] This invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges. Compared with existing computer vision detection techniques that rely solely on visible damage images, this method introduces labeled images. By analyzing these labeled images, second life-influencing parameters (such as concrete carbonation depth and steel corrosion rate) are obtained. Combined with first life-influencing parameters (such as geometric features like cracks and spalling) determined from visible damage images, a complete life assessment system is constructed from two dimensions: apparent damage and material degradation mechanisms. This multi-source data fusion approach can capture visible surface damage and deeply analyze the internal degradation process of the material, overcoming the shortcomings of traditional methods that only focus on visible features. This improves the comprehensiveness of concrete bridge damage detection and the accuracy of remaining life prediction.

[0010] Other advantages, objectives, and features of the invention will be set forth in the following description and will be apparent to those skilled in the art in some respects, or may be learned by practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0011] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following figures are provided for illustration:

[0012] Figure 1 This is a flowchart illustrating a specific example of a machine vision-based method for damage detection and remaining life prediction of concrete bridges in this invention.

[0013] Figure 2 This is a flowchart illustrating a specific example of comparing a digital model of a concrete bridge with a standard model of a concrete bridge to determine multiple structural differences in this invention.

[0014] Figure 3 This is a flowchart illustrating a specific example of estimating the remaining lifespan of a concrete bridge based on various first lifespan influence parameters, various second lifespan influence parameters, and multiple structural difference points in this invention.

[0015] Figure 4 This is a schematic block diagram of a specific example of an electronic device in an embodiment of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can also refer to the internal connection of two components; and they can refer to a wireless connection or a wired connection. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0018] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0019] This invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges, such as... Figure 1 As shown, it includes:

[0020] S101, Obtain the surface image of the concrete bridge and its corresponding structural coordinates;

[0021] S102, the structural coordinates are explicitly embedded into the concrete bridge surface image to obtain a coordinateized surface image;

[0022] S103, Input the coordinated surface image into the image recognition model to identify the image type and the corresponding structural coordinates. The image types include visible damage images and marked images. The marked images are marked images generated by human testing.

[0023] S104. Based on the images of each type of visible damage and the corresponding structural coordinates, determine each first lifetime influence parameter.

[0024] S105, Based on each marker class image and the corresponding structural coordinates, determine each second lifetime influence parameter;

[0025] S106. Estimate the remaining life of the concrete bridge based on the various first life-affecting parameters and the various second life-affecting parameters.

[0026] For example, images of the concrete bridge surface can be acquired using a drone equipped with an RTK module. A drone featuring a high-precision RTK-GPS module, a high-definition camera, and an IMU is selected. Drone mapping software plans the flight path and tilts the bridge at multiple angles. At each shutter trigger, the three-dimensional coordinates (X, Y, Z) and attitude data (heading angle, pitch angle, roll angle) at the moment of capture are simultaneously recorded. After acquisition, the image sequence is processed using Pix4Dmapper software. Feature point matching and bundle adjustment are used to generate a three-dimensional point cloud of the bridge surface, establishing the correspondence between image pixels and structural coordinates. Simultaneously, ground control points are used for coordinate calibration to obtain the concrete surface image and its corresponding structural coordinates.

[0027] Next, the obtained concrete surface image and the corresponding structural coordinates are fused together. That is, the structural coordinate information is incorporated into the concrete surface image to construct a coordinate-based surface image. Specifically, the mapping table between image pixels and structural coordinates can be read, and the three-dimensional coordinates (X,Y,Z) corresponding to each target pixel can be superimposed on the non-critical area of ​​the image as white semi-transparent text. The target pixels can be pixels determined according to preset rules. For example, if the image is divided into grids of the target size, then the target pixels are the pixels at the four corners of the grid lines. This embodiment does not limit the method of determining the target pixels, and those skilled in the art can determine it as needed.

[0028] The coordinate-based surface image is segmented into multiple images. These segmented images are then input into an image recognition model, which outputs the image type and coordinates of each segmented image. The output includes images representing visible damage and marked images. Visible damage images represent visual losses to the bridge, such as cracks and spalling, while marked images represent images obtained through artificial testing, such as the color change of phenolphthalein reagent at a specific location. This image recognition model can be based on the YOLO model, with modifications to the output layer to include both visible damage and marked images. An OCR branch is added after the YOLO model to achieve coordinate recognition of the coordinate-based surface image. In constructing the YOLO model, 5000 bridge surface images are used as the original dataset, with 3000 images containing visible damage (e.g., cracks, spalling), 1000 images containing various artificially marked patterns, and 1000 images with clean backgrounds. Data augmentation operations are performed on the dataset, including random flipping, brightness adjustment, Gaussian blur, and random cropping, expanding the dataset to 20,000 images. Training was performed on an NVIDIA RTX4090 GPU with a batch size of 16 and 300 epochs. A cosine annealing decay strategy was used, with the validation set loss as the basis for early stopping.

[0029] For images showing visible damage, geometric features such as crack length, width, and spalling area are extracted, and spatial distribution features are also extracted. Coordinates are mapped to bridge structural parts (e.g., mid-span, supports, web), generating a multi-dimensional feature vector containing damage type, geometric parameters, and location, which serves as the first life-influencing parameter. For labeled images, relevant parameters and corresponding structural coordinates are extracted based on the testing purpose of the labeled images, serving as the second life-influencing parameter. The remaining life of a concrete bridge can be estimated by merging the first and second life-influencing parameters into a single dataset. A prediction model is constructed using XGBoost or LSTM neural networks. The remaining life label is the bridge's design life minus the years already served, combined with expert evaluation corrections. A sample is constructed using data of the same type as the first and second life-influencing parameters. This sample and corresponding remaining life label are input into the prediction model, and the root mean square error (RMSE) is used as the error metric for training. A trained prediction model is then obtained. The dataset formed by merging the first and second life-influencing parameters is input into this prediction model to obtain the predicted remaining life.

[0030] This invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges. Compared with existing computer vision detection techniques that rely solely on visible damage images, this method introduces labeled images. By analyzing these labeled images, second life-influencing parameters (such as concrete carbonation depth and steel corrosion rate) are obtained. Combined with first life-influencing parameters (such as geometric features like cracks and spalling) determined from visible damage images, a complete life assessment system is constructed from two dimensions: apparent damage and material degradation mechanisms. This multi-source data fusion approach can capture visible surface damage and deeply analyze the internal degradation process of the material, overcoming the shortcomings of traditional methods that only focus on visible features. This improves the comprehensiveness of concrete bridge damage detection and the accuracy of remaining life prediction.

[0031] As an optional implementation, based on each visible damage type image and the corresponding structural coordinates, each first lifetime influence parameter is determined, including:

[0032] Extract geometric feature parameters from images of visible damage;

[0033] Determine whether there are identified random speckle images at the structural coordinates of the visible damage images; random speckle images are one type of labeled images.

[0034] If a random speckle pattern exists, the mechanical parameters are analyzed based on the random speckle pattern.

[0035] The mechanical parameters, geometric feature parameters, and corresponding structural coordinates at the structural coordinates of each visible damage image are input into a pre-established first damage attention model for importance ranking to determine the key visible damage images.

[0036] If no random speckle image exists, the geometric feature parameters and corresponding structural coordinates of each visible damage image are input into a pre-established second damage attention model for importance ranking to determine the key visible damage images.

[0037] The mechanical parameters, geometric feature parameters, and damage location at the corresponding coordinates of the key visible damage image are used as the first lifetime influence parameters.

[0038] For example, for crack damage, the crack contour is extracted using an edge detection algorithm, and the crack length, width, area, and orientation angle are calculated; for spalling damage, the spalling area is separated using threshold segmentation, and parameters such as spalling area, perimeter, and depth are calculated; for exposed rebar damage, the rebar contour is refined using morphological operations, and parameters such as exposed rebar length and corrosion area ratio are calculated. These calculated geometric feature parameters are stored in a data structure and associated with the structural coordinates (X, Y, Z) of the corresponding visible damage image.

[0039] Then, it is determined whether a random speckle image exists at the structural coordinates of the visible damage images. Random speckle images are marker images with randomly distributed speckle patterns, manually sprayed or printed on the surface of concrete bridges, primarily used for non-contact measurement of structural deformation and mechanical parameters. Specifically, random speckle images are selected from the output of the image recognition model, and their structural coordinates are extracted. For each structural coordinate of a visible damage image, its Euclidean distance to the structural coordinates of all random speckle images is calculated. A distance threshold, such as 0.5 meters, is set. If the distance between the coordinates of a random speckle image and the coordinates of a visible damage image is less than this threshold, then a random speckle image is determined to exist at the structural coordinates of the visible damage image; otherwise, it does not exist.

[0040] When random speckle images are available, mechanical parameters are analyzed based on them. Specifically, firstly, subpixel-level corner detection is performed on the random speckle images to obtain the coordinates of feature points in the speckle pattern. Then, the random speckle images acquired during the previous capture are retrieved. Based on the bridge loading state at the time of the previous capture and the current bridge loading state, random speckle images of the same area are compared. The displacement vector field of the feature points is calculated through feature point matching, thereby obtaining the strain distribution in that area. According to Hooke's law, combined with the elastic modulus of concrete and steel reinforcement, strain is converted into stress, yielding mechanical parameters such as the magnitude and direction of the principal stresses and shear stresses in that area. Simultaneously, by analyzing the displacement over time, dynamic parameters such as the structure's vibration frequency and damping ratio are calculated, and these mechanical parameters are correlated with the corresponding coordinates of the visible damage images.

[0041] The mechanical parameters, geometric feature parameters, and corresponding structural coordinates at the structural coordinates of each visible damage image are input into a pre-established first damage attention model for importance ranking to determine key visible damage images. The first damage attention model adopts a Transformer-based structure. The input layer normalizes the mechanical parameters, geometric feature parameters, and structural coordinates and concatenates them into a feature vector as the model input. The intermediate layers of the model contain multiple multi-head attention mechanism layers and feedforward neural network layers. The multi-head attention mechanism is used to capture the correlation between different features, highlighting key features by calculating the attention weights of different features. The feedforward neural network layer performs further nonlinear transformations and feature extraction on the features. The output layer outputs the importance score of each visible damage image through a fully connected layer and a softmax function. During model training, the actual reduction in the bridge's load-bearing capacity or the severity of damage assessed by experts is used as labels. Training is performed using a cross-entropy loss function and the Adam optimizer, iterating until the loss function converges. Based on the importance scores output by the model, all visible damage images are ranked, and a certain proportion of visible damage images with higher scores are selected as key visible damage images, such as the top 30% of visible damage images.

[0042] When no random speckle images are available, the geometric feature parameters and corresponding structural coordinates of each visible lesion image are input into a pre-established second lesion attention model for importance ranking to determine key visible lesion images. The second lesion attention model has a similar structure to the first lesion attention model, but its input layer only contains geometric feature parameters and structural coordinates. After normalization and concatenation, these parameters are input into the model, where feature processing is performed through a multi-head attention mechanism and a feedforward neural network layer. The output layer outputs the importance score for each visible lesion image. During training, the lesion development speed predicted based on geometric features and historical lesion data is used as a label, and training is performed using a mean squared error loss function and an Adam optimizer until the loss converges. Key visible lesion images are selected based on their importance scores.

[0043] Finally, the mechanical parameters, geometric feature parameters, and damage location (mapped to specific parts of the bridge, such as the mid-span web, support plate, etc.) at the coordinates corresponding to the key visible damage images are used as the first life-affecting parameters.

[0044] This invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges. It not only focuses on the geometric features of visible damage but also incorporates random speckle image analysis to determine mechanical parameters, overcoming the limitations of single geometric analysis. Furthermore, based on the presence of random speckle images, it flexibly utilizes first and second damage attention models for importance ranking, avoiding assessment bias caused by missing data and significantly improving assessment efficiency and accuracy. Compared to traditional methods that rely solely on human experience to judge damage severity, this approach combines geometric features, mechanical parameters, and intelligent algorithms to provide more scientific and reliable first-stage life-influencing parameters for assessing the remaining life of concrete bridges.

[0045] As an optional implementation, the artificially generated marker image is the color image produced when the target reagent is encountered at the target location hole. Based on each marker image and its corresponding structural coordinates, various second lifetime influence parameters are determined, including:

[0046] Image preprocessing is performed on the marked images to obtain the target images. Color regions are extracted from the target images and divided into levels according to different color values ​​to obtain various color level regions. Statistical analysis is performed on each color level region to determine the proportion of each color level region to the entire hole. The various color level regions and their proportions to the entire hole are input into a pre-trained machine learning model to predict the carbonation depth of the benchmark concrete and the corrosion rate of the benchmark steel reinforcement. The temperature and humidity of the environment where the concrete bridge is located are obtained. The environmental temperature and humidity and the carbonation depth of the benchmark concrete are input into a pre-established concrete carbonation depth correction model to obtain the corrected concrete carbonation depth. The environmental temperature and humidity and the corrosion rate of the benchmark steel reinforcement are input into a pre-established steel reinforcement corrosion rate correction model to obtain the corrected steel reinforcement corrosion rate. The corrected concrete carbonation depth, the corrected steel reinforcement corrosion rate, and the corresponding structural coordinates are used as the second lifetime influence parameters.

[0047] For example, the marked images are preprocessed by converting them from the RGB color space to the HSV color space using image processing software, reducing noise through Gaussian blurring, and then performing binarization using an adaptive thresholding method to highlight the marked areas. Next, contour detection is used to extract the hole contours and filter out interference, cropping the regions of interest (ROIs) for subsequent analysis. Color regions are extracted within the ROIs of the target image. All pixels within the ROI are traversed, and RGB values ​​are extracted and converted to hue values ​​H in the HSV space. Based on the typical color characteristics of concrete carbonization and steel corrosion, multiple color level ranges are pre-defined. For example, reddish-brown represents severe corrosion, yellowish-brown represents medium corrosion, bluish-gray represents no carbonization, and grayish-white represents carbonization. This can be achieved using threshold-based segmentation methods or clustering algorithms (such as K-means). A color level mask image is generated, and the number of pixels in each color level region is counted and divided by the total number of pixels in the ROI to obtain the proportion of each color level region to the holes.

[0048] The proportion of each color level region is used as a feature vector and input into a pre-trained machine learning model, such as a random forest regression model. This model is trained based on laboratory standard specimen data, which includes different carbonation depths and steel corrosion rates. The model outputs the benchmark concrete carbonation depth and benchmark steel corrosion rate by analyzing the correlation between image color features and measured carbonation depth and corrosion rate.

[0049] Meanwhile, ambient temperature and relative humidity are collected by temperature and humidity sensors deployed at the bridge site. This temperature and humidity data, along with a baseline carbonization depth, are input into a carbonization depth correction model. For example, based on an empirical formula using Fick's diffusion law, the corrected carbonization depth is obtained. Specifically, the carbonization depth correction model is as follows:

[0050] d corr =d0·f T ·f RH ;

[0051] Where, d corr The corrected carbonization depth is given by d0, where d0 is the baseline carbonization depth and f is the base carbonization depth. T f is the temperature correction factor. RH Humidity correction factor;

[0052]

[0053] E a R is the activation energy of the carbonization reaction, R is the gas constant, which is generally 40 kJ / mol, T0 is the standard temperature, and T is the actual ambient temperature.

[0054]

[0055] RH0 is the standard humidity, RH is the actual ambient humidity, and n is an empirical coefficient, which is generally taken as 1.5 to 2.0.

[0056] Temperature and humidity data, along with the baseline steel corrosion rate, are input into the steel corrosion rate correction model. For example, by combining electrochemical theory, the corrected steel corrosion rate is obtained. Specifically:

[0057] r corr =r0·g T ·g RH ;

[0058] Where, r corr The corrected corrosion rate is given by r0, where r0 is the baseline corrosion rate and g is the base rate. T For temperature correction factor, g RH This is the humidity correction factor.

[0059]

[0060] β is an empirical parameter, usually taken as 5000 to 6000, T0 is the standard temperature, and T is the actual temperature.

[0061]

[0062] RH0 is the standard humidity, RH is the actual humidity, and m is an empirical coefficient, which is generally taken as 2.0 to 3.0.

[0063] Finally, the corrected carbonization depth, the corrected steel corrosion rate, and their corresponding structural coordinates are integrated as a second lifetime influence parameter.

[0064] This invention provides a machine vision-based method for damage detection and remaining life prediction of concrete bridges. It achieves non-destructive detection through the color features of marked images, avoiding the destructive nature of traditional detection methods. It quantifies the relationship between color grade proportions and material degradation parameters, and combines real-time temperature and humidity to dynamically correct baseline parameters, accurately reflecting the actual impact of the environment on concrete carbonation and steel corrosion, thus improving the accuracy of prediction results. Furthermore, by integrating marked image and visible damage image data, it constructs a complete life-affecting parameter system from both appearance features and material degradation mechanisms, resulting in a more comprehensive assessment.

[0065] As an optional implementation, before estimating the remaining life of a concrete bridge based on various first life-affecting parameters and various second life-affecting parameters, the following steps are taken: acquiring three-dimensional point cloud data of the concrete bridge; reconstructing a digital model of the concrete bridge based on the three-dimensional point cloud data; comparing the digital model of the concrete bridge with a standard model of the concrete bridge to determine multiple structural difference points; and estimating the remaining life of the concrete bridge based on various first life-affecting parameters and various second life-affecting parameters, including estimating the remaining life of the concrete bridge based on various first life-affecting parameters, various second life-affecting parameters, and multiple structural difference points.

[0066] For example, the 3D point cloud data of a concrete bridge can be obtained through laser scanning. The point cloud data is imported into professional 3D modeling software, which automatically identifies the spatial distribution characteristics of the point cloud and connects the discrete points into a triangular mesh using a triangulation algorithm, thus initially constructing a bridge surface model. For complex structural parts of the bridge, such as the connection between piers and main beams, and the prestressed anchorage area, the point cloud positions are manually adjusted, and missing data points are supplemented to ensure the completeness of model details. To make the model more realistic, corresponding material properties are assigned to the digital model based on the actual color and texture of the bridge concrete. Simultaneously, the continuity and smoothness of the model surface are checked, and areas with holes or overlaps are repaired and optimized, ultimately generating a digital model that accurately reflects the actual geometry of the bridge.

[0067] Then, the digital model of the concrete bridge is compared with the standard model. Specifically, as an optional implementation, the digital model of the concrete bridge is compared with the standard model of the concrete bridge to identify multiple structural differences, such as... Figure 2 As shown, it includes:

[0068] S201, For each point in the point cloud, determine a local region centered on that point, and represent the point cloud within the local region as a graph structure;

[0069] S202 uses a graph convolutional network to encode the graph structure and learn local feature descriptors of the point cloud, which contain local geometric features.

[0070] S203, the local feature descriptor of the point cloud and the semantic label information of the point cloud are fused to obtain the context feature descriptor, and the semantic label information of the point cloud is used to characterize the environmental information of the bridge where the point cloud is located;

[0071] S204: By calculating the local feature descriptors of points, coarse matching based on feature similarity is performed to establish a preliminary point-to-point correspondence.

[0072] S205, calculate the similarity between context feature descriptors of corresponding point pairs, including local geometric feature similarity and semantic label consistency;

[0073] S206, In the digital model of the concrete bridge and the standard model of the concrete bridge, respectively, obtain the local geometric features of the point cloud corresponding to the point pairs with similarity exceeding the first threshold.

[0074] S207. Based on the local geometric features of the point cloud, determine the geometric statistical features of the local region where the point cloud is located. The geometric statistical features include at least one of the region average normal vector, curvature distribution, and point density.

[0075] S208. By combining the geometric and statistical features of the point cloud in the digital model of the concrete bridge with the geometric and statistical features of the point cloud in the standard model of the concrete bridge, the degree of matching of the point cloud pairs is determined, and the matching point cloud pairs are obtained.

[0076] S209, Based on the matching point cloud pairs, determine multiple structural difference points.

[0077] For example, when comparing a digital model of a concrete bridge with a standard model to determine structural differences, a spherical region with radius r is first defined around each point in the point cloud of the digital model as a local region. This radius r is set based on the point cloud density and the complexity of the bridge structure, typically between 0.1 and 0.5 meters, ensuring that the local region contains a sufficient number of points and reflects structural details. The point cloud within the local region is then constructed as a graph structure, with each point as a node. The connection relationships between edges are determined by calculating the Euclidean distance between nodes; two points with a distance less than a threshold d are connected, with the value of d determined by referencing the average spacing of the point cloud.

[0078] Next, a graph convolutional network is used to encode the constructed graph structure. The network input is the 3D coordinates of the nodes. Through multiple convolutional operations, the spatial relationship between each node and its neighboring nodes is learned, and local feature descriptors of the point cloud are output. These descriptors contain geometric feature information such as the geometric shape and concavity / convexity of the local region. To further enrich the feature representation, semantic label information of the point cloud is integrated. The semantic label information is obtained through manual annotation or semantic segmentation model prediction, covering environmental information such as the location of the point cloud on the bridge (e.g., main beam, pier, railing) and functional areas (e.g., stress zone, non-stress zone). The semantic labels are encoded into vector form and then concatenated with the local feature descriptors to obtain the context feature descriptors.

[0079] Then, coarse matching based on feature similarity is performed by calculating the Euclidean distance or cosine similarity between the local feature descriptors of the point cloud. Coarse matching can quickly establish preliminary point-to-point correspondences in a large amount of point cloud data, narrowing the search range for subsequent fine matching and improving overall matching efficiency. Specifically, a similarity threshold T1 is set, and preliminary correspondences are established for point pairs with similarity greater than T1. Fine matching is then performed on the corresponding point pairs after coarse matching, calculating the similarity between their context feature descriptors. This similarity includes two parts: first, the similarity of local geometric features, calculated by comparing the differences in geometric parameters in the local feature descriptors; and second, the consistency of semantic labels.

[0080] The determination of semantic label consistency can be achieved by: inputting the semantic label of any point into a pre-established semantic library of standard concrete bridge models for querying; when a standard semantic label is matched, the semantic label similarity is set to the first similarity; the semantic library of standard concrete bridge models contains standard semantic labels and a semantic association graph; when no standard semantic label is matched, the semantic label is broken down into the smallest semantic units, and the associated terms of each smallest semantic unit are queried in the semantic association graph; the associated terms of each semantic unit are exhaustively combined and matched with the standard semantic label a second time; when a standard semantic label is matched a second time, the semantic label similarity is set to the second similarity; when no standard semantic label is matched a second time, the semantic label similarity is set to the third similarity.

[0081] Specifically, when performing semantic tag similarity calculation, the semantic tag of any point is first input into the semantic database of the standard model of concrete bridges for querying. For example, if the semantic tag "precast T-beam" is input, and there is a completely identical standard semantic tag "precast T-beam" in the semantic database, then the similarity of that semantic tag is directly determined as the first similarity (e.g., set to 1, indicating a complete match). If the input semantic tag "steel-concrete composite beam flange" does not find a match in the semantic database, it needs to be broken down into the smallest semantic units, namely "steel-concrete", "composite beam", and "flange". Subsequently, the associated terms of each smallest semantic unit are queried in the semantic association graph. For example, the associated terms of "steel-concrete" include "reinforced concrete", "composite beam" is associated with "composite beam", and "flange" is associated with "edge member", etc. Then, these associated terms are exhaustively combined to generate a series of combined results such as "reinforced concrete composite beam edge member" and "steel-concrete composite beam flange", and a second matching is performed with the standard semantic tag. If the combined result "reinforced concrete composite beam flange" has a corresponding standard semantic tag in the semantic database, then the similarity of this semantic tag is set to the second similarity (e.g., 0.8, indicating a high match). If no standard semantic tag is matched in all exhaustive combinations, such as when the generated combined results have no corresponding items in the semantic database, then the semantic tag similarity is set to the third similarity (e.g., 0, indicating no match). Through this method of first splitting, then exhaustively combining, and finally matching twice, the similarity between the input semantic tag and the standard semantic tag is systematically and comprehensively determined.

[0082] For local geometric features, geometric attributes such as the normal vector and curvature of points can be calculated. Measurement methods such as Euclidean distance and cosine similarity can be used to compare the differences in normal vectors, curvature, and other attributes between point pairs. In geometric feature similarity calculation, the degree of difference and similarity have an inverse mapping relationship. Through mathematical transformation, numerical differences are converted into similarity values ​​in the 0-1 interval. Specifically, for continuous scalar features such as curvature and point density, Euclidean distance can be used to calculate the difference in feature values ​​between point pairs, d = |x1 - x2|, and the maximum possible difference value D can be pre-set. max After normalization, the similarity is obtained as follows: For example, if the curvature of point A is 0.05 and the curvature of point B is 0.03, then if D... max =0.1, then the similarity is 0.8; for vector-type features such as normal vectors, the cosine similarity is used to calculate the cosine value of the angle θ between the vectors. The result directly reflects the degree of directional similarity, with a value range of [-1,1]. Usually, the absolute value or... Adjust to the [0,1] interval, such as the cosine similarity between the normal vector of point A (0,0,1) and the normal vector of point B (0,0.6,0.8) being 0.8; in addition, a tolerance threshold δ can be set for specific features. If the difference d≤δ, the similarity is 1; otherwise, it is adjusted to the [0,1] interval. Linear decay is employed; for example, when the point cloud density difference threshold is set to 5 points / cubic meter, a point pair with a difference of 3 points has a similarity of 1. Using the above method, the degree of difference in geometric features between point pairs is quantified into a similarity value in the [0,1] interval. Then, the similarity values ​​of various geometric features are averaged to obtain the local geometric feature similarity. The calculation of local geometric feature similarity quickly eliminates obviously mismatched point pairs by comparing geometric features at the single-point level (such as normal vector direction and curvature magnitude), avoiding subsequent invalid calculations.

[0083] The similarity between context feature descriptors of corresponding point pairs can be represented by a weighted sum of the similarity of local geometric features and semantic consistency. Then, based on the overall similarity obtained by the weighted sum, point pairs with similarity exceeding T2 are selected in the digital model and standard model of concrete bridges, and the local geometric features of the corresponding point clouds in these point pairs are obtained.

[0084] Furthermore, based on the local geometric features of the point cloud, the geometric statistical features of the local region where the point cloud is located are calculated. This embodiment, based on the achievement of local geometric feature similarity standards, further analyzes the statistical regularities of the regional geometric structure to avoid mismatches caused by single-point noise or local anomalies. Specifically, principal component analysis can be used to calculate the regional average normal vector, quadratic surface fitting can be used to calculate the curvature distribution, and the number of points per unit volume can be counted to obtain the point density. Combining the geometric statistical features of the corresponding point clouds in the digital model and standard model of the concrete bridge, the matching degree of the point cloud pairs is comprehensively evaluated. Specifically, by setting weights for different geometric statistical features, the weighted total difference value is calculated. Point pairs with a difference value less than a set threshold T3 are determined as matched point cloud pairs. Finally, based on the matched point cloud pairs, the positional differences and geometric shape differences of the corresponding points in the two models are analyzed to determine multiple structural difference points.

[0085] Traditional methods rely solely on geometric coordinates or simple shape feature comparisons, making them suitable for scenarios with low matching accuracy requirements. However, when high matching accuracy is required, they tend to overlook subtle local features and the influence of semantic environment, potentially misjudging similar geometric shapes in different functional areas as differences. In contrast, the machine vision-based method for detecting damage and predicting the remaining life of concrete bridges provided in this invention fuses features and semantics, considering both structural details and environmental semantics. Through multi-level matching filtering, it first performs a coarse screening, then combines contextual features to eliminate false matches, and finally uses geometric statistical features to accurately locate substantial differences, thus avoiding misjudgments.

[0086] As an optional implementation, the first life-affecting parameter, the second life-affecting parameter, and the structural difference points all include corresponding structural coordinates. Based on each of the first life-affecting parameters, each of the second life-affecting parameters, and multiple structural difference points, the remaining life of the concrete bridge is estimated, such as... Figure 3 As shown, it includes:

[0087] S301, based on the structural coordinates, associate the first lifetime influence parameter, the second lifetime influence parameter, and the structural difference point;

[0088] S302, Based on the correlation, analyze the coupling relationship between the first lifetime influence parameter, the second lifetime influence parameter and the structural difference point to obtain the coupling relationship characteristics;

[0089] S303, combining the first lifetime influence parameter, the second lifetime influence parameter, structural differences and coupling relationship characteristics, and based on the linear degradation assumption, calculates the instantaneous rate of damage propagation and material degradation;

[0090] S304 inputs the instantaneous rates of damage propagation and material degradation into the nonlinear model to simulate the accelerated process of long-term damage accumulation and generate a predicted sequence of parameters changing over time.

[0091] S305, compare the predicted sequence with the preset structural failure threshold, and combine the synergistic effect of coupling relationship characteristics on the threshold to determine the time when each parameter reaches the threshold.

[0092] S306, take the minimum value as the predicted remaining life of the concrete bridge.

[0093] For example, when determining the predicted remaining life of a concrete bridge, the first step is to establish the correlation between different parameters and structural coordinates. Based on the first life-influencing parameters (such as mechanical parameters, geometric feature parameters, and damage locations corresponding to key visible damage images), the second life-influencing parameters (corrected concrete carbonation depth, steel corrosion rate, and corresponding structural coordinates), and structural difference points (obtained by comparing the bridge digital model with the standard model), all parameters are stored in the same data table using the structural coordinates (X, Y, Z) as the index. For example, for the bridge section at coordinates (12.34, 5.67, 3.21), the crack width, carbonation depth, and deviation between design and actual dimensions at that location are recorded one by one to ensure that each coordinate point is associated with complete parameter information.

[0094] Next, the coupling relationships between various parameters are analyzed. The data tables are traversed, and for each set of coordinates, the interactions between the first life-influencing parameter, the second life-influencing parameter, and structural difference points are studied. For example, a structural difference point at a certain location indicates that the beam size is smaller than the design value, while a wide crack (first life-influencing parameter) and a high carbonation depth (second life-influencing parameter) exist at that location. By statistically analyzing multiple similar data points, it is found that insufficient structural dimensions lead to stress concentration, which accelerates crack propagation. Furthermore, increased carbonation depth weakens concrete strength, further exacerbating crack and structural deformation. In this way, coupling relationship patterns under different parameter combinations are summarized, forming coupling relationship characteristics. For example, there is a positive correlation between "structural dimension deviation - crack width - carbonation depth," meaning that a change in one parameter accelerates the change in other parameters. The above analysis process can first observe the distribution pattern of variables through scatter plots. If a linear trend is observed, the Pearson coefficient is used to verify the strength of the linear positive correlation. If a curved or clustered distribution is observed, mutual information is used to calculate the linear / nonlinear correlation between the parameters as coupling relationship characteristics.

[0095] Then, based on the linear degradation assumption, for each parameter combination at each coordinate point, the initial degradation rate of each parameter is first determined. For concrete carbonation depth and steel corrosion rate, the previously corrected instantaneous rate is used as the initial value. Then, the initial rate is adjusted based on the coupling relationship characteristics. For example, if the coupling relationship shows a strong positive correlation between carbonation depth and steel corrosion rate, and the carbonation depth at the current coordinate point is large, then the initial value of steel corrosion rate is increased accordingly. Specifically, if the Pearson correlation coefficient between carbonation depth and steel corrosion rate is 0.8, it indicates that for every 1 mm increase in carbonation depth, the steel corrosion rate may increase proportionally. Therefore, multiplicative correction can be used. For example, if the initial rate of a certain parameter A is v... A The current value of its coupling parameter B is x. B And given that the correlation coefficient between the two is ρ, then the correction rate for parameter A is:

[0096]

[0097] Where, μ B and σ B Let B be the mean and standard deviation of parameter B. This formula standardizes the impact of the current state of parameter B on A and incorporates it into the correction. In this formula, the correlation coefficient is used as a scaling factor to reflect the linkage between variables. This is to eliminate the influence of dimensions and make the coupling effects of different parameters comparable.

[0098] For nonlinear couplings (such as exponential acceleration relationships), a nonlinear correction factor needs to be introduced. For example, in the calculation of the instantaneous rate of material degradation, when the crack width exceeds a critical value, the corrosion rate of the steel reinforcement may increase exponentially. In this case, a correction factor can be set as follows:

[0099] Correction factor = e k·MI(A,B)·f(B) ;

[0100] Where MI(A,B) is the mutual information value of parameters A and B (range [0,1], the larger the value, the stronger the dependence); f(B) is the nonlinear function of parameter B, such as This represents the rate of change of B relative to the baseline value B0, where k is the calibration coefficient, which can be obtained by fitting historical data. Finally, the initial rate is multiplied by this correction factor to obtain the corrected steel corrosion rate.

[0101] In the above formula, when B exceeds the benchmark value, the mutual information value amplifies its influence on A through an exponential function. For example, if B is the crack width and A is the steel corrosion rate, when the crack width exceeds the critical value B0, the corrosion rate accelerates exponentially with the crack width, and the degree of acceleration is positively correlated with the mutual information value (i.e., coupling strength) between the two. The above method calculates the instantaneous rate of damage propagation and material degradation at each coordinate point by comprehensively considering various parameters and their coupling relationships.

[0102] Instantaneous rates are input into a nonlinear model to simulate the long-term damage accumulation process. A model capable of reflecting nonlinear parameter changes, such as a neural network model or a nonlinear equation model based on physicochemical principles, is selected. The instantaneous rate and coupling characteristics at each coordinate point are used as model input. Based on a set time step (e.g., in years), the model predicts the changes in each parameter over a future period, generating a predicted sequence of parameter changes over time. For example, the model predicts parameters such as crack width, carbonation depth, and steel corrosion rate at different time points in the next 5, 10, and 15 years.

[0103] The predicted sequence is compared with preset structural failure thresholds. Failure thresholds for each parameter are pre-set according to bridge design specifications and safety standards, such as crack width exceeding 1.5 mm, carbonation depth reaching the thickness of the steel reinforcement protective layer, and steel reinforcement corrosion rate exceeding 20%. The synergistic effect of coupling characteristics on the thresholds is considered to determine the time when each parameter reaches its threshold. For example, due to the coupling relationship between crack width and carbonation depth, their combined effect may accelerate the steel reinforcement corrosion rate, causing the corrosion rate to reach the threshold earlier. By analyzing the predicted sequence, the specific time when each parameter reaches its corresponding threshold is determined.

[0104] Finally, the minimum value among the times when all parameters reach their thresholds is selected as the predicted remaining life of the concrete bridge. In other words, if even one critical parameter reaches its failure threshold, it may severely impact the structural safety of the bridge, rendering it unusable. This method comprehensively considers multiple parameters and their interrelationships, achieving a relatively accurate prediction of the remaining life of concrete bridges and providing an important basis for bridge maintenance and repair decisions.

[0105] This invention provides a machine vision-based device for damage detection and remaining life prediction of concrete bridges, comprising: an acquisition module for acquiring surface images of the concrete bridge and their corresponding structural coordinates; a coordinateization module for embedding the structural coordinates into the surface images of the concrete bridge to obtain a coordinateized surface image; a recognition module for inputting the coordinateized surface image into an image recognition model to recognize the image type and the corresponding structural coordinates, wherein the image type includes visible damage images and marked images, and the marked images are marked images generated by human testing; a first life influence parameter determination module for determining each first life influence parameter based on each visible damage image and its corresponding structural coordinate; a second life influence parameter determination module for determining each second life influence parameter based on each marked image and its corresponding structural coordinate; and a remaining life determination module for estimating the remaining life of the concrete bridge based on each first life influence parameter and each second life influence parameter.

[0106] This application also provides an electronic device, such as... Figure 4 As shown, processor 501 and memory 502 are connected via a bus or other means.

[0107] Processor 501 can be a central processing unit (CPU). Processor 501 can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or combinations of the above types of chips.

[0108] The memory 502, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the machine vision-based concrete bridge damage detection and remaining life prediction method in this embodiment of the invention. The processor executes various functional applications and data processing by running the non-transitory software programs, instructions, and modules stored in the memory.

[0109] Memory 502 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created by the processor, etc. Furthermore, the memory may include high-speed random access memory and non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory 502 may optionally include memory remotely located relative to the processor, which can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0110] The one or more modules are stored in the memory 502, and when executed by the processor 501, they perform actions such as... Figure 1 The illustrated embodiment presents a machine vision-based method for detecting damage and predicting the remaining life of concrete bridges.

[0111] For specific details regarding the aforementioned electronic devices, please refer to the relevant documentation. Figure 1 The relevant descriptions and effects in the illustrated embodiments are for understanding purposes only and will not be repeated here.

[0112] This embodiment also provides a computer storage medium storing computer-executable instructions that can execute the machine vision-based concrete bridge damage detection and remaining life prediction methods described in any of the above method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.

[0113] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made to it in form and detail without departing from the scope defined by the claims of the present invention.

Claims

1. A method for damage detection and remaining life prediction of concrete bridges based on machine vision, characterized in that, include: Obtain images of the surface of a concrete bridge and its corresponding structural coordinates; The structural coordinates are explicitly embedded into the surface image of the concrete bridge to obtain a coordinateized surface image. The coordinate-based surface image is input into the image recognition model to identify the image type and the corresponding structural coordinates. The image types include visible damage images and marked images, where marked images are marked images generated by human testing. Based on the images of each type of visible damage and the corresponding structural coordinates, determine each first lifetime influence parameter; Based on each marker class image and its corresponding structural coordinates, determine each second lifetime influence parameter; The remaining life of the concrete bridge is estimated based on the first life influence parameters and the second life influence parameters. Among them, the artificially generated marker image is the color image produced when the target reagent is encountered at the target location hole. Based on each marker image and its corresponding structural coordinates, the various second lifetime influence parameters are determined, including: Image preprocessing is performed on the labeled images to obtain the target images; Extract color regions from the target image, and divide the color regions into levels according to different color values ​​to obtain each color level region. Statistical analysis was conducted on each color grade area to determine the proportion of each color grade area to the entire hole. The color level regions and the proportion of each color level region to the entire hole are input into a pre-trained machine learning model to predict the carbonation depth of the benchmark concrete and the corrosion rate of the benchmark steel bars. Obtain the temperature and humidity of the environment where the concrete bridge is located; The ambient temperature and humidity, and the carbonation depth of the reference concrete are input into the pre-established carbonation depth correction model of concrete to obtain the corrected carbonation depth of concrete. The ambient temperature and humidity, and the baseline steel corrosion rate are input into the pre-established steel corrosion rate correction model to obtain the corrected steel corrosion rate. The corrected carbonation depth of concrete, the corrected corrosion rate of steel bars, and the corresponding structural coordinates are used as the second life-affecting parameters.

2. The method for damage detection and remaining life prediction of concrete bridges based on machine vision according to claim 1, characterized in that, Based on the images of each type of visible damage and the corresponding structural coordinates, the parameters affecting the first lifetime are determined, including: Extract geometric feature parameters from images of visible damage; Determine whether there are identified random speckle images at the structural coordinates of the visible damage images; random speckle images are one type of labeled images. If a random speckle pattern exists, the mechanical parameters are analyzed based on the random speckle pattern. The mechanical parameters, geometric feature parameters, and corresponding structural coordinates at the structural coordinates of each visible damage image are input into a pre-established first damage attention model for importance ranking to determine the key visible damage images. If no random speckle image exists, the geometric feature parameters and corresponding structural coordinates of each visible damage image are input into a pre-established second damage attention model for importance ranking to determine the key visible damage images. The mechanical parameters, geometric feature parameters, and damage location at the corresponding coordinates of the key visible damage image are used as the first lifetime influence parameters.

3. A method for damage detection and remaining life prediction of concrete bridges based on machine vision according to claim 1 or 2, characterized in that, Before estimating the remaining life of a concrete bridge based on the various first-life-influence parameters and the various second-life-influence parameters, the following steps are taken: Acquire 3D point cloud data of a concrete bridge; Reconstruct a digital model of a concrete bridge based on 3D point cloud data; The digital model of the concrete bridge was compared with the standard model of the concrete bridge to identify several structural differences. Based on the various first-life-affecting parameters and various second-life-affecting parameters, the remaining life of the concrete bridge is estimated, including: The remaining life of the concrete bridge is estimated based on the various first life-affecting parameters, the various second life-affecting parameters, and multiple structural difference points.

4. The method for damage detection and remaining life prediction of concrete bridges based on machine vision according to claim 3, characterized in that, The digital model of the concrete bridge was compared with the standard model of the concrete bridge to identify several structural differences, including: For each point in the point cloud, a local region is defined with that point as the center, and the point cloud within the local region is represented as a graph structure; By using graph convolutional networks to encode the graph structure, local feature descriptors of the point cloud are learned, which contain local geometric features. The local feature descriptor and the semantic label information of the point cloud are fused to obtain the context feature descriptor, and the semantic label information of the point cloud is used to characterize the environmental information of the bridge where the point cloud is located. By calculating the local feature descriptors of points, coarse matching based on feature similarity is performed to establish a preliminary point-to-point correspondence. Calculate the similarity between context feature descriptors of corresponding point pairs, including local geometric feature similarity and semantic label consistency; In the digital model and standard model of the concrete bridge, the local geometric features of the corresponding point clouds in point pairs with similarity exceeding the first threshold are obtained respectively. Based on the local geometric features of the point cloud, determine the geometric statistical features of the local region where the point cloud is located. The geometric statistical features include at least one of the region average normal vector, curvature distribution, and point density. By combining the geometric and statistical features of the point cloud in the digital model of the concrete bridge with the geometric and statistical features of the point cloud in the standard model of the concrete bridge, the degree of matching of the point cloud pairs is determined, and the matching point cloud pairs are obtained. Based on the matching point cloud pairs, multiple structural difference points are identified.

5. The method for damage detection and remaining life prediction of concrete bridges based on machine vision according to claim 3, characterized in that, The first life-affecting parameters, the second life-affecting parameters, and the structural difference points all include corresponding structural coordinates. Based on each of the first life-affecting parameters, each of the second life-affecting parameters, and multiple structural difference points, the remaining life of the concrete bridge is estimated, including: Based on the structural coordinates, the first lifetime influence parameter, the second lifetime influence parameter, and the structural difference point are correlated; Based on the correlation, the coupling relationship between the first lifetime influence parameter, the second lifetime influence parameter and the structural difference point is analyzed to obtain the coupling relationship characteristics; Combining the first lifetime influence parameter, the second lifetime influence parameter, structural differences and coupling relationship characteristics, and based on the linear degradation assumption, the instantaneous rate of damage propagation and material degradation is calculated. The instantaneous rates of damage propagation and material degradation are input into a nonlinear model to simulate the accelerated process of long-term damage accumulation and generate a predicted sequence of parameters changing over time. The predicted sequence is compared with the preset structural failure threshold, and the time when each parameter reaches the threshold is determined by combining the synergistic effect of coupling relationship characteristics on the threshold. The minimum value is taken as the predicted remaining life of the concrete bridge.

6. The method for damage detection and remaining life prediction of concrete bridges based on machine vision according to claim 4, characterized in that, Calculate the similarity between contextual feature descriptors of corresponding point pairs. This similarity includes local geometric feature similarity and semantic label consistency, including: Input the semantic label of any point into the pre-established semantic library of the standard model of concrete bridges for query. When a standard semantic label is matched, the semantic label similarity is the first similarity. The semantic library of the standard model of concrete bridges contains standard semantic labels and semantic association graphs. If no standard semantic tag is matched, the semantic tag is broken down into the smallest semantic units, and the associated terms of each smallest semantic unit are queried in the semantic association graph. The associated terms of each semantic unit are exhaustively combined and then matched with the standard semantic tags. When a standard semantic tag is matched in the second match, the semantic tag similarity is set as the second similarity. If the second match fails to match the standard semantic tag, the semantic tag similarity is set to the third similarity.

7. A machine vision-based device for damage detection and remaining life prediction of concrete bridges, characterized in that, include: The acquisition module is used to acquire images of the surface of concrete bridges and their corresponding structural coordinates. The coordinate transformation module is used to explicitly embed structural coordinates into the surface image of a concrete bridge to obtain a coordinate-transformed surface image. The recognition module is used to input the coordinated surface image into the image recognition model, identify the image type and the corresponding structural coordinates. The image types include visible damage images and marked images, where marked images are marked images generated by human testing. The first lifetime influence parameter determination module is used to determine each first lifetime influence parameter based on each visible damage image and the corresponding structural coordinates. The second lifetime influence parameter determination module is used to determine each second lifetime influence parameter based on each marker class image and the corresponding structural coordinates. The remaining life determination module is used to estimate the remaining life of concrete bridges based on each first life influence parameter and each second life influence parameter. Among them, the artificially generated marker image is the color image produced when the target reagent is encountered at the target location hole. Based on each marker image and its corresponding structural coordinates, the various second lifetime influence parameters are determined, including: Image preprocessing is performed on the labeled images to obtain the target images; Extract color regions from the target image, and divide the color regions into levels according to different color values ​​to obtain each color level region. Statistical analysis was conducted on each color grade area to determine the proportion of each color grade area to the entire hole. The color level regions and the proportion of each color level region to the entire hole are input into a pre-trained machine learning model to predict the carbonation depth of the benchmark concrete and the corrosion rate of the benchmark steel bars. Obtain the temperature and humidity of the environment where the concrete bridge is located; The ambient temperature and humidity, and the carbonation depth of the reference concrete are input into the pre-established carbonation depth correction model of concrete to obtain the corrected carbonation depth of concrete. The ambient temperature and humidity, and the baseline steel corrosion rate are input into the pre-established steel corrosion rate correction model to obtain the corrected steel corrosion rate. The corrected carbonation depth of concrete, the corrected corrosion rate of steel bars, and the corresponding structural coordinates are used as the second life-affecting parameters.

8. An electronic device, the device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor performs the steps of the machine vision-based method for detecting damage and predicting the remaining life of concrete bridges according to any one of claims 1-6.

9. A computer storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the machine vision-based method for damage detection and remaining life prediction of concrete bridges as described in any one of claims 1-6.

Images