Method and System for Detecting and Evaluating Damage of Concrete Components

Abstract

The present application relates to a method and system for detecting and evaluating damage to concrete components, including: obtaining a first image and cropping the first image to obtain a second image, where the first image is a partial image of a damaged concrete component in existing tests and actual projects. Generating an annotation file and an annotation image after annotating the second image with a polyline. Processing the annotation file to adapt it to a damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtaining corresponding optimal weight parameters. Obtaining a third image, and identifying the third image based on the optimal weight parameters and the damage target segmentation model to obtain a mask of the damage to the concrete component and the damage type in the third image. When the damage type is a crack, extracting the pixel points corresponding to the actual crack of the concrete component based on the mask. Obtaining the pixel resolution of the image acquisition device, and obtaining the geometric parameters corresponding to the pixel points and other damage masks according to the pixel resolution.

Description

Technical Field

[0001] This application relates to the technical field of damage detection for damaged building structures, and particularly to a method and system for detecting and evaluating damage to concrete components. Background Art

[0002] Reinforced Concrete (RC) structures are widely used in building structures in China. The problem of deterioration of such structures during service is relatively prominent, and they are prone to significant damage after experiencing destructive earthquakes. Detecting the damage to damaged RC structures and evaluating the safety of the post-earthquake structures can provide a scientific basis for judging the safety of RC structures and for the maintenance, repair, reinforcement or demolition of RC structures, which is of great significance for avoiding and controlling disasters.

[0003] Currently, the main method for detecting damage to RC structures is manual inspection, relying on structural engineering professionals to conduct inspections or investigations at the project or earthquake disaster site. By means of visual observation, photographing, etc., combined with tools such as crack gauges and crack observation instruments, the apparent damage phenomena of damaged RC structures are recorded. However, using this method is prone to misdetection and missed detection, and due to the strong subjectivity of the detection, the detection accuracy is relatively low. At the same time, the detection process is time-consuming and laborious, making it difficult for this method to meet the requirements of rapid detection of post-earthquake structural damage. With the increasing maturity of digital image processing technology and machine learning methods, after the inspection personnel collect images of damaged structural components on site, they can detect the damage targets in the images through operations such as image preprocessing, noise reduction, feature extraction, and feature classification. However, the images collected by the inspection personnel often have a lot of interference and complex background information, and the damage targets are also complex and diverse. The detection methods based on image processing and traditional machine learning generally use manually designed features, which have a relatively low hierarchical level, insufficient representativeness and separability, making it difficult for the damage detection methods based on traditional computer vision to adapt to complex and diverse structural damage detection.

[0004] In recent years, to improve the efficiency and accuracy of building structure damage detection, engineers have used deep learning methods based on Convolutional Neural Network (CNN for short) to identify structural damage. Compared with traditional machine learning methods, deep learning methods have stronger feature learning and expression capabilities, and can extract more diverse and abstract features in images. Currently, for the damage identification of RC structures, CNN-based methods are mainly divided into object detection-based methods and image segmentation-based methods. The typical representative of the former is Faster R-CNN, which has been used for the detection of steel structure diseases and RC structure cracks, but it can only classify and locate the damage. If we want to analyze the physical parameters of the damage, such as crack width, length, etc., we need to additionally adopt a large number of complex image processing operations on the basis of obtaining the damage type and location information using the CNN model. The typical representative of the latter is U-Net, which has been used for the detection of RC structure damage, but it can only classify and semantically segment the damage, and cannot directly determine the location of the damage in the structure.

[0005] Therefore, for the large number of RC structural components, the traditional methods for detecting and evaluating RC component damage cannot complete the location, classification, and segmentation of common damage types at one time. A large amount of labor costs are required for the detection and evaluation of RC component damage, and the detection and evaluation results are often not accurate enough. Summary of the Invention

[0006] Based on this, it is necessary to provide a method and system for detecting and evaluating concrete component damage that can effectively save labor costs, complete damage location, classification, and segmentation at one time, and have relatively accurate detection and evaluation results for the above technical problems.

[0007] In the first aspect, the present application provides a method for detecting and evaluating concrete component damage, the method comprising:

[0008] Obtain a first image and crop the first image to obtain a second image, where the first image is a partial image of a damaged concrete component in existing tests and actual projects, and the second image is a damage image of a concrete component adapted to a damage target segmentation model;

[0009] Obtain an annotation file and an annotation image, where the annotation file is generated by annotating the second image with a polyline, and the annotation image is an image generated by annotating the second image with the polyline;

[0010] Process the annotation file to adapt it to the damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtain the corresponding optimal weight parameters;

[0011] Obtain a third image, and identify the third image based on the optimal weight parameters and the damage target segmentation model to obtain a mask of the damage of the concrete member and the damage type in the third image, where the third image is a damage image of the concrete member to be detected;

[0012] When the damage type of the concrete member to be detected is a crack, extract the first pixel points based on the mask of the crack, where the first pixel points are the pixel points corresponding to the actual crack of the concrete member to be detected;

[0013] Obtain the pixel resolution of the image acquisition device, and obtain the geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

[0014] In one embodiment, the obtaining the third image and identifying the third image based on the optimal weight parameters and the damage target segmentation model includes:

[0015] Take pictures of the concrete member to be detected multiple times through the image acquisition device to obtain multiple images with overlapping areas;

[0016] Stitch the multiple images with overlapping areas into a panoramic image based on the image stitching method of the SURF algorithm to eliminate the overlapping areas;

[0017] Obtain the third image by cropping the panoramic image.

[0018] In one embodiment, the method further includes:

[0019] Identify the third image through the Mask R-CNN convolutional neural network to obtain the damage type, bounding box, and mask within the bounding box of the concrete member to be detected;

[0020] Establish a damage level standard corresponding to the damage degree, where the damage degree is a pre-set configuration parameter for determining the damage level;

[0021] Create a database based on the damage type, damage degree, and damage level standard.

[0022] In one embodiment, the damage types include concrete cracks, cover spalling, concrete crushing, steel bar exposure, steel bar buckling, and fracture;

[0023] The database includes the damage type, damage level standard, and the corresponding relationship between the damage degree and the damage level standard, and is used to evaluate the third image to obtain the damage level of the concrete member to be detected.

[0024] In one embodiment, after obtaining the third image and identifying the third image based on the optimal weight parameter and the damage target segmentation model, the following steps are included:

[0025] Crop the image within the bounding box where the crack mask is located;

[0026] Filter and denoise the image within the bounding box, remove the pixel points outside the crack mask, and perform threshold segmentation and morphological analysis to extract the first pixel points.

[0027] In one embodiment, obtaining the pixel resolution of the image acquisition device includes:

[0028] Obtain the internal and external parameters of the image acquisition device after calibration;

[0029] Calculate according to the calibrated internal and external parameters to obtain the pixel resolution.

[0030] In one embodiment, the method further includes:

[0031] After extracting the first pixel points, determine the center line of the first pixel points through morphological analysis;

[0032] Based on the distance between the center line and the edge of the first pixel points, obtain the geometric parameters of the first pixel points;

[0033] Obtain the inclination angle of the first pixel points through the inclination angle of the minimum circumscribed rectangle of the first pixel points;

[0034] Wherein, the geometric parameters include the width and length of the first pixel points.

[0035] In a second aspect, the present application provides a concrete component damage detection and evaluation system, and the system includes:

[0036] An image acquisition module, configured to acquire a first image and crop the first image to obtain a second image, where the first image is a partial image of a damaged concrete component in existing tests and actual projects, and the second image is a damage image of the concrete component adapted to the damage target segmentation model;

[0037] A labeling module, configured to acquire a labeling file and a labeling image, where the labeling file is generated after labeling the second image through a polyline, and the labeling image is an image generated after the second image is labeled by the polyline;

[0038] A model training module, configured to process the labeling file to adapt to the damage target segmentation model, and after training the target segmentation model with the labeling image and the labeling file, obtain the corresponding optimal weight parameter;

[0039] A damage recognition module, configured to obtain a third image, and recognize the third image based on the optimal weight parameters and a damage target segmentation model to obtain a mask and a damage type of the concrete member damage in the third image, where the third image is a damage image of a concrete member to be detected;

[0040] An extraction module, configured to, when the damage type of the concrete member to be detected is a crack, extract first pixel points based on the mask of the crack, where the first pixel points are pixel points corresponding to the real crack of the concrete member to be detected;

[0041] A damage analysis module, configured to obtain the pixel resolution of an image acquisition device, and obtain geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

[0042] In a third aspect, the present application provides an electronic device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0043] Obtain a first image and crop the first image to obtain a second image, where the first image is a partial image of a damaged concrete member in existing tests and actual projects, and the second image is a damage image of a concrete member adapted to a damage target segmentation model;

[0044] Obtain an annotation file and an annotation image, where the annotation file is generated after annotating the second image with a polyline, and the annotation image is an image generated after annotating the second image with the polyline;

[0045] Process the annotation file to adapt it to the damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtain corresponding optimal weight parameters;

[0046] Obtain a third image, and recognize the third image based on the optimal weight parameters and a damage target segmentation model to obtain a mask and a damage type of the concrete member damage in the third image, where the third image is a damage image of a concrete member to be detected;

[0047] When the damage type of the concrete member to be detected is a crack, extract first pixel points based on the mask of the crack, where the first pixel points are pixel points corresponding to the real crack of the concrete member to be detected;

[0048] Obtain the pixel resolution of an image acquisition device, and obtain geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

[0049] Fourthly, the present application provides a computer storage medium storing a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0050] Obtain a first image and crop the first image to obtain a second image, where the first image is a partial image of a damaged concrete member in existing tests and actual projects, and the second image is a damage image of the concrete member adapted to the damage target segmentation model;

[0051] Obtain an annotation file and an annotation image, where the annotation file is generated by annotating the second image with a polyline, and the annotation image is an image generated by annotating the second image with the polyline;

[0052] Process the annotation file to adapt it to the damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtain the corresponding optimal weight parameters;

[0053] Obtain a third image, and identify the third image based on the optimal weight parameters and the damage target segmentation model to obtain a mask and a damage type of the concrete member damage in the third image, where the third image is a damage image of a concrete member to be detected;

[0054] When the damage type of the concrete member to be detected is a crack, extract a first pixel point based on the mask of the crack, where the first pixel point is a pixel point corresponding to the actual crack of the concrete member to be detected;

[0055] Obtain the pixel resolution of the image acquisition device, and obtain geometric parameters corresponding to the first pixel point and other damage masks according to the pixel resolution.

[0056] The above concrete component damage detection and evaluation method and system obtain damaged concrete component images adapted to the target segmentation model by cropping local images of damaged concrete components in existing experiments and actual projects. Subsequently, the damaged images of the concrete components are labeled with polylines to obtain corresponding annotation files and annotation images, and then the damaged target segmentation model is trained with the annotation files and annotation images that meet the training requirements of the damaged target segmentation model to obtain the optimal weight parameters of the damaged concrete component images. Then, by obtaining the damaged images of the concrete components to be detected and identifying the damaged images of the concrete components to be detected based on the optimal weight parameters and the trained damaged target segmentation model, the mask of the damage to the concrete components to be detected can be segmented, and the true crack pixel points of the concrete components to be detected can be extracted from the crack mask. Finally, according to the pixel resolution of the image acquisition device, the true damage pixel points of the damage to the concrete components to be detected and the geometric parameters corresponding to other damage masks are obtained, and the staff can complete the detection and evaluation of the damage to the concrete components based on the geometric parameters of the damage pixels of the concrete component damage. This method is based on the segmentation results of the damaged target segmentation model and combines corresponding image processing operations to obtain the damage type and location of the concrete components, and automatically counts the damage pixel information on the surface of the concrete components, without the need for on-site measurement and recording by inspectors, and can complete damage location, classification, and segmentation at one time. To a certain extent, it saves manpower, reduces the errors caused by manual measurement, and improves the accuracy of concrete component damage detection and evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is one of the flowcharts of the concrete component damage detection and evaluation method provided by this application;

[0058] Figure 2 is another flowchart of the concrete component damage detection and evaluation method provided by this application;

[0059] Figure 3 is yet another flowchart of the concrete component damage detection and evaluation method provided by this application;

[0060] Figure 4 is still another flowchart of the concrete component damage detection and evaluation method provided by this application;

[0061] Figure 5 is one of the flowcharts of the concrete component damage detection and evaluation method provided by this application;

[0062] Figure 6 is another flowchart of the concrete component damage detection and evaluation method provided by this application;

[0063] Figure 7 is a schematic diagram of the flow of the concrete component damage detection and evaluation method in a specific embodiment provided by this application;

[0064] Figure 8 Schematic diagram of surface damage marking of concrete components in the specific embodiment provided for this application for the concrete component damage detection and evaluation method

[0065] Figure 9 Schematic diagram of damage recognition results of partial local photos of the front elevation of the beam at the joint of concrete components in the specific embodiment provided for this application for the concrete component damage detection and evaluation method

[0066] Figure 10 Schematic diagram of the extraction process and effect of crack pixels in the partial front elevation photo of the beam at the joint of concrete components in the specific embodiment provided for this application for the concrete component damage detection and evaluation method

[0067] Figure 11 Schematic diagram of the interface of the intelligent detection software for concrete component damage in the specific embodiment provided for this application

[0068] Figure 12 Schematic diagram of the structure of the concrete component damage detection and evaluation system provided for this application

[0069] Figure 13 Internal structure diagram of the computer device provided for this application Detailed implementation manners

[0070] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0071] As Figure 1 shown, in one embodiment, a concrete component damage detection and evaluation method includes the following steps:

[0072] Step S110, obtaining a first image and cropping the first image to obtain a second image, where the first image is a partial image of a damaged concrete component in existing experiments and actual projects, and the second image is a damage image of the concrete component adapted to the damage target segmentation model.

[0073] Specifically, the server crops the partial image of the damaged concrete component in existing experiments and actual projects to obtain a damage image of the concrete component adapted to the target segmentation model, so as to train the damage target segmentation model.

[0074] Step S120: Obtain the annotation file and the annotated image. The annotation file is generated after annotating the second image with polylines, and the annotated image is the image generated after annotating the second image with polylines.

[0075] Specifically, the server annotates the damage image of the concrete member obtained by cropping in step S110 with polylines, that is, encloses the damage image of the concrete member with polylines to form a closed polygon, and makes the polylines closely fit the damage edge of the concrete member, thereby generating the corresponding annotation file and annotated image.

[0076] Step S130: Process the annotation file to adapt it to the damage target segmentation model, and after training the target segmentation model with the annotated image and the annotation file, obtain the corresponding optimal weight parameters.

[0077] Specifically, the server processes the annotation file generated in step S120 to make it meet the training requirements of the damage target segmentation model, and trains the target segmentation model with the annotated image generated in step S120 and the processed annotation file to obtain the optimal weight parameters.

[0078] Step S140: Obtain the third image, and identify the third image based on the optimal weight parameters and the damage target segmentation model to obtain the mask and damage type of the concrete member damage in the third image. The third image is the damage image of the concrete member to be detected.

[0079] Specifically, the server identifies the damage image of the concrete member to be detected collected in real time by the image acquisition device based on the optimal weight parameters obtained in step S130 and the damage target segmentation model, and obtains the mask matrix corresponding to the surface damage of the concrete member to be detected and its damage type.

[0080] Step S150: When the damage type of the concrete member to be detected is a crack, extract the first pixel points based on the crack mask. The first pixel points are the pixel points corresponding to the real cracks of the concrete member to be detected.

[0081] Specifically, when the damage type of the concrete member to be detected is a crack, the server extracts the pixel points corresponding to the real cracks of the concrete member to be detected based on the crack mask matrix obtained in step S140.

[0082] Step S160: Obtain the pixel resolution of the image acquisition device, and obtain the geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

[0083] Specifically, the server obtains the crack pixels of the concrete member to be detected and the geometric parameters corresponding to other types of damage according to the pixel resolution of the image acquisition device.

[0084] The above concrete component damage detection and evaluation method obtains damaged concrete component images suitable for the target segmentation model by cropping local images of damaged concrete components in existing experiments and actual projects. Subsequently, the damaged images of the concrete components are labeled with polylines to obtain corresponding annotation files and annotation images, and then the damage target segmentation model is trained with the annotation files and annotation images that meet the training requirements of the damage target segmentation model to obtain the optimal weight parameters of the damaged concrete component images. Then, by obtaining the damaged images of the concrete components to be detected and identifying the damaged images of the concrete components to be detected based on the optimal weight parameters and the trained damage target segmentation model, the mask of the damage of the concrete components to be detected can be segmented, and the real crack pixel points of the concrete components to be detected can be extracted from the crack mask. Finally, according to the pixel resolution of the image acquisition device, the real damage pixel points of the damage of the concrete components to be detected and the geometric parameters corresponding to other damage masks are obtained, and the staff can complete the detection and evaluation of the damage of the concrete components according to the geometric parameters of the damage pixels of the concrete component damage. This method is based on the segmentation results of the damage target segmentation model and combines corresponding image processing operations to obtain the damage type and location of the concrete components, and automatically counts the damage pixel information on the surface of the concrete components, without the need for on-site measurement and recording by the inspectors, and can complete damage location, classification and segmentation at one time. To a certain extent, it saves manpower, reduces the errors caused by manual measurement, and improves the accuracy of concrete component damage detection and evaluation.

[0085] As Figure 2 shown, the concrete component damage detection and evaluation method provided by this application obtains a third image and performs recognition on the third image with the optimal weight parameters and the damage target segmentation model, including the following steps:

[0086] Step S142, the image acquisition device takes pictures of the concrete components to be detected multiple times to obtain multiple images with overlapping areas.

[0087] Specifically, the server takes pictures of the concrete components to be detected multiple times through the image acquisition device to obtain multiple local images of the concrete components to be detected with overlapping areas.

[0088] Step S144, an image stitching method based on the SURF algorithm stitches multiple images with overlapping areas into a panoramic image to eliminate the overlapping areas.

[0089] Among them, SURF is the full name of SpeededUp RobustFeatures, that is, accelerated robust features, which is an algorithm for robust local feature point detection and description.

[0090] Specifically, the server stitches together the multiple local images of concrete components with overlapping regions obtained in step S142 into a panoramic image through an image stitching method using the SURF algorithm to eliminate the overlapping regions of the multiple local images of concrete components.

[0091] Step S146: Crop the panoramic image to obtain a third image.

[0092] Specifically, the server crops the panoramic image stitched in step S144 to obtain a third image, that is, the damage image of the concrete component to be detected.

[0093] As Figure 3 shown, the concrete component damage detection and evaluation method provided by this application further includes the following steps:

[0094] Step S310: Identify the third image through the Mask R-CNN convolutional neural network to obtain the damage type, bounding box, and mask located within the bounding box of the concrete component to be detected.

[0095] Specifically, the server identifies the damage image obtained after cropping the local image of the concrete component to be detected through the Mask R-CNN convolutional neural network to obtain the damage type, bounding box, and damage mask located within the bounding box of the concrete component to be detected.

[0096] Among them, the damage types of the concrete component to be detected include concrete cracks, cover spalling, concrete crushing, steel bar exposure, steel bar buckling, and fracture, which are stored in the database for the recognition of the damage image of the concrete component to be detected.

[0097] Step S320: Establish a damage grade standard corresponding to the damage degree, where the damage degree is a configuration parameter preset for determining the damage grade.

[0098] Specifically, the server automatically establishes a damage grade standard corresponding to the damage degree on the basis of step S310, where the damage degree is a configuration parameter preset by the user for determining the damage grade.

[0099] Step S330: Create a database based on the damage type, damage degree, and damage grade standard.

[0100] Specifically, the server creates a database based on the damage type of the concrete component and its corresponding damage grade standard.

[0101] It should be noted that the database includes the damage types of concrete components, the damage degrees, and the damage grade standards corresponding to the damage degrees. There is a corresponding relationship between the damage degrees of concrete components and their corresponding damage grade standards, which is used to evaluate the damage images of concrete components to obtain the damage grades of the damage images of concrete components.

[0102] As Figure 4 shown, the concrete component damage detection and evaluation method provided by this application obtains the third image and performs recognition on the third image based on the optimal weight parameters and the damage target segmentation model. Then, it includes the following steps:

[0103] Step S410: Crop the image within the bounding box where the crack mask is located.

[0104] Specifically, the server crops the image within the bounding box where the crack mask is located.

[0105] Step S420: Filter and denoise the image within the bounding box, remove the pixel points outside the crack mask, and perform threshold segmentation and morphological analysis to extract the first pixel points.

[0106] Specifically, the server filters and denoises the image within the bounding box where the crack mask is located, removes the pixel points outside the crack mask, and performs threshold segmentation and morphological analysis to extract the first pixel points, that is, the pixel points of the real cracks of the concrete component to be detected, so as to achieve accurate detection of the cracks of the concrete component.

[0107] As Figure 5 shown, the concrete component damage detection and evaluation method provided by this application obtains the pixel resolution of the image acquisition device, including the following steps:

[0108] Step S162: Obtain the internal and external parameters of the image acquisition device after calibration.

[0109] Specifically, the server obtains the internal and external parameters of the image acquisition device after calibration.

[0110] Step S164: Calculate according to the calibrated internal and external parameters to obtain the pixel resolution.

[0111] Specifically, the server calculates the internal and external parameters of the image acquisition device after calibration obtained in step S162 to obtain the pixel resolution of the image acquisition device.

[0112] As Figure 6 shown, the concrete component damage detection and evaluation method provided by this application further includes the following steps:

[0113] Step S610: After extracting the first pixel points, determine the center line of the first pixel points through morphological analysis.

[0114] Specifically, when the damage type of the concrete component to be detected is a crack, after the server extracts the first pixel point of the concrete component, the center line of the first pixel point of the concrete component to be detected is determined through morphological analysis.

[0115] Step S620: Based on the distance between the center line and the edge of the first pixel point, obtain the geometric parameters of the first pixel point.

[0116] Specifically, the server obtains the geometric parameters of the crack of the concrete component to be detected based on the distance between the center line and the edge of the real crack pixel point.

[0117] Step S630: Obtain the inclination angle of the first pixel point through the inclination angle of the minimum circumscribed rectangle of the first pixel point.

[0118] Specifically, the server obtains the inclination angle of the damage through the inclination angle of the minimum circumscribed rectangle of the concrete component damage.

[0119] It should be noted that the geometric parameters of the concrete component damage include the length and width of the damage, which are used to evaluate the damage degree of the concrete component.

[0120] See Figure 7 As shown, in a specific embodiment, a method for detecting and evaluating the damage of a concrete component, the concrete component, abbreviated as an RC component. This method is based on the damage segmentation result combined with image processing, obtains the geometric information of the damage of the RC component through the pixel resolution of the image acquisition device, and automatically counts the damage information on the surface of the RC component. There is no need for inspectors to measure and record on-site, and it can also automatically quantify and evaluate the damage level of the post-earthquake RC component according to the established standards.

[0121] First, use the relevant image acquisition device to take pictures of the front elevation of the RC component to be detected to obtain multiple partial front pictures of the component with overlapping areas, and correct the distortion of the partial front pictures of the component obtained by the front shooting. Subsequently, use image stitching technology to obtain the panoramic picture of the RC component and appropriately crop it to meet the requirements of the damage target segmentation model. Finally, obtain multiple non-overlapping partial pictures of the component, and all partial pictures need to include the front elevation of the RC component, and the front elevation is the surface to be detected.

[0122] Secondly, the locally cropped images of RC components are used as input image data, and a damage target segmentation model is employed to automatically identify various types of damages in the input image data. The recognizable damage types of RC components include: concrete cracks, cover spalling, concrete crushing, steel bar exposure, steel bar buckling, and fracture. Among them, the damage target segmentation model adopts the Mask R-CNN image instance segmentation convolutional neural network architecture. This model first proposes target candidate regions and then performs target detection and segmentation, including an image feature extraction part composed of a backbone network and a feature pyramid network, and a target prediction part composed of a region proposal network, branches for target classification and localization, and a mask branch for instance segmentation of targets.

[0123] After identifying the local image data, the damage target segmentation model can generate the vertex coordinates of the bounding box of the damage target, damage labels, mask matrix, and confidence. After being processed by the model, corresponding damage bounding boxes, damage types, and masks for segmenting damages can be generated in the image data. After obtaining the damage recognition results of the RC components, for the extraction of pixel points corresponding to the surface cracks of the concrete, the image within the bounding box where the crack mask is located is intercepted. Operations such as filtering and denoising, removing pixels outside the crack mask, etc. are performed on the image within this bounding box, and a threshold segmentation algorithm combined with morphological analysis operations are used to extract the real crack pixels. Subsequently, physical parameters such as crack width and length are calculated. The morphological analysis method is used to extract the crack skeleton line, that is, the center line, and the distance between the intersection points of the perpendicular line of the center line and the crack edge is calculated at the calculation point to obtain the crack width at the calculation point, and the maximum crack width is obtained by analyzing along the crack length. At the same time, the crack length can be obtained by calculating the length of the crack center line, and the crack inclination angle is calculated based on the inclination angle of the minimum circumscribed rectangle of the crack. The areas of concrete cover spalling, concrete crushing, steel bar exposure, and buckling can be obtained by directly calculating the physical areas of the masks of this type of damage, and the quantity information of this type of damage is obtained by counting the number of target recognition results of steel bar fractures.

[0124] Finally, by encapsulating the programs involved in damage target segmentation, image processing, and calculation of damage physical parameters, and based on PyQt, a graphical user interface is written and the designed program is packaged and output as an intelligent detection software for RC component damages. This intelligent detection software for RC component damages includes functions for identifying damages in single and batch locally cropped photos of RC components, calculating damage physical parameters, and statistically analyzing the damage information on the component surface. The damage physical parameters include the maximum crack width, crack length, crack inclination angle, and the areas of cover spalling, concrete crushing, steel bar exposure, and buckling, as well as the number of steel bar fractures, and the main damage information can be extracted from the statistical information to quantitatively evaluate the damage level of the post-earthquake components according to the established standards.

[0125] In this embodiment, before using the damage target segmentation model for damage identification, it is necessary to train the damage target segmentation model using the labeled dataset, and the dataset used in the training process is mainly composed of photos of past pseudo-static tests of RC components. The process of establishing the dataset is as follows:

[0126] First, collect photos of past pseudo-static tests of RC components, including RC shear walls, RC columns, RC beams, RC joints, etc., and at the same time collect photos of damaged RC components in actual projects. The experimental photos should include different shooting distances, lighting conditions, shooting angles, and diverse background information as much as possible. Make appropriate cropping of the collected photos according to the computing power of the graphics processor and the requirements of the damage segmentation model for the input image size.

[0127] Secondly, formulate identification criteria for visible damage on the surfaces of various typical RC components, and organize relevant professionals to mark various damages in the photos to form a dataset of surface damages of RC components. When marking the damages in the photos, professional software such as Labelme and Photoshop can be used, and a polyline is used to enclose the damage along the damage edge. For the marking of cracks, there is a small gap between the polyline and the crack edge, as Figure 8 shown. For other types of damages, it is advisable to make the polyline fit the damage edge as much as possible. When the number of labeled pictures in the dataset is insufficient, data augmentation techniques can be used to increase the number and diversity of labeled pictures by cropping, rotating, mirroring, and adjusting the chromaticity and brightness of the existing labeled pictures. Parameters such as the number of training epochs, the number of iteration steps, the starting layer of weight training, the size of the pictures used for training, and the initial learning rate can be set for the training of the damage segmentation model. The learning rate decay strategy and the weight optimization algorithm used for training can also be adjusted, such as the SGD method (stochastic gradient descent method) and the Adam method (a derivative of the stochastic gradient descent method). When identifying damages, it is necessary to use the damage segmentation model combined with the optimal weight parameters obtained through training. The orthographic front partial photo of the component after distortion correction used for damage identification is obtained by cropping and splicing the panoramic orthographic front image of the component. When splicing the partial photos, the SURF (Speeded Up Robust Features) algorithm is used to splice the partial photos, and the operation steps are as follows:

[0128] First, construct the Hessian matrix and the image scale space for subsequent pixel processing. Subsequently, in the three-dimensional scale space, locate the feature points based on three-dimensional linear interpolation, and then determine the main direction of each feature point by calculating the Harr wavelet response of the feature points one by one. Subsequently, count the Harr wavelet features of the pixels around the feature points to generate the feature point descriptors, and use the feature point descriptors to match the feature points according to the Euclidean distance between two feature points and the trace of the Hessian matrix. Finally, perform geometric transformation on the local photo according to the positional relationship of the matched feature points to complete image registration, and perform local photo fusion to complete local photo stitching.

[0129] In the process of extracting the pixel points corresponding to the real cracks in concrete, first read the original grayscale image of the image within the damage bounding box, then smooth the image using the two-dimensional median filtering method, and take the difference between the original grayscale image and the smoothed image. Remove the pixels outside the crack mask, and use the Otsu automatic clustering threshold segmentation method to preliminarily extract the crack pixels. Finally, perform dilation and erosion operations combined with the operation of removing isolated connected components to obtain the real crack pixels. When calculating the damage physical parameters, it is necessary to first calculate the pixel size of the damage, such as the pixel width at the crack calculation point, the pixel area of spalling, etc., and then calculate the damage physical parameters according to the pixel resolution. For the calculation of the photo pixel resolution, when the image acquisition device is facing the front elevation of the component, that is, the angle between the lens optical axis and the normal of the front elevation is zero, or although the lens optical axis is not orthogonal to the front elevation, but the lens field of view area is small, so that the distance difference between the object points corresponding to each pixel point to the projection center in the camera coordinate system is very small, the pixel resolution of the whole image is taken as the pixel resolution of the imaging center, and the pixel resolution of the imaging center is taken as the average of the horizontal and vertical resolutions of this point.

[0130] When performing batch detection with the RC component damage intelligent detection software written according to the PyQt interface, it will automatically call the image processing and crack geometric parameter calculation programs, analyze each crack identified in the local photos of the batch RC components one by one, and finally export the crack geometric parameters of the front elevation of the RC components to the damage parameter summary table. The content in the table includes the summary of damage types, crack geometric information, and the areas of spalling, crushing, exposed reinforcement, and steel bar buckling, as well as the number of steel bar fractures.

[0131] In this embodiment, based on the statistical results of the RC component damage information, quantitatively evaluate the damage level of the RC component according to the corresponding relationship between the damage level and the damage degree. The corresponding relationship between the damage level and the damage degree of the RC component is shown in Table 1:

[0132] Table 1 Corresponding relationship between the damage level and the damage degree of RC components

[0133]

[0134]

[0135] In another specific embodiment, a method for detecting and evaluating the damage of concrete components is applied to the detection and evaluation of the damage of concrete components (referred to as RC components for short). First, collect the photos of the pseudo-static loading tests of past RC components and the photos of damaged RC components in actual projects or earthquake disaster sites. The resolution of such photos of damaged RC components generally ranges from 3000×4000 to 8000×6000. The RC components in the tests include RC shear walls, RC columns, RC beams, RC joints, RC slabs, and RC precast stairs, etc. The damaged RC components in actual projects include RC components deteriorated due to concrete carbonization, salt damage, etc., with damages such as cracks, exposed reinforcement corrosion, etc., and earthquake-damaged RC structural components. The collected photos of damaged components should have diverse background information, lighting conditions, different shooting distances, and shooting angles. Crop the collected photos so that their resolution generally ranges from 512×512 to 1500×1500, and the specific size can be determined according to the computing power of the hardware used for training the model, such as the video memory size.

[0136] Secondly, compile the annotation criteria for six typical surface damages of RC components, including concrete cracks, cover spalling, concrete crushing, steel bar exposure, steel bar buckling, and fracture. Organize relevant professionals to use professional software such as Labelme and Photoshop to enclose the damages along the edges of the above damages with polylines to form closed polygons. See Figure 8 As shown, there is a small gap between the polyline and the crack edge, and the polyline fits closely with other damage edges. When the number of marked photos is limited and the diversity is insufficient, the existing marked pictures can be cropped, rotated, mirrored, and adjusted in brightness and chromaticity to expand the quantity of the annotation data used for training. The partial photos of RC components with damages marked form the dataset of the surface damages of RC components. This dataset is split into a training set, a validation set, and a test set according to the ratio of 70%, 15%, and 15% respectively, which are used to optimize the weight parameters of the damage segmentation model and verify and test the generalization ability and damage recognition performance of the model. When optimizing the weight parameters, the training algebra (epoch), iteration steps, starting layer for weight optimization, size of the photos used for training, and initial learning rate of the model can be set, and a suitable learning rate decay strategy and optimization algorithm can be adopted.

[0137] After training is completed, use the damage segmentation model combined with the optimal weight parameters obtained through training to perform damage recognition. The damage segmentation model adopts the Mask R-CNN model that can perform object classification, localization, and segmentation. Taking the damage recognition result of a partial photo of the front elevation of the beam of a certain RC joint as an example, as Figure 9As shown, the recognition results include damage bounding boxes, damage types, damage segmentation masks, and confidence levels. Among them, the input images for damage recognition are obtained in the following way:

[0138] Use a camera bracket to erect a consumer digital camera so that the camera is kept horizontal and facing the front elevation of the RC member. Take partial photos of the front elevation of the RC member at different positions in the horizontal and vertical directions. The overlap between the partial photos should be in the range of 60% - 80%. During shooting, use a laser rangefinder to measure the distance from the lens to the front elevation of the RC member to calculate the pixel resolution of the photo. Subsequently, perform distortion correction on each partial photo and use an image stitching method based on the SURF algorithm to stitch the partial front-facing photos of the front elevation of each RC member into a panoramic photo including the front elevation of the RC member to delete the overlapping areas between the partial photos. According to the requirements of the damage segmentation model for the input image size, take a panoramic photo of the front elevation of the RC member and appropriately crop the panoramic photo to generate a partial image of the front elevation of the RC member for damage recognition.

[0139] After obtaining the partial image of the front elevation of the RC member for damage recognition, obtain the physical parameters of the damage in the partial image through image processing. Taking the extraction process of a crack in the damage recognition result of the front elevation of the beam of a certain RC node as an example, the specific method is as follows:

[0140] First, intercept the image within the bounding box of the mask of a certain crack and read its grayscale image. As Figure 10 shown, successively adopt median filtering for smoothing, difference between the grayscale image and the smoothed image, removing pixels outside the mask, Otsu threshold segmentation, dilation and erosion, removing isolated connected components, and skeletonization operations to extract real crack pixels and obtain the crack centerline. After obtaining the crack centerline, obtain the crack pixel width at the calculation point by calculating the distance between the intersection points of the central perpendicular line and the crack edge line, and further obtain the maximum crack pixel width along the length of the crack. At the same time, obtain the length of the crack by calculating the length of the crack centerline, and obtain the inclination angle of the crack by calculating the angle between the edge of the minimum circumscribed rectangle of the crack and the horizontal axis. In addition, directly calculate the masked area of cover spalling, concrete crushing, steel bar exposure, and steel bar buckling to obtain the approximate pixel area of such damage. Finally, convert the pixel sizes of cracks, spalling, crushing, steel bar exposure, and buckling into real physical parameters through the pixel resolution of the image.

[0141] In this embodiment, the intelligent damage detection software for RC members is developed using the PyQt interface. The software interface is as Figure 11As shown in the figure. The software integrates the function of recognizing damage in local photos of single RC components and the functions of recognizing damage in local photos of batch components, calculating damage physical parameters, and statistics of damage information. The damage statistics information includes the summary of damage types in the image, the maximum width of cracks, the length of cracks, the spalling area, the crushing area, the area of exposed reinforcement, the area of buckled reinforcement, and the number of broken steel bars. Taking the damage detection of batch local photos of the front elevation of a beam of a certain RC joint as an example, the damage information statistics results are shown in Tables 2 and 3:

[0142] Table 2 Summary of Local Damage Parameters of RC Components

[0143]

[0144] Table 3 Summary of Crack Parameters of RC Components

[0145]

[0146]

[0147] Further obtain the damage information for the quantitative assessment of the damage level of RC components from the damage information statistics results, as shown in Table 4:

[0148] Table 4 Summary of Damage Parameters of RC Components

[0149]

[0150] Subsequently, according to the corresponding relationship between the damage degree and damage level of RC components given in Table 1, evaluate the damage level of RC components. After evaluation, the damage level of the beam components of this RC joint is "IV Severe Damage".

[0151] The above concrete component damage detection and evaluation method, that is, the method for surface damage detection and damage level evaluation of RC components based on the convolutional neural network model, can realize the automatic recognition of various common damages on the surface of RC components in tests, projects, and earthquake disaster sites. The damages that can be recognized include concrete cracks, spalling of the protective layer, concrete crushing, exposed reinforcement, buckling of steel bars, and fracture. Using this method can eliminate the need for inspectors to manually detect and record the damage of RC components on site, reduce the workload, and effectively improve the detection efficiency and accuracy. In addition, this method can realize the automatic positioning, classification, and instance segmentation of surface damages of RC components at one time. The output damage segmentation results are convenient for calculating damage physical parameters. On this basis, combined with a small amount of image processing operations, the accurate calculation of damage geometric parameters can be realized. In addition, this method can automatically generate the damage information statistics results of the surface of RC components based on the graphical user interface, and use the statistical results to quantitatively evaluate the damage level of RC components according to the corresponding relationship between the damage level and damage degree of RC components.

[0152] Such as Figure 12As shown in the figure, in one embodiment, a concrete component damage detection and evaluation system includes an image acquisition module 1210, an annotation module 1220, a model training module 1230, a damage recognition module 1240, an extraction module 1250, and a damage analysis module 1260.

[0153] The image acquisition module 1210 is used to acquire a first image and crop the first image to obtain a second image. The first image is a partial image of a damaged concrete component in existing tests and actual projects, and the second image is a damage image of the concrete component adapted to the damage target segmentation model.

[0154] The annotation module 1220 is used to acquire an annotation file and an annotation image. The annotation file is generated after annotating the second image with a polyline, and the annotation image is the image generated after annotating the second image with the polyline.

[0155] The model training module 1230 is used to process the annotation file to adapt to the damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtain the corresponding optimal weight parameters.

[0156] The damage recognition module 1240 is used to acquire a third image, and based on the optimal weight parameters and the damage target segmentation model, identify the third image to obtain the mask and damage type of the concrete component damage in the third image. The third image is the damage image of the concrete component to be detected.

[0157] The extraction module 1250 is used to, when the damage type of the concrete component to be detected is a crack, extract the first pixel points based on the mask of the crack. The first pixel points are the pixel points corresponding to the actual cracks of the concrete component to be detected.

[0158] The damage analysis module 1260 is used to acquire the pixel resolution of the image acquisition device, and based on the pixel resolution, obtain the geometric parameters corresponding to the first pixel points and other damage masks.

[0159] In this embodiment, for the concrete component damage detection and evaluation system provided by this application, the damage recognition module is specifically used for:

[0160] Taking pictures of the concrete component to be detected multiple times through an image acquisition device to obtain multiple images with overlapping areas.

[0161] Using an image stitching method based on the SURF algorithm to stitch multiple images with overlapping areas into a panoramic image to eliminate the overlapping areas.

[0162] Obtaining the third image by cropping the panoramic image.

[0163] In this embodiment, the concrete component damage detection and evaluation system provided by the present application further includes a data creation module, which is used for:

[0164] Identify the third image through the Mask R-CNN convolutional neural network to obtain the damage type, bounding box, and mask within the bounding box of the concrete component to be detected.

[0165] Establish a damage grade standard corresponding to the damage degree, where the damage degree is a pre-set configuration parameter for determining the damage grade.

[0166] Create a database based on the damage type, damage degree, and damage grade standard.

[0167] In this embodiment, the concrete component damage detection and evaluation system provided by the present application further includes a cropping module and an image processing module.

[0168] The cropping module is used to crop the image within the bounding box where the mask is located.

[0169] The image processing module is used to filter and denoise the image within the bounding box, remove pixel points outside the crack mask, and perform threshold segmentation and morphological analysis to extract the first pixel points.

[0170] In this embodiment, the damage analysis module of the concrete component damage detection and evaluation system provided by the present application is specifically used for:

[0171] Obtain the internal and external parameters of the image acquisition device after calibration.

[0172] Calculate based on the calibrated internal and external parameters to obtain the pixel resolution.

[0173] In this embodiment, the concrete component damage detection and evaluation system provided by the present application further includes a geometric analysis module, which is used for:

[0174] After extracting the first pixel points, determine the center line of the first pixel points through morphological analysis.

[0175] Based on the distance between the center line and the edge of the first pixel points, obtain the geometric parameters of the first pixel points.

[0176] Obtain the inclination angle of the first pixel points through the inclination angle of the minimum circumscribed rectangle of the first pixel points.

[0177] In one embodiment, a computer device is provided. This computer device can be an intelligent terminal, and its internal structure diagram can be as Figure 13As shown. The computer device includes a processor, a memory, and a network interface connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements a method for detecting and evaluating damage to concrete components.

[0178] Those skilled in the art can understand that Figure 13 the structure shown in [the figure] is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different component layout.

[0179] In one embodiment, a computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it implements the steps in the above method embodiments.

[0180] In one embodiment, a computer storage medium stores a computer program. When the computer program is executed by a processor, it implements the steps in the above method embodiments.

[0181] In one embodiment, a computer program product or a computer program is provided. The computer program product or the computer program includes computer instructions. The computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the above method embodiments.

[0182] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in this application can include non-volatile and / or volatile memories. Non-volatile memories can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memories can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc.

[0183] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0184] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention should be subject to the appended claims.

Claims

1. A method for detecting and evaluating damage of concrete components, characterized in that, the method includes: Obtaining a first image and cropping the first image to obtain a second image. The first image is a partial image of a damaged concrete component in existing tests and actual projects, and the second image is a damage image of the concrete component adapted to the damage target segmentation model; Obtaining an annotation file and an annotation image. The annotation file is generated after annotating the second image with polylines, and the annotation image is an image generated after annotating the second image with the polylines; Processing the annotation file to adapt it to the damage target segmentation model, and after training the target segmentation model with the annotation image and the annotation file, obtaining corresponding optimal weight parameters; Obtaining a third image, and identifying the third image based on the optimal weight parameters and the damage target segmentation model to obtain a mask and a damage type of the concrete component damage in the third image. The third image is a damage image of the concrete component to be detected; When the damage type of the concrete component to be detected is a crack, extracting first pixel points based on the mask of the crack. The first pixel points are pixel points corresponding to the actual cracks of the concrete component to be detected; Obtaining the pixel resolution of the image acquisition device, and obtaining geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

2. The method for detecting and evaluating damage of concrete components according to claim 1, characterized in that, the obtaining the third image and identifying the third image based on the optimal weight parameters and the damage target segmentation model includes: Taking pictures of the concrete component to be detected multiple times by the image acquisition device to obtain multiple images with overlapping areas; Using an image stitching method based on the SURF algorithm to stitch the multiple images with overlapping areas into a panoramic image to eliminate the overlapping areas; Obtaining the third image by cropping the panoramic image.

3. The method for detecting and evaluating damage of concrete components according to claim 1, characterized in that, the method further includes: Identifying the third image through a MaskR-CNN convolutional neural network to obtain the damage type, bounding box and mask located within the bounding box of the concrete component to be detected; Establishing a damage grade standard corresponding to the damage degree. The damage degree is a pre-set configuration parameter for determining the damage grade; Creating a database based on the damage type, damage degree and damage grade standard.

4. The method for detecting and evaluating damage of concrete components according to claim 3, characterized in that, the damage types include concrete cracks, cover spalling, concrete crushing, steel bar exposure, steel bar buckling and fracture; The database includes the damage types, damage grade standards and the corresponding relationship between the damage degree and the damage grade standards, and is used to evaluate the third image to obtain the damage grade of the concrete component to be detected.

5. The method for detecting and evaluating damage of concrete components according to claim 1, characterized in that, The step of obtaining the third image and identifying the third image based on the optimal weight parameters and the damage target segmentation model is followed by: Cropping the image within the bounding box where the crack mask is located; Filtering and denoising the image within the bounding box, removing pixel points outside the crack mask, and performing threshold segmentation and morphological analysis to extract the first pixel points.

6. The method for detecting and evaluating damage of concrete members according to claim 1, wherein, The step of obtaining the pixel resolution of the image acquisition device includes: Obtaining the calibrated internal and external parameters of the image acquisition device; Calculating based on the calibrated internal and external parameters to obtain the pixel resolution.

7. The method for detecting and evaluating damage of concrete members according to any one of claims 1 to 6, wherein, The method further includes: After extracting the first pixel points, determining the center line of the first pixel points through morphological analysis; Obtaining the geometric parameters of the first pixel points based on the distance between the center line and the edge of the first pixel points; Obtaining the inclination angle of the first pixel points through the inclination angle of the minimum circumscribed rectangle of the first pixel points; wherein, the geometric parameters include the width and length of the first pixel points.

8. A system for detecting and evaluating damage of concrete members, wherein, The system includes: An image acquisition module, configured to acquire a first image and crop the first image to obtain a second image, where the first image is a partial image of a damaged concrete member in existing tests and actual projects, and the second image is a damage image of a concrete member adapted to the damage target segmentation model; A labeling module, configured to acquire a labeling file and a labeling image, where the labeling file is generated after labeling the second image with a polyline, and the labeling image is an image generated after labeling the second image with the polyline; A model training module, configured to process the labeling file to adapt to the damage target segmentation model, and after training the target segmentation model with the labeling image and the labeling file, obtain the corresponding optimal weight parameters; A damage identification module, configured to acquire a third image and identify the third image based on the optimal weight parameters and the damage target segmentation model to obtain the mask and damage type of the damage of the concrete member in the third image, where the third image is a damage image of a concrete member to be detected; An extraction module, configured to, when the damage type of the concrete member to be detected is a crack, extract the first pixel points based on the mask of the crack, where the first pixel points are pixel points corresponding to the real cracks of the concrete member to be detected; A damage analysis module, configured to obtain the pixel resolution of the image acquisition device and obtain the geometric parameters corresponding to the first pixel points and other damage masks according to the pixel resolution.

9. An electronic device, including a memory and a processor, where the memory stores a computer program, wherein, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer storage medium, storing a computer program, wherein, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images