An RC structure seismic damage identification and segmentation method, device and electronic equipment

Abstract

The present invention relates to the technical field of seismic damage identification of concrete structures, and particularly to a method, device, electronic device and storage medium for seismic damage identification and segmentation of RC structures. The method includes: acquiring images of damaged RC structures and components, and constructing a target detection data set and a semantic segmentation data set; constructing a damage location model and a semantic segmentation model and training them, and fine-tuning the image super-resolution algorithm using the target detection data set and the semantic segmentation data set; fusing the trained damage location model and the trained semantic segmentation model based on the fine-tuned image super-resolution algorithm to obtain a final damage identification and segmentation model. It breaks through the limitation of the traditional damage segmentation model on the shooting distance of the input image, helps to promote the practical application of advanced computer vision algorithms in actual projects, realizes the long-distance identification and segmentation of seismic damage of RC structures, and improves the efficiency of post-earthquake assessment work.

Description

Technical Field

[0001] This application relates to the technical field of seismic damage identification of concrete structures, and particularly relates to a method, device, electronic device and storage medium for seismic damage identification and segmentation of RC structures. Background Art

[0002] After an earthquake occurs, it is crucial to conduct a reasonable and accurate safety assessment of the damaged RC structure. In recent years, the rapid development of computer vision technology has provided many new paradigms for many problems in the engineering field and has great development potential in the post-earthquake assessment field. The damage location and damage segmentation of RC structures based on computer vision can achieve non-contact rapid damage assessment and have become an important development trend in the post-earthquake assessment field.

[0003] Currently, the damage assessment methods based on computer vision can be mainly divided into damage location (object detection) and damage segmentation (semantic segmentation). The damage location based on object detection can locate and label the damaged parts in the picture with a rectangular box, and the damage segmentation based on semantic segmentation can perform pixel-level segmentation on different damage types in the picture (such as concrete spalling, steel bar exposure, etc.). Currently, the damage location model can achieve damage location at a farther distance (the input of the model can be a picture of the overall structure scale) and has relatively high accuracy. However, due to the lack of damage segmentation, such models cannot provide further guidance for more refined post-earthquake assessment work. Although the semantic segmentation model can further subdivide the damage types of the structure, the input of such models is often high-definition damage pictures taken at a close distance (i.e., pictures of component scale). When inputting a picture of the structure scale into the semantic segmentation model, it is difficult to accurately segment different damage types at the local damage positions on the overall structure. This is because when the local damage on the picture of the structure scale is enlarged, the local damage picture will become very blurred and does not meet the requirements of the existing semantic segmentation model for the quality of the input picture. Therefore, when using the existing semantic segmentation model to quantitatively evaluate the structural damage, the limitation of the model for the input picture results in the need for the evaluator to perform multiple image acquisition operations at the component level, which greatly reduces the efficiency of the post-earthquake assessment work. Summary of the Invention

[0004] This application aims to solve at least one of the technical problems in the related art to some extent.

[0005] To this end, the first object of this application is to propose a method for seismic damage identification and segmentation of RC structures to solve problems such as the difficulty in accurately segmenting different damage types at the local damage positions on the overall structure and not meeting the requirements of the existing semantic segmentation model for the quality of the input picture in the prior art means.

[0006] The second object of the present application is to propose a device.

[0007] The third object of the present application is to propose an electronic device.

[0008] The fourth object of the present application is to propose a computer-readable storage medium.

[0009] To achieve the above object, the first aspect embodiment of the present application proposes an RC structure seismic damage identification and segmentation method, including:

[0010] Obtain images of damaged RC structures and components, and construct a target detection data set and a semantic segmentation data set;

[0011] Construct a damage location model and a semantic segmentation model, and use the target detection data set and the semantic segmentation data set to train the damage location model and the semantic segmentation model respectively, to obtain a trained damage location model and a trained semantic segmentation model;

[0012] Use the target detection data set and the semantic segmentation data set to fine-tune the image super-resolution algorithm;

[0013] Based on the fine-tuned image super-resolution algorithm, fuse the trained damage location model and the trained semantic segmentation model to obtain a final damage identification and segmentation model.

[0014] Preferably, the obtaining images of damaged RC structures and components, and constructing a target detection data set and a semantic segmentation data set includes:

[0015] Obtain images of damaged RC structures and components, extract features from the images of the damaged RC structures and components, and perform annotation processing, and construct a target detection data set and a semantic segmentation data set based on the feature maps after annotation processing.

[0016] Preferably, the constructing a damage location model and a semantic segmentation model, and using the target detection data set and the semantic segmentation data set to train the damage location model and the semantic segmentation model respectively, to obtain a trained damage location model and a trained semantic segmentation model includes:

[0017] Construct a damage location model based on the YOLO V5 network, and use the target detection data set to train the damage location model to obtain a trained damage location model;

[0018] Use a network architecture of U-shaped with skip connections based on pyramid multi-scale feature fusion, Haar wavelet downsampling and attention module to construct a semantic segmentation model, and use the semantic segmentation data set to train the semantic segmentation model to obtain a trained semantic segmentation model.

[0019] Preferably, it further includes:

[0020] The loss function used in training is a weighted multi-class cross-entropy loss function, and its calculation formula is:

[0021]

[0022] where E is the weighted cross-entropy loss, α is the weight, k is the total number of classes, α s is the scaled weight, t k is the label, and y k is the predicted value.

[0023] Preferably, the fine-tuning of the image super-resolution algorithm using the object detection dataset and the semantic segmentation dataset includes:

[0024] Re-screen the object detection dataset and the semantic segmentation dataset, summarize the pictures that meet the preset values, and obtain a summarized dataset;

[0025] Use the summarized dataset to fine-tune the weights of the image super-resolution model StableSR to obtain a fine-tuned image super-resolution algorithm.

[0026] Preferably, the construction of the object detection dataset includes:

[0027] Obtain high-definition pictures of building complexes or bridge landscapes, reduce the pictures of damaged RC components at the component level and paste them at random positions on the high-definition pictures of building complexes or bridge landscapes, and use rectangular frames to label the local damages in the pictures.

[0028] Preferably, the annotation processing includes: 8 types of pixels, namely background, concrete spalling, concrete cracking, concrete crushing, longitudinal bar exposure, longitudinal bar buckling, stirrup exposure, and concrete fragments.

[0029] To achieve the above object, the second aspect embodiment of the present application proposes an RC structure seismic damage identification and segmentation device, including:

[0030] An image acquisition module that acquires images of damaged RC structures and components, and constructs an object detection dataset and a semantic segmentation dataset;

[0031] A model construction module that constructs a damage location model and a semantic segmentation model, and uses the object detection dataset and the semantic segmentation dataset to train the damage location model and the semantic segmentation model respectively to obtain a trained damage location model and a trained semantic segmentation model;

[0032] An algorithm fine-tuning module that fine-tunes the image super-resolution algorithm using the object detection dataset and the semantic segmentation dataset;

[0033] The model fusion module fuses the trained damage location model and the trained semantic segmentation model based on the fine-tuned image super-resolution algorithm to obtain the final damage recognition and segmentation model.

[0034] To achieve the above object, an embodiment of the third aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0035] The memory stores computer-executable instructions;

[0036] The processor executes the computer-executable instructions stored in the memory to implement the method described in any one of the above.

[0037] To achieve the above object, an embodiment of the fourth aspect of the present application provides a computer-readable storage medium, including computer-executable instructions stored in the computer-readable storage medium, and the computer-executable instructions are used to implement the method described in any one of the above when executed by a processor.

[0038] A method for RC structure seismic damage recognition and segmentation provided by the present application first uses an object detection algorithm to locate local damage on the structural scale, then separately extracts the local damage area. The extracted local damage pictures are input into an image super-resolution algorithm for repair, and then the repaired local damage pictures are input into a semantic segmentation model for damage segmentation. Finally, the segmentation results are re-overlaid at the corresponding positions of the original pictures as the output. It breaks through the limitation of the traditional damage segmentation model on the shooting distance of the input pictures, helps to promote the practical application of advanced computer vision algorithms in actual engineering, realizes long-distance recognition and segmentation of RC structure seismic damage, and improves the efficiency of post-earthquake assessment work.

[0039] Some of the additional aspects and advantages of the present application will be given in the following description, some will become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and / or additional aspects and advantages of the present application will become obvious and easy to understand from the following description of the embodiments in conjunction with the drawings, where:

[0041] Figure 1 is a flowchart of the first specific embodiment of a method for RC structure seismic damage recognition and segmentation provided by the present invention;

[0042] Figure 2 is a schematic diagram of the semantic segmentation network model architecture;

[0043] Figure 3Schematic diagram of the framework of the seismic damage long-distance recognition and segmentation method for RC structures based on image super-resolution;

[0044] Figure 4 Structural block diagram of a seismic damage recognition and segmentation device for RC structures provided by an embodiment of the present invention. Detailed implementation manners

[0045] The core of the present invention is to provide a seismic damage recognition and segmentation method, device, electronic device and storage medium for RC structures. By constructing a damage location model and a semantic segmentation model, the long-distance recognition and segmentation of seismic damage of RC structures are realized, and the efficiency of post-earthquake assessment work is improved.

[0046] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention will be further described in detail below with reference to the drawings and specific implementation manners. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. 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.

[0047] Please refer to Figure 1 , Figure 1 which is a flowchart of the first specific embodiment of a seismic damage recognition and segmentation method for RC structures provided by the present invention; the specific operation steps are as follows:

[0048] Step S101: Obtain images of damaged RC structures and components, and construct a target detection data set and a semantic segmentation data set;

[0049] Obtain images of damaged RC structures and components, extract features from the images of the damaged RC structures and components, and perform annotation processing. Based on the feature maps after annotation processing, construct a target detection data set and a semantic segmentation data set.

[0050] The construction of the target detection data set includes:

[0051] Obtain high-definition building complex or bridge landscape pictures, paste the pictures of damaged RC components at the component level after shrinking them at random positions on the high-definition building complex or the bridge landscape pictures, and use rectangular frames to label the local damages in the pictures.

[0052] The annotation processing includes: 8 types of pixels, namely background, concrete spalling, concrete cracking, concrete crushing, exposed longitudinal reinforcement, buckling of longitudinal reinforcement, exposed stirrups, and concrete fragments.

[0053] Step S102: Construct a damage location model and a semantic segmentation model, and use the object detection dataset and the semantic segmentation dataset to train the damage location model and the semantic segmentation model respectively to obtain a trained damage location model and a trained semantic segmentation model;

[0054] Construct a damage location model based on the YOLO V5 network, and use the object detection dataset to train the damage location model to obtain a trained damage location model;

[0055] Construct a semantic segmentation model using a network architecture with pyramid multi-scale feature fusion, Haar wavelet downsampling, and an attention module and a U-shaped network with skip connections, and use the semantic segmentation dataset to train the semantic segmentation model to obtain a trained semantic segmentation model.

[0056] The loss function used in training is a weighted multi-class cross-entropy loss function, and its calculation formula is:

[0057]

[0058] where E is the weighted cross-entropy loss, α is the weight, k is the total number of classes, α s is the scaled weight, t k is the label, and y k is the predicted value.

[0059] Step S103: Fine-tune the image super-resolution algorithm using the object detection dataset and the semantic segmentation dataset;

[0060] Re-screen the object detection dataset and the semantic segmentation dataset, summarize the images that meet the preset values, and obtain a summary dataset;

[0061] Use the summary dataset to fine-tune the weights of the image super-resolution model StableSR to obtain a fine-tuned image super-resolution algorithm.

[0062] Step S104: Based on the fine-tuned image super-resolution algorithm, fuse the trained damage location model and the trained semantic segmentation model to obtain a final damage recognition and segmentation model.

[0063] This embodiment provides a method for seismic damage identification and segmentation of RC structures. First, a target detection algorithm is used to locate local damages at the structural scale. Subsequently, the local damage areas are separately extracted. The extracted local damage pictures are input into an image super-resolution algorithm for restoration. Then, the restored local damage pictures are input into a semantic segmentation model for damage segmentation. Finally, the segmentation results are re-overlaid at the corresponding positions of the original pictures as the output. This breaks through the limitation of the traditional damage segmentation model on the shooting distance of the input pictures, helps to promote the practical application of advanced computer vision algorithms in actual engineering, realizes the long-distance identification and segmentation of seismic damages of RC structures, and improves the efficiency of post-earthquake assessment work.

[0064] Based on the above embodiment, this embodiment describes the method for seismic damage identification and segmentation of RC structures as follows:

[0065] Construct a new picture dataset of damaged RC structures and components that can be used for target detection and semantic segmentation;

[0066] As Figure 2 shown, the sources of the pictures used mainly include the following: test pictures taken in recent years during the RC beam, column, joint and other tests carried out by the research team to which the applicant belongs, paper illustrations in existing RC component test studies, actual damage pictures of damaged RC structures in actual earthquake disaster areas shown in existing earthquake damage reports, and search engines (such as Google Image). The pictures in the dataset mainly include various different RC structural components (beams, columns, joints, walls, etc.). The pictures in the dataset are taken in many different environments (laboratories, environments with dim light, earthquake disaster areas, etc.), and contain various complex background information and random occlusions (plant occlusions, concrete fragments, etc.). A total of 8 types of pixels are labeled in the dataset pictures, namely background, concrete spalling, concrete cracking, concrete crushing, longitudinal bar exposure, longitudinal bar buckling, stirrup exposure, and concrete fragments. The manual annotation process is based on the labelme annotation tool, and each pixel point in the feature map exported after annotation has a unique category;

[0067] The specific method for making the picture dataset for target detection is as follows: First, several high-definition building complex or bridge landscape pictures mainly of urban street scenes are collected. Subsequently, the damaged RC component pictures at the component level are reduced in size and pasted at random positions on the high-definition building complex or bridge landscape pictures to simulate local damages in the pictures at the structural level. Then, rectangular frames are used to label the local damages in the pictures.

[0068] The manual annotation process of the picture dataset for semantic segmentation is based on the labelme annotation tool, and each pixel point in the feature map exported after annotation has a unique category;

[0069] As Figure 3 shown, the constructed object detection dataset is used to train the damage location model based on the YOLO V5 network structure. Subsequently, a semantic segmentation model with a U-shaped structure with skip connections, including pyramid multi-scale feature fusion, Haar wavelet downsampling, and an attention module, is proposed, and the constructed semantic segmentation dataset is used to train the proposed semantic segmentation model;

[0070] The learning rate scheduler adopts an adaptive learning rate automatic scheduler. The core idea of the adaptive learning rate automatic scheduler algorithm is to sacrifice training time and dynamically adjust the learning rate according to the loss change during the training process. When the loss change is small, that is, when the model tends to be stable, the learning rate is decreased for more fine-tuning; when the loss change is large, the learning rate is increased to accelerate convergence;

[0071] The object detection dataset and the semantic segmentation dataset are used to fine-tune the weights of the image super-resolution algorithm StableSR;

[0072] The object detection dataset and the semantic segmentation dataset are re-screened, and the images with a resolution of 2K and above are aggregated as a high-quality image dataset. The high-quality image dataset is used to fine-tune the weights of the image super-resolution model StableSR to enhance its super-resolution accuracy in the field of damage recognition and avoid excessive artifacts;

[0073] The damage location model and the semantic segmentation model are organically integrated to form a new technical solution, realizing the localization - extraction - super-resolution - segmentation of structural damage and the final output;

[0074] A method for seismic damage recognition and segmentation of RC structures provided by an embodiment of the present invention first uses an object detection algorithm to locate local damage on the structural scale, then separately extracts the local damage area. The extracted local damage images are input into an image super-resolution algorithm for repair, and then the repaired local damage images are input into a semantic segmentation model for damage segmentation. Finally, the segmentation results are re-overlaid at the corresponding positions of the original images as the output. It breaks through the limitation of the traditional damage segmentation model on the shooting distance of the input images, helps to promote the practical application of advanced computer vision algorithms in actual projects, realizes the long-distance recognition and segmentation of seismic damage of RC structures, and improves the efficiency of post-earthquake assessment work.

[0075] Please refer to Figure 4 , Figure 4 which is a structural block diagram of a device for seismic damage recognition and segmentation of RC structures provided by an embodiment of the present invention; the specific device may include:

[0076] The image acquisition module 100 acquires images of damaged RC structures and components, and constructs a target detection dataset and a semantic segmentation dataset;

[0077] The model construction module 200 constructs a damage location model and a semantic segmentation model, and trains the damage location model and the semantic segmentation model respectively using the target detection dataset and the semantic segmentation dataset to obtain a trained damage location model and a trained semantic segmentation model;

[0078] The algorithm fine-tuning module 300 fine-tunes the image super-resolution algorithm using the target detection dataset and the semantic segmentation dataset;

[0079] The model fusion module 400 fuses the trained damage location model and the trained semantic segmentation model based on the fine-tuned image super-resolution algorithm to obtain a final damage recognition and segmentation model.

[0080] An RC structure earthquake damage recognition and segmentation device in this embodiment is used to implement the foregoing RC structure earthquake damage recognition and segmentation method. Therefore, the specific implementation manners in an RC structure earthquake damage recognition and segmentation device can be seen in the embodiment part of the foregoing RC structure earthquake damage recognition and segmentation method. For example, the image acquisition module 100, the model construction module 200, the algorithm fine-tuning module 300, and the model fusion module 400 are respectively used to implement steps S101, S102, S103, and S104 in the foregoing RC structure earthquake damage recognition and segmentation method. Therefore, its specific implementation manners can refer to the descriptions of the corresponding individual part embodiments and will not be elaborated herein.

[0081] To implement the above embodiments, the present application also proposes an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method provided in the foregoing embodiments.

[0082] To implement the above embodiments, the present application also proposes a computer-readable storage medium storing computer execution instructions, and the computer execution instructions are used to implement the method provided in the foregoing embodiments when executed by a processor.

[0083] To implement the above embodiments, the present application also proposes a computer program product including a computer program, and the computer program implements the method provided in the foregoing embodiments when executed by a processor.

[0084] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the present application and other processes all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0085] It should be noted that personal information from users should be collected for legal and reasonable purposes and should not be shared or sold outside of these legal uses. In addition, such collection / sharing should be carried out after obtaining the informed consent of the user, including but not limited to notifying the user to read the user agreement / user notice before using the function and signing an agreement / authorization including authorizing the relevant user information. In addition, any necessary steps should be taken to defend and safeguard access to such personal information data and to ensure that others with access to the personal information data comply with their privacy policies and procedures.

[0086] This application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by restricting data collection and deleting the data. In addition, when applicable, personal identifiers are removed from such personal information to protect the privacy of the user.

[0087] In the description of the foregoing embodiments, the descriptions referring to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0088] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" can explicitly or implicitly include at least one of the features. In the description of the present application, "a plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0089] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a customized logical function or process, and the scope of the preferred implementation of the present application includes additional implementations, where the functions can be executed in a manner that is not in the order shown or discussed, including in a substantially simultaneous manner according to the functions involved or in a reverse order, which should be understood by those skilled in the art of the embodiments of the present application.

[0090] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or in conjunction with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0091] It should be understood that various parts of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc.

[0092] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0093] In addition, each functional unit in various embodiments of the present application may be integrated into one processing module, or each unit may exist physically alone, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0094] The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disc, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present application.

Claims

1. A seismic damage identification and segmentation method for RC structures, characterized in that, Including: Obtain images of damaged RC structures and components, and construct an object detection dataset and a semantic segmentation dataset; Construct a damage location model and a semantic segmentation model, and use the object detection dataset and the semantic segmentation dataset to train the damage location model and the semantic segmentation model respectively to obtain a trained damage location model and a trained semantic segmentation model. Among them, the damage location model is constructed based on the YOLO V5 network, and the object detection dataset is used to train the damage location model to obtain a trained damage location model. The semantic segmentation model is constructed using a network architecture of U-shaped with skip connections based on pyramid multi-scale feature fusion, Haar wavelet downsampling, and attention module, and the semantic segmentation dataset is used to train the semantic segmentation model to obtain a trained semantic segmentation model; Fine-tune the image super-resolution algorithm using the object detection dataset and the semantic segmentation dataset; Fuse the trained damage location model and the trained semantic segmentation model based on the fine-tuned image super-resolution algorithm to obtain a final damage recognition and segmentation model. Among them, the locally damaged images extracted separately are input into the image super-resolution algorithm for repair, and then the repaired locally damaged images are input into the semantic segmentation model for damage segmentation, and then the segmentation results are re-overlaid on the corresponding positions of the original images as the output.

2. The seismic damage identification and segmentation method for RC structures according to claim 1, characterized in that The obtaining of images of damaged RC structures and components and the construction of the object detection dataset and the semantic segmentation dataset include: Obtain images of damaged RC structures and components, extract features from the images of damaged RC structures and components, and perform annotation processing, and construct an object detection dataset and a semantic segmentation dataset based on the feature maps after annotation processing.

3. The seismic damage identification and segmentation method for RC structures according to claim 1, characterized in that Also including: The loss function used in training is a weighted multi-class cross-entropy loss function, and its calculation formula is: Among them, E is the weighted cross-entropy loss, α is the weight, k is the total number of categories, and α s is the scaled weight, t k is the label, and y k is the predicted value.

4. The seismic damage identification and segmentation method for the RC structure according to claim 1, characterized in that, The fine-tuning of the image super-resolution algorithm using the object detection dataset and the semantic segmentation dataset includes: Re-screen the object detection dataset and the semantic segmentation dataset, summarize the images that meet the preset values, and obtain a summary dataset; Use the summary dataset to fine-tune the weights of the image super-resolution model StableSR to obtain a fine-tuned image super-resolution algorithm.

5. The seismic damage identification and segmentation method for RC structures according to claim 2, characterized in that The construction of the object detection dataset includes: Obtain high-definition building complex or bridge landscape pictures, paste the reduced damaged RC component pictures at random positions on the high-definition building complex or the bridge landscape pictures, and use rectangular frames to label the local damages in the pictures.

6. The seismic damage identification and segmentation method for the RC structure according to claim 2, characterized in that The annotation processing includes: 8 types of pixels, namely background, concrete spalling, concrete cracking, concrete crushing, longitudinal bar exposure, longitudinal bar buckling, stirrup exposure, and concrete fragments.

7. An earthquake damage identification and segmentation device for RC structures, characterized in that, Including: An image acquisition module that obtains images of damaged RC structures and components and constructs an object detection dataset and a semantic segmentation dataset; The model construction module constructs a damage location model and a semantic segmentation model, and trains the damage location model and the semantic segmentation model by using the object detection data set and the semantic segmentation data set respectively, so as to obtain a trained damage location model and a trained semantic segmentation model. Among them, the damage location model is constructed based on the YOLO V5 network, and the damage location model is trained by using the object detection data set to obtain a trained damage location model. The semantic segmentation model is constructed by using a network architecture of a U-shaped network with skip connections based on pyramid multi-scale feature fusion, Haar wavelet downsampling and attention module, and the semantic segmentation model is trained by using the semantic segmentation data set to obtain a trained semantic segmentation model; The algorithm fine-tuning module fine-tunes the image super-resolution algorithm by using the object detection data set and the semantic segmentation data set; The model fusion module fuses the trained damage location model and the trained semantic segmentation model based on the fine-tuned image super-resolution algorithm to obtain a final damage recognition and segmentation model. Among them, the locally damaged image extracted separately is input into the image super-resolution algorithm for repair, and then the repaired locally damaged image is input into the semantic segmentation model for damage segmentation, and then the segmentation result is re-overlaid at the corresponding position of the original image as the output.

8. An electronic device, characterized in that, Comprising: A processor, and a memory communicatively connected to the processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to implement the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Computer execution instructions are stored in the computer-readable storage medium, and when the computer execution instructions are executed by the processor, they are used to implement the method according to any one of claims 1-6.

Images