Steel bar signal suppression and reinforced concrete structure defect reconstruction method and system

Abstract

The invention relates to the technical field of nondestructive testing and intelligent signal processing, in particular to a reinforcing steel bar signal suppression and reinforced concrete structure defect reconstruction method and system. Feature modeling is carried out through a Transform architecture, a dynamic attention mask mechanism is introduced, explicit modeling and suppression are carried out on steel bar echoes, meanwhile, defect related features are enhanced, and the problem that defect information is prone to being lost due to the fact that a traditional method depends on artificial experience parameters or simple filtering is solved. Meanwhile, weak reflection defect signals such as cracks and cavities in the concrete are enhanced and spatially reconstructed on the basis of inhibiting the interference of the reinforcing steel bars, so that a reconstruction result has higher definition and interpretability, and the imaging quality and defect identification reliability of the ground penetrating radar in a complex reinforcing steel bar environment are improved. Besides, an end-to-end trainable network structure is adopted, so that the method has good generalization ability and anti-noise performance, can adapt to detection scenes in different structural forms and different construction environments, and has high engineering application value.

Description

Technical Field

[0001] This invention relates to the fields of nondestructive testing and intelligent signal processing technology, specifically to a method and system for suppressing steel bar signals and reconstructing defects in reinforced concrete structures. Background Technology

[0002] Ground-penetrating radar (GPR), as a core technology in the field of non-destructive testing of concrete structures, has been widely used in the identification of internal defects in bridges, tunnels, and various building structures due to its non-destructive and high-efficiency advantages, providing important technical support for infrastructure safety assessment. However, in actual engineering inspection scenarios, the steel bars that are widely distributed within concrete structures will strongly scatter radar waves, forming high-amplitude, strong-reflection echoes. These echoes exhibit significant hyperbolic characteristics in GPR images, and their energy is much higher than the weak reflection signals generated by deep defects such as cracks and voids. This can easily cover or interfere with defect signals, significantly increasing the difficulty of defect identification, significantly reducing imaging quality, and seriously affecting the accuracy of inspection results.

[0003] Existing technologies have conducted research on the identification of road and underground structural defects using GPR images, but significant limitations remain. For example, patent CN120783024A proposes a method for detecting hidden defects in urban roads based on YOLO11n-GPR, which achieves automatic identification of cracks and cavities through a lightweight target detection model, but does not address the issue of steel reinforcement interference, and the accuracy of defect identification is significantly affected by steel reinforcement reflection. Patent CN119206354A proposes an image discrimination method for underground structures based on KAN, which can only achieve image-level identification of defect categories and cannot output precise location information of defects, making it difficult to meet practical engineering needs. Patent CN111965711B proposes a method for simulating the development of pavement reflection crack depth based on GPR image forward modeling, which only analyzes the crack evolution law from a theoretical perspective and does not consider the problem of strong steel reinforcement reflection obscuring defect signals, thus limiting its practicality.

[0004] In summary, existing technical solutions either focus on target detection and classification or on idealized forward modeling analysis. They generally fail to effectively distinguish the characteristic differences between strong reflections from reinforcing bars and weak reflections from defects at the signal modeling level. This results in defect information in reinforced concrete structures being easily obscured and difficult to accurately reconstruct, limiting overall recognition accuracy and engineering applicability. Consequently, they cannot meet the high-precision detection requirements of complex structures such as bridges and urban roads. Summary of the Invention

[0005] The purpose of this invention is to propose a method and system for suppressing steel bar signals and reconstructing defects in reinforced concrete structures, thereby solving the problems of identifying defects caused by strong reflection interference from steel bars and the difficulty in effectively separating and reconstructing defect signals in ground-penetrating radar detection of reinforced concrete structures.

[0006] According to a first aspect of the present disclosure, a method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures is provided, comprising the following steps: The original B-scan GPR image is divided into two-dimensional patches of fixed size. Linear projection processing is performed on each patch and position encoding is incorporated to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input. The sequence embedding vector is input into the Transformer encoder, and global feature modeling is carried out with the help of a multi-head self-attention module and a feedforward neural network to extract the rebar response features. Based on the feature information output by the encoder, feature mapping is performed through convolutional layers or multilayer perceptrons to generate rebar energy maps. Based on the rebar energy map, the rebar echo region is determined by an adaptive threshold algorithm and a rebar mask is constructed. Dynamic mask attention is used to apply controllable weights to weaken the features of the rebar echo region, while enhancing the feature response of defect-related regions. The features processed by dynamic mask attention are input into the Transformer decoder, and the GPR image with enhanced defect signal is reconstructed through multi-layer feature fusion and spatial relation recovery operations. A joint loss function is used to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

[0007] In one embodiment, the sequence embedding vector is: in: The initial input sequence, Original B-scan GPR image; For patch, embed functions; For position encoding.

[0008] In one embodiment, the multi-head self-attention module and the feedforward neural network perform global feature modeling in the following way: in, yes Features of the layer yes Features of the layer It is a multi-head self-attention module; It is a feed-forward network that utilizes the output features of the encoder. ; , , These represent the query vector, key vector, and value vector obtained after linear mapping of the input features, respectively. This represents the dimension of the key vector. This represents a function that normalizes the attention weights.

[0009] In one embodiment, the reinforcement energy diagram is generated as follows: in, It is the encoder output feature. Represents the weight matrix. Indicates the bias amount. Represents the activation function. This is a diagram showing the energy distribution of reinforcing steel.

[0010] In one embodiment, the rebar mask is constructed as follows: in, For adaptive threshold, For indicator functions; Dynamic mask attention is: in, This is the steel reinforcement strength inhibition factor.

[0011] In one embodiment, the GPR image after defect signal enhancement is as follows: in, This is a Transformer decoder, and its output is the reconstructed defect image.

[0012] In one embodiment, the joint loss function is: in, This refers to defect reconstruction error; For reinforcement identification and masking accuracy; For noise suppression, These are the weighting coefficients.

[0013] According to a second aspect of the present disclosure, a system for suppressing rebar signals and reconstructing defects in reinforced concrete structures is provided, comprising: The image sequence embedding module divides the original B-scan GPR image into fixed-size two-dimensional patches, performs linear projection processing on each patch and incorporates position encoding to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input. The rebar feature extraction and energy map generation module inputs the sequence embedding vector into the Transformer encoder, and uses a multi-head self-attention module and a feedforward neural network to perform global feature modeling and extract the rebar response features. Based on the feature information output by the encoder, feature mapping is performed through convolutional layers or multilayer perceptrons to generate the rebar energy map. The rebar mask construction and dynamic attention suppression module determines the rebar echo region and constructs a rebar mask based on the rebar energy map using an adaptive threshold algorithm; it then applies controllable weights to weaken the features of the rebar echo region using dynamic mask attention, while simultaneously enhancing the feature response of defect-related regions. The defect image reconstruction module inputs the features after dynamic mask attention processing into the Transformer decoder, and reconstructs the GPR image after the defect signal is enhanced through multi-layer feature fusion and spatial relationship recovery operations. The multi-objective joint optimization module uses a joint loss function to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

[0014] According to a third aspect of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the memory, wherein the processor executes the program to implement the aforementioned method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures.

[0015] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the aforementioned method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures.

[0016] The advantages of the above technical solutions adopted in this invention compared with the prior art are as follows: 1. In the feature modeling stage, a rebar area masking mechanism is introduced to specifically weaken the influence of strong rebar reflection signals. This addresses the core problem of weak reflection of defects caused by rebar echoes in traditional methods, significantly reducing interference on defect identification and allowing deep defect signals to be effectively highlighted.

[0017] 2. By using the Transformer decoder to specifically enhance and spatially reconstruct the weak reflection signals related to defects, the clarity of defect images is significantly improved, and the integrity and interpretability of defect contours are optimized, providing accurate and reliable imaging basis for subsequent engineering interpretation and quantitative analysis.

[0018] 3. The entire processing flow does not rely on manual experience to set thresholds, filtering parameters, etc. It adapts to different burial depths and different arrangement of steel bars through an adaptive mechanism, which can flexibly meet the testing needs of various reinforced concrete structures such as bridges and roads, effectively reducing the threshold for on-site operation.

[0019] 4. Employing an end-to-end trainable network architecture, it eliminates the need for cumbersome steps such as preprocessing, feature extraction, and defect identification. It can automatically learn feature patterns in complex scenarios and possesses excellent noise resistance and generalization capabilities. It can be directly integrated into existing ground-penetrating radar detection systems, adapting to complex detection environments in real-world engineering projects, and providing efficient and reliable technical support for non-destructive testing of infrastructure. Attached Figure Description

[0020] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application.

[0021] Figure 1 Flowchart of the method for suppressing rebar signals and reconstructing defects in reinforced concrete structures; Figure 2 A schematic diagram of the GPR image sequence embedding process; Figure 3 A schematic diagram illustrating the use of dynamic mask attention to suppress strong reflections from reinforcing bars and the enhancement and reconstruction of weak reflection defect signals by the Transformer decoder; Figure 4 The diagram shows a comparison between the method of this invention and the traditional GPR image processing method. The traditional method refers to the time-domain or frequency-domain signal processing method based on manually set parameters in existing ground-penetrating radar image processing. Detailed Implementation

[0022] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0023] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0024] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0025] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of this disclosure. It should be noted that each block in a flowchart or block diagram may represent a module, segment, or portion of code, which may include one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutively represented blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, may be implemented using a dedicated hardware-based system that performs the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0026] Example 1: like Figure 1 As shown, this embodiment provides a method for suppressing rebar signals and reconstructing defects in reinforced concrete structures, including the following steps: S1. Original B-scan GPR image The system divides the data into fixed-size two-dimensional patches. For each patch, linear projection processing is performed, and positional encoding is incorporated to generate a sequence embedding vector. Figure 3 As shown, the format conversion from two-dimensional imaging data to serialized input is completed; Specifically, the two-dimensional GPR imaging data is converted into a serialized input format suitable for Transformer encoder processing, so as to achieve effective embedding and expression of features while maintaining the spatial structure of the image.

[0027] S2. Input the sequence embedding vector into the Transformer encoder, and use a multi-head self-attention module and a feedforward neural network to perform global feature modeling and extract the rebar response features; based on the feature information output by the encoder, perform feature mapping through a convolutional layer or a multilayer perceptron to generate a rebar energy map; Specifically, self-attention computation is used to adaptively model the correlation between features at different spatial locations, thereby effectively extracting rebar response features and suppressing background interference. A feedforward neural network (FFN) is used for feature mapping and enhancement. The rebar energy map is used to locate and characterize the spatial position of rebar echoes in the GPR image, where high-energy regions correspond to strong reflection signals generated by the rebar, providing a basis for subsequent rebar suppression and defect information extraction.

[0028] S3. Based on the rebar energy map, the rebar echo region is determined by an adaptive threshold algorithm and a rebar mask is constructed; dynamic mask attention is used to apply controllable weights to weaken the features of the rebar echo region, while enhancing the feature response of the defect-related region, thereby achieving precise suppression of the rebar signal; Specifically, under the dynamic mask attention mechanism, the feature weights of the area containing the reinforcing bars are weakened, while the weak reflection areas associated with defects are enhanced. Unlike traditional filtering or simple energy reduction methods, this invention introduces adjustable suppression intensity to achieve controllable suppression of the reinforcing bar signal, avoiding excessive weakening of defect information and thus improving the reliability of defect imaging.

[0029] S4. Input the features after dynamic mask attention processing into the Transformer decoder, and reconstruct the GPR image after defect signal enhancement through multi-layer feature fusion and spatial relationship recovery operations; Specifically, this step can effectively enhance weak reflection defect signals such as internal cracks and voids in concrete, making them clearer in the reconstructed image, which is beneficial for subsequent defect identification and engineering interpretation.

[0030] S5. A joint loss function is used to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

[0031] Specifically, during the model training phase, this invention employs a multi-objective joint optimization strategy, simultaneously constraining both the rebar suppression effect and the defect reconstruction quality. By introducing defect reconstruction error, rebar recognition and mask accuracy constraints, and noise suppression terms, the model suppresses rebar interference while maintaining the integrity of defect features during training, improving feature reconstruction quality and overall generalization ability, thereby stably outputting high-quality defect images in complex construction environments.

[0032] Depend on Figure 4 The comparison results show that, after adopting the method of the present invention, the reflection of the reinforcing bars is significantly suppressed, the defect outline is clearer, and the contrast between the defect and the background is significantly improved, which verifies the effectiveness and practicality of the method of the present invention in ground penetrating radar detection of reinforced concrete structures.

[0033] Example 2: This embodiment provides a system for suppressing rebar signals and reconstructing defects in reinforced concrete structures, including: The image sequence embedding module divides the original B-scan GPR image into fixed-size two-dimensional patches, performs linear projection processing on each patch and incorporates position encoding to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input. The rebar feature extraction and energy map generation module inputs the sequence embedding vector into the Transformer encoder, and uses a multi-head self-attention module and a feedforward neural network to perform global feature modeling and extract the rebar response features. Based on the feature information output by the encoder, feature mapping is performed through convolutional layers or multilayer perceptrons to generate the rebar energy map. The rebar mask construction and dynamic attention suppression module determines the rebar echo region and constructs a rebar mask based on the rebar energy map using an adaptive threshold algorithm; it then applies controllable weights to weaken the features of the rebar echo region using dynamic mask attention, while simultaneously enhancing the feature response of defect-related regions. The defect image reconstruction module inputs the features after dynamic mask attention processing into the Transformer decoder, and reconstructs the GPR image after the defect signal is enhanced through multi-layer feature fusion and spatial relationship recovery operations. The multi-objective joint optimization module uses a joint loss function to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

[0034] The above modules can be deployed on the same device or distributed devices; the division of modules is only a functional logic description and does not limit the specific physical boundaries or implementation order.

[0035] Example 3: An electronic device is provided for running the aforementioned "Method for Suppressing Reinforcing Steel Signals and Reconstructing Defects in Reinforced Concrete Structures". The electronic device includes a processor, a memory, and optional communication interfaces / display devices / input devices, etc.; the memory stores a computer program that can run on the processor, and when the processor executes the program, it implements steps S1 to S5 of the method described in Embodiment 1, specifically including but not limited to: S1. Divide the original B-scan GPR image into two-dimensional patches of fixed size, perform linear projection processing on each patch and incorporate position encoding to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input; S2. Input the sequence embedding vector into the Transformer encoder, and use a multi-head self-attention module and a feedforward neural network to perform global feature modeling and extract the rebar response features; based on the feature information output by the encoder, perform feature mapping through a convolutional layer or a multilayer perceptron to generate a rebar energy map; S3. Based on the rebar energy map, the rebar echo region is determined by an adaptive threshold algorithm and a rebar mask is constructed; dynamic mask attention is used to apply controllable weights to weaken the features of the rebar echo region, while enhancing the feature response of the defect-related region. S4. Input the features after dynamic mask attention processing into the Transformer decoder, and reconstruct the GPR image after defect signal enhancement through multi-layer feature fusion and spatial relationship recovery operations; S5. A joint loss function is used to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

[0036] The electronic device hardware can be one of a server, personal computer, workstation, industrial controller, edge computing device, or mobile terminal; the processor can be a general-purpose CPU, GPU, NPU, FPGA, or a combination thereof; the memory can be RAM, ROM, flash memory, or disk array. The device can interact with local / remote data storage (acquiring observation data and outputting inversion results) through a communication interface. The above hardware configuration does not constitute a limitation of the present invention.

[0037] Example 4: A computer-readable storage medium storing a computer program, which, when run on a processor of an electronic device, causes the program to execute the method steps S1 to S5 described in Embodiment 1; the storage medium may be a disk, optical disk, flash memory, solid-state drive, read-only memory, random access memory, or any combination of the above media.

[0038] Those skilled in the art will understand that the modules or steps described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, which can then be stored in a storage device for execution by a computer device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. This disclosure is not limited to any particular combination of hardware and software.

[0039] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0040] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. A method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures, characterized in that, Includes the following steps: The original B-scan GPR image is divided into two-dimensional patches of fixed size. Linear projection processing is performed on each patch and position encoding is incorporated to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input. The sequence embedding vector is input into the Transformer encoder, and global feature modeling is carried out with the help of a multi-head self-attention module and a feedforward neural network to extract the rebar response features. Based on the feature information output by the encoder, feature mapping is performed through convolutional layers or multilayer perceptrons to generate rebar energy maps. Based on the rebar energy map, the rebar echo region is determined by an adaptive threshold algorithm and a rebar mask is constructed. Dynamic mask attention is used to apply controllable weights to weaken the features of the rebar echo region, while enhancing the feature response of defect-related regions. The features processed by dynamic mask attention are input into the Transformer decoder, and the GPR image with enhanced defect signal is reconstructed through multi-layer feature fusion and spatial relation recovery operations. A joint loss function is used to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

2. The method for suppressing rebar signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The sequence embedding vector is: in: The initial input sequence, Original B-scan GPR image; For patch, embed functions; For position encoding.

3. The method for suppressing rebar signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The multi-head self-attention module and the feedforward neural network perform global feature modeling in the following way: in, yes Features of the layer yes Features of the layer It is a multi-head self-attention module; It is a feedforward neural network that utilizes the encoder output features. ; , , These represent the query vector, key vector, and value vector obtained after linear mapping of the input features, respectively. This represents the dimension of the key vector. This represents a function that normalizes the attention weights.

4. The method for suppressing rebar signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The method for generating the steel reinforcement energy diagram is as follows: in, It is the encoder output feature. Represents the weight matrix. Indicates the bias amount. Represents the activation function. This is a diagram showing the energy distribution of reinforcing steel.

5. The method for suppressing rebar signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The method for constructing the steel reinforcement mask is as follows: in, For adaptive threshold, For indicator functions; Dynamic mask attention is: in, This is the steel reinforcement strength inhibition factor.

6. The method for suppressing rebar signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The GPR image after defect signal enhancement is as follows: in, This is a Transformer decoder, and its output is the reconstructed defect image.

7. The method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures according to claim 1, characterized in that, The joint loss function is: in, This refers to defect reconstruction error; For reinforcement identification and masking accuracy; For noise suppression, These are the weighting coefficients.

8. A system for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures, characterized in that, include: The image sequence embedding module divides the original B-scan GPR image into fixed-size two-dimensional patches, performs linear projection processing on each patch and incorporates position encoding to generate a sequence embedding vector, thus completing the format conversion from two-dimensional imaging data to serialized input. The rebar feature extraction and energy map generation module inputs the sequence embedding vector into the Transformer encoder, and uses a multi-head self-attention module and a feedforward neural network to perform global feature modeling and extract the rebar response features. Based on the feature information output by the encoder, feature mapping is performed through convolutional layers or multilayer perceptrons to generate the rebar energy map. The rebar mask construction and dynamic attention suppression module determines the rebar echo region and constructs a rebar mask based on the rebar energy map using an adaptive threshold algorithm; it then applies controllable weights to weaken the features of the rebar echo region using dynamic mask attention, while simultaneously enhancing the feature response of defect-related regions. The defect image reconstruction module inputs the features after dynamic mask attention processing into the Transformer decoder, and reconstructs the GPR image after the defect signal is enhanced through multi-layer feature fusion and spatial relationship recovery operations. The multi-objective joint optimization module uses a joint loss function to simultaneously optimize the reinforcement suppression effect and the defect reconstruction quality.

9. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements the method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for suppressing steel reinforcement signals and reconstructing defects in reinforced concrete structures as described in any one of claims 1-7.

Images