Reinforced concrete defect detection method, system, equipment and medium

Abstract

The invention discloses a reinforced concrete defect detection method, system and equipment and a medium. The method comprises the following steps: acquiring ground penetrating radar data; inputting the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module; performing feature extraction processing on the ground penetrating radar data through a multi-scale feature extraction module to obtain a multi-channel feature vector; performing weighted adjustment processing of channel and spatial dimensions on the multi-channel feature vector through a residual enhancement module to obtain an enhanced feature vector; performing feature reconstruction processing on the enhanced feature vector through a feature fusion module, and outputting to obtain clutter suppression data; and performing defect detection and positioning processing on the clutter suppression data to obtain a defect detection result. The embodiment of the invention can improve the disease identification precision of the ground penetrating radar under the complex working condition in the reinforced concrete structure, and can be widely applied to the technical field of radar detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar detection, and in particular to a reinforced concrete defect detection method, system, device and medium. BACKGROUND

[0002] As a leading non-destructive monitoring technology, ground penetrating radar has been widely used in the health detection of reinforced concrete structures. However, ground penetrating radar often encounters the interference of steel bar clutter and concrete aggregate clutter during the detection of reinforced concrete structures. In related technologies, there are detection methods designed for a single type of clutter, which do not consider the complex superposition effect when steel bar clutter and concrete aggregate clutter act simultaneously. Moreover, these algorithms generally ignore the weakening of the steel shielding effect on the penetration of electromagnetic waves, and fail to effectively enhance the amplitude of the disease signal, resulting in problems such as missed diagnosis and misdiagnosis in disease identification, which affects the defect detection efficiency of reinforced concrete.

[0003] In summary, the technical problems existing in related technologies need to be improved. SUMMARY

[0004] The main purpose of the embodiments of the present application is to provide a reinforced concrete defect detection method, system, device and medium, which can improve the accuracy of defect detection.

[0005] To achieve the above-mentioned purpose, one aspect of the embodiments of the present application provides a reinforced concrete defect detection method, which comprises:

[0006] Obtaining ground penetrating radar data;

[0007] Inputting the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module;

[0008] Performing feature extraction processing on the ground penetrating radar data through the multi-scale feature extraction module to obtain a multi-channel feature vector;

[0009] Performing channel and spatial dimension weighting adjustment processing on the multi-channel feature vector through the residual enhancement module to obtain an enhanced feature vector;

[0010] Performing feature reconstruction processing on the enhanced feature vector through the feature fusion module to output clutter suppression data;

[0011] Performing defect detection and positioning processing on the clutter suppression data to obtain a defect detection result.

[0012] In some embodiments, the performing feature extraction processing on the ground penetrating radar data through the multi-scale feature extraction module to obtain a multi-channel feature vector comprises:

[0013] performing two-dimensional convolution and maximum pooling processing on the ground penetrating radar data to obtain initial spatial features;

[0014] performing down-sampling processing on the initial spatial features to obtain down-sampled features;

[0015] performing multi-channel high-dimensional feature space mapping processing on the down-sampled features to obtain the multi-channel feature vector.

[0016] In some embodiments, the weighting adjustment processing of the multi-channel feature vector in the channel and spatial dimension by the residual enhancement module to obtain an enhanced feature vector comprises:

[0017] inputting the multi-channel feature vector into the residual enhancement module, the residual enhancement module comprising a plurality of convolution attention units combined by residual connection;

[0018] performing channel weighting and spatial weighting processing on the multi-channel feature vector by the convolution attention unit to obtain the enhanced feature vector.

[0019] In some embodiments, the feature reconstruction processing of the enhanced feature vector by the feature fusion module to output to obtain clutter suppression data comprises:

[0020] performing residual convolution, multi-scale feature fusion and chained residual pooling processing on the enhanced feature vector to obtain a fusion feature;

[0021] performing spatial size recovery processing on the fusion feature to obtain a recovered feature;

[0022] performing structure feature adjustment processing on the recovered feature to obtain an adjusted feature;

[0023] performing feature mapping restoration processing on the adjusted feature to obtain the clutter suppression data.

[0024] In some embodiments, the defect detection and positioning processing of the clutter suppression data to obtain a defect detection result comprises:

[0025] performing field quantity iterative updating processing on the clutter suppression data to obtain updated data;

[0026] performing back propagation processing on the updated data with time equal to zero to obtain two-dimensional imaging;

[0027] performing target positioning processing on the two-dimensional imaging to obtain the defect detection result.

[0028] In some embodiments, before the inputting the ground penetrating radar data into the clutter suppression model, the method further comprises pre-training the clutter suppression model, and the pre-training the clutter suppression model comprises:

[0029] performing synthetic data pair construction processing based on forward simulation to obtain a training data set;

[0030] inputting the training data set into a clutter suppression model to obtain a prediction result;

[0031] performing loss calculation processing on the prediction result according to a hybrid loss function to obtain a loss value;

[0032] performing parameter adjustment processing on the clutter suppression model according to the loss value.

[0033] In some embodiments, the performing synthetic data pair construction processing based on forward simulation to obtain a training data set comprises:

[0034] randomly generating a two-dimensional synthetic model pair;

[0035] performing electromagnetic wave propagation modeling processing on the two-dimensional synthetic model pair according to a finite difference time domain method combined with a perfect matched boundary to obtain training radar data;

[0036] performing slicing processing on the training radar data according to a sliding window strategy to obtain the training data set.

[0037] To achieve the above object, another aspect of the embodiment of the present application proposes a reinforced concrete defect detection system, which comprises:

[0038] an acquisition module configured to acquire ground penetrating radar data;

[0039] an input module configured to input the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module;

[0040] the multi-scale feature extraction module is configured to perform feature extraction processing on the ground penetrating radar data to obtain a multi-channel feature vector;

[0041] the residual enhancement module is configured to perform weighted adjustment processing on the channel and spatial dimension of the multi-channel feature vector to obtain an enhanced feature vector;

[0042] the feature fusion module is configured to perform feature reconstruction processing on the enhanced feature vector through the feature fusion module to output clutter suppression data;

[0043] a detection module configured to perform defect detection and positioning processing on the clutter suppression data to obtain a defect detection result.

[0044] To achieve the above object, another aspect of the embodiments of the present application provides an electronic device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the method described above when executing the computer program.

[0045] To achieve the above object, another aspect of the embodiments of the present application provides a computer readable storage medium, which stores a computer program, and the computer program implements the method described above when executed by a processor.

[0046] To achieve the above object, another aspect of the embodiments of the present application provides a computer program product, comprising a computer program, and the computer program implements the method described above when executed by a processor.

[0047] The embodiments of the present application at least have the following beneficial effects: the present application provides a reinforced concrete defect detection method, system, device and medium, the scheme obtains ground penetrating radar data; inputs the ground penetrating radar data into a clutter suppression model, the clutter suppression model includes a multi-scale feature extraction module, a residual enhancement module and a feature fusion module; the multi-scale feature extraction module is used for feature extraction processing of the ground penetrating radar data, and a multi-channel feature vector is obtained; the residual enhancement module is used for weighting adjustment processing of channel and spatial dimension of the multi-channel feature vector, and an enhanced feature vector is obtained; the feature fusion module is used for feature reconstruction processing of the enhanced feature vector, and clutter suppression data is output; the clutter suppression data is subjected to defect detection and positioning processing, and a defect detection result is obtained. The embodiments of the present application suppress the reinforcement and concrete aggregate clutter through the clutter suppression model, and simultaneously enhance the disease weak signal amplitude, which significantly improves the disease identification accuracy, imaging clarity and positioning robustness of the ground penetrating radar under complex working conditions in the reinforced concrete structure. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a flowchart of a reinforced concrete defect detection method provided by the embodiments of the present application;

[0049] Figure 2 is a structural diagram of a clutter suppression model provided by the embodiments of the present application;

[0050] Figure 3 is a flowchart of data set construction provided by the embodiments of the present application;

[0051] Figure 4 is a graph of original radar data processed by non-negative matrix factorization provided by the embodiments of the present application;

[0052] Figure 5 is a graph of radar data processed by the clutter suppression model provided by the embodiments of the present application;

[0053] Figure 6 is a radar data residual graph provided by an embodiment of the present application;

[0054] Figure 7 is a result graph of imaging processing on radar data of Figure 4 ;

[0055] Figure 8 is a result graph of imaging processing on radar data of Figure 5 ;

[0056] Figure 9 is a structural schematic diagram of a reinforced concrete defect detection system provided by an embodiment of the present application;

[0057] Figure 10 is a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not used to limit the present application. When the following description refers to the accompanying drawings, the same numbers in different drawings represent the same or similar elements unless otherwise specified. The implementation described in the following exemplary embodiments does not represent all the implementations consistent with the embodiments of the present application. They are only examples of systems and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0059] It can be understood that the terms “first”, “second”, and the like used in the present application can be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as the first information. Depending on the context, the word “if” as used herein can be interpreted as “when” or “when” or “in response to determining”.

[0060] The terms “at least one”, “multiple”, “each”, “any”, and the like used in the present application include one, two or more than two, multiple includes two or more than two, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0062] Before explaining the embodiments of the present application in detail, some of the nouns and terms involved in the embodiments of the present application are first explained. The nouns and terms involved in the embodiments of the present application are subject to the following explanations.

[0063] 1) Ground Penetrating Radar (GPR): A non-destructive testing technology that uses high-frequency electromagnetic waves to detect underground structures.

[0064] 2) Rebar Clutter: The steel bars inside the concrete generate high-amplitude reflection interference on electromagnetic waves.

[0065] 3) Aggregate Clutter: Random scattering signal caused by uneven particle size and distribution of coarse aggregate.

[0066] 4) RefineNet: A multi-scale feature fusion residual enhancement network framework for preserving high-resolution details.

[0067] 5) PSNR / SSIM: Peak signal-to-noise ratio and structural similarity index, respectively, used to quantify signal fidelity and structural integrity.

[0068] 6) Two-dimensional convolution (Conv2D): A basic operation for extracting local area features from an input image, usually used for spatial feature encoding.

[0069] 7) Two-dimensional transposed convolution (TransConv2D): also known as deconvolution, is used to restore the feature map upsampling to a higher spatial resolution.

[0070] 8) Batch Normalization: Normalizes the feature maps to speed up convergence and stabilize the training process.

[0071] 9) Activation function ReLU / LeakyReLU: A nonlinear function used to introduce activation and sparsity; LeakyReLU retains a weak response in the negative range.

[0072] 10) MaxPooling2D: Takes the maximum value of a local area for dimensionality reduction and extracting dominant features.

[0073] 11) Upsampling / Downsampling: used to enlarge or compress the spatial size of the feature map.

[0074] 12) Sigmoid activation function (Sigmoid): Maps the output to the interval [0, 1], commonly used for final probability output.

[0075] 13) RCU (Residual Convolution Unit): A sub-unit composed of two consecutive convolutions and a residual connection, used for detail enhancement and channel alignment.

[0076] 14) CBAM (Convolutional Block Attention Module): A convolutional block attention module composed of channel attention and spatial attention mechanisms.

[0077] 15) Multi-resolution Fusion: Integrates multi-scale feature information through different scale convolutions and upsampling operations.

[0078] 16) Chained Residual Pooling: Enhances the ability to obtain context semantic information through a series of pooling and convolution units.

[0079] 17) Sum: Element-wise addition operation on multiple branch feature maps, used for feature information fusion.

[0080] 18) Channel attention module: Channel attention module, weights the channel dimension of the feature map, highlights the key channel response.

[0081] 19) Spatial attention module: Spatial attention module, weights the spatial dimension of the feature map, emphasizes the structure of the salient region.

[0082] 20) Global MaxPooling / Global AveragePooling: Global MaxPooling / Global AveragePooling, used to extract global-level channel or spatial statistical features.

[0083] 21) Multi-Layer Perceptron (MLP): Usually contains a set of fully connected layers, used to model nonlinear channel relationships.

[0084] As a leading non-destructive monitoring technology, ground penetrating radar (GPR) has been widely applied in the health detection of reinforced concrete structures, such as disease monitoring behind tunnel lining wall, bridge structure defect detection, airport runway subsidence evaluation, and urban road cavity positioning. GPR transmits high-frequency electromagnetic waves into the concrete structure, captures the reflected electromagnetic wave signals from the internal medium interface and defects, and then evaluates the internal structure state. However, GPR often faces two main types of clutter interference in the detection process of reinforced concrete structures: (1) Steel bar clutter: The steel bar mesh produces strong reflection signals to the incident electromagnetic waves, usually showing high amplitude and continuous distribution of reflection characteristics. This type of clutter not only obscures the target signals near and behind the steel bar, but also makes it difficult for electromagnetic waves to effectively penetrate the steel bar layer due to the significant shielding effect of the steel bar mesh, resulting in extremely weak or even undetectable defect signals in the deep or bottom layers, which severely limits the accurate detection of deep defects in reinforced concrete structures by GPR. (2) Concrete aggregate clutter: There are a large number of randomly distributed aggregate particles inside the concrete, and the dielectric constant of these aggregates differs significantly from that of the concrete matrix, causing significant scattering and reflection of electromagnetic waves during transmission, resulting in incoherent clutter signals. The combined effect of this random clutter and the above-mentioned steel bar clutter greatly reduces the signal-to-noise ratio of the GPR signal and increases the difficulty of defect identification, especially when the target defect size is small or the depth is large, the defect signal may be completely obscured or even disappear, making it difficult to accurately detect.

[0085] In related technologies, related clutter suppression methods such as background averaging and principal component analysis can only effectively suppress early direct waves (surface reflection and antenna coupling signals), and have limited processing capability for incoherent clutter caused by steel bars and concrete aggregates. Although some deep learning methods can improve the quality of GPR signals to some extent, most of these methods are designed for a single type of clutter and do not consider the complex superposition effect when steel bar clutter and concrete aggregate clutter act simultaneously. More importantly, these algorithms generally ignore the weakening of the steel shielding effect on electromagnetic wave penetration, and fail to effectively enhance the amplitude of the defect signal, resulting in problems such as missed detection and false detection in defect identification, especially when the defect occurs in a deep and small volume location. In actual engineering applications such as ballastless track structures, steel bars often exhibit irregular and uneven distribution characteristics, and the robustness and generalization ability of related algorithms in such complex structures are severely insufficient, making it difficult to apply them widely.

[0086] In view of this, the application provides a reinforced concrete defect detection method, system, device and medium. The method comprises the following steps: acquiring ground penetrating radar data; inputting the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module; performing feature extraction processing on the ground penetrating radar data through the multi-scale feature extraction module to obtain a multi-channel feature vector; performing weighted adjustment processing on the multi-channel feature vector through the residual enhancement module in the channel and spatial dimension to obtain an enhanced feature vector; performing feature reconstruction processing on the enhanced feature vector through the feature fusion module to output clutter suppression data; and performing defect detection and positioning processing on the clutter suppression data to obtain a defect detection result. The application suppresses the reinforcement and concrete aggregate clutter through the clutter suppression model, simultaneously enhances the weak signal amplitude of the disease, and significantly improves the disease identification accuracy, imaging clarity and positioning robustness of the ground penetrating radar in the reinforced concrete structure under complex working conditions.

[0087] The application can be used in many general or specific computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0088] Figure 1 is an optional flowchart of a reinforced concrete defect detection method provided by the application, Figure 1 The method in the application can include but is not limited to steps S101 to S106.

[0089] Step S101, acquiring ground penetrating radar data;

[0090] Step S102, inputting the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module;

[0091] Step S103, performing feature extraction processing on the ground penetrating radar data through the multi-scale feature extraction module to obtain a multi-channel feature vector;

[0092] Step S104, the residual enhancement module is used to perform channel and spatial dimension weighting adjustment processing on the multi-channel feature vector, to obtain an enhanced feature vector;

[0093] Step S105, the feature fusion module is used to perform feature reconstruction processing on the enhanced feature vector, to output clutter suppression data;

[0094] Step S106, defect detection and positioning processing are performed on the clutter suppression data, to obtain a defect detection result.

[0095] The steps S101 to S106 shown in the embodiments of the present application can be used to obtain the ground penetrating radar data to be processed in real time by a radar device, or can be used to obtain the ground penetrating radar data to be processed by a database or the like. The ground penetrating radar data is input into the clutter suppression model. Please refer to Figure 2 The embodiments of the present application construct an end-to-end deep convolutional network taking an image semantic segmentation model as a backbone and fusing a channel-spatial attention mechanism. The clutter suppression model includes a multi-scale feature extraction module, a residual enhancement module and a feature fusion module, can highlight the echo features of the disease area from the strong clutter background, and cooperatively suppress the steel continuous mirror reflection signal and the aggregate scattering clutter. The clutter suppression model of the embodiments of the present application performs feature extraction processing on the ground penetrating radar data by the multi-scale feature extraction module, performs channel and spatial dimension weighting adjustment processing on the multi-channel feature vector by the residual enhancement module, performs feature reconstruction processing on the enhanced feature vector by the feature fusion module, and finally outputs the clutter suppression data. Defect detection and positioning processing are performed on the clutter suppression data output by the model, to obtain a defect detection result. The embodiments of the present application can improve the disease identification precision, imaging clarity and positioning robustness of the ground penetrating radar under complex working conditions in the reinforced concrete structure, and meet the engineering detection requirements of high precision and high reliability.

[0096] In some embodiments, the multi-scale feature extraction module is used to perform feature extraction processing on the ground penetrating radar data, to obtain a multi-channel feature vector, including:

[0097] The ground penetrating radar data is subjected to two-dimensional convolution and maximum pooling processing, to obtain an initial spatial feature;

[0098] The initial spatial feature is subjected to down-sampling processing, to obtain a down-sampled feature;

[0099] The down-sampled feature is subjected to multi-channel high-dimensional feature space mapping processing, to obtain the multi-channel feature vector.

[0100] In the embodiments of the present application, the clutter suppression model of the present application can simultaneously suppress the clutter caused by steel bars and concrete aggregates in complex ground penetrating radar images and enhance the weak defect echo signal to improve its recognizability. The multi-scale feature extraction module includes multiple two-dimensional convolutions, maximum pooling layers, down-sampling layers, mapping layers, etc. The multi-scale feature extraction module of the present application extracts the initial spatial features of the input single-channel ground penetrating radar data through multiple two-dimensional convolutions (Conv2D) and maximum pooling layers, and completes down-sampling and maps them to a multi-channel high-dimensional feature space.

[0101] In some embodiments, the weighting adjustment processing of the channel and spatial dimensions of the multi-channel feature vector by the residual enhancement module includes:

[0102] The multi-channel feature vector is input into the residual enhancement module, and the residual enhancement module includes multiple convolution attention units combined through residual connection.

[0103] The multi-channel feature vector is processed by the convolution attention unit for channel weighting and spatial weighting to obtain the enhanced feature vector.

[0104] In the embodiments of the present application, the multi-channel feature vector is input into the residual enhancement module to perform weighting adjustment of the channel and spatial dimensions of the down-sampled feature map to improve the expression ability of the clutter and defect signal. Specifically, the residual enhancement module can improve the capture and expression ability of the network for feature information by fusing residual connection and attention mechanism. The residual enhancement module is composed of two sequential convolution attention units through residual connection. The convolution attention unit can effectively enhance the feature expression ability of the convolutional neural network by combining channel attention and spatial attention mechanism. Specifically, the convolution attention unit first extracts the channel information of the input feature map through global average pooling and maximum pooling, and calculates the channel attention weight through a multi-layer perceptron to perform channel weighting on the input feature. Subsequently, the enhanced feature map is input into the spatial attention module, the spatial maps of the average pooling and the maximum pooling are spliced, and the spatial attention weight is obtained through a convolution layer with a kernel size of 7 to highlight the significant spatial region. Under the action of this double attention mechanism, the convolution attention unit can adaptively strengthen the important channels and key spatial positions in the input feature map, thereby improving the ability of the network to capture key information, effectively suppressing redundant features and noise interference, and improving the overall performance of the model.

[0105] In some embodiments, the feature reconstruction processing of the enhanced feature vector by the feature fusion module includes:

[0106] perform residual convolution, multi-scale feature fusion and chained residual pooling processing on the enhanced feature vector to obtain fused features;

[0107] perform spatial size recovery processing on the fused features to obtain recovered features;

[0108] perform structure feature adjustment processing on the recovered features to obtain adjusted features;

[0109] perform feature mapping restoration processing on the adjusted features to obtain the clutter suppression data.

[0110] In the embodiments of the present application, the feature fusion module further integrates local and global context information through residual convolution units (RCUs), multi-scale feature fusion and chained residual pooling operations, which can enhance the completeness of feature expression. Moreover, the clutter suppression model uses multiple two-dimensional deconvolution layers to gradually restore the feature maps to the original spatial size, and applies a residual enhancement module again to emphasize key structure features. Finally, a two-dimensional convolution layer with a convolution kernel size of 1 and a Sigmoid activation function are used to restore the feature mapping to a single-channel output, which realizes accurate reconstruction of the target defect. The clutter suppression model of the embodiments of the present application can accurately recover the spatial distribution information of structure defects such as voids while effectively suppressing clutter.

[0111] In some embodiments, the defect detection and positioning processing on the clutter suppression data to obtain a defect detection result comprises:

[0112] perform field quantity iterative update processing on the clutter suppression data to obtain updated data;

[0113] perform back propagation processing on the updated data with time being zero to obtain two-dimensional imaging;

[0114] perform target positioning processing on the two-dimensional imaging to obtain the defect detection result.

[0115] In the embodiments of the present application, the clutter-suppressed data is processed for defect detection and positioning using an explosive reflector imaging condition-based reverse time migration algorithm, which can perform high-precision positioning on the radar data after clutter suppression and improve the imaging accuracy of internal defects in the structure. The explosive reflector imaging condition is a special reverse time migration imaging principle based on the assumption of time consistency, which considers that all reflection events occur simultaneously on the explosive reflector. This method does not need to store complete wave field data, so it has high computational efficiency and low memory requirements, and is suitable for engineering applications. The embodiments of the present application accurately simulate the propagation and reflection process of electromagnetic waves in the medium by updating the electric field and magnetic field components in each direction in the electromagnetic field. The algorithm considers parameters such as the dielectric constant, conductivity and magnetic permeability of the material at each spatial grid point, and iteratively updates the field quantities through the finite difference time domain (FDTD) format. The boundary conditions are effectively handled by including additional terms to ensure numerical stability. Finally, by backward propagating the wave field at time zero, the reflected wave field distribution on the target interface is obtained, thereby realizing two-dimensional high-precision imaging of defects. This imaging result can effectively overcome the electromagnetic shielding effect caused by steel bars, enhance the detection capability of deep and small defects, and improve the defect identification and positioning accuracy of the ground penetrating radar system in complex reinforced concrete structures.

[0116] In some embodiments, before the ground penetrating radar data is input into the clutter suppression model, the method further comprises pre-training the clutter suppression model, and the pre-training the clutter suppression model comprises:

[0117] constructing a training data set based on forward simulation and synthetic data;

[0118] inputting the training data set into the clutter suppression model to obtain a prediction result;

[0119] performing loss calculation processing on the prediction result according to a hybrid loss function to obtain a loss value;

[0120] performing parameter adjustment processing on the clutter suppression model according to the loss value.

[0121] In the embodiments of the present application, considering that it is difficult to obtain paired "clutter-containing" and "clutter-removed" data in actual ground penetrating radar detection, in order to meet the training needs of supervised learning, the embodiments of the present application adopt a method based on forward simulation to construct a plurality of synthetic data pairs, and expand the data scale through a sliding window to enhance the network's ability to identify and separate the characteristics of reinforcement and aggregate clutter. Then the training data set is input into the clutter suppression model, and the clutter suppression model adopts a hybrid loss function for network training and optimization. The loss function combines pyramid loss (Pyramid Loss), mean square error loss (MSE Loss) and structural similarity loss (SSIM Loss) to simultaneously consider the consistency of image overall structure, detailed texture and pixel intensity, ensuring that the reconstructed GPR image has good target fidelity and visual clarity. The expression of the hybrid loss function is as follows:

[0122] L Hybrid (X,Y)=L Pyramid (X,Y)+L MSE (X,Y)+L SSIM (X,Y);

[0123] In the formula, X represents the network prediction output, and Y represents the clutter-removed label image.

[0124] The embodiments of the present application also set network training parameters, the optimizer selects an Adam optimizer, the initial learning rate is set to 0.0002, the batch size (Batch Size) is set to 32, the total number of training rounds (Epochs) is set to 250, and the data division ratio is divided into a training set, a validation set and a test set according to an 8:1:1 ratio. Through the reasonable combination of the above loss function and the fine setting of the training strategy, the clutter suppression model of the embodiments of the present application can not only accurately remove the multi-source clutter caused by reinforcement and concrete aggregate in the ground penetrating radar image, but also effectively enhance the echo amplitude and structural clarity of the disease target, can robustly identify and high-fidelity reconstruct the disease information in a complex reinforced concrete environment, and improve the readability of the radar image and the reliability of the engineering application.

[0125] In some embodiments, the synthetic data pair construction process based on forward simulation obtains a training data set, including:

[0126] randomly generating a two-dimensional synthetic model pair;

[0127] performing electromagnetic wave propagation modeling on the two-dimensional synthetic model pair according to the finite difference time domain method combined with a perfect matched boundary to obtain training radar data;

[0128] performing slicing processing on the training radar data according to a sliding window strategy to obtain the training data set.

[0129] In the embodiments of the present application, 200 pairs of two-dimensional synthetic models containing cavity defects are randomly generated. Please refer to Figure 3 In each pair of models, one contains a reinforcement arrangement structure and a non-uniform concrete medium to simulate a radar clutter environment of a complex structure (such as a ballastless track) in reality; the other does not contain reinforcement and the background is a uniform medium, which is used to provide an ideal cavity signal without clutter as a network training label. The number, size and position of the cavities are randomly set among the models to cover various actual working conditions. The finite difference time domain method combined with a perfect matched boundary is used for electromagnetic wave propagation modeling. The excitation signal is a Ricker wavelet with a center frequency of 1.5 GHz and 2.0 GHz, and the simulation duration is randomly set in the range of 10-15 ns to generate radar data with clear waveforms and various echoes. Finally, 400 pairs of radar B-scan images containing reinforcement and aggregate clutter, and corresponding radar data without clutter are obtained, which are used as training radar data. In order to further expand the data volume and improve the defect detail extraction capability, the radar images are cut into pieces using a sliding window strategy. The window size is set to 256x256, and the step size is 64. Finally, 15134 groups of training samples are extracted from the 400 pairs of images, and the training data set is obtained and input into the clutter suppression model for model training.

[0130] Next, combined with specific application examples, the scheme of the embodiments of the present application is described and explained in detail:

[0131] The embodiments of the present application can be applied to the processing scene of ground penetrating radar data, can detect defects of reinforced concrete structures, and can accurately identify defects and suppress multi-source clutter in complex engineering structures. Based on a full-size experimental platform of a ballastless track slab, different size foam blocks are embedded in the concrete mortar layer to simulate cavity defects. Please refer to Figure 4 The collected ground penetrating radar data is obtained after background removal (RNMF processing), in which the clutter produced by the reinforcement and concrete aggregate seriously interferes with the identification of defect echoes, and part of the cavity signal is covered or blurred. Please refer to Figure 5 After applying the clutter suppression model provided by the embodiments of the present application, the radar data is generated, the clutter is effectively suppressed, the cavity defect echo is obviously enhanced, and the outlines of the nine simulated cavities are clearly distinguishable, effectively improving the detection accuracy of defects. Please refer to Figure 6 The embodiments of the present application can also display the clutter suppression model successfully distinguishing the clutter and defect signals through a residual graph, enhancing the visibility and positioning ability of the defect signal. The embodiments of the present application further utilize the reverse time migration algorithm to perform inversion imaging on the de-cluttered data. Please refer to Figure 7 , Figure 7 is a result image of the imaging processing of the radar data corresponding to Figure 4 Please refer to Figure 8 , Figure 8 is a result image of the imaging processing of the radar data corresponding toFigure 5 The contrast can know that the embodiment of the application can improve the spatial resolution and imaging quality of defects. Compared with the unprocessed data, the de-cluttering combined with reverse time migration imaging can accurately locate and clearly display the void distribution in the concrete mortar layer, significantly improving the defect detection effect and engineering application value of ground penetrating radar in reinforced concrete complex structures.

[0132] Please refer to Figure 9 The embodiment of the application also provides a reinforced concrete defect detection system, which can realize the above method, and the system comprises:

[0133] The acquisition module 901 is configured to acquire ground penetrating radar data.

[0134] The input module 902 is configured to input the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model comprises a multi-scale feature extraction module, a residual enhancement module and a feature fusion module.

[0135] The multi-scale feature extraction module 903 is configured to perform feature extraction processing on the ground penetrating radar data to obtain a multi-channel feature vector.

[0136] The residual enhancement module 904 is configured to perform weighted adjustment processing on the channel and spatial dimension of the multi-channel feature vector to obtain an enhanced feature vector.

[0137] The feature fusion module 905 is configured to perform feature reconstruction processing on the enhanced feature vector through the feature fusion module to output clutter suppression data.

[0138] The detection module 906 is configured to perform defect detection and positioning processing on the clutter suppression data to obtain a defect detection result.

[0139] It can be understood that the contents in the above method embodiments are all applicable to the present system embodiment, the present system embodiment specifically realizes the same functions as the above method embodiments, and achieves the same beneficial effects as the above method embodiments.

[0140] The embodiment of the application also provides an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the processor realizes the above method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0141] It can be understood that the contents in the above method embodiments are all applicable to the present device embodiment, the present device embodiment specifically realizes the same functions as the above method embodiments, and achieves the same beneficial effects as the above method embodiments.

[0142] Please refer to Figure 10 , Figure 10 The hardware structure of an electronic device is illustrated, and the electronic device includes:

[0143] The processor 1001 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, etc., and is configured to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0144] The memory 1002 can be implemented by a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory), etc. The memory 1002 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 1002 and are called and executed by the processor 1001 to implement the above-mentioned methods of the embodiments of the present application.

[0145] The input / output interface 1003 is configured to implement information input and output.

[0146] The communication interface 1004 is configured to implement the communication interaction between the device and other devices. The communication can be implemented by a wired manner (for example, a USB, a network cable, etc.) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, etc.).

[0147] The bus 1005 is configured to transmit information between various components (for example, the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004) of the device.

[0148] The processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004 are connected to each other by the bus 1005 to realize the communication connection between them in the device.

[0149] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the above-mentioned method.

[0150] It can be understood that the contents in the above-mentioned method embodiments are all applicable to the present storage medium embodiments. The present storage medium embodiments specifically implement the same functions as the above-mentioned method embodiments, and achieve the same beneficial effects as the above-mentioned method embodiments.

[0151] The embodiment of the present application further provides a computer program product comprising a computer program which, when executed by a processor, implements the method described above.

[0152] It can be understood that the contents in the method embodiments described above are applicable to the program product embodiments, the program product embodiments specifically implement the functions same as those of the method embodiments described above, and achieve the same beneficial effects as those of the method embodiments described above.

[0153] The memory, as a non-transitory computer readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory can include a high-speed random access memory, and can also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged with respect to the processor, and these remote memories can be connected to the processor through a network. Examples of the above network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0154] The method, system, device and medium for detecting defects of reinforced concrete provided by the embodiment of the present application, the scheme obtains ground penetrating radar data; inputs the ground penetrating radar data into a clutter suppression model, the clutter suppression model includes a multi-scale feature extraction module, a residual enhancement module and a feature fusion module; the multi-scale feature extraction module is used for performing feature extraction processing on the ground penetrating radar data to obtain a multi-channel feature vector; the residual enhancement module is used for performing weighted adjustment processing on the channel and the spatial dimension of the multi-channel feature vector to obtain an enhanced feature vector; the feature fusion module is used for performing feature reconstruction processing on the enhanced feature vector to output clutter suppression data; and the clutter suppression data is subjected to defect detection and positioning processing to obtain a defect detection result. The embodiment of the present application suppresses the reinforcement and concrete aggregate clutter through the clutter suppression model, simultaneously enhances the weak signal amplitude of the disease, and significantly improves the disease identification accuracy, imaging clarity and positioning robustness of the ground penetrating radar under complex working conditions of the reinforced concrete structure.

[0155] The clutter suppression model constructed by the embodiment of the present application can effectively identify and remove two types of clutter by fusing channel and spatial attention mechanisms and multi-scale feature fusion, overcoming the limitations of filtering and single clutter suppression algorithms in related technologies. The clutter suppression model provided by the embodiment of the present application not only realizes clutter suppression, but also enhances the echo signal amplitude of the disease area, improves the signal-to-noise ratio, ensures that the disease features are clearly retained in the image after clutter suppression, and improves the accuracy and reliability of defect identification.

[0156] The embodiments of the present application can be suitable for complex track plates and concrete structures with irregularly distributed steel bars by introducing diversified training data and structural design, and have good generalization ability and engineering adaptability. The inverse time migration (RTM) algorithm based on the explosion reflection surface imaging condition can improve high-resolution imaging and accurate positioning of the disease, and the clutter suppression model can overcome the imaging difficulty caused by the shielding effect of the steel bar, thereby improving the engineering practical value of the ground penetrating radar disease detection. Through the synthesis and measured data verification, the embodiments of the present application can effectively reduce the risk of missed judgment and misjudgment, enhance the disease detection capability of the ground penetrating radar in the reinforced concrete structure, and promote the development of nondestructive testing technology of underground structure.

[0157] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0158] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than the figures shown, or combine certain steps, or different steps.

[0159] The system embodiments described above are only schematic, and the units described as separate components can or can not be physically separate, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0160] Those skilled in the art can understand that all or some steps in the above disclosed method, the function modules / units in the system and the device can be implemented as software, firmware, hardware and their appropriate combinations.

[0161] The terms "first", "second", "third", "fourth" and the like (if any) in the specification of the present application and the above-described figures are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0162] It should be understood that, in the application, "at least one" refers to one or more, and "multiple" refers to two or more. "And / or" is used to describe the association relationship of the associated objects, which means that there can be three relationships, for example, "A and / or B" can represent three cases of only A, only B, and A and B existing at the same time, wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can represent a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, and c can be single or multiple.

[0163] In several embodiments provided in the application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the above units is only a logical function division, and actual implementation can have another division manner, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection between systems or units, which can be electrical, mechanical or other forms.

[0164] The units described above as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0165] In addition, each functional unit in each embodiment of the application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0166] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0167] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, and are not limited to the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. A reinforced concrete defect detection method, characterized in that: The method comprises the following steps: Acquisition of ground penetrating radar data; Inputting the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model includes a multi-scale feature extraction module, a residual enhancement module, and a feature fusion module; Performing feature extraction processing on the ground penetrating radar data by the multi-scale feature extraction module to obtain a multi-channel feature vector; Performing a weighted adjustment process of channel and spatial dimensions on the multi-channel feature vector by the residual enhancement module to obtain an enhanced feature vector; Performing feature reconstruction processing on the enhanced feature vector by the feature fusion module, and outputting clutter suppression data; Defect detection and positioning processing are performed on the clutter suppression data to obtain a defect detection result.

2. The method according to claim 1, characterized in that The step of performing feature extraction processing on the ground penetrating radar data by the multi-scale feature extraction module to obtain a multi-channel feature vector includes: Performing two-dimensional convolution and maximum pooling processing on the ground penetrating radar data to obtain initial spatial features; Performing downsampling processing on the initial spatial features to obtain downsampled features; Perform multi-channel high-dimensional feature space mapping processing on the downsampled features to obtain the multi-channel feature vector.

3. The method according to claim 1, characterized in that The step of performing a weighted adjustment process of channel and spatial dimensions on the multi-channel feature vector by the residual enhancement module to obtain an enhanced feature vector includes: Inputting the multi-channel feature vector into the residual enhancement module, wherein the residual enhancement module includes a plurality of convolutional attention units, and the plurality of convolutional attention units are combined via residual connections; The multi-channel feature vector is subjected to channel weighting and spatial weighting processing by the convolutional attention unit to obtain the enhanced feature vector.

4. The method according to claim 1, wherein The step of performing feature reconstruction processing on the enhanced feature vector by the feature fusion module and outputting clutter suppression data includes: Performing residual convolution, multi-scale feature fusion and chain residual pooling on the enhanced feature vector to obtain a fused feature; Performing spatial scale restoration processing on the fused features to obtain restored features; Performing structural feature adjustment processing on the restored features to obtain adjusted features; Performing feature mapping restoration processing on the adjustment features to obtain the clutter suppression data.

5. The method according to claim 1, wherein The performing defect detection and positioning processing on the clutter suppression data to obtain a defect detection result includes: performing field quantity iterative updating processing on the clutter suppression data to obtain updated data; Performing back propagation processing on the update data at time zero to obtain a two-dimensional image; Target positioning processing is performed on the two-dimensional imaging to obtain the defect detection result.

6. The method according to any one of claims 1 to 5, characterized in that Before inputting the ground penetrating radar data into the clutter suppression model, the method further includes pre-training the clutter suppression model, wherein pre-training the clutter suppression model includes: Based on the forward simulation, synthetic data pairs are constructed and processed to obtain a training data set; Inputting the training data set into the clutter suppression model to obtain a prediction result; Performing loss calculation processing on the prediction result according to the hybrid loss function to obtain a loss value; Parameter adjustment processing is performed on the clutter suppression model according to the loss value.

7. The method according to claim 6, characterized in that The synthetic data pair construction process based on forward simulation to obtain a training data set includes: Randomly generate pairs of 2D synthetic models; Performing electromagnetic wave propagation modeling processing on the two-dimensional synthetic model pair according to a finite difference time domain method combined with a perfect matching boundary to obtain training radar data; The training radar data is sliced ​​according to a sliding window strategy to obtain the training data set.

8. A reinforced concrete defect detection system, characterized in that: The system comprises: Acquisition module, used to acquire ground penetrating radar data; An input module, configured to input the ground penetrating radar data into a clutter suppression model, wherein the clutter suppression model includes a multi-scale feature extraction module, a residual enhancement module, and a feature fusion module; A multi-scale feature extraction module is used to perform feature extraction processing on the ground penetrating radar data to obtain a multi-channel feature vector; A residual enhancement module, configured to perform a weighted adjustment process on the multi-channel feature vector in terms of channels and spatial dimensions to obtain an enhanced feature vector; a feature fusion module, configured to perform feature reconstruction processing on the enhanced feature vector through the feature fusion module, and output clutter suppression data; The detection module is used to perform defect detection and positioning processing on the clutter suppression data to obtain a defect detection result.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images