Pipe joint water seepage prediction method and system based on scattered wave field interpretation

Abstract

The application belongs to the technical field of nondestructive testing and image data processing of rail transit tunnel, and specifically discloses a segment joint water seepage prediction method and system based on scattered wave field interpretation, which comprises the following steps: after a radar wave field image is acquired, a metal hyperbolic feature is detected to generate a mask, a metal area is shielded, and a purified wave field is generated by utilizing anisotropic diffusion filling; morphological component separation is performed under a curvelet and wavelet redundant dictionary, independent scattering anomaly components are extracted by eliminating structural textures; a two-dimensional Hilbert transform is performed on the components to extract instantaneous phase features; according to the dielectric constant difference and polarity inversion law, the phase change gradient is calculated to accurately distinguish water seepage and air cavity areas; and finally, engineering diagnosis maps are generated by combining color mapping and geometric measurement. The application effectively eliminates metal and structure interference, and realizes automatic qualitative distinction and quantitative evaluation of air-water diseases.

Description

Technical Field

[0001] This invention belongs to the field of non-destructive testing and image data processing technology for rail transit tunnels, and relates to a method and system for predicting water seepage in tunnel segment joints based on scattered wave field interpretation. Background Technology

[0002] Joints in shield tunnel segments are high-risk areas for groundwater leakage. Ground-penetrating radar (GPR), a non-destructive testing method, can acquire two-dimensional grayscale wavefield images reflecting the electromagnetic reflection characteristics within the medium along the joint direction; these images are commonly referred to as B-Scan images. Due to the differences in dielectric constants between air, water, and concrete, the amplitude and phase of reflected waves generated at different medium interfaces carry crucial information about the type of damage. Therefore, interpreting radar wavefield images to predict joint seepage has significant engineering value.

[0003] Currently, common radar image interpretation methods in engineering practice mainly rely on technicians manually interpreting B-Scan images, observing the shape and amplitude of strongly reflective areas to determine the presence of cavities or water-bearing anomalies. More automated technical solutions include thresholding grayscale images to extract bright anomaly areas, or using pre-trained deep convolutional networks to directly classify the entire image or image blocks into three categories: air, water, and normal, thereby achieving qualitative identification of defects.

[0004] The aforementioned existing technologies have significant technical shortcomings when dealing with the specific scenario of segment joints. First, the metal bolts used to connect the segments at the joint generate a strong hyperbolic diffraction response in radar images, masking any weak seepage or void signals that may exist around them. Manual interpretation or simple thresholding is insufficient to effectively eliminate this metal interference. Second, the multi-layered structure of the concrete joint itself produces linear reflective textures. These textures overlap with void scattering patches in the image, making it easy for classification networks to misjudge structural textures as defects, resulting in numerous false alarms. Most critically, both air voids and water seepage areas appear as bright patches in wavefield amplitude images. Grayscale amplitude alone cannot reliably distinguish the medium properties of the two. Existing solutions lack the ability to deeply interpret signal phase, resulting in a persistent bottleneck in the accurate identification of seepage areas. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above objectives, the present invention proposes the following technical solution: a method for predicting water seepage in pipe segment joints based on scattered wave field interpretation, comprising: S1, acquiring the original two-dimensional wave field grayscale image collected along the pipe segment joint, performing hyperbolic feature detection on it, extracting the pixel coordinate set of the hyperbolic reflection area generated by the metal part and performing morphological dilation operation to generate a metal mask matrix.

[0006] S2. Use a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. Extract the background wave field pixels at the edge of the metal region and perform structural diffusion filling to repair the defective region along the seam extension direction to generate the first purified wave field image.

[0007] S3. Perform sparse decomposition and morphological component separation on the first purified wave field image under a redundant representation dictionary containing a structure dictionary and anomaly dictionary, remove the structure component corresponding to the structure dictionary, and retain the independent scattering anomaly component image corresponding to the anomaly dictionary.

[0008] S4. Perform a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image, and extract the instantaneous phase features of each pixel.

[0009] S5. Based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave, analyze the phase change gradient of the edge of the abnormal region in the instantaneous phase characteristics, identify the pixel clusters that have undergone phase polarity reversal, and determine them as water seepage areas. The remaining pixel clusters belonging to the independent scattering abnormal components are determined as air void areas.

[0010] S6. Map the water seepage area and air void area to different display color channels respectively, and combine them with the instantaneous amplitude parameters obtained from the complex anomalous wave field image to generate an engineering diagnostic map.

[0011] The second aspect of the present invention provides a segment joint seepage prediction system based on scattered wave field interpretation, comprising: a metal mask generation module, which acquires the original two-dimensional wave field grayscale image collected along the segment joint, performs hyperbolic feature detection on it, extracts the pixel coordinate set of the hyperbolic reflection area generated by the metal part and performs morphological dilation operation to generate a metal mask matrix.

[0012] The defective wave field repair module uses a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. It extracts the background wave field pixels at the edge of the metal region and performs structural diffusion filling to repair the defective region along the seam extension direction, generating the first purified wave field image.

[0013] The morphological component separation module performs sparse decomposition and morphological component separation on the first purified wavefield image under a redundant representation dictionary containing a structure dictionary and an anomaly dictionary. It removes the structural components corresponding to the structure dictionary and retains the independent scattering anomaly component images corresponding to the anomaly dictionary.

[0014] The instantaneous phase extraction module performs a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image and extracts the instantaneous phase features of each pixel.

[0015] The air-water medium determination module analyzes the phase change gradient at the edge of the abnormal region in the instantaneous phase characteristics based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave. It identifies the pixel clusters that have undergone phase polarity reversal and determines them as water seepage areas. The remaining pixel clusters belonging to the independent scattering abnormal components are determined as air void areas.

[0016] The engineering diagnostic output module maps water seepage areas and air void areas to different display color channels, and generates an engineering diagnostic map by combining the instantaneous amplitude parameters obtained from complex anomalous wave field images.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) By constructing a metal mask matrix and anisotropically diffusing and filling the defective wave field, the present invention can effectively suppress the hyperbolic diffraction interference generated by metal parts such as connecting bolts, while maintaining the continuity of the background texture of the pipe segment joint, avoiding the masking effect caused by strong metal reflection on subsequent image interpretation, and improving the signal purity of the wave field image.

[0018] (2) This invention uses a redundant representation dictionary to perform morphological component separation on the purified wave field image, decomposes the linear structural texture of the concrete joint and the point-like patch scattering anomalies caused by air and water into different sub-dictionaries, and the independent scattering anomaly component image obtained after removing the structural component eliminates the strong phase interference of the concrete interface, providing a pure wave field without structure wrapping and interference for subsequent phase analysis, and solving the problem that amplitude information cannot distinguish between air and water media.

[0019] (3) This invention constructs a complex anomalous wave field image and extracts instantaneous phase features by performing a two-dimensional Hilbert transform on the image of independent scattering anomalous components. It combines the polarity difference of the reflection coefficient of electromagnetic waves at the interface between concrete and air and water, and uses the phase change gradient as a criterion to identify phase polarity reversal. It can hard distinguish between water seepage areas and air void areas. Compared with the traditional method that simply relies on amplitude intensity, it reduces the false alarm and missed detection probability of air and water identification.

[0020] (4) This invention maps the qualitatively differentiated gas-water medium distribution to different hue channels of the HSV color space, and maps the instantaneous amplitude parameter to the brightness channel. It also calculates quantitative indicators such as the cross-sectional area and extension depth of the seepage area based on the physical resolution, generating an engineering diagnostic map that combines the visualization of disease attributes and the measurement of geometric dimensions. This realizes the integrated automatic processing from image interpretation to engineering identification, providing a reliable and intuitive decision-making basis for tunnel operation and maintenance. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0023] Figure 2 This is a schematic diagram of the system module connections of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 As shown, the method for predicting water seepage in pipe segment joints based on scattered wave field interpretation proposed in this invention includes: S1, acquiring the original two-dimensional wave field grayscale image collected along the pipe segment joint, performing hyperbolic feature detection on it, extracting the pixel coordinate set of the hyperbolic reflection area generated by the metal part and performing morphological dilation operation to generate a metal mask matrix.

[0026] In a preferred embodiment, the pixel coordinate set of the hyperbolic reflection region generated by the metal part is extracted and morphological dilation is performed to generate a metal mask matrix. Specifically, this includes: performing edge enhancement processing on the original two-dimensional wave field grayscale image to generate an edge enhancement image; performing hyperbolic template matching based on Hough transform on the edge enhancement image to detect and extract the set of hyperbolic vertex coordinates and asymptote pixel coordinates that satisfy the reflection characteristics of the metal part, as the pixel coordinate set of the hyperbolic reflection region; and performing dilation operation by expanding a preset morphological structure element outward from the pixel coordinate set as the center to generate a binary matrix with the same size as the original image, wherein the metal region is mapped as the mask value and the non-metal region is mapped as the retained value, thus obtaining the metal mask matrix.

[0027] Specifically, edge enhancement processing is performed on the original two-dimensional wavefield grayscale image acquired along the segment joint. The original two-dimensional wavefield grayscale image is a B-Scan record obtained by ground penetrating radar scanning along the segment joint. The grayscale value of each pixel corresponds to the amplitude intensity of the radar reflected wave at that spatial location. Diffraction waves generated by metal parts such as connecting bolts will form hyperbolic high-brightness reflection clusters. Edge enhancement processing uses spatial gradient masks in two directions, including a horizontal Sobel mask. With vertical Sobel mask Compared with the original two-dimensional wavefield grayscale image, respectively Perform two-dimensional convolution to obtain the horizontal gradient components. and vertical gradient components Edge enhancement image The gradient magnitude is given by the following formula:

[0028] In the formula, and These are the x and y coordinates of the pixel, respectively. This represents the original two-dimensional wavefield grayscale image in coordinates. grayscale value at that location , , Represents discrete two-dimensional convolution operation. , This process sharpens the hyperbolic contours generated by the diffraction of the metal part, while relatively suppressing the background texture, resulting in an edge-enhanced image.

[0029] After obtaining the edge-enhanced image, hyperbolic template matching based on Hough transform is performed on the edge-enhanced image to detect and extract the reflection features of the metal part. The hyperbolic trajectory of the metal part in the original two-dimensional wavefield grayscale image, and its temporal distance relationship mapped to the pixel coordinate system, can be represented by the equation:

[0030] in, Let these be the coordinates of the hyperbola's vertex. The asymptote opening factor is related to the medium wave velocity, sampling time interval, and spatial sampling spacing. It is a dimensionless positive number, which determines the slope of the hyperbola branch. The Hough transform addresses edge enhancement images where the gradient magnitude exceeds a preset threshold. The pixels are considered as candidate points, and for each candidate point Based on the local gradient direction constraint, that is, using the horizontal gradient component and vertical gradient components Calculate the actual gradient direction at the candidate point, and require that the angle deviation between this actual gradient direction and the theoretical normal direction obtained by differentiating the hyperbola equation at that point is less than a preset angle tolerance. Only for candidate points that satisfy this direction constraint, [further details are needed]. , , In the three-dimensional accumulator space with axes as the axis, according to the parametric relationship Voting is accumulated under the condition that the expression within the square root is non-negative. When the accumulated value of a certain parameter combination exceeds a preset peak threshold, it is determined that the parameter set corresponds to a hyperbola generated by a real metal part. For each detected hyperbola, the vertex coordinates are used as the basis for the calculation. Based on, along The axis traverses the horizontal range of the image in integer pixel steps, using the hyperbola equation. Calculate the corresponding The coordinates, after rounding, yield all pixel positions on the hyperbola's trajectory; simultaneously, along the two asymptotes of the hyperbola... Extending outward from the vertex to the image boundary, the pixel positions traversed by the asymptote are also included in the record. The union of the pixel positions of the hyperbola's trajectory and the pixel positions of the extended asymptote constitutes the set of the hyperbola's vertex coordinates and asymptote pixel coordinates, which is the set of pixel coordinates of the hyperbola's reflection region, denoted as . This set fully characterizes the main energy distribution range of diffracted waves from metallic components in the images.

[0031] Subsequently, the pixel coordinate set of the hyperbolic reflection region. Based on this, a morphological dilation operation is performed. The preset morphological structural element has a radius of... A disk-shaped structural element of one pixel. The expansion operation is defined as:

[0032] In the formula, For set Any pixel coordinate in the, structural element The relative offset coordinates within the area. This operation causes the pixel region associated with the metal part to expand uniformly outward, incorporating diffraction sidelobes and energy leakage into the shielding range. The dilation result generates a binarized matrix with the same size as the original two-dimensional wavefield grayscale image, denoted as the metal mask matrix. For any coordinate in the matrix If it belongs to the dilated set of pixel coordinates ,but This is the shielding value corresponding to the metallic region; if it does not belong to the expansion set, then... This is the retention value corresponding to the non-metallic region. Thus, the metal mask matrix used for subsequent metal interference suppression is obtained.

[0033] For example, taking measured radar data of a shield tunnel segment joint as the processing object, the original two-dimensional wavefield grayscale image was acquired along the joint direction by a ground-penetrating radar antenna with a center frequency of 900MHz. The acquisition time window was set to 20ns, the number of time sampling points was 512, the spatial sampling interval was 0.005m, and the resulting image size was 512 pixels × 300 pixels. Edge enhancement threshold. Take 120, in the Hough transform parameter space The search range is 1 to 512. For numbers 1 to 300 Discretization was performed within the range of 0.2 to 2.0 with a step size of 0.05. The peak threshold was set to 80°, and the angle tolerance was set to 10°. Two hyperbolas satisfying the reflection characteristics of the metal part were detected and extracted, with their parameters being the vertex coordinates. , and vertex coordinates , The set of pixel coordinates of the extracted hyperbolic reflection region. It covers a total of 1582 pixel locations. The radius of the disk structural element in morphological dilation. The radius is set to 5 pixels. This value was determined based on wavefield image samples collected by 200 industrial sensors under similar shield tunnel segment conditions. This ensures that the expanded area completely covers the hyperbolic diffraction main lobe and secondary sidelobe energy generated by the metal bolts. A metal mask matrix is ​​generated after expansion. The metallic regions correspond to a shielding value of 0, which visually appears as black shielding patches. The non-metallic seam background regions correspond to a retention value of 1. In subsequent steps, the original two-dimensional wavefield grayscale image will be compared with... Multiplying pixel by pixel will reset the metallic reflection component to zero.

[0034] S2. Use a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. Extract the background wave field pixels at the edge of the metal region and perform structural diffusion filling to repair the defective region along the seam extension direction to generate the first purified wave field image.

[0035] In a preferred embodiment, background wavefield pixels at the edge of the metal region are extracted, and structural diffusion is performed along the seam extension direction to fill and repair the defective area, generating a first purified wavefield image. Specifically, this includes: multiplying the metal mask matrix with the original two-dimensional wavefield grayscale image pixel by pixel to mask the pixel values ​​corresponding to the metal region, resulting in a defective wavefield image containing holes; in the defective wavefield image, pixel sampling is performed along the edge of the metal region defined by the metal mask to extract the set of pixels reflecting the seam background texture as background wavefield pixels; using the background wavefield pixels as the constraint boundary, an anisotropic diffusion partial differential equation is established and iteratively solved; during the iteration process, the diffusion conduction coefficient is controlled so that its value in the direction parallel to the seam extension direction is greater than its value in the direction perpendicular to the seam, until the defective area is repaired, generating a first purified wavefield image with continuous and consistent wavefield texture.

[0036] Specifically, following the metal mask matrix generated in step S1, this metal mask matrix is ​​multiplied pixel-by-pixel with the original two-dimensional wavefield grayscale image to mask the pixel values ​​corresponding to the metal regions. The original two-dimensional wavefield grayscale image is a B-Scan record acquired by ground-penetrating radar along the joint of the tunnel lining segments, denoted as... ,in For spatial sampling column index along the survey line direction, The time-sampling row index represents the grayscale value of each pixel, corresponding to the amplitude intensity of the reflected echo when an electromagnetic wave propagates within the concrete medium. Metal mask matrix. This is the binarized matrix output from step S1, where metallic regions are mapped to a mask value of 0, and non-metallic regions are mapped to a retained value of 1. The pixel-wise multiplication operation is defined as:

[0037] In the formula, Indicates the missing wavefield image in coordinates The pixel values ​​at the location are determined by the metal mask matrix. Since the pixels corresponding to the hyperbolic diffraction region of the metal bolt are assigned a value of 0, multiplying them by the original non-zero grayscale value forces the result to zero, forming black holes. Pixels in the non-metallic seam region, however, retain their original background texture and amplitude information completely after being multiplied by 1. The defective wavefield image is thus a defective wavefield image containing holes. The hole locations in the image correspond to the removed metal diffraction interference areas, whose internal pixel values ​​are zero. The hole boundaries are determined by the outer contour of the shielding area defined by the metal mask.

[0038] Next, in the defective wavefield image, pixel sampling is performed along the edges of the metal regions defined by the metal mask to extract a set of pixels reflecting the seam background texture. The method for determining the edges of the metal regions is as follows: in the metal mask matrix... In the middle, retrieve all that satisfy And at least one pixel exists in its four-neighbor or eight-neighbor domain that satisfies coordinates This is taken as the boundary pixel of the hole. For each pixel coordinate located on the hole boundary... By calculating the metal mask matrix The Euclidean distance transformation is performed and the spatial gradient is calculated to determine the direction of the normal to the outer side of the non-metallic region at the boundary pixel of the hole, and the distance is extended outward. Each pixel is used to read the grayscale value at the corresponding location in the defective wavefield image. Sampling distance Integer values ​​between 3 and 7 pixels are selected to ensure that the sampling subset falls within the effective signal area outside the metal region, while simultaneously capturing the wave field variation trend highly correlated with the texture of the defective area, close to the edge of the hole. All sampled pixel grayscale values, along with their coordinate positions, constitute the background wave field pixel set, denoted as... ,in The total number of sampling points. The grayscale values ​​in the background wavefield pixel set inherit the background texture features such as interlayer reflection and aggregate scattering of the concrete layer at the joint interface, and provide the basis for texture constraints for subsequent hole filling.

[0039] Then, using the background wavefield pixel set as the constraint boundary, an anisotropic diffusion partial differential equation is established and iteratively solved to repair the hole defect area. The initial values ​​of all pixels inside the hole in the defective wavefield image are all zero. The repair process simulates the diffusion propagation of information along a specific dominant direction, gradually extending the background texture at the hole boundary into the hole. The anisotropic diffusion partial differential equation is established in the following form:

[0040] In the formula, Indicates the diffusion time The image grayscale field at time t, with initial conditions as follows: , For spatial gradient operators, For divergence operators, It is the spatially invariant diffusion conduction coefficient tensor within the void region. To control diffusion behavior and preferentially propagate information along the seam extension direction, the diffusion conduction coefficient tensor is defined as a... Positive definite symmetric matrix:

[0041] in, It is a two-dimensional rotation matrix. This is the angle between the seam extension direction and the horizontal axis of the image. The seam extension direction is the survey line direction along which the radar antenna moves circumferentially along the tube segment, corresponding to the horizontal axis direction in the B-Scan image. The rotation matrix is ​​simplified to an identity matrix, and the diffusion conduction coefficient tensor degenerates into a diagonal matrix. The diffusion conductivity is parallel to the direction of seam extension. This represents the diffusion conductivity perpendicular to the seam direction. To achieve structural diffusion, we set... Greater than , specifically take , This results in the wave field texture spreading at a rate 20 times faster along the horizontal survey line than at a rate 20 times faster along the vertical depth direction. The diffusion process only affects pixels inside the hole region; pixel values ​​outside the hole and in the background wave field pixel set remain locked throughout the iteration process. This serves as a constraint boundary, directing useful information from the seam texture into the missing areas.

[0042] Iterative solution employs an explicit finite difference scheme at a time step. Discrete propagation is performed on top. Let... For the first coordinates at the next iteration For the grayscale value at a given location, the iterative update formula for pixels inside the hole region is as follows:

[0043] The second-order partial derivatives in the formula are approximated using the central difference. , , Pixels. Time step Setting it to 0.2 satisfies the stability convergence condition. The iteration termination condition is set to the point where the average change in grayscale value of all pixels within the hole region between two adjacent iterations is lower than a preset threshold. The diffusion process ends when the number of iterations reaches the upper limit of 5000. After anisotropic diffusion iteration and filling, the pixel grayscale values ​​inside the hole region are gradually replaced from zero to smooth transition values ​​that are consistent with the surrounding seam background texture. There are no longer grayscale jumps at the hole boundary, generating a first purified wavefield image with a continuous and consistent wavefield texture. The first purified wavefield image completely preserves the original concrete interlayer reflection and medium scattering information, while the hyperbolic diffraction interference caused by metal bolts has been effectively eliminated, and the electromagnetic wave propagation response near the seam interface is realistically presented.

[0044] For example, continuing from the example obtained in step S1, the metal mask matrix The masking value for the two metal bolt regions in the mask image is 0. The metal mask matrix is ​​compared with the original two-dimensional wavefield grayscale image. After pixel-by-pixel multiplication, the defective wavefield image Two irregular black holes appeared, their outlines perfectly matching the boundary of the expanded hyperbolic region. The sampling distance was used when sampling the edge of the metal region. The pixel value is calculated based on the width of the transition zone in the metal diffraction area of ​​200 industrial sensors under shield tunnel segment conditions. It represents the set of pixels extracted from the outside of the hole boundary to the background wave field. Total number of pixels The diffusion conductivity is 847 in the direction parallel to the seam extension. Take a diffusion conductivity coefficient of 1.0, perpendicular to the seam direction. With a parameter configuration of 0.05, the anisotropic diffusion iterations converged after 1834 iterations. The convergence threshold was... In the first cleaned wavefield image after restoration, the original metal hole area was filled with a texture pattern that naturally continues the reflection phase axis of the adjacent seam layer. The mean gray value inside and outside the hole area was improved from the sharp difference between 142.7 and 0 inside the hole before restoration to a continuous transition state with a deviation of less than 1.5 gray levels between the mean gray value of 139.8 in the hole-filled area and the mean gray value of 141.3 in the neighborhood after restoration. This confirms that structural diffusion filling can effectively restore the texture continuity of the seam wavefield.

[0045] S3. Perform sparse decomposition and morphological component separation on the first purified wave field image under a redundant representation dictionary containing a structure dictionary and anomaly dictionary, remove the structure component corresponding to the structure dictionary, and retain the independent scattering anomaly component image corresponding to the anomaly dictionary.

[0046] In a preferred embodiment, the structural components corresponding to the structural dictionary are discarded, and the independent scattering anomaly component images corresponding to the anomaly dictionary are retained. Specifically, this includes: obtaining a curve wave dictionary constructed based on the linear geometric prior of the segment joint as the structural dictionary, and obtaining a wavelet dictionary constructed based on the isotropic features of local void scattering in the medium as the anomaly dictionary, and combining them to form a redundant representation dictionary; taking the first purified wavefield image as input, and using the alternating projection optimization algorithm, performing iterative sparse coding and image reconstruction under the curve wave dictionary and the wavelet dictionary respectively to minimize the reconstruction error; when the convergence condition is met, the image components reconstructed from the curve wave dictionary are determined to be concrete joint structural components and discarded, while the image components reconstructed from the wavelet dictionary are retained, and the output is the independent scattering anomaly component image containing only the scattering signal of the gas-water anomaly.

[0047] Specifically, a pre-constructed redundant representation dictionary is obtained, which is composed of two sub-dictionaries. The structure dictionary is a curvelet dictionary constructed based on the linear geometric priors of the segment joints, and the anomaly dictionary is a wavelet dictionary constructed based on the isotropic characteristics of scattering from local voids in the medium. The essence of the curvelet dictionary is a family of basis functions divided into multiple scales and directions. Each scale corresponds to a different spatial frequency bandwidth, and each scale is further subdivided into several angular sectors. The basis functions are elongated or wedge-shaped, with their major axis pointing in the same direction as the angular sector. This geometric construction enables the curvelet dictionary to sparsely represent the linear reflection texture of the concrete interface of the segment joints, meaning that the joint structure texture can be reconstructed using only a linear combination of a small number of basis functions. The wavelet dictionary consists of basis functions from the classical two-dimensional discrete wavelet transform, including a low-pass scale function and high-pass wavelet functions in three directions. The basis functions are dot-shaped or square patches, providing a sparse representation of scattering anomalies generated by local isotropic medium voids, bubbles, or water aggregates. The curvelet dictionary is denoted as... The dictionary notation for wavelet is: The two together form a redundant representation dictionary, which is an overcomplete dictionary with a total number of atoms exceeding the dimension of the signal being represented. It can simultaneously provide sparse representation bases that are different from each other for signal components with different morphological features.

[0048] The first purified wave field image generated in step S2 is used as the input image to be decomposed. Its dimension is the number of rows. With column number The product of these factors, where the grayscale value of each pixel corresponds to the wave field amplitude after eliminating metallic interference, is used. The alternating projection optimization algorithm will... Modeled as structural components with abnormal components The superposition, with an additional residual component of finite energy. :

[0049] Structural components The requirement is that it can be derived from the curvilinear dictionary. It is composed of a small number of linearly combined atoms, and the anomalous component The requirement is that it can be derived from a wavelet dictionary. It is composed of a small number of linearly combined atoms. Morphological component separation under this modeling premise is achieved through iterative projection and shrinkage thresholding operations. Initial structural components are set. The zero matrix, initial outlier components For a zero matrix, use iterative indexing. Start from 0.

[0050] In the In this iteration, the current residual for the structural components is first calculated:

[0051] The residual contains image information that has not yet been captured by the two components. The structural components are then... With residual The sum is projected onto the curved substrate for updating. The update process is as follows: Performing a curvelet forward transform yields the curvelet coefficient vector. ,in For the curvelet analysis operator (adjoint operator), the image is mapped to the curvelet coefficient domain; for A soft thresholding process is applied to each coefficient; the soft thresholding function... Defined as:

[0052] In the formula, For a single curve coefficient value, Indicates taking symbols, The preset threshold for the structural components is set based on the noise standard deviation estimated from the uniform background region of the first purified wavefield image. ,Pick This coefficient was determined based on the wavefield images collected by 200 industrial sensors under tunnel segment conditions, after statistical analysis of the curve domain noise characteristics. It effectively distinguishes between coherent information of the linear structure at the joints and random noise. The coefficient vector after soft thresholding is denoted as... Then through the curve wave synthesis operator Perform an inverse transformation to obtain the updated structural components:

[0053] The forward and inverse transformations of the curve wave can be implemented using a fast discretization algorithm of the second-generation curve wave transform. The number of scales is 5, and the second to fourth scales are divided into 8, 16, and 16 angular sectors, respectively, to meet the capture requirements of the linear texture of the pipe segment joint under various orientations.

[0054] Next, the anomalous components are updated. The latest reconstructed structural components are then utilized. Update the residuals to obtain the residuals for the outlier components. Abnormal components With the updated residual The sum is projected onto the wavelet basis. The wavelet dictionary is then used. In terms of analysis operators The image was decomposed into wavelet coefficients of different scales and orientations using a two-dimensional discrete wavelet transform. The selected wavelet type was the Cohen-Daubechies-Feauveau 9 / 7 wavelet, which possesses compact support and bioorthogonality properties. The decomposition level was set to 4 levels. Soft thresholding was applied to the wavelet coefficients. Set as This value, also based on statistics from the aforementioned 200 sets of industrial sensor data, is used to separate point scattering anomalies from background texture. The processed coefficients are then analyzed using a wavelet synthesis operator. Reconstructed into the updated anomalous components:

[0055] A complete alternating projection iteration consists of the above-mentioned structural component updates and outlier component updates. The new total residual is then calculated, and the convergence condition is determined. The convergence condition is based on the rate of decrease of the relative residual energy: the iteration terminates if the relative change in the reconstructed image between two adjacent iterations satisfies the following formula or if the number of iterations reaches the upper limit of 100.

[0056]

[0057] In the formula, This represents the Frobenius norm, which is the square root of the sum of the squares of all elements of the matrix. After iterative convergence, the final decomposed structural components are obtained. and abnormal components .Will The component identified as a concrete joint structure exhibits a continuous horizontal layered texture consistent with the direction of the segment joint, corresponding to linear geometric information reflected between concrete layers, and is therefore discarded in subsequent analysis. The image is identified as an independent scattering anomaly component image, where the gray values ​​of all pixels are composed solely of isotropic scattering signals generated by local voids in media such as air and water, and no longer contain the linear reflection components of the concrete structure. Within the independent scattering anomaly component image, both air void regions and water seepage regions appear as strong amplitude patches caused by differences in dielectric constant relative to the background concrete medium; amplitude information alone cannot directly distinguish the medium properties of the two.

[0058] For example, using the first purified wavefield image with a size of 300 rows × 512 columns generated in steps S1-S2, its noise standard deviation... The value is 4.7, calculated from statistics of the air-to-the-wave region at the top of the image. Curved wave domain threshold. Set to 15.0, wavelet domain threshold. Take 13.2. When the iteration reaches the 26th iteration, the relative residual energy decreases to... The algorithm terminates when the convergence condition is met. The structural components obtained from the decomposition... The layered interface texture of the segment joints was preserved, and its energy accounted for 82.3% of the total energy of the original first cleaned wavefield image. Anomaly components. The image contains three distinct patchy bright areas located near the column coordinate intervals [112,148], [263,301], and [389,423], respectively. These three areas are intertwined with the interlayer reflections of concrete in the original image. However, after morphological component separation, the linear structural interference has been removed, leaving only the point-like and patchy scattering anomalies. This provides a high signal-to-noise ratio wavefield without structural interference for subsequent Hilbert transform and phase polarity determination.

[0059] S4. Perform a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image, and extract the instantaneous phase features of each pixel.

[0060] In a preferred embodiment, a two-dimensional Hilbert transform is performed on the independent scattering anomaly component image to generate a complex anomaly wavefield image, and the instantaneous phase features of each pixel are extracted. Specifically, this includes: performing a one-dimensional fast Fourier transform on each column of the independent scattering anomaly component image to convert the spatial domain image to the frequency domain, obtaining the spectrum of each column of signals; in the frequency domain, performing single-sideband filtering on the spectrum of each column of signals to retain positive frequency components and suppress negative frequency components, and doubling the amplitude of positive frequency components; performing an inverse fast Fourier transform on the processed single-sideband spectrum to obtain a complex matrix after frequency domain transformation; using the independent scattering anomaly component image as the real part and the imaginary part of the complex matrix after frequency domain transformation as the imaginary part, combining them to generate a complex anomaly wavefield image; performing an arctangent operation on the imaginary part value and the real part value of each pixel in the complex anomaly wavefield image to resolve the angle value ranging from -π to π, which is used as the instantaneous phase feature of each pixel.

[0061] Specifically, a two-dimensional Hilbert transform is performed on the independent scattering anomaly component images generated in step S3. Here, the two-dimensional Hilbert transform is essentially a spatial set of images on which a one-dimensional Hilbert transform is independently applied column-by-column along the radar echo time-depth direction. The independent scattering anomaly component images are real-valued grayscale images that retain only the air and water scattering patch signals after morphological component separation, denoted as... ,in This is the column index along the direction of the segment joint survey line, with a value range of [value range missing]. , The row index of the radar echo in the time-depth direction, with a value range of [value missing]. , Representing coordinates The real grayscale value at point is a dimensionless digital amplitude value quantized by the radar data acquisition system. A one-dimensional fast Fourier transform is performed column-by-column on this real grayscale image, converting the spatial domain image to the frequency domain along the time depth direction. For the ... Column signal, i.e., along Column vectors of direction Its one-dimensional discrete Fourier transform is defined as:

[0062] In the formula, The imaginary unit, This is a frequency index, with a value range of [value range missing]. , For the first Column signals in frequency index The complex spectrum value at that location. This complex spectrum contains positive and negative frequency components, symmetrically distributed at... to and to Within the range.

[0063] In the frequency domain of each signal, a single-sideband filter is applied to the spectrum of each signal to retain positive frequency components and suppress negative frequency components. Single-sideband filtering is achieved by constructing a frequency domain filter. Implementation, defined as:

[0064] In the formula, The corresponding DC component is set to zero to eliminate mean offset; to For the positive frequency range, the filter value is set to 2 to compensate for the energy loss after negative frequency suppression and ensure that the ratio of the real and imaginary parts of the analyzed signal is correct. The value is set to 0 for the corresponding Nyquist frequency to suppress high frequencies; to For the corresponding negative frequency range, the filter value is set to 0, resulting in complete suppression. The filtered spectrum is as follows:

[0065] This operation ensures that each signal retains only one-sided positive frequency information in the frequency domain, and the corresponding result in the time domain is the Hilbert transform of that real signal.

[0066] Processed single-sideband spectrum Performing a one-dimensional inverse fast Fourier transform restores the frequency domain signal to the spatial domain. The inverse transform is defined as:

[0067] Inverse transform result It is a complex number, whose real and imaginary parts both vary with spatial coordinates. For all After performing the above operations, the complex matrix after frequency domain transformation is obtained, denoted as . Its dimension is consistent with the independent scattering anomaly component image, which is Line × A column, where each element is a complex number.

[0068] Image of independent scattering anomalies As the real part, the complex matrix after frequency domain transformation As the imaginary part, it is used to generate a complex anomalous wavefield image. :

[0069] In the formula, Indicates taking the complex number The imaginary part of the value is the two-dimensional Hilbert transform result of the image of the independent scattering anomaly components. Complex anomaly wavefield image. It is a two-dimensional matrix with complex elements. Its real part preserves the amplitude distribution of the original anomalous scattering, while its imaginary part encodes the phase shift information of the anomalous scattering signal in space. The real and imaginary parts together constitute the extension of the analytic signal in two-dimensional space.

[0070] Instantaneous phase features are extracted by performing arctangent calculations on the imaginary and real part values ​​of each pixel in the complex anomalous wavefield image. Instantaneous phase features Defined as:

[0071] In the formula, For complex anomalous wavefield images in coordinates The real part of the place, The imaginary part at the same coordinate; Given the arctangent function in the four quadrants, the angle is determined by the ratio of its imaginary to real parts, and the quadrant in which the angle lies is determined by the signs of the two parts. The range of values ​​is thus analytically derived. arrive The angle between them. Instantaneous phase characteristics. The unit of measurement is radians, which reflects the electromagnetic wave's position on the coordinate system. The instantaneous phase state at the time of reflection: a positive value indicates that the reflected wave is in phase with the reference signal, and a negative value indicates that it is out of phase. The continuous change or abrupt change in phase is directly related to the medium properties of the reflecting interface.

[0072] For example, the independent scattering anomaly component image obtained from step S3 has an image size of 300 rows × 512 columns. The transform length of the column-by-column one-dimensional fast Fourier transform is... This equals 300 rows in the image. After single-sideband filtering, the positive frequency range... to Multiply by 2 for compensation, Nyquist frequency Zero cases handled. to The corresponding negative frequency components were completely suppressed to zero. An inverse fast Fourier transform yielded a 300-row × 512-column complex matrix, where the imaginary parts of the three anomalous patch regions exhibited independent distribution characteristics different from their real parts. The resulting complex anomalous wavefield image was generated by combining these components. In the diagram, the instantaneous phase characteristics of the patch located in the column coordinate interval [112, 148] after arctangent calculation range from -2.41 to -1.92 radians in the central region of the patch, exhibiting a sharp jump from positive to negative values ​​at the boundary. The instantaneous phase characteristics of the patch located in the column coordinate interval [263, 301] are concentrated between 1.65 and 2.78 radians, maintaining a positive phase trend at the boundary. This difference in instantaneous phase characteristics stems from the comparison of the dielectric constants of the media on both sides of the reflective interface, providing a calculable physical basis for the subsequent step S5 to distinguish between the seepage zone and the air void zone.

[0073] S5. Based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave, analyze the phase change gradient of the edge of the abnormal region in the instantaneous phase characteristics, identify the pixel clusters that have undergone phase polarity reversal, and determine them as water seepage areas. The remaining pixel clusters belonging to the independent scattering abnormal components are determined as air void areas.

[0074] In a preferred embodiment, pixel clusters exhibiting phase polarity reversal are identified and classified as water seepage areas, while the remaining pixel clusters belonging to independent scattering anomalous components are classified as air void areas. Specifically, this includes: traversing all connected pixel clusters in the image of independent scattering anomalous components, and at the boundary of each pixel cluster, extracting the phase change gradient of the instantaneous phase feature from the background area to the center of the anomalous area; constructing a judgment logic based on the physical laws that the reflection coefficient is positive and the phase does not reverse when concrete enters the air medium, and that the reflection coefficient is negative and the phase undergoes half-wave loss, i.e., polarity reversal, when concrete enters the water medium; marking pixel clusters exhibiting abrupt changes from positive to negative phase and polarity reversal as water seepage areas; and marking pixel clusters exhibiting changes from negative to positive phase or maintaining a continuous positive phase as air void areas.

[0075] Specifically, iterate through all connected pixel clusters in the independent scattering anomaly component image. The independent scattering anomaly component image is a real-valued grayscale image output from step S3, denoted as... This image contains only scattered amplitude signals generated by local voids in media such as air and water; the linear structural texture of the concrete joints has been separated and removed. To identify potential air-water anomaly regions from this image, the following steps were first performed... Apply amplitude threshold segmentation, amplitude threshold The value is based on the root mean square (RMS) value of background noise in the independent scattering anomaly component image after morphological component separation, obtained from wavefield image samples collected by 200 industrial sensors under shield tunnel segment conditions. ,Pick This threshold can reliably distinguish between strongly scattering patches caused by real voids and residual background undulations. Medium grayscale value greater than or equal to The pixel is marked as a foreground pixel and assigned a value of 1; the grayscale value is less than The pixels marked as background pixels are assigned a value of 0, resulting in a binarized anomaly mask. An 8-connected region labeling is performed on the binarized anomaly mask to obtain a set of unconnected connected pixel clusters. Each connected pixel cluster corresponds to the projection range of an anomaly volume, which may be an air void or a water seepage area. The index of the connected pixel cluster is denoted as... , No. The set of pixel coordinates contained in a connected pixel cluster is denoted as .

[0076] For each connected pixel cluster This involves determining the set of boundary pixels for the pixel cluster and calculating the center of the anomalous region. Boundary pixels are obtained through morphological boundary detection: [The text then abruptly shifts to a different topic:] ...binarized anomaly mask... The Middle Local masks corresponding to each connected component ,use A dilation operation is performed on a cross-shaped structuring element of a certain size, and the dilation result is compared with the original local mask. Subtracting the two, the non-zero pixels in the difference image represent the boundary pixel set of the connected pixel cluster, denoted as . The coordinates of the center of the abnormal region are obtained by calculating the arithmetic mean of the coordinates of all pixels within the connected pixel cluster:

[0077] In the formula, For connected pixel clusters The total number of pixels included. For each boundary pixel in the boundary pixel set. Calculate the unit direction vector from the background region to the center of the anomaly region. Its expression is:

[0078] like If the boundary pixel itself is the geometric center, then skip that pixel. Continue along the direction pointing towards the center, starting from the boundary pixel... Starting from the instantaneous phase feature image obtained in step S4 Above, with With the origin as the starting point, along the direction Internal collection Each sampling point, simultaneously along the reverse direction External collection There are 1 sampling point, with a sampling step size of 1 pixel. Take 5. Thus we obtain A sequence of instantaneous phase values ​​for each sampled pixel. Instantaneous phase characteristics. The range of values ​​is arrive The radians are obtained from the real and imaginary parts of the complex anomalous wavefield image in step S4 through a four-quadrant arctangent operation. Phase change gradient. Defined as the average phase difference as it passes through the boundary pixels from the outside to the inside along this direction, using center difference combined with nearest neighbor rounding and phase wrapper operator:

[0079] In the formula, Indicates moving forward from the boundary pixel along the direction pointing to the center. The instantaneous phase value mapped to the effective index at a pixel after rounding. Indicates moving backward in a direction away from the center. The instantaneous phase value at a pixel. This indicates a rounding operation, ensuring that floating-point coordinates are converted into valid discrete image matrix indices; The phase wrapper operator is defined as follows: Used to normalize the phase difference to The interval is used to eliminate the phase entanglement effect.

[0080] For connected pixel clusters After calculating the phase change gradient of all boundary pixels, a decision logic is constructed based on physical laws. The propagation speed of electromagnetic waves in concrete depends on the relative permittivity of the concrete. Its typical value is between 6 and 8; the relative permittivity of air The relative permittivity of water When radar waves are incident perpendicularly from concrete to another medium, the reflection coefficient at the interface... It is determined by the following relation:

[0081] in, and These are the relative permittivity of the media on the incident and transmission sides, respectively. When concrete is the incident side and air is the transmission side, Reflectance coefficient When the phase is positive, the reflected wave is in phase with the incident wave, meaning the phase does not reverse; when the concrete is the incident side and the water is the transmitting side, Reflectance coefficient When the value is negative, the reflected wave experiences a half-wave loss relative to the incident wave, i.e., a phase polarity reversal. In terms of instantaneous phase characteristics, this manifests as a shift in phase from the stable phase in the background region towards the interior of the anomalous region. The drastic jump in phase change is as follows: If the phase change gradient at the boundary of a connected pixel cluster is predominantly negative (i.e., along the direction from the background to the center, the instantaneous phase difference after wrapping shows a sudden negative change, indicating a polarity reversal), the pixel cluster is rigidly marked as a water seepage area. If the phase change gradient shows a positive change or remains continuously positive, it indicates that no polarity reversal has occurred, and the pixel cluster is rigidly marked as an air void area. Specific quantization conditions use a boundary pixel voting mechanism: statistically satisfying... Number of boundary pixels and satisfy Number of boundary pixels ,in As the threshold for determining phase change, take radians. If If more than 60% of the boundary pixels exhibit a significant negative phase abrupt change, then the connected pixel cluster is determined to be a water seepage area; otherwise, if If most phases within a cluster remain positive, it is determined to be an air void region. After performing the above traversal and marking on all connected pixel clusters, step S5 outputs a marking matrix of the same size as the image of the independent scattering anomaly components, where pixels corresponding to water seepage regions are assigned a marking value of 1, pixels corresponding to air void regions are assigned a marking value of 2, and background regions are assigned a marking value of 0.

[0082] For example, the instantaneous phase feature image obtained in step S4 and the independent scattering anomaly component image obtained in step S3 The image size is 300 rows × 512 columns. The root mean square value of background noise in the image of independent scattering anomaly components. The statistical value is 3.8, and the amplitude threshold is... Set to 17.1. Three connected pixel clusters were detected in the binarized anomaly mask, corresponding to patches near the column coordinate intervals [112,148], [263,301], and [389,423], respectively, with center coordinates of (130,184), (282,196), and (406,175), respectively. For the first cluster, the total number of boundary pixels... After sampling step size The phase change gradient is calculated along the center direction, including the 73 boundary pixels. Less than The first cluster, accounting for 77.7%, exceeded the 60% threshold and was therefore rigidly marked as a water seepage area. For the second cluster, with a total of 88 boundary pixels, 71 pixels exhibited positive phase change gradients, accounting for 80.7%. The instantaneous phase within the cluster was continuously positively distributed between 1.65 and 2.78 radians, and it was rigidly marked as an air void region. The third cluster exhibited a mixed ratio of positive and negative phase change gradients in its boundary pixels, but 49 pixels showed positive gradients, accounting for 62.0%, exceeding the 60% threshold, and it was also rigidly marked as an air void region. This completed the qualitative determination of the medium properties of all anomalous patches in the independent scattering anomaly component image.

[0083] S6. Map the water seepage area and air void area to different display color channels respectively, and combine them with the instantaneous amplitude parameters obtained from the complex anomalous wave field image to generate an engineering diagnostic map.

[0084] In a preferred embodiment, generating an engineering diagnostic map specifically includes: performing pseudo-color mapping in the HSV color space; setting the hue channel of pixel clusters marked as water seepage areas to a first preset value; setting the hue channel of pixel clusters marked as air void areas to a second preset value; calculating the modulus of each pixel in the complex anomalous wave field image as an instantaneous amplitude parameter characterizing the scattering intensity of the disease; normalizing the instantaneous amplitude parameter and mapping it to the lightness channel of the HSV color space; and calculating the cross-sectional area and extension depth of the seepage area by extracting the number of pixels marked as water seepage areas and combining it with the physical resolution calibrated during digital acquisition, thereby generating an engineering diagnostic map containing qualitative and quantitative information.

[0085] Specifically, pseudo-color mapping is performed in the HSV color space. The HSV color space consists of three independent components: hue (H), saturation (S), and lightness (V). Hue represents the color type using angle values, saturation represents the purity of the color, and lightness represents the brightness of the color. The hue channel of the pixel clusters corresponding to the water seepage area in the marker matrix output in step S5 is set to a first preset value, and the hue channel of the pixel clusters corresponding to the air void area is set to a second preset value. The first preset value is H=0°, corresponding to the red area, used to visually warn of seepage problems in the engineering diagnostic diagram; the second preset value is H=240°, corresponding to the blue area, used to represent air voids. The saturation channel is uniformly set to a constant value S=1.0 to ensure high color purity and a clear visual distinction from the background grayscale.

[0086] Next, the magnitude of each pixel in the complex anomalous wavefield image is calculated. The complex anomalous wavefield image is generated in step S4 and is denoted as... ,in For the image of independent scattering anomaly components in coordinates The real gray value at that location, This represents the imaginary part of the complex matrix after frequency domain transformation at the same coordinate system. Instantaneous amplitude parameter. Defined as the modulus of the complex number:

[0087] Instantaneous amplitude parameters characterize the electromagnetic wave in coordinate systems. The instantaneous energy intensity at the time of reflection, also known as the envelope amplitude, is measured in digital amplitude values ​​from the radar data acquisition system and is dimensionless. The instantaneous amplitude parameter is normalized. The values ​​of are linearly mapped to the interval 0 to 1, and the normalization formula is:

[0088] In the formula, This represents the global minimum value in the instantaneous amplitude parameter matrix. This is the global maximum value. The normalized instantaneous amplitude parameter. The value ranges from 0 to 1. The normalized instantaneous amplitude parameter is mapped to the lightness channel V of the HSV color space:

[0089] The brightness channel V has a value of 0 corresponding to complete darkness and a value of 1 corresponding to the brightest. This makes areas with higher scattering intensity appear brighter in the engineering diagnostic map, while background areas with lower scattering intensity appear darker. Observers can intuitively judge the strength of reflected energy in the diseased area based on the brightness.

[0090] For background pixels, which are neither water seepage areas nor air cavities, the brightness channel is still mapped by the normalized instantaneous amplitude parameter, the hue channel is set to 0° to visually present a gray tone, and the saturation channel is set to 0, thus degenerating into a grayscale image in pseudo-color rendering. Only abnormal areas are highlighted with color to achieve a visually differentiated presentation of abnormal medium properties.

[0091] After generating the hue, saturation, and lightness components, the three channels of the HSV color space are converted to the RGB color space for display output. The conversion process follows the standard HSV to RGB conversion algorithm: for hue H represented in degrees, H is first mapped to the [0, 360) interval, and the hue sector index is calculated. and sector offset Set three intermediate variables: p, q, and t.

[0092] Based on the value of sector index i, the RGB three-channel values ​​are determined by the following segmented assignments:

[0093] The R, G, and B channels are linearly mapped to integer values ​​from 0 to 255, and combined into 24-bit color pixel values ​​to form a pixel matrix of the pseudo-color engineering diagnostic map.

[0094] To generate an engineering diagnostic map containing both qualitative and quantitative information, the geometric indices of the seepage area are further calculated. The independent connected pixel clusters with a flag value of 1 are extracted from the flag matrix in step S5. For the m-th water body seepage connected pixel cluster, the total number of pixels it contains is extracted and denoted as . The physical resolution calibrated during digital acquisition includes the spatial sampling interval in the horizontal direction. and vertical time sampling interval ,in The speed of electromagnetic wave propagation in concrete Physical spacing converted to depth direction ,coefficient Corresponding electromagnetic wave round-trip path. Cross-sectional area of ​​the m-th seepage zone. Depend on The calculations are in square meters.

[0095] The extension depth index of the m-th seepage zone Defined as the maximum span of the seepage area in the depth direction: extract the coordinates of all pixels in the connected pixel cluster from the marker matrix and calculate its row coordinate range. to ,but The unit is meters. The values ​​of cross-sectional area and depth of penetration are superimposed on the corresponding disease area of ​​the engineering diagnostic map as text annotations, forming an engineering diagnostic map that simultaneously includes a pseudo-color mapping for qualitative differentiation of air and water media and quantitative values ​​of the seepage range.

[0096] For example, following the label matrix obtained in step S5 containing three anomaly markers, the first cluster column coordinate interval [112, 148] is labeled as a water seepage area, and the second cluster [263, 301] and the third cluster [389, 423] are labeled as air void areas. Complex anomaly wavefield image. Corresponding instantaneous amplitude parameter matrix The global minimum value is calculated. The global maximum value is 2.3. The value is 187.6. After normalization, the average brightness of the first cluster of water bodies is 0.73, the average brightness of the second cluster of air voids is 0.58, and the average brightness of the third cluster of air voids is 0.41. In the HSV color space, the first cluster's hue channel 0° corresponds to red with a saturation of 1.0, the second and third clusters' hue channels 240° correspond to blue with a saturation of 1.0, and the background area has a saturation of 0, a hue of 0°, and a brightness determined by the normalized instantaneous amplitude parameter. After conversion to RGB, the red water seepage area and the blue air void area form a sharp contrast against the gray background. Spatial sampling interval during digital acquisition. The electromagnetic wave velocity in concrete is 0.005m. Take 0.118 m / ns as the time sampling interval. The value is 0.039 ns, which translates to the vertical physical distance. The value is 0.0023m. This refers to the total number of pixels in the first cluster of water seepage areas. Given 847, calculate the cross-sectional area. Square meters, with row coordinates in the depth direction ranging from 181 to 207, and extended depth index. Meters. These quantitative indicators, together with the pseudo-color mapping, output an engineering diagnostic map, allowing engineering technicians to directly read the distribution of air and water defects and the scale of seepage inside the joints.

[0097] It should be noted that the independent scattering anomaly component image generated in step S3 only contains patch signals formed by scattering from air and water media, and does not contain real grayscale images of linear concrete joint structure texture; step S4 performs a two-dimensional Hilbert transform on the real grayscale image to upgrade it to a complex anomaly wave field image, so that step S5 can resolve the instantaneous phase features used to distinguish air and water from the complex domain, wherein air and water are both presented as bright real amplitude values ​​in the independent scattering anomaly component image.

[0098] Furthermore, since the first purified wave field image generated in step S2 is an image that has removed the interference of metal hyperbolic diffraction and maintained the continuity of the joint texture; step S3 separates the morphological components of the image, removes the linear reflection components generated by the concrete structure, and generates an independent scattering anomaly component image, so that when performing Hilbert transform in step S4, the strong phase of the concrete structure interface and the weak phase of the disease area are avoided from phase wrapping and interference, providing a high signal-to-noise ratio wave field without structural interference for the phase polarity reversal determination in step S5.

[0099] Please see Figure 2 As shown, the second aspect of the present invention provides a segment joint seepage prediction system based on scattered wave field interpretation, comprising: a metal mask generation module, a defect wave field repair module, a morphological component separation module, an instantaneous phase extraction module, a gas-water medium determination module, and an engineering diagnostic output module.

[0100] The metal mask generation module acquires the original two-dimensional wave field grayscale image collected along the joint of the pipe segment, performs hyperbolic feature detection on it, extracts the pixel coordinate set of the hyperbolic reflection area generated by the metal part, and performs morphological dilation operation to generate a metal mask matrix.

[0101] The defective wave field repair module uses a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. It then extracts the background wave field pixels at the edge of the metal region and performs structural diffusion filling along the seam extension direction to repair the defective region, generating a first purified wave field image.

[0102] The morphological component separation module performs sparse decomposition and morphological component separation on the first purified wavefield image under a redundant representation dictionary containing a structure dictionary and an anomaly dictionary, removes the structural component corresponding to the structure dictionary, and retains the independent scattering anomaly component image corresponding to the anomaly dictionary.

[0103] The instantaneous phase extraction module performs a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image and extracts the instantaneous phase features of each pixel.

[0104] The gas-water medium determination module analyzes the phase change gradient of the edge of the abnormal region in the instantaneous phase characteristics based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave, identifies the pixel clusters that have undergone phase polarity reversal and determines them as water seepage areas, and determines the remaining pixel clusters belonging to independent scattering abnormal components as air void areas.

[0105] The engineering diagnostic output module maps the water seepage area and the air void area to different display color channels, and generates an engineering diagnostic map by combining the instantaneous amplitude parameters obtained from the complex anomalous wave field image.

[0106] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting leakage at pipe segment joints based on scattered wave field interpretation, characterized in that, include: S1. Acquire the original two-dimensional wave field grayscale image collected along the joint of the pipe segment, perform hyperbolic feature detection on it, extract the pixel coordinate set of the hyperbolic reflection area generated by the metal part and perform morphological dilation operation to generate a metal mask matrix. S2. Use a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. Extract the background wave field pixels at the edge of the metal region and perform structural diffusion filling to repair the defective region along the seam extension direction to generate the first purified wave field image. S3. Perform sparse decomposition and morphological component separation on the first purified wave field image under a redundant representation dictionary containing a structure dictionary and anomaly dictionary, remove the structure component corresponding to the structure dictionary, and retain the independent scattering anomaly component image corresponding to the anomaly dictionary. S4. Perform a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image, and extract the instantaneous phase features of each pixel. S5. Based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave, analyze the phase change gradient of the edge of the abnormal region in the instantaneous phase characteristics, identify the pixel clusters that have undergone phase polarity reversal, and determine them as water seepage areas. The remaining pixel clusters belonging to the independent scattering abnormal components are determined as air void areas. S6. Map the water seepage area and air void area to different display color channels respectively, and combine them with the instantaneous amplitude parameters obtained from the complex anomalous wave field image to generate an engineering diagnostic map.

2. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 1, characterized in that, Extract the pixel coordinate set of the hyperbolic reflection region generated by the metal part and perform morphological dilation to generate a metal mask matrix, specifically including: Edge enhancement processing is performed on the original two-dimensional wavefield grayscale image to generate an edge-enhanced image; Hyperbola template matching based on Hough transform is performed on the edge enhancement image to detect and extract the set of hyperbola vertex coordinates and asymptote pixel coordinates that satisfy the reflection characteristics of the metal part, which is used as the set of pixel coordinates of the hyperbola reflection region. Centered on the set of pixel coordinates, a dilation operation is performed by expanding outwards using preset morphological structural elements to generate a binary matrix with the same size as the original image. Metal regions are mapped as masking values, and non-metal regions are mapped as retained values, resulting in a metal mask matrix.

3. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 2, characterized in that, Background wavefield pixels at the edge of the metal region are extracted, and structural diffusion filling is performed along the seam extension direction to repair the defective area, generating the first purified wavefield image, specifically including: The metal mask matrix is ​​multiplied pixel by pixel with the original two-dimensional wave field grayscale image to mask the pixel values ​​corresponding to the metal region, thus obtaining a defective wave field image with holes. In the defective wavefield image, pixel sampling is performed along the edge of the metal region defined by the metal mask to extract the set of pixels that reflect the background texture of the seam, which is then used as the background wavefield pixels. Using the background wavefield pixels as the constraint boundary, an anisotropic diffusion partial differential equation is established and solved iteratively. During the iteration process, the diffusion conduction coefficient is controlled so that its value in the direction parallel to the seam extension is greater than its value in the direction perpendicular to the seam, until the defective area is repaired, generating a first cleaned wavefield image with continuous and consistent wavefield texture.

4. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 1, characterized in that, Remove the structural components corresponding to the structure dictionary, and retain the independent scattering anomaly component images corresponding to the anomaly dictionary, specifically including: A curve wave dictionary based on the linear geometric prior of the segment joint is obtained as a structural dictionary, and a wavelet dictionary based on the isotropic characteristics of local void scattering of the medium is obtained as an anomaly dictionary. The two dictionaries are combined to form a redundant representation dictionary. Using the first purified wavefield image as input, the alternating projection optimization algorithm is used to perform iterative sparse coding and image reconstruction under the curve wave dictionary and wavelet dictionary respectively, so as to minimize the reconstruction error. Once the convergence condition is met, the image components reconstructed from the curve dictionary are identified as concrete joint structure components and discarded, while the image components reconstructed from the wavelet dictionary are retained. The output is an image of independent scattering anomaly components containing only the scattering signals of gas and water anomalies.

5. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 1, characterized in that, A two-dimensional Hilbert transform is performed on the independent scattering anomaly component images to generate complex anomaly wavefield images, and the instantaneous phase features of each pixel are extracted, specifically including: One-dimensional fast Fourier transform is performed column by column on the image of independent scattering anomaly components to convert the spatial domain image to the frequency domain and obtain the spectrum of each column of signal. In the frequency domain, the spectrum of each signal is processed by single-sideband filtering, retaining positive frequency components and suppressing negative frequency components, and the amplitude of positive frequency components is doubled for compensation. The processed single-sideband spectrum is subjected to inverse fast Fourier transform to obtain a complex matrix after frequency domain transformation; the independent scattering anomaly component image is used as the real part, and the imaginary part of the complex matrix after frequency domain transformation is used as the imaginary part, and the complex anomaly wave field image is generated by combining them. The arctangent operation is performed on the imaginary and real part values ​​of each pixel in the complex anomalous wave field image to extract the angle value between -π and π, which is used as the instantaneous phase feature of each pixel.

6. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 1, characterized in that, Identify pixel clusters exhibiting phase polarity reversal and classify them as water seepage areas. Classify the remaining pixel clusters belonging to independent scattering anomalous components as air void areas, specifically including: Traverse all connected pixel clusters in the image of the independent scattering anomaly component, and at the boundary of each pixel cluster, extract the phase change gradient of the instantaneous phase feature from the background region to the center of the anomaly region. Based on the physical laws that the reflection coefficient is positive and the phase does not reverse when concrete enters the air medium, and that the reflection coefficient is negative and the phase undergoes half-wave loss (i.e., polarity reversal) when concrete enters the water medium, a judgment logic is constructed. Clusters of pixels that show a sudden change from positive to negative phase and a reversal of polarity will be marked as water seepage areas; clusters of pixels that show a change from negative to positive phase or a continuous positive phase will be marked as air void areas.

7. The method for predicting leakage at pipe segment joints based on scattered wave field interpretation according to claim 1, characterized in that, Generate engineering diagnostic diagrams, specifically including: In the HSV color space, pseudo-color mapping is performed, and the hue channel of the pixel cluster marked as the water seepage area is set to the first preset value, and the hue channel of the pixel cluster marked as the air cavity area is set to the second preset value. The magnitude of each pixel in the complex anomalous wave field image is calculated as the instantaneous amplitude parameter characterizing the scattering intensity of the disease; after normalizing the instantaneous amplitude parameter, it is mapped to the lightness channel of the HSV color space. By extracting the number of pixels marked as water seepage areas and combining them with the physical resolution calibrated during digital acquisition, the cross-sectional area and depth of the seepage areas are calculated to generate an engineering diagnostic map containing qualitative and quantitative information.

8. A segment joint seepage prediction system based on scattered wave field interpretation, characterized in that, include: The metal mask generation module acquires the original two-dimensional wave field grayscale image collected along the joint of the pipe segment, performs hyperbolic feature detection on it, extracts the pixel coordinate set of the hyperbolic reflection area generated by the metal part, and performs morphological dilation operation to generate a metal mask matrix. The defective wave field repair module uses a metal mask matrix to shield the metal region pixels in the original two-dimensional wave field grayscale image to obtain a defective wave field image. It extracts the background wave field pixels at the edge of the metal region and performs structural diffusion filling to repair the defective region along the seam extension direction, generating the first cleaned wave field image. The morphological component separation module performs sparse decomposition and morphological component separation on the first purified wave field image under a redundant representation dictionary containing a structure dictionary and anomaly dictionary, removes the structural component corresponding to the structure dictionary, and retains the independent scattering anomaly component image corresponding to the anomaly dictionary. The instantaneous phase extraction module performs a two-dimensional Hilbert transform on the independent scattering anomaly component image to generate a complex anomaly wave field image and extracts the instantaneous phase features of each pixel. The air-water medium determination module analyzes the phase change gradient at the edge of the abnormal region in the instantaneous phase characteristics based on the mapping relationship between the difference in dielectric constant and the phase polarity of the reflected wave. It identifies the pixel clusters that have undergone phase polarity reversal and determines them as water seepage areas. The remaining pixel clusters belonging to the independent scattering abnormal components are determined as air void areas. The engineering diagnostic output module maps water seepage areas and air void areas to different display color channels, and generates an engineering diagnostic map by combining the instantaneous amplitude parameters obtained from complex anomalous wave field images.

Images