An underground pipeline intelligent identification method and device

Abstract

The application relates to an underground pipeline intelligent identification method and device, and belongs to the technical field of pipeline identification, wherein the method comprises the following steps: generating a bottom-layer electromagnetic waveform energy representation vector matrix based on an electromagnetic echo signal matrix; performing nonlinear fitting operation on the bottom-layer electromagnetic waveform energy representation vector matrix and initial medium parameters of a stratum to determine a dielectric constant gradient correction coefficient and an underground dynamic dielectric constant distribution sequence; determining echo matrix hyperbolic curve feature point coordinates based on the underground dynamic dielectric constant distribution sequence; determining a geometric shape curvature derivative of the echo matrix hyperbolic curve based on the echo matrix hyperbolic curve feature point coordinates and determining a curvature shape distortion difference value; correcting the echo matrix hyperbolic curve based on the curvature shape distortion difference value, performing space vector mapping on the corrected echo matrix hyperbolic curve based on local wave speed, and determining an underground pipeline coordinate identification result. The application effectively improves the accuracy of underground pipeline intelligent identification.

Description

Technical Field

[0001] This invention relates to the field of pipeline identification technology, and in particular to an intelligent identification method and device for underground pipelines. Background Technology

[0002] Traditional intelligent identification methods for underground pipelines refer to the spatial location investigation of water supply, gas, and communication pipelines buried in urban underground spaces. This involves using ground-penetrating radar to emit high-frequency electromagnetic waves into the ground and receiving the reflected echoes. Technicians then export a radar profile image composed of the electromagnetic wave travel time, amplitude, and phase. Discrete wavelet transform is used to denoise the original radar waveform. Next, an edge detection operator is used to extract the vertex pixel coordinates of the hyperbolic reflection waveform in the profile image. Finally, based on the empirical value of the soil dielectric constant and the bidirectional travel time of the electromagnetic waves, the burial depth and horizontal coordinates of the pipeline are calculated.

[0003] Traditional intelligent identification methods for underground pipelines rely on discrete wavelet transform and edge operator processing to extract surface waveform features from radar profiles. However, they neglect the phase and amplitude coupling evolution of deep energy when electromagnetic waves propagate in complex strata, which limits the identification of weak reflection signals. Furthermore, they heavily rely on fixed empirical values ​​of dielectric constants to estimate target burial depth, making it difficult to adapt to the gradual changes in the physical properties of multi-layered media. This results in a significant deviation between the estimated propagation wave velocity and the actual physical state, leading to a large accumulated mapping error when converting travel time to spatial depth. Consequently, this causes distortion in the three-dimensional spatial coordinate positioning of pipelines and serious misjudgments of their physical morphological properties.

[0004] Therefore, improving the accuracy of intelligent identification of underground pipelines has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of this, it is necessary to provide an intelligent identification method and device for underground pipelines to solve the problem of insufficient accuracy of existing intelligent identification methods for underground pipelines.

[0006] To address the aforementioned problems, in a first aspect, the present invention provides an intelligent identification method for underground pipelines, comprising: Obtain the electromagnetic echo signal matrix generated by ground penetrating radar, and generate the underlying electromagnetic waveform energy characterization vector matrix based on the electromagnetic echo signal matrix; The energy characterization vector matrix of the underlying electromagnetic waveform is nonlinearly fitted to the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient. Based on the dielectric constant gradient correction coefficient, the initial medium parameters of the formation are corrected to determine the dynamic dielectric constant distribution sequence of the subsurface. The coordinates of the characteristic points of the echo matrix hyperbola are determined by local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence. The geometric curvature derivative of the echo matrix hyperbola is determined based on the coordinates of the characteristic points of the echo matrix hyperbola. The curvature distortion difference is determined based on the geometric curvature derivative and the curvature of the standard medium. The hyperbola of the echo matrix is ​​corrected based on the curvature distortion difference, and the corrected hyperbola of the echo matrix is ​​spatially vector-mapped based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

[0007] In one possible implementation, generating the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix includes: Scan the electromagnetic echo signal matrix, locate the center pixel position of the hyperbola echo peak in the two-dimensional spatiotemporal domain, extract the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, perform signal time-domain discretization analysis and amplitude normalization mapping on the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, and obtain the basic characterization component data of the vertex waveform. Based on the vertex waveform basic characterization component data, the phase variation ratio of edge pixels in the eight neighborhood directions of the center pixel is analyzed. The phase variation ratio of edge pixels is distributed with the radar signal attenuation factor. The spatial phase gradient deviation value is obtained by linear aggregation of edge phase gradient and center offset. Based on the vertex waveform basic characterization component data and the spatial phase gradient deviation variation value, the energy amplitude square root constraint is applied to the echo amplitude sequence value. The constraint result is mapped with the spatial phase gradient deviation variation value through tensor dot product to generate the underlying electromagnetic waveform energy characterization vector matrix.

[0008] In one possible implementation, the step of performing a nonlinear fitting operation between the underlying electromagnetic waveform energy characterization vector matrix and the initial formation medium parameters to determine the dielectric constant gradient correction coefficient includes: Based on the vector dot product mapping between the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation, the inner product of waveform and medium energy response is obtained. The inner product of waveform and medium energy response is compared with the fitted correlation benchmark value under a preset threshold constraint. The coordinates of elements that meet the response conditions are extracted and aggregated into a two-dimensional array to obtain the initial medium waveform energy correlation matrix. Based on the initial dielectric waveform energy correlation matrix and the node weights of the deep neural network, the deviation response is mapped to obtain the node-mapped response. The node-mapped response is then processed using a nonlinear activation function to determine the dielectric constant gradient correction coefficient.

[0009] In one possible implementation, the step of correcting the initial formation medium parameters based on the dielectric constant gradient correction coefficient to determine the subsurface dynamic dielectric constant distribution sequence includes: The gradient correction coefficient of dielectric constant is fused with the initial medium parameters of the formation to determine the weighted intermediate values ​​of the layer. Differential evolution analysis is performed on the interlayer differences of the weighted intermediate values ​​of the layer to determine the differential values. The differential values ​​of the layer differential benchmark values ​​are compared with the differential benchmark values ​​of the layer. High-frequency disturbances are removed and the continuous intermediate values ​​that are smoothed and retained are combined to determine the distribution sequence of the underground dynamic dielectric constant.

[0010] In one possible implementation, the step of determining the coordinates of hyperbolic feature points of the echo matrix by local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence includes: Electromagnetic wave dynamics inversion is performed based on the underground dynamic dielectric constant distribution sequence to solve for the propagation velocity value. The spatial horizontal and vertical coordinate components of the characteristic points of the electromagnetic echo hyperbola are separated. Based on the propagation velocity value, the coordinates of the characteristic points are recombined and offset mapped to determine the coordinates of the characteristic points of the echo matrix hyperbola.

[0011] In one possible implementation, determining the geometric curvature derivative of the echo hyperbola based on the coordinates of its feature points includes: The spatial horizontal and vertical coordinate components of the echo matrix hyperbola feature point coordinates are continuously differentiated by second-order difference operation. The intermediate terms of the first difference between adjacent trajectory points are extracted. The intermediate terms of the first difference are differentiated again and spatial smoothing filtering is performed based on the curvature change benchmark value to determine the geometric curvature derivative of the echo matrix hyperbola.

[0012] In one possible implementation, determining the curvature distortion difference based on the geometric curvature derivative and the standard medium curvature includes: Deformation deviation analysis is performed on the geometric curvature derivative and standard medium curvature. The absolute amount of deformation is extracted and a distortion assessment sequence is constructed. After removing sparse outliers in the absolute amount distribution density characteristics of the distortion assessment sequence, the retained terms are fused to generate the curvature morphology distortion difference.

[0013] In one possible implementation, the correction of the hyperbola echo matrix based on the curvature distortion difference includes: Based on the curvature morphology distortion difference and time delay sensitive operator, convolution mapping operation is performed in the discrete time domain to determine the travel time disturbance deviation. The travel time disturbance deviation is verified by time domain offset to determine the time deviation adjustment value. Based on the time deviation adjustment value and the ground penetrating radar sampling frequency, the travel time delay compensation increment is generated. The original bidirectional travel time values ​​of the waveform vertices of the echo matrix hyperbola are extracted. The original bidirectional travel time values ​​of the waveform vertices are offset and recombined based on the travel time delay compensation increment to obtain the corrected travel time parameters.

[0014] In one possible implementation, the step of performing spatial vector mapping on the corrected echo matrix hyperbola based on local wave velocity to determine the underground pipeline coordinate identification result includes: Based on the corrected travel time parameters and the local real electromagnetic wave propagation speed, a one-way propagation distance spatial mapping is performed. Combined with the position coordinates of the ground penetrating radar antenna, spatial geometric projection analysis is performed. The three-dimensional spatial mapping relationship of depth and horizontal displacement is transformed to determine the underground pipeline coordinate identification result.

[0015] On the other hand, the present invention also provides an intelligent identification device for underground pipelines, comprising: The acquisition module is used to acquire the electromagnetic echo signal matrix generated by the ground penetrating radar, and generate the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix. The correction module is used to perform nonlinear fitting calculations on the energy characterization vector matrix of the underlying electromagnetic waveform and the initial medium parameters of the formation, determine the dielectric constant gradient correction coefficient, and correct the initial medium parameters of the formation based on the dielectric constant gradient correction coefficient to determine the underground dynamic dielectric constant distribution sequence. The inversion module is used to determine the coordinates of the feature points of the echo matrix hyperbola by performing local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence, determine the geometric curvature derivative of the echo matrix hyperbola based on the coordinates of the feature points of the echo matrix hyperbola, and determine the curvature distortion difference based on the geometric curvature derivative and the curvature of the standard medium. The determination module is used to correct the hyperbola of the echo matrix based on the difference in curvature morphology distortion, and to perform spatial vector mapping on the corrected hyperbola of the echo matrix based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

[0016] The beneficial effects of this invention are as follows: The intelligent identification method and device for underground pipelines provided by this invention, by extracting the hyperbolic reflection wave amplitude sequence and phase shift distribution to analyze the edge variation law, and integrating energy characterization to dynamically correct the initial stratum parameters through network mapping, effectively overcomes the limitations of fixed empirical values, enhances the adaptive calculation capability for the evolution of physical properties of multi-layer media, and improves the capture accuracy of deep weak echo signals. By dynamically distributing and inverting wave velocity and analyzing the waveform curvature morphology distortion difference, the accuracy of propagation feature restoration is significantly optimized, the cumulative mapping error of cross-layer refraction is reduced, and the high fidelity of travel time to spatial depth conversion is ensured, thereby improving the accuracy of three-dimensional spatial coordinate positioning under complex structures. This invention effectively improves the accuracy of intelligent identification of underground pipelines. Attached Figure Description

[0017] Figure 1 A schematic flowchart of an embodiment of the intelligent identification method for underground pipelines provided by the present invention; Figure 2A schematic flowchart of an embodiment of the intelligent identification process for underground pipelines provided by the present invention; Figure 3 This is a schematic diagram of an embodiment of the intelligent identification device for underground pipelines provided by the present invention. Detailed Implementation

[0018] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] This invention provides a method and device for intelligent identification of underground pipelines, which will be described below.

[0023] Figure 1 This is a schematic flowchart of an embodiment of the intelligent identification method for underground pipelines provided by the present invention, as shown below. Figure 1 As shown, the intelligent identification method for underground pipelines includes: S101. Obtain the electromagnetic echo signal matrix generated by the ground penetrating radar, and generate the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix.

[0024] It should be noted that the intelligent identification method for underground pipelines provided by this invention can be applied to the identification of urban underground pipelines.

[0025] When conducting intelligent identification of underground pipelines, the identification device (such as a desktop or portable computer) can first acquire the electromagnetic echo signal matrix generated by ground penetrating radar, and then generate a low-level electromagnetic waveform energy representation vector matrix through the electromagnetic echo signal matrix, providing a data foundation for subsequent pipeline identification.

[0026] S102. Perform nonlinear fitting calculations on the energy characterization vector matrix of the underlying electromagnetic waveform and the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient. Based on the dielectric constant gradient correction coefficient, correct the initial medium parameters of the formation to determine the underground dynamic dielectric constant distribution sequence.

[0027] It should be noted that after determining the underlying electromagnetic waveform energy characterization vector matrix, a nonlinear fitting operation can be performed on the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient. Then, the initial medium parameters of the formation are corrected by the dielectric constant gradient correction coefficient to determine the underground dynamic dielectric constant distribution sequence. This fully considers the influence of the underground dielectric constant on pipeline identification and improves the accuracy of subsequent pipeline identification.

[0028] S103. Based on the underground dynamic dielectric constant distribution sequence, local wave velocity inversion is performed to determine the coordinates of the characteristic points of the echo matrix hyperbola. Based on the coordinates of the characteristic points of the echo matrix hyperbola, the geometric curvature derivative of the echo matrix hyperbola is determined. Based on the geometric curvature derivative and the curvature of the standard medium, the curvature distortion difference is determined.

[0029] It should be noted that after determining the underground dynamic dielectric constant distribution sequence, the coordinates of characteristic points of the echo matrix hyperbola can be determined through local wave velocity inversion using this sequence. Then, based on these coordinates, the geometric curvature derivative of the echo matrix hyperbola can be further determined. Finally, the curvature distortion difference is determined using the geometric curvature derivative and the standard medium curvature, thereby optimizing the accuracy of propagation feature reconstruction and further improving pipeline identification accuracy.

[0030] S104. The hyperbola of the echo matrix is ​​corrected based on the difference in curvature morphology distortion, and the corrected hyperbola of the echo matrix is ​​spatially vector-mapped based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

[0031] It should be noted that: Finally, the hyperbola of the echo matrix can be corrected by the difference in curvature morphology distortion, and then the corrected hyperbola of the echo matrix can be spatially vector-mapped by the local wave velocity to determine the coordinate identification result of the underground pipeline. This ensures the accuracy of three-dimensional spatial coordinate positioning under complex structures and achieves accurate underground pipeline identification.

[0032] In summary, the intelligent underground pipeline identification method provided by this invention extracts the hyperbolic reflection wave amplitude sequence and phase shift distribution to analyze the edge variation law, integrates energy characterization to perform network mapping and dynamically correct the initial stratum parameters, effectively overcomes the limitations of fixed empirical values, enhances the adaptive calculation capability for the evolution of physical properties of multi-layer media, improves the capture accuracy of deep weak echo signals, significantly optimizes the accuracy of propagation feature restoration by dynamically distributing and inverting wave velocity and analyzing waveform curvature and distortion differences, reduces cross-layer refraction cumulative mapping errors, and ensures high fidelity in travel time to spatial depth conversion, thereby improving the accuracy of three-dimensional spatial coordinate positioning under complex structures. This invention effectively improves the accuracy of intelligent underground pipeline identification.

[0033] In some embodiments of the present invention, the step of generating a low-level electromagnetic waveform energy representation vector matrix based on an electromagnetic echo signal matrix includes: Scan the electromagnetic echo signal matrix, locate the center pixel position of the hyperbola echo peak in the two-dimensional spatiotemporal domain, extract the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, perform signal time-domain discretization analysis and amplitude normalization mapping on the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, and obtain the basic characterization component data of the vertex waveform. Based on the vertex waveform basic characterization component data, the phase variation ratio of edge pixels in the eight neighborhood directions of the center pixel is analyzed. The phase variation ratio of edge pixels is distributed with the radar signal attenuation factor. The spatial phase gradient deviation value is obtained by linear aggregation of edge phase gradient and center offset. Based on the vertex waveform basic characterization component data and the spatial phase gradient deviation variation value, the energy amplitude square root constraint is applied to the echo amplitude sequence value. The constraint result is mapped with the spatial phase gradient deviation variation value through tensor dot product to generate the underlying electromagnetic waveform energy characterization vector matrix.

[0034] It should be noted that when generating the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix, the electromagnetic echo signal matrix is ​​first scanned to locate the center pixel position of the hyperbolic echo peak in the two-dimensional spatiotemporal domain. The echo amplitude sequence value and initial phase offset value corresponding to the center pixel position are extracted. The echo amplitude sequence value and initial phase offset value corresponding to the center pixel position are then subjected to signal time-domain discretization analysis and amplitude normalization mapping to obtain the vertex waveform basic representation component data. Then, using the vertex waveform basic representation component data, the phase variation ratio of the edge pixels in the eight-neighbor direction of the center pixel is analyzed. The phase variation ratio of the edge pixels is distributed by the product term of the radar signal attenuation factor. Through the linear aggregation of the edge phase gradient and the center offset, the spatial phase gradient deviation value is obtained. Finally, using the vertex waveform basic representation component data and the spatial phase gradient deviation value, the energy amplitude square root constraint is applied to the echo amplitude sequence value. The constraint result is then mapped to the spatial phase gradient deviation value by tensor dot product mapping to generate the underlying electromagnetic waveform energy representation vector matrix.

[0035] In some embodiments of the present invention, the step of performing nonlinear fitting operations on the bottom electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient includes: Based on the vector dot product mapping between the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation, the inner product of waveform and medium energy response is obtained. The inner product of waveform and medium energy response is compared with the fitted correlation benchmark value under a preset threshold constraint. The coordinates of elements that meet the response conditions are extracted and aggregated into a two-dimensional array to obtain the initial medium waveform energy correlation matrix. Based on the initial dielectric waveform energy correlation matrix and the node weights of the deep neural network, the deviation response is mapped to obtain the node-mapped response. The node-mapped response is then processed using a nonlinear activation function to determine the dielectric constant gradient correction coefficient.

[0036] It should be noted that when determining the dielectric constant gradient correction coefficient, we can first perform a vector dot product mapping between the underlying electromagnetic waveform energy characterization vector matrix and the initial formation medium parameters to obtain the inner product of the waveform and medium energy response. This inner product is then compared with the fitted correlation benchmark value under a preset threshold constraint. The coordinates of elements meeting the response conditions are extracted and aggregated into a two-dimensional array to obtain the initial medium waveform energy correlation matrix. Next, we perform a deviation response mapping using the initial medium waveform energy correlation matrix and the node weights of a deep neural network to obtain the node-mapped response. Finally, we process the node-mapped response using a nonlinear activation function to determine the dielectric constant gradient correction coefficient.

[0037] In some embodiments of the present invention, the step of correcting the initial medium parameters of the formation based on the dielectric constant gradient correction coefficient to determine the dynamic dielectric constant distribution sequence of the subsurface includes: The gradient correction coefficient of dielectric constant is fused with the initial medium parameters of the formation to determine the weighted intermediate values ​​of the layer. Differential evolution analysis is performed on the interlayer differences of the weighted intermediate values ​​of the layer to determine the differential values. The differential values ​​of the layer differential benchmark values ​​are compared with the differential benchmark values ​​of the layer. High-frequency disturbances are removed and the continuous intermediate values ​​that are smoothed and retained are combined to determine the distribution sequence of the underground dynamic dielectric constant.

[0038] It should be noted that when determining the distribution sequence of underground dynamic dielectric constant, the dielectric constant gradient correction coefficient and the initial medium parameters of the formation can be weighted and fused to determine the weighted intermediate values ​​of the layers. The differences between the layers of the weighted intermediate values ​​of the layers can be determined by differential evolution analysis to determine the difference values. The difference values ​​of the layers are compared with the differential benchmark values ​​of the layers, high-frequency disturbances are eliminated and the continuous intermediate values ​​that are smoothed and retained are combined to determine the distribution sequence of underground dynamic dielectric constant.

[0039] In some embodiments of the present invention, the step of determining the coordinates of hyperbolic feature points of the echo matrix by local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence includes: Electromagnetic wave dynamics inversion is performed based on the underground dynamic dielectric constant distribution sequence to solve for the propagation velocity value. The spatial horizontal and vertical coordinate components of the characteristic points of the electromagnetic echo hyperbola are separated. Based on the propagation velocity value, the coordinates of the characteristic points are recombined and offset mapped to determine the coordinates of the characteristic points of the echo matrix hyperbola.

[0040] It should be noted that when determining the coordinates of the feature points of the hyperbola echo matrix by performing local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence, electromagnetic wave dynamics inversion can be performed through the underground dynamic dielectric constant distribution sequence to solve for the propagation velocity values, separate the spatial horizontal and vertical coordinate components of the feature points of the electromagnetic echo hyperbola, and then reconstruct and offset the feature point coordinates through the propagation velocity values ​​to determine the coordinates of the feature points of the hyperbola echo matrix.

[0041] In some embodiments of the present invention, determining the geometric curvature derivative of the echo matrix hyperbola based on the coordinates of its feature points includes: The spatial horizontal and vertical coordinate components of the echo matrix hyperbola feature point coordinates are continuously differentiated by second-order difference operation. The intermediate terms of the first difference between adjacent trajectory points are extracted. The intermediate terms of the first difference are differentiated again and spatial smoothing filtering is performed based on the curvature change benchmark value to determine the geometric curvature derivative of the echo matrix hyperbola.

[0042] It should be noted that when determining the geometric curvature derivative of the echo matrix hyperbola, continuous second-order difference differentiation can be performed on the spatial horizontal and vertical coordinate components of the feature point coordinates of the echo matrix hyperbola. The intermediate terms of the first difference between adjacent trajectory points can be extracted, and the intermediate terms of the first difference can be differenced again and spatial smoothing filtering can be performed based on the curvature abrupt change reference value to determine the geometric curvature derivative of the echo matrix hyperbola.

[0043] In some embodiments of the present invention, determining the curvature distortion difference based on the geometric curvature derivative and the standard medium curvature includes: Deformation deviation analysis is performed on the geometric curvature derivative and standard medium curvature. The absolute amount of deformation is extracted and a distortion assessment sequence is constructed. After removing sparse outliers in the absolute amount distribution density characteristics of the distortion assessment sequence, the retained terms are fused to generate the curvature morphology distortion difference.

[0044] It should be noted that when determining the curvature distortion difference based on the geometric curvature derivative and the standard medium curvature, deformation deviation analysis can be performed on the geometric curvature derivative and the standard medium curvature to extract the absolute amount of deformation and construct a distortion assessment sequence. After removing sparse outliers in the absolute amount distribution density characteristics of the distortion assessment sequence, the retained terms are merged to generate the curvature distortion difference.

[0045] In some embodiments of the present invention, the correction of the hyperbola echo matrix based on the curvature distortion difference includes: Based on the curvature morphology distortion difference and time delay sensitive operator, convolution mapping operation is performed in the discrete time domain to determine the travel time disturbance deviation. The travel time disturbance deviation is verified by time domain offset to determine the time deviation adjustment value. Based on the time deviation adjustment value and the ground penetrating radar sampling frequency, the travel time delay compensation increment is generated. The original bidirectional travel time values ​​of the waveform vertices of the echo matrix hyperbola are extracted. The original bidirectional travel time values ​​of the waveform vertices are offset and recombined based on the travel time delay compensation increment to obtain the corrected travel time parameters.

[0046] It should be noted that when correcting the hyperbola echo matrix, convolution mapping operations can be performed in the discrete time domain based on the curvature distortion difference and the time delay sensitive operator to determine the travel time disturbance deviation. Then, a time domain offset verification is performed on the travel time disturbance deviation to determine the time deviation adjustment value. Next, based on the time deviation adjustment value and the ground-penetrating radar sampling frequency, a travel time delay compensation increment is generated. Then, the original bidirectional travel time values ​​at the waveform vertices of the echo matrix hyperbola can be extracted. The original bidirectional travel time values ​​at the waveform vertices are then offset and reconstructed using the travel time delay compensation increment to obtain the corrected travel time parameters.

[0047] In some embodiments of the present invention, the step of performing spatial vector mapping on the corrected echo matrix hyperbola based on local wave velocity to determine the underground pipeline coordinate identification result includes: Based on the corrected travel time parameters and the local real electromagnetic wave propagation speed, a one-way propagation distance spatial mapping is performed. Combined with the position coordinates of the ground penetrating radar antenna, spatial geometric projection analysis is performed. The three-dimensional spatial mapping relationship of depth and horizontal displacement is transformed to determine the underground pipeline coordinate identification result.

[0048] It should be noted that when determining the coordinate identification results of underground pipelines, the spatial mapping of unidirectional propagation distance can be performed by correcting the travel time parameters and the local real electromagnetic wave propagation speed. Combined with the position coordinates of the ground penetrating radar antenna, spatial geometric projection analysis can be performed to transform the three-dimensional spatial mapping relationship of depth and horizontal displacement, thereby determining the coordinate identification results of underground pipelines.

[0049] The following describes the intelligent identification process for underground pipelines provided by the present invention through specific embodiments, combined with Figure 2 Here are the specific steps involved in the intelligent identification process for underground pipelines: 1. Using a two-dimensional sliding window step scan, the electromagnetic echo signal matrix generated by the ground penetrating radar is retrieved, the amplitude sequence of the hyperbolic reflected wave vertex and the local phase shift distribution are extracted, the phase variation law of the center and edge pixels is analyzed, and the amplitude sequence and phase variation components are fused to generate the underlying electromagnetic waveform energy characterization vector matrix.

[0050] 2. Call the underlying electromagnetic waveform energy characterization vector matrix and perform nonlinear fitting operation with the initial medium parameters of the formation. Introduce nonlinear mapping processing of deep neural network nodes to obtain the dielectric constant gradient correction coefficient. Use the dielectric constant gradient correction coefficient to perform dynamic weighted correction on the initial medium parameters of the formation to obtain the numerical value of the underground dynamic dielectric constant distribution sequence.

[0051] 3. Based on the numerical inversion of the local wave velocity from the underground dynamic dielectric constant distribution sequence, the coordinates of the feature points of the hyperbola echo matrix are extracted and the derivative of the geometric curvature is obtained. Combined with the standard medium curvature, intelligent distortion analysis is performed to generate the numerical value of the curvature distortion difference.

[0052] 4. Call the curvature distortion difference value, convolve the curvature distortion difference value with the delay-sensitive operator to compensate, generate the travel time increment and correct the original travel time of the hyperbola vertex, combine the local medium wave velocity to perform spatial vector mapping, perform pipeline accuracy spatial coordinate identification, and generate a pipeline spatial coordinate intelligent identification list.

[0053] 5. Based on the intelligent identification list of pipeline spatial coordinates, locate the radial reflection zone in the echo matrix, extract the time difference of positive and negative polarity wave peaks and troughs and determine the wall reflection travel time. After wave velocity coupling and frequency scaling, invert the radial physical dimensions of the pipeline. Through reconstruction residual matching of the standard pipe diameter library, output the pipeline entity physical attribute label dataset.

[0054] The underlying electromagnetic waveform energy characterization vector matrix includes peak energy concentration vector, waveform polarization feature matrix elements, and gridded electromagnetic scattering intensity distribution. The underground dynamic dielectric constant distribution sequence includes discrete stratum polarizability readings and gridded microwave attenuation coefficients. The curvature morphology distortion difference values ​​include reflection arc wing symmetry deviation, vertex widening distortion index, and hyperbola asymptote tilt error. The pipeline spatial coordinate intelligent identification list includes a three-dimensional orientation node positioning matrix and anomaly inflection point spatial coordinate cluster. The pipeline entity physical attribute label dataset includes the target pipe segment outer diameter estimation value, internal cavity effective diameter, and pipe wall structure thickness classification label.

[0055] The specific steps for generating the underlying electromagnetic waveform energy representation vector matrix are as follows: 1. Scan the electromagnetic echo signal matrix, locate the center pixel position of the hyperbola echo peak in the two-dimensional spatiotemporal domain, extract the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, and obtain the basic characterization component data of the vertex waveform through signal time-domain discretization analysis and amplitude normalization mapping.

[0056] A line-by-line scanning operation is performed on the two-dimensional matrix of electromagnetic echo signals generated by radar detection equipment after analog-to-digital conversion. The amplitude values ​​of pixels within the matrix are extracted, and extreme points with amplitude values ​​greater than the eight surrounding pixels are identified as the center pixel location of the hyperbolic echo peak. For this center pixel location, discrete amplitude values ​​within the sampling period are extracted along the time axis and arranged to form an echo amplitude sequence. The initial phase shift value corresponding to this point is extracted through complex signal analysis. The echo amplitude sequence values ​​are then divided by an amplitude threshold to perform amplitude normalization mapping. Setting the amplitude of a certain point in the echo amplitude sequence to 30000, dividing it by the amplitude threshold of 60000 yields an amplitude normalization value of 0.5. The amplitude normalization values ​​of this point and neighboring time nodes are collected and combined with the initial phase shift value to construct the basic characteristic component data of the peak waveform. The advantage of this operational logic is that the amplitude normalization mapping operation compresses the amplitude distribution to the 0-1 range, eliminating the influence of absolute energy value differences caused by medium reflection, and allowing the normalized signal sequence to be fed into the subsequent feature analysis process.

[0057] 2. Call the vertex waveform basic representation component data, analyze the phase change ratio of edge pixels in the eight neighborhood directions of the center pixel, distribute it with the radar signal attenuation factor, and obtain the spatial phase gradient deviation value through linear aggregation of edge phase gradient and center offset.

[0058] The system retrieves the basic representation component data of the vertex waveform and extracts the phase values ​​of eight adjacent pixels in the surrounding eight directions in a two-dimensional coordinate system, using the center pixel as a reference. The difference between the phase values ​​of the adjacent pixels and the center pixel is calculated, and this difference is divided by the center pixel's phase value to obtain the phase variation ratio of the edge pixels in the eight neighboring directions. The radar signal attenuation factor is retrieved, and the edge pixel phase variation ratio is multiplied by the attenuation factor to complete the product term allocation operation. For the product result, the spatial distance deviation corresponding to the edge pixels is extracted as the center offset. The product result is multiplied by the edge phase gradient and summed with the center offset, then linearly aggregated to obtain the spatial phase gradient deviation value.

[0059] 3. Call the vertex waveform basic characterization component data and the spatial phase gradient deviation variation value, apply the energy amplitude square root constraint to the echo amplitude sequence value, perform tensor dot product mapping on the constraint result and the spatial phase gradient deviation variation value, implement the dimensional aggregation of phase shift and amplitude enhancement components, and generate the underlying electromagnetic waveform energy characterization vector matrix after spatial feature coordinate system mapping correction.

[0060] The algorithm calls upon the vertex waveform fundamental representation component data and the spatial phase gradient deviation variation values. For the echo amplitude sequence values ​​in the vertex waveform fundamental representation component data, it performs a square root operation on the elements, applying a square root constraint on the energy amplitude. The amplitude components obtained after the square root constraint and the spatial phase gradient deviation variation values ​​are expanded into spatial tensors of the same dimension. Element-by-element multiplication is performed on the corresponding coordinate elements to perform tensor dot product mapping. The dot product result is summed and superimposed with the initial phase offset value to complete the aggregation of the phase offset and amplitude enhancement component dimensions. A preset spatial feature coordinate system transformation matrix is ​​obtained. The aggregated vector is multiplied by the transformation matrix to perform mapping correction, outputting the underlying electromagnetic waveform energy representation vector matrix.

[0061] The specific steps for determining the numerical values ​​of the underground dynamic dielectric constant distribution sequence are as follows: 1. Based on the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation, perform vector dot product mapping to obtain the inner product of waveform and medium energy response. Compare the inner product with the fitted correlation benchmark value under a preset threshold constraint, extract and aggregate the coordinates of elements that meet the response conditions into a two-dimensional array, and obtain the initial medium waveform energy correlation matrix.

[0062] The process of obtaining the inner product of waveform and medium energy response refers to expanding the underlying electromagnetic waveform energy characterization vector matrix into a one-dimensional pre-vector, expanding the initial medium parameters of the formation into a one-dimensional post-vector, performing numerical multiplication of the corresponding elements of the one-dimensional pre-vector and the one-dimensional post-vector, and summing the product terms output by the multiplication operation to obtain the inner product of waveform and medium energy response.

[0063] The initial medium parameters of the formation are obtained. The electromagnetic waveform energy characterization vector matrix of the bottom layer is expanded and concatenated into a pre-vector. The initial medium parameters of the formation are expanded into a post-vector along the same dimensions. Multiplication is performed on the values ​​at the same index positions in the pre-vector and post-vector, and the product terms are summed to derive the inner product of the waveform and medium energy response. A preset fitting correlation benchmark value is used. The inner product is compared with the fitting correlation benchmark value. If the inner product is greater than or equal to the benchmark value, the horizontal and vertical coordinate elements of the corresponding spatial grid are extracted and written into a two-dimensional spatial array to form the initial medium waveform energy correlation degree matrix. The pre-vector element is set to 0.8, the post-vector element to 2.0, and the product is 1.6. Assuming the summation of the entire array is 12.0, and the preset fitting correlation benchmark value is 10.0. Comparing 12.0 with 10.0, since 12.0 is greater than 10.0, the response condition is met. The horizontal value of 100 and the vertical value of 50 are stored in the initial medium waveform energy correlation matrix. The advantage of this operation logic is that it filters background noise and weakly correlated signals through threshold constraints and inner product mapping, and the generated correlation matrix serves as the distribution feature of the candidate medium.

[0064] 2. Call the initial dielectric waveform energy correlation matrix and deep neural network node weights, perform deviation response mapping to obtain node mapping response quantities, introduce a nonlinear activation function for the node mapping response quantities, remove low confidence response nodes, reconstruct and normalize high confidence mapping parameters, and construct dielectric constant gradient correction coefficients.

[0065] A deep neural network framework is obtained, and the initial dielectric waveform energy correlation matrix is ​​called to extract the feature distribution as input data for the input layer. The weights of the trained deep neural network nodes in the hidden layers are obtained, and the input features are multiplied by the node weights to obtain the node mapping response. A linear rectified activation function is introduced into each hidden layer to process the node mapping response, filtering out response values ​​less than 0 and removing nodes with low confidence responses. For the remaining mapping parameters greater than 0, the sum of all parameters is calculated, and each parameter is divided by the sum for normalization. The normalized data sequence is then used. With an input feature value of 4.0 and a deep neural network node weight of 1.5, the multiplication yields a node mapping response of 6.0, which is maintained at 6.0 after activation. The remaining response in the same layer is 4.0. The sum of 6.0 and 4.0 is 10.0, and dividing 6.0 by 10.0 yields a normalized value of 0.6, which is used as the dielectric constant gradient correction coefficient at the corresponding position. The advantage of this operational logic is that it uses activation functions to block the transmission of negative features and strengthens positive correlation physical features, while a correction coefficient of 0.6 is used to fine-tune the basic parameters.

[0066] 3. Call the dielectric constant gradient correction coefficient and the initial medium parameters of the formation to perform weighted parameter fusion, solve the weighted intermediate quantity of the layer, perform differential evolution analysis on the interlayer difference of the weighted intermediate quantity to obtain the difference value, compare with the layer difference benchmark value to perform high frequency disturbance removal, combine the smooth and retained continuous intermediate quantity, and obtain the underground dynamic dielectric constant distribution sequence value.

[0067] The dielectric constant gradient correction coefficient is called and multiplied with the initial medium parameter of the corresponding formation. The product is then added to the initial medium parameter to complete the weighted parameter fusion and obtain the weighted intermediate value of the layer. The weighted intermediate values ​​of adjacent layers in the depth direction are subtracted to obtain the difference value, which reflects the rate of change of the interlayer medium. A preset layer difference benchmark value is obtained, and it is determined whether the difference value is greater than the benchmark value. If it is greater, it is modified to the average of adjacent normal difference values, and high-frequency disturbance removal is performed. The continuous intermediate values ​​after anomaly removal are smoothly connected in depth order to obtain the subsurface dynamic dielectric constant distribution sequence value. The initial medium parameter of the formation is set to 5.0, and the dielectric constant gradient correction coefficient is set to 0.6. Multiplying them gives 3.0, and adding 5.0 gives a weighted intermediate value of 8.0. Assuming the weighted average of adjacent depth segments is 7.5, the difference is 0.5. The segment difference baseline is 0.8. Comparing 0.5 and 0.8, since 0.5 is less than 0.8, it is considered a reasonable fluctuation and retained. The advantage of this operational logic is that it weakens the abrupt interference caused by shallow multiple reflections through difference comparison and threshold filtering. The smoothly retained continuous sequence values ​​are used for inverse inference of the formation medium environment.

[0068] The specific steps for determining the value of curvature distortion difference are as follows: 1. Based on the numerical distribution sequence of underground dynamic dielectric constant, perform electromagnetic wave dynamics inversion, solve for the propagation velocity value, separate the spatial horizontal and vertical coordinate components of the feature points of the electromagnetic echo hyperbola, perform feature point coordinate reconstruction and offset mapping based on the propagation velocity value, filter the drift error, and construct the feature point coordinate distribution sequence.

[0069] The propagation velocity constant of electromagnetic waves in a vacuum is obtained. The numerical values ​​of the underground dynamic dielectric constant distribution sequence are retrieved, and the square root of this dielectric constant sequence value is calculated. The propagation velocity constant of electromagnetic waves in a vacuum is divided by the square root result, and electromagnetic wave dynamics inversion calculation is performed to derive the electromagnetic wave propagation velocity value at the corresponding depth. Hyperbolic feature points are extracted from the energy characterization vector matrix of the underlying electromagnetic waveform, and their spatial horizontal and vertical coordinate values ​​are decomposed. The vertical coordinate value is multiplied by the propagation velocity value to calculate the physical depth offset, completing the feature point coordinate reconstruction and offset mapping. The median filtering algorithm is used to calculate the mean coordinates within a local window, eliminating drift errors that deviate from the mean by more than a specified pixel distance, thus constructing the feature point coordinate distribution sequence. The dielectric constant distribution sequence value is set to 9.0, and its square root is calculated to be 3.0. Dividing 300,000,000 meters per second by 3.0 yields a propagation speed of 100,000,000 meters per second. The longitudinal travel time of the feature point is 0.00000004 seconds. Multiplying 100,000,000 by 0.00000004 gives a physical depth offset of 4.0 meters. The advantage of this computational logic is that it introduces the dielectric constant to invert the real-time wave velocity, correcting the deep-level positioning distortion caused by the constant wave velocity assumption.

[0070] 2. Call the spatial horizontal and vertical coordinate components in the feature point coordinate distribution sequence, perform continuous second-order difference differentiation operation, extract the first difference micro-variation law of adjacent trajectory points to form the first difference intermediate term, difference the intermediate term again and introduce the curvature mutation benchmark value optimized by deep learning self-adaptation to perform spatial smoothing filtering, and obtain the geometric curvature derivative.

[0071] The feature point coordinate distribution sequence is invoked, and the longitudinal physical depth difference between two adjacent points is extracted along the horizontal coordinate direction. This difference is divided by the horizontal distance step size to calculate the slope change rate, which is used as the first-order difference intermediate term. The intermediate terms of adjacent first-order differences are subtracted and divided by the horizontal distance step size to obtain an approximate second-order derivative. A pre-existing curvature abrupt change baseline value is obtained. If the approximate second-order derivative value is greater than this baseline value, a smoothing fit is performed on the point to obtain the geometric curvature derivative. The depths of adjacent points are set to 4.0 meters, 4.2 meters, and 4.1 meters, with a distance step size of 0.1 meters. The first segment depth difference of 0.2 meters divided by 0.1 meters yields a first-order difference intermediate term of 2.0. The second segment depth difference of -0.1 meters divided by 0.1 meters yields an intermediate term of -1.0. Subtracting 2.0 from -1.0 yields -3.0, which is divided by 0.1 to obtain an approximate second-order derivative value of -30.0. The baseline value for the mutation is 20.0. If the absolute value of -30.0 is greater than 20.0, a curvature mutation is determined to exist. Smoothing is then performed to adjust it to 15.0. The advantage of this operation logic is that it combines mathematical difference with adaptive threshold to identify curvature anomalies.

[0072] 3. Call the geometric curvature derivative and the standard medium curvature constant to perform deformation deviation analysis, extract the absolute amount of deformation and arrange them to form a distortion assessment sequence, analyze the absolute amount distribution density characteristics of the assessment sequence, remove sparse outliers and merge the retained terms to generate the curvature morphology distortion difference value.

[0073] The standard medium curvature constant is obtained from a pre-defined data set. The geometric curvature derivative is then used, and the deformation difference is calculated by subtracting the standard medium curvature constant from the geometric curvature derivative. The absolute value of this difference is extracted as the absolute deformation quantity. The absolute deformation quantities at continuous spatial points are arranged in lateral coordinate order to form a distortion assessment sequence. The number of elements with absolute deformation quantities greater than zero within a predetermined length window is counted, and their ratio to the total number of elements is calculated as the distribution density feature. If the distribution density feature is lower than the ratio constant, the absolute deformation quantity in that region is identified as a sparse outlier and removed. The remaining absolute deformation quantity values ​​are then averaged and merged. The standard medium curvature constant is set to 10.0. The calculated geometric curvature derivative of 15.0 is subtracted from 10.0, resulting in a deformation difference of 5.0 and an absolute deformation quantity of 5.0. Within an assessment window of length 20, 14 points meet the condition, and the calculated distribution density feature is 0.7. The preset ratio constant is 0.2. Since 0.7 is greater than 0.2, sequence elements including 5.0 are retained. Assuming the sum of the retained sequence elements is 50.0, there are 10 valid elements. Dividing 50.0 by 10 yields a curvature distortion difference value of 5.0. The advantage of this operational logic is that it introduces distribution density features to filter out isolated calculation errors.

[0074] The specific steps for determining the pipeline spatial coordinate intelligent identification list are as follows: 1. Call the curvature distortion difference value and the time delay sensitive operator to perform convolution mapping operation in the discrete time domain, capture the travel time disturbance deviation, use the differential fluctuation law to verify the time domain offset, determine the time deviation adjustment value, combine the radar sampling frequency to reconstruct the discrete deviation domain, and generate the travel time delay compensation increment.

[0075] A one-dimensional time delay sensitive operator sequence is obtained. The curvature distortion difference value is used as a weighting factor and substituted into the discrete time domain. Point-by-point multiplication and accumulation operations are performed with the radar travel time recording matrix to extract the travel time disturbance deviation. The fluctuation slope of the travel time disturbance deviation over time is calculated. The differential fluctuation law is compared with the preset synchronization check clock to verify the time domain offset center, and the time deviation adjustment value is calculated and established. The radar equipment sampling frequency parameter is extracted, and the unit sampling time interval is calculated using the sampling frequency. The time deviation adjustment value is divided by the unit sampling time interval. The travel time disturbance deviation at a certain sampling point is set to 0.00000002 seconds, and the verified time deviation adjustment value is determined to be 0.00000001 seconds. The radar sampling frequency is set to 2000000000 Hz, and the unit sampling time interval is calculated to be 0.0000000005 seconds. Dividing 0.00000001 seconds by 0.0000000005 seconds yields a travel time delay compensation increment of 20 sampling points. The advantage of this operational logic lies in converting continuous time deviations into adjustable discrete sampling point increments through sampling frequency discretization and reconstruction.

[0076] 2. Extract the original bidirectional timekeeping values ​​at the vertices of the hyperbola waveform, call the timekeeping delay compensation increment to perform offset reconstruction correction on the original bidirectional timekeeping values, perform phase alignment operation based on the period of the detection signal at the reflection interface and the tolerance reference to eliminate the nonlinear clock wander effect, and reconstruct the response time sequence of the reflection point through time delay accumulation to obtain the corrected timekeeping parameters.

[0077] The system reads the hyperbolic waveform vertex data recorded by the radar equipment, extracts the original bidirectional timekeeping value, calls the time delay compensation increment, multiplies it by the unit sampling time interval to restore the time increment, and performs addition and subtraction correction operations on the time increment and the original bidirectional timekeeping value to complete the offset reconstruction. It obtains the radar transmit signal period width, divides the corrected timekeeping value by the signal period and takes the remainder, and determines whether the remainder is within the preset tolerance reference range. If it deviates, it fine-tunes the timekeeping to align to the nearest integer multiple of the period node, performs phase alignment operations to eliminate nonlinear clock runaway effects, and accumulates and combines the corrected time nodes to reconstruct the reflection point response time sequence. The original bidirectional timekeeping value is set to 0.0000002 seconds, and after subtracting the time increment of 0.00000001 seconds, the reconstructed timekeeping is 0.00000019 seconds. The transmission signal period is 0.00000002 seconds, and the tolerance reference is 0.000000002 seconds. Dividing 0.00000019 by 0.00000002 leaves a remainder of 0.00000001 seconds. Since 0.00000001 exceeds the tolerance reference range, it is fine-tuned towards the nearest integer period node of 0.0000002 seconds. This value corrects the timing parameters. The advantage of this operational logic is that it uses period phase alignment to correct clock drift.

[0078] 3. Call the corrected travel time parameters and the local real electromagnetic wave propagation speed values ​​to perform unidirectional propagation distance spatial mapping, combine the ground penetrating radar antenna position coordinates to perform spatial geometric projection analysis, transform the three-dimensional spatial mapping relationship of depth and horizontal displacement, and generate a pipeline spatial coordinate intelligent identification list.

[0079] The corrected travel time parameter is called, and the value is divided by 2 to obtain the one-way transmission time. The local real electromagnetic wave propagation speed value is then called, and the one-way transmission time is multiplied by the propagation speed to derive the one-way propagation distance of the detected target. The ground-penetrating radar antenna position coordinates and surface elevation data are obtained. Using this coordinate system as the reference origin, the calculated one-way propagation distance is used to analyze the detection angle through sine and cosine geometric trigonometric functions, decomposing it into vertically downward depth coordinates and horizontal displacement coordinates along the ground surface. The depth and horizontal position information are integrated and output to a three-dimensional spatial mapping table, exporting a structured storage pipeline spatial coordinate intelligent recognition list. The corrected travel time parameter is set to 0.0000002 seconds, divided by 2 to obtain a one-way transmission time of 0.0000001 seconds. Multiplying 0.0000001 seconds by the propagation speed value of 100,000,000 meters per second, the one-way propagation distance is calculated to be 10.0 meters. Given that the antenna's elevation on the ground is 100.0 meters, and the detection angle is offset vertically downwards by 10 degrees, the calculated cosine component yields a vertical depth offset of 9.8 meters. Therefore, the target's absolute depth coordinates are 90.2 meters, and this record is added to the identification list. The advantage of this computational logic lies in its comprehensive correction of travel time and dynamic velocity to perform geometric spatial projection analysis.

[0080] The specific steps for determining the pipeline entity physical attribute label dataset are as follows: 1. Based on the intelligent identification list of pipeline spatial coordinates, the radial reflection area in the electromagnetic echo signal matrix is ​​delineated, the peak time of the positive polarity main reflection wave and the trough time of the negative polarity wave are extracted, the time delay interval between the polarity reversal points is analyzed, and the round-trip time span of the reflection interface is statistically calculated by combining the discrete sampling clock frequency to generate the wall reflection travel time difference.

[0081] The system extracts the horizontal and depth target ranges from the intelligent identification list of pipeline spatial coordinates, and extracts a radial reflection area image block covering the physical boundary of the pipe wall from the electromagnetic echo signal matrix. It then traverses the signal time axis of this area along the depth direction, locating and extracting the time point of the first positive oscillation of the main reflection peak as the pipe top reflection time, and scanning downwards to locate the time point of the phase reversal negative polarity trough as the pipe bottom reflection time. The time of the main reflection peak is subtracted from the negative polarity trough time to analytically calculate the time delay interval between the two polarity reversal points. The reciprocal of the obtained discrete sampling clock frequency is multiplied by the number of sampling points to verify the comparison time delay. The positive polarity main reflection peak time is set to 0.0000001 seconds, and the negative polarity trough time obtained by scanning downwards is 0.000000104 seconds. Subtracting these values ​​yields a time delay interval of 0.000000004 seconds between the two polarity reversal points. The discrete sampling clock frequency is 2000000000 Hz, with a reciprocal of 0.0000000005 seconds. A total of 8 sampling points are used at intervals, and the product is 0.000000004 seconds. Verification confirms this as the travel time difference caused by wall reflection. The advantage of this operational logic lies in utilizing the phase reversal law of electromagnetic waves at interfaces with different dielectric constants to extract the travel time segment.

[0082] 2. The propagation distance is analyzed by calling the wall reflection travel time difference and the local real electromagnetic wave propagation speed values. Wave velocity dispersion correction and compensation are performed by combining the radar operating frequency and medium wavelength attenuation parameters. The interlayer thickness distribution characteristics of the upper and lower reflection interfaces are calculated, the internal cross-sectional shape of the pipe wall is analyzed, and the radial characteristic dimensions of the pipeline are obtained.

[0083] The wall reflection travel time difference and the local true electromagnetic wave propagation speed are used. The wall reflection travel time difference is divided by 2 to obtain the one-way internal travel time. Then, the one-way internal travel time is multiplied by the propagation speed to complete the basic propagation distance analysis. The radar operating frequency is read, and the corresponding medium wavelength attenuation parameter is retrieved according to the filling medium inside the pipeline. The calculated basic propagation distance is multiplied by the reciprocal of the attenuation parameter to perform wave velocity dispersion correction compensation, eliminating the deceleration tail error when high-frequency signals are transmitted inside the dense pipe wall, and outputting the calculated interlayer thickness distribution characteristics of the upper and lower reflection interfaces. The wall reflection travel time difference is set to 0.000000004 seconds, and divided by 2 to obtain the one-way internal travel time of 0.000000002 seconds. 0.000000002 seconds is multiplied by the local true electromagnetic wave propagation speed of 100000000 meters per second to calculate the basic propagation distance of 0.2 meters. The wavelength attenuation parameter of polyethylene material at the radar operating frequency is obtained as 1.05. Dividing 0.2 meters by 1.05 yields a calculated interlayer thickness of 0.19 meters after dispersion correction, which is considered the radial characteristic dimension of the pipeline. The advantage of this calculation logic is that it considers the high-frequency attenuation and wave velocity hysteresis occurring inside the thinner pipe wall, resulting in a more objective reconstruction of the radial thickness.

[0084] 3. Call the radial feature dimension of the pipeline, perform error matching in the standard pipe diameter library dataset, lock the target nominal parameter item with the smallest difference, determine the pipeline material type, outer diameter and wall thickness parameter set, combine equipment management code and spatial geographic location analysis to analyze multi-dimensional IoT association relationship, and summarize to generate pipeline entity physical attribute label dataset.

[0085] Obtain the pre-entered standard pipe diameter database dataset, which stores the factory-specified inner and outer diameters and wall thickness combinations for various types of pipes. Retrieve the parsed radial feature dimensions of the pipes, traverse the standard pipe diameter database dataset, and calculate the absolute difference between the feature dimensions and the nominal wall thickness items of each pipe in the database. By sorting and comparing, identify the target nominal parameter item corresponding to the minimum absolute difference, and extract the pipe material type description, preset outer diameter value, and preset wall thickness parameter set bound to this nominal parameter item. Retrieve the equipment management code bound to the current location, combine it with the three-dimensional spatial geographic coordinates, encapsulate the material and size data, analyze the IoT association between multi-dimensional physical attributes and identification codes, and summarize and output a full-element pipeline entity physical attribute label dataset. Assume the standard pipe diameter database dataset contains model A with a wall thickness of 0.15 meters, model B with a wall thickness of 0.20 meters, and model C with a wall thickness of 0.25 meters. Subtracting the nominal value from the feature dimension of 0.19 meters yields absolute differences of 0.04 meters, 0.01 meters, and 0.06 meters, respectively. The smallest difference of 0.01 meters corresponds to model B. The pipeline material is determined to be polyethylene, with an outer diameter of 1.0 meter and a wall thickness of 0.20 meters. The equipment code is extracted as 10086. The parameters are combined and packaged to generate entity attribute labels. The beneficial aspect of this calculation logic is the use of error matching to eliminate minor imperfections.

[0086] The aforementioned intelligent identification process for underground pipelines can also be achieved through an intelligent underground pipeline identification system based on ground-penetrating radar and deep learning. This system specifically includes: The bellows extraction module scans the radar signal matrix, extracts the amplitude sequence of the target pipe's circumferential reflection zone and the phase variation law at the socket, analyzes the spatial phase distribution of the center and edge pixels, fuses the amplitude and phase variation components, and generates the underlying electromagnetic waveform energy characterization vector matrix. The parameter inversion module calls the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation to perform nonlinear fitting and weight mapping, inverts the gradient correction coefficient of the enhanced polyethylene medium, performs segment weighting on the target reflection zone, and obtains the numerical sequence of underground dynamic dielectric constant distribution. The wave velocity analysis module calls the underground dynamic dielectric constant distribution sequence numerical to perform electromagnetic wave dynamics inversion, obtains the local real electromagnetic wave propagation velocity numerical value, extracts the target reflection feature coordinates, performs continuous second-order difference to obtain the morphological curvature derivative, and performs deformation deviation analysis with the standard curvature constant to generate curvature morphological distortion difference numerical value. The interface mapping module calls the curvature shape distortion difference value and the local real electromagnetic wave propagation speed value to perform time delay compensation convolution mapping, performs offset reconstruction correction on the original bidirectional time, and performs spatial geometric projection analysis in combination with the antenna position to generate a pipeline spatial coordinate intelligent identification list. The attribute recognition module calls the pipeline spatial coordinate intelligent recognition list and the local real electromagnetic wave propagation speed value to locate the radial reflection area, analyzes the wall reflection travel time difference between polarity reversal points, obtains the pipeline radial feature size after dispersion correction, and outputs the pipeline entity physical attribute label dataset.

[0087] This invention extracts the hyperbolic reflection wave amplitude sequence and phase shift distribution to analyze the edge variation law, integrates energy characterization to perform network mapping and dynamically correct the initial formation parameters, effectively overcomes the limitations of fixed empirical values, enhances the adaptive calculation capability for the evolution of physical properties of multi-layer media, improves the capture accuracy of deep weak echo signals, inverts wave velocity based on dynamic distribution and analyzes the waveform curvature and distortion difference, combines delay operator to perform convolution compensation to generate increments to verify the original travel time, significantly optimizes the accuracy of propagation feature restoration, reduces cross-layer refraction cumulative mapping error, ensures high fidelity conversion of travel time to spatial depth, greatly improves the three-dimensional spatial coordinate positioning accuracy under complex structures, and realizes efficient restoration and reconstruction of pipeline physical entity position and feature parameters.

[0088] To better implement the intelligent underground pipeline identification method in this invention embodiment, based on the intelligent underground pipeline identification method, correspondingly, as follows: Figure 3 As shown, this embodiment of the invention also provides an intelligent identification device for underground pipelines. The intelligent identification device 300 for underground pipelines includes: The acquisition module 301 is used to acquire the electromagnetic echo signal matrix generated by the ground penetrating radar and generate the underlying electromagnetic waveform energy characterization vector matrix based on the electromagnetic echo signal matrix. The correction module 302 is used to perform nonlinear fitting calculations on the energy characterization vector matrix of the underlying electromagnetic waveform and the initial medium parameters of the formation, determine the dielectric constant gradient correction coefficient, and correct the initial medium parameters of the formation based on the dielectric constant gradient correction coefficient to determine the underground dynamic dielectric constant distribution sequence. The inversion module 303 is used to determine the coordinates of the feature points of the echo matrix hyperbola by performing local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence, determine the geometric curvature derivative of the echo matrix hyperbola based on the coordinates of the feature points of the echo matrix hyperbola, and determine the curvature distortion difference based on the geometric curvature derivative and the curvature of the standard medium. The determination module 304 is used to correct the hyperbola of the echo matrix based on the curvature distortion difference, and to perform spatial vector mapping on the corrected hyperbola of the echo matrix based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

[0089] The underground pipeline intelligent identification device 300 provided in the above embodiments can realize the technical solutions described in the above underground pipeline intelligent identification method embodiments. The specific implementation principles of each module or unit can be found in the corresponding content in the above underground pipeline intelligent identification method embodiments, which will not be repeated here.

[0090] The above provides a detailed description of the intelligent identification method and device for underground pipelines provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent identification of underground pipelines, characterized in that, include: Obtain the electromagnetic echo signal matrix generated by ground penetrating radar, and generate the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix; The energy characterization vector matrix of the underlying electromagnetic waveform is nonlinearly fitted to the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient. Based on the dielectric constant gradient correction coefficient, the initial medium parameters of the formation are corrected to determine the underground dynamic dielectric constant distribution sequence. The coordinates of the characteristic points of the echo matrix hyperbola are determined by local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence. The geometric curvature derivative of the echo matrix hyperbola is determined based on the coordinates of the characteristic points of the echo matrix hyperbola. The curvature distortion difference is determined based on the geometric curvature derivative and the curvature of the standard medium. The hyperbola of the echo matrix is ​​corrected based on the curvature distortion difference, and the corrected hyperbola of the echo matrix is ​​spatially vector-mapped based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

2. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The generation of the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix includes: Scan the electromagnetic echo signal matrix, locate the center pixel position of the hyperbola echo peak in the two-dimensional spatiotemporal domain, extract the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, perform signal time-domain discretization analysis and amplitude normalization mapping on the echo amplitude sequence value and initial phase offset value corresponding to the center pixel position, and obtain the basic characterization component data of the vertex waveform. Based on the vertex waveform basic characterization component data, the phase variation ratio of edge pixels in the eight neighborhood directions of the center pixel is analyzed. The phase variation ratio of edge pixels is distributed with the radar signal attenuation factor. The spatial phase gradient deviation value is obtained by linear aggregation of edge phase gradient and center offset. Based on the vertex waveform basic characterization component data and the spatial phase gradient deviation variation value, the energy amplitude square root constraint is applied to the echo amplitude sequence value. The constraint result is mapped with the spatial phase gradient deviation variation value through tensor dot product to generate the underlying electromagnetic waveform energy characterization vector matrix.

3. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The process of performing nonlinear fitting calculations on the energy characterization vector matrix of the underlying electromagnetic waveform and the initial medium parameters of the formation to determine the dielectric constant gradient correction coefficient includes: Based on the vector dot product mapping between the underlying electromagnetic waveform energy characterization vector matrix and the initial medium parameters of the formation, the inner product of waveform and medium energy response is obtained. The inner product of waveform and medium energy response is compared with the fitted correlation benchmark value under a preset threshold constraint. The coordinates of elements that meet the response conditions are extracted and aggregated into a two-dimensional array to obtain the initial medium waveform energy correlation matrix. Based on the initial dielectric waveform energy correlation matrix and the node weights of the deep neural network, the deviation response is mapped to obtain the node-mapped response. The node-mapped response is then processed using a nonlinear activation function to determine the dielectric constant gradient correction coefficient.

4. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The process of correcting the initial medium parameters of the formation based on the dielectric constant gradient correction coefficient to determine the dynamic dielectric constant distribution sequence in the subsurface includes: The gradient correction coefficient of dielectric constant is fused with the initial medium parameters of the formation to determine the weighted intermediate values ​​of the layer. Differential evolution analysis is performed on the interlayer differences of the weighted intermediate values ​​of the layer to determine the differential values. The differential values ​​of the layer differential benchmark values ​​are compared with the differential benchmark values ​​of the layer. High-frequency disturbances are removed and the continuous intermediate values ​​that are smoothed and retained are combined to determine the distribution sequence of the underground dynamic dielectric constant.

5. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The determination of the coordinates of hyperbolic feature points of the echo matrix by local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence includes: Electromagnetic wave dynamics inversion is performed based on the underground dynamic dielectric constant distribution sequence to solve for the propagation velocity value. The spatial horizontal and vertical coordinate components of the characteristic points of the electromagnetic echo hyperbola are separated. Based on the propagation velocity value, the coordinates of the characteristic points are recombined and offset mapped to determine the coordinates of the characteristic points of the echo matrix hyperbola.

6. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The determination of the geometric curvature derivative of the hyperbola based on the coordinates of its feature points includes: The spatial horizontal and vertical coordinate components of the echo matrix hyperbola feature point coordinates are continuously differentiated by second-order difference operation. The intermediate terms of the first difference between adjacent trajectory points are extracted. The intermediate terms of the first difference are differentiated again and spatial smoothing filtering is performed based on the curvature change benchmark value to determine the geometric curvature derivative of the echo matrix hyperbola.

7. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The determination of curvature distortion difference based on the geometric curvature derivative and the standard medium curvature includes: Deformation deviation analysis is performed on the geometric curvature derivative and standard medium curvature. The absolute amount of deformation is extracted and a distortion assessment sequence is constructed. After removing sparse outliers in the absolute amount distribution density characteristics of the distortion assessment sequence, the retained terms are fused to generate the curvature morphology distortion difference.

8. The intelligent identification method for underground pipelines according to claim 1, characterized in that, The correction of the hyperbola echo matrix based on the curvature distortion difference includes: Based on the curvature morphology distortion difference and time delay sensitive operator, convolution mapping operation is performed in the discrete time domain to determine the travel time disturbance deviation. The travel time disturbance deviation is verified by time domain offset to determine the time deviation adjustment value. Based on the time deviation adjustment value and the ground penetrating radar sampling frequency, the travel time delay compensation increment is generated. The original bidirectional travel time values ​​of the waveform vertices of the echo matrix hyperbola are extracted. The original bidirectional travel time values ​​of the waveform vertices are offset and recombined based on the travel time delay compensation increment to obtain the corrected travel time parameters.

9. The intelligent identification method for underground pipelines according to claim 8, characterized in that, The process of spatial vector mapping of the corrected echo matrix hyperbola based on local wave velocity to determine the underground pipeline coordinate identification result includes: Based on the corrected travel time parameters and the local real electromagnetic wave propagation speed, a one-way propagation distance spatial mapping is performed. Combined with the position coordinates of the ground penetrating radar antenna, spatial geometric projection analysis is performed. The three-dimensional spatial mapping relationship of depth and horizontal displacement is transformed to determine the underground pipeline coordinate identification result.

10. An intelligent identification device for underground pipelines, characterized in that, include: The acquisition module is used to acquire the electromagnetic echo signal matrix generated by the ground penetrating radar, and generate the underlying electromagnetic waveform energy representation vector matrix based on the electromagnetic echo signal matrix; The correction module is used to perform nonlinear fitting calculations on the energy characterization vector matrix of the underlying electromagnetic waveform and the initial medium parameters of the formation, determine the dielectric constant gradient correction coefficient, and correct the initial medium parameters of the formation based on the dielectric constant gradient correction coefficient to determine the underground dynamic dielectric constant distribution sequence. The inversion module is used to determine the coordinates of the feature points of the echo matrix hyperbola by performing local wave velocity inversion based on the underground dynamic dielectric constant distribution sequence, determine the geometric curvature derivative of the echo matrix hyperbola based on the coordinates of the feature points of the echo matrix hyperbola, and determine the curvature distortion difference based on the geometric curvature derivative and the curvature of the standard medium. The determination module is used to correct the hyperbola of the echo matrix based on the difference in curvature morphology distortion, and to perform spatial vector mapping on the corrected hyperbola of the echo matrix based on the local wave velocity to determine the coordinate identification result of the underground pipeline.

Images