Underground pipeline distribution detection method and system based on ground penetrating radar

Abstract

The invention discloses an underground pipeline distribution detection method and system based on a ground penetrating radar, and relates to the field of underground pipe network engineering. The method comprises the steps of performing orthogonal grid division on a to-be-detected interval based on grid preset parameters, performing GPR orthogonal scanning on a grid detection area, and determining an initial trend of an underground pipeline according to a first vertex and a second vertex; adjusting the initial trend of the underground pipeline based on the initial vertical scanning direction; on the basis of the adjusted direction of the underground pipeline, continuing to perform GPR scanning on the grid detection area until scanning of the grid detection area is completed, and inputting obtained radar original data into the image detection model to obtain a first type of parameter value corresponding to the underground pipeline; and inputting the first type parameter value and the second type parameter value into a pipe diameter prediction model to obtain a pipe diameter value. According to the invention, through linkage of the three modules, intelligent and efficient cooperative one-stop service of detection image analysis, detection path planning and detection remote control can be realized.

Description

Technical Field

[0001] This application relates to the field of underground pipeline engineering, and in particular to a method and system for detecting the distribution of underground pipelines based on ground-penetrating radar. Background Technology

[0002] The rapid pace and improvement of urbanization have driven the construction of municipal underground pipe networks, resulting in a large number of networks, wide distribution, and high investment. Multiple factors, including rapid urban expansion and a focus on underground infrastructure while neglecting above-ground infrastructure, have led to problems such as insufficient overall planning, unclear construction data, inadequate daily maintenance, and prominent safety hazards in China's municipal underground pipe network facilities. The "old," "dense," and "disorganized" state of these facilities is becoming increasingly prominent. Pipeline network renovation is a crucial measure to ensure the safety of urban lifelines.

[0003] The primary technical challenge in pipeline renovation is to locate and investigate unknown underground pipelines. The following methods are typically employed: (1) reviewing the initial planning and design drawings or the as-built drawings of the pipeline to be renovated; (2) excavating or manually drilling wells (with workers going down into the wells) for exploration; and (3) using equipment such as pipeline detectors for geophysical exploration.

[0004] However, research has revealed the following drawbacks of existing technologies, specifically: (1) Poor accuracy of pipeline drawings: Design or construction changes, manual measurement, etc. may cause errors between the initial design drawings or as-built drawings of the pipeline network that needs to be modified and the actual pipeline network location. These errors are difficult to control precisely, thus affecting the accuracy of pipeline network location detection.

[0005] (2) Excavation and drilling exploration is highly risky: Excavation is destructive and may damage existing pipelines; manual drilling exploration is dangerous to workers' safety, and when the distance between two wells is large, it is very challenging to determine the location of the pipeline network between the wells, resulting in low exploration efficiency.

[0006] (3) Low automation of pipeline detection equipment: Most pipeline detection equipment on the market has problems such as relying on manual analysis of detection images, low degree of automation of equipment drive, and fragmented detection record management information, resulting in subjective dependence on accuracy and efficiency.

[0007] Therefore, there is an urgent need for a new method for detecting the distribution of underground pipelines to solve the problems existing in the current technology. Summary of the Invention

[0008] The purpose of this application is to address at least one of the aforementioned technical deficiencies.

[0009] On the one hand, embodiments of this application provide a method for detecting the distribution of underground pipelines based on ground-penetrating radar, the method comprising: Obtain the grid preset parameters, and divide the area to be detected into orthogonal grids based on the grid preset parameters to obtain the grid detection area, which includes longitudinal survey line marks and transverse survey line marks; Based on the longitudinal and transverse survey line marks, the grid detection area is subjected to GPR orthogonal scanning to determine the first and second vertices of the hyperbola, and the initial direction of the underground pipeline is determined based on the first and second vertices. Determine the initial vertical scanning direction, and adjust the initial orientation of the underground pipeline along the longitudinal and transverse survey line marks based on the initial vertical scanning direction to obtain the adjusted underground pipeline orientation. The initial vertical scanning direction is perpendicular to the initial orientation of the underground pipeline. Based on the adjusted underground pipeline route, GPR scanning continues in the grid detection area until the grid detection area is scanned. The original radar data corresponding to the section to be detected is obtained, and the original radar data is input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline. The first type of parameter values ​​include the material, burial depth and hyperbola equation of the underground pipeline hyperbola feature fitting. The original radar data includes at least one GPR image and GPS data. Obtain the second type of parameter values ​​corresponding to the underground pipeline, and input the first type of parameter values ​​and the second type of parameter values ​​into the pipe diameter prediction model to obtain the pipe diameter value corresponding to the underground pipeline; Based on the first type of parameter values, the second type of parameter values, the pipe diameter value, and GPS data corresponding to the underground pipelines, the distribution of underground pipelines in the area to be inspected is determined.

[0010] Optionally, based on the longitudinal and transverse survey line marks, a GPR orthogonal scan is performed on the grid detection area to determine the first and second vertices of the hyperbola, including: The GPR device is driven at a constant speed along the longitudinal and transverse survey lines to perform GPR orthogonal scanning, thereby obtaining the hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line. The hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line are used to determine the survey line with the strongest signal, and the position of the strongest survey line is taken as the first vertex of the hyperbola. Based on the first vertex of the hyperbola, the GPR device is continuously pushed at a constant speed along the longitudinal survey line to perform GPR orthogonal scanning, thus obtaining the second vertex of the hyperbola.

[0011] Optionally, based on the first vertex and the second vertex, the initial direction of the underground pipeline is determined, including: Connect the first vertex and the second vertex to obtain a connecting line, and determine the angle between the connecting line and the due north direction; The target scanning direction is determined based on the connection angle, and GPR orthogonal scanning is continued according to the target scanning direction until the vertices of the hyperbolas included in the GPR scan image meet the preset requirements, thus obtaining the initial direction of the underground pipeline. The target scanning direction is 90 degrees away from the connection angle. If the target scanning direction is offset, the target scanning direction is adjusted according to a preset step size, and GPR orthogonal scanning continues based on the adjusted target scanning direction.

[0012] Optionally, the initial orientation of the underground pipeline is adjusted based on the initial vertical scanning direction along the longitudinal and transverse survey line marks to obtain the adjusted underground pipeline orientation, including: Based on the initial vertical scanning direction, the GPR device is driven at a constant speed along the longitudinal and transverse survey line marks to perform GPR orthogonal scanning and obtain GPR scan images; If the vertex position of the hyperbola included in the GPR scan image shifts, the initial direction of the underground pipeline is adjusted to obtain the adjusted underground pipeline direction.

[0013] Optionally, the image detection model includes a target detection network. Raw radar data is input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, including: The GPR image is input into the target detection network to obtain the predicted target bounding box of the underground pipeline based on its material and hyperbolic features; The predicted target bounding box is cropped in the GPR image to obtain a cropped image. The cropped image is then preprocessed to obtain a processed cropped image. The image preprocessing includes grayscale conversion, Gaussian denoising, contrast enhancement, adaptive binarization, opening operation denoising, and closing operation hole filling. The processed cropped image is subjected to hyperbola fitting to obtain the hyperbola equation. Hyperbola fitting includes extracting the maximum connected component, skeletonization, and skeleton pruning. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0014] Optionally, the image detection model includes an instance segmentation network. The raw radar data is input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, including: The GPR image is input into the instance segmentation network to obtain the material and hyperbolic features corresponding to the underground pipeline; The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0015] Optionally, the image detection model includes a CNN classification network. The raw radar data is input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, including: The GPR images are input into a CNN classification network to obtain GPR images with hyperbolic features, GPR images without hyperbolic features, and the material corresponding to the underground pipeline. A sliding window exhaustive search process is performed on GPR images with hyperbolic features to obtain the hyperbolic features corresponding to underground pipelines. The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0016] Optionally, after obtaining the first type of parameter values ​​corresponding to the underground pipeline, the following may also be included: The matching degree of the hyperbola equation is evaluated by using the asymptote weighted vertex error formula or the adaptive deformation similarity formula, and the matching result is obtained.

[0017] Optionally, the method further includes: Determine the relative errors for underground pipelines separately. The relative errors include the relative error of burial depth and the relative error of pipe diameter. Obtain a preset engineering risk threshold, determine the number of relative errors less than the engineering risk threshold based on the engineering risk threshold, and determine the result evaluation result based on the number of relative errors less than the engineering risk threshold. The predicted pipeline and the actual pipeline are matched based on the Hungarian algorithm to determine the optimal matching pipeline. Based on the burial depth, pipe diameter and preset relative error threshold of the optimal matching pipeline, the number of qualified matching indicators is determined, and the correlation accuracy assessment result is determined based on the number of qualified matching indicators.

[0018] On the other hand, embodiments of this application provide an underground pipeline distribution detection system based on ground-penetrating radar. This system includes an image processing module, a work trolley module, and a control platform module. This ground-penetrating radar-based underground pipeline distribution detection system is used to execute any method for underground pipeline distribution detection based on ground-penetrating radar, wherein: The image processing module is used to receive the radar raw data and obtain the first type of parameter value, the second type of parameter value, and the pipe diameter value corresponding to the underground pipeline based on the radar raw data. According to the first type of parameter value, the second type of parameter value, and GPS data corresponding to the underground pipeline, the distribution result of the underground pipeline in the area to be detected is determined. The first type of parameter value includes the material, burial depth, and hyperbolic equation of the pipeline hyperbolic feature fitting. The radar raw data includes at least one GPR image and GPS data. The work vehicle module includes a GPR device and a drive platform. The GPR device is mounted on the drive platform via a keyway and relies on the drive platform to drive the scanning forward to obtain the raw radar data. The GPR device includes GPS positioning, device information and wireless transmission functions, and transmits the scanned raw radar data to the image processing module. The control platform module is used to send control commands to the trolley module, monitor the operation status of the trolley module, and determine the underground pipeline distribution results based on the first type of parameter values, the second type of parameter values, the pipe diameter value, and the GPS data sent by the image processing module.

[0019] In another aspect, embodiments of this application provide an electronic device, including a processor and a memory: The memory is configured to store machine-readable instructions that, when executed by the processor, cause the processor to perform any one of the methods in a ground-penetrating radar-based method for detecting the distribution of underground pipelines.

[0020] The beneficial effects of the technical solutions provided in this application include at least the following: In this embodiment, the area to be detected can be divided into grids, and GPR scanning can be performed along the marked lines of the grid division. Furthermore, the scanning direction angle is adjusted in real time during scanning to obtain the final direction of the underground pipeline, thus achieving automatic optimization control of the scanning path and ensuring the accuracy of the subsequently obtained underground pipeline parameter values. After scanning, the obtained raw radar data can be input into the image detection model and pipe diameter prediction model to obtain the final parameter prediction data. It is evident that the detection of underground pipelines in this embodiment utilizes intelligent image analysis, independent of human subjective experience, achieving remote, fully automated analysis of the detected images, improving detection and identification efficiency and accuracy. Simultaneously, it enables efficient and unified detection, identification, and management of wide-area municipal underground pipe networks, reducing human interference and information entry errors, thereby ensuring the correctness and consistency of the data source.

[0021] In this embodiment, the AWE method dynamically weights vertex coordinate errors with λ, which enhances vertex sensitivity. It also introduces the logarithmic error of the asymptote slope ratio, thereby capturing the consistency of the opening direction. ADS, on the other hand, strengthens vertex region error penalty through Gaussian terms and adds an opening compensation mechanism, thus reducing evaluation errors caused by missing far-end data. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A schematic flowchart illustrating an underground pipeline distribution detection method based on ground-penetrating radar provided in this application embodiment; Figure 2 A schematic diagram of the scanning direction provided for an embodiment of this application; Figure 3 A schematic diagram of direction finding provided for an embodiment of this application; Figure 4 This is a schematic diagram of vertex coordinates in direction finding 1 provided in an embodiment of this application; Figure 5 This is a schematic diagram of vertex coordinates in direction finding 2 provided in the embodiments of this application; Figure 6 A schematic diagram illustrating the sliding window exhaustive search process provided in an embodiment of this application; Figure 7 A schematic diagram of the structure of an underground pipeline distribution detection device based on ground penetrating radar provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting the invention.

[0025] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this application means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0027] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0028] Specifically, such as Figure 1 As shown, the method may include: Step S101: Obtain the grid preset parameters, and divide the area to be detected into orthogonal grids based on the grid preset parameters to obtain the grid detection area, which includes longitudinal survey line marks and transverse survey line marks.

[0029] Optionally, for areas to be inspected where underground pipelines may exist, an orthogonal network can be laid out based on preset grid parameters to obtain a grid inspection area. For example, if the area to be inspected is 2m×2m, the area can be divided into grids, such as drawing a line every 0.5m, forming a 4×4 grid. The longitudinal (X-axis) and transverse (Y-axis) survey lines must be strictly perpendicular. The longitudinal survey lines are arranged along the north-south direction, with a line painted every 2m and labeled (S1, S2, S3...). The transverse survey lines are arranged along the east-west direction, with a line painted every 2m (labeled T1, T2, T3...).

[0030] Step S102: Perform GPR orthogonal scanning on the grid detection area based on the longitudinal and transverse survey line marks to determine the first and second vertices of the hyperbola, and determine the initial direction of the underground pipeline based on the first and second vertices.

[0031] Optionally, after marking the area to be measured with a grid, the grid detection area can be orthogonally scanned with GPR (Ground Penetrating Radar) based on the marked longitudinal and transverse survey lines to obtain the first and second vertices of the hyperbola containing hyperbolic features. Then, based on the determined first and second vertices, the initial direction of the underground pipeline can be determined.

[0032] Optionally, based on the longitudinal and transverse survey line marks, a GPR orthogonal scan is performed on the grid detection area to determine the first and second vertices of the hyperbola, including: The GPR device is driven at a constant speed along the longitudinal and transverse survey lines to perform GPR orthogonal scanning, thereby obtaining the hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line. The hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line are used to determine the survey line with the strongest signal, and the position of the strongest survey line is taken as the first vertex of the hyperbola. Based on the first vertex of the hyperbola, the GPR device is continuously pushed at a constant speed along the longitudinal survey line to perform GPR orthogonal scanning, thus obtaining the second vertex of the hyperbola.

[0033] Optionally, when performing GPR orthogonal scanning on the grid detection area, the GPR device can be pushed at a constant speed according to the marked longitudinal and transverse survey lines, and the hyperbola signal intensity on the GPR device screen can be observed in real time during the pushing process. The hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line can be obtained based on the observed signal intensity. Furthermore, the position of the survey line with the strongest hyperbola signal can be taken as the first vertex of the hyperbola. Then, based on the determined first vertex, the GPR device can be pushed at a constant speed along the longitudinal survey line to perform GPR orthogonal scanning to obtain the second vertex of the hyperbola.

[0034] For example, the GPR device can be pushed at a constant speed along the longitudinal survey lines (S1, S2, S3...) while keeping the antenna direction parallel to the survey lines. During this pushing process, the GPR screen is observed in real time, and the locations where hyperbolic signals appear are recorded. This process continues until all longitudinal survey lines have been scanned. Then, the scanning is repeated along the transverse survey lines (T1, T2, T3...), with the GPR screen observed in real time and the locations where hyperbolic signals appear recorded, until all transverse survey lines have been scanned to cover the same area. During the scanning process, if a strong and continuous hyperbolic signal is observed in the longitudinal scan, while the transverse signal is weak, the pipeline is determined to be closer to the longitudinal direction; conversely, if the hyperbolic signal is weak, the pipeline is determined to be closer to the transverse direction.

[0035] Furthermore, when the survey line with the strongest hyperbola signal is detected, the scanning is paused, and the first vertex of the hyperbola (point A) is marked on the ground with a mark. The coordinates of point A are recorded. Then, starting from point A, the ground-penetrating radar equipment is pushed along the longitudinal survey line at a 45° angle to scan a certain distance, observe the position of the new second vertex of the hyperbola (point B), and mark it on the ground with a mark.

[0036] In an optional embodiment of this application, determining the initial direction of the underground pipeline based on the first vertex and the second vertex includes: Connect the first vertex and the second vertex to obtain a connecting line, and determine the angle between the connecting line and the due north direction; The target scanning direction is determined based on the connection angle, and GPR orthogonal scanning is continued according to the target scanning direction until the vertices of the hyperbolas included in the GPR scan image meet the preset requirements, thus obtaining the initial direction of the underground pipeline. The target scanning direction is 90 degrees away from the connection angle. If the target scanning direction deviates, the target scanning direction is adjusted according to the preset step size, and GPR orthogonal scanning continues based on the adjusted target scanning direction.

[0037] Optionally, the first and second vertices can be connected to obtain a connecting line, and the angle between this connecting line and the due north direction can be calculated. This angle direction is the target scanning direction. Further, the ground-penetrating radar equipment is driven along the target scanning direction to perform GPR orthogonal scanning, while observing whether the vertices of the hyperbola meet preset requirements, such as symmetry and vertex stability. Vertex stability means that after a set number of consecutive scans, the position of the hyperbola vertex in the obtained GPR image meets the preset requirements, such as the coordinate difference between several positions being less than a threshold. This indicates that the position of the hyperbola vertex has almost no shift, and can be considered that the hyperbola vertex is in a stable state. Further, if these conditions are met, it represents the initial direction of the underground pipeline. In practical applications, if the target scanning direction may shift, the target scanning direction is adjusted according to a preset step size, such as fine-tuning the angle in 5° steps, until the hyperbola is symmetrical and the vertices are stable, thus obtaining the initial direction of the underground pipeline.

[0038] Step S103: Determine the initial vertical scanning direction, and adjust the initial direction of the underground pipeline along the longitudinal and transverse survey lines based on the initial vertical scanning direction to obtain the adjusted underground pipeline direction. The initial vertical scanning direction is perpendicular to the initial direction of the underground pipeline.

[0039] In actual detection, ground penetrating radar needs to complete a relatively long detection route (referred to as the survey line). Along this survey line, the direction of underground pipelines may change, so it is necessary to correct the detection direction of the ground penetrating radar in real time (referred to as direction finding) to ensure that it is perpendicular to the direction of the underground pipelines.

[0040] In an optional embodiment of this application, the initial orientation of the underground pipeline is adjusted based on the initial vertical scanning direction along the longitudinal and transverse survey line marks to obtain the adjusted underground pipeline orientation, including: Based on the initial vertical scanning direction, the GPR device is driven at a constant speed along the longitudinal and transverse survey line marks to perform GPR orthogonal scanning and obtain GPR scan images; If the vertex position of the hyperbola included in the GPR scan image shifts, the initial direction of the underground pipeline is adjusted to obtain the adjusted underground pipeline direction.

[0041] Optionally, after determining the initial direction of the underground pipeline, an initial vertical scanning direction can be set, which is perpendicular to the initial direction of the underground pipeline. For example, ... Figure 2 As shown, the initial orientation of the underground pipeline forms a 45° angle with due north. Therefore, the initial vertical scan direction... It should be 135°.

[0042] Furthermore, at regular intervals (e.g., every 2m) along the survey lines (i.e., longitudinal and transverse survey line marks), the ground-penetrating radar equipment is driven at a constant speed along the current vertical scanning direction to perform GPR orthogonal scanning, obtaining GPR scan images (also known as GPR-scan images). The position of the apex of the hyperbola in the GPR scan image is used to determine whether the pipeline's orientation has changed. If the apex of the hyperbola in the GPR scan image shifts to the left or right, it indicates a change in the pipeline's orientation, requiring adjustment of the scanning direction to obtain the adjusted underground pipeline orientation. The formula for adjusting the angle is: in, Indicates adjusting the angle. is the offset distance of the vertex of the hyperbola in the GPR scan image. For scanning speed, For time.

[0043] For example, such as Figure 3 As shown, direction finder 4 is perpendicular to pipeline A, and direction finder 2 is located a distance away from direction finder 1. The direction of direction finder 2 is parallel to direction finder 1. At this time, as... Figure 4 The coordinates of the vertex of the hyperbola of the underground pipeline shown in the GPR scan image - Direction Finding 1 are: ,like Figure 5 As shown, the coordinates of the vertex of the hyperbola of the underground pipeline in the GPR scan image - Direction Finding 2 are... Clearly, the coordinates of the vertex of the hyperbola of the underground pipeline have shifted, therefore the direction-finding angle (i.e., scanning direction) of the ground-penetrating radar needs to be adjusted. The angle that needs to be adjusted is: Step S104: Based on the adjusted underground pipeline route, continue to perform GPR scanning on the grid detection area until the grid detection area is scanned, obtain the radar raw data corresponding to the section to be detected, and input the radar raw data into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline. The first type of parameter values ​​include the material, burial depth and hyperbolic equation of the pipeline hyperbolic feature fitting. The radar raw data includes at least one GPR image and GPS (Global Positioning System) data.

[0044] Optionally, after obtaining the adjusted underground pipeline route, GPR scanning can be continued on the grid detection area based on the adjusted route until the entire grid detection area is scanned and covered. At this point, the raw radar data corresponding to the section to be detected can be obtained. This raw radar data includes at least one GPR image and GPS data. Furthermore, this raw radar data can be uploaded to the backend and input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline. These first type of parameter values ​​can specifically include the material, burial depth, and hyperbolic equation fitted by the hyperbolic feature of the underground pipeline.

[0045] In an optional embodiment of this application, the image detection model includes a target detection network. Raw radar data is input into the image detection model to obtain first-type parameter values ​​corresponding to the underground pipeline, including: The GPR image is input into the target detection network to obtain the predicted target bounding box of the underground pipeline based on its material and hyperbolic features; The predicted target bounding box is cropped in the GPR image to obtain a cropped image. The cropped image is then preprocessed to obtain a processed cropped image. The image preprocessing includes grayscale conversion, Gaussian denoising, contrast enhancement, adaptive binarization, opening operation denoising, and closing operation hole filling. The processed cropped image is subjected to hyperbola fitting to obtain the hyperbola equation. Hyperbola fitting includes extracting the maximum connected component, skeletonization, and skeleton pruning. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0046] Optionally, the image detection model may include an object detection network, which can be a YOLO (YouOnly Look Once) series, an R-CNN (Regions with Convolutional Neural Networks) series, or an SSD (Single Shot MultiBox Detector) series, as specified in this embodiment. In practical applications, a dataset (referred to as dataset 1) can be established by collecting original GPR B-scan images (i.e., GPR images) of underground pipelines whose material, burial depth (parameter 2), pipe diameter, dielectric constant of the buried medium, conductivity of the buried medium, spacing of radar transceiver antennas, and hyperbolic equation fitted to the hyperbolic features of the pipeline are all known. Then, the target boxes and materials of the hyperbolic features of the underground pipelines included in all original GPR B-scan images in dataset 1 are labeled to obtain a training dataset (referred to as dataset 2). Further, the initial object detection network is trained based on dataset 2 until the corresponding loss function converges to obtain the object detection network.

[0047] Correspondingly, the GPR image from the original radar data can be input into the target detection network to obtain the predicted target bounding boxes of the underground pipeline's material and hyperbolic features. Then, based on the coordinates of the predicted target bounding boxes of the underground pipeline's hyperbolic features, the original GPR image is cropped to obtain a cropped image. Further, all the obtained cropped images are sequentially processed using grayscale conversion, Gaussian denoising, contrast enhancement, adaptive binarization, opening denoising, and closing hole filling to obtain the processed cropped image. Then, hyperbolic fitting is performed on the processed cropped image to obtain the hyperbolic equation: Where a, b, k, h are constants that are set in advance.

[0048] Furthermore, the vertex coordinates of the hyperbola equation can be determined, and then the burial depth of the underground pipeline can be calculated based on the vertex coordinates of the hyperbola equation.

[0049] Specifically, the top-left pixel of the original GPR-scan image is used as the origin (0,0). The horizontal direction to the right of the origin is the positive X-axis, and the vertical direction downwards is the positive Y-axis, establishing a coordinate system. There are two common types of GPR B-scan images: the range-two-way travel time GPR image and the range-depth GPR image. The range-two-way travel time GPR image represents the X-axis of the ground-penetrating radar's range measurement, and the Y-axis represents the ground-penetrating radar's two-way travel time. Range-two-way travel time refers to the time it takes for a radar wave to travel from the transmitting antenna, be reflected by the underground medium, and return to the receiving antenna. The range-depth GPR image represents the X-axis of the ground-penetrating radar's range measurement, and the Y-axis represents the ground-penetrating radar's depth.

[0050] Suppose that the coordinates of the vertex of the fitted hyperbola are... If the GPR B-scan image is a distance-depth GPR image, then the formula for calculating the burial depth of underground pipelines is: in, This represents the true depth of the probe in the GPR image. This represents the height in pixels of the GPR image.

[0051] If the GPR B-scan image is a distance-two-way travel GPR image, then the formula for calculating the burial depth of underground pipelines is: in, Indicating two-way travel, Indicates radar wave speed. This represents the height in pixels of the GPR image.

[0052] The skeleton pruning strategy is that each x value on the skeleton has only one corresponding y value. The actual operation methods of grayscale processing, Gaussian denoising processing, contrast enhancement processing, adaptive binarization processing, opening operation denoising processing and closing operation hole filling processing, maximum connected component extraction processing, skeletonization processing and skeleton pruning processing can refer to existing technologies, and will not be elaborated here.

[0053] In an optional embodiment of this application, the image detection model includes an instance segmentation network. Raw radar data is input into the image detection model to obtain first-class parameter values ​​corresponding to the underground pipeline, including: The GPR image is input into the instance segmentation network to obtain the material and hyperbolic features corresponding to the underground pipeline; The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0054] Optionally, the image detection model can also include an instance segmentation network, such as MaskR-CNN (Mask Region-based Convolutional Neural Networks for mask prediction) or YOLACT (You Only Look At Coefficients, a neural network model). In practical applications, a dataset (referred to as dataset 1) can be created by collecting original GPR B-scan images (i.e., GPR images) of underground pipelines, where the material, burial depth (parameter 2), pipe diameter, dielectric constant of the buried medium, conductivity of the buried medium, spacing of radar transceiver antennas, and hyperbolic equation fitted to the hyperbolic features of the pipeline are all known. Then, the target boxes, hyperbolic feature regions, and materials of the underground pipeline signals in all the original GPR B-scan images of dataset 1 are labeled to obtain dataset 3. The initial instance segmentation network is then trained based on dataset 3 until the corresponding loss function converges, resulting in the segmentation network.

[0055] Furthermore, the GPR image is input into a pre-trained instance segmentation network. This allows for the segmentation of hyperbolic features belonging to underground pipelines within the GPR image and the identification of the pipeline material. The obtained hyperbolic features are then sequentially processed through skeletonization, skeleton pruning, and hyperbolic fitting algorithms to obtain the hyperbolic equation. Finally, the burial depth of the underground pipeline is determined based on the vertex coordinates of the obtained hyperbolic equation. The specific implementation methods for processing the obtained hyperbolic features through skeletonization, skeleton pruning, and hyperbolic fitting algorithms to obtain the hyperbolic equation, and for determining the burial depth of the underground pipeline based on the vertex coordinates of the obtained hyperbolic equation, are described above and will not be repeated here.

[0056] In an optional embodiment of this application, the image detection model includes a CNN (Convolutional Neural Network) classification network. Raw radar data is input into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, including: The GPR images are input into a CNN classification network to obtain GPR images with hyperbolic features, GPR images without hyperbolic features, and the material corresponding to the underground pipeline. A sliding window exhaustive search process is performed on GPR images with hyperbolic features to obtain the hyperbolic features corresponding to underground pipelines. The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0057] Optionally, the image detection model can also include a CNN classification network. In practical applications, after obtaining Dataset 1 mentioned above, the GPR images in Dataset 1 can be cropped into smaller images. These smaller images are then classified into those with hyperbolic features and those without. The smaller images with hyperbolic features are then labeled with their corresponding pipeline materials, resulting in Dataset 4. The initial CNN classification network is then trained on Dataset 4 until the corresponding loss function converges, thus obtaining the CNN classification network.

[0058] Furthermore, the scanned GPR images are input into a trained CNN classification network, which can classify GPR images into those with hyperbolic features, those without hyperbolic features, and identify the material corresponding to the underground pipeline.

[0059] Furthermore, a sliding window exhaustive search process is performed on the GPR image with hyperbolic features to separate the hyperbolic features from the GPR image, thus obtaining the hyperbolic features corresponding to the underground pipelines. For example, ... Figure 6 As shown, the sliding window is set to a small square, with a side length that is 1 / 10 of the smaller value between the image width (pixels) and height (pixels). Then, a scanning is performed based on a sliding window. After the first scan is completed, a second scan is performed with the position shifted up and down by n pixels from the position of the first scan. n is generally taken as 1 / 2 of the side length m of the small window. This process is repeated until the GPR image containing hyperbolic features is scanned, thus obtaining the hyperbolic features corresponding to the underground pipeline.

[0060] Furthermore, the separated hyperbolic features are sequentially processed through skeletonization, skeleton pruning, and hyperbolic fitting algorithms to obtain the hyperbolic equation. Then, the burial depth of the underground pipeline is obtained based on the vertex coordinates of the obtained hyperbolic equation. The specific implementation methods for processing the obtained hyperbolic features through skeletonization, skeleton pruning, and hyperbolic fitting algorithms to obtain the hyperbolic equation, and for obtaining the burial depth of the underground pipeline based on the vertex coordinates of the obtained hyperbolic equation, are described above and will not be repeated here.

[0061] In an optional embodiment of this application, after obtaining the first type of parameter values ​​corresponding to the underground pipeline, the method further includes: The matching degree of the hyperbola equation is evaluated by using the asymptote weighted vertex error formula or the adaptive deformation similarity formula, and the matching result is obtained.

[0062] Optionally, in this embodiment, the material of the pipeline and the hyperbolic features of the pipeline are obtained through three different methods. Subsequently, image morphology processing, hyperbolic fitting and pipeline burial depth calculation can be performed based on the obtained material of the pipeline and the hyperbolic features of the pipeline. In practical applications, the hyperbolic feature images separated by these three methods can be combined to obtain more accurate recognition results.

[0063] Optionally, after fitting the hyperbola, the matching procedure between the fitted hyperbola and the real hyperbola can be evaluated based on the asymptote weighted vertex error (AWE) formula and the adaptive deformation similarity (ADS) formula.

[0064] The formula for the asymptote weighted vertex error is: in, This is the vertex position error. This refers to the asymptote slope error. , , , This represents the constant value in the hyperbola simulated during model training. , , , The values ​​of a, b, k, and h in the obtained hyperbola are determined because when the model training dataset is constructed, a hyperbola that matches the features of the pipeline hyperbola in the GPR image is first drawn manually. Then, a hyperbola fitting algorithm is used to fit the manually drawn hyperbola. At this point, the specific values ​​of a, b, k, and h in the hyperbola equation will be obtained. The hyperbola obtained earlier is a skeleton with a shape close to the hyperbola obtained after the algorithm processes the image. The hyperbola equation obtained after fitting the skeleton by the algorithm mainly consists of four constants a, b, k, and h and two variables x and y. The fundamental purpose of performing asymptote weighted vertex error calculation is to ensure that the skeleton shape obtained after the algorithm processes the image can closely approximate the shape of the manually drawn hyperbola.

[0065] The formula for Adaptive Deformation Similarity (ADS) is: in, For vertical matching degree, The vertex Gaussian weights, For adaptive compensation of the opening. Values Values.

[0066] In this embodiment of the application, the AWE method is implemented through... Dynamically weighted vertex coordinate errors enhance vertex sensitivity, while the introduction of logarithmic error in the asymptote slope ratio helps capture the consistency of the opening direction. ADS can further strengthen vertex region error penalty through Gaussian terms and simultaneously add an opening compensation mechanism, thereby reducing evaluation errors caused by missing far-end data.

[0067] Step S105: Obtain the second type of parameter values ​​corresponding to the underground pipeline, and input the first type of parameter values ​​and the second type of parameter values ​​into the pipe diameter prediction model to obtain the pipe diameter value corresponding to the underground pipeline.

[0068] Optionally, the pipe diameter value corresponding to the underground pipeline usually needs to be inferred from the material of the underground pipeline, the burial depth, the hyperbolic equation fitted with the hyperbolic characteristics of the pipeline, the dielectric constant and conductivity of the buried medium, and the spacing of the radar transceiver antennas. Therefore, this embodiment of the application also needs to obtain the second type of parameter values ​​corresponding to the underground pipeline (i.e., the dielectric constant and conductivity of the buried medium and the spacing of the radar transceiver antennas), and then input the first type of parameter values ​​and the second type of parameter values ​​into the pipe diameter prediction model to obtain the pipe diameter value corresponding to the underground pipeline.

[0069] The dielectric constant and conductivity of the buried medium can be measured and calibrated on-site using existing geophysical techniques, and the spacing between the ground-penetrating radar transceiver antennas can be set and recorded before detection, so it can be obtained without interpreting GPR images.

[0070] Step S106: Determine the distribution results of underground pipelines in the area to be inspected based on the first type of parameter values, the second type of parameter values, the pipe diameter value, and GPS data corresponding to the underground pipelines.

[0071] In this embodiment, after the area to be detected is divided into grids, GPR scanning can be performed along the marked lines of the grid division. During scanning, the angle of the scanning direction is adjusted in real time to obtain the final direction of the underground pipeline, thus achieving automatic optimization control of the scanning path and ensuring the accuracy of the subsequently obtained underground pipeline parameter values. After scanning, the obtained raw radar data can be input into the image detection model and the pipe diameter prediction model to obtain the final parameter monitoring data. It is evident that the detection of underground pipelines in this embodiment utilizes intelligent image analysis, independent of human subjective experience, achieving remote, fully automated analysis of the detected images, improving detection and identification efficiency and accuracy. Simultaneously, it enables efficient and unified detection, identification, and management of wide-area municipal underground pipe networks, reducing human interference and information entry errors, thereby ensuring the correctness and consistency of the data source.

[0072] Optionally, the positions of the pipeline vertices in each direction can be marked based on the dielectric constant and conductivity of the buried medium, the spacing of the radar transceiver antennas, the material and burial depth of the underground pipeline, the hyperbolic equation fitted with the hyperbolic characteristics of the pipeline, the pipe diameter, and GPS information. The marked pipeline vertices are then connected in sequence, and the actual pipeline routing curve is generated by combining the GPS information, which is the result of the underground pipeline distribution in the area to be detected.

[0073] In this embodiment of the application, the method further includes: Determine the relative errors for underground pipelines separately. The relative errors include the relative error of burial depth and the relative error of pipe diameter. Obtain a preset engineering risk threshold, determine the number of relative errors that are less than the engineering risk threshold based on the engineering risk threshold, and determine the result evaluation result based on the number of relative errors that are less than the engineering risk threshold. The predicted pipeline and the actual pipeline are matched using the Hungarian algorithm to determine the optimal matching pipeline. Based on the burial depth, pipe diameter and preset relative error threshold of the optimal matching pipeline, the number of qualified matching indicators is determined, and the correlation accuracy assessment result is determined based on the number of qualified matching indicators.

[0074] Optionally, the accuracy of the obtained burial depth and pipe diameter values ​​is typically evaluated using relative error (RE). The relative error for pipe diameter is: in, Indicates the predicted pipe diameter. This is the actual pipe diameter; The relative error of burial depth is: in, Indicates the predicted burial depth. To determine the true burial depth, and in order to evaluate the accuracy of the model used to obtain the burial depth and pipe diameter, this application proposes an evaluation method based on Engineering Risk Sensitive Accuracy (ERA) and Multi-Objective Parameter Matching Rate (MPMR). The specific details are as follows: Engineering Risk Sensitive Accuracy (ERA): Optionally, in engineering practice, errors in model predictions may lead to certain risk consequences. Based on this, in this embodiment, the relative errors of burial depth and pipe diameter can be obtained using the formulas mentioned above. Then, the relative errors of pipe diameter and burial depth are compared using a risk sensitivity threshold defined by engineering specifications (this threshold characterizes the tolerance for relative errors in pipe diameter and burial depth). If the relative errors of pipe diameter and burial depth are both within the threshold, they are classified as low-risk samples; otherwise, they are classified as high-risk samples. Furthermore, the number of low-risk samples and the total number of samples are used to calculate the ERA (Earnings Reduction). The formula for calculating ERA is: in, Indicates the number of low-risk samples. This represents the total number of samples. In practical applications, a higher calculated ERA indicates that the model used in the application is more secure, usable, and reliable.

[0075] (2) Multi-objective parameter matching rate (i.e., MPMR) In practical applications, GPR images often contain multiple pipelines. In such cases, MPMR can be used to comprehensively evaluate the correlation accuracy in multi-object scenes. First, the Hungarian algorithm can be used to match the predicted pipelines with the real pipelines, and the set of predicted pipelines in the model is denoted as . The actual pipeline set is Then, the matching condition is defined as the positional offset distance between the predicted pipeline target box and the actual pipeline target box, and the optimal matching pipeline is found using the Hungarian algorithm. For this optimal matching pipeline, it is checked whether the detected burial depth and pipe diameter values ​​both meet the relative error threshold. If both are met, it is counted in the dual-index qualified matching count. The MPMR calculation formula is as follows: in, This indicates the number of qualified matches for both indicators. This represents the total number of matches. In practical applications, a higher MPMR indicates better overall model performance in multi-object scenarios and more accurate detection results.

[0076] Optionally, embodiments of this application also provide an underground pipeline distribution detection system based on ground-penetrating radar. This system includes an image processing module, a work trolley module, and a control platform module, wherein: The image processing module is used to receive the raw radar data and obtain the first type of parameter values, the second type of parameter values, and the pipe diameter value corresponding to the underground pipeline based on the raw radar data. According to the first type of parameter values, the second type of parameter values, and GPS data corresponding to the underground pipeline, the distribution result of the underground pipeline in the area to be detected is determined. The first type of parameter values ​​include the material, burial depth, and hyperbolic equation fitted by the hyperbolic feature of the underground pipeline. The raw radar data includes at least one GPR image and GPS data.

[0077] The work vehicle module includes a GPR device and a drive platform. The GPR device is mounted on the drive platform via a keyway and relies on the drive platform to drive the scanning forward to obtain raw radar data. The GPR device includes GPS positioning, device information, and wireless transmission functions, and transmits the scanned raw radar data to the image processing module.

[0078] The control platform module is used to send control commands to the trolley module, monitor the operation status of the trolley module, and determine the underground pipeline distribution results based on the first type of parameter values, the second type of parameter values, the pipe diameter value, and the GPS data sent by the image processing module.

[0079] In practical applications, the image processing module can be deployed on a cloud server and communicate with the work vehicle module. The work vehicle module's drive platform uses tracked wheels, which effectively reduces the contact pressure between the vehicle and the ground, enhancing its ability to navigate complex terrains such as sand, snow, and mud, while ensuring strong climbing and obstacle-crossing capabilities even in harsh conditions. Simultaneously, the drive platform integrates wireless remote control, allowing operators to control the work vehicle in real time via a dedicated remote control or a remote control terminal (integrated into the management platform), achieving precise operation from a distance. It also features a one-button stop function to ensure rapid stopping in emergencies, improving safety. Furthermore, the drive platform employs a high-efficiency energy system, enhancing its range and enabling a longer driving distance on a single charge, meeting the needs of extended operation.

[0080] The control platform module simultaneously features operation monitoring, result display, recording, and control command transmission functions. It can receive path information (Information 2), equipment information (Information 3), and recognition result information (Information 1) from the image processing module, and automatically generate models, perform 3D visualization, and record the results. It also monitors the operation status of the trolley using the path information (Information 2) and equipment information (Information 3). The control platform is equipped with a remote control terminal and offers three modes: detection start / end point (automatically planning the optimal path based on the start / end point), detection time (automatically planning the optimal path if the detection time is met), and detection distance (automatically planning the optimal path if the detection distance is met). Operators can switch between these three modes on the control platform to send commands to the drive platform to control the trolley's detection operation. Furthermore, the control platform's user interface can provide one-stop service to pipeline inspection units, construction units, and pipeline operation units.

[0081] As can be seen, in this embodiment, three modules can work together to achieve intelligent, efficient, collaborative, and one-stop service that integrates detection image analysis, detection equipment driving, detection path planning, detection remote control, and detection management.

[0082] like Figure 7 As shown in the figure, this application embodiment also provides an underground pipeline distribution detection device based on ground penetrating radar. The device includes a grid division module 701, an initial orientation determination module 702, an initial orientation adjustment module 703, a parameter value determination module 704, and a distribution result determination module 705. Its characteristic is that it includes: The grid division module is used to obtain the preset grid parameters and perform orthogonal grid division on the area to be detected based on the preset grid parameters to obtain the grid detection area, which includes longitudinal survey line marks and transverse survey line marks. The initial orientation determination module is used to perform GPR orthogonal scanning on the grid detection area based on longitudinal and transverse survey line marks, determine the first and second vertices of the hyperbola, and determine the initial orientation of the underground pipeline based on the first and second vertices. The initial orientation adjustment module is used to determine the initial vertical scanning direction and adjust the initial orientation of the underground pipeline along the longitudinal and transverse survey line marks based on the initial vertical scanning direction to obtain the adjusted underground pipeline orientation. The initial vertical scanning direction is perpendicular to the initial orientation of the underground pipeline. The parameter value determination module is used to continue GPR scanning of the grid detection area based on the adjusted underground pipeline route until the grid detection area is scanned, obtain the original radar data corresponding to the section to be detected, and input the original radar data into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline. The first type of parameter values ​​include the material, burial depth, and hyperbolic equation fitted by the hyperbolic feature of the underground pipeline. The original radar data includes at least one GPR image and GPS data. The module also obtains the second type of parameter values ​​corresponding to the underground pipeline and inputs the first and second type of parameter values ​​into the pipe diameter prediction model to obtain the pipe diameter value corresponding to the underground pipeline. The distribution result determination module is used to determine the distribution results of underground pipelines in the area to be inspected based on the first type of parameter values, the second type of parameter values, the pipe diameter value, and GPS data corresponding to the underground pipelines.

[0083] Optionally, the initial orientation determination module, when performing GPR orthogonal scanning of the grid detection area based on longitudinal and transverse survey line marks to determine the first and second vertices of the hyperbola, is specifically used for: The GPR device is pushed at a constant speed along the longitudinal and transverse survey lines to perform GPR orthogonal scanning, thereby obtaining the signal corresponding to the longitudinal survey line and the hyperbolic signal corresponding to the transverse survey line. The hyperbola signal corresponding to the longitudinal survey line and the hyperbola signal corresponding to the transverse survey line are used to determine the survey line with the strongest signal, and the position of the strongest survey line is taken as the first vertex of the hyperbola. Based on the first vertex of the hyperbola, the GPR device is continuously pushed at a constant speed along the longitudinal survey line to perform GPR orthogonal scanning, thus obtaining the second vertex of the hyperbola.

[0084] Optionally, the initial routing determination module, when determining the initial routing of the underground pipeline based on the first vertex and the second vertex, is specifically used for: Connect the first vertex and the second vertex to obtain a connecting line, and determine the angle between the connecting line and the due north direction; The target scanning direction is determined based on the connection angle, and GPR orthogonal scanning is continued according to the target scanning direction until the vertices of the hyperbolas included in the GPR scan image meet the preset requirements, thus obtaining the initial direction of the underground pipeline. The target scanning direction is 90 degrees away from the connection angle. If the target scanning direction is offset, the target scanning direction is adjusted according to a preset step size, and GPR orthogonal scanning continues based on the adjusted target scanning direction.

[0085] Optionally, when the initial orientation adjustment module adjusts the initial orientation of the underground pipeline based on the initial vertical scanning direction along the longitudinal and transverse survey line marks to obtain the adjusted underground pipeline orientation, it is specifically used for: Based on the initial vertical scanning direction, the GPR device is driven at a constant speed along the longitudinal and transverse survey line marks to perform GPR orthogonal scanning and obtain GPR scan images; If the vertex position of the hyperbola included in the GPR scan image shifts, the initial direction of the underground pipeline is adjusted to obtain the adjusted underground pipeline direction.

[0086] Optionally, the image detection model includes a target detection network. The parameter value determination module, when inputting raw radar data into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, is specifically used for: The GPR image is input into the target detection network to obtain the predicted target bounding box of the underground pipeline based on its material and hyperbolic features; The predicted target bounding box is cropped in the GPR image to obtain a cropped image. The cropped image is then preprocessed to obtain a processed cropped image. The image preprocessing includes grayscale conversion, Gaussian denoising, contrast enhancement, adaptive binarization, opening operation denoising, and closing operation hole filling. The processed cropped image is subjected to hyperbola fitting to obtain the hyperbola equation. Hyperbola fitting includes extracting the maximum connected component, skeletonization, and skeleton pruning. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0087] Optionally, the image detection model includes an instance segmentation network. The parameter value determination module, when inputting raw radar data into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, is specifically used for: The GPR image is input into the instance segmentation network to obtain the material and hyperbolic features corresponding to the underground pipeline; The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0088] Optionally, the image detection model includes a CNN classification network. The parameter value determination module, when inputting raw radar data into the image detection model to obtain the first type of parameter values ​​corresponding to the underground pipeline, is specifically used for: The GPR images are input into a CNN classification network to obtain GPR images with hyperbolic features, GPR images without hyperbolic features, and the material corresponding to the underground pipeline. A sliding window exhaustive search process is performed on GPR images with hyperbolic features to obtain the hyperbolic features corresponding to underground pipelines. The hyperbola features are subjected to skeleton fitting to obtain the hyperbola equation. The skeleton fitting process includes skeletonization, skeleton pruning and hyperbola fitting algorithm. The burial depth of underground pipelines can be obtained from the coordinates of the vertex of the hyperbola equation.

[0089] Optionally, the device also includes a model evaluation module, specifically used for: After obtaining the first type of parameter values ​​corresponding to the underground pipeline, the matching degree of the hyperbola equation is evaluated by using the asymptote weighted vertex error formula or the adaptive deformation similarity formula to obtain the matching result.

[0090] Optionally, the model evaluation module is also used for: Determine the relative errors for underground pipelines separately. The relative errors include the relative error of burial depth and the relative error of pipe diameter. Obtain a preset engineering risk threshold, determine the number of relative errors less than the engineering risk threshold based on the engineering risk threshold, and determine the result evaluation result based on the number of relative errors less than the engineering risk threshold. The predicted pipeline and the actual pipeline are matched based on the Hungarian algorithm to determine the optimal matching pipeline. Based on the burial depth, pipe diameter and preset relative error threshold of the optimal matching pipeline, the number of qualified matching indicators is determined, and the correlation accuracy assessment result is determined based on the number of qualified matching indicators.

[0091] The underground pipeline distribution detection device based on ground penetrating radar in this embodiment can execute the underground pipeline distribution detection method based on ground penetrating radar shown in the embodiment of this application. The implementation principle is similar, and will not be described again here.

[0092] This application provides an electronic device, which includes a processor and a memory configured to store machine-readable instructions that, when executed by the processor, cause the processor to perform a ground-penetrating radar-based method for detecting the distribution of underground pipelines.

[0093] This application provides an electronic device, such as... Figure 8 As shown, Figure 8 The illustrated electronic device includes a processor 2001 and a memory 2003. The processor 2001 and the memory 2003 are connected, for example, via a bus 2002. Optionally, the electronic device 2000 may further include a transceiver 2004. It should be noted that in practical applications, the transceiver 2004 is not limited to one type, and the structure of this electronic device 2000 does not constitute a limitation on the embodiments of this application.

[0094] Processor 2001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 2001 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0095] Bus 2002 may include a pathway for transmitting information between the aforementioned components. Bus 2002 may be a PCI bus or an EISA bus, etc. Bus 2002 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0096] The memory 2003 may be ROM or other type of static storage device capable of storing static information and instructions, RAM or other type of dynamic storage device capable of storing information and instructions, or EEPROM, CD-ROM or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0097] The memory 2003 stores the application code that executes the scheme of this application, and its execution is controlled by the processor 2001. The processor 2001 executes the application code stored in the memory 2003 to implement... Figure 7 The embodiment shown illustrates the operation of an underground pipeline distribution detection device based on ground-penetrating radar.

[0098] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0099] The above description is only a partial embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting the distribution of underground pipelines based on ground penetrating radar, characterized by, The method comprises the following steps: acquiring grid preset parameters, and performing orthogonal grid division on a to-be-detected interval based on the grid preset parameters to obtain a grid detection area, the grid detection area comprising longitudinal measuring line marks and transverse measuring line marks; performing GPR orthogonal scanning on the grid detection area based on the longitudinal measuring line marks and the transverse measuring line marks, determining a first vertex and a second vertex of a hyperbolic curve, and determining an initial direction of an underground pipeline according to the first vertex and the second vertex; determining an initial vertical scanning direction, and adjusting the initial direction of the underground pipeline based on the initial vertical scanning direction along the longitudinal measuring line marks and the transverse measuring line marks to obtain an adjusted direction of the underground pipeline, the initial vertical scanning direction being perpendicular to the initial direction of the underground pipeline; continuing to perform GPR scanning on the grid detection area based on the adjusted direction of the underground pipeline until the scanning of the grid detection area is completed, obtaining radar raw data corresponding to the to-be-detected interval, and inputting the radar raw data into an image detection model to obtain a first type of parameter value corresponding to the underground pipeline, the first type of parameter value comprising the material, the buried depth and the hyperbolic curve equation fitted by the hyperbolic curve characteristics of the underground pipeline, the radar raw data comprising at least one GPR image and GPS data; acquiring a second type of parameter value corresponding to the underground pipeline, and inputting the first type of parameter value and the second type of parameter value into a pipe diameter prediction model to obtain a pipe diameter value corresponding to the underground pipeline; determining an underground pipeline distribution result of the to-be-detected interval according to the first type of parameter value, the second type of parameter value, the pipe diameter value and the GPS data corresponding to the underground pipeline.

2. The method of claim 1, wherein, The method comprises the following steps: uniformly pushing a GPR device along the longitudinal measuring line and the transverse measuring line respectively to perform GPR orthogonal scanning, obtaining a hyperbolic curve signal corresponding to the longitudinal measuring line and a hyperbolic curve signal corresponding to the transverse measuring line; determining a strongest measuring line based on the hyperbolic curve signal corresponding to the longitudinal measuring line and the hyperbolic curve signal corresponding to the transverse measuring line, and taking the position of the strongest measuring line as the first vertex of the hyperbolic curve; continuing to uniformly push the GPR device along the longitudinal measuring line based on the first vertex of the hyperbolic curve to perform GPR orthogonal scanning, obtaining the second vertex of the hyperbolic curve.

3. The method of claim 1, wherein, The method comprises the following steps: connecting the first vertex and the second vertex to obtain a connecting line, and determining a connecting included angle between the connecting line and the north direction; determining a target scanning direction according to the connecting included angle, and continuing to perform GPR orthogonal scanning according to the target scanning direction until the vertex of the hyperbolic curve included in the GPR scanning image meets a preset requirement, obtaining the initial direction of the underground pipeline; The target scanning direction is 90 degrees different from the connecting angle, and if the target scanning direction is offset, the target scanning direction is adjusted according to a preset step, and GPR orthogonal scanning is continued based on the adjusted target scanning direction.

4. The method of claim 1, wherein, The initial vertical scanning direction is used to adjust the initial orientation of the underground pipeline along the longitudinal survey line mark and the transverse survey line mark to obtain an adjusted underground pipeline orientation. The initial vertical scanning direction is used to push the GPR device at a constant speed along the longitudinal survey line mark and the transverse survey line mark to obtain a GPR scan image. If the vertex position of the hyperbola included in the GPR scan image is offset, the initial orientation of the underground pipeline is adjusted to obtain an adjusted underground pipeline orientation.

5. The method of claim 1, wherein, The image detection model includes a target detection network, and the radar raw data is input into the image detection model to obtain a first type of parameter value corresponding to the underground pipeline, including: The GPR image is input into the target detection network to obtain a predicted target box of the hyperbolic feature and the material of the underground pipeline; The predicted target box is intercepted in the GPR image to obtain a cropped image, and the cropped image is preprocessed to obtain a processed cropped image, and the image preprocessing includes grayscale processing, Gaussian denoising processing, contrast enhancement processing, adaptive binarization processing, open operation denoising processing, and closed operation hole filling processing; The processed cropped image is subjected to hyperbolic fitting processing to obtain a hyperbolic equation, and the hyperbolic fitting processing includes maximum connected domain extraction processing, skeletonization processing, and skeleton pruning processing; The vertex coordinates of the hyperbolic equation are used to obtain the burial depth of the underground pipeline.

6. The method of claim 1, wherein, The image detection model includes an instance segmentation network, and the radar raw data is input into the image detection model to obtain a first type of parameter value corresponding to the underground pipeline, including: The GPR image is input into the instance segmentation network to obtain the hyperbolic feature and the material of the underground pipeline; The hyperbolic feature is subjected to skeleton fitting processing to obtain a hyperbolic equation, and the skeleton fitting processing includes skeletonization processing, skeleton pruning processing, and hyperbolic fitting algorithm processing; The vertex coordinates of the hyperbolic equation are used to obtain the burial depth of the underground pipeline.

7. The method of claim 1, wherein, The image detection model includes a CNN classification network, and the radar raw data is input into the image detection model to obtain a first type of parameter value corresponding to the underground pipeline, including: The GPR image is input into the CNN classification network to obtain a GPR image with a hyperbolic feature, a GPR image without a hyperbolic feature, and a material corresponding to the underground pipeline; The GPR image with the hyperbolic feature is subjected to sliding window exhaustive search processing to obtain a hyperbolic feature corresponding to the underground pipeline; The hyperbolic feature is subjected to skeleton fitting processing to obtain a hyperbolic equation, and the skeleton fitting processing includes skeletonization processing, skeleton pruning processing, and hyperbolic fitting algorithm processing; The vertex coordinates of the hyperbolic equation are used to obtain the burial depth of the underground pipeline.

8. The method of claim 1, wherein, After the first type of parameter value corresponding to the underground pipeline is obtained, the method further comprises: The matching degree of the hyperbolic equation is evaluated by using an asymptote weighted vertex error formula or an adaptive deformation similarity formula to obtain a matching result.

9. The method of claim 1, wherein, The method further comprises: The relative errors corresponding to the underground pipeline are determined respectively, and the relative errors include a relative error of the buried depth and a relative error of the pipe diameter value; A preset engineering risk threshold is obtained, and the number of relative errors less than the engineering risk threshold is determined based on the engineering risk threshold, and a result evaluation result is determined based on the number of relative errors less than the engineering risk threshold; The predicted pipeline and the real pipeline are matched based on the Hungarian algorithm to determine an optimal matching pipeline, and the number of index qualified matches is determined based on the buried depth, the pipe diameter value of the optimal matching pipeline, and a preset relative error threshold, and an association accuracy evaluation result is determined based on the number of index qualified matches.

10. A ground penetrating radar based underground pipeline distribution detection system, characterized by, The system comprises an image processing module, a work trolley module, and a management and control platform module, and the underground pipeline distribution detection system based on the ground penetrating radar is used to execute the method of any one of claims 1-9, wherein: The image processing module is used to receive radar raw data, and obtain the first type of parameter value, the second type of parameter value, and the pipe diameter value corresponding to the underground pipeline based on the radar raw data, and determine the underground pipeline distribution result of the to-be-detected interval based on the first type of parameter value, the second type of parameter value, and GPS data of the underground pipeline, wherein the first type of parameter value includes the material, the buried depth, and the hyperbolic equation fitted by the hyperbolic curve characteristics of the underground pipeline, and the radar raw data includes at least one GPR image and GPS data; The work trolley module comprises a GPR device and a driving platform, the GPR device is installed on the driving platform through a key groove and is driven to scan and move forward by the driving platform to obtain the radar raw data, the GPR device comprises GPS positioning, device information, and wireless transmission functions, and transmits the scanned radar raw data to the image processing module; The management and control platform module is used to send a work trolley module control instruction to the work trolley module, monitor the work state of the work trolley module, and determine the underground pipeline distribution result based on the first type of parameter value, the second type of parameter value, the pipe diameter value, and the GPS data sent by the image processing module.

Images