A Ground Penetrating Radar Parameter Inversion Method Based on Graphical User Interface

Abstract

This invention belongs to the field of ground-penetrating radar (GPR) data interpretation technology, specifically a method for inverting GPR underground pipeline parameters based on a graphical user interface (GUI). The method includes designing a dedicated GUI, divided into: a display and interaction area for GPR B-Scan signals, threshold adjustment, input of basic GPR parameters, a start button, and display of fitting results. Through the GUI, key points of the hyperbolic features in the GPR B-Scan image are annotated. Combined with the input sampling time interval and lateral inter-channel distance of the GPR system, the burial depth and radius parameters of the underground pipeline are inverted, yielding the underground pipeline's burial depth and radius parameters. Experimental results show that this invention can correctly invert the burial depth and radius of underground pipelines using GPR, solving problems such as insufficient versatility, susceptibility to noise, high computational load, and lack of radius inversion results in traditional methods.

Description

Technical Field

[0001] This invention belongs to the field of ground-penetrating radar data interpretation technology, specifically relating to a ground-penetrating radar underground pipeline parameter inversion method. Background Technology

[0002] Ground-penetrating radar (GPR), as a non-destructive testing technology, offers advantages over other conventional underground detection methods, including faster detection speed, continuous detection process, high resolution, convenient and flexible operation, and lower cost, making it widely used in engineering surveying and detection. In underground pipeline inspection, a key application of GPR, underground tubular targets exhibit hyperbolic characteristics in B-Scan signals, based on the imaging principle of GPR. However, the lack of a direct correlation with the original pipeline shape makes accurate interpretation of GPR data difficult.

[0003] Traditional methods employ threshold segmentation, morphological processing, data point extraction, and hyperbola fitting, but they suffer from the following problems: the threshold for threshold segmentation depends on empirical coefficients and needs to be reset according to different scenarios; morphological processing cannot eliminate relevant residual noise, affecting the extraction of subsequent data points; traditional hyperbola extraction algorithms require fitting all data points, resulting in high computational cost; furthermore, existing research can only locate the target but does not invert the pipe radius. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a ground-penetrating radar method for retrieving underground pipeline parameters that is highly versatile, unaffected by noise, requires minimal computation, and can simultaneously obtain both the pipeline's burial depth and radius.

[0005] The ground-penetrating radar (GPR) underground pipeline parameter inversion method provided by this invention is based on graphical user interface (GUI) technology. The designed GUI annotates key points of target features in the GPR B-Scan image, and combines the input GPR system parameters (sampling time interval and lateral inter-channel distance) to obtain the burial depth and radius parameters of the underground pipeline through physical model fitting and inversion. The specific steps are as follows.

[0006] (a) Graphical User Interface

[0007] The designed dedicated graphical user interface, such as Figure 1 As shown, it includes five parts:

[0008] (1) The display and interaction area of ​​the ground-penetrating radar B-Scan signal, such as Figure 1As shown in rectangle 1 at the top left; this area can display the original B-Scan signal or the local B-Scan signal after filtering such as object detection; interactive, including common interactive functions built into Matplotlib: drag, zoom, undo, etc.; at the same time, you can use the right mouse button to mark key points on the image, the system records the coordinates of the mouse click, and a red dot is also displayed on the image to indicate the marked position;

[0009] (2) Threshold adjustment region, such as Figure 1 As shown in scroll bar area 2 on the right, it is used to adjust the threshold size for threshold segmentation; Figure 2 The results show the effects of B-Scan when different thresholds are selected. Since subsequent processing is not performed on the results after thresholding, the choice of threshold is relatively flexible. It is only necessary to ensure that the outline of the hyperbola can be clearly seen and that key points can be labeled.

[0010] (3) Input area for basic parameters of ground-penetrating radar, such as Figure 1 As shown in rectangular area 3 in the lower left of the middle, the parameters include: the relative permittivity of the underground medium, the sampling time interval of the ground penetrating radar system, and the lateral interchannel distance;

[0011] (4) Start button, such as Figure 1 As shown in the small rectangular area 4 in the lower right corner; press the Start button, and the program will fit a hyperbola and invert the parameters in the background based on the key point coordinates and the input parameters;

[0012] (5) The display area of ​​the fitting results, such as Figure 1 The rectangular area 5 in the lower right corner is used to display the target burial depth and radius parameters obtained from the fitting inversion.

[0013] (II) Physical Model Construction

[0014] Ground penetrating radar scenario such as Figure 3 As shown in the diagram, the red dot represents the transmitting and receiving antenna, and the blue circle represents the cross-section of the underground pipe. Electromagnetic waves propagate from the transmitting antenna to the pipe surface, are reflected, and return to the receiving antenna. Based on the electromagnetic wave propagation mechanism and the geometric relationship of the golden triangle, the two-way propagation time t of the electromagnetic wave is:

[0015]

[0016] In the formula, x is the lateral distance between the antenna and the center of the pipe, z is the burial depth of the pipe (the distance between the upper surface of the pipe and the ground), defined as the distance between the ground surface and the upper tangent of the underground pipe, r is the radius of the underground pipe, and v is the wave speed of electromagnetic waves propagating in the current medium.

[0017] (III) Key Point Coordinate Data Processing

[0018] Use the mouse to annotate key points on the interface. The pixel coordinates of the mouse position are displayed in the upper right corner, and the system records the coordinates each time the right mouse button is clicked. Before officially starting curve fitting, based on the physical model formula, the sampling time interval of the ground penetrating radar system parameters input to the graphical user interface, and the lateral interchannel distance, the obtained key point coordinates are mapped from pixel positions to their actual physical locations. The specific steps are as follows:

[0019] (1) Taking the top left corner of the image as the origin, the horizontal axis as x, and the vertical axis as y, find the point with the smallest y-value among all keypoint coordinates. This point corresponds to the vertex of the hyperbola. Assume the coordinates of this point are (x... k ,y k );

[0020] (2) Calculate the x-coordinates and x-coordinates of other key points. k The difference, multiplied by the lateral inter-channel distance of the ground-penetrating radar system, yields x. k The origin is x, and each key point is located at x in the physical model;

[0021] (3) Based on x and other parameters v, z and r obtained in step (2), t can be calculated according to different z and r during the gradient descent process.

[0022] (iv) Two-stage curve fitting and inversion parameters

[0023] Based on the physical model of ground-penetrating radar, the burial depth of the pipeline only affects the vertical position of the hyperbola in the ground-penetrating radar B-Scan image, and the pipeline radius only changes the expansion angle of the hyperbola. Therefore, this invention performs curve fitting in two stages, as follows:

[0024] (1) First, keep the radius unchanged and only invert the burial depth. The loss function used in this stage is the relative error function, i.e.

[0025]

[0026] In the formula, t0 is the propagation time corresponding to the marked hyperbola vertex, and t1 is the propagation time corresponding to the hyperbola vertex calculated using the inversion parameters and model formula.

[0027] (2) Keeping the burial depth constant, only the radius is inverted. The loss function used in this stage is the mean square error function, i.e.

[0028]

[0029] In the formula, t0 is the two-way propagation time sequence of the marked key points converted according to the electromagnetic wave propagation speed, t1 is the two-way propagation time calculated using inversion parameters and model formulas, and n represents the number of sampling channels along the track of the ground penetrating radar.

[0030] Pressing the Start button will cause the background system to continuously reduce the error of formula (2-3) using the gradient descent method according to the above strategy until the error reaches the given error requirement. The fitted and inverted target parameters will then be displayed. Figure 1 The two text boxes in area 5 of the middle section.

[0031] Experimental results show that the present invention can correctly invert the burial depth and radius of underground pipelines of ground penetrating radar, and can solve the problems of insufficient universality, susceptibility to noise, large computational load, and lack of radius inversion results of traditional methods. Attached Figure Description

[0032] Figure 1 The designed graphical user interface includes: signal display and interaction area 1, threshold adjustment area 2, radar parameter input area 3, start button 4, and fitting result display area 5.

[0033] Figure 2 The images are B-Scan images from actual measurements, where (a) is the original image, (b) is the image segmented with threshold = -10, and (c) is the image segmented with threshold = -20.

[0034] Figure 3 This is a schematic diagram of the physical model, where the red dots represent transmitting and receiving antennas, and the blue dots and circles represent the cross-sections of underground pipes.

[0035] Figure 4 To Figure 2 The diagram shows the key point annotation of the measured B-Scan image. The coordinates of the key points indicated by the red arrows in the image are displayed in the upper right corner of the interface.

[0036] Figure 5 This is a schematic diagram of the hyperbola fitting effect. The blue curve represents manually marked points, and the orange curve represents points fitted by gradient descent.

[0037] Figure 6 The three buried target models used in numerical simulation are, from top to bottom: (a) burial depth 0.3m, radius 0.1m; (b) burial depth 0.3m, radius 0.05m; (c) burial depth 0.5m, radius 0.1m.

[0038] Figure 7 These are illustrations of the three key point annotations for the hyperbola in simulation data, from left to right: the top edge, the midpoint, and the bottom edge.

[0039] Figure 8 This is a schematic diagram of the detection scenario using the measured data from TU1208.

[0040] Figure 9 for Figure 8The B-Scan image of the measured data for the scene shown is shown, with 1-4 in the figure being four hyperbolas used for parameter inversion.

[0041] Figure 10-13 They are respectively the corresponding Figure 9 The key point annotation effect diagram and the hyperbola fitting effect diagram of hyperbolas No. 1-4. Detailed Implementation

[0042] The present invention will be further described below through specific embodiments.

[0043] Example 1: Simulation Data Inversion

[0044] Numerical simulation models such as Figure 6 As shown, the scene size is 2m*1m, the spatial resolution is 0.001m, the simulation duration is 15ns, the relative permittivity of the underground medium is 6, and the excitation source is a Ricker wavelet with a center frequency of 1GHz. A cylindrical metal pipe is buried in each underground medium with the following burial depths and radii: (a) 0.3m depth, 0.1m radius; (b) 0.3m depth, 0.05m radius; (c) 0.5m depth, 0.1m radius. The three models are processed using the graphical user interface and algorithm steps mentioned above. Due to the time-varying nature of the excitation source signal, this invention tests the upper edge, midpoint, and lower edge of the hyperbola signal in the B-Scan image. The key points are marked as follows: Figure 7 As shown.

[0045] The inversion results are shown in Table 1. When the midpoint of the hyperbola is selected for annotation, the burial depth inversion results of the three models all have an offset of approximately 0.08m relative to the true value, and the radius inversion results all have an offset of approximately 0.05m relative to the true value. The offset in the burial depth inversion results is because the time corresponding to the hyperbola vertex includes the two-way propagation time of the electromagnetic wave and the time taken for the excitation source amplitude to gradually increase from 0 to its maximum value. When the excitation source remains unchanged, this offset is constant. The offset in the radius inversion results is determined by the distance between the sampling channels, and this offset is also constant when the antenna step remains unchanged. Therefore, the offset of one of the models can be used as a correction reference. In this implementation case, the offset obtained from the model with a burial depth of 0.3m and a radius of 0.1m is used as the reference offset to correct the other models. After correction, the relative error of the burial depth is only about 0.6%, and the relative error of the radius is only about 4.4%, that is, the inversion accuracy of both burial depth and radius reaches the millimeter level, verifying the feasibility of the method.

[0046] The table shows negative results for the radius inversion. Analysis of the detection model reveals that the expansion angle of the hyperbola changes monotonically as the radius changes from negative to positive. This indicates that the negative number is merely a numerical solution derived from the shape of the hyperbola and does not represent the actual physical result. Correcting the radius will yield the correct positive result.

[0047] Table 1 Simulation data inversion results

[0048]

[0049]

[0050] Example 2: Inversion of Measured Data

[0051] This example uses the GPR profile dataset TU1208, recorded at the full-scale geophysical test site in Nantes, France, for testing. The actual test scenario is as follows: Figure 8 As shown, the scene is filled with gneiss, and the relative permittivity of the entire scene is estimated to be 3. Three layers of pipes are buried on the left side, each containing three pipes: from left to right, an empty steel pipe, a PVC pipe filled with water, and an empty PVC pipe. A concrete pipe is also buried on the right side. The detected B-Scan signal is as follows... Figure 9 As shown, this example selects the hyperbolic signal with the strongest signal in each layer for inversion, where hyperbolas 1 to 3 represent the three layers of pipes respectively, and hyperbolas 4 represents the buried concrete pipe.

[0052] Key points were labeled and inverted for each of the four hyperbolas, as follows: Figure 10-13 As shown in Table 2, the inversion results of hyperbola No. 1 were used to correct the inversion depth, and the inversion results of hyperbola No. 4 were used to correct the inversion radius. The final fitting results are shown in Table 2. Since the pipe diameter in this measured data is between 8 cm and 10 cm, much smaller than the excitation source wavelength, it is difficult to invert the pipe radius from the shape of the hyperbola. Therefore, we only inverted the pipe depth. The results show that the relative errors of the depth inversion results are 11.3% and 1.4%, respectively, and the shapes of the four hyperbolas also fit well.

[0053] Table 2. Results of the inversion of measured data

[0054]

Claims

1. A ground penetrating radar underground pipeline parameter inversion method based on a graphical user interface, characterized by, Through the key point labeling of hyperbolic curve features in ground penetrating radar B-Scan image by a graphical user interface, combined with the input ground penetrating radar system parameters including sampling time interval and lateral trace interval, the buried depth and radius parameters of underground pipeline are obtained by physical model fitting inversion; the specific steps are as follows: (I) Design a special graphical user interface The special graphical user interface is divided into: ground penetrating radar B-Scan signal display and interaction area, threshold adjustment, ground penetrating radar basic parameter input column, start button, fitting result display; among them: (1) The ground penetrating radar B-Scan signal display and interaction area is used to display the original B-Scan signal or the local B-Scan signal after target detection screening processing; the interaction includes the common interaction functions of Matplotlib: drag, zoom in, undo; at the same time, the mouse right button is used to mark the key points on the picture, and the system records the coordinates of the mouse clicks, and the red dots are also displayed on the picture to mark the marked positions; (2) Threshold adjustment: adjust the threshold value size of threshold segmentation through the scroll bar to clearly see the hyperbolic curve contour, which is convenient for key point labeling; (3) The input column of ground penetrating radar basic parameters is used to input the basic parameters required for inversion: the relative permittivity of underground medium, the sampling time interval and the lateral trace interval of ground penetrating radar system; (4) Start button: press the start button, and the program fits the hyperbolic curve according to the key point coordinates and the input parameters in the background, and inverts the parameters; (5) Fitting result display: the parameter fitting result is displayed in the text box; (II) Build a physical model Modeling of ground penetrating radar detecting underground targets, that is, deducing the equation satisfied by related important parameters; electromagnetic wave propagates to the pipe surface and reflects, and the double-path propagation time t of electromagnetic wave is obtained according to the electromagnetic wave propagation excitation and the golden triangle geometric relationship: ， （1）； In the formula, x is the lateral distance between the antenna and the center of the pipeline, z is the buried depth of the pipeline, which is defined as the distance between the ground surface and the upper tangent surface of the underground pipeline, r is the radius of the underground pipeline, and v is the wave velocity of electromagnetic wave in the current medium; (III) Process key point coordinate data The key points are labeled on the interface using the mouse, and the pixel coordinates of the labeled points are displayed in the upper right corner. The system records the coordinates of each mouse right click, and then maps the labeled key point coordinates from the pixel position to the real physical position according to the physical model and the input radar system parameters sampling time interval and lateral trace interval; (IV) Two-stage curve fitting and inversion parameters The specific algorithm steps of two-stage curve fitting are as follows: (1) Inversion of the burial depth while keeping the radius unchanged; the loss function used in this stage is the relative error between the labeled vertex and the vertex of the fitted curve, as shown in formula (2), where is the propagation time corresponding to the labeled key point, is the propagation time corresponding to the hyperbolic vertex calculated using the inversion parameters and the model formula; ； (2) Keep the buried depth unchanged to invert the radius; the loss function used in this stage is the mean square error function, as shown in formula (3), wherein is the two-way travel time sequence converted by labeling the key points, is the two-way travel time sequence calculated using the inversion parameters and the model formula; ； Press the start button, and the background continuously optimizes according to the above two-stage fitting strategy to reduce the loss function of the two stages, and finally the hyperbolic curve fitting parameters are displayed in the rectangular box in the interface, and n represents the number of sampling traces of ground penetrating radar along the track.

2. The ground penetrating radar underground pipeline parameter inversion method according to claim 1, wherein the specific steps of processing key point coordinate data in step (III) are as follows: (1) take the upper left corner of the picture as the origin, the horizontal axis to the right as the axis, and the lower axis as the axis to establish a scene coordinate system, find the point with the smallest value in all key points , which is the vertex corresponding to the hyperbola;​ (2) Calculate the horizontal coordinate of other key points The difference between the horizontal coordinate of the key point and the horizontal coordinate of the origin of the physical model, multiplied by the lateral trace interval of the ground penetrating radar system, is the horizontal coordinate of the key point in the physical model. The difference between the horizontal coordinate of the key point and the horizontal coordinate of the origin of the physical model, multiplied by the lateral trace interval of the ground penetrating radar system, is the horizontal coordinate of the key point in the physical model.​​ (3) During the gradient descent process, according to the result obtained in step (2) and other parameters Using formula (1) to calculate .

Images