Ground penetrating radar target identification method based on polarization decomposition three-dimensional template

Abstract

The invention relates to the technical field of ground penetrating radars, in particular to a ground penetrating radar target recognition method based on a polarization decomposition three-dimensional template, and mainly solves the technical problem that an existing underground target recognition method is low in classification accuracy of pipeline targets. According to the ground penetrating radar target identification method based on the polarization decomposition three-dimensional template, a parameter Hm used for representing the characteristics of a ball target is constructed, and the parameter Hm can effectively distinguish the ball target from other targets, so that the classification accuracy of pipeline targets can be improved; meanwhile, the three-dimensional template adopted by the invention formulates classification areas for four scattering mechanisms of underground Bragg surface scattering, dipole scattering, double-bounce scattering and anisotropic scattering by using three-dimensional vectors (H, alpha, Hm), and can be used for detecting underground pipelines.

Description

Technical Field

[0001] The present invention relates to the technical field of ground penetrating radar, and in particular to a ground penetrating radar target recognition method based on polarization decomposition three-dimensional template. Background Technique

[0002] Ground Penetrating Radar (GPR) is a prospecting technology that uses electromagnetic waves to characterize underground target information and is widely used in fields such as water conservancy construction, archaeological research, and military investigations. Compared with single-polarization ground penetrating radar, full-polarization ground penetrating radar can obtain more comprehensive underground target information and has better effects in underground target detection.

[0003] Polarization decomposition is a method for obtaining various attributes of targets from full-polarization data and classifying and identifying different targets, mainly including Pauli decomposition, Freeman decomposition, and H-Alpha decomposition, etc. Among them, H-Alpha decomposition can identify linear targets and plays an important role in underground pipeline detection.

[0004] H-Alpha decomposition is mainly used to classify four types of underground targets: spherical, pipeline, dihedral angle, and multi-branch. However, since the B-scan images and polarization decomposition parameters of pipeline targets and spherical targets are relatively similar, the confusion between spherical targets and pipeline targets in underground targets is the most serious. As a result, the existing underground target recognition methods based on H-Alpha decomposition have a low recognition rate for pipeline targets and cannot guarantee the classification accuracy of pipeline targets. Summary of the Invention

[0005] To overcome the technical defect of the low classification accuracy of pipeline targets existing in the existing underground target recognition methods, the present invention provides a ground penetrating radar target recognition method based on polarization decomposition three-dimensional template.

[0006] The ground penetrating radar target recognition method based on polarization decomposition three-dimensional template provided by the present invention includes the following steps:

[0007] S1. Collect full-polarization data of four types of targets: spherical, pipeline, dihedral angle, and multi-branch;

[0008] S2. Perform H-Alpha decomposition on the full-polarization data to obtain parameters H and α;

[0009] S3. Based on the difference between the full-polarization data of underground spherical targets and other targets, create a parameter Hm for characterizing the characteristics of spherical targets;

[0010] S4. Construct a three-dimensional vector (H, α, Hm) as a data set, and randomly divide the data set into a training set and a test set;

[0011] S5. Obtain the sample center of the target and the optimal separating surface to form a three-dimensional template;

[0012] S6. Perform target recognition through the three-dimensional template.

[0013] Optionally, in step S3, the calculation formula of Hm is as follows:

[0014] ;

[0015] ;

[0016] ;

[0017] where t represents the delay of the received signal relative to the transmitted signal, x represents the coordinate in the survey line direction, m represents the coordinate of the target center point in the survey line direction, σ represents the standard deviation in the Gaussian function, S represents the B-scan data received by the ground penetrating radar, and d represents the scaling factor for adjusting the correspondence between the input and output of the arctangent function.

[0018] Optionally, step S5 is divided into the following sub-steps:

[0019] S51. Use the particle swarm optimization algorithm to find the point with the minimum sum of distances between the spherical target and all three-dimensional samples as the sample center of the spherical target;

[0020] S52. Treat the other three types of targets as a whole, and use the particle swarm optimization algorithm to find the point with the minimum sum of distances between this whole and all three-dimensional samples as the sample center of this whole;

[0021] S53. Find the optimal separating surface perpendicular to the line connecting the sample center of the spherical target and the sample center of this whole;

[0022] S54. Define the area below the optimal separating surface as the spherical target area, and divide the area above the optimal separating surface into a pipeline target area, a dihedral angle target area, and a multi-branch target area according to the H-Alpha template area division rules.

[0023] Optionally, in step S53, the optimal separating surface is found according to the following formula:

[0024] ;

[0025] where and respectively represent the classification accuracies of the spherical target and other targets under a separating surface at a certain position;

[0026] When F takes the maximum value, the corresponding position is the position of the optimal separating surface.

[0027] Optionally, the value range of H is 0 - 1, the value range of α is 0 - 90, and the value range of Hm is 0 - 1;

[0028] The area above the optimal separation interface, where H is in the range of 0 - 0.5 and α is in the range of 0 - 47.5, is defined as the pipeline - type target area;

[0029] The area above the optimal separation interface, where H is in the range of 0 - 0.5 and α is in the range of 47.5 - 90, is defined as the dihedral - angle - type target area;

[0030] The area above the optimal separation interface, where H is in the range of 0.5 - 1 and α is in the range of 0 - 90, is defined as the multi - branch - type target area.

[0031] The technical solution provided by the present invention has the following advantages compared with the prior art:

[0032] The ground - penetrating radar target recognition method based on the polarization decomposition three - dimensional template provided by the present invention constructs a parameter Hm for characterizing the characteristics of spherical targets. The parameter Hm can effectively distinguish spherical targets from other targets, thereby improving the classification accuracy of pipeline - type targets; at the same time, the three - dimensional template adopted by the present invention uses three - dimensional vectors (H, α, Hm) to define classification regions for four scattering mechanisms: underground Bragg surface scattering, dipole scattering, double - bounce scattering, and anisotropic scattering, and can be used for the detection of underground pipelines. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present invention, and are used together with the specification to explain the principles of the present invention.

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0035] Figure 1 It represents the flowchart of the target recognition method in the embodiments of the present invention;

[0036] Figure 2 It represents the schematic diagram of the three - dimensional template in the embodiments of the present invention;

[0037] Figure 3 It represents the projection diagram of four types of targets on the three - dimensional template in the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the solution of the present invention will be further described below. It should be noted that, without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0039] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0040] The following combines Figures 1 to 3 to detail the specific embodiments of the present invention.

[0041] This embodiment provides a ground penetrating radar target recognition method based on a polarization decomposition three-dimensional template, including steps S1 to S6.

[0042] S1. Collect the full polarization data of four types of targets: spherical, pipeline, dihedral, and multi-branch.

[0043] It is easy to understand that the spherical target represents Bragg surface scattering, the pipeline target represents dipole scattering, the dihedral target represents double bounce scattering, and the multi-branch target represents anisotropic scattering.

[0044] Specifically, use a full polarization ground penetrating radar to collect the full polarization data of four types of targets: indoor metal balls, metal pipes, metal dihedrals, and tree-like multi-branches.

[0045] S2. Perform H-Alpha decomposition on the full polarization data to obtain parameters H and α.

[0046] It is easy to understand that H-Alpha decomposition is a polarization interference synthetic aperture radar data processing method. By decomposing the coherence matrix of the scattering matrix, parameters such as entropy H, average scattering angle α, and anisotropy A are extracted to characterize the scattering mechanism of ground objects.

[0047] S3. Based on the difference between the full polarization data of underground spherical targets and other targets, create a parameter Hm to characterize the characteristics of spherical targets.

[0048] It should be noted that the difference between the full polarization data of spherical targets and other targets is the data directly above the target. Above the ideal center point of the ball, the distance and shape of the upper surface of the ball are exactly the same when observed from the perspective of the horizontal survey line direction and the perspective of the vertical survey line direction. Therefore, under ideal conditions, the two co-polarization data at the center of the spherical target are the same. Moreover, since the gradient in all directions at the center point directly above the ball is 0, which is approximately a plane, the electric field will not produce a scattering component perpendicular to the transmission direction, so the horizontal receiving antenna will not receive the signal from the vertical transmitting antenna, and the cross-polarization data at this time is zero. After H-Alpha decomposition, the theoretical polarization entropy H at the center point directly above the spherical target is 0. The H value in the area near the center of the spherical target tends to 0, which is much smaller than the central H value of other targets.

[0049] The parameter Hm described in this embodiment is constructed based on the above differences, and its core principle is to use the Gaussian function to integrate the reciprocal of H along the measurement line. The specific formula is:

[0050] ;

[0051] ;

[0052] ;

[0053] Wherein, t represents the delay of the received signal relative to the transmitted signal, x represents the coordinate in the survey line direction, m represents the coordinate of the target center point in the survey line direction, σ represents the standard deviation in the Gaussian function, S represents the B-scan data received by the ground penetrating radar, and d represents the proportional factor for adjusting the corresponding relationship between the input and output of the inverse tangent function.

[0054] It should be noted that setting the lower limit of the integral interval to x can make the output parameters correspond to H and α one by one. After experimental measurement, the best selection of σ for shallow underground targets is 0.03. Taking the reciprocal of the H value can make the smaller H value at the center of the ball target become the dominant factor in the integration process. Finally, if the detection target is a ball, the output parameter Hmc tends to +∞, otherwise Hmc is a small positive number.

[0055] It is easy to understand that Hmc is an intermediate quantity when obtaining Hm. In order to make the parameter Hmc more effectively combine H and α to identify underground targets and construct a three-dimensional template, this embodiment nonlinearly normalizes Hmc to the interval [0,1] to convert it into Hm.

[0056] In this embodiment, through experimental measurement, the value of H at the center of the ball target is less than 0.2, while that of other types of targets is greater than 0.2. Therefore, the parameter value d with H value of 0.2 and Hm value of 0.5 is calculated, and d = 0.625. Thus, the boundary of the three-dimensional template can be determined. The value range of H is 0 - 1, the value range of α is 0 - 90, and the value range of Hm is 0 - 1.

[0057] S4. Construct a three-dimensional vector (H, α, Hm) as the data set, and randomly divide the data set into a training set and a test set.

[0058] Specifically, the data set is divided into a training set and a test set according to a ratio of 7:3, that is, the training set accounts for 70% of the data set, and the test set accounts for 30% of the data set.

[0059] S5. Obtain the sample center of the target and the optimal separation interface to form a three-dimensional template.

[0060] Specifically, step S5 is divided into the following sub-steps: S51. Use the particle swarm optimization algorithm to find the point with the minimum sum of distances between the ball target and all three-dimensional samples as the sample center of the ball target; S52. Regard the other three types of targets as a whole, and use the particle swarm optimization algorithm to find the point with the minimum sum of distances between this whole and all three-dimensional samples as the sample center of this whole; S53. Find the optimal separation interface perpendicular to the line connecting the sample center of the ball target and the sample center of this whole; S54. Define the area below the optimal separation interface as the ball target area, and divide the area above the optimal separation interface into a pipeline target area, a dihedral angle target area, and a multi-branch target area according to the H-Alpha template area division rule.

[0061] More specifically, in step S53, the optimal separation interface is found according to the following formula:

[0062] ;

[0063] where and respectively represent the classification accuracies of the ball target and other targets under the separation interface at a certain position;

[0064] When F takes the maximum value, the corresponding position is the position of the optimal separation interface.

[0065] More specifically, in the area above the optimal separation interface and within the range of H from 0 to 0.5 and α from 0 to 47.5, it is defined as the pipeline target area;

[0066] In the area above the optimal separation interface and within the range of H from 0 to 0.5 and α from 47.5 to 90, it is defined as the dihedral angle target area;

[0067] The area above the optimal interface, where H ranges from 0.5 to 1 and α ranges from 0 to 90, is defined as the multi-branch type target area.

[0068] As Figure 2 shown, the red part in the figure represents the ball type target area, the blue part represents the pipeline type target area, the green part represents the dihedral angle type target area, and the brown part represents the multi-branch type target area. The projection of the corresponding characteristic parameters of the four types of targets in the 3D template is as Figure 3 shown.

[0069] S6. Perform target recognition through the 3D template.

[0070] The ground penetrating radar target recognition method based on the polarization decomposition 3D template in this embodiment can be used for underground pipeline detection. Among underground targets, the confusion between ball type targets and pipeline targets is the most serious. For example, the B-scan images of underground pipelines and spherical cavities and the existing polarization decomposition parameters are relatively similar. This method can effectively distinguish the two through the introduction of Hm, and combining with the 3D template can improve the classification accuracy of pipelines.

[0071] The above are only specific implementation manners of the present invention, enabling those skilled in the art to understand or implement the present invention. Although described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments, and they should all be covered by the protection scope of the claims.

Claims

1. A ground penetrating radar target recognition method based on a three-dimensional polarization decomposition template, characterized in that It includes the following steps: S1. Collect the full-polarization data of four types of targets, namely spherical targets, pipeline targets, dihedral angle targets, and multi-branch targets; S2. Perform H-Alpha decomposition on the full-polarization data to obtain parameters H and α; S3. Based on the differences in the full-polarization data between underground spherical targets and other targets, create a parameter Hm to characterize the characteristics of spherical targets; S4. Construct a three-dimensional vector (H, α, Hm) as a data set, and randomly divide the data set into a training set and a test set; S5. Obtain the sample center and the optimal separation interface of the target to form a three-dimensional template; S6. Perform target recognition through the three-dimensional template.

2. The ground penetrating radar target recognition method based on a three-dimensional template of polarization decomposition according to claim 1, characterized in that, In step S3, the calculation formula of Hm is as follows: ； ； ； Where, t represents the delay of the received signal relative to the transmitted signal, x represents the coordinate in the survey line direction, m represents the coordinate of the target center point in the survey line direction, σ represents the standard deviation in the Gaussian function, S represents the B-scan data received by the ground penetrating radar, and d represents the scaling factor for adjusting the correspondence between the input and output of the arctangent function.

3. The ground penetrating radar target recognition method based on a three-dimensional template of polarization decomposition according to claim 1, characterized in that, Step S5 is divided into the following sub-steps: S51. Use the particle swarm optimization algorithm to find the point with the minimum sum of distances between the spherical target and all three-dimensional samples as the sample center of the spherical target; S52. Regard the other three types of targets as a whole, and use the particle swarm optimization algorithm to find the point with the minimum sum of distances between this whole and all three-dimensional samples as the sample center of this whole; S53. Find the optimal separation interface perpendicular to the line connecting the sample center of the spherical target and the sample center of this whole; S54. Define the area below the optimal separation interface as the spherical target area, and divide the area above the optimal separation interface into a pipeline target area, a dihedral angle target area, and a multi-branch target area according to the H-Alpha template area division rules.

4. The ground penetrating radar target recognition method based on a three-dimensional template of polarization decomposition according to claim 3, characterized in that, In step S53, find the optimal separation interface according to the following formula: ； Among them, and respectively represent the classification accuracies of ball targets and other targets under a certain position interface; When F takes the maximum value, the corresponding position is the position of the optimal separation interface.

5. The ground penetrating radar target recognition method based on a three-dimensional template of polarization decomposition according to claim 3, characterized in that The value range of H is 0 - 1, the value range of α is 0 - 90, and the value range of Hm is 0 - 1; The area above the optimal separation interface and within the range of H from 0 to 0.5 and α from 0 to 47.5 is defined as the pipeline target area; The area above the optimal separation interface and within the range of H from 0 to 0.5 and α from 47.5 to 90 is defined as the dihedral angle target area; The area above the optimal separation interface and within the range of H from 0.5 to 1 and α from 0 to 90 is defined as the multi-branch target area.

Images