A geological radar echo multi-scale analytic processing method based on Gabor transform

Abstract

The application discloses a geological radar echo multi-scale analytical processing method based on Gabor transformation, and utilizes adjustment of Gabor filter parameters to obtain feature enhancement images of different targets under different scales, matches the feature enhancement images with disease maps to identify targets in ground penetrating radar echo data. The application effectively solves the technical problem that the existing echo analytical technology of ground penetrating radar detection has high difficulty in identifying different disease targets and high subjective factors, filters radar echoes by adjusting different bandwidths, different amplitude type ratios, different wavelengths, different directions and other parameters, obtains a data enhancement method under different parameters, realizes identification of different feature targets, and effectively extracts features of different disease targets under different scales.

Description

Technical Field

[0001] This invention relates to the technical field of radar signal processing, and in particular to a multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform. Background Technology

[0002] In the past two decades, with the rapid development of my country's economy and society, China has ushered in a peak period of transportation infrastructure construction. Reports indicate that by the end of 2020, the total length of highways in my country had reached 5.19 million kilometers. The completion acceptance of highway structures such as roadbeds, pavements, bridges, culverts, and tunnels requires filling quality testing. Furthermore, considering that highway structures, as they age, are subjected to the weight of the land, traffic loads, rainwater erosion, and other external stimuli year-round, abnormal defects such as loosening, cracks, potholes, and collapses can occur, becoming significant hidden dangers to highway traffic safety.

[0003] Currently, ground-penetrating radar (GPR) technology is commonly used for detecting hidden defects in highway subgrades. GPR is characterized by high efficiency, continuous operation, speed, high resolution, and non-destructive testing, and has been widely applied in non-destructive testing of infrastructure such as highways, urban roads, and tunnels. It has also become a key technology for evaluating engineering quality. GPR detection technology is primarily based on B-scan scanning. The resulting scan images are typically obtained through multi-directional and multi-angle measurements of the same scene, and the target image is reconstructed based on the surrounding dielectric environment to determine the target's relative position. However, it cannot distinguish targets of different materials based on signal amplitude or hyperbolic peak values. Furthermore, the identification and interpretation of various hidden defects beneath the road surface, such as voids, loosening, and cavities, still present certain challenges and have not yet reached a fully practical stage. In practical applications, it is difficult to accurately analyze targets using only acquired B-scan images, resulting in high identification difficulty and a high degree of subjectivity. Summary of the Invention

[0004] To address the limitations and defects of existing technologies, this invention provides a multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform. The analytical processing method uses an analytical processing system, which includes a ground-penetrating radar data preprocessing platform, a Gabor transform multi-scale processing platform, a feature extraction platform, and an output platform.

[0005] The ground-penetrating radar data preprocessing platform includes an echo data zero-point correction module, an echo data direct wave removal module, and a bad channel data removal module. The echo data zero-point correction module corrects the zero-point position of the input ground-penetrating radar echoes to the corresponding position on the ground, enabling subsequent image processing for calculating target depth. The echo data processed by the echo data zero-point correction module is input to the echo data direct wave removal module to remove direct waves from the radar electromagnetic waves that travel directly from the transmitter to the receiver, facilitating subsequent data processing. The data output from the direct wave removal module is then input to the bad channel data removal module to remove bad channel data generated during acquisition due to jitter, facilitating subsequent data processing. The data after bad channel data removal is input to the Gabor transform multi-scale processing platform for multi-scale processing. The multi-scale processed data is then input to the feature extraction platform, which matches the multi-scale processed data with feature maps to identify the type of defects in the echo data to be detected. The output platform displays the matching and identification results.

[0006] The Gabber transform multi-scale processing platform includes a bandwidth setting module, an amplitude ratio setting module, a wavelength setting module, and a direction setting module. The Gabber transform multi-scale processing platform sets the bandwidth value, amplitude ratio, wavelength, and direction parameters of the Gabber filter for the data after removing bad sector data, and obtains local enhancement information of the data image in the feature selection direction.

[0007] The feature extraction platform includes a pipeline feature matching module, a void feature matching module, a loose feature matching module, and a water-rich feature matching module. The feature maps of the pipeline feature matching module, the void feature matching module, the loose feature matching module, and the water-rich feature matching module are preset highway distress maps. The features formed by the echo data to be detected through Gabor transform multi-scale filtering are matched and identified with the highway distress map. The distress type of the echo data to be detected is determined based on the matching and identification results.

[0008] The output platform displays a disease type identification interface for the echo data to be detected.

[0009] Optionally, the weak signal identification method based on phase difference is used to segment the echo data to be detected, and the Euclidean distance matching method based on feature center points is used to perform similarity matching with the pipeline feature map of the pipeline feature matching module, the void feature map of the void feature matching module, the loose feature map of the loose feature matching module, and the water-rich feature map of the water-rich feature matching module.

[0010] Optionally, the steps of multi-scale processing performed by the Gabor transform multi-scale processing platform include:

[0011] The number of target features M to be detected is estimated based on the preprocessed ground-penetrating radar echo image;

[0012] The bandwidth, amplitude ratio, wavelength, and direction of the Gabor filter are set to form a multi-scale parameter X1;

[0013] The preprocessed echo data is filtered under the multi-scale parameter X1 to obtain the enhanced echo data of the multi-scale parameter X1.

[0014] The bandwidth, amplitude ratio, wavelength, and direction of the Gabor filter are set to form a multi-scale parameter X2;

[0015] The preprocessed echo data is filtered under the multi-scale parameter X2 to obtain the enhanced echo data under the multi-scale parameter X2.

[0016] Continue setting the bandwidth, amplitude ratio, wavelength, and direction of the Gabor filter to form the multi-scale parameter X. n The preprocessed echo data is subjected to the multi-scale parameter X. n The filtering process is performed to obtain the multi-scale parameter X. n The enhanced echo data continues until n = M.

[0017] Optionally, the weak signal identification method for the phase difference includes:

[0018] The preprocessed echo data is converted into phase image data;

[0019] Lateral differential is performed on the phase image data to form a lateral phase difference, thereby highlighting weak target signals;

[0020] The preprocessed echo data is segmented based on prominent weak target signals.

[0021] Optionally, the feature center point Euclidean distance matching method includes:

[0022] Calculate the center position of the feature in the data after feature segmentation;

[0023] Calculate the data feature center positions of the pipeline feature map, the void feature map, the loose feature map, and the water-rich feature map, respectively;

[0024] Using the center position of the feature in the segmented data as the coincidence point with the center position of the data feature in the pipeline feature map, the void feature map, the loose feature map, and the water-rich feature map, the Euclidean distance between the segmented data and the map data is calculated respectively.

[0025] The average Euclidean distance is calculated by traversing the segmented data and all map data, and then normalized to obtain the average normalized Euclidean distance value.

[0026] If the average normalized Euclidean distance value is greater than 0.8, it is confirmed that the segmented data and the map data have similar features; if the average normalized Euclidean distance value is less than or equal to 0.8, it is confirmed that the segmented data and the map data have dissimilar features.

[0027] The present invention has the following beneficial effects:

[0028] This invention provides a multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform. The analytical processing method uses an analytical processing system, which includes a ground-penetrating radar data preprocessing platform, a Gabor transform multi-scale processing platform, a feature extraction platform, and an output platform. The ground-penetrating radar (GPR) data preprocessing platform consists of an echo data zero-point correction module, an echo data direct wave removal module, and a "bad channel" data removal module. It processes the acquired GPR echo data through these modules to form effective acquisition data. The Gabber transform multi-scale processing platform consists of a bandwidth setting module, an amplitude ratio setting module, a wavelength setting module, and a direction setting module. It processes the effective acquisition data through multi-scale settings and processing to form radar data with different target enhancements. The feature extraction platform consists of a pipeline feature matching module, a void feature matching module, a loose feature matching module, and a water-rich feature matching module. Typical radar echoes with features of prefabricated pipelines, voids, looseness, and water richness are input into the feature extraction platform to form a feature database. The radar echo data processed by Gabber transform multi-scale processing is matched with different features in the feature database to obtain the features of the target to be detected. By adjusting the set values ​​on the Gabber transform multi-scale processing platform, different target features are obtained. The output platform displays the detection and recognition interface for different target features. This invention employs a multi-scale analysis technique for ground-penetrating radar echoes based on Gabor transform. By adjusting parameters such as bandwidth, amplitude ratio, wavelength, and direction to filter radar echoes, data augmentation methods under different parameters are obtained, enabling the identification of targets with different characteristics. This solves the technical problems of high difficulty in target analysis and identification in existing ground-penetrating radar echoes. Attached Figure Description

[0029] Figure 1 This is a flowchart illustrating the multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform provided in Embodiment 1 of the present invention.

[0030] Figure 2 This is a schematic diagram of road echo data acquisition using a ground-coupled method of ground-penetrating radar, as provided in Embodiment 1 of the present invention.

[0031] Figure 3This is a schematic diagram of zero-point correction for road echo data acquired by ground-coupled ground-penetrating radar according to Embodiment 1 of the present invention.

[0032] Figure 4 This is a schematic diagram illustrating the removal of direct waves from road echo data acquired using the ground coupling method of ground-penetrating radar according to Embodiment 1 of the present invention.

[0033] Figure 5 The image shows the bandwidth, direction, wavelength, and angle of the Gabor filter provided in Embodiment 1 of the present invention, which are 1.15, 0.3, 1.3, and 90 degrees, respectively.

[0034] Figure 6 This is a schematic diagram of zero-point correction for road echo data acquired by the air coupling method of ground-penetrating radar according to Embodiment 1 of the present invention.

[0035] Figure 7 This is a schematic diagram illustrating the removal of direct waves from road echo data acquired using the air coupling method of ground-penetrating radar according to Embodiment 1 of the present invention.

[0036] Figure 8 The image shows the bandwidth, direction, wavelength, and angle of the Gabor filter provided in Embodiment 1 of the present invention, which are 4.5, 0.1, 0.9, and 180, respectively.

[0037] Figure 9 The image shows the bandwidth, direction, wavelength, and angle of the Gabor filter provided in Embodiment 1 of the present invention, which are 0.95, 0.1, 0.9, and 90 degrees, respectively.

[0038] The attached figures are labeled as follows: ① Non-surface echo of raw data acquisition; ② Direct wave echo of raw data acquisition; ③ Zero-point correction and removal of direct wave echo; ④ Pipeline characteristics; ⑤ Void characteristics; ⑥ Loose characteristics; ⑦ Crack characteristics. Detailed Implementation

[0039] To enable those skilled in the art to better understand the technical solution of the present invention, the following detailed description of the multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform provided by the present invention is given in conjunction with the accompanying drawings.

[0040] Example 1

[0041] This embodiment addresses the challenge of target analysis in ground-penetrating radar (GPR) echo data by combining a multi-scale signal detection method based on Gabor transform. By adjusting Gabor filter parameters, enhanced images of different targets at different scales are obtained, and then matched with the disease map to identify targets in the GPR echo data.

[0042] The purpose of this embodiment is to solve the technical problems of high difficulty in identifying different disease targets and high subjective factors in existing ground-penetrating radar (GPR) echo analysis technology. Therefore, it provides a multi-scale analytical processing method for GPR echoes based on Gabor transform. This method uses Gabor filter parameters to obtain feature-enhanced images of different targets at different scales, and then matches these images with disease maps to identify targets in the GPR echo data. The specific steps are as follows:

[0043] (1) Establish a ground-penetrating radar data preprocessing platform. This platform consists of an echo data zero-point correction module, an echo data direct wave removal module, and a "bad channel" removal module. The specific process is as follows:

[0044] (1-1) Input the collected ground radar echo data into the echo data zero-point correction module to locate the ground surface features in the radar echo at the echo starting point, which facilitates subsequent target depth calculation and display;

[0045] (1-2) Input the echo data processed by the echo data zero-point correction module into the echo data removal direct echo module to remove the direct echo of radar electromagnetic waves from the transmitter to the receiver.

[0046] (1-3) The output of the echo data removal module is the "removal of direct wave data" module, which removes "bad channel" data generated during the acquisition process due to jitter, etc.

[0047] (2) Establish a Gabor transform multi-scale processing platform. This platform consists of a bandwidth setting module, an amplitude ratio setting module, a wavelength setting module, and a direction setting module. The specific process is as follows:

[0048] (2-1) Estimate the number of target features M based on the preprocessed ground-penetrating radar echo image;

[0049] (2-2) Set the bandwidth, amplitude ratio, wavelength and direction of the Gabor filter to form the i-th multi-scale parameter;

[0050] (2-3) The preprocessed echo data is filtered under multi-scale parameter i to obtain enhanced echo data with multi-scale parameter i;

[0051] (2-4) Continue to set the bandwidth, amplitude ratio, wavelength and direction values ​​of the Gabor filter to form a multi-scale parameter i+1. Perform filtering processing on the pre-processed echo data under the multi-scale parameter i+1 to obtain enhanced echo data with multi-scale parameter i+1 until i+1 = M.

[0052] (3) Feature extraction platform, which consists of pipeline feature matching module, void feature matching module, loose feature matching module, and water-rich feature matching module. The specific process is as follows:

[0053] (3-1) Convert the preprocessed echo data into phase image data, perform lateral differential on the phase image data to form a lateral phase difference, which can highlight the weak target signal, and segment the preprocessed echo data according to the highlighted target signal.

[0054] (3-2) Calculate the center position of features in the feature segmentation data, as well as the center position of features in pipeline feature maps, void feature maps, loose feature maps, and water-rich feature maps;

[0055] (3-3) Using the location of the data center segmentation data as the coincidence point with the feature center locations of pipeline feature maps, void feature maps, loose feature maps, and water-rich feature maps, calculate the 2-norm of the segmented data and the map data respectively;

[0056] (3-4) Traverse the segmented data and calculate the average 2-norm of all the map data and perform normalization. If the average normalized 2-norm value is greater than the threshold ρ, the segmented data and the map data are considered to be similar in features; otherwise, they are not similar.

[0057] (4) The output platform outputs and displays the defect type identification interface of the echo data to be detected. Further, in step (2), the expression for the kernel function of the two-dimensional Gabor filter is as follows:

[0058]

[0059] Where u is the direction, v is the scale, z = (x, y) are the coordinates of each energy point in the echo data, and ||*|| represents the modulo operation on *. k u,v It is the center frequency of the filter, and k v =k max f v , Where k max Where f is the maximum frequency, and k is the space factor. u,v The response of the Gabor filter at different directions and scales is described by changing k. u,v The value of σ can be used to obtain a set of Gabor filter functions. σ is the radius of the Gabor filter kernel function, ||k u,v || 2 / σ 2 Used to compensate for the attenuation of the energy spectrum determined by the frequency, exp(-σ 2 / 2) is the DC component.

[0060] Furthermore, in step (3-1), the Hilbert transform method is used to obtain the phase image H(x,y) of the acquired echo data z, and the phase change of the phase image H(x,y) is calculated, that is, the phase image data is subjected to lateral difference. The echo data corresponding to the lateral differential phase image with a difference value greater than δ are segmented, as shown in the following expression:

[0061] (X,Y)={(x,y)D(x,y)>δ,x=1,...,P; y=1,...,Q}.

[0062] Furthermore, in step (3-3), the similarity measurement method between the segmented data and the spectral data is as follows: The target echo segmented data (X...) I ,Y I ) and spectral data (P J Q J By taking the difference between each element in the expression and calculating its L2 norm, we obtain ||(X) I ,Y I )-(P J Q J )||2, where X and Y represent the lateral distance and time depth coordinates of the target in the echo, respectively, I represents the I-th target, P and Q represent the lateral distance and time depth coordinates of the target in the map, respectively, and J represents the map of the J-th target.

[0063] Furthermore, in step (3-4), the method for determining the threshold ρ is as follows: the value of the threshold ρ is generally an empirical value of 0.65, 0.7, 0.75, or 0.8. If the test environment is complex and the data noise interference is large, a lower empirical value of 0.65 or 0.7 can be selected. If the test environment is good and the data noise interference is small, a higher empirical value of 0.75 or 0.8 can be selected.

[0064] Figure 1 This is a flowchart illustrating the multi-scale analytical processing method for ground-penetrating radar echoes based on Gabor transform provided in Embodiment 1 of the present invention. This embodiment develops a ground-penetrating radar data preprocessing platform to perform zero-point correction, direct wave removal, and "bad road" data elimination on the acquired raw ground-penetrating radar echo data. Then, a Gabor transform multi-scale processing platform is developed. By setting the bandwidth, amplitude ratio, wavelength, and direction parameters of the Gabor filter, local enhancement information of the data image along the feature selection direction is obtained. The locally enhanced echo image is input into the developed feature extraction platform, matched with typical highway defect maps, and finally, the defect targets are identified and output.

[0065] This embodiment analyzes ground-penetrating radar (GPR) echo data of internal road structures and damaged targets. GPR echoes of internal road structures and damaged targets are acquired via ground coupling. This echo data is preprocessed, and the original images are shown below. Figure 2 As shown, the image contains both direct waves and non-surface echoes. A zero-point correction algorithm removes the echoes between the radar and the ground surface, ensuring that the "0ns" time in the image corresponds to the actual ground surface. This facilitates subsequent target location and identification. The zero-point corrected image is shown below. Figure 3 As shown, the direct wave in the image is then removed using a direct wave removal algorithm, such as... Figure 4 As shown, the direct surface wave at the "0ns" time was removed.

[0066] By adjusting the bandwidth, direction, wavelength, and angle of the Gabor filter to 1.15, 0.3, 1.3, and 90 degrees respectively, this embodiment obtains the image under these parameters as shown below. Figure 5 As shown, by matching with pipeline feature maps, void feature maps, loose feature maps, and water-rich feature maps, features such as... Figure 5 The pipes marked in the Chinese box are identified as follows: Figure 5 The empty space marked by the triangle in the middle.

[0067] Ground-penetrating radar echoes of internal road structures and damaged targets were acquired using air coupling. These echo data were preprocessed, and the original images were obtained as follows: Figure 6 As shown, in addition to the direct wave, the image also contains electromagnetic wave echoes generated by air coupling, which are not ground surface echoes. A zero-point correction algorithm removes the echoes between the radar and the ground surface in the image, ensuring that the "0ns" time in the image corresponds to the actual ground surface. This facilitates subsequent target location and identification. The zero-point corrected image is shown below. Figure 7 As shown, the direct wave removal algorithm removes direct waves from the image, such as... Figure 8 As shown, the direct surface wave at the "0ns" time was removed.

[0068] By adjusting the bandwidth, direction, wavelength, and angle of the Gabor filter to 4.5, 0.1, 0.9, and 180 degrees respectively, this embodiment obtains the image under these parameters as shown below. Figure 8 As shown, by matching with pipeline feature maps, void feature maps, loose feature maps, and water-rich feature maps, features such as... Figure 8 The loosening disease is marked by the pentagonal box in the middle. By adjusting the bandwidth, direction, wavelength, and angle of the Gabor filter to 0.95, 0.1, 0.9, and 90 degrees respectively, this embodiment obtains the image under these parameters as shown below. Figure 9 As shown, by matching with pipeline feature maps, void feature maps, loose feature maps, and water-rich feature maps, features such as... Figure 9 The cracks marked by the oval frame in the middle are identified as follows: Figure 9 The hollowing disease is marked by the triangle in the middle.

[0069] This embodiment provides a multi-scale analytical processing method for ground-penetrating radar (GPR) echoes based on Gabor transform, comprising a GPR data preprocessing platform, a Gabor transform multi-scale processing platform, a feature extraction platform, and an output platform. This embodiment uses a GPR echo multi-scale analytical processing method based on Gabor transform, utilizing Gabor filter parameters to obtain feature-enhanced images of different targets at different scales, and then matching them with disease atlases to identify targets in the GPR echo data. This embodiment effectively solves the technical problems of high difficulty in identifying different disease targets and high subjective factors in existing GPR echo analysis techniques, effectively extracting features of different disease targets at different scales.

[0070] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A method for multi-scale analytical processing of geological radar echoes based on the Gabor transform, characterized in that, The analysis processing method uses an analysis processing system, which comprises a geological radar data preprocessing platform, a Gabor transform multi-scale processing platform, a feature extraction platform and an output platform; The geological radar data preprocessing platform comprises an echo data zero point correction module, an echo data direct wave removal module and a bad channel data removal module; the echo data zero point correction module is used to correct the zero point position corresponding to the ground of the input geological radar echo, so as to calculate the target depth position of the subsequent processing image; the echo data zero point correction module processed echo data is input to the echo data direct wave removal module, the direct wave of the radar electromagnetic wave directly reaching the receiver from the transmitter is removed, so as to facilitate subsequent data processing; the data output by the echo data direct wave removal module after the direct wave is removed is input to the bad channel data removal module, the bad channel data generated in the acquisition process due to jitter is removed, so as to facilitate subsequent data processing; The data after the bad channel data is removed is input to the Gabor transform multi-scale processing platform for multi-scale processing, and the data after the multi-scale processing is input to the feature extraction platform; the feature extraction platform matches the data after the multi-scale processing with a feature map to identify the disease type of the to-be-detected echo data; and the output platform displays the matching identification result; The Gabor transform multi-scale processing platform comprises a bandwidth value setting module, an amplitude type ratio setting module, a wavelength setting module and a direction setting module; the Gabor transform multi-scale processing platform sets the bandwidth value, amplitude type ratio, wavelength and direction parameters of the Gabor filter for the data after the bad channel data is removed, and obtains the local enhancement information of the data image in the feature selection direction; The feature extraction platform comprises a pipeline feature matching module, a void feature matching module, a loose feature matching module and a water-rich feature matching module; the feature map of the pipeline feature matching module, the void feature matching module, the loose feature matching module and the water-rich feature matching module is a pre-set highway disease map; the features formed by the Gabor transform multi-scale filtering of the to-be-detected echo data are matched with the highway disease map for identification; and the disease type of the to-be-detected echo data is determined according to the matching identification result; The output platform displays a disease type identification interface of the to-be-detected echo data; The multi-scale processing step of the Gabor transform multi-scale processing platform comprises: Estimating the number M of to-be-detected target features according to the preprocessed geological radar echo image; Setting the bandwidth value, amplitude type ratio, wavelength and direction value of the Gabor filter to form a multi-scale parameter X1; Filtering the preprocessed echo data under the multi-scale parameter X1 to obtain enhanced echo data of the multi-scale parameter X1; Setting the bandwidth value, amplitude type ratio, wavelength and direction value of the Gabor filter to form a multi-scale parameter X2; Filtering the preprocessed echo data under the multi-scale parameter X2 to obtain enhanced echo data of the multi-scale parameter X2; Setting the bandwidth value, amplitude type ratio, wavelength and direction value of the Gabor filter to form a multi-scale parameter X2; Filtering the preprocessed echo data under the multi-scale parameter X2 to obtain enhanced echo data of the multi-scale parameter X2; Continue to set the bandwidth value, the amplitude ratio, the wavelength and the direction value of the Gab filter, form the multi-scale parameter X n , and perform filtering processing on the preprocessed echo data under the multi-scale parameter X n , to obtain enhanced echo data of the multi-scale parameter X n , until n=M.

2. The Gabor transform-based multi-scale analytical processing method of geological radar echoes according to claim 1, characterized in that, The weak signal recognition method using phase difference is used to perform feature segmentation on the to-be-detected echo data, and a feature center point Euclidean distance matching method is used to perform similarity matching of the feature segmented data with a pipeline feature atlas of the pipeline feature matching module, a void feature atlas of the void feature matching module, a loose feature atlas of the loose feature matching module, and a water-rich feature atlas of the water-rich feature matching module.

3. The Gabor transform-based multi-scale analytical processing method of geological radar echoes according to claim 2, characterized in that, The weak signal recognition method using phase difference comprises: Converting the preprocessed echo data into phase image data; Performing transverse difference on the phase image data to form a transverse phase difference, so as to highlight a weak target signal; Segmenting the preprocessed echo data according to the highlighted weak target signal.

Images