Intelligent imaging method and device for deep underground cavity, and medium

Abstract

The invention relates to a deep underground cavity intelligent imaging method and device and a medium, and the method comprises the steps: reading the single-channel data or multi-channel data of a to-be-imaged cross-hole radar, calculating the first arrival travel time data of each antenna pair based on the single-channel data or multi-channel data, and carrying out the normalization processing; constructing a travel time two-dimensional image fingerprint based on the normalized first arrival travel time data; inputting the travel time two-dimensional image fingerprint into a trained UNET model, and outputting an underground cavity distribution image; and performing pixel transformation processing on the underground cavity distribution image to obtain an actual spatial distribution diagram of the deep underground cavity. The equipment and the medium are used for implementing the method. Compared with the prior art, the method has the advantages that the calculation cost for performing underground cavity intelligent imaging by using the UNET model is reduced, and the calculation efficiency and the imaging precision are improved.

Description

Technical Field

[0001] The invention relates to the field of deep underground target detection in civil engineering construction, and in particular to a deep underground cavity intelligent imaging method, equipment and medium. Background Art

[0002] In recent years, with the vigorous development of urban construction, the construction of large-scale urban transportation hubs, large-scale urban complexes and other projects has led to the development of foundation pit projects in a super-large, super-deep and grouped manner. The timely and accurate detection of deep underground cavities in the foundation pit area is becoming a problem that needs to be solved urgently. For underground targets with large burial depths and small scales, ground geophysical detection methods have great limitations. Cross-Borehole Ground Penetrating Radar (CB-GPR), as a geophysical detection technology that is not limited by the depth of the target body, can accurately detect the spatial distribution and properties of deep underground cavities in deep foundation pit areas.

[0003] Traditional cross-hole radar data imaging generally includes tomography methods based on travel time and attenuation, and inversion methods based on full waveform. However, these imaging inversion methods face great challenges in terms of stability, adaptability to engineering environments, and computational efficiency, and cannot meet the real-time, efficient, and accurate requirements for underground cavity detection imaging in engineering.

[0004] In recent years, artificial intelligence such as deep learning algorithms have made significant progress in the field of imaging. Deep learning shows unique advantages in characterizing nonlinear mapping relationships between data sets. The "U" type neural network (UNET) is a neural network that fuses data features at the same level through jump connections. It can more accurately segment image features and achieve "end-to-end" learning of images. It is theoretically feasible to use the UNET model to train cross-hole radar data and establish a nonlinear mapping relationship between detection data and cavity models to achieve target imaging. However, since the input source data of the UNET model is an image, there are many problems and limitations in directly using the cross-hole radar single-channel data set or multi-channel profile as the model input. For example, when constructing sample labels, it is necessary to use the finite-difference time-domain algorithm to forward calculate the multi-channel radar profile data of different models. For the deep underground cavity detection model, this requires a huge computing cost. Similarly, due to the large drilling depth (more than 50 meters), a multi-channel data profile may exceed tens of thousands of channels, and the input image size of a single sample exceeds 5MB, which will lead to slow training of the UNET model. In addition, the waveform features in the multi-channel radar profile are greatly affected by the surrounding environment, making feature extraction difficult.

[0005] Therefore, it is a problem that needs to be solved to provide an imaging method based on the UNET model that can both ensure imaging accuracy and reduce computing costs and efficiency. Summary of the invention

[0006] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a deep underground cavity intelligent imaging method, equipment and medium.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] According to a first aspect of the present invention, a method for intelligent imaging of deep underground cavities is provided, the method comprising the following steps:

[0009] Reading single-channel data or multi-channel data of the cross-hole radar to be imaged, calculating the first arrival travel time data of each antenna pair based on the single-channel data or multi-channel data, and performing normalization processing;

[0010] Construct a two-dimensional travel time image fingerprint based on the normalized first arrival travel time data;

[0011] Input the travel time two-dimensional image fingerprint into the trained UNET model, and output the underground cavity distribution image;

[0012] The underground cavity distribution image is processed by pixel transformation to obtain the actual spatial distribution map of deep underground cavities.

[0013] As a preferred technical solution, it is characterized in that the normalization processing is to convert the first arrival travel time data into a normalized value between [0,1].

[0014] As a preferred technical solution, the method for constructing a travel time two-dimensional image fingerprint is:

[0015] Numbering the transmitting antenna and the receiving antenna in the antenna pair respectively;

[0016] Using the point serial number of the transmitting antenna as the ordinate and the point serial number of the receiving antenna as the abscissa, a travel time two-dimensional image fingerprint coordinate system is constructed, wherein the coordinate points in the travel time two-dimensional image fingerprint coordinate system represent corresponding antenna pairs;

[0017] Determine the pixel color of the corresponding coordinate point in the two-dimensional travel time image fingerprint coordinate system based on the normalized first arrival travel time data of each antenna pair to form a first arrival travel time two-dimensional image;

[0018] The first arrival travel time two-dimensional image is scaled and a travel time two-dimensional image fingerprint of a preset specification is output.

[0019] As a preferred technical solution, the preset specification is determined based on the number of transmitting antennas and the number of receiving antennas in the paired antennas.

[0020] As a preferred technical solution, the pixel transformation processing method is:

[0021] Acquire actual cross-hole spacing and drilling depth based on the single-channel data or multi-channel data;

[0022] scaling the underground cavity distribution image in the horizontal direction based on the actual cross-hole spacing;

[0023] The underground cavity distribution image is scaled vertically based on the actual drilling depth.

[0024] As a preferred technical solution, the training process of the UNET model is:

[0025] Construct multiple formation electrical parameter models containing voids, and in each model, voids of different sizes and shapes are randomly distributed;

[0026] Calculating the model first arrival travel time data of each antenna pair in each formation electrical parameter model, and generating a travel time two-dimensional image fingerprint of the corresponding formation electrical parameter model based on the model first arrival travel time data;

[0027] The travel time two-dimensional image fingerprint is used as the training data set of the UNET model, and the training data set is divided into a training set and a test set according to a preset ratio, and the formation electrical parameter model corresponding to the travel time two-dimensional image fingerprint is used as a verification sample;

[0028] Use the training set to train the UNET model, and use the test set to test the UNET model trained by the training set;

[0029] If the underground cavity distribution image output by the UNET model during training and detection matches the corresponding verification sample to a preset degree after pixel transformation, the UNET model is saved; otherwise, training continues.

[0030] As a preferred technical solution, the method for calculating the model's first arrival travel time data is to adopt a rapid scanning method.

[0031] As a preferred technical solution, the preset ratio is 9:1, and the preset index is 95%.

[0032] According to a second aspect of the present invention, there is provided an electronic device, comprising a memory and a processor, wherein a computer program is stored in the memory, and the method described above is implemented when the processor executes the program.

[0033] According to a third aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, wherein the program implements the method described above when executed by a processor.

[0034] Compared with the prior art, the present invention has the following advantages:

[0035] 1) Convert the acquired image data into a two-dimensional image fingerprint of travel time, which can not only comprehensively cover all the first arrival travel time information but also greatly reduce the size of the image, solving the problem of high computational cost and low imaging accuracy caused by directly using the cross-aperture radar multi-channel waveform image as the input of the UNET model, and ensuring high computational efficiency and imaging accuracy.

[0036] 2) Using the travel time two-dimensional image fingerprint as the training data set of the UNET model can effectively solve the problem of slow training process when using cross-aperture radar multi-channel waveform images as training sets, and improve the training efficiency of the model.

[0037] 3) The fast scanning method is used to directly calculate the first arrival travel time data. There is no need to first perform forward modeling on the radar profile and then pick up the first arrival travel time data. This simplifies the process of obtaining the first arrival travel time data, shortens the time of constructing the travel time two-dimensional image fingerprint, and speeds up the entire method process. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flow chart of the method of the present invention;

[0039] Figure 2 Schematic diagram of the cavity numerical model of the present invention and its corresponding travel time two-dimensional image fingerprint; (2a) is a schematic diagram of the cavity numerical model; (2b) is a schematic diagram of the corresponding travel time two-dimensional image fingerprint;

[0040] Figure 3 3a) is a schematic diagram of the test model of the present invention; (3a) is a schematic diagram of the underground cavity model of the test model; (3b) is a schematic diagram of the travel time two-dimensional image fingerprint of the test model; (3c) is a predicted actual spatial distribution diagram of the test model;

[0041] Figure 4 4a) is a schematic diagram related to an example of the present invention; (4a) is a multi-channel radar waveform diagram obtained by actual measurement; (4b) is a schematic diagram of a travel time two-dimensional image fingerprint reconstructed based on the example; (4c) is an underground cavity distribution image output by the UNET model of the present invention; (4d) is the actual spatial distribution diagram of the predicted example. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0043] In order to solve the problem in the prior art that the UNET model working mode cannot guarantee the imaging accuracy and will incur a huge computational cost for imaging the deep underground cavity cross-hole radar detection data, this embodiment provides a deep underground cavity intelligent imaging method, the flow chart of which is as follows: Figure 1 As shown, the specific steps include:

[0044] S1. Data normalization processing:

[0045] S11. Reading single-channel data or multi-channel data of the cross-hole radar to be imaged.

[0046] S12. Based on the single-channel data or the multi-channel data, the first arrival travel time data of each antenna pair is calculated using a fast scanning method, and the first arrival travel time data is normalized to be converted into a normalized value between [0, 1].

[0047] S2. Constructing the two-dimensional image fingerprint of travel time:

[0048] S21. Number the transmitting antenna and the receiving antenna in the antenna pair respectively.

[0049] S22. Using the point serial number of the transmitting antenna as the ordinate and the point serial number of the receiving antenna as the abscissa, a travel time two-dimensional image fingerprint coordinate system is constructed, wherein the coordinate points in the travel time two-dimensional image fingerprint coordinate system represent corresponding antenna pairs.

[0050] S23, determining the pixel color of the corresponding coordinate point in the two-dimensional travel time image fingerprint coordinate system based on the normalized first-arrival travel time data of each antenna pair, and forming a first-arrival travel time two-dimensional image.

[0051] S24, scaling the first arrival travel time two-dimensional image, and outputting the travel time two-dimensional image fingerprint, the size of which is based on the number of transmitting antennas and receiving antennas in the paired antenna pair.

[0052] S3. Output underground cavity distribution image:

[0053] The travel time two-dimensional image fingerprint obtained from step S2 is input into the trained UNET model, and an underground cavity distribution image with a size of 200×200 is output.

[0054] S4. Generate the actual spatial distribution map of deep underground cavities:

[0055] S41. Acquire actual cross-hole spacing and drilling depth based on single-channel data or multi-channel data.

[0056] S42, scaling the underground cavity distribution image in the horizontal direction to the actual cross-hole spacing size.

[0057] S43, scaling the underground cavity distribution image in the vertical direction to the actual drilling depth.

[0058] The underground cavity distribution image after horizontal and vertical scaling is the actual spatial distribution map of deep underground cavities.

[0059] The training process of the above UNET model is:

[0060] A1. Construct 5000 images such as Figure 2 The electrical parameter model of the formation with voids shown in (2a) is shown in Figure 2. In each model, the size of the randomly distributed voids is 0.5. 2 ~2.0 2 m 2 and cavities that are inconsistent and irregular in shape.

[0061] A2. Calculate the model first arrival travel time data of each antenna pair in each formation electrical parameter model, and generate the travel time two-dimensional image fingerprint of the corresponding formation electrical parameter model based on the first arrival travel time data of the formation electrical parameter model. Taking the formation electrical parameter model constructed in Figure (2a) as an example, the generated travel time two-dimensional image fingerprint is shown in Figure (2b). In this scenario, the number of paired receiving antennas and transmitting antennas is 160, so the size of Figure (2b) is 160×160.

[0062] A3. 5000 travel time 2D image fingerprints of the formation electrical parameter model containing voids are used as the training data set of the UNET model, and the training data set is divided into a training set and a test set according to a ratio of 9:1. The formation electrical parameter model corresponding to the travel time 2D image fingerprint is used as a verification sample.

[0063] A4. Set the network epoch number to 500, the batch size to 16, the learning rate to 1.0e-3, and use the Adam gradient descent algorithm for training and testing.

[0064] A5. Use the training set to train the UNET model, and use the test set to test the UNET model trained by the training set.

[0065] Among them, select Figure 3(3a) in the figure is a sample example of the test set to show the model training result. The sample example is processed in step S2 of the above method to generate a travel time two-dimensional image fingerprint as shown in Figure (3b). In this scenario, the number of transmitting antennas in the paired antenna pairs is 160, and the number of receiving antennas is 150, so the size of Figure (3b) is 150×160. The travel time two-dimensional image fingerprint is input into the trained UNET model, and the output of the UNET model is pixel transformed to generate the actual spatial distribution map predicted by Figure (3c). Comparing Figure (3a) with Figure (3b), it can be seen that the UNET model performs well under the sample example of the test set.

[0066] A6. If the underground cavity distribution image output by the UNET model matches the corresponding verification sample by 95% after pixel transformation during training and detection, the UNET model is saved, otherwise training is continued.

[0067] In order to further verify the feasibility of the method provided by the present invention, cross-hole radar detection data of an underground cavity in a bridge foundation is selected and cavity imaging is performed according to steps S1 to S4 using the method of this embodiment.

[0068] Specifically, in this situation, we can obtain Figure 4 The multi-channel radar waveform shown in (4a) is processed to generate a two-dimensional travel time image fingerprint as shown in (4b), which is used as the input of the UNET model and outputs a 200×200 underground cavity distribution image as shown in (4c). In this scenario, the cross-hole distance is 15m and the drilling depth is 65m. According to the actual values ​​of the cross-hole distance and the drilling depth, the underground cavity distribution image is converted into a 15×65 image as shown in (4d), which is the actual cavity distribution map.

[0069] The electronic device of the present invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) or loaded from a storage unit into a random access memory (RAM). In the RAM, various programs and data required for device operation can also be stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0070] Multiple components in the device are connected to the I / O interface, including: input units, such as keyboards, mice, etc.; output units, such as various types of displays, speakers, etc.; storage units, such as disks, optical disks, etc.; and communication units, such as network cards, modems, wireless communication transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunication networks.

[0071] The processing unit performs the various methods and processes described above, such as methods S1 to S4. For example, in some embodiments, methods S1 to S4 may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via a ROM and / or a communication unit. When the computer program is loaded into the RAM and executed by the CPU, one or more steps of methods S1 to S4 described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute methods S1 to S4 in any other appropriate manner (e.g., by means of firmware).

[0072] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0073] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0074] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0075] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A deep underground cavity intelligent imaging method, characterized in that: The method includes: Reading single-channel data or multi-channel data of the cross-hole radar to be imaged, calculating the first arrival travel time data of each antenna pair based on the single-channel data or multi-channel data, and performing normalization processing; Construct a two-dimensional travel time image fingerprint based on the normalized first arrival travel time data; Input the travel time two-dimensional image fingerprint into the trained UNET model, and output the underground cavity distribution image; The underground cavity distribution image is processed by pixel transformation to obtain the actual spatial distribution map of deep underground cavities.

2. The deep underground cavity intelligent imaging method according to claim 1, characterized in that: The normalization process is to convert the first arrival travel time data into a normalized value between [0, 1].

3. The deep underground cavity intelligent imaging method according to claim 1, characterized in that: The method for constructing the travel time two-dimensional image fingerprint is: Numbering the transmitting antenna and the receiving antenna in the antenna pair respectively; Using the point serial number of the transmitting antenna as the ordinate and the point serial number of the receiving antenna as the abscissa, a travel time two-dimensional image fingerprint coordinate system is constructed, wherein the coordinate points in the travel time two-dimensional image fingerprint coordinate system represent corresponding antenna pairs; Determine the pixel color of the corresponding coordinate point in the two-dimensional travel time image fingerprint coordinate system based on the normalized first arrival travel time data of each antenna pair to form a first arrival travel time two-dimensional image; The first arrival travel time two-dimensional image is proportionally transformed to output a travel time two-dimensional image fingerprint of a preset specification.

4. A deep underground cavity intelligent imaging method according to claim 3, characterized in that: The preset specification is determined based on the number of transmitting antennas and the number of receiving antennas in the paired antennas.

5. The deep underground cavity intelligent imaging method according to claim 1, characterized in that: The pixel transformation processing method is: Acquire actual cross-hole spacing and drilling depth based on the single-channel data or multi-channel data; scaling the underground cavity distribution image in the horizontal direction based on the actual cross-hole spacing; The underground cavity distribution image is scaled vertically based on the actual drilling depth.

6. The deep underground cavity intelligent imaging method according to claim 1, characterized in that: The training process of the UNET model is: Construct multiple formation electrical parameter models containing voids, and in each model, voids of different sizes and shapes are randomly distributed; Calculating the model first arrival travel time data of each antenna pair in each formation electrical parameter model, and generating a travel time two-dimensional image fingerprint of the corresponding formation electrical parameter model based on the model first arrival travel time data; The travel time two-dimensional image fingerprint is used as the training data set of the UNET model, and the training data set is divided into a training set and a test set according to a preset ratio. The formation electrical parameter model corresponding to the travel time two-dimensional image fingerprint is used as a verification sample; Use the training set to train the UNET model, and use the test set to test the UNET model trained by the training set; If the underground cavity distribution image output by the UNET model during training and detection matches the corresponding verification sample to a preset degree after pixel transformation, the UNET model is saved; otherwise, training continues.

7. A deep underground cavity intelligent imaging method according to claim 6, characterized in that: The preset ratio is 9:1, and the preset index is 95%.

8. The deep underground cavity intelligent imaging method according to claim 1, characterized in that: The method for calculating the first arrival travel time data of the model is to adopt a fast scanning method.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the program, the method according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images