CN-118759593-B - Method, system, equipment and medium for processing ground penetrating radar data of water delivery tunnel

CN118759593BCN 118759593 BCN118759593 BCN 118759593BCN-118759593-B

Abstract

The application discloses a method, a system, equipment and a medium for processing data of a ground penetrating radar of a water conveying tunnel, wherein the method comprises the steps of constructing a water conveying tunnel detection simulation scene and establishing an interference suppression network; the method and the device for simulating the structure disease and interference factors of the tunnel detection simulation scene can be used for obtaining a training data set, performing model training on an interference suppression network according to the training data set to obtain a trained target network, obtaining initial ground penetrating radar data, inputting the initial ground penetrating radar data into the target network to perform interference suppression processing, and outputting the target ground penetrating radar data.

Inventors

TAN CAI
YUAN MINGDAO
LV LI
SHI YONGSHENG
ZHANG XUHUI
BAO HEXIN
Peng Ruxuan
HUANG WENKANG
LIU YIJIE

Assignees

广东省水利水电科学研究院

Dates

Publication Date: 20260505
Application Date: 20240612

Claims (9)

1. The data processing method for the ground penetrating radar of the water conveying tunnel is characterized by comprising the following steps of: Constructing a water conveyance tunnel detection simulation scene, and establishing an interference suppression network; performing structural disease and interference factor simulation processing on the water conveyance tunnel detection simulation scene to obtain a training data set; Performing model training processing on the interference suppression network according to the training data set to obtain a trained target network; Acquiring initial ground penetrating radar data, inputting the initial ground penetrating radar data into the target network for interference suppression processing, and outputting to obtain target ground penetrating radar data; The structural disease and interference factor simulation processing is carried out on the water conveyance tunnel detection simulation scene, obtaining a training data set, comprising the steps of: Performing cavity simulation processing on the water conveyance tunnel detection simulation scene, determining a disease distribution area by generating two-dimensional Gaussian distribution sample data during modeling, generating a cavity at a sample data point to simulate a cavity or a loose area in the water conveyance tunnel, and obtaining reference data; carrying out heterogeneous treatment on a concrete structure of the water conveyance tunnel detection simulation scene, and carrying out random water addition simulation treatment on the surface of the concrete structure to obtain an interference model; Performing time domain finite difference simulation processing on the interference model to obtain ground penetrating radar simulation data; Carrying out channel deletion and high-frequency noise treatment on the ground penetrating radar analog data, randomly generating a disorder rate, then calculating the channel number to be disorder according to the total channel number and the disorder rate, randomly selecting the size of a disorder block, randomly selecting the starting position and the exchanged position of the disorder block according to the size of the disorder block, and simulating the random disorder of a signal channel to obtain interference data; the reference data and the interference data are determined as the training data set.
2. The method according to claim 1, wherein said establishing an interference suppression network comprises the steps of: constructing and obtaining a shallow layer feature extraction structure according to the first convolution module; Constructing and obtaining an encoder-decoder structure according to a preprocessing module, a sparse window transform module, a partition fusion downsampling module, a double-layer upsampling module and a jump connection module; constructing and obtaining a reconstruction structure according to the second convolution module; And carrying out sequential connection processing on the shallow characteristic extraction structure, the encoder-decoder structure and the reconstruction structure to obtain the interference suppression network.
3. The method according to claim 1, wherein the training data set includes reference data and interference data, and the model training process is performed on the interference suppression network according to the training data set to obtain a trained target network, and the method comprises the following steps: inputting the interference data into the interference suppression network, and outputting to obtain network recovery data; constructing and obtaining a network loss function according to the average absolute error, the peak signal-to-noise ratio and the structural similarity; performing difference calculation processing on the reference data and the network recovery data according to the network loss function to obtain a loss value; And updating the parameters of the interference suppression network according to the loss value to obtain the target network after training is completed.
4. The method according to claim 1, wherein the inputting the initial ground penetrating radar data into the target network for interference suppression processing, and outputting to obtain target ground penetrating radar data, includes the following steps: Inputting the initial ground penetrating radar data into the target network, the target network comprising a shallow feature extraction structure, an encoder-decoder structure, and a reconstruction structure; Performing preliminary extraction processing on the initial ground penetrating radar data through the shallow feature extraction structure to obtain initial image features; the initial image features are subjected to feature extraction processing through the encoder-decoder structure to obtain target image features, wherein the encoder-decoder structure comprises a preprocessing module, a sparse window transform module, a partition fusion downsampling module, a double-layer upsampling module and a jump connection module; and carrying out reconstruction processing on the target image features through the reconstruction structure to obtain target ground penetrating radar data.
5. The method according to claim 4, wherein the performing feature extraction processing on the initial image feature by the encoder-decoder structure to obtain a target image feature comprises the steps of: Partitioning and linear transformation processing is carried out on the initial image characteristics through the preprocessing module, so that first characteristics are obtained; performing image feature extraction processing on the first features through the sparse window transform module to obtain second features; Performing layer-by-layer downsampling processing on the second feature through the partition fusion downsampling module to obtain a third feature; transmitting the third feature to the double-layer upsampling module through the skip connection module; and carrying out interpolation and convolution processing on the third feature through the double-layer upsampling module to obtain a processing result, and then carrying out upsampling processing on the processing result to obtain the target image feature.
6. The method according to claim 5, wherein the sparse window transform module includes a plurality of transform layers, and the image feature extraction processing is performed on the first feature by the sparse window transform module to obtain a second feature, and the method includes the steps of: performing feature segmentation processing on the first features according to a segmentation window to obtain window features; performing multi-head self-attention calculation processing on the window characteristics according to a sparse self-correlation algorithm to obtain a calculation result; And performing offset processing on the segmentation window, adding the calculation result and the first feature, updating the segmentation window and the first feature, returning to the feature segmentation processing on the first feature according to the segmentation window to obtain a window feature until the number of circulation times reaches the number of layers of the transducer layer, and determining the calculation result as the second feature.
7. A water tunnel ground penetrating radar data processing system, the system comprising: The first module is used for constructing a water delivery tunnel detection simulation scene and establishing an interference suppression network; The second module is used for carrying out structural disease and interference factor simulation on the water delivery tunnel detection simulation scene to obtain a training data set; The third module is used for carrying out model training treatment on the interference suppression network according to the training data set to obtain a target network after training; the fourth module is used for acquiring initial ground penetrating radar data, inputting the initial ground penetrating radar data into the target network for interference suppression processing, and outputting the initial ground penetrating radar data to obtain target ground penetrating radar data; The second module is configured to perform structural disease and interference factor simulation processing on the water conveyance tunnel detection simulation scene, to obtain a training data set, and includes: Performing cavity simulation processing on the water conveyance tunnel detection simulation scene, determining a disease distribution area by generating two-dimensional Gaussian distribution sample data during modeling, generating a cavity at a sample data point to simulate a cavity or a loose area in the water conveyance tunnel, and obtaining reference data; carrying out heterogeneous treatment on a concrete structure of the water conveyance tunnel detection simulation scene, and carrying out random water addition simulation treatment on the surface of the concrete structure to obtain an interference model; Performing time domain finite difference simulation processing on the interference model to obtain ground penetrating radar simulation data; Carrying out channel deletion and high-frequency noise treatment on the ground penetrating radar analog data, randomly generating a disorder rate, then calculating the channel number to be disorder according to the total channel number and the disorder rate, randomly selecting the size of a disorder block, randomly selecting the starting position and the exchanged position of the disorder block according to the size of the disorder block, and simulating the random disorder of a signal channel to obtain interference data; the reference data and the interference data are determined as the training data set.
8. An electronic device comprising a memory storing a computer program and a processor implementing the method of any of claims 1 to 6 when the computer program is executed by the processor.
9. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 6.

Description

Method, system, equipment and medium for processing ground penetrating radar data of water delivery tunnel Technical Field The application relates to the technical field of signal processing, in particular to a method, a system, equipment and a medium for processing data of a ground penetrating radar of a water delivery tunnel. Background In the related art, along with the growth of the service time, cavities, cracks and void diseases can appear in the tunnel structure, and the use of the water delivery tunnel can be influenced by the diseases in the structure, so that the structural safety of the water delivery tunnel can be threatened even. Therefore, the ground penetrating radar is adopted to detect the water delivery tunnel structure, so that diseases in the tunnel are found in time and repaired. However, when the ground penetrating radar is adopted to detect the water delivery tunnel, various interferences often occur to influence the effectiveness of the ground penetrating radar data, so that the echo of diseases in the water delivery tunnel is interfered, and the diseases in the water delivery tunnel are difficult to accurately locate. In summary, the technical problems in the related art are to be improved. Disclosure of Invention The embodiment of the application mainly aims to provide a method, a system, equipment and a medium for processing data of a ground penetrating radar of a water delivery tunnel, which can improve the data quality. In order to achieve the above object, an aspect of an embodiment of the present application provides a method for processing data of a ground penetrating radar of a water conveyance tunnel, the method including: Constructing a water conveyance tunnel detection simulation scene, and establishing an interference suppression network; performing structural disease and interference factor simulation processing on the water conveyance tunnel detection simulation scene to obtain a training data set; Performing model training processing on the interference suppression network according to the training data set to obtain a trained target network; And acquiring initial ground penetrating radar data, inputting the initial ground penetrating radar data into the target network for interference suppression processing, and outputting to obtain target ground penetrating radar data. In some embodiments, the establishing an interference suppression network includes the steps of: constructing and obtaining a shallow layer feature extraction structure according to the first convolution module; Constructing and obtaining an encoder-decoder structure according to a preprocessing module, a sparse window transform module, a partition fusion downsampling module, a double-layer upsampling module and a jump connection module; constructing and obtaining a reconstruction structure according to the second convolution module; And carrying out sequential connection processing on the shallow characteristic extraction structure, the encoder-decoder structure and the reconstruction structure to obtain the interference suppression network. In some embodiments, the performing structural disease and interference factor simulation processing on the water conveyance tunnel detection simulation scene to obtain a training data set includes the following steps: carrying out cavity simulation processing on the water conveyance tunnel detection simulation scene to obtain reference data; carrying out heterogeneous treatment on a concrete structure of the water conveyance tunnel detection simulation scene, and carrying out random water addition simulation treatment on the surface of the concrete structure to obtain an interference model; Performing time domain finite difference simulation processing on the interference model to obtain ground penetrating radar simulation data; Performing channel deletion and high-frequency noise treatment on the ground penetrating radar simulation data to obtain interference data; the reference data and the interference data are determined as the training data set. In some embodiments, the training data set includes reference data and interference data, and the model training process is performed on the interference suppression network according to the training data set to obtain a trained target network, which includes the following steps: inputting the interference data into the interference suppression network, and outputting to obtain network recovery data; constructing and obtaining a network loss function according to the average absolute error, the peak signal-to-noise ratio and the structural similarity; performing difference calculation processing on the reference data and the network recovery data according to the network loss function to obtain a loss value; And updating the parameters of the interference suppression network according to the loss value to obtain the target network after training is completed. In some embodiments, the inputting the initial ground pen