A multi-source underground cavity ground penetrating radar automatic classification database construction method and system

Abstract

The application provides a multi-source underground empty area ground penetrating radar automatic classification database construction method and system, which comprises the following steps: acquiring actual engineering underground empty area disease engineering detection radar data; acquiring stratum parameters where underground empty area disease bodies are located; acquiring model test radar data; acquiring numerical simulation radar data; pre-processing the engineering detection radar data, the model test radar data and the numerical simulation radar data; training an automatic classification model and performing data classification; and based on the classification result, constructing an engineering detection-model test-numerical simulation ground penetrating radar comprehensive database. The application can construct a more accurate and complete large batch of underground empty area disease body radar atlas comprehensive database, can comprehensively collect and couple underground empty area radar atlases of various sources, and can provide sufficient data basis for underground empty area disease body identification and prevention.

Description

Technical Field

[0001] This invention relates to the field of underground airspace radar map database construction, and in particular to a method and system for constructing an automatic classification database of multi-source underground airspace ground-penetrating radar. Background Technology

[0002] Identification and treatment of underground void defects are crucial issues in engineering maintenance and safety management. Currently, underground void identification is limited by data acquisition constraints. Radar maps obtained from engineering surveys are relatively limited, and radar maps obtained from model tests are difficult to scale up for mass production, thus failing to meet the needs for accurate identification and analysis of underground void defects. Therefore, using numerical simulation to obtain underground void maps has become an important supplementary method. Furthermore, numerical simulation-generated radar maps offer advantages such as large-scale generation, low cost, and detailed study of the characteristics of underground void defects. Establishing a multi-source radar map database and system will help collect and store a large number of underground void maps from engineering surveys and void model maps from laboratory tests, supplemented by maps generated from large-scale numerical simulations. This will enable efficient detection, accurate classification, and effective management of underground void defects. This database construction method allows for better identification, analysis, and evaluation of underground void defects, providing more comprehensive and detailed data support for engineering maintenance and management, which is of great significance for engineering maintenance and safety. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for constructing an automatic classification database for multi-source underground airspace ground-penetrating radar, which solves the aforementioned problems in existing technologies and is implemented through the technical solutions described below.

[0004] According to a first aspect of the technical solution of the present invention, a method for constructing an automatic classification database of multi-source underground airspace ground-penetrating radar is provided, comprising the following steps:

[0005] Step S1 uses ground-penetrating radar equipment to acquire engineering detection radar data of underground void defects in actual engineering projects;

[0006] Step S2 is based on engineering detection radar data of underground void defects in actual engineering projects to determine the location of the underground void defects and to obtain samples from different strata for testing and detection to obtain the stratum parameters where the underground void defects are located.

[0007] Step S3 uses the geological parameters of the underground void disease body as the parameters of the underground void model test, builds the underground void test model, and uses ground penetrating radar equipment to obtain radar data of the model test.

[0008] Step S4 uses the stratigraphic parameters of the underground void disease body as the stratigraphic parameters of the numerical model to construct a numerical model simulating the underground void, integrates the software and calculates the constructed numerical model, and obtains numerical simulation radar data.

[0009] Step S5 involves preprocessing the data from the three sources: the engineering detection radar data, the model test radar data, and the numerical simulation radar data.

[0010] Step S6 divides the data from the three sources into training set, validation set and test set based on the data partitioning. Based on the partitioned data, an automatic classification model is trained to classify the data into three types.

[0011] Step S7 constructs a comprehensive database of ground-penetrating radar based on the classification results of the automatic classification model, encompassing engineering detection, model testing, and numerical simulation.

[0012] In addition to the aspects and any possible implementations described above, a further implementation is provided in which, in step S2, the geological parameters of the underground void disease body include geological type, geological thickness, geological dielectric constant, and geological conductivity.

[0013] In addition to the aspects and any possible implementations described above, a further implementation is provided in which, in step S1, the engineering detection radar data includes raw waveform data and radar spectra.

[0014] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the preprocessing in step S5 includes: data filtering, data labeling, and data partitioning.

[0015] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the data is marked using outline marking and tabular records.

[0016] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the data is divided into a training set, a validation set, and a test set in a ratio of 8:1:1.

[0017] In addition to the aspects described above and any possible implementation, a further implementation is provided in which, in step S6, the automatic classification model divides the data into three types: engineering exploration, model experiment, and numerical simulation.

[0018] As described above, and in accordance with any possible implementation, a further implementation is provided, wherein for the automatic classification model, the sample is defined as (x i ,y i ), where x i It is the input sample, y iThese are the corresponding labels, and the output of the model is defined as f(x). i Considering the three types of imbalanced samples, for sample x i The objective function of the automatic classification model is:

[0019]

[0020] Among them, w j The weights of category j are adaptively adjusted based on the performance of the automatic classification model during training. ij It is sample x i Does it belong to category j? If it belongs to category j, then y ij f is 1 if it is not 0 otherwise j (x i ) is the model for sample x i The predicted probability of belonging to category j.

[0021] In addition to the aspects and any possible implementations described above, a further implementation is provided in which, in step S4, the numerical model of the underground void includes:

[0022] Regular-shaped empty area models include circular, elliptical, triangular, rectangular, trapezoidal, and arbitrary polygonal empty area models;

[0023] The irregular void model includes the void shape collected from actual engineering underground void cases and the irregular void shape generated by the inversion model.

[0024] As described above, and in accordance with any possible implementation, a further implementation is provided, wherein the inversion model function is:

[0025] y = f(W·x + b)

[0026] Where f is the activation function, W is the weight matrix, b is the bias vector used to adjust the model's output to better fit the training data and adapt to the prediction task; x is the input vector, which is the actual collected mine detection data, including engineering detection radar data and model test radar data; y is the predicted cavity shape; · represents matrix multiplication.

[0027] In addition to the aspects and any possible implementations described above, a further implementation is provided in which the numerical model construction steps include:

[0028] Python-gprMax software integration;

[0029] Matplotlib functions are used to draw corresponding rectangles to define the size and range of the numerical model;

[0030] Establish graphical functions for regular and irregular void areas to construct numerical models of underground void areas with different sample characteristics;

[0031] An algorithm for randomly generating underground void models of different sizes, shapes, and burial depths is designed by combining graphical functions with a random number generator.

[0032] Batch generation of numerical models of underground void disease structures;

[0033] The integrated software is run using Python to calculate batch airspace model radar data and obtain ground-penetrating radar data for the target number of underground airspaces.

[0034] According to a second aspect of the technical solution of the present invention, an automatic classification database system for multi-source underground airspace ground penetrating radar is provided, comprising a data acquisition module, a data marking module, a data classification module and a data storage module;

[0035] The data acquisition module includes an engineering exploration data acquisition module, a model test data acquisition module, and a numerical simulation data acquisition module, which are used to collect data obtained from engineering exploration, model tests, and numerical simulations.

[0036] The data marking module uses contour marking and tabular recording to input underground airspace radar data information, including three types of information: airspace shape, burial depth and equivalent diameter.

[0037] The data classification module uses an automatic classification model to analyze and classify the acquired underground airspace radar data into three types: engineering detection data, model test data, and numerical simulation data.

[0038] The data storage module includes three storage tables: an engineering exploration data storage table, a model test data storage table, and a numerical simulation data storage table, which store the radar spectrum data that has been marked, recorded, and classified.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) The database integrates multi-source radar maps, indoor test maps, engineering detection maps and numerical simulation maps, which can provide richer and more diverse ground-penetrating radar data on road cavity defects;

[0041] (2) Multi-source data fusion can improve the accuracy of road cavity identification, which helps to accurately identify and treat road defects and improve the service life and safety of urban roads. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of a method and system for constructing an automatic classification database for multi-source underground airspace ground-penetrating radar. Detailed Implementation

[0043] To better understand the technical solution of this invention, the content of this invention includes, but is not limited to, the specific embodiments described below. Similar technologies and methods should be considered within the scope of protection of this invention. To make the technical problems to be solved, the technical solutions, and advantages of this invention clearer, a detailed description will be provided below in conjunction with the accompanying drawings and specific embodiments.

[0044] It should be understood that the embodiments described in this invention are merely some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without inventive effort are within the scope of protection of this invention.

[0045] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0046] The technical solution of this invention first provides a method for constructing an automatic classification database for multi-source underground airspace ground-penetrating radar, such as... Figure 1 As shown, the method for constructing an automatic classification database for multi-source underground airspace ground-penetrating radar includes the following steps:

[0047] Step S1 uses ground-penetrating radar equipment to acquire engineering detection radar data of underground void defects in actual engineering projects, and acquires waveform data and map data of underground voids.

[0048] Step S2 determines the location of the underground void disease based on engineering detection radar data of actual engineering underground void disease, and obtains samples from different strata for testing and detection to obtain the stratum parameters of the underground void disease, including stratum type, stratum thickness, stratum dielectric constant and stratum conductivity.

[0049] Step S3 uses the geological parameters of the underground void disease area as the parameters for the underground void model test, builds the underground void test model, and uses ground penetrating radar equipment to acquire radar data for the model test.

[0050] Step S4 uses the stratigraphic parameters of the underground void as the stratigraphic parameters of the numerical model to construct a numerical model simulating the underground void. The software is integrated and the constructed numerical model is calculated to obtain numerical simulation radar data.

[0051] The numerical model construction steps include:

[0052] Python-gprMax software integration;

[0053] Matplotlib functions are used to draw corresponding rectangles to define the size and range of the numerical model;

[0054] Establish graphical functions for regular and irregular void areas to construct numerical models of underground void areas with different sample characteristics;

[0055] An algorithm for randomly generating underground void models of different sizes, shapes, and burial depths is designed by combining graphical functions with a random number generator.

[0056] Batch generation of numerical models of underground void disease structures;

[0057] The integrated software is run using Python to calculate batch airspace model radar data and obtain ground-penetrating radar data for the target number of underground airspaces.

[0058] The numerical model of the underground void includes:

[0059] Regular-shaped empty area models include circular, elliptical, triangular, rectangular, trapezoidal, and arbitrary polygonal empty area models;

[0060] The irregular void model includes the void shape collected from actual engineering underground void cases, as well as the irregular void shape generated by combining the collected void shape with the inversion model.

[0061] The inversion model function is as follows:

[0062] y = f(W·x + b)

[0063] Where f is the activation function, W is the weight matrix, b is the bias vector, x is the input vector, and · represents matrix multiplication.

[0064] Step S5 involves preprocessing data from three sources: actual engineering radar data of underground airspace, indoor test radar data, and numerical simulation radar data. This includes data filtering, data labeling, and data segmentation.

[0065] Step S6 divides the data from the three sources into training set, validation set and test set based on the data partitioning. Based on the partitioned data, an automatic classification model is trained to classify the data into three types.

[0066] During training, different weights are set according to the number of samples to reduce errors caused by sample imbalance. The automatic classification model divides the data into three types: engineering exploration, model experiment and numerical simulation.

[0067] Specifically, for the automatic classification model, the sample is defined as (x i ,y i), where x i It is the input sample, y i These are the corresponding labels, and the output of the model is defined as f(x). i Considering the three types of imbalanced samples, for sample x i The objective function of the automatic classification model is:

[0068]

[0069] Among them, w j The weights of category j are adaptively adjusted based on the performance of the automatic classification model during training. ij It is sample x i Does it belong to category j? If it belongs to category j, then y ij f is 1 if it is not 0 otherwise j (x i ) is the model for sample x i The predicted probability of belonging to category j.

[0070] Step S7 constructs a comprehensive database of ground-penetrating radar based on the classification results of the automatic classification model, encompassing engineering detection, model testing, and numerical simulation.

[0071] The technical solution of this invention also provides an automatic classification database system for multi-source underground airspace ground-penetrating radar, including a data acquisition module, a data marking module, a data classification module and a data storage module;

[0072] The data acquisition module includes an engineering exploration data acquisition module, a model test data acquisition module, and a numerical simulation data acquisition module, which are used to collect data obtained from engineering exploration, model tests, and numerical simulations.

[0073] The data marking module uses contour marking and tabular recording to mark underground airspace radar data information, including three types of information: airspace shape, burial depth, and equivalent diameter.

[0074] The data classification module uses an automatic classification model to analyze and classify the acquired underground airspace radar data into three types: engineering detection data, model test data, and numerical simulation data.

[0075] The data storage module includes three storage tables: an engineering exploration data storage table, a model test data storage table, and a numerical simulation data storage table, which store the radar spectrum data that has been marked, recorded, and classified.

[0076] Example

[0077] Data Acquisition

[0078] Engineering detection data acquisition: Using ground-penetrating radar equipment to acquire radar data of underground void defects in actual engineering projects, including waveform data and map data of underground voids.

[0079] Model test data acquisition: Based on radar data of underground void defects in actual engineering projects, the location of underground void defects is determined, and samples from different strata are obtained for testing to obtain the stratum parameters where the underground void defects are located, including stratum type, stratum thickness, stratum dielectric constant and stratum conductivity. The stratum parameters where the underground void defects are located are used as the parameters for underground void model tests, an underground void test model is built, and ground penetrating radar equipment is used to acquire radar data for the model test.

[0080] Numerical simulation data acquisition: The stratigraphic parameters of the obtained underground void disease body are used as the stratigraphic parameters of the numerical model to construct a numerical model of the underground void. The integrated software is run in Python to calculate the radar data of the batch void model and obtain the ground penetrating radar data of the target number of underground voids.

[0081] b. Data tagging

[0082] The raw data was initially manually screened to remove unclear data. The radar data information of the underground airspace was marked by contour marking and tabular recording, including three types of information: airspace shape, burial depth and equivalent diameter. The data from the three sources were divided into training set, validation set and test set in a ratio of 8:1:1.

[0083] c. Data Classification

[0084] Based on the imbalance of samples, an automatic classification model is trained using the pre-defined training, validation, and test sets to classify the data into engineering exploration, model experiment, and numerical simulation categories.

[0085] d. Data storage

[0086] The categorized data is stored in the corresponding database tables, including: engineering exploration data storage table, model test data storage table, and numerical simulation data storage table, thus completing the construction of the integrated database of engineering exploration-model test-numerical simulation ground penetrating radar.

[0087] Therefore, the multi-source underground airspace ground-penetrating radar data includes:

[0088] Engineering detection data on underground void defects in actual engineering projects, collected using ground-penetrating radar equipment;

[0089] The indoor void test model is constructed using model test data collected by ground penetrating radar equipment. The void test model is constructed using the geological parameters of the underground void disease body obtained in actual engineering.

[0090] Numerical simulation data of the underground void numerical model obtained by ground-penetrating radar forward modeling using integrated numerical simulation software. The model parameters used in the numerical simulation are based on the stratum parameters of the underground void disease body obtained in actual engineering.

[0091] The ground-penetrating radar data for the underground airspace includes raw waveform data and radar maps.

[0092] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A method for constructing an automatic classification database for multi-source underground airspace ground-penetrating radar, characterized in that, Includes the following steps: Step S1: Obtain engineering detection radar data for underground void defects in actual engineering projects; Step S2: Based on the engineering detection radar data, obtain the geological parameters of the underground void disease body; Step S3: Use the geological parameters of the underground void diseased body as the parameters for the underground void model test, build the underground void test model, and obtain the radar data of the model test; Step S4: Use the stratigraphic parameters of the underground void as the stratigraphic parameters of the numerical model, construct a numerical model simulating the underground void, integrate the software and calculate the constructed numerical model, and obtain numerical simulation radar data. The numerical model of the underground void includes: Regular void models include circular, elliptical, triangular, rectangular, trapezoidal, and arbitrary polygonal void models; irregular void models include void shapes collected from actual engineering underground void cases and irregular void shapes generated using inversion models. The inversion model function is as follows: in, f It is an activation function. W It is a weight matrix. b It is a bias vector. x It is the input vector, and · represents matrix multiplication; The numerical model construction steps include: Python-gprMax software integration; Matplotlib functions are used to draw corresponding rectangles to define the size and range of the numerical model; Establish graphical functions for regular and irregular void areas to construct numerical models of underground void areas with different sample characteristics; An algorithm for randomly generating underground void models of different sizes, shapes, and burial depths is designed by combining graphical functions with a random number generator. Batch generation of numerical models of underground void disease structures; The integrated software is run using Python to calculate batch airspace model radar data and obtain ground-penetrating radar data of the target number of underground airspaces. Step S5: Preprocess the engineering detection radar data, the model test radar data, and the numerical simulation radar data; Step S6: Train an automatic classification model and classify the data. The automatic classification model divides the data into three types: engineering exploration, model experiment, and numerical simulation. Step S7: Based on the classification results, construct a comprehensive database of ground-penetrating radar, including engineering detection, model testing, and numerical simulation.

2. The method according to claim 1, characterized in that, In step S1, the engineering detection radar data includes raw waveform data and radar spectrum; in step S2, the geological parameters of the underground void disease body include geological type, geological thickness, geological dielectric constant and geological conductivity.

3. The method according to claim 1, characterized in that, In step S5, the preprocessing includes: data filtering, data labeling, and data partitioning.

4. The method according to claim 3, characterized in that, The data is marked using outline marking and tabular records.

5. The method according to claim 1, characterized in that, For the aforementioned automatic classification model, the sample is defined as... ,in It is the input sample. These are the corresponding labels, defining the model's output as... For the sample The objective function of the automatic classification model is: in, It is a category The weights are adaptively adjusted based on the performance of the automatic classification model during training. It is a sample Does it belong to a category? If it belongs to a category , It is 1 if it is true, otherwise it is 0. It is the model on the sample Category The predicted probability.

6. An automatic classification database system for multi-source underground airspace ground-penetrating radar, characterized in that, The system is based on the method according to any one of claims 1 to 5, and the system includes a data acquisition module, a data labeling module, a data classification module, and a data storage module; The data acquisition module includes an engineering detection radar data acquisition module, a model test radar data acquisition module, and a numerical simulation radar data acquisition module, which are used to collect and acquire engineering detection radar data, model test radar data, and numerical simulation radar data, respectively. The data marking module uses outline marking and tabular records to mark data; The data classification module uses an automatic classification model to analyze and classify data, which is divided into three types: engineering exploration data, model test data, and numerical simulation data. The data storage module includes an engineering detection radar data storage table, a model test radar data storage table, and a numerical simulation radar data storage table, which are used for data storage.

Images