A method and system for monitoring underground obstacles based on geological radar

Abstract

The present invention discloses an underground obstacle monitoring method and system based on geological radar, and relates to the field of underground monitoring. An underground obstacle monitoring system based on geological radar includes: a local area feature acquisition module, an underground three-dimensional model acquisition module and an obstacle classification module. The present invention divides the target monitoring area into several local areas, and adopts a simulated annealing algorithm to determine the size of each local area. In each local area, by calculating the signal attenuation index and the approximate depth of the obstacle, a preliminary evaluation can be made on the existence of underground obstacles and the underground characteristics of the local area; this method not only improves the pertinence and efficiency of monitoring, but also provides reliable basic data for subsequent precise scanning and three-dimensional modeling through refined local area analysis, thereby significantly improving the accuracy and reliability of the entire monitoring process.

Description

Technical Field

[0001] The present invention relates to the field of underground monitoring, and in particular to an underground obstacle monitoring method and system based on geological radar. Background Art

[0002] With the acceleration of urbanization, the development and utilization of underground space has become increasingly important; the existence of underground obstacles poses a major challenge to the safe construction of underground projects; traditional underground obstacle detection methods, such as manual excavation and ground penetrating radar, have problems such as low efficiency, high cost, and insufficient accuracy; therefore, the development of an efficient, accurate, and low-cost underground obstacle monitoring technology has important practical significance and broad application prospects.

[0003] At present, the monitoring of underground obstacles mainly relies on ground-penetrating radar technology; this technology detects underground structures by emitting high-frequency electromagnetic waves and receiving reflected signals; however, ground-penetrating radar technology has the following limitations: limited resolution, making it difficult to accurately identify small or deeply buried obstacles; large environmental impact, and greatly affected by surface conditions, such as soil moisture and density, which can affect the propagation and reflection of radar waves.

[0004] In order to overcome the shortcomings of the existing technology, this technical solution proposes an underground obstacle monitoring method based on geological radar. Summary of the invention

[0005] The present invention divides the target monitoring area into several local areas and adopts a simulated annealing algorithm to determine the size of each local area. In each local area, by calculating the signal attenuation index and the approximate depth of the obstacle, a preliminary evaluation can be made on the existence of underground obstacles and the underground characteristics of the local area. This method not only improves the pertinence and efficiency of monitoring, but also provides reliable basic data for subsequent precise scanning and three-dimensional modeling through refined local area analysis, thereby significantly improving the accuracy and reliability of the entire monitoring process.

[0006] A method for monitoring underground obstacles based on geological radar, comprising:

[0007] The target area to be monitored is divided into several local areas in a grid shape, and the spacing of the grids is selected and obtained by the swarm optimization algorithm; the target area is preliminarily scanned using a geological radar according to a preliminary scanning route, and the preliminary scanning route includes several parallel and evenly spaced scanning paths covering the entire target area. Each local area is passed by a scanning path to obtain several low-frequency radar signal data;

[0008] For any local area, the low-frequency radar signal data passing through the local area is analyzed, the signal attenuation index of the local area is calculated, and the obstacle reference depth of the local area is calculated;

[0009] Based on the signal attenuation index and obstacle reference depth of each local area, the ideal radar transmission frequency corresponding to each local area is obtained; the ideal radar transmission frequency is used to use the geological radar to perform precise scanning along the preset route to obtain a number of precise radar signal data; all the obtained precise radar signal data are subjected to time-depth processing to obtain a number of two-dimensional underground profiles; all two-dimensional underground profiles are superimposed and registered to obtain the underground three-dimensional model of the target area;

[0010] For any obstacle, the local point cloud area of ​​the current obstacle is extracted from the underground three-dimensional model, the extracted local point cloud area is used as the input of the obstacle classification model, and the obstacle classification result is output; the obstacle information is obtained according to the underground three-dimensional model and the obstacle classification result.

[0011] Preferably, the signal attenuation index of the local area is calculated and obtained, and the specific operation is as follows:

[0012] For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area, and the starting point, center point and end point of the local scanning path are used as reference points; for any reference point, the strongest and first received reflection signal strength of the current reference point is extracted from the low-frequency radar signal data corresponding to the current local scanning path. , and record the propagation time t of the reflected signal, using the formula Calculate and obtain the one-way propagation distance d of the reflected signal, where c is the speed of light; is the dielectric constant corresponding to the current local area; then use the formula Calculate and obtain the single-point attenuation index a of the current reference point, where: is the initial signal strength emitted by the radar; after obtaining the single-point attenuation indexes of the three reference points, the average value of the single-point attenuation indexes of the three reference points is calculated, and the calculated average value is used as the signal attenuation index of the current local area.

[0013] Preferably, the obstacle reference depth of the local area is calculated and obtained, and the specific operation is as follows:

[0014] For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area, and the local radar signal data corresponding to the current local scanning path is intercepted in the low-frequency radar signal data; a signal strength threshold is set, and for any reflected signal in the local radar signal data that is higher than the signal strength threshold, the formula is used Calculate the one-way propagation distance of the current reflected signal , as the rough obstacle depth corresponding to the current reflection signal, where T represents the propagation time of the current reflection signal; after obtaining the rough obstacle depths of all reflection signals above the signal strength threshold in the local radar signal data, calculate the average value of all obtained rough obstacle depths to obtain the obstacle reference depth of the current local area.

[0015] Preferably, the ideal radar transmission frequency corresponding to each local area is obtained, and the specific operation is as follows:

[0016] Set the base transmit frequency , for any local area, using the formula Calculate and obtain the ideal radar transmission frequency f in the current local area, where D is the obstacle reference depth in the current local area, A is the signal attenuation index in the current local area, and They represent the adjustment coefficients of obstacle reference depth and signal attenuation index respectively; and The value of is obtained through genetic algorithm selection.

[0017] Preferably, the group optimization algorithm selected to obtain the grid spacing is a simulated annealing algorithm, and the specific operation is as follows:

[0018] Step 1: Randomly select an initial grid spacing as the current solution, set the initial temperature, maximum number of iterations and temperature attenuation coefficient; use the objective function Calculate the fitness value of the current solution, where L represents the length of the grid spacing; H represents the degree of difference of each local area when the current grid spacing is applied, and the degree of difference H is expressed by the sum of the standard deviation of the signal attenuation index of all local areas and the standard deviation of the obstacle reference depth; and are the weights of the grid spacing length and the degree of local area difference, respectively;

[0019] Step 2: Add or subtract a random value to the current solution to generate a neighborhood solution; use the objective function to calculate the fitness value of the neighborhood solution and compare it with the fitness value of the current solution; if the fitness value of the current solution is less than the fitness value of the neighborhood solution, update the current solution to the selected neighborhood solution; if the fitness value of the current solution is greater than the fitness value of the neighborhood solution, update the neighborhood solution according to the probability Decide whether to accept the neighborhood solution as the current solution; where represents the fitness value of the neighborhood solution, represents the fitness value of the current solution, Indicates the current temperature; using the formula Update the temperature, where is the temperature attenuation coefficient;

[0020] Step 3: Repeat step 2 until the maximum number of iterations is reached, and the current solution finally obtained is the optimal grid spacing.

[0021] Preferably, the obstacle classification model is established based on the PointNet model.

[0022] Preferably, the obstacle classification model is trained, and the specific operations are as follows:

[0023] Obtain several point cloud data samples with obstacle classification results marked; divide all point cloud data samples into training sets and test sets, use the training sets to train the obstacle classification model, and then input the test sets into the obstacle classification model to obtain the obstacle classification test results; set the training conditions, if the obstacle classification test results meet the training conditions, output the trained obstacle classification model; otherwise, continue to use the training set to train the obstacle classification model.

[0024] Preferably, an underground obstacle monitoring system based on geological radar comprises:

[0025] The local area feature acquisition module is used to divide the target area to be monitored into several local areas in a grid shape; use the geological radar to perform a preliminary scan on the target area according to the preliminary scanning route to obtain several low-frequency radar signal data; analyze the low-frequency radar signal data of all local areas, calculate the signal attenuation index of the local area, and calculate the obstacle reference depth of the local area;

[0026] The underground three-dimensional model acquisition module is used to obtain the ideal radar transmission frequency corresponding to each local area according to the signal attenuation index and obstacle reference depth of each local area; use the ideal radar transmission frequency to scan along the preset route using the geological radar to obtain a number of precise radar signal data; perform time-depth processing on all the acquired precise radar signal data to obtain a number of two-dimensional underground profiles; superimpose and align all two-dimensional underground profiles to obtain the underground three-dimensional model of the target area;

[0027] The obstacle classification module is used to extract the local point cloud area of ​​all obstacles from the underground three-dimensional model, use the extracted local point cloud area as the input of the obstacle classification model, and output the obstacle classification result; obtain obstacle information based on the underground three-dimensional model and the obstacle classification result.

[0028] The present invention has the following advantages:

[0029] The present invention divides the target monitoring area into several local areas and adopts a simulated annealing algorithm to determine the size of each local area. In each local area, by calculating the signal attenuation index and the approximate depth of the obstacle, a preliminary evaluation can be made on the existence of underground obstacles and the underground characteristics of the local area. This method not only improves the pertinence and efficiency of monitoring, but also provides reliable basic data for subsequent precise scanning and three-dimensional modeling through refined local area analysis, thereby significantly improving the accuracy and reliability of the entire monitoring process.

[0030] The present invention determines the ideal radar transmission frequency based on the preliminary scanning results, the signal attenuation index of the local area and the reference depth of the obstacle, performs precise scanning and obtains precise radar signal data; this step ensures that the radar signal can effectively detect underground obstacles, and improves the quality and availability of the data; using these precise radar signal data, time-depth processing is performed to generate a two-dimensional underground profile, and by superimposing and aligning these profiles, an underground three-dimensional model of the target area is constructed; this improves the efficiency of data processing, and makes the monitoring results of underground obstacles more intuitive and accurate, providing strong technical support for the planning and construction of underground projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram of the structure of an underground obstacle monitoring system based on geological radar used in an embodiment of the present invention;

[0032] Figure 2 A schematic flow chart of the underground obstacle monitoring method based on geological radar adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0034] Embodiment 1, a method for monitoring underground obstacles based on geological radar, such as Figure 2 As shown, including:

[0035] The target area to be monitored is divided into several local areas in a grid shape, and each grid represents a local area to be monitored. This division method helps to process monitoring data in a systematic and modular manner, so that each local area can be analyzed and processed in a targeted manner, improving the accuracy and efficiency of monitoring. The grid spacing is selected and obtained by the group optimization algorithm. The appropriate grid spacing is crucial to ensure monitoring efficiency; the optimization algorithm can help find the best grid configuration to adapt to different geological conditions and monitoring needs; the target area is preliminarily scanned using a geological radar according to a preliminary scanning route. The preliminary scanning route includes several parallel and evenly spaced scanning paths covering the entire target area. Each local area is passed by a scanning path to obtain several low-frequency radar signal data; the purpose of the preliminary scan is to collect basic data and provide necessary information for subsequent detailed analysis; this method can quickly cover the entire area and provide guidance for further precise scanning; low-frequency radar signal data is the key to analyzing underground structures and identifying obstacles; by analyzing these data, it is possible to preliminarily determine whether there are obstacles underground and their approximate locations;

[0036] For any local area, the low-frequency radar signal data passing through the local area is analyzed to identify and quantify the changes in the signal during underground propagation. Analyzing the radar signal data helps to understand the response of the underground medium to the signal, which is a key step in identifying underground obstacles. The signal attenuation index of the local area is calculated, and the obstacle reference depth of the local area is calculated at the same time. The signal attenuation index refers to the measure of energy loss of the radar signal during underground propagation. By analyzing the attenuation of the signal, the electromagnetic properties of the underground medium can be inferred. The signal attenuation index is crucial to determining the propagation characteristics of the radar wave in the underground medium, which affects the detection depth and resolution of the radar wave. The obstacle reference depth refers to the depth range of underground obstacles that may exist inferred based on the radar signal data. This parameter helps to determine the approximate location of the obstacle and provides important information for subsequent precise scanning.

[0037] Based on the signal attenuation index and obstacle reference depth of each local area, the ideal radar transmission frequency corresponding to each local area is obtained. Selecting the appropriate transmission frequency can optimize the penetration and resolution of the radar signal and improve the efficiency and accuracy of obstacle detection. The ideal radar transmission frequency is used to use the geological radar to perform precise scanning along the preset route to obtain a number of precise radar signal data. All the acquired precise radar signal data are subjected to time-depth processing to convert the time domain signal into the depth domain signal to obtain a number of two-dimensional underground profiles. All two-dimensional underground profiles are superimposed and registered to obtain the underground three-dimensional model of the target area. This step synthesizes multiple profiles into a complete three-dimensional model, providing a comprehensive view to understand and analyze the underground structure, which helps to more accurately identify and evaluate underground obstacles.

[0038] For any obstacle, the local point cloud area of ​​the current obstacle is extracted from the underground 3D model, and the extracted local point cloud area is used as the input of the obstacle classification model to output the obstacle classification results, such as rocks, pipes, cables, etc. This is crucial for understanding underground structures and planning construction, and provides clear obstacle information for the construction team to help them make more informed construction decisions, such as avoiding certain areas or adopting specific construction techniques; Obstacle information is obtained based on the underground 3D model and obstacle classification results.

[0039] Calculate and obtain the signal attenuation index of the local area. The specific operations are as follows:

[0040] For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area. This step is to focus the analysis on a specific local area so as to more accurately analyze the radar signal characteristics of the area. By intercepting the local scanning path, it can be ensured that the analyzed data is directly related to the specific area, and the accuracy of the signal attenuation index calculation is improved; the starting point, center point and end point of the local scanning path are used as reference points. Since the purpose of the preliminary scan is only to obtain approximate information about the underground, the three reference points selected here can reduce the amount of calculation while having a more reasonable representation of the underground information; for any reference point, the strongest and first received reflection signal strength of the current reference point is extracted from the low-frequency radar signal data corresponding to the current local scanning path. , and record the propagation time t of the reflected signal. The intensity and propagation time of the reflected signal are the basic data for calculating the signal attenuation index. They reflect the characteristics of radar signal propagation in underground media. Using the formula Calculate and obtain the one-way propagation distance d of the reflected signal, where c is the speed of light; is the dielectric constant corresponding to the current local area; then use the formula Calculate and obtain the single-point attenuation index a of the current reference point, where: is the initial signal strength emitted by the radar; after obtaining the single-point attenuation index of the three reference points, the average value of the single-point attenuation index of the three reference points is calculated, and the average value obtained by calculation is used as the signal attenuation index of the current local area. By calculating the average value, a comprehensive indicator representing the signal attenuation characteristics of the entire local area can be obtained. This indicator is of great significance for subsequent radar signal analysis and obstacle detection.

[0041] Calculate and obtain the obstacle reference depth in the local area. The specific operations are as follows:

[0042] For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area, and the local radar signal data corresponding to the current local scanning path is intercepted in the low-frequency radar signal data. This step is to focus the analysis on a specific local area to ensure that the data of subsequent analysis is directly related to the area, thereby improving the accuracy of obstacle reference depth calculation; set a signal strength threshold, which helps to filter out signals caused by noise or other non-obstacle factors, ensuring that only those reflected signals that truly represent underground obstacles are analyzed. For any reflected signal in the local radar signal data that is higher than the signal strength threshold, use the formula Calculate the one-way propagation distance of the current reflected signal , as the rough depth of the obstacle corresponding to the current reflection signal, where T represents the propagation time of the current reflection signal. By calculating the propagation distance of each effective reflection signal, the approximate depth of the obstacle can be estimated; this step provides preliminary information for determining the precise location of the obstacle; after obtaining the rough depth of the obstacle for all reflection signals above the signal strength threshold in the local radar signal data, calculate the average of all obtained rough depths of the obstacle to obtain the obstacle reference depth of the current local area. Calculating the average can reduce the impact of individual abnormal data and provide a more stable and reliable obstacle depth estimation. The obstacle reference depth is crucial for subsequent precise scanning and detailed analysis of obstacles.

[0043] Get the ideal radar transmission frequency corresponding to each local area. The specific operations are as follows:

[0044] Set the base transmit frequency This reference transmission frequency is determined based on experience or preliminary experimental data and is used as the starting point for subsequent optimization to ensure that the radar system can start working at a known frequency and provide a basis for further optimization. For any local area, the formula Calculate and obtain the ideal radar transmission frequency f in the current local area, where D is the obstacle reference depth in the current local area, A is the signal attenuation index in the current local area, and They represent the adjustment coefficients of obstacle reference depth and signal attenuation index respectively; and The value of is obtained through genetic algorithm selection; by calculating the ideal radar transmission frequency, the most suitable radar transmission frequency can be customized for each local area; this helps to optimize the penetration and resolution of radar signals and improve the efficiency and accuracy of obstacle detection.

[0045] The swarm optimization algorithm selected for obtaining the grid spacing is the simulated annealing algorithm. The simulated annealing algorithm is a probabilistic global optimization algorithm inspired by the annealing process of solid materials in physics. The algorithm simulates the process of particles gradually cooling down and reaching the lowest energy state during the metal annealing process to find the global optimal solution of the problem. During the execution of the algorithm, it accepts a worse solution with a certain probability to avoid falling into the local optimal solution, and gradually lowers the "temperature" to reduce the probability of accepting a bad solution. Finally, when the "temperature" is low enough, the algorithm tends to be stable, and the solution is close to the global optimal solution. The specific operations are as follows:

[0046] Step 1: Randomly select an initial grid spacing as the current solution, set the initial temperature, maximum number of iterations and temperature attenuation coefficient; use the objective function Calculate the fitness value of the current solution, where L represents the length of the grid spacing; H represents the degree of difference of each local area when the current grid spacing is applied, and the degree of difference H is expressed by the sum of the standard deviation of the signal attenuation index of all local areas and the standard deviation of the obstacle reference depth; and are the weights of the grid spacing length and the local area difference degree respectively. If you pay more attention to the computational efficiency and hope to reduce the number and time of initial scanning, you can If the grid spacing is set larger, a larger grid spacing is preferred. If the local area division accuracy is more important, A larger setting tends to choose a grid division that reflects more underground differences;

[0047] Step 2: Add or subtract a random value to the current solution to generate a neighborhood solution; use the objective function to calculate the fitness value of the neighborhood solution and compare it with the fitness value of the current solution; if the fitness value of the current solution is less than the fitness value of the neighborhood solution, update the current solution to the selected neighborhood solution; if the fitness value of the current solution is greater than the fitness value of the neighborhood solution, update the neighborhood solution according to the probability Decide whether to accept the neighborhood solution as the current solution; where represents the fitness value of the neighborhood solution, represents the fitness value of the current solution, Indicates the current temperature; using the formula Update the temperature, where is the temperature attenuation coefficient;

[0048] Step 3: Repeat step 2 until the maximum number of iterations is reached, and the current solution finally obtained is the optimal grid spacing.

[0049] The obstacle classification model is built on the PointNet model, which is a deep learning network architecture for processing point cloud data. It can directly take raw point cloud data as input without any manual feature engineering or complex preprocessing steps. The model can capture the global features in point cloud data by directly learning the relationship between points using deep neural networks, thereby achieving classification and segmentation of three-dimensional objects. PointNet consists of a sequence of multi-layer perceptrons (MLPs) that can learn the local features of each point in the point cloud and obtain the global descriptor of the entire point cloud through global feature aggregation. This architecture enables PointNet to perform well in processing irregular data such as point cloud data, and is widely used in fields such as three-dimensional object recognition, scene understanding, and environmental perception of autonomous vehicles.

[0050] Train the obstacle classification model. The specific operations are as follows:

[0051] Obtain several point cloud data samples with obstacle classification results marked; divide all point cloud data samples into training sets and test sets, use the training sets to train the obstacle classification model, and then input the test sets into the obstacle classification model to obtain the obstacle classification test results; set the training conditions, if the obstacle classification test results meet the training conditions, output the trained obstacle classification model; otherwise, continue to use the training set to train the obstacle classification model.

[0052] Embodiment 2, an underground obstacle monitoring system based on geological radar, such as Figure 1 As shown, including:

[0053] The local area feature acquisition module is used to divide the target area to be monitored into several local areas in a grid shape; use the geological radar to perform a preliminary scan on the target area according to the preliminary scanning route to obtain several low-frequency radar signal data; analyze the low-frequency radar signal data of all local areas, calculate the signal attenuation index of the local area, and calculate the obstacle reference depth of the local area;

[0054] The underground three-dimensional model acquisition module is used to obtain the ideal radar transmission frequency corresponding to each local area according to the signal attenuation index and obstacle reference depth of each local area; use the ideal radar transmission frequency to scan along the preset route using the geological radar to obtain a number of precise radar signal data; perform time-depth processing on all the acquired precise radar signal data to obtain a number of two-dimensional underground profiles; superimpose and align all two-dimensional underground profiles to obtain the underground three-dimensional model of the target area;

[0055] The obstacle classification module is used to extract the local point cloud area of ​​all obstacles from the underground three-dimensional model, use the extracted local point cloud area as the input of the obstacle classification model, and output the obstacle classification result; obtain obstacle information based on the underground three-dimensional model and the obstacle classification result.

[0056] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for monitoring underground obstacles based on geological radar, characterized in that: include: The target area to be monitored is divided into several local areas in a grid shape, and the spacing of the grids is selected and obtained by the swarm optimization algorithm; the target area is preliminarily scanned using a geological radar according to a preliminary scanning route, and the preliminary scanning route includes several parallel and evenly spaced scanning paths covering the entire target area. Each local area is passed by a scanning path to obtain several low-frequency radar signal data; For any local area, the low-frequency radar signal data passing through the local area is analyzed, the signal attenuation index of the local area is calculated, and the obstacle reference depth of the local area is calculated; Based on the signal attenuation index and obstacle reference depth of each local area, the ideal radar transmission frequency corresponding to each local area is obtained; the ideal radar transmission frequency is used to use the geological radar to perform precise scanning along the preset route to obtain a number of precise radar signal data; all the obtained precise radar signal data are subjected to time-depth processing to obtain a number of two-dimensional underground profiles; all two-dimensional underground profiles are superimposed and registered to obtain the underground three-dimensional model of the target area; For any obstacle, extract the local point cloud area of ​​the current obstacle from the underground 3D model, use the extracted local point cloud area as the input of the obstacle classification model, and output the obstacle classification result; Obstacle information is obtained based on the underground 3D model and obstacle classification results.

2. The underground obstacle monitoring method based on geological radar according to claim 1 is characterized in that: Calculate and obtain the signal attenuation index of the local area. The specific operations are as follows: For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area, and the starting point, center point and end point of the local scanning path are used as reference points; for any reference point, the strongest and first received reflection signal strength of the current reference point is extracted from the low-frequency radar signal data corresponding to the current local scanning path. , and record the propagation time t of the reflected signal, using the formula Calculate and obtain the one-way propagation distance d of the reflected signal, where c is the speed of light; is the dielectric constant corresponding to the current local area; then use the formula Calculate and obtain the single-point attenuation index a of the current reference point, where: is the initial signal strength emitted by the radar; after obtaining the single-point attenuation indexes of the three reference points, the average value of the single-point attenuation indexes of the three reference points is calculated, and the calculated average value is used as the signal attenuation index of the current local area.

3. The underground obstacle monitoring method based on geological radar according to claim 2 is characterized in that: Calculate and obtain the obstacle reference depth in the local area. The specific operations are as follows: For any local area, the local scanning path corresponding to the current local area is intercepted on the scanning path passing through the current local area, and the local radar signal data corresponding to the current local scanning path is intercepted in the low-frequency radar signal data; a signal strength threshold is set, and for any reflected signal in the local radar signal data that is higher than the signal strength threshold, the formula is used Calculate the one-way propagation distance of the current reflected signal , as the rough depth of the obstacle corresponding to the current reflection signal, where T represents the propagation time of the current reflection signal; After obtaining the rough obstacle depths of all reflected signals above the signal strength threshold in the local radar signal data, the average value of all obtained rough obstacle depths is calculated to obtain the obstacle reference depth of the current local area.

4. The underground obstacle monitoring method based on geological radar according to claim 3 is characterized in that: Get the ideal radar transmission frequency corresponding to each local area. The specific operations are as follows: Set the base transmit frequency , for any local area, using the formula Calculate and obtain the ideal radar transmission frequency f in the current local area, where D is the obstacle reference depth in the current local area, A is the signal attenuation index in the current local area, and They represent the adjustment coefficients of obstacle reference depth and signal attenuation index respectively; and The value of is obtained through genetic algorithm selection.

5. The underground obstacle monitoring method based on geological radar according to claim 4 is characterized in that: Select the simulated annealing algorithm as the group optimization algorithm for obtaining the grid spacing. The specific operations are as follows: Step 1: Randomly select an initial grid spacing as the current solution, set the initial temperature, maximum number of iterations and temperature attenuation coefficient; use the objective function Calculate the fitness value of the current solution, where L represents the length of the grid spacing; H represents the degree of difference of each local area when the current grid spacing is applied, and the degree of difference H is expressed by the sum of the standard deviation of the signal attenuation index of all local areas and the standard deviation of the obstacle reference depth; and are the weights of the grid spacing length and the degree of local area difference, respectively; Step 2: Add or subtract a random value to the current solution to generate a neighborhood solution; use the objective function to calculate the fitness value of the neighborhood solution and compare it with the fitness value of the current solution; if the fitness value of the current solution is less than the fitness value of the neighborhood solution, update the current solution to the selected neighborhood solution; if the fitness value of the current solution is greater than the fitness value of the neighborhood solution, update the neighborhood solution according to the probability Decide whether to accept the neighborhood solution as the current solution; where represents the fitness value of the neighborhood solution, represents the fitness value of the current solution, Indicates the current temperature; using the formula Update the temperature, where is the temperature attenuation coefficient; Step 3: Repeat step 2 until the maximum number of iterations is reached, and the current solution finally obtained is the optimal grid spacing.

6. The underground obstacle monitoring method based on geological radar according to claim 5 is characterized in that: The obstacle classification model is built based on the PointNet model.

7. The underground obstacle monitoring method based on geological radar according to claim 6 is characterized in that: Train the obstacle classification model. The specific operations are as follows: Obtain several point cloud data samples with obstacle classification results marked on them; divide all point cloud data samples into a training set and a test set, use the training set to train the obstacle classification model, and then input the test set into the obstacle classification model to obtain the obstacle classification test results; Set the training conditions. If the obstacle classification test results meet the training conditions, output the trained obstacle classification model; otherwise, continue to use the training set to train the obstacle classification model.

8. An underground obstacle monitoring system based on geological radar, characterized in that: The system is applied to the underground obstacle monitoring method based on geological radar as described in any one of claims 1 to 7, comprising: The local area feature acquisition module is used to divide the target area to be monitored into several local areas in a grid shape; use the geological radar to perform a preliminary scan on the target area according to the preliminary scanning route to obtain several low-frequency radar signal data; analyze the low-frequency radar signal data of all local areas, calculate the signal attenuation index of the local area, and calculate the obstacle reference depth of the local area; The underground three-dimensional model acquisition module is used to obtain the ideal radar transmission frequency corresponding to each local area according to the signal attenuation index and obstacle reference depth of each local area; use the ideal radar transmission frequency to scan along the preset route using the geological radar to obtain a number of precise radar signal data; perform time-depth processing on all the acquired precise radar signal data to obtain a number of two-dimensional underground profiles; superimpose and align all two-dimensional underground profiles to obtain the underground three-dimensional model of the target area; The obstacle classification module is used to extract the local point cloud area of ​​all obstacles from the underground three-dimensional model, use the extracted local point cloud area as the input of the obstacle classification model, and output the obstacle classification result; obtain obstacle information based on the underground three-dimensional model and the obstacle classification result.

Images