Ai-based underground utility detection and digital twin implementation system and method

Abstract

According to various embodiments, an AI-based automatic detection and digital twin implementation system for underground buried objects may include a processor, wherein the processor receives underground detection data acquired from a Ground Penetrating Radar (GPR) device, generates a detection result including the presence and location information of underground buried objects within the underground detection data using a pre-trained artificial intelligence model, generates a 3D digital twin model of the underground buried objects based on the detection result, visualizes the 3D digital twin model and provides it through a user interface, and is configured to provide location information of the underground buried objects in the form of map data using GPS data. According to various embodiments, an AI-based method for automatic detection of underground buried objects and implementation of a digital twin may include: generating a detection result including information on the presence and location of underground buried objects within the underground detection data using a pre-trained artificial intelligence model; generating a three-dimensional digital twin model of the underground buried objects based on the detection result; visualizing the three-dimensional digital twin model and providing it through a user interface; and providing location information of the underground buried objects in the form of map data using GPS data. Various other embodiments are also possible.

Description

Technology Field

[0001] The present invention relates to underground buried object detection and design technology, and specifically to a system and method for automatically interpreting detection data acquired by Ground Penetrating Radar (GPR) using artificial intelligence (AI) and implementing a three-dimensional digital twin of an underground buried object based thereon. Background Technology

[0003] Various underground infrastructures, such as water and sewage pipes, communication lines, power lines, gas pipes, and heating pipelines, serve as essential elements of urban and industrial infrastructure. Since these buried facilities are closely related to road construction, building construction, and maintenance work, prior detection and accurate location identification are absolutely required.

[0004] To this end, Ground Penetrating Radar (GPR) has traditionally been primarily used. GPR is a method that emits high-frequency electromagnetic waves into the ground and receives the reflected signals to represent the underground structure as a two-dimensional or cross-sectional image; it has the advantages of being non-destructive and relatively simple. However, due to the complexity of the signals, the resulting radargrams require manual interpretation by experts, and there is a problem in that the results vary depending on the interpreter's experience or subjectivity.

[0005] Furthermore, because GPR data is highly sensitive to external environmental factors such as electromagnetic interference, ground conditions, and moisture content, signal distortion and overlap frequently occur, making object identification or shape estimation difficult. In particular, in environments where buried structures are dense, such as urban areas, noise is severe and signal overlap makes accurate judgment challenging.

[0006] Another limitation of GPR is that it is mostly provided as two-dimensional data. Three-dimensional information is required to understand the actual shape and spatial positional relationships of underground structures, but existing systems faced many limitations in implementing this. Consequently, it was difficult to utilize it directly for construction design or digital simulation. Furthermore, the low level of digitization and visualization of GPR data is also pointed out as a major problem, as it makes it difficult to link with systems such as CAD and GIS or to operate them in a digital twin environment.

[0007] Attempts are being made to apply artificial intelligence technology to GPR data analysis to solve these problems, but the technical completeness is still lacking in terms of object detection accuracy, real-time processing capability, 3D transformation of data, and integration with digital twins.

[0008] Therefore, there is an urgent need for technology that can provide intuitive and precise information on buried underground structures by analyzing GPR-based detection data with AI, visualizing it in 3D in real time, and implementing it on a digital twin. Prior art literature

[0010] Korean Registered Patent Publication No. 10-2743130 (Dec. 11, 2024) Korean Registered Patent Publication No. 10-2525894 (April 21, 2023) Korean Published Patent Publication No. 10-2023-0084145 (June 15, 2023) The problem to be solved

[0011] Accordingly, the technical problem that the present invention aims to solve is created to resolve the aforementioned problem, and aims to provide an AI-based automatic detection of underground buried structures and a digital twin implementation system and method.

[0012] The present invention aims to solve the aforementioned problems by providing a technology that automatically analyzes the location and shape of underground buried objects using artificial intelligence based on GPR detection data, and visualizes and implements this in a digital twin environment in real time.

[0013] Furthermore, the present invention aims to provide a technology that improves the accuracy of identification and classification of buried objects based on detection data, implements a data processing method robust against noise or interference, and enables precise analysis without expert intervention by automating the entire process.

[0014] Furthermore, the present invention aims to implement a technology that maximizes real-time field usability by ensuring geographic information-based consistency through the combination of GPR data and GPS-based location information, and by enabling the detection results to be applied to a virtual reality or augmented reality-based interface.

[0015] The technical problems that the present invention aims to solve are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art to which the present invention belongs from the description below. means of solving the problem

[0017] The AI-based automatic detection of underground buried objects and digital twin implementation system according to the present invention for solving the above technical problem includes a processor, and the processor receives underground detection data acquired from a Ground Penetrating Radar (GPR) device, generates a detection result including the presence and location information of underground buried objects within the underground detection data using a pre-trained artificial intelligence model, generates a 3D digital twin model of the underground buried objects based on the detection result, visualizes the 3D digital twin model and provides it through a user interface, and can be configured to provide location information of the underground buried objects in the form of map data using GPS data.

[0018] The AI-based automatic detection of underground buried objects and digital twin implementation method according to the present invention may include: a step of generating a detection result including the presence and location information of underground buried objects within the underground detection data using a pre-trained artificial intelligence model; a step of generating a three-dimensional digital twin model of the underground buried objects based on the detection result; a step of visualizing the three-dimensional digital twin model and providing it through a user interface; and a step of providing the location information of the underground buried objects in the form of map data using GPS data. Effects of the invention

[0020] According to the present invention, by analyzing GPR detection results using artificial intelligence technology, automatic identification and location estimation of underground buried objects are possible, and rapid and accurate analysis results can be derived without expert interpretation. This enables standardization and consistency in detection.

[0021] In addition, information on underground utilities can be intuitively checked in a digital twin environment through 3D visualization based on artificial intelligence analysis results, and this can be directly linked to and utilized in construction design and management systems.

[0022] In addition, the present invention can contribute to the construction of smart city-based infrastructure by ensuring the accuracy of spatial information through the alignment processing of GPR data and GPS location information, and by enabling integration with various sensors and simulation systems.

[0023] Consequently, the present invention can significantly improve the accuracy and efficiency of underground buried object detection and provide various practical effects, such as ensuring construction safety, preventing construction delays, and reducing maintenance costs.

[0024] The effects of the present invention are not limited to those described above, and potential effects expected from the technical features of the present invention will be clearly understood from the description below. Brief explanation of the drawing

[0026] FIG. 1 illustrates a block diagram of a system according to various embodiments of the present invention. FIG. 2 is a flowchart illustrating how a system operates according to various embodiments of the present invention. FIG. 3 is a diagram of a GPR detection result screen and a GPR exploration principle according to various embodiments of the present invention. FIG. 4 is a diagram of a Confusion Matrix, which is the result of evaluating the performance of an AI model according to various embodiments of the present invention. FIG. 5 is a diagram briefly illustrating the overall process of a system according to various embodiments of the present invention. FIG. 6 is a diagram showing the accuracy of an AI model by class according to various embodiments of the present invention. FIG. 7 is a diagram illustrating a GPR detection process and an underground condition mapping process according to various embodiments of the present invention. Specific details for implementing the invention

[0027] Hereinafter, various embodiments of this document are described with reference to the accompanying drawings. The embodiments and the terms used therein are not intended to limit the technology described in this document to specific embodiments and should be understood to include various modifications, equivalents, and / or substitutions of said embodiments. In relation to the description of the drawings, similar reference numerals may be used for similar components. A singular expression may include a plural expression unless the context clearly indicates otherwise. In this document, expressions such as "A or B" or "at least one of A and / or B" may include all possible combinations of items listed together. Expressions such as "first," "second," "first," or "second" may modify said components regardless of order or importance and are used only to distinguish one component from another and do not limit said components. When it is mentioned that a certain (e.g., 1st) component is "(functionally or telecommunicationally) connected" or "connected" to another (e.g., 2nd) component, said certain component may be directly connected to said other component or connected through another component (e.g., 3rd component).

[0028] In this document, "configured to" may be used interchangeably with, depending on the context, for example, hardware- or software-wise, "suitable for," "capable of," "modified to," "made to," "capable of," or "designed to." In some cases, the expression "device configured to" may mean that the device is "capable of" in conjunction with other devices or components. For example, the phrase "processor configured to perform A, B, and C" may mean a dedicated processor for performing the corresponding operations (e.g., an embedded processor), or a general-purpose processor capable of performing the corresponding operations by executing one or more software programs stored in a memory device (e.g., a CPU or application processor).

[0029] An electronic device according to various embodiments of the present document may include, for example, at least one of a smartphone, a tablet PC, a desktop PC, a laptop PC, a netbook computer, a workstation, and a server.

[0030] Referring to FIG. 1, an electronic device (101) within a network environment (100) in various embodiments is described. The electronic device (101) may include a bus (110), a processor (120), a memory (130), an input / output interface (150), a display (160), and a communication interface (170). In some embodiments, the electronic device (101) may omit at least one of the components or additionally include other components. The bus (110) may include a circuit that connects the components (110-170) to each other and transmits communication (e.g., control messages or data) between the components. The processor (120) may include one or more of a central processing unit, an application processor, or a communication processor (CP). The processor (120) may, for example, perform operations or data processing regarding the control and / or communication of at least one other component of the electronic device (101).

[0031] Memory (130) may include volatile and / or non-volatile memory. Memory (130) may store instructions or data related to at least one other component of the electronic device (101), for example. According to one embodiment, memory (130) may store software and / or programs (140). Programs (140) may include, for example, a kernel (141), middleware (143), an application programming interface (API) (145), and / or application programs (or "applications") (147), etc. At least part of the kernel (141), middleware (143), or API (145) may be referred to as an operating system. The kernel (141) can control or manage system resources (e.g., bus (110), processor (120), or memory (130), etc.) used to execute operations or functions implemented in other programs (e.g., middleware (143), API (145), or application program (147)). Additionally, the kernel (141) can provide an interface that controls or manages system resources by accessing individual components of the electronic device (101) from the middleware (143), API (145), or application program (147).

[0032] Middleware (143) may act as an intermediary to enable, for example, an API (145) or an application program (147) to communicate with the kernel (141) to exchange data. Additionally, middleware (143) may process one or more work requests received from the application program (147) according to priority. For example, middleware (143) may process the one or more work requests by assigning a priority to at least one of the application programs (147) to use the system resources of the electronic device (101) (e.g., bus (110), processor (120), or memory (130), etc.). The API (145) is an interface for the application (147) to control functions provided by the kernel (141) or middleware (143), and may include at least one interface or function (e.g., command) for, for example, file control, window control, image processing, or character control. The input / output interface (150) can, for example, transmit commands or data input from a user or other external device to other component(s) of the electronic device (101), or output commands or data received from other component(s) of the electronic device (101) to the user or other external device.

[0033] The display (160) may include, for example, a liquid crystal display (LCD), a light-emitting diode (LED) display, an organic light-emitting diode (OLED) display, a micro-electromechanical system (MEMS) display, or an electronic paper display. The display (160) may display various content (e.g., text, images, videos, icons, and / or symbols, etc.) to a user, for example. The display (160) may include a touch screen and may receive touch, gesture, proximity, or hovering input using, for example, an electronic pen or a part of the user's body. The communication interface (170) may establish communication between, for example, the electronic device (101) and an external device (e.g., a first external electronic device (102), a second external electronic device (104), or a server (106)). For example, the communication interface (170) can be connected to a network (162) via wireless or wired communication to communicate with an external device (e.g., a second external electronic device (104) or a server (106)).

[0034] Wireless communication may include cellular communication using at least one of, for example, LTE, LTE-A (LTE Advance), CDMA (code division multiple access), WCDMA (wideband CDMA), UMTS (universal mobile telecommunications system), WiBro (Wireless Broadband), or GSM (Global System for Mobile Communications). According to one embodiment, wireless communication may include at least one of, for example, WiFi (wireless fidelity), Bluetooth, Bluetooth Low Energy (BLE), Zigbee, NFC (near field communication), Magnetic Secure Transmission, Radio Frequency (RF), or Body Area Network (BAN). According to one embodiment, wireless communication may include GNSS. GNSS may be, for example, GPS (Global Positioning System), Glonass (Global Navigation Satellite System), Beidou Navigation Satellite System (hereinafter "Beidou"), or Galileo, the European global satellite-based navigation system. Hereinafter, in this document, "GPS" may be used interchangeably with "GNSS". Wired communication may include at least one of, for example, USB (universal serial bus), HDMI (high definition multimedia interface), RS-232 (recommended standard 232), power line communication, or POTS (plain old telephone service).The network (162) may include at least one of a telecommunications network, for example, a computer network (e.g., LAN or WAN), the Internet, or a telephone network.

[0035] Each of the first and second external electronic devices (102, 104) may be the same or a different type of device as the electronic device (101). According to various embodiments, all or part of the operations performed on the electronic device (101) may be performed on one or more other electronic devices (e.g., electronic devices (102, 104), or a server (106). According to one embodiment, when the electronic device (101) needs to perform a function or service automatically or upon request, the electronic device (101) may request at least some of the associated functions from another device (e.g., electronic devices (102, 104), or a server (106)) instead of performing the function or service itself or additionally. The other electronic device (e.g., electronic devices (102, 104), or a server (106)) may perform the requested function or additional functions and transmit the result to the electronic device (101). The electronic device (101) may provide the requested function or service by processing the received result as is or additionally. For this purpose, for example, cloud computing, distributed computing, or client-server computing technologies may be used.

[0037] FIG. 2 is a flowchart illustrating how a system operates according to various embodiments of the present invention.

[0038] FIG. 3 is a diagram of a GPR detection result screen and a GPR exploration principle according to various embodiments of the present invention.

[0039] FIG. 4 is a diagram of a Confusion Matrix, which is the result of evaluating the performance of an AI model according to various embodiments of the present invention.

[0040] FIG. 5 is a diagram briefly illustrating the overall process of a system according to various embodiments of the present invention.

[0041] FIG. 6 is a diagram showing the accuracy of an AI model by class according to various embodiments of the present invention.

[0042] FIG. 7 is a diagram illustrating a GPR detection process and an underground condition mapping process according to various embodiments of the present invention.

[0044] In operation 201, according to various embodiments, the processor (120) of the system (100) may receive underground detection data acquired from a Ground Penetrating Radar (GPR) device. Here, Ground Penetrating Radar (GPR) is a non-destructive geophysical exploration technology that radiates high-frequency electromagnetic waves (generally 10 MHz to 2.6 GHz) into the ground and receives and analyzes signals reflected by changes in the electrical properties (mainly permittivity) of the underground medium to image the underground structure. GPR is utilized in various fields such as detection of underground buried objects, ground structure investigation, rock mass exploration, and groundwater exploration, and may refer to a technology essential for safety diagnosis of existing buried objects, particularly due to its non-destructive characteristics.

[0045] According to a specific embodiment, the processor (120) of the system (100) may receive a reflected signal in the time domain or frequency domain received through the antenna of the GPR equipment in the form of digital data, the time domain data is data obtained by sampling the amplitude of the received signal over time, and the frequency domain data is amplitude and phase information for each frequency component obtained by performing a Fourier Transform on the time domain data, and the GPR data can generally be expressed in the form of B-scan, C-scan, or 3D volume data, each of which can be expressed as follows.

[0046] B-scan: A two-dimensional cross-sectional image of the reflected signal measured while moving the GPR antenna at regular intervals, which can show the distribution of reflectors according to the depth underground.

[0047] C-scan: Represents reflection signals measured at a specific depth as a horizontal cross-sectional image, which can show the distribution of reflectors at a specific underground depth.

[0048] 3D volume data: This can refer to data that represents the distribution of reflectors throughout the entire underground space in three dimensions by arranging B-scan data in three dimensions.

[0049] According to one embodiment, the processor (120) of the system (100) can receive a B-scan image acquired from a GPR device and determine the location and depth of a pipe buried underground. Referring to FIG. 3, a hyperbola-shaped reflection pattern in the GPR survey result screen is a typical signal generated by a cylindrical buried object (e.g., a pipe), and the processor (120) can estimate the horizontal location and depth of the pipe by analyzing the apex position and curvature of this pattern. At this time, the travel distance (x0) to the apex of the hyperbola and the distance (d) from the antenna to the buried object can be expressed by the following relationship.

[0050]

[0051] Here, is the propagation speed of electromagnetic waves, is reflection time, can mean the distance between the antenna and the ground surface.

[0053] In operation 203, according to various embodiments, the processor (120) of the system (100) may generate detection results including information on the presence and location of underground buried objects within underground detection data using a pre-trained artificial intelligence model. Here, the pre-trained artificial intelligence model may refer to a deep learning model that is pre-trained using a large amount of GPR data and correct information (e.g., location, type, depth, etc. of the buried objects) for said data, and deep learning may refer to a technology that automatically learns abstracted features from data based on an artificial neural network and thereby demonstrates excellent performance in solving complex pattern recognition and prediction problems.

[0054] According to specific embodiments, the artificial intelligence model may include an object detection model based on YOLOv11 (You Only Look Once version 11). Here, YOLO (You Only Look Once) refers to a family of deep learning models developed for real-time object detection, and YOLOv11 may refer to the 11th version of YOLO. The YOLO model processes an image in a single forward pass to simultaneously predict the bounding box and class probability of an object within the image, providing fast processing speed and high real-time performance. This can be very important in fields requiring rapid decision-making, such as underground utilities detection.

[0055] According to one embodiment, the processor (120) of the system (100) inputs the received GPR data into a pre-trained artificial intelligence model to obtain detection results including information such as the presence, type, location, and depth of underground buried objects. The artificial intelligence model analyzes the features of the GPR data, identifies the buried objects based on learned patterns, and estimates their location. At this time, the features of the GPR data can be defined in various ways, such as signal amplitude, frequency, phase, time delay, and spatial distribution pattern, and the deep learning model can learn these features hierarchically to derive the final detection results. For example, the processor (120) of the system (100) can use the YOLOv11 model to detect underground buried objects such as pipes, cables, and voids in a GPR image and output the location coordinates (x, y, z), type, and confidence score of each buried object. The YOLOv11 model is a deep learning model developed for real-time object detection. It predicts the location and type of an object by looking at an image only once (Look Once), providing fast processing speed and high accuracy. Referring to Figure 4, the Confusion Matrix visually represents the object detection performance of the YOLOv11 model and demonstrates how accurately the model classifies metallic, non-metallic, and background objects. At this time, performance indicators such as accuracy, recall, and F1-score for each class are provided to evaluate the reliability of the model.

[0056] According to one embodiment, the processor (120) of the system (100) can classify underground buried objects into metallic, non-metallic, and background types. This classification can play an important role in improving the detection accuracy and reliability of the AI ​​model by distinguishing various objects existing underground according to their material characteristics during the Ground Penetrating Radar (GPR) data analysis process and by clearly identifying underground buried objects and the surrounding environment (background). Specifically, the processor (120) can classify objects within GPR images or 3D volume data into three main categories—metallic, non-metallic, and background—through an AI model, particularly a deep learning-based object detection model. Since each category exhibits different unique reflection characteristics for GPR signals, the AI ​​model can determine the type of object and clearly distinguish it from the background by analyzing the differences in these reflection characteristics. Each category can be described as follows.

[0057] Metallic: This refers to underground utilities made of metallic materials such as steel, copper, and aluminum. Metallic objects cause strong reflection of GPR signals and tend to exhibit distinct hyperbolas or other characteristic reflection patterns. This is because most electromagnetic waves are reflected due to the high electrical conductivity of metals.

[0058] Non-Metallic: This refers to underground utilities made of non-metallic materials such as PVC (polyvinyl chloride), PE (polyethylene), and concrete. Compared to metallic objects, non-metallic objects exhibit weaker reflection intensity for GPR signals, and their reflection patterns may be relatively indistinct or difficult to distinguish from the surrounding soil. This is due to the low electrical conductivity and dielectric constant characteristics of non-metallic materials.

[0059] Background: Refers to the surrounding environment, such as soil, bedrock, and groundwater, where no underground utilities exist. Background areas generally tend to exhibit weak or uniform reflection characteristics for GPR signals.

[0060] According to one embodiment, the processor (120) of the system (100) can classify objects exhibiting strong reflection signals and distinct hyperbolic patterns within a GPR image as metallic pipes through an AI model, classify objects exhibiting weak reflection signals and indistinct reflection patterns as non-metallic cables, and classify areas with almost no reflection signals or uniform reflection characteristics as background. Referring to FIG. 4, the Confusion Matrix visually shows how accurately the AI ​​model classifies metallic, non-metallic, and background objects, and can be used to evaluate the classification performance of the model. At this time, performance indicators such as precision, recall, and F1-score for each class can be provided together to quantitatively evaluate the classification performance of the model.

[0062] In operation 205, according to various embodiments, the processor (120) of the system (100) can generate a three-dimensional digital twin model of an underground buried object based on detection results. Here, a digital twin refers to a model that accurately replicates a physical object or system in the real world in a digital form in a virtual world, and the digital twin can be utilized to integrate sensor data, simulation results, historical data, etc., to reflect the current state of the object or system in the real world in real time, predict future states, and support optimal operation and control.

[0063] According to one embodiment, the processor (120) of the system (100) can implement a three-dimensional digital twin model in the form of a 3D object using the Unity engine. Here, Unity is a cross-platform game engine for developing 2D and 3D games, simulations, visualization content, etc. Unity provides a powerful rendering pipeline, physics engine, animation system, user interface (UI) system, etc., and supports various platforms (PC, mobile, VR / AR, etc.) so it can be utilized in various fields.

[0064] According to a specific embodiment, the processor (120) can convert the detection results of the AI ​​model into three-dimensional spatial coordinate information (x, y, z) and generate a three-dimensional shape model of an underground buried object based thereon. At this time, different three-dimensional modeling techniques may be applied depending on the type of buried object, or CAD (Computer-Aided Design) data may be utilized to improve the accuracy of the model. For example, a pipe may be represented as a cylindrical model, a cable as a linear model, and a cavity as an irregular volume model. Additionally, the digital twin model may include attribute information such as the material, size, and connection relationship of the underground buried object, which can be utilized for construction simulations, maintenance planning, etc. For example, the processor (120) can generate a three-dimensional pipe model using information such as the location, diameter, and curvature of the detected pipe, and model the trajectory of the cable as a three-dimensional line. These three-dimensional models can comprehensively represent underground spatial information and can be utilized for construction simulations, interference analysis, visualization, etc. Referring to Fig. 5, the 3D design and construction simulation screen can show an example of virtually implementing an underground space based on a digital twin model and simulating the construction process.

[0065] According to one embodiment, the processor (120) of the system (100) can provide a virtual design and construction simulation to virtually place a new structure (e.g., a building, an underground parking lot, a subway line, etc.) based on a digital twin model composed of 3D objects, and to review and resolve problems that may occur during the construction process (e.g., interference between buried objects, appropriateness of the excavation path, accessibility of construction equipment, etc.). The virtual design and construction simulation is a process of virtually implementing the actual construction environment and performing design changes, establishing construction plans, reviewing the construction sequence, and evaluating safety, and can be utilized in conjunction with Building Information Modeling (BIM) technology.

[0066] According to a specific embodiment, the processor (120) of the system (100) can utilize the functions of the Unity engine to virtually place a new structure on a digital twin model composed of 3D objects and simulate the construction process over time. At this time, a physics engine can be utilized to consider physical characteristics such as the weight, collision, and gravity of the structure, and to simulate the movement of construction equipment to build a virtual environment similar to the actual construction environment. The simulation results can be provided in various forms such as 2D drawings, 3D models, animations, and reports, and can be utilized for design changes, modification of construction plans, safety training, etc. For example, the processor (120) can virtually simulate the excavation path for the foundation work of a new building and support decision-making such as changing the excavation path or adjusting the construction sequence by reviewing in advance the possibility of collision with existing buried pipes. In addition, by simulating the movement of construction equipment (e.g., excavator, crane, etc.), the efficiency of the workspace can be evaluated and the possibility of safety accidents can be predicted to establish preventive measures. Referring to Fig. 5, the 3D design and construction simulation screen can represent a virtual design process using a digital twin model and can support the user in performing tasks such as reviewing construction plans and making design changes in a virtual environment.

[0068] In operation 207, according to various embodiments, the processor (120) of the system (100) may visualize a three-dimensional digital twin model and provide it through a user interface. Here, the user interface may refer to a set of hardware and software elements provided to enable a user to interact with the system, and may refer to any device or component that performs the role of transmitting user input to the system and providing the system's output to the user. Specifically, the processor (120) may render the three-dimensional digital twin model in the form of 2D or 3D graphics and provide it through a user interface such as a monitor, touchscreen, virtual reality (VR) device, or augmented reality (AR) device. At this time, depending on user operation, functions such as changing the viewpoint, zooming in / out, viewing cross-sections, and looking up information may be provided to support the user in effectively exploring the digital twin model and obtaining information. In addition, visual elements such as color, transparency, and animation may be utilized during the visualization process to enhance the information delivery capability.

[0069] According to one embodiment, the processor (120) of the system (100) may be further configured to provide accuracy information regarding the analysis results of an artificial intelligence model in the form of a confusion matrix or an accuracy graph, together with a three-dimensional digital twin model. Here, the confusion matrix is ​​a widely used metric for evaluating the performance of an AI model, particularly a classification model, and can represent the relationship between the model's prediction results and the actual correct answers in a table form, and the confusion matrix can visually show how accurately the model classifies each class and which classes it tends to confuse.

[0070] According to a specific embodiment, the processor (120) of the system (100) can generate a confusion matrix by comparing the prediction result of the AI ​​model with the actual ground truth and provide it through a user interface. Each row of the confusion matrix represents the actual class, each column represents the class predicted by the model, and the value of each cell represents the number (or ratio) of data classified into the corresponding class. Key performance indicators that can be calculated through the confusion matrix include accuracy, precision, recall, F1-score, etc., and each performance indicator can be defined as follows.

[0071] Accuracy: Can refer to the proportion of the total data where the model correctly identifies the answer.

[0072] Precision: Can represent the proportion of data that actually belongs to a specific class among the data predicted by the model to be that class.

[0073] Recall: Can represent the proportion of data that the model predicted as belonging to a specific class out of the total data that actually belongs to that class.

[0074] F1-score: Represents the harmonic mean of precision and recall, and can be used to comprehensively evaluate the two indicators.

[0075] According to one embodiment, with reference to FIG. 4, the Confusion Matrix can indicate how accurately an AI model classifies three classes: metallic, non-metallic, and background. Here, diagonal elements represent cases where the model gets the correct answer, and off-diagonal elements represent cases where the model gets the wrong answer. In this case, the user can use the Confusion Matrix to identify the strengths and weaknesses of the model and utilize it for decision-making, such as adding data to improve the model, changing the model structure, or tuning hyperparameters.

[0076] According to one embodiment, the processor (120) of the system (100) may provide accuracy information regarding the analysis results of an artificial intelligence model in the form of an accuracy graph, along with a three-dimensional digital twin model. Here, the accuracy graph refers to various forms of graphs that visually represent the prediction accuracy of an AI model and may show overall accuracy or class-wise accuracy. Specifically, the processor (120) may calculate the overall accuracy or class-wise accuracy by comparing the prediction results of the AI ​​model with the actual correct answer, and visualize this in the form of a bar graph, line graph, pie chart, etc., and provide it through a user interface. The accuracy graph may help the user intuitively understand the performance of the model, evaluate the reliability of the model, and utilize it for model selection or improvement. For example, referring to Fig. 6, the Class-wise Accuracy graph shows the accuracy of the AI ​​model for each class (metallic, non-metallic, background) in the form of a bar graph. Through this graph, the user can visually compare which class the model classifies better and which class it struggles with, and analyze the model's performance.

[0078] In operation 209, according to various embodiments, the processor (120) of the system (100) can provide location information of underground buried objects in the form of map data using Global Positioning System (GPS) data. Here, the Global Positioning System (GPS) is a satellite navigation system developed and managed by the U.S. Department of Defense, and refers to a system that calculates the user's current location using signals received from about 30 GPS satellites orbiting the Earth. GPS can be utilized in various fields such as car navigation, smartphone location tracking, surveying, and mapmaking. Specifically, the processor (120) can generate location information of underground buried objects in the form of map data based on a geographic coordinate system (latitude, longitude) by linking GPS data collected during the GPR exploration process with a digital twin model. In this case, GPS data includes location information of the GPR antenna and can be represented on map data by assigning geographic coordinate information to each element of the digital twin model (e.g., pipes, cables, etc.). The map data can be stored in the form of vector data (points, lines, surfaces) or raster data (grid-shaped images) and can be provided through a user interface by utilizing map APIs such as the Google Maps API and OpenStreetMap API.

[0079] According to one embodiment, the processor (120) of the system (100) maps location information obtained from the GPS receiver of the GPR exploration vehicle to the corresponding point of the digital twin model and can provide this through Google Maps or a self-developed mapping system. Referring to FIG. 7, the AI ​​solution provides visual information as digital twin data of underground buried objects by linking GPR detection data analyzed by AI with GPS data, thereby allowing the user to accurately identify the location of underground buried objects in conjunction with actual geographical information and utilize it at the construction site.

[0080] According to one embodiment, the processor (120) of the system (100) may perform labeling and augmentation processing on the underground detection data using a Roboflow platform and perform a preprocessing process to remove noise from the GPR data before receiving the underground detection data. Here, Roboflow is a cloud-based platform that supports image dataset management, preprocessing, model training, etc., labeling is a task of displaying information about objects within an image in the form of annotations, and augmentation may refer to a data augmentation technique that creates a new image by modifying an existing image.

[0081] According to a specific embodiment, the processor (120) of the system (100) can artificially increase the size and diversity of the dataset by labeling information such as bounding boxes, class labels, and masks in the form of annotations to underground buried objects within each image included in the GPR image dataset through the Roboflow platform, and by applying various enhancement techniques such as image rotation, translation, scaling, cropping, flipping, adding noise, and adjusting brightness / contrast. This can contribute to improving the generalization performance of the model, preventing overfitting, and increasing the robustness of the model. For example, the processor (120) can perform a labeling task through the interface of the Roboflow platform to designate the pipe area in the GPR image as a polygon or rectangle and assign a class label of "pipe", and can also increase the model's adaptability to various noise environments that may occur in actual GPR data by simulating pipe images taken from various angles through data augmentation or by generating images with added noise.

[0082] According to one embodiment, the processor (120) of the system (100) may perform a preprocessing process to remove noise from GPR data. Here, noise is an unnecessary signal component included in GPR data and is a major factor that impedes the accuracy of data analysis and interpretation. Various types of noise may be mixed in GPR data, which may occur due to environmental factors (e.g., electromagnetic interference, surface reflection, changes in ground conditions, etc.), characteristics of the equipment itself (e.g., antenna noise, system noise, etc.), or errors in the data collection process. Specifically, the processor (120) may apply various preprocessing techniques to the GPR data to remove noise and improve the signal-to-noise ratio (SNR) of the data. Representative preprocessing techniques include filtering, smoothing, background removal, time gain control (TGC), and normalization, and the definitions of each technique can be explained as follows.

[0083] Filtering: A technique that removes or emphasizes specific frequency components; noise can be eliminated by designing and applying filters in the frequency domain. (Examples: bandpass filter, bandstop filter, Wiener filter)

[0084] Smoothing: This refers to a technique that reduces irregular fluctuations in data by replacing the values ​​of adjacent data points with the mean or median. (Examples: Moving average filter, median filter, Gaussian filter)

[0085] Background removal: This can refer to a technique for removing background signals that appear in a consistent pattern in GPR data, such as surface reflections and system noise.

[0086] Time gain control: Since GPR signals attenuate as they propagate underground, this refers to a technique that compensates for signal strength according to depth to better detect buried objects at deep depths.

[0087] Normalization: This refers to a technique that adjusts the range of data to a specific interval (e.g., 0–1, -1–1) to improve the stability of model training and accelerate convergence. (Examples: min-max normalization, Z-score normalization)

[0088] According to an additional embodiment, the processor (120) of the system (100) can remove impulse noise or salt-and-pepper noise by applying a median filter to the GPR data, or remove high-frequency noise by applying a Gaussian filter, and can also compensate for signal attenuation with depth through time gain control and increase model training efficiency by adjusting the data range through normalization.

[0089] According to one embodiment, underground detection data includes a hyperbola-shaped reflection signal, and an artificial intelligence model can extract location information of underground buried objects by analyzing the hyperbola-shaped reflection signal. Here, a hyperbola refers to a specific type of reflection signal pattern generated by cylindrical or spherical objects (e.g., pipes, cables, manholes, etc.) buried underground during Ground Penetrating Radar (GPR) exploration. When the reflection signal measured as a GPR antenna passes over the buried object is represented as a B-scan image (GPR cross-sectional image), it is common for such objects to appear in a hyperbolic reflection pattern, i.e., in the form of a hyperbola. Specifically, GPR is a technology that images underground structures by radiating high-frequency electromagnetic waves into the ground and receiving signals reflected due to changes in the electrical properties (mainly permittivity) of the underground medium. Since cylindrical or spherical buried objects have a different permittivity from the surrounding soil or bedrock, they reflect the GPR signal, and this reflected signal may appear in the form of a hyperbola in the B-scan image. At this time, the apex of the hyperbola represents the closest point (or shallowest point) of the buried object, and the curvature of the hyperbola may be related to the depth of the buried object. For example, referring to FIG. 3, a hyperbolic reflection pattern can be clearly seen on the GPR survey result screen, which indicates that a cylindrical buried object, such as a pipe or cable, is buried underground, and the processor (120) can analyze this hyperbola-shaped reflection signal to extract information such as the horizontal position (x, y coordinates) and depth (z coordinate) of the buried object. At this time, various algorithms can be applied to estimate the location information of the buried object by analyzing features such as the apex position, curvature, and width of the hyperbola.

[0090] According to one embodiment, an artificial intelligence model can extract location information of an underground buried object by analyzing a hyperbola-shaped reflection signal. This may mean that an AI model, particularly a deep learning-based object detection or image segmentation model, automatically recognizes a hyperbola pattern within GPR data and analyzes the features of the pattern to estimate the location coordinates (x, y, z) of the buried object. Specifically, the processor (120) can detect a hyperbola-shaped reflection signal from a GPR image or 3D volume data through an AI model and extract feature points or feature vectors such as the apex position, curvature, width, and amplitude of the hyperbola, and the deep learning model can be designed to learn the complex relationship between the hyperbola pattern and the location of the buried object using a large amount of GPR data and the location information of the buried object (ground truth data) for the data, and to accurately predict the location of the buried object for new GPR data. For example, the processor (120) can detect the bounding box of the hyperbola in the GPR image using an object detection model such as an AI model, YOLO (You Only Look Once) or Faster R-CNN, and estimate the center coordinates of the bounding box as the horizontal position (x, y coordinates) of the buried object, and also use an image segmentation model, such as a model such as Mask R-CNN or U-Net, to segment the hyperbola area into pixels, and analyze the shape and size information of the segmented area to estimate the depth (z coordinate) of the buried object.

[0091] According to one embodiment, the system (100) can be applied to various types of construction work, maintenance work, and exploration work where pre-detection of underground utilities is essential. Pre-detection of underground utilities can play a very important role in accurately identifying the location and depth of pipes, cables, conduits, tanks, etc. buried underground before excavation or drilling work, thereby preventing accidents that may occur during work (e.g., damage to utilities, gas leaks, electric shock, etc.), establishing an efficient construction plan, and improving the safety and productivity of the work. Specifically, the system (100) is mounted on a Ground Penetrating Radar (GPR) exploration device to detect underground buried objects in real time at the site and provides underground space information to the worker through AI-based analysis and visualization functions, thereby enabling the worker to make various decisions such as planning excavation paths, determining construction sequences, and establishing equipment operation plans, and can be particularly useful for the following types of construction and work.

[0092] Gas Construction: In construction work involving the installation of new gas pipelines or the replacement and repair of existing ones, accurately identifying the location of existing pipelines can prevent gas leakage accidents during excavation work and support safe construction.

[0093] Water Supply Construction: In construction work involving the installation of new water pipes or the replacement and repair of existing ones, identifying the location of existing pipes can help prevent water outages during excavation work and facilitate efficient construction.

[0094] Communication Construction: In construction work involving the installation of new communication cables or the replacement and repair of existing ones, identifying the location of existing cables can prevent communication disruptions during excavation work and increase work efficiency.

[0095] According to one embodiment, the processor (120) of the system (100) can modify an existing construction plan or establish a new, more efficient and safe construction plan using the results of underground buried object detection. This may mean predicting potential problems that may occur during the construction process in advance by utilizing AI-based accurate and reliable underground buried object information, and deriving an optimal construction plan to resolve them. Specifically, the processor (120) can analyze the detection results of the AI ​​model to identify problems in the existing construction plan (e.g., interference between the excavation path and the buried object, inefficiency of the construction sequence, etc.), propose a new construction method (e.g., non-excavation method, modular method, etc.), and establish a plan to optimize construction resources (e.g., equipment, manpower, time, cost, etc.). In this process, various technologies such as 3D simulation, interference analysis, and optimization algorithms may be utilized. For example, the processor (120) can be used to establish a construction plan that plans the optimal path for excavation work based on the results of underground buried object detection, calculates the required equipment and manpower, and predicts the construction period, and can also contribute to predicting risk factors that may occur during the construction process and establishing safety measures to prevent them (e.g., changing the excavation method, installing support structures, etc.).

[0097] According to various embodiments, an AI-based automatic detection and digital twin implementation system for underground buried objects may include a processor, wherein the processor receives underground detection data acquired from a Ground Penetrating Radar (GPR) device, generates a detection result including the presence and location information of underground buried objects within the underground detection data using a pre-trained artificial intelligence model, generates a three-dimensional digital twin model of the underground buried objects based on the detection result, visualizes the three-dimensional digital twin model and provides it through a user interface, and is configured to provide location information of the underground buried objects in the form of map data using GPS data.

[0098] According to various embodiments, the artificial intelligence model may include a YOLOv11-based object detection model, and the YOLOv11 model may be characterized by being generated by fine-tuning a pre-trained model trained on the COCO dataset to the GPR data.

[0099] According to various embodiments, the processor may be further configured to perform labeling and augmentation processing on the underground detection data using a Roboflow platform and to perform a preprocessing process to remove noise from the GPR data before receiving the underground detection data.

[0100] According to various embodiments, the processor may be further configured to implement the three-dimensional digital twin model in the form of a 3D object using the Unity engine and to virtually place the 3D object to provide a virtual design simulation prior to actual construction.

[0101] According to various embodiments, the processor may be further configured to classify the types of the underground buried objects into metallic, non-metallic, and background types.

[0102] According to various embodiments, the underground detection data may include a hyperbolic reflection signal, and the artificial intelligence model may be configured to analyze the hyperbolic reflection signal to extract location information of the underground buried object.

[0103] According to various embodiments, the processor may be further configured to provide accuracy information regarding the analysis results of the artificial intelligence model in the form of a confusion matrix or an accuracy graph, together with the three-dimensional digital twin model.

[0104] According to various embodiments, the system may be characterized as being applied to gas, water, and telecommunications construction projects requiring prior detection of underground buried objects.

[0105] According to various embodiments, the processor may be further configured to establish a new construction plan using the underground buried object detection results.

[0106] According to various embodiments, a method for AI-based automatic detection of underground buried objects and implementation of a digital twin may include: receiving underground detection data acquired from a Ground Penetrating Radar (GPR) device; generating a detection result including information on the presence and location of underground buried objects within the underground detection data using a pre-trained artificial intelligence model; generating a three-dimensional digital twin model of the underground buried objects based on the detection result; visualizing the three-dimensional digital twin model and providing it through a user interface; and providing location information of the underground buried objects in the form of map data using GPS data.

[0108] As used in this document, the terms “module” or “part” include a unit composed of hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit, for example. “Module” or “part” may be a component formed integrally or a minimum unit or part thereof that performs one or more functions. “Module” or “part” may be implemented mechanically or electronically and may include, for example, an application-specific integrated circuit (ASIC) chip, field-programmable gate arrays (FPGAs), or programmable logic device known or to be developed that performs certain operations, and may be executed by a processor (120). At least part of the device (e.g., modules or functions thereof) or method (e.g., operations) according to various embodiments may be implemented as instructions stored in a computer-readable storage medium (e.g., memory (130)) in the form of a program module. When the above instruction is executed by a processor (e.g., processor (120)), the processor may perform a function corresponding to the above instruction. Computer-readable recording media may include a hard disk, a floppy disk, a magnetic medium (e.g., magnetic tape), an optical recording medium (e.g., CD-ROM, DVD, magneto-optical medium (e.g., floptical disk), built-in memory, etc. Instructions may include code generated by a compiler or code that can be executed by an interpreter. A module or program module according to various embodiments may include at least one of the aforementioned components, some of which may be omitted, or additionally include other components. Operations performed by a module, program module, or other components according to various embodiments may be executed sequentially, in parallel, iteratively, or heuristically, or at least some operations may be executed in a different order, omitted, or other operations may be added.

[0109] Furthermore, the embodiments disclosed in this document are presented for the purpose of explaining and understanding the disclosed technical content and are not intended to limit the scope of this disclosure. Accordingly, the scope of this disclosure should be interpreted to include all modifications or various other embodiments based on the technical concept of this disclosure.

Claims

Claim 1 A system for AI-based automatic detection of underground buried objects and digital twin implementation, comprising a processor, wherein the processor receives underground detection data acquired from a Ground Penetrating Radar (GPR) device, generates a detection result including the presence and location information of underground buried objects within the underground detection data using a pre-trained artificial intelligence model, generates a 3D digital twin model of the underground buried objects based on the detection result, visualizes the 3D digital twin model and provides it through a user interface, and is configured to provide location information of the underground buried objects in the form of map data using GPS data. Claim 2 A system according to claim 1, wherein the artificial intelligence model includes a YOLOv11-based object detection model, and the YOLOv11 model is generated by fine-tuning a pre-trained model trained on the COCO dataset onto GPR data. Claim 3 A system according to claim 1, characterized in that the processor is further configured to perform labeling and augmentation processing on the underground detection data using a Roboflow platform and to perform a preprocessing process to remove noise from the GPR data before receiving the underground detection data. Claim 4 A system according to claim 1, characterized in that the processor is further configured to implement the three-dimensional digital twin model in the form of a 3D object using the Unity engine and to virtually place the 3D object to provide a virtual design simulation prior to actual construction. Claim 5 A system according to claim 1, wherein the processor is further configured to classify the types of the underground buried objects into metallic, non-metallic, and background. Claim 6 A system according to claim 1, characterized in that the underground detection data includes a hyperbolic reflection signal, and the artificial intelligence model is configured to analyze the hyperbolic reflection signal to extract location information of an underground buried object. Claim 7 A system according to claim 1, characterized in that the processor is further configured to provide accuracy information regarding the analysis results of the artificial intelligence model in the form of a confusion matrix or an accuracy graph, together with the three-dimensional digital twin model. Claim 8 In claim 1, the system is characterized by being applied to gas, water, and telecommunications construction projects requiring prior detection of underground buried objects. Claim 9 A system according to claim 1, characterized in that the processor is further configured to establish a new construction plan using the underground buried object detection results. Claim 10 A method for AI-based automatic detection of underground buried objects and implementation of a digital twin, comprising: receiving underground detection data acquired from a Ground Penetrating Radar (GPR) device; generating a detection result including the presence and location information of an underground buried object within the underground detection data using a pre-trained artificial intelligence model; generating a three-dimensional digital twin model of the underground buried object based on the detection result; visualizing the three-dimensional digital twin model and providing it through a user interface; and providing the location information of the underground buried object in the form of map data using GPS data.

Images