KR-102960297-B1 - Method and Apparatus for Artificial Intelligence Processing of 3D GPR Data Based on MFCC Image

Abstract

The present invention relates to a method and apparatus for artificial intelligence processing of 3D GPR data based on MFCC images. The method according to the present invention comprises the steps of: extracting 3D GPR data of a predetermined size from 3D GPR data acquired from a multi-channel GPR device; reconstructing the extracted 3D GPR data into 1D signal waveform data to generate super waveform data; converting the super waveform data into MFCC image data using an MFCC technique; and inputting the MFCC image data into a pre-trained artificial intelligence model to output a detection result of an underground target. According to the present invention, by using an artificial intelligence model trained based on MFCC images, it is possible to detect underground targets accurately by increasing the detection rate and lowering the false positive rate compared to conventional methods.

Inventors

• 김상욱
• 최종성
• 성현모

Assignees

• 이성 주식회사

Dates

Publication Date

20260507

Application Date

20241125

Priority Date

20231218

Claims (6)

1. A step of extracting 3D GPR data of a predetermined size from 3D GPR data acquired from a multi-channel GPR device, A step of generating superwaveform data by reconstructing the extracted 3D GPR data into 1D signal waveform data, A step of converting the above superwaveform data into MFCC image data through the MFCC technique, A step of inputting the above MFCC image data into a pre-trained artificial intelligence model to output a detection result for an underground target. MFCC image-based 3D GPR data artificial intelligence processing method including
2. In Article 1, The step of extracting 3D GPR data of a predetermined size from 3D GPR data acquired from the multi-channel GPR device above is: A step of setting an extraction region of a preset size in three-dimensional GPR data acquired from the above multi-channel GPR device, A step of extracting 3D GPR data within the above extraction area, A step of setting the next extraction area by moving the extraction area by a predetermined interval in the sweep direction of the multi-channel GPR sensor equipped in the multi-channel GPR device, Step of extracting the next 3D GPR data from the above next extraction area Includes, The size of the above extraction area is set to a size that can include the largest target among the detected targets, and MFCC image-based 3D GPR data artificial intelligence processing method in which the movement interval of the above extraction area is set such that there exists an overlapping section between the previous extraction area and the next extraction area.
3. In Paragraph 2, It further includes the step of training the above artificial intelligence model, and The step of training the above artificial intelligence model is, A step of extracting a region labeled with a target class from 3D GPR data collected using the above-described multi-channel GPR device as training 3D GPR data, A step of converting multiple GPR cross-sectional image data included in the extracted 3D GPR training data into 1D waveform signal data, sequentially concatenating them in array order to generate 1D superwaveform data, and then generating MFCC images for each target class converted through the MFCC technique. A method for artificial intelligence processing 3D GPR data based on MFCC images, comprising the step of training an artificial intelligence model using the above-mentioned MFCC images for each target class.
4. In Paragraph 3, The above artificial intelligence model is, Artificial intelligence processing method for 3D GPR data based on MFCC images, which is any one of CNN, ResNet, DenseNet, Inception, VGGNet, EfficientNet, ANN, and YOLO.
5. A computer-readable recording medium having a computer program that executes the MFCC image-based three-dimensional GPR data artificial intelligence processing method described in any one of claims 1 to 4 when executed by a processor.
6. In computing devices, Includes a processor and memory, A computing device that executes the MFCC image-based 3D GPR data artificial intelligence processing method described in any one of claims 1 to 4 when a computer program stored in the memory is executed by the processor.

Description

Method and Apparatus for Artificial Intelligence Processing of 3D GPR Data Based on MFCC Image The present invention relates to a method and apparatus for artificial intelligence processing of 3D GPR data, and more specifically, to a method and apparatus for artificial intelligence processing of 3D GPR data based on MFCC images. [National Project Information] [Ministry Name] Ministry of Science and ICT [Name of Project Management (Specialized) Agency] Korea Information & Communication Technology Industry Promotion Agency [Research Project Name] AI-Converged Coastal Guard and Mine Detection System [Research Project Title] Development and Demonstration Project of AI-Integrated Mine Detection System [Contribution Rate] 1/1 [Name of Project Performing Organization] Lee Sung Co., Ltd. [Research Period] May 1, 2021 – December 31, 2023 Target detection techniques utilizing Ground Penetrating Radar (GPR) data are employed in various fields and play a crucial role, particularly in detecting dangerous objects such as landmines. Recently, with the introduction of Artificial Intelligence (AI) technology into this target detection process, active efforts are being made to improve the accuracy of target detection. In single-channel GPR, artificial intelligence is applied based on longitudinal images, but in multi-channel GPR, a technique is used to apply artificial intelligence based on 3D GPR data, including planar, longitudinal, and transverse images. In multi-channel GPR, images of a target appear in planar, longitudinal, and transverse planes. Therefore, there is a risk that applying artificial intelligence based on cross-sectional images may result in insufficient target identification. Accordingly, AI techniques must be applied independently to each of the three planes. However, this approach requires preparing separate training data models for each plane during the learning process, which may lead to reduced training efficiency. Furthermore, as artificial intelligence algorithms for target detection are applied independently to each cross-section, different inference results may arise from each section. This is a major cause of increased false positive rates and decreased detection rates. In other words, if consistent results are not obtained across all three cross-sections for the same target, detection accuracy declines. This issue is a major challenge facing existing GPR data-based target detection technologies, and additional technological development is required to address it. FIG. 1 is a drawing provided to explain the process of detecting buried landmines with a multi-channel GPR device according to one embodiment of the present invention. FIG. 2 is a configuration diagram of a multi-channel GPR device according to one embodiment of the present invention. FIG. 3 is a diagram exemplarily illustrating the process of generating three-dimensional GPR data in a multi-channel GPR device according to an embodiment of the present invention. FIG. 4 is a flowchart provided to explain the artificial intelligence model learning process according to one embodiment of the present invention. Figure 5 is a diagram provided to explain the process of converting three-dimensional GPR data containing a target signal into a single MFCC image. Figure 6 is a diagram showing the detailed process of an MFCC operation that generates an MFCC image from superwaveform data. FIG. 7 is a drawing provided to explain the process of extracting three-dimensional GPR data of a certain size for input to an artificial intelligence model according to one embodiment of the present invention. Then, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. The terms used in this specification are for describing embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms "comprises" and/or "comprising" used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and "and/or" includes each of the mentioned components and all combinations of one or more. Although terms such as "first," "second," etc., are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another. Therefore, the first component mentioned below may be the second component within the technical scope of the invention. In this specification, the term "computing device" includes all various devices capable of performing computational processing and providing results to a user. For example, a computing device may include desktop PCs, notebook computers, and server computers, as well as sm