• The Journal of the Acoustical Society of Korea
• /
• v.41 no.4
• /
• pp.426-434
• /
• 2022

Bubbles generated by various causes in the ocean are known to persist for long periods of time. Although the volume occupied by bubbles in the ocean is small, the presence of bubbles in ocean due to resonance and attenuation greatly affects the acoustic properties. Accordingly, bistatic reverberation experiment was performed in the ocean where artificial bubbles exist. A number of transducers and receivers were installed on 6 buoys arranged in a hexagonal shape, and blowing agents were dropped in the center of the buoy to generate bubbles. For reverberation modeling that reflects acoustic characteristics changed by bubbles, the spatial distribution of bubbles was estimated using video data and received signals. A measurement-based bubble spectral shape was used, and it was assumed that the bubble density within the spatial distribution of the estimated bubble was the same. As a result, it was confirmed that the bubble reverberation was simulated in a time similar to the measured data regardless of the bubble density, and the bubble reverberation level similar to the measured data was simulated at a void fraction of about 10^-7 ~ 10^-6.8.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_4
• /
• pp.1889-1900
• /
• 2022

Roads are an essential factor in the physical functioning of modern society. The spatial information of the road has much longer update cycle than the traffic situation information, and it is necessary to generate the information faster and more accurately than now. In this study, as a way to achieve that goal, the Pan-sharpening technique was applied to satellite images of Kompsat 3 and 3A to improve spatial resolution. Then, the data were used for road extraction using the semantic segmentation technique, which has been actively researched recently. The acquired Kompsat 3/3A pan-sharpened images were trained by putting it into a U-Net based segmentation model along with Massachusetts road data, and the applicability of the images were evaluated. As a result of training and verification, it was found that the model prediction performance was maintained as long as certain conditions were maintained for the input image. Therefore, it is expected that the possibility of utilizing satellite images such as Kompsat satellite will be even higher if rich training data are constructed by applying a method that minimizes the impact of surrounding environmental conditions affecting models such as shadows and surface conditions.

• The Journal of the Korea Contents Association
• /
• v.22 no.3
• /
• pp.354-360
• /
• 2022

Metabolic disease is a disease with a prevalence of 26% in Korean, and has three of the five states of abdominal obesity, hypertension, hunger glycemic disorder, high neutral fat, and low HDL cholesterol at the same time. This paper links the consumer panel data of the Rural Development Agency(RDA) and the medical care data of the National Health Insurance Service(NHIS) to generate a classification model that can be divided into a metabolic disease group and a control group through food consumption characteristics, and attempts to compare the differences. Many existing domestic and foreign studies related to metabolic diseases and food consumption characteristics are disease correlation studies of specific food groups and specific ingredients, and this paper is logistic considering all food groups included in the general diet. We created a classification model using regression, a decision tree-based classification model, and a classification model using XGBoost. Of the three models, the high-precision model is the XGBoost classification model, but the accuracy was not high at less than 0.7. As a future study, it is necessary to extend the observation period for food consumption in the patient group to more than 5 years and to study the metabolic disease classification model after converting the food consumed into nutritional characteristics.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.39 no.6
• /
• pp.429-435
• /
• 2021

The integrated underground space map shows the underground facility(water supply, sewage, gas, electric power, communication, heating), underground structures (subway, underpass, underground walkway, underground parking lot, underground shopping mall, common ward), ground(drilling, coffin, geology) refers to a map constructed so that a total of 15 types of underground information can be checked at a glance on a three-dimensional basis. The purpose of this study is to develop a technology to correct the problem of curved surface processing and the unevenness of underground facility pipelines that occur in converting 2D underground facility data into 3D-based underground space integrated map(3D underground facility model). do it with. To this end, we first investigated and reviewed the domestic and foreign status of technologies that generate data on underground facilities based on three dimensions, and developed a surface correction algorithm and an unevenness correction algorithm to solve practical problems. Algorithms to verify the developed algorithm This applied correction program was developed. Based on the above process, the three-dimensional model of the underground facility could be produced identically to reality. This study is judged to have significance as a basic study to improve the utilization of the underground spatial integration map.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_1
• /
• pp.485-496
• /
• 2022

This study is for the software implementation to generate atmospheric and surface reflectance products from RapidEye satellite imagery. The software is an extension based on Orfeo Toolbox (OTB) and an open-source remote sensing software including calibration modules which use an absolute atmospheric correction algorithm. In order to verify the performance of the program, the accuracy of the product was validated by a test image on the Radiometric Calibration Network (RadCalNet) site. In addition, the accuracy of the surface reflectance product generated from the KOMPSAT-3A image, the surface reflectance of Landsat Analysis Ready Data (ARD) of the same site, and near acquisition date were compared with RapidEye-based one. At the same time, a comparative study was carried out with the processing results using QUick Atmospheric Correction (QUAC) and Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) tool supported by a commercial tool for the same image. Similar to the KOMPSAT-3A-based surface reflectance product, the results obtained from RapidEye Extension showed accuracy of agreement level within 5%, compared with RadCalNet data. They also showed better accuracy in all band images than the results using QUAC or FLAASH tool. As the importance of the Red-Edge band in agriculture, forests, and the environment applications is being emphasized, it is expected that the utilization of the surface reflectance products of RapidEye images produced using this program will also increase.

• Journal of Korean Society of Forest Science
• /
• v.111 no.3
• /
• pp.435-449
• /
• 2022

Forest canopy height is an indispensable vertical structure parameter that can be used for understanding forest biomass and carbon storage as well as for managing a sustainable forest ecosystem. Plot-based field surveys, such as the national forest inventory, have been conducted to provide estimates of the forest canopy height. However, the comprehensive nationwide field monitoring of forest canopy height has been limited by its cost, lack of spatial coverage, and the inaccessibility of some forested areas. These issues can be addressed by remote sensing technology, which has gained popularity as a means to obtain detailed 2- and 3-dimensional measurements of the structure of the canopy at multiple scales. Here, we reviewed both international and domestic studies that have used remote sensing technology approaches to estimate the forest canopy height. We categorized and examined previous approaches as: 1) LiDAR approach, 2) Stereo or SAR image-based point clouds approach, and 3) combination approach of remote sensing data. We also reviewed upscaling approaches of utilizing remote sensing data to generate a continuous map of canopy height across large areas. Finally, we provided suggestions for further advancement of the Korean forest canopy height estimation system through the use of various remote sensing technologies.

• Journal of Intelligence and Information Systems
• /
• v.29 no.3
• /
• pp.419-437
• /
• 2023

As the number of thin filers in Korea surpasses 12 million, there is a growing interest in enhancing the accuracy of assessing their credit default risk to generate additional revenue. Specifically, researchers are actively pursuing the development of default prediction models using machine learning and deep learning algorithms, in contrast to traditional statistical default prediction methods, which struggle to capture nonlinearity. Among these efforts, Graph Neural Network (GNN) architecture is noteworthy for predicting default in situations with limited data on thin filers. This is due to their ability to incorporate network information between borrowers alongside conventional credit-related data. However, prior research employing graph neural networks has faced limitations in effectively handling diverse categorical variables present in credit information. In this study, we introduce the Transformer embedded Graph Convolutional Network (TeGCN), which aims to address these limitations and enable effective default prediction for thin filers. TeGCN combines the TabTransformer, capable of extracting contextual information from categorical variables, with the Graph Convolutional Network, which captures network information between borrowers. Our TeGCN model surpasses the baseline model's performance across both the general borrower dataset and the thin filer dataset. Specially, our model performs outstanding results in thin filer default prediction. This study achieves high default prediction accuracy by a model structure tailored to characteristics of credit information containing numerous categorical variables, especially in the context of thin filers with limited data. Our study can contribute to resolving the financial exclusion issues faced by thin filers and facilitate additional revenue within the financial industry.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.36 no.5
• /
• pp.323-330
• /
• 2023

The development of an analysis model that reflects the microstructure characteristics of polyvinyl alcohol (PVA) fiber-reinforced cementitious composites, which have a highly complex microstructure, enables synergy between efficient material design and real experiments. PVA fiber orientations are an important factor that influences the mechanical behavior of PVA fiber-reinforced cementitious composites. Owing to the difficulty in distinguishing the gray level value obtained from micro-CT images of PVA fibers from adjacent phases, fiber segmentation is time-consuming work. In this study, a micro-CT test with a voxel size of 0.65 ㎛³ was performed to investigate the three-dimensional distribution of fibers. To segment the fibers and generate training data, histogram, morphology, and gradient-based phase-segmentation methods were used. A U-net model was proposed to segment fibers from micro-CT images of PVA fiber-reinforced cementitious composites. Data augmentation was applied to increase the accuracy of the training, using a total of 1024 images as training data. The performance of the model was evaluated using accuracy, precision, recall, and F1 score. The trained model achieved a high fiber segmentation performance and efficiency, and the approach can be applied to other specimens as well.

• Journal of the Institute of Convergence Signal Processing
• /
• v.24 no.3
• /
• pp.160-165
• /
• 2023

In this study, we propose an SNS sentence writing assistance system that utilizes YOLO and GPT to assist users in writing texts with images, such as SNS. We utilize the YOLO model to extract objects from images inserted during writing, and also extract meta-information such as GPS information and creation time information, and use them as prompt values for GPT. To use the YOLO model, we trained it on form image data, and the mAP score of the model is about 0.25 on average. GPT was trained on 1,000 blog text data with the topic of 'restaurant reviews', and the model trained in this study was used to generate sentences with two types of keywords extracted from the images. A survey was conducted to evaluate the practicality of the generated sentences, and a closed-ended survey was conducted to clearly analyze the survey results. There were three evaluation items for the questionnaire by providing the inserted image and keyword sentences. The results showed that the keywords in the images generated meaningful sentences. Through this study, we found that the accuracy of image-based sentence generation depends on the relationship between image keywords and GPT learning contents.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.7
• /
• pp.11-20
• /
• 2024

In this paper, we propose the BAG (Bit-width Aware Generator) and the Intermediate Layer Knowledge Distillation using Channel-wise Attention to reduce the knowledge gap between a quantized network, a full-precision network, and a generator in GDFQ (Generative Data-Free Quantization). Since the generator in GDFQ is only trained by the feedback from the full-precision network, the gap resulting in decreased capability due to low bit-width of the quantized network has no effect on training the generator. To alleviate this problem, BAG is quantized with same bit-width of the quantized network, and it can generate synthetic images, which are effectively used for training the quantized network. Typically, the knowledge gap between the quantized network and the full-precision network is also important. To resolve this, we compute channel-wise attention of outputs of convolutional layers, and minimize the loss function as the distance of them. As the result, the quantized network can learn which channels to focus on more from mimicking the full-precision network. To prove the efficiency of proposed methods, we quantize the network trained on CIFAR-100 with 3 bit-width weights and activations, and train it and the generator with our method. As the result, we achieve 56.14% Top-1 Accuracy and increase 3.4% higher accuracy compared to our baseline AdaDFQ.