• Journal of Korea Water Resources Association
• /
• v.54 no.12
• /
• pp.1255-1263
• /
• 2021

While it could become an alternative water resource, fog could undermine traffic safety and operational performance of infrastructures. To reduce such adverse impacts, it is necessary to have spatially continuous fog risk information. In this work, tree-based machine-learning models were developed in order to quantify fog risks with routine meteorological observations alone. The Extreme Gradient Boosting (XGB), Light Gradient Boosting (LGB), and Random Forests (RF) were chosen for the regional fog models using operational weather and visibility observations within the Jeollabuk-do province. Results showed that RF seemed to show the most robust performance to categorize between fog and non-fog situations during the training and evaluation period of 2017-2019. While the LGB performed better than in predicting fog occurrences than the others, its false alarm ratio was the highest (0.695) among the three models. The predictability of the three models considerably declined when applying them for an independent period of 2020, potentially due to the distinctively enhanced air quality in the year under the global lockdown. Nonetheless, even in 2020, the three models were all able to produce fog risk information consistent with the spatial variation of observed fog occurrences. This work suggests that the tree-based machine learning models could be used as tools to find locations with relatively high fog risks.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.20 no.6
• /
• pp.1161-1175
• /
• 2018

An unexpected event could be easily followed by a large secondary accident due to the limitation in sight of drivers in road tunnels. Therefore, a series of automated incident detection systems have been under operation, which, however, appear in very low detection rates due to very low image qualities on CCTVs in tunnels. In order to overcome that limit, deep learning based tunnel incident detection system was developed, which already showed high detection rates in November of 2017. However, since the object detection process could deal with only still images, moving direction and speed of moving vehicles could not be identified. Furthermore it was hard to detect stopping and reverse the status of moving vehicles. Therefore, apart from the object detection, an object tracking method has been introduced and combined with the detection algorithm to track the moving vehicles. Also, stopping-reverse discrimination algorithm was proposed, thereby implementing into the combined incident detection processes. Each performance on detection of stopping, reverse driving and fire incident state were evaluated with showing 100% detection rate. But the detection for 'person' object appears relatively low success rate to 78.5%. Nevertheless, it is believed that the enlarged richness of image big-data could dramatically enhance the detection capacity of the automatic incident detection system.

• Journal of the Korea Convergence Society
• /
• v.10 no.7
• /
• pp.15-21
• /
• 2019

Research in detecting fake information gained a lot of interest after the US presidential election in 2016. Information from unknown sources are produced in the shape of news, and its rapid spread is fueled by the interest of public drawn to stimulating and interesting issues. In addition, the wide use of mass communication platforms such as social network services makes this phenomenon worse. Poynter Institute created the International Fact Checking Network (IFCN) to provide guidelines for judging the facts of skilled professionals and releasing "Code of Ethics" for fact check agencies. However, this type of approach is costly because of the large number of experts required to test authenticity of each article. Therefore, research in automated fake news detection technology that can efficiently identify it is gaining more attention. In this paper, we investigate fake news detection systems and researches that are rapidly developing, mainly thanks to recent advances in deep learning technology. In addition, we also organize shared tasks and training corpus that are released in various forms, so that researchers can easily participate in this field, which deserves a lot of research effort.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.23 no.1
• /
• pp.34-45
• /
• 2021

Weather is one of the important factors in the agricultural industry as it affects the price, production, and quality of crops. Upland crops are directly exposed to the natural environment because they are mainly grown in mountainous areas. Therefore, it is necessary to provide accurate weather for upland crops. This study examined the effectiveness of 12 forest soil factors to interpolate the weather in mountainous areas. The daily temperature and precipitation were collected by the Korea Meteorological Administration between January 2009 and December 2018. The Generalized Additive Model (GAM), Kriging, and Random Forest (RF) were considered to interpolate. For evaluating the interpolation performance, automatic weather stations were used as training data and automated synoptic observing systems were used as test data for cross-validation. Unfortunately, the forest soil factors were not significant to interpolate the weather in the mountainous areas. GAM with only geography aspects showed that it can interpolate well in terms of root mean squared error and mean absolute error. The significance of the factors was tested at the 5% significance level in GAM, and the climate zone code (CLZN_CD) and soil water code B (SIBFLR_LAR) were identified as relatively important factors. It has shown that CLZN_CD could help to interpolate the daily average and minimum daily temperature for upland crops.

• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.5
• /
• pp.157-166
• /
• 2022

Exercise is necessary for a healthy life, but it is recommended that it be conducted in a non-face-to-face environment in the context of an epidemic such as COVID-19. However, in the existing non-face-to-face exercise content, it is possible to recognize exercise movements, but the process of interpreting and providing feedback information is not automated. Therefore, in this paper, to solve this problem, we propose a method of creating a formalized rule to track the contents of exercise and the motions that constitute it. To make such a rule, first make a rule for the overall exercise content, and then create a tracking rule for the motions that make up the exercise. A motion tracking rule can be created by dividing the motion into steps and defining a key frame pose that divides the steps, and creating a transition rule between states and states represented by the key frame poses. The rules created in this way are premised on the use of posture and motion recognition technology using motion capture equipment, and are used for logical development for automation of application of these technologies. By using the rules proposed in this paper, not only recognizing the motions appearing in the exercise process, but also automating the interpretation of the entire motion process, making it possible to produce more advanced contents such as an artificial intelligence training system. Accordingly, the quality of feedback on the exercise process can be improved.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.22 no.2
• /
• pp.164-181
• /
• 2023

In a low speed limit environment, the speed limit tolerance of automated traffic enforcement devices is very high, which is one of the main factors for the low compliance rate. Therefore, in this study, we aimed to the improve the speed limit tolerance and to present a new standard. The effects of the operator and user errors that can cause speeding by drivers were analyzed. Based on the results of the analysis, an improvement of the tolerance was proposed by applying an error in the enforcement device and GPS speed. In addition, long-term expected safety effects such as the accident rate and severity were estimated from the operator's perspective when improving the tolerance. As a result of the estimation, the speed limit compliance rate, accident rate, and change rate of a number of severe accidents due to speed change, and pedestrian traffic accident mortality rate were all improved in all speed limit environments. The introduction of the proposed improvement is expected to improve road safety significantly.

• Journal of The Korean Association For Science Education
• /
• v.43 no.6
• /
• pp.573-582
• /
• 2023

This study aimed to explore the potential use of artificial intelligence in science education for gifted students by analyzing the structure of abstracts written by students at a gifted science academy and comparing the performance of various elements extracted using AI. The study involved an analysis of 263 graduation theses from S Science High School over five years (2017-2021), focusing on the frequency and types of background, objectives, methods, results, and discussions included in their abstracts. This was followed by an evaluation of their accuracy using AI classification methods with fine-tuning and prompts. The results revealed that the frequency of elements in the abstracts written by gifted students followed the order of objectives, methods, results, background, and discussions. However, only 57.4% of the abstracts contained all the essential elements, such as objectives, methods, and results. Among these elements, fine-tuned AI classification showed the highest accuracy, with background, objectives, and results demonstrating relatively high performance, while methods and discussions were often inaccurately classified. These findings suggest the need for a more effective use of AI, through providing a better distribution of elements or appropriate datasets for training. Educational implications of these findings were also discussed.

• Journal of Navigation and Port Research
• /
• v.47 no.6
• /
• pp.437-447
• /
• 2023

By actively adopting technologies from the Fourth Industrial Revolution, the port industry is trending toward new types of ports, such as automated and smart ports. However, behind the development of these ports, there is an increasing risk of cyber security incidents and threats within ports and container terminals, including information leakage through cargo handling equipment and ransomware attacks leading to disruptions in terminal operations. Despite the necessity of research to enhance cyber security within ports, there is a lack of such studies in the domestic context. This study focuses on Busan Port, a representative port in South Korea that actively incorporates technology from the Fourth Industrial Revolution, in order to discover variables for improving cyber security in container terminals. The research results categorized factors for enhancing cyber security in Busan Port's container terminals into network construction and policy support, standardization of education and personnel training, and legal and regulatory factors. Subsequently, multiple regression analysis was conducted based on these factors, leading to the identification of detailed factors for securing and enhancing safety, reliability, performance, and satisfaction in Busan Port's container terminals. The significance of this study lies in providing direction for enhancing cyber security in Busan Port's container terminals and addressing the increasing incidents of cyber security attacks within ports and container terminals.

• Imaging Science in Dentistry
• /
• v.54 no.1
• /
• pp.81-91
• /
• 2024

Purpose: The objective of this study was to propose a deep-learning model for the detection of the mandibular canal on dental panoramic radiographs. Materials and Methods: A total of 2,100 panoramic radiographs (PANs) were collected from 3 different machines: RAYSCAN Alpha (n=700, PAN A), OP-100 (n=700, PAN B), and CS8100 (n=700, PAN C). Initially, an oral and maxillofacial radiologist coarsely annotated the mandibular canals. For deep learning analysis, convolutional neural networks (CNNs) utilizing U-Net architecture were employed for automated canal segmentation. Seven independent networks were trained using training sets representing all possible combinations of the 3 groups. These networks were then assessed using a hold-out test dataset. Results: Among the 7 networks evaluated, the network trained with all 3 available groups achieved an average precision of 90.6%, a recall of 87.4%, and a Dice similarity coefficient (DSC) of 88.9%. The 3 networks trained using each of the 3 possible 2-group combinations also demonstrated reliable performance for mandibular canal segmentation, as follows: 1) PAN A and B exhibited a mean DSC of 87.9%, 2) PAN A and C displayed a mean DSC of 87.8%, and 3) PAN B and C demonstrated a mean DSC of 88.4%. Conclusion: This multi-device study indicated that the examined CNN-based deep learning approach can achieve excellent canal segmentation performance, with a DSC exceeding 88%. Furthermore, the study highlighted the importance of considering the characteristics of panoramic radiographs when developing a robust deep-learning network, rather than depending solely on the size of the dataset.

• Korean Journal of Radiology
• /
• v.22 no.3
• /
• pp.476-488
• /
• 2021

Objective: We aimed to develop a deep neural network for segmenting lung parenchyma with extensive pathological conditions on non-contrast chest computed tomography (CT) images. Materials and Methods: Thin-section non-contrast chest CT images from 203 patients (115 males, 88 females; age range, 31-89 years) between January 2017 and May 2017 were included in the study, of which 150 cases had extensive lung parenchymal disease involving more than 40% of the parenchymal area. Parenchymal diseases included interstitial lung disease (ILD), emphysema, nontuberculous mycobacterial lung disease, tuberculous destroyed lung, pneumonia, lung cancer, and other diseases. Five experienced radiologists manually drew the margin of the lungs, slice by slice, on CT images. The dataset used to develop the network consisted of 157 cases for training, 20 cases for development, and 26 cases for internal validation. Two-dimensional (2D) U-Net and three-dimensional (3D) U-Net models were used for the task. The network was trained to segment the lung parenchyma as a whole and segment the right and left lung separately. The University Hospitals of Geneva ILD dataset, which contained high-resolution CT images of ILD, was used for external validation. Results: The Dice similarity coefficients for internal validation were 99.6 ± 0.3% (2D U-Net whole lung model), 99.5 ± 0.3% (2D U-Net separate lung model), 99.4 ± 0.5% (3D U-Net whole lung model), and 99.4 ± 0.5% (3D U-Net separate lung model). The Dice similarity coefficients for the external validation dataset were 98.4 ± 1.0% (2D U-Net whole lung model) and 98.4 ± 1.0% (2D U-Net separate lung model). In 31 cases, where the extent of ILD was larger than 75% of the lung parenchymal area, the Dice similarity coefficients were 97.9 ± 1.3% (2D U-Net whole lung model) and 98.0 ± 1.2% (2D U-Net separate lung model). Conclusion: The deep neural network achieved excellent performance in automatically delineating the boundaries of lung parenchyma with extensive pathological conditions on non-contrast chest CT images.