• Journal of the korean academy of Pediatric Dentistry
• /
• v.49 no.2
• /
• pp.131-139
• /
• 2022

This study aimed to evaluate the effectiveness of deep convolutional neural networks (CNNs) for diagnosis of interproximal caries in pediatric intraoral radiographs. A total of 500 intraoral radiographic images of first and second primary molars were used for the study. A CNN model (Resnet 50) was applied for the detection of proximal caries. The diagnostic accuracy, sensitivity, specificity, receiver operating characteristic (ROC) curve, and area under ROC curve (AUC) were calculated on the test dataset. The diagnostic accuracy was 0.84, sensitivity was 0.74, and specificity was 0.94. The trained CNN algorithm achieved AUC of 0.86. The diagnostic CNN model for pediatric intraoral radiographs showed good performance with high accuracy. Deep learning can assist dentists in diagnosis of proximal caries lesions in pediatric intraoral radiographs.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.25 no.6
• /
• pp.509-523
• /
• 2023

The high-level nuclear waste (HLW) repository is exposed to complex environmental conditions consisting of high temperature, high humidity, and radiation, resulting in structural deterioration. Therefore, structural health monitoring is essential, and piezo sensors are used to detect cracks and estimate strength. However, since the monitoring sensors installed in the disposal tunnel and disposal container cannot be replaced or removed, the quantitative life of the monitoring sensor and its suitability must be assessed. In this study, the life of a piezo sensor for monitoring was assessed using an accelerated life test (ALT). The failure mode and mechanism of the piezo sensor under high temperature conditions were determined, and temperature stress's influence on the piezo sensor's life was analyzed. ALT was conducted on temperature stress and the relationship between temperature stress and piezo sensor life was suggested. The life of the piezo sensor was assessed using the Weibull probability distribution and the Arrhenius acceleration model. The suggested relationship can be used in multiple stress ALT designs for more precise life assessment.

• Korean Journal of Remote Sensing
• /
• v.36 no.5_3
• /
• pp.1109-1123
• /
• 2020

In South Korea with forest as a major land cover class (over 60% of the country), many wildfires occur every year. Wildfires weaken the shear strength of the soil, forming a layer of soil that is vulnerable to landslides. It is important to identify the severity of a wildfire as well as the burned area to sustainably manage the forest. Although satellite remote sensing has been widely used to map wildfire severity, it is often difficult to determine the severity using only the temporal change of satellite-derived indices such as Normalized Difference Vegetation Index (NDVI) and Normalized Burn Ratio (NBR). In this study, we proposed an approach for determining wildfire severity based on machine learning through the synergistic use of Sentinel-1A Synthetic Aperture Radar-C data and Sentinel-2A Multi Spectral Instrument data. Three wildfire cases-Samcheok in May 2017, Gangreung·Donghae in April 2019, and Gosung·Sokcho in April 2019-were used for developing wildfire severity mapping models with three machine learning algorithms (i.e., Random Forest, Logistic Regression, and Support Vector Machine). The results showed that the random forest model yielded the best performance, resulting in an overall accuracy of 82.3%. The cross-site validation to examine the spatiotemporal transferability of the machine learning models showed that the models were highly sensitive to temporal differences between the training and validation sites, especially in the early growing season. This implies that a more robust model with high spatiotemporal transferability can be developed when more wildfire cases with different seasons and areas are added in the future.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.23 no.3
• /
• pp.149-155
• /
• 2021

Early warning systems for weather risk management in the agricultural sector have been developed to predict potential wind damage to crops. These systems take into account the daily maximum wind speed to determine the critical wind speed that causes fruit drops and provide the weather risk information to farmers. In an effort to increase the accuracy of wind risk predictions, an artificial neural network for binary classification was implemented. In the present study, the daily wind speed and other weather data, which were measured at weather stations at sites of interest in Jeollabuk-do and Jeollanam-do as well as Gyeongsangbuk- do and part of Gyeongsangnam- do provinces in 2019, were used for training the neural network. These weather stations include 210 synoptic and automated weather stations operated by the Korean Meteorological Administration (KMA). The wind speed data collected at the same locations between January 1 and December 12, 2020 were used to validate the neural network model. The data collected from December 13, 2020 to February 18, 2021 were used to evaluate the wind risk prediction performance before and after the use of the artificial neural network. The critical wind speed of damage risk was determined to be 11 m/s, which is the wind speed reported to cause fruit drops and damages. Furthermore, the maximum wind speeds were expressed using Weibull distribution probability density function for warning of wind damage. It was found that the accuracy of wind damage risk prediction was improved from 65.36% to 93.62% after re-classification using the artificial neural network. Nevertheless, the error rate also increased from 13.46% to 37.64%, as well. It is likely that the machine learning approach used in the present study would benefit case studies where no prediction by risk warning systems becomes a relatively serious issue.

• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.11
• /
• pp.471-480
• /
• 2023

The Korean construction order volume in South Korea grew significantly from 91.3 trillion won in public orders in 2013 to a total of 212 trillion won in 2021, particularly in the private sector. As the size of the domestic and overseas markets grew, the scale and complexity of EPC (Engineering, Procurement, Construction) projects increased, and risk management of project management and ITB (Invitation to Bid) documents became a critical issue. The time granted to actual construction companies in the bidding process following the EPC project award is not only limited, but also extremely challenging to review all the risk terms in the ITB document due to manpower and cost issues. Previous research attempted to categorize the risk terms in EPC contract documents and detect them based on AI, but there were limitations to practical use due to problems related to data, such as the limit of labeled data utilization and class imbalance. Therefore, this study aims to develop an AI model that can categorize the contract terms based on the FIDIC Yellow 2017(Federation Internationale Des Ingenieurs-Conseils Contract terms) standard in detail, rather than defining and classifying risk terms like previous research. A multi-text classification function is necessary because the contract terms that need to be reviewed in detail may vary depending on the scale and type of the project. To enhance the performance of the multi-text classification model, we developed the ELECTRA PLM (Pre-trained Language Model) capable of efficiently learning the context of text data from the pre-training stage, and conducted a four-step experiment to validate the performance of the model. As a result, the ensemble version of the self-developed ITB-ELECTRA model and Legal-BERT achieved the best performance with a weighted average F1-Score of 76% in the classification of 57 contract terms.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.26 no.5
• /
• pp.285-291
• /
• 2014

The statistical information of the coastal water temperature monitoring data can be biased because of outliers and missing intervals. Though a number of outlier detection methods have been developed, their applications are very limited to the in-situ monitoring data because of the assumptions of the a prior information of the outliers and no-missing condition, and the excessive computational time for some methods. In this study, the practical robust method is developed that can be efficiently and effectively detect the outliers in case of the big-data. This model is composed of these two parts, one part is the construction part of the approximate components of the monitoring data using the robust smoothing and data re-sampling method, and the other part is the main iterative outlier detection part using the detailed components of the data estimated by the approximate components. This model is tested using the two-years 5-minute interval water temperature data in Lake Saemangeum. It can be estimated that the outlier proportion of the data is about 1.6-3.7%. It shows that most of the outliers in the data are detected and removed with satisfaction by the model. In order to effectively detect and remove the outliers, the outlier detection using the long-span smoothing should be applied earlier than that using the short-span smoothing.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.12
• /
• pp.889-898
• /
• 2013

Due to the popularization of high-performance capturing equipments and the emergence of powerful image-editing softwares, it is easy to make high-quality counterfeit money. However, the probability of detecting counterfeit money to the general public is extremely low and the detection device is expensive. In this paper, a counterfeit money detection algorithm using a general purpose scanner and computer system is proposed. First, the printing features of color printers are calculated using morphological operations and gray-level co-occurrence matrix. Then, these features are used to train a support vector machine classifier. This trained classifier is applied for identifying either original or counterfeit money. In the experiment, we measured the detection rate between the original and counterfeit money. Also, the printing source was identified. The proposed algorithm was compared with the algorithm using wiener filter to identify color printing source. The accuracy for identifying counterfeit money was 91.92%. The accuracy for identifying the printing source was over 94.5%. The results support that the proposed algorithm performs better than previous researches.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.27 no.4
• /
• pp.785-802
• /
• 2017

As cyber threats using malicious code continue to increase, many security and vaccine companies are putting a lot of effort into analysis and detection of malicious codes. However, obfuscation techniques that make software analysis more difficult are applied to malicious codes, making it difficult to respond quickly to malicious codes. In particular, commercial obfuscation tools can quickly and easily generate new variants of malicious codes so that malicious code analysts can not respond to them. In order for analysts to quickly analyze the actual malicious behavior of the new variants, reverse obfuscation(=de-obfuscation) is needed to disable obfuscation. In this paper, general analysis methodology is proposed to de-obfuscate the software used by a commercial obfuscation tool, Themida. First, We describe operation principle of Themida by analyzing obfuscated executable file using Themida. Next, We extract original code and data information of executable from obfuscated executable using Pintool, DBI(Dynamic Binary Instrumentation) framework, and explain the implementation results of automated analysis tool which can deobfuscate to original executable using the extracted original code and data information. Finally, We evaluate the performance of our automated analysis tool by comparing the original executable with the de-obfuscated executable.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.4
• /
• pp.50-58
• /
• 2017

Recently, some methods of converging heterogeneous fire sensor data have been proposed for effective fire detection, but the rule-based methods have low adaptability and accuracy, and the fuzzy inference methods suffer from detection speed and accuracy by lack of consideration for images. In addition, a few image-based deep learning methods were researched, but it was too difficult to rapidly recognize the fire event in absence of cameras or out of scope of a camera in practical situations. In this paper, we propose a novel fire detection system combining a deep learning algorithm based on CNN and fuzzy inference engine based on heterogeneous fire sensor data including temperature, humidity, gas, and smoke density. we show it is possible for the proposed system to rapidly detect fire by utilizing images and to decide fire in a reliable way by utilizing multi-sensor data. Also, we apply distributed computing architecture to fire detection algorithm in order to avoid concentration of computing power on a server and to enhance scalability as a result. Finally, we prove the performance of the system through two experiments by means of NIST's fire dynamics simulator in both cases of an explosively spreading fire and a gradually growing fire.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.8
• /
• pp.986-994
• /
• 2016

In this paper, we proposed ultrasonic radar module and fixed module for the 2D indoor map building and from each of the modules, we can see the possibilities, limitations and considerations. And finally show the result of building actual 2D indoor map from the modules. Recently there are lots of works for the building indoor map by spotlight on the simultaneous localization and mapping (SLAM). And the LiDAR, ultrasonic, camera sensors are usually used for this work. Especially the LiDAR sensor have a higher resolution and wider detection range more than the ultrasonic sensor, but also there are limitation in the size of module, higher cost, much more throughput of processing data, and weaker to use in various indoor environment noises. So from these reasons, in this paper we could verify that proposed modules and schemes have a enough performance to build the 2D indoor map instead of using LiDAR and camera sensor with minimum number of ultrasonic sensors and less throughput of processing data.