• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.55-62
• /
• 2021

In this paper, we propose a human-content interface that allows users to quickly and efficiently search data through friction-based scrolling with ROI(Regions of interests). Our approach, conceived from the behavior of finding information or content of interest to users, efficiently calculates ROI for a given content. Based on the kernel developed by conceiving from GMM(Gaussian mixture model), information is searched by moving the screen smoothly and quickly to the location of the information of interest to the user. In this paper, linear interpolation is applied to make one softer inertia, and this is applied to scrolls. As a result, unlike the existing approach in which information is searched according to the user's input, our method can more easily and intuitively find information or content that the user is interested in through friction-based scrolling. For this reason, the user can save search time.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.12
• /
• pp.1-8
• /
• 2022

As an efficient file system for SSD(Solid State Drive), F2FS is employed in the kernel of Linux operating system. F2FS applies various methods to improve performance by reflecting the characteristics of flash memory. One of them is improvement of the index structure that contains addresses of data blocks for each file. This paper presents a method for further improving performance by modifying the index structure of F2FS. F2FS manages all index blocks as logical numbers, and an address mapping table is used to find the physical block addresses of index blocks on flash memory. This paper shows performance improvement by applying logical numbers to the last level index blocks only. The count of mapping table search for a data block access is reduced to 1~2 from 1~4.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.174-174
• /
• 2023

The impact of global warming on the south Asian summer monsoon is of critical importance for the large population of this region. This study aims to investigate the future changes of the precipitation extremes during pre-monsoon and monsoon, across this region in a more organized regional structure. The study area is divided into six major divisions based on the Köppen-Geiger's climate structure and 10 sub-divisions considering the geographical locations. The future changes of extreme precipitation indices are analyzed for each zone separately using five indices from ETCCDI (Expert Team on Climate Change Detection and Indices); R10mm, Rx1day, Rx5day, R95pTOT and PRCPTOT. 10 global climate model (GCM) outputs from the latest CMIP6 under four combinations of SSP-RCP scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5) are used. The GCMs are bias corrected using nonparametric quantile transformation based on the smoothing spline method. The future period is divided into near future (2031-2065) and far future (2066-2100) and then the changes are compared based on the historical period (1980-2014). The analysis is carried out separately for pre-monsoon (March, April, May) and monsoon (June, July, August, September). The methodology used to compare the changes is probability distribution functions (PDF). Kernel density estimation is used to plot the PDFs. For this study we did not use a multi-model ensemble output and the changes in each extreme precipitation index are analyzed GCM wise. From the results it can be observed that the performance of the GCMs vary depending on the sub-zone as well as on the precipitation index. Final conclusions are made by removing the poor performing GCMs and by analyzing the overall changes in the PDFs of the remaining GCMs.

• Journal of the Institute of Convergence Signal Processing
• /
• v.23 no.4
• /
• pp.234-240
• /
• 2022

This paper proposes a customized AI exercise recommendation service for balancing the relative amount of exercise according to the working environment by each occupation. WISDM database is collected by using acceleration and gyro sensors, and is a dataset that classifies physical activities into 18 categories. Our system recommends a adaptive exercise using the analyzed activity type after classifying 18 physical activities into 3 physical activities types such as whole body, upper body and lower body. 1 Dimensional convolutional neural network is used for classifying a physical activity in this paper. Proposed model is composed of a convolution blocks in which 1D convolution layers with a various sized kernel are connected in parallel. Convolution blocks can extract a detailed local features of input pattern effectively that can be extracted from deep neural network models, as applying multi 1D convolution layers to input pattern. To evaluate performance of the proposed neural network model, as a result of comparing the previous recurrent neural network, our method showed a remarkable 98.4% accuracy.

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.6
• /
• pp.39-51
• /
• 2024

This paper proposes a novel approach to enhance the security of container-based systems by analyzing system calls to dynamically detect race conditions without modifying the kernel. Container escape attacks allow attackers to break out of a container's isolation and access other systems, utilizing vulnerabilities such as race conditions that can occur in parallel computing environments. To effectively detect and defend against such attacks, this study utilizes eBPF to observe system call patterns during attack attempts and employs a AdaBoost model to detect them. For this purpose, system calls invoked during the attacks such as Dirty COW and Dirty Cred from popular applications such as MongoDB, PostgreSQL, and Redis, were used as training data. The experimental results show that this method achieved a precision of 99.55%, a recall of 99.68%, and an F1-score of 99.62%, with the system overhead of 8%.

• The Korean Journal of Applied Statistics
• /
• v.37 no.1
• /
• pp.13-38
• /
• 2024

As economic uncertainties have increased recently due to COVID-19, there is a growing need to quickly grasp private consumption trends that directly reflect the economic situation of private economic entities. This study proposes a method of estimating private consumption in real-time by comprehensively utilizing big data as well as existing macroeconomic indicators. In particular, it is intended to improve the accuracy of private consumption estimation by comparing and analyzing various machine learning methods that are capable of fitting ultra-high-dimensional big data. As a result of the empirical analysis, it has been demonstrated that when the number of covariates including big data is large, variables can be selected in advance and used for model fit to improve private consumption prediction performance. In addition, as the inclusion of big data greatly improves the predictive performance of private consumption after COVID-19, the benefit of big data that reflects new information in a timely manner has been shown to increase when economic uncertainty is high.

• Food Science and Preservation
• /
• v.31 no.4
• /
• pp.554-564
• /
• 2024

The effects of extracting solvents on the physicochemical properties of vegetable oils extracted from four oil seed plants, namely Dennettia tripetala, Dacryodes edulis, Cola rostrata, and Persea americana, were studied. Vegetable oils were extracted using the Soxhlet method. The oils were used for determining % yield, acid value (AV), iodine value (IV), saponification value (SV), electrical conductivity (EC), and pH. The results revealed that the range of the mean % yield of oils extracted using hexane, carbon tetrachloride (CCl₄), petroleum ether, acetone, and methanol, respectively, were 7.5-12.0, 9.0-22.0, 7.5-27.5 and 12.0-37.5 for the four oil Seeds respectively. Mean AVs of oils in mg KOH/g across the solvents were 3.1-3.7, 3.1-3.8, 2.5-3.9 and 2.4-2.8 for Cola rostrata, Dacryodes edulis, Dennettia tripetala and Persea americana respectively. Mean IVs of oils in gI2/100 g across the solvents were 33.25-33.97, 33.06-33.35, 32.06-33.76 and 33.00-34.00 for the four oil seeds, respectively. Mean SVs in mg KOH/g across the solvents were 191.00-197.44, 190.74-198.31, 194.11-202.52, and 182.23-199.44, respectively. Mean EC values ranged 0.31-0.32, 0.30-0.33, 0.30-0.33, and 0.31-0.32 µS/cm across the solvents, respectively. Mean pH values ranged from 6.1-6.4, 4.6-6.3, 6.2-6.4, and 6.1-6.3 across the solvents for the oils, respectively. The AVs of the oils suggest that they are edible oils, the IVs classify the oils as non-drying oils suitable for paint making, and SVs reveal that the oils are good for soap making. Hexane, petroleum ether, and CCl4 are suitable for oil extraction industrially, while D. edulis, D. tripetala, and P. Americana oils are economically viable oil resources due to their high percentage yield, SV and IV.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.37 no.3
• /
• pp.199-208
• /
• 2019

As the number of available satellites increases and technology advances, image information outputs are becoming increasingly diverse and a large amount of data is accumulating. In this study, we propose a change detection method for high-resolution satellite images that uses transfer learning and a deep learning network to overcome the limit caused by insufficient training data via the use of pre-trained information. The deep learning network used in this study comprises convolutional layers to extract the spatial and spectral information and convolutional long-short term memory layers to analyze the time series information. To use the learned information, the two initial convolutional layers of the change detection network are designed to use learned values from 40,000 patches of the ISPRS (International Society for Photogrammertry and Remote Sensing) dataset as initial values. In addition, 2D (2-Dimensional) and 3D (3-dimensional) kernels were used to find the optimized structure for the high-resolution satellite images. The experimental results for the KOMPSAT-3A (KOrean Multi-Purpose SATllite-3A) satellite images show that this change detection method can effectively extract changed/unchanged pixels but is less sensitive to changes due to shadow and relief displacements. In addition, the change detection accuracy of two sites was improved by using 3D kernels. This is because a 3D kernel can consider not only the spatial information but also the spectral information. This study indicates that we can effectively detect changes in high-resolution satellite images using the constructed image information and deep learning network. In future work, a pre-trained change detection network will be applied to newly obtained images to extend the scope of the application.

• Journal of Cadastre & Land InformatiX
• /
• v.48 no.1
• /
• pp.201-213
• /
• 2018

In recent years, the number of traffic accidents caused by car accidents has been decreasing steadily due to traffic accident prevention activities in Korea. However, the number of accidents in Seoul is higher than that of other regions. Various studies have been conducted to prevent traffic accidents, which are human disasters. In particular, previous studies have performed the spatial analysis of traffic accidents by counting the number of traffic accidents by administrative districts or by estimating the density through kernel density method in order to identify the traffic accident cluster areas. However, since traffic accidents take place along the road, it would be more meaningful to investigate them concentrated on the road network. In this study, traffic accidents were assigned to the nearest road network in two ways and analyzed by hotspot analysis using Getis-Ord Gi* statistics. One of them was investigated with a fixed road link of 10m unit, and the other by computing the average traffic accidents per unit length per road section. As a result by the first method, it was possible to identify the specific road sections where traffic accidents are concentrated. On the other hand, the results by the second method showed that the traffic accident concentrated areas are extensible depending on the characteristic of the road links. The methods proposed here provide different approaches for visualizing the traffic accidents and thus, make it possible to identify those sections clearly that need improvement as for the traffic environment.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.1 no.1
• /
• pp.41-51
• /
• 2002

As Advanced Traveler Information System(ATIS) is the kernel of the Intelligent Transportation System, it is very important how to manage data from traffic information collectors on a road and have at borough grip of the travel time's change quickly and exactly for doing its part. Link travel time can be obtained by two method. One is measured by area detection systems and the other is estimated by point detection systems. Measured travel time by area detection systems has the limitation for real time information because it Is calculated by the probe which has already passed through the link. Estimated travel time by point detection systems is calculated by the data on the same time of each. section, this is, it use the characteristic of the various cars of each section to estimate travel time. For this reason, it has the difference with real travel time. In this study, Artificial Neural Networks is used for estimating link travel time concerned about the relationship with vehicle detector data and link travel time. The method of estimating link travel time are classified according to the kind of input data and the Absolute value of error between the estimated and the real are distributed within 5$\~$15minute over 90 percent with the result of testing the method using the vehicle detector data and AVI data of Namsan Tunnel $\#$1. It also reduces Time lag of the information offered time and draws late delay generation and dissolution.