• The Bulletin of The Korean Astronomical Society
• /
• v.35 no.1
• /
• pp.83.1-83.1
• /
• 2010

In the last meeting of KAS, we reported the first statistical study of Faraday rotation measure (RM) in the large-scale structure of the universe using the data of cosmological structure formation simulations. With a turbulence dynamo model for the intergalactic magnetic field (IGMF), we predicted that the root mean square of RM through filaments is \sim 1 rad/m^2. Future radio observatories such as the Square Kilometer Array (SKA) could detect this signal level. However, it is known that the typical foreground galactic RM is a few tens and less than ten rad/m^2 in the low and high galactic latitudes, respectively. So the RM in the large-scale structure could be detected only after the foreground galactic RM is removed. In this talk, we show how we remove the foreground galactic RM and what we obtain from the masked data, by using some noise models and masking techniques. Our results can be used to simulate future RM observations by SKA, and eventually to constrain the origin and evolution of the IGMF in the large-scale structure.

• Journal of Korea Multimedia Society
• /
• v.24 no.2
• /
• pp.208-214
• /
• 2021

Since face masks in public were mandated during COVID-19, more people have taken temperature checks, with their masks on. The study has developed a contactless thermal camera that accurately measures temperatures of people wearing different kinds of masks, detect people wearing masks wrong, and record the temperature data. The built-in system that identifies people wearing masks wrong is what masks our contactless thermal camera differentiated from other thermal cameras. Also our contactless thermal camera can keep track of the number of mask wearers in different regions and their temperatures. Thus, the analysis of such regional data can significantly contribute to stemming the spread of the virus.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.25 no.2
• /
• pp.253-260
• /
• 2015

The side-channel attack is widely known as an attack on implementations of cryptographic algorithms using additional side-channel information such as power traces, electromagnetic waves and sounds. As a countermeasure of side channel attack, the masking method is usually used, however full-round masking makes the efficiency of ciphers dramatically decreased. In order to avoid such a loss of efficiency, one can use reduced-round masking. In this paper, we describe a side channel attack on the lightweight block cipher LEA with the first one~six rounds masked. Our attack is based on differentials and power traces which provide knowledge of Hamming weight for the intermediate data computed during the enciphering of plaintexts. According to our experimental result, it is possible to recover 25 bits of the first round key in LEA-128.

• Smart Structures and Systems
• /
• v.31 no.2
• /
• pp.183-197
• /
• 2023

For long-span bridges with a structural health monitoring (SHM) system, environmental temperature-driven responses are proved to be a main component in measurements. However, anomalous structural behavior may be hidden incomplicated recorded data. In order to receive reliable assessment of structural performance, it is important to study therelationship between temperature and monitoring data. This paper presents an application of the cointegration based methodology to detect anomalies that may be masked by temperature effects and then forecast the temperature-induced deflection (TID) of long-span suspension bridges. Firstly, temperature effects on girder deflection are analyzed with fieldmeasured data of a suspension bridge. Subsequently, the cointegration testing procedure is conducted. A threshold-based anomaly detection framework that eliminates the influence of environmental temperature is also proposed. The cointegrated residual series is extracted as the index to monitor anomaly events in bridges. Then, wavelet separation method is used to obtain TIDs from recorded data. Combining cointegration theory with autoregressive moving average (ARMA) model, TIDs for longspan bridges are modeled and forecasted. Finally, in-situ measurements of Xihoumen Bridge are adopted as an example to demonstrate the effectiveness of the cointegration based approach. In conclusion, the proposed method is practical for actual structures which ensures the efficient management and maintenance based on monitoring data.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.104-107
• /
• 2022

autoencoder deep learning model has excellent ability to restore abnormal data to normal data, so it is not appropriate for anomaly detection. In addition, the Inpainting method, which is a method of restoring hidden data after masking (masking) a part of the data, has a problem in that the restoring ability is poor for noisy images. In this paper, we use a method of modifying and improving the MLP-Mixer model to mask the image at a certain ratio and to reconstruct the image by delivering compressed information of the masked image to the model. After constructing a model learned with normal data from the MVTec AD dataset, a reconstruction error was obtained by inputting normal and abnormal images, respectively, and anomaly detection was performed through this. As a result of the performance evaluation, it was found that the proposed method has superior anomaly detection performance compared to the existing method.

• Journal of Information Technology Services
• /
• v.19 no.5
• /
• pp.83-91
• /
• 2020

Language Models such as BERT has been an important factor of deep learning-based natural language processing. Pre-training the transformer-based language models would be computationally expensive since they are consist of deep and broad architecture and layers using an attention mechanism and also require huge amount of data to train. Hence, it became mandatory to do fine-tuning large pre-trained language models which are trained by Google or some companies can afford the resources and cost. There are various techniques for fine tuning the language models and this paper examines three techniques, which are data augmentation, tuning the hyper paramters and partly re-constructing the neural networks. For data augmentation, we use no-answer augmentation and back-translation method. Also, some useful combinations of hyper parameters are observed by conducting a number of experiments. Finally, we have GRU, LSTM networks to boost our model performance with adding those networks to BERT pre-trained model. We do fine-tuning the pre-trained korean-based language model through the methods mentioned above and push the F1 score from baseline up to 89.66. Moreover, some failure attempts give us important lessons and tell us the further direction in a good way.

• 제어로봇시스템학회:학술대회논문집
• /
• 1995.10a
• /
• pp.303-306
• /
• 1995

The present paper deals with extraction of figures and characters from their background using the knowledge of color. At each pixel of the image on the CRT sent from a video camera, RGB values are transformed into the values in another color system, HSI, where "H" denotes hue;"S" denotes saturation;"I" denotes intensity. Representing color in HSI color space is advantageous, since a human feels color mainly in hue with the aid of brightness and purity. Comparing HSI data thus obtained with the masked original image detects noise-free edges included in the orginal image. Then setting a set of HSI thresholds and changing it identifies the portion of image of the same color. This color information is used in recongnizing characters and figures as an auxiliary system of a hierachical figure categorization method for characters and figures recognition.cters and figures recognition.

• Journal of the Korean Society of Visualization
• /
• v.14 no.2
• /
• pp.46-51
• /
• 2016

In this study, a Masked Omni-Directional Integration(MODI) method for pressure calculation is proposed using the Particle Image Velocimetry (PIV) data. To obtain the velocity field, the Affine PIV method was adopted. Synthetic images were generated for a solid body rotation. Calculation on the pressure was based on the Navier-Stokes equation. The results obtained by the MODI were compared with those obtained by theoretical pressure and by the Omni-Directional Integration(ODI) method. It was shown that the minimum error by the proposed MODI method was attained when the mask size was 1.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2003.05a
• /
• pp.202-207
• /
• 2003

The purpose of this study was to investigate the effect of background noise on loudness magnitude estimation of Heavy-weight impact noise. Relationship between loudness magnitude estimation and estimation methods about floor impact noise had appeared low in apartment which receive much effect of background noise. Then, to need new estimation method abut effect of background noise. Masking effects by background noise is increased steadily, there is a continuous transition between an audible impact noise and one that is totally masked. Result 1 hat analyze interrelationship of phychoacoustical data and values through Zwicker Parameters, to Investigate that an estimation experiment about Annoyance need.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.244-247
• /
• 2003

Multi-temporal data is used to determine the rate of deforestation between the years 1976, 1987 and 2000. Three Landsat TM images, for each period, are pre-processed, mosaicked and normalized difference vegetation index (NDVI) values computed. Based on the values, totally non-forested areas are masked out. The forested areas, both partially and wholly, show a very high degree of correlation between all the bands (reflective), thus necessitating application of principal component analysis. The first two principal components and NDVI values (scaled to 0 ? 255) are used in K-means unsupervised classification to distinguish forest from non-forest areas (that appeared as forest at first). Comparison of the resulting thematic maps gives an annual deforestation rate of roughly 15 0000ha. or 2% between any two epochs.