• Geophysics and Geophysical Exploration
• /
• v.6 no.1
• /
• pp.40-52
• /
• 2003

Despite the various opening models of the southwestern part of the East Sea (Japan Sea) between the Korean Peninsula and the Japan Arc, the continental margin of the Korean Peninsula remains unknown in crustal structure. As a result, continental rifting and subsequent seafloor spreading processes to explain the opening of the East Sea have not been adequately addressed. We investigated crustal and sedimentary velocity structures across the Korean margin into the adjacent Ulleung Basin from multichannel seismic reflection and ocean bottom seismometer data. The Ulleung Basin shows crustal velocity structure typical of oceanic although its crustal thickness of about 10 km is greater than normal. The continental margin documents rapid transition from continental to oceanic crust, exhibiting a remarkable decrease in crustal thickness accompanied by shallowing of Moho over a distance of about 50 km. The crustal model of the margin is characterized by a high-velocity (up to 7.4 km/s) lower crustal (HVLC) layer that is thicker than 10 km under the slope base and pinches out seawards. The HVLC layer is interpreted as magmatic underplating emplaced during continental rifting In response to high upper mantle temperature. The acoustic basement of the slope base shows an igneous stratigraphy developed by massive volcanic eruption. These features suggest that the evolution of the Korean margin can be explained by the processes occurring at volcanic rifted margins. Global earthquake tomography supports our interpretation by defining the abnormally hot upper mantle across the Korean margin and in the Ulleung Basin.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.45 no.9
• /
• pp.794-806
• /
• 2017

It is essential to protect the national space assets and space environment safely as a space development country from the continuously increasing space debris. And Active Debris Removal(ADR) is the most active way to solve this problem. In this paper, we studied the Artificial Neural Network(ANN) for a stable recognition model of vision-based space debris tracking system. We obtained the simulated image of the space environment by the KARICAT which is the ground-based space debris clearing satellite testbed developed by the Korea Aerospace Research Institute, and created the vector which encodes structure and color-based features of each object after image segmentation by depth discontinuity. The Feature Vector consists of 3D surface area, principle vector of point cloud, 2D shape and color information. We designed artificial neural network model based on the separated Feature Vector. In order to improve the performance of the artificial neural network, the model is divided according to the categories of the input feature vectors, and the ensemble technique is applied to each model. As a result, we confirmed the performance improvement of recognition model by ensemble technique.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2009.05a
• /
• pp.57-65
• /
• 2009

Selection of good mineralized area is a combination of the integration of all the available geo-scientific (i.e., geological, geochemical, and geophysical) information, extrapolation of likely features from known mineralized terrenes and the ability to be predictive. The time-space relationships of the hydrothermal deposits in the East Asia are closely related to the changing plate motions. Also, two distinctive hydrothermal systems during Mesozoic occurred in Korea: the Jurassic/Early Cretaceous deep-level ones during the Daebo orogeny and the Late Cretaceous/Tertiary shallow geothermal ones during the Bulguksa event. Both the Mesozoic geothermal system and the mineralization document a close spatial and temporal relationship with syn- to post-tectonic magmatism. The Jurassic mineral deposits were formed at the relatively high temperature and deep-crustal level from the mineralizing fluids characterized by the relatively homogeneous and similar ranges of ${\delta}^{18}O$ values, suggesting that ore-forming fluids were principally derived from spatially associated Jurassic granitoid and related pegmatite. Most of the Jurassic auriferous deposits (ca. 165-145 Ma) show fluid characteristics typical of an orogenic-type gold deposits, and were probably generated in a compressional to transpressional regime caused by an orthogonal to oblique convergence of the Izanagi Plate into the East Asian continental margin. On the other hand, Late Cretaceous ferroalloy, base-metal and precious-metal deposits in the Taebaeksan, Okcheon and Gyeongsang basins occurred as vein, replacement, breccia-pipe, porphyry-style and skarn deposits. Diverse mineralization styles represent a spatial and temporal distinction between the proximal environment of sub-volcanic activity and the distal to transitional condition derived from volcanic environments. However, Cu (-Au) or Fe-Mo-W deposits are proximal to a magmatic source, whereas polymetallic or precious-metal deposits are more distal to transitional. Strike-slip faults and caldera-related fractures together with sub-volcanic activity are associated with major faults reactivated by a northward (oblique) to northwestward (orthogonal) convergence, and have played an important role in the formation of the Cretaceous Au-Ag lode deposits (ca. 110-45 Ma) under a continental arc setting. The temporal and spatial distinctions between the two typical Mesozoic deposit styles in Korea reflect a different thermal episodes (i.e., late orogenic and post-orogenic) and ore-forming fluids related to different depths of emplacement of magma (i.e., plutonic and sub-volcanic) due to regional changes in tectonic settings.

• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.959-971
• /
• 2019

CNN (Convolutional Neural Network) is one representative deep learning algorithm, which can extract high-level spatial and spectral features, and has been applied for hyperspectral image classification. However, one significant drawback behind the application of CNNs in hyperspectral images is the high dimensionality of the data, which increases the training time and processing complexity. To address this problem, several CNN based hyperspectral image classification studies have exploited PCA (Principal Component Analysis) for dimensionality reduction. One limitation to this is that the spectral information of the original image can be lost through PCA. Although it is clear that the use of PCA affects the accuracy and the CNN training time, the impact of PCA for CNN based hyperspectral image classification has been understudied. The purpose of this study is to analyze the quantitative effect of PCA in CNN for hyperspectral image classification. The hyperspectral images were first transformed through PCA and applied into the CNN model by varying the size of the reduced dimensionality. In addition, 2D-CNN and 3D-CNN frameworks were applied to analyze the sensitivity of the PCA with respect to the convolution kernel in the model. Experimental results were evaluated based on classification accuracy, learning time, variance ratio, and training process. The size of the reduced dimensionality was the most efficient when the explained variance ratio recorded 99.7%~99.8%. Since the 3D kernel had higher classification accuracy in the original-CNN than the PCA-CNN in comparison to the 2D-CNN, the results revealed that the dimensionality reduction was relatively less effective in 3D kernel.

• Economic and Environmental Geology
• /
• v.31 no.3
• /
• pp.235-247
• /
• 1998

A new technique of simultaneons inversion for 3-D seismic velocity structure by using direct, reflected, and refracted waves is applied to the center of the Korean Peninsula including Pyongnam Basin, Kyonggi Massif, Okchon Fold Zone, Taebaeksan Fold Zone, Ryongnam Massif and Kyongsang Basin. Pg, Sg, PmP, SmS, Pn, and Sn arrival times of 32 events with 404 seismic rays are inverted for locations and crustal structure. 5 ($1^{\circ}$ along the latitude)${\times}6$ ($0.5^{\circ}$ along the longitude) ${\times}8$ block (4 km each layer) model was inverted. 3-D seismic crustal velocity tomography including eight sections from the surface to the Moho, eight profiles along latitude and longitude and the Moho depth distribution was determined. The results are as follows: (1) the average velocity and thickness of sediment are 5.15 km/sec and 3-4 km, and the velocity of basement is 6.12 km/sec. (2) the velocities fluctuate strongly in the upper crust, and the velocity distribution of the lower crust under Conrad appears basically horizontal. (3) the average depth of Moho is 29.8 km and velocity is 7.97 km/sec. (4) from the sedimentary depth and velocity, basement thickness and velocity, form of the upper crust, the Moho depth and form of the remarkable crustal velocity differences among Pyongnam Basin, Kyonggi Massif, Okchon Zone, Ryongnam Massif and Kyongsang Basin can be found. (5) The different crustal features of ocean and continent crust are obvious. (6) Some deep index of the Chugaryong Rift Zone can be located from the cross section profiles. (7) We note that there are big anisotropy bodies near north of Seoul and Hongsung in the upper crust, implying that they may be related to the Chugaryong Rift Zone and deep fault systems.

• Journal of KIISE:Information Networking
• /
• v.37 no.5
• /
• pp.317-326
• /
• 2010

Nowadays, transmission bandwidth for network traffic is increasing and the type is varied such as peer-to-peer (PZP), real-time video, and so on, because distributed computing environment is spread and various network-based applications are developed. However, as PZP traffic occupies much volume among Internet backbone traffics, transmission bandwidth and quality of service(QoS) of other network applications such as web, ftp, and real-time video cannot be guaranteed. In previous research, the port-based technique which checks well-known port number and the Deep Packet Inspection(DPI) technique which checks the payload of packets were suggested for solving the problem of the P2P traffics, however there were difficulties to apply those methods to detection of P2P traffics because P2P applications are not used well-known port number and payload of packets may be encrypted. A proposed algorithm for identifying P2P heavy traffics based on flow transport parameters and behavioral characteristics can solve the problem of the port-based technique and the DPI technique. The focus of this paper is to identify P2P heavy traffic flows rather than all P2P traffics. P2P traffics are consist of two steps i)searching the opposite peer which have some contents ii) downloading the contents from one or more peers. We define P2P flow patterns on these P2P applications' features and then implement the system to classify P2P heavy traffics.

• Korean Journal of Remote Sensing
• /
• v.9 no.2
• /
• pp.51-69
• /
• 1993

Precipitable Water(PW) are retrieved over the tropical and subtropical Pacific Ocean from TOVS infrared and microwave channel brightness temperature and OLR observations by means of stepwise linear regression. The retrieved TOVS PW fields generated by PW$_{sfc}$(71.1 % of the variance and 0.62 g cm$^{-2}$ standard error over the surface) and PW$_{700500}$(71.7 % and 0.17 g cm$^{-2}$ over the 700 - 500 hPa layer) revealed more evolving synoptic signals over the tropical and subtropical Pacific Ocean. The PW$_{sfc}$ dose not show significantly the TP feature because of the representation of the lower PW for high-level clouds not associated with deep convection. There exists some elusion to trace the TP on the PW$_{sfc}$ field if any supplementary information does not provide. But ECMWF analysis has a general tendency of drying the subtropics and moistening the ITCZ (InterTropical Convergence Zone) and SPCZ(South Pacific Convergence Zone). However, although ECMWF analysis is fairly successful in capturing mean patterms, it is unsuccessful in following active synoptic signal like a tropical plume. Similarly, SMMR-PW does not represent the TP well which consists of the highand middle-level clouds, but PW$_{sfc}$ shows underestimated moistness of TP and does not depict significant signal of TP. In the PW field derived from microwave observations, the TP can not be recognized well. Furthermore, the signature of PW$_{sfc}$ was different from OLR for the TP, which implies the presence of high- and middle-layer thin clouds, but in a closer agreement for deep and active convection areas which contain thick middle- and lower-layer clouds; though OLR represented the cloudiness in the tropics well. In synoptically active regions, it differed from OLR analysis, primarily bacause of actual differences in water vapor and cloud features. The signature of PW$_{sfc}$ was different from OLR for the TP.

• Journal of the Korean Society of Radiology
• /
• v.15 no.5
• /
• pp.613-620
• /
• 2021

The purpose of this study is to propose a convolutional neural network model that can classify normal and abnormal(cardiomegaly) in chest X-ray images. The training data and test data used in this paper were used by acquiring chest X-ray images of patients diagnosed with normal and abnormal(cardiomegaly). Using the proposed deep learning model, we classified normal and abnormal(cardiomegaly) images and verified the classification performance. When using the proposed model, the classification accuracy of normal and abnormal(cardiomegaly) was 99.88%. Validation of classification performance using normal images as test data showed 95%, 100%, 90%, and 96% in accuracy, precision, recall, and F1 score. Validation of classification performance using abnormal(cardiomegaly) images as test data showed 95%, 92%, 100%, and 96% in accuracy, precision, recall, and F1 score. Our classification results show that the proposed convolutional neural network model shows very good performance in feature extraction and classification of chest X-ray images. The convolutional neural network model proposed in this paper is expected to show useful results for disease classification of chest X-ray images, and further study of CNN models are needed focusing on the features of medical images.

• Journal of the Korea Convergence Society
• /
• v.10 no.3
• /
• pp.1-6
• /
• 2019

In modern society, lifestyle and individuality are important, and personalized lifestyle and patterns are emerging. The number of people with articulation diseases is increasing due to wrong living habits. In addition, as the number of households increases, there is a case where emergency care is not received at the appropriate time. We need information that can be managed by ourselves through accurate analysis according to the individual's condition for health and disease management, and care appropriate to the emergency situation. It is effectively used for classification and prediction of data using CNN in deep learning. CNN differs in accuracy and processing time according to the data features. Therefore, it is necessary to improve processing speed and accuracy for real-time healthcare. In this paper, we propose motion monitoring using Mask R-CNN for articulation disease management. The proposed method uses Mask R-CNN which is superior in accuracy and processing time than CNN. After the user's motion is learned in the neural network, if the user's motion is different from the learned data, the control method can be fed back to the user, the emergency situation can be informed to the guardian, and appropriate methods can be taken according to the situation.

• Journal of Broadcast Engineering
• /
• v.24 no.3
• /
• pp.472-484
• /
• 2019

This paper proposes a time-domain sound event detection algorithm using DNN (Deep Neural Network). In this system, time domain sound waveform data which is not converted into the frequency domain is used as input to the DNN. The overall structure uses CRNN structure, and GLU, ResNet, and Squeeze-and-excitation blocks are applied. And proposed structure uses structure that considers features extracted from several layers together. In addition, under the assumption that it is practically difficult to obtain training data with strong labels, this study conducted training using a small number of weakly labeled training data and a large number of unlabeled training data. To efficiently use a small number of training data, the training data applied data augmentation methods such as time stretching, pitch change, DRC (dynamic range compression), and block mixing. Unlabeled data was supplemented with insufficient training data by attaching a pseudo-label. In the case of using the neural network and the data augmentation method proposed in this paper, the sound event detection performance is improved by about 6 %(based on the f-score), compared with the case where the neural network of the CRNN structure is used by training in the conventional method.