• 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.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.5
• /
• pp.169-176
• /
• 2020

Recently, deep neural networks (DNNs) have brought revolutions to many mobile and embedded devices by providing human-level machine intelligence for various applications. However, high inference accuracy of such DNNs comes at high computational costs, and, hence, there have been significant efforts to reduce computational overheads of DNNs either by compressing off-the-shelf models or by designing a new small footprint DNN architecture tailored to resource constrained devices. One notable recent paradigm in designing small footprint DNN models is sharing parameters in several layers. However, in previous approaches, the parameter-sharing techniques have been applied to large deep networks, such as ResNet, that are known to have high redundancy. In this paper, we propose a parameter-sharing method for already parameter-efficient small networks such as ShuffleNetV2. In our approach, small templates are combined with small layer-specific parameters to generate weights. Our experiment results on ImageNet and CIFAR100 datasets show that our approach can reduce the size of parameters by 15%-35% of ShuffleNetV2 while achieving smaller drops in accuracies compared to previous parameter-sharing and pruning approaches. We further show that the proposed approach is efficient in terms of latency and energy consumption on modern embedded devices.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.12 no.1
• /
• pp.249-256
• /
• 2012

ePUB is one of the most pervasive eBook formats since it was announced as a 2009 standard in IDPF(International Digital Publishing Forum). ePUB is compressed as a ZIP Archive format and structured as a standard format of OCF. Domestically and internationally, eBook services applying DRM into ePUB have currently been active, while it does not satisfy various needs from business because it does not support preview service which is due to the encryption of the whole ePUB as a single key. This study suggests a way to apply diverse business models by providing eBook content based on ePUB multiple key encryption method, not by encrypting it simply as a single key. The encryption method which applies a multiple key encryption method is suggested as a practically appliable form together with licence issuing and delivery method and decryption method in eBook readers. The multiple key encryption method suggested in this study will make it possible to support partial preview services of ePUB based eBook content. It will be applied to diverse service models and enhance the security level better then single key based encryption method.

• Journal of the Korean Society for Railway
• /
• v.20 no.4
• /
• pp.521-528
• /
• 2017

In this study, vertical loading tests were conducted to investigate the nonlinear behaviors of the fastening systems that have generally been used in the concrete track of domestic railway lines. Nonlinear load-displacement curve models were derived based on the test results. The uplift forces generated in the fastening systems were evaluated by applying the derived nonlinear models as well as the existing linear models. The influence of the factors on the maximum uplift force of the fastening system was analyzed through a parameter study on the distance between neighboring sleepers, the horizontal distance between the center of the bearing and the nearest fastening system from the deck end, and the height of the bridge girder. From the evaluation results it is known that, for economical track and bridge design, due to deck end deformation, it is necessary to consider the nonlinear behavior of the fastening system in the calculation of the uplift force of the fastening systems.

• Journal of the Korea Concrete Institute
• /
• v.24 no.2
• /
• pp.157-164
• /
• 2012

This study evaluates the shear performance of reinforced concrete beams with recycled coarse aggregates. A total of six specimens with various replacement ratios of recycled coarse aggregates (0%, 50%, and 100%) and different amount of shear reinforcement were cast and tested in this study. A finite element analysis was performed to predict the shear behavior of the specimens with natural or recycled coarse aggregates. The FE analysis was performed using a two-dimensional nonlinear FE analysis program based on the disturbed stress field model (DSFM), which is an extension of the modified compression field theory (MCFT). Experimental results showed that the specimens with 50% and 100% replacement ratios of recycled coarse aggregates had the similar shear strength compared to the specimen with natural aggregates, regardless of the replacement ratios of recycled coarse aggregates and the amount of the shear reinforcement. Furthermore, the comparison between experimental and analytical results showed that the proposed numerical modeling methods and the analytical model, DSFM, can be successfully used to predict the shear behavior of reinforced concrete beams with recycled coarse aggregates.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.19 no.1 s.71
• /
• pp.93-103
• /
• 2006

It is absolutely essential that safety assessment of the containment buildings during service life because containment buildings are last barrier to protect radioactive substance due to the accidents. Therefore, this study describes an enhanced degenerated shell finite element(FE) which has been developed for nonlinear FE analysis of reinforced concrete(RC) containment buildings with elasto-plastic material model. For the purpose of the material nonlinear analysis, Drucker-Prager failure criteria is adapted in compression region and material parameters which determine the shape of the failure envelop are derived from biaxial stress tests. Reissner-Mindlin(RM) assumptions are adopted to develop the degenerated shell FE so that transverse shear deformation effects is considered. However, it is found that there are serious defects such as locking phenomena in RM degenerated shell FE since the stiffness matrix has been overestimated in some situations. Therefore, shell formulation is provided in this paper with emphasis on the terms related to the stiffness matrix based on assumed strain method. Finally, the performance of the present shell element to analysis RC containment buildings is tested and demonstrated with several numerical examples. From the numerical tests, the present results show a good agreement with experimental data or other numerical results.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.22 no.4
• /
• pp.355-362
• /
• 2009

Many of residential buildings, which have pilotis in lower stories to meet the architectural needs, are recently constructed in Korea. Usually, infill walls located in the upper stories of these buildings may cause a soft first story, which is very weak from the earthquake resistance. In the design of the buildings, the infill walls of upper stories are usually considered as non-structural elements and thus they are not included in the analytical model. However, the infill walls may affect the seismic behavior of the residential buildings. Therefore, the differences in seismic behaviors of RC buildings with and without masonry infill walls are required to be investigated. In this study, seismic fragility analyses were performed for masonry infilled low-rise RC moment-resisting frames. And seismic behaviors of RC moment-resisting frame with/without masonry infill walls were evaluated. Two types of structural system with the same frame and different allocation of infill walls are used to evaluate the influence of masonry infill walls on seismic behavior of RC moment-resisting frames. The infill walls were modeled as bi-equivalent diagonal struts. The fragility analyses show that the seismic performance of RC moment-resisting frames with soft story is below the desirable building seismic performance level recommended by current seismic codes, indicating high vulnerability of RC moment-resisting frames with soft story.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.31 no.5
• /
• pp.251-258
• /
• 2018

In construction site, conventional methods such as jackhammer or explosive methods(dynamite) have been often used for the demolition of structures. Use of those methods are more carefully treated in environmentally and historically sensitive area. For those reasons, use of Soundless Chemical Demolition Agent(SCDA) is getting the spotlight. The SCDA is a powder which has expansive strength when it is mixed with water. In these Characteristics, SCDA can destroy the concrete or rock as it is poured into boreholes of the concrete or rock structures. However, there is no industrial standard for the use of SCDA effectively yet. In this study, experimental study to measure the expansive pressure was conducted depending on various boundary conditions such as waterproof, length of the steel pipe, submerged of steel pipe. Furthermore, computational analysis using damage plasticity model to predict the minimum required pressure of the SCDA for the concrete demolition depending on spacing between holes(k-factor) and compressive strength of the concrete was conducted. Obtained results indicates that water heat dissipation with submerged steel pipe shows the stable pressure for measuring the SCDA and hole distance(k-factor) is the most important factor for crack initiation of concrete.

• Journal of the Korean Geotechnical Society
• /
• v.19 no.3
• /
• pp.53-64
• /
• 2003

Many transmission towers, high-rise buildings and bridges are constructed near steep slopes and are supported by large-diameter piles. These structures may be subjected to large lateral loads, such as violent winds and earthquakes. Widely used types of foundations for these structures are pier foundations, which have large-diameters with high stiffness. The behavior of a pier foundation subjected to lateral loads is similar to that of a short rigid pile because both elements seem to fail by rotation developing passive resistance on opposite faces above and below the rotation point, unlike the behavior of a long flexible pile. This paper describes the results of several numerical studies performed with a three-dimensional finite element method (FEM) of model tests of a laterally loaded short pile located near slopes, respectively. In this paper, the results of model tests of single piles and pile groups subjected to lateral loading, in homogeneous sand with 30$^{\circ}$ slopes and horizontal ground were analyzed by the 3-D FE analyses. The pile was assumed to be linearly elastic. The sand was assumed to have non-associative characteristics, following the MC-DP model. The failure criterion is governed by the Mohr-Coulomb equation and the plastic potential is given by the Drucker-Prager equation. The main purpose of this paper is the validation of the 3-D elasto-plastic FEM by comparisons with the experimental data.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.16 no.11
• /
• pp.7990-8000
• /
• 2015

Plant factories are plant cultivation systems which produce farm products uniformly under the controlled environmental condition regardless of seasons and places. Thermal flow in the plant factory is an important parameter in cultivating plants. In this research, we study thermal flow characteristics for a hybrid plant factory with multi-layer cultivation shelves using computer simulation techniques. In order to obtain numerical solutions for thermal flow characteristics, a finite volume method was applied. We consider a low-Reynolds-number ${\kappa}-{\epsilon}$ turbulence model, incompressible viscous flows, and pressure boundary conditions for numerical simulation. Commercial software Solid Works Flow Simulation is then used to investigate characteristics of thermal flows in the plant factory applying several different inflow air velocities and arrangements of cultivation shelves. From numerical analysis results, we found that temperatures in cultivation shelves were uniformly distributed for Case 3 when the inflow air velocity was 1.6 m/s by using a blower in the plant factory. However in Case 1 lower temperature distributions were observed in test beds, TB2 and TB3, which indicated that additional temperature control efforts would be required. Average shelf temperature increased by $3^{\circ}C$ using artificial light source (DYLED47) with 50% blue and 50% red LED ratios. Korea Academia-Industrial cooperation Society.