• Korean Journal of Remote Sensing
• /
• v.39 no.3
• /
• pp.283-295
• /
• 2023

With the development of sensor and satellite technology, numerous high-resolution and multi-spectral satellite images have been available. Due to their wavelength-dependent reflection, transmission, and scattering characteristics, multi-spectral satellite images can provide complementary information for earth observation. In particular, the short-wave infrared (SWIR) band can penetrate certain types of atmospheric aerosols from the benefit of the reduced Rayleigh scattering effect, which allows for a clearer view and more detailed information to be captured from hazed surfaces compared to the visible band. In this study, we proposed a multi-resolution transform-based image fusion method to combine visible and SWIR satellite images. The purpose of the fusion method is to generate a single integrated image that incorporates complementary information such as detailed background information from the visible band and land cover information in the haze region from the SWIR band. For this purpose, this study applied the Laplacian pyramid-based multi-resolution transform method, which is a representative image decomposition approach for image fusion. Additionally, we modified the multiresolution fusion method by combining a haze-guided weight map based on the prior knowledge that SWIR bands contain more information in pixels from the haze region. The proposed method was validated using very high-resolution satellite images from Worldview-3, containing multi-spectral visible and SWIR bands. The experimental data including hazed areas with limited visibility caused by smoke from wildfires was utilized to validate the penetration properties of the proposed fusion method. Both quantitative and visual evaluations were conducted using image quality assessment indices. The results showed that the bright features from the SWIR bands in the hazed areas were successfully fused into the integrated feature maps without any loss of detailed information from the visible bands.

• Computational Structural Engineering
• /
• v.6 no.4
• /
• pp.67-72
• /
• 1993

This paper deals with the response spectra analysis method of the secondary structures including the correlation effect between the nonuniform multiple support excitations. Based on the random vibration theory, the multiple floor response spectra and the cross-correlation coefficient spectra of the floor motions are derived from the design ground response spectra. The example analysis results show that the proposed method yields more accurate results than those by the conventional multiple floor response spectra method without the correlation effects of the support motions. The present method may be easily employed in the seismic design of the secondary structures in engineering practice.

• Proceedings of the KIEE Conference
• /
• 1999.11c
• /
• pp.603-605
• /
• 1999

본 논문은 통계적 Regression방법인 Alternating Conditional Expectation(ACE) 방법을 적용하여 다중계측기를 이용한 공간 변수의 분포에 대한 추정 혹은 재구성 문제를 분석하는 방법을 제시하고 있다. 한다. 계측기 설치 비용 및 설치 위치의 한계로 인해 완벽하게 이루어지기 힘든 공간 변수의 연속적인 분포 추정은 공정 시스템이나 안전성 관련 변수의 감시분야에 많이 응용되고 있다. 본 논문은 계측기 추가에 따르는 비용을 줄이거나 동일한 수의 계측기로 측정오차를 감소시킬 목적으로 가상계측기의 개념을 도입하고 이를 적용하기 위한 통계적 추정 방법론을 기술하고 있다. 수치모사를 통한 분석결과 본 방법은 비선형성이 큰 변수분포에 대해서도 Robust한 결과를 주는 것을 확인하였다.

• Journal of KIISE:Software and Applications
• /
• v.30 no.7_8
• /
• pp.719-726
• /
• 2003

This paper proposes a multimodal registration method for segmentation of the Visible Human color images, in which color characteristics of bones are very similar to those of its surrounding fat areas. Bones are initially segmented in CT images, and then registered into color images to lineate their boundaries in the color images. For the segmentation of bones in CT images, a thresholding method is developed. The registration method registers boundaries of bodies in CT and color images using a cross-correlation approach, in which the boundaries of bodies are extracted by thresholding segmentation methods. The proposed method has been applied to segmentation of bones in a head and legs whose boundary is ambiguous due to surrounding fat areas with similar color characteristics, and produced promising results.

• ICROS
• /
• v.19 no.2
• /
• pp.34-38
• /
• 2013

자연계에서 빈번히 목격되는 군집 현상과 그 효용성의 고찰에 기인하여, 최근 다중 이동 로봇의 협업에 대한 연구가 활발히 수행되고 있다. 그 중, 본 논문에서는 다중 이동 로봇의 군집 제어 방법론들을 설명하고, 그와 관련된 최신 결과들도 소개한다. 특히 군집 제어 문제를 해결하기 위한 대표적인 방식인 행동 기반 접근법, 가상 구조 접근법, 선도-추종 접근법, 그래프 이론 기반 접근법 위주로 소개한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2006.11a
• /
• pp.543-546
• /
• 2006

MDA 는 OMG 에서 제안한 소프트웨어 개발 방법론으로써, 플랫폼 독립적인 모델에서 플랫폼 종속적인 모델로의 모델 변환을 사용하여 소프트웨어의 이식성, 재사용성 등을 향상 시킬 수 있다. 다중 에이전트 기반 시스템을 개발하는 데 MDA 를 적용하는 것은, 다양한 유비쿼터스 환경에서 동작하는 지능형 시스템 개발을 효율적으로 지원한다. 최근 MDA 를 적용하여 다중 에이전트 기반 시스템을 개발한 연구가 있다. 그러나 이 연구는 에이전트 플랫폼이 고려되지 않고 모델 변환이 자동화 되지 못한 단점이 있다. 본 논문에서는 OMG 에서 표준화한 방법으로 UML 을 모델을 기술하고 모델 변환 규칙을 적용한다. 이를 통해 하나의 모델로부터 다양한 에이전트 플랫폼으로의 이식성을 얻을 수 있다. 이를 평가하기 위해 이전 연구에서 개발한 전자상거래 시스템의 시나리오를 가지고 에이전트 시스템을 구현하고 그 유효성을 입증하였다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2007.06c
• /
• pp.422-426
• /
• 2007

음성 질의에서 효율적으로 onset을 검출하기 위한 연구는 다양하게 이루어져 왔다. 특히 대부분의 연구는 확률론적 모델에서 큰 성과를 나타내고 있다. 그러나 이러한 모델들은 변화나 확장이 쉽지 않다는 단점을 가지고 있다. 본 논문에서는 동적 확장 가능한 다중 계층 신경망(Dynamic Expansible MLP)을 제안하여, 기존 방법론의 확장 가능성을 모색한다. 또한, 음성 질의의 onset을 검출하기 위해 MLP를 활용하기 위한 모델을 제시한다.

• Journal of Intelligence and Information Systems
• /
• v.12 no.2
• /
• pp.17-32
• /
• 2006

Recently, community computing has been proposed for group formation and group decision-making. However, legacy community computing systems do not support group need identification for ad hoc group formation, which would be one of key features of ubiquitous decision support systems and services. Hence, this paper aims to provide a multi-agent based methodology to enable nomadic community computing which supports ad hoc need identification and group formation. Focusing on supporting group decision-making of relatively small sized multiple individual in a community, the methodology copes with the following three characteristics: (1) ad hoc group formation, (2) context-aware group need identification and (3) using mobile devices working in- and out-doors. NAMA-US, an RFID-based prototype system has been developed to show the feasibility of the idea proposed in this paper.

• Journal of Nuclear Fuel Cycle and Waste Technology(JNFCWT)
• /
• v.5 no.1
• /
• pp.29-37
• /
• 2007

A conventional single compartment model cannot simulate reasonably the migration phenomenon of contaminants through unsaturated zone, due to the intrinsic unrealistic assumption of the compartment model that contaminants entering a compartment are immediately and uniformly mixed. Although, a multi-compartments model, in which even physically identical layer is divided into multiple compartments, may be used for explaining the retardation of contaminant mass flux along with increasing number of compartments, its numerical modeling is usually time-consuming and appropriate analytical solutions have not been reported yet. In order to improve the conventional compartment models on contaminant migration through unsaturated zone, a series of analytical solutions for multi-compartments model were derived and a generalized constraint under which the results from multi-compartments model can be simply approximated by single compartment model was proposed. The simplified approximation method was verified by a simple numerical analysis on the constraint under hypothetical conditions. It was also proved that the influent contaminant transfer rate from the bulk unsaturated zone can be generally represented into a time-dependent nominal transfer rate rather than a constant. In addition, the nominal transfer rate turned out to be very sensitive to the contaminant transfer rate between compartments in unsaturated zone, but to be almost insensitive to the transfer rate from contaminated zone. It is expected that the simplified approximation method developed in this study can be used for rapid and reasonable estimation of the migration phenomenon of contaminant through unsaturated zone, instead of time-consuming multi-compartments modeling.

• Annual Conference on Human and Language Technology
• /
• 2017.10a
• /
• pp.56-59
• /
• 2017

CNN(Convolutional Neural Network)을 이용하여 발화 주제 다중 분류 task를 multi-labeling 방법과, cluster 방법을 이용하여 수행하고, 각 방법론에 MSE(Mean Square Error), softmax cross-entropy, sigmoid cross-entropy를 적용하여 성능을 평가하였다. Network는 음절 단위로 tokenize하고, 품사정보를 각 token의 추가한 sequence와, Naver DB를 통하여 얻은 named entity 정보를 입력으로 사용한다. 실험결과 cluster 방법으로 문제를 변형하고, sigmoid를 output layer의 activation function으로 사용하고 cross entropy cost function을 이용하여 network를 학습시켰을 때 F1 0.9873으로 가장 좋은 성능을 보였다.