• 수산해양기술연구
• /
• 제58권3호
• /
• pp.193-198
• /
• 2022

In this study, for the purpose of reducing the catch of small giant octopus in a net pot, an escape experiment of octopus was performed on five types of escape rings of different sizes. As a result of the experiment, the smallest giant octopus with a weight of 406 g was found to escape from an escape ring with a diameter of 30 mm or larger, and 592 g octopus, a weight similar to the octopus of the current minimum landing weight (600 g), escaped from an escape ring with a diameter of larger than 40 mm. An individual weight with 406 g becomes 39 mm when converted from a diameter of 25 mm circular escape vent; that is, the circumference to the inner diameter of the mesh. It can be inferred that the converted mesh size of 39 mm cannot escape. Logistic regression analysis was performed using a generalized linear model (GLM) to investigate the correlation between the ratio of escape ring size/Mantle diameter (R/MD) and the escape rate. As a result, it was found that there was a significant correlation between the R/MD ratio and the escape rate and that the higher the R/MD ratio, the greater the escape rate. As a result of logistic regression analysis, the R/MD value was denoted 0.520 with the 50% escape rate. In addition, it can be estimated to be about 50 mm when converted to the mesh size. Therefore, in this study, the diameter of the escape ring and the size of the escape possible of the octopus were experimentally considered. It was found that there was a significant correlation.

• 대한원격탐사학회지
• /
• 제39권5_3호
• /
• pp.979-995
• /
• 2023

산불은 예측이 어려운 재해이기 때문에 실시간 모니터링을 통해 빠르게 대응하는 것이 중요하며, 정지 궤도 위성 영상은 광역을 짧은 시간 간격으로 모니터링할 수 있어 산불 탐지 분야에 활발히 이용되고 있다. 기존의 위성 영상 기반 산불 탐지 알고리즘은 밝기 온도의 통계량 분석을 통한 임계값 기반으로 이상치를 탐지하는 방향으로 진행되어 왔다. 그러나 강도가 약한 산불을 탐지하기 어렵거나, 적절한 임계값 설정의 어려움으로 일반화 성능이 저하되는 한계점이 있어 최근에는 기계학습을 이용한 산불 탐지 알고리즘들이 제시되고 있다. 현재까지는 random forest, VanillaConvolutional neural network (CNN), U-net 구조 등의 비교적 간단한 기법이 적용되고 있다. 따라서, 본 연구에서는 정지궤도 위성인 Advanced Himawari Imager를 이용하여 동아시아와 호주를 대상으로 State of the Art (SOTA)딥러닝 기법을 적용한 산불 탐지 알고리즘을 개발하고자 하였다. SOTA 모델은 EfficientNet과 lion optimizer를 적용하여 개발하고, Vanilla CNN 구조를 사용한 모델과 산불 탐지 결과를 비교하였다. EfficientNet은 동아시아와 호주에서 0.88 및 0.83의 F1-score를 기록함으로써 CNN (동아시아: 0.83, 호주: 0.78)에 비해 뛰어난 성능을 입증하였다. EfficientNet에 불균형 문제 해결을 위한 weighted loss, equal sampling, image augmentation 기법 적용 시, 동아시아와 호주에서 각각 0.92와 0.84의 F1-score를 기록함으로써 적용 전(동아시아: 0.88, 호주: 0.83)에 비하여 성능이 향상되었음을 확인하였다. 본 연구를 통하여 제시된 SOTA 딥러닝 기법의 산불 탐지에의 적용 가능성과 딥러닝 모델의 성능 향상을 위해 고려해야 할 방향은 향후 산불탐지 분야에 대한 딥러닝 적용에 도움이 될 것으로 기대된다.

• 한국수자원학회논문집
• /
• 제53권6호
• /
• pp.409-416
• /
• 2020

최근 증가하고 있는 이상기후현상으로 인한 사회·경제적 피해를 줄이기 위해 이상기후 감시가 필수적이다. 이 연구의 목적은 Noah 3.3 지면모형으로 추정한 토양수분자료를 활용하여 준실시간 주간 근역층 토양수분지수(Soil Moisture Index, SMI)를 산정하는데 있다. 동아시아영역(15-60°N, 70-150°E)에 대해 Noah 3.3 지면모형의 적용성을 평가하기 위해 양쯔강유역을 선정하였으며, 해당 유역에서 증발산과 현열을 FluxNet, FluxCom, Global Land Evaporation Amsterdam Model (GLEAM), ERA-5, Generalized Complementary Relationship (GCR)자료를 이용하여 비교·평가하였다. 양쯔강 유역에서 Noah 지면모형으로 추정한 증발산은 FluxNet, FluxCom, GLEAM, ERA-5, GCR에 의한 증발산과 0.96이상의 매우 높은 결정계수의 값을 보였으며, 현열의 경우에는 FluxNet 현열 자료와 0.71의 결정계수로 증발산 보다 다소 낮은 값을 보였다. 주간 근역층 SMI 시계열로부터 2019년 7월부터 10월까지 중국의 동부지역에서 극한가뭄(Extreme drought)이 확장되는 현상이 관측되었다. 월별 극한가뭄 발생일수의 트렌드 분석결과, 우리나라의 경우 봄철에는 극한가뭄이 지난 20년 동안 대체로 감소하는 경향이 나타났으나, 가을철에는 한반도 전역에 걸쳐 증가하는 경향이 나타났다. 이 연구가 가뭄의 시·공간적 지속성 및 확장성과 최근 가뭄발생의 경향성 등을 종합적으로 분석하고 판단하여, 가뭄으로 인한 사회·경제적 피해를 줄이기 위한 적절한 대책 마련에 활용성이 클 것으로 기대된다.

• Nuclear Engineering and Technology
• /
• 제41권5호
• /
• pp.747-752
• /
• 2009

A vulnerability assessment is essential for the efficient operation of a physical protection system (PPS). Previous assessment codes have used a simple model called an adversary sequence diagram. In this study, the use of a two-dimensional (2D) map of a facility as a model for a PPS is suggested as an alternative approach. The analysis of a 2D model, however, consumes a lot of time. Accordingly, a generalized heuristic algorithm has been applied to address this issue. The proposed assessment method was implemented to a computer code; Systematic Analysis of physical Protection Effectiveness (SAPE). This code was applied to a variety of facilities and evaluated for feasibility by applying it to various facilities. To help upgrade a PPS, a sensitivity analysis of all protection elements along a chosen path is proposed. SAPE will help to accurately and intuitively assess a PPS.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2003년도 ICCAS
• /
• pp.2033-2038
• /
• 2003

This paper describes an integrated navigation strategy for the autonomous service robot PSR. The PSR is under development at the KIST for service tasks in indoor public environments. The PSR is a multi-functional mobile-manipulator typed agent, which works in daily life. Major advantages of proposed navigation are as follows: 1) Structured control architecture for a systematic integration of various software modules. A Petri net based configuration design enables stable control flow of a robot. 2) A range sensor based generalized scheme of navigation. Any range sensor can be selectively applied using the proposed navigation scheme. 3) No need for modification of environments. (No use of artificial landmarks.) 4) Hybrid approaches combining reactive behavior as well as deliberative planner, and local grid maps as well as global topological maps. A presented experimental result shows that the proposed navigation scheme is useful for mobile service robot in practical applications.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 1989년도 한글날기념 학술대회 발표논문집
• /
• pp.101-106
• /
• 1989

음성 및 문자를 통한 컴퓨터와의 정보 교환을 위한 통합 사용자 인터페이스 (Intelligent Man- Machine interface) 시스템의 일환으로 한국어 단모음의 인식을 위한 시스템을 인공 신경망 모델을 사용하여 구현하였으며 인식시스템의 상위 접속부에 필요한 단어 인식 모듈에 있어서의 인지 실험도 행하였다. 모음인식의 입력으로는 제1, 제2, 제3 포르만트가 사용되었으며 실험대상은 한국어의 [아, 어, 오, 우, 으, 이, 애, 에]의 8 개의 단모음으로 하였다. 사용한 인공 신경망 모델은 Multilayer Perceptron 이며, 학습 규칙은 Generalized Delta Rule 이다. 1 인의 남성 화자에 대하여 약 94%의 인식율을 나타내었다. 그리고 음성 인식시의 인지 현상 실험을 위하여 약 20개의 단어를 인공신경망의 어휘레벨에 저장하여 음성의 왜곡, 인지시의 lexical 영향, categorical percetion등을 실험하였다. 이때의 인공 신경망 모델은 Interactive Activation and Competition Model을 사용하였으며, 음성 입력으로는 가상의 음성 피쳐 데이타를 사용하였다.

• 제어로봇시스템학회논문지
• /
• 제3권1호
• /
• pp.101-108
• /
• 1997

The initial token value required to the optimal performance of discrete event systems can be decided by Sum of Delay Time and Synchronic Time ratio, which are new synchronic variables in Timed Place Petri Nets. For the system consisting of two Live-and-Bounded circuits(LB-circuits) fused in common Transition-Transition-Path or common Place-Place-Path, we prove that the Synchronic Time Ratio is the initial token ratio between two LB-circuits to optimally perform system functions. These results are generalized and formulated as a theorem. The initial tokens of a specific place can imply shared resources. Using the theorem, we can decide the minimum number of the shared resources to obtain the optimal performance, and minimize the idling time of resources. As an example, an automated assembly system is modeled by Timed Place Petri Net, and the initial tokens to achieve the optimal system performance are identified. All the values are verified by simulation.

• Nuclear Engineering and Technology
• /
• 제52권2호
• /
• pp.223-229
• /
• 2020

The present work proposes a solution to the static Boltzmann transport equation approximated by the simplified P₃ (SP₃) on angular, and the analytic coarse mesh finite difference (ACMFD) for spatial variables. Multi-group SP₃-ACMFD equations in 3D rectangular geometry are solved using the GMRES solution technique. As the core time dependent analysis necessitates the solution of an eigenvalue problem for an initial condition, this work is hence devoted to development and verification of the proposed static SP₃-ACMFD solver. A 3D multi-group static diffusion solver is also developed as a byproduct of this work to assess the improvement achieved using the SP₃ technique. Static results are then compared against transport benchmarks to assess the proximity of SP₃-ACMFD solutions to their full transport peers. Results prove that the approach can be considered as an acceptable interim approximation with outputs superior to the diffusion method, close to the transport results, and with the computational costs less than the full transport approach. The work would be further generalized to time dependent solutions in Part II.

• The Journal of Asian Finance, Economics and Business
• /
• 제7권9호
• /
• pp.105-115
• /
• 2020

The study empirically examines the effects of loan portfolio diversification on bank risk and return in the nascent banking market of Vietnam. Loan portfolio diversification is captured through the Hirschman-Herfindahl index and the Shannon Entropy with sectoral exposures. We access each bank's financial reports to collect the required data, especially the breakdown of sectoral loan portfolios, thus constituting a unique dataset. To compute bank return, we use the traditional accounting indicators, including return-on-assets, return-on-equity, and net-interest margin. For bank risk, we utilize the loan-loss provisions and non-performing loans relative to gross customer loans. Using a sample of 30 commercial banks over the period from 2008 to 2019 and the system generalized method of moments estimator for the dynamic panel, we indicate the downsides of portfolio diversification. Concretely, we observe that all diversification measures exhibit significantly negative signs in all regressions across different bank return proxies. At the same time, the estimates display the significant and positive impact of diversification on the non-performing loan ratio. Hence, sectoral loan portfolio diversification significantly hampers bank performance in both aspects of lower return and higher credit risk. The results are robust across a rich set of bank performance and portfolio diversification measures.

• The Journal of Asian Finance, Economics and Business
• /
• 제7권9호
• /
• pp.11-18
• /
• 2020

This paper analyzes factors affecting enterprise's financial risk listed on the Vietnam stock market. The panel data of research sample includes 524 non-financial listed enterprises on the Vietnam stock market for a period of eleven years, from 2009 to 2019. The Generalized Least Square (GLS) is employed to address econometric issues and to improve the accuracy of the regression coefficients. In this research, financial risk is measured by the Alexander Bathory model. Debt structure, Solvency, Profitability, Operational ability, Capital structure are independent variables in the study. Firm Size, firm age, growth rate are control variables. The model results show that in order to prevent and limit financial risk for enterprises listed on the Vietnam Stock Market, attention should be paid to variables reflecting Liability structure ratio, Quick Ratio, Return on Assets, Total asset turnover, Accounts receivable turnover, Net assets ratio and Fixed assets ratio. The empirical results show that there are differences in the impact of these factors on the financial risk in state-owned enterprises and non-state enterprises listed on the Vietnam stock market. The findings of this article are useful for business administrators, helping business managers make the right financial decisions to improve the efficiency of financial risk management in enterprises.