• Title/Summary/Keyword: LIG process

검색결과 105건 처리시간 0.029초

온-보드에서의 딥러닝을 활용한 드론의 실시간 객체 인식 연구 (A Study on Realtime Drone Object Detection Using On-board Deep Learning)

  • 이장우;김주영;김재경;권철희
    • 한국항공우주학회지
    • /
    • 제49권10호
    • /
    • pp.883-892
    • /
    • 2021
  • 본 논문에서는 드론을 활용한 감시정찰 임무의 효율성을 향상하기 위해 드론 탑재장비에서 실시간으로 구동 가능한 딥러닝 기반의 객체 인식 모델을 개발하는 연구를 수행하였다. 드론 영상 내 객체 인식 성능을 높이는 목적으로 학습 단계에서 학습 데이터 전처리 및 증강, 전이 학습을 수행하였고 각 클래스 별 성능 편차를 줄이기 위해 가중 크로스 엔트로피 방법을 적용하였다. 추론 속도를 개선하기 위해 양자화 기법이 적용된 추론 가속화 엔진을 생성하여 실시간성을 높였다. 마지막으로 모델의 성능을 확인하기 위해 학습에 참여하지 않은 드론 영상 데이터에서 인식 성능 및 실시간성을 분석하였다.

연속형 시스템의 신뢰성 성장 관리 시험 설계 방안 (A Method of Reliability Growth Management Test Design for Continuous System)

  • 서양우;윤정환;이승상;엄천섭
    • 시스템엔지니어링학술지
    • /
    • 제16권2호
    • /
    • pp.87-96
    • /
    • 2020
  • In this paper we proposed the test design method of reliability growth management. First, we presented the process for establishing the reliability growth management test design considering the number of failures and the termination test time. Reliability growth analysis of continuous system was performed in accordance with the test design process presented. In case the reliability test result is not met with the reliability target value after more than three failures occurred, the required test times were analyzed that 1,725 hrs for one corrective action, 1,950 hrs for two corrective actions. If the number of failures is less than three, design a reliability demonstration test according to confidence level 80% and 90% was performed using RGA 11 Software. As a result, it is possible to establish the reliability growth management test design with sufficient use of available resources. The results of this study can be used when establishing a test design for assessment of reliability growth management of all continuous systems.

무기체계의 안전 설계를 위한 DFMEA 적용에 관한 연구 (A Study on the Application of DFMEA for Safety Design of Weapon System)

  • 서양우;오영일;김희욱;김소정
    • 시스템엔지니어링학술지
    • /
    • 제18권1호
    • /
    • pp.46-57
    • /
    • 2022
  • In this paper, we proposed the DFMEA Implementation Method for safety design of Weapon System. First, we presented the process for DFMEA. And then, the case analysis of OOO missile was performed in accordance with the process presented. After defining the system requirements of OOO missile, failure definition scoring criteria was set. In order to clarify the definition of failure, the failure was classified into safety, reliability, maintainability and others. After performing the function analysis, the relationship matrix analysis was performed to identify the failure mode according to the function without omission. After clarifying the failure classification, mode of failure, cause of failure and effect were analyzed to calculate the severity, occurrence and detection values. After the action priority was judged, the recommended action according to the failure classification was identified for the determined action priority. The results of this study can be used as a relevant basis for the design reflection and resource re-allocation of stakeholders.

지상전투차량에서 표적정보 처리 및 공유 방안 구현 (An Implementation of Target Information Management and its Sharing Process among Ground Fighting Vehicles)

  • 최일호;노해환;손원기
    • 한국군사과학기술학회지
    • /
    • 제23권1호
    • /
    • pp.66-75
    • /
    • 2020
  • Enemy information has significant value when it comes to the process of military actions in battle field. Our Army now uses Battlefield Management Systems(BMSs) equipped in Ground Fighting Vehicles(GFVs) and we need to make research on what kind of role enemy information can play in such systems. Also, enemy information can be shared among GFVs and target information shall be extracted from it in view of KVMF scheme. Because KVMF becomes requisite standard in modern BMSs, we need to implement target information handling process in KVMF standard. In this article, we will focus on how target information and its sharing process can be managed efficiently without information conflicts. Also, situation map produced by it will be noted.

물리적 아키텍처 설계에 대한 DSM 방법론 적용 사례 연구 (On the design method of physical architecture based on the Design Structure Matrix (DSM) approach)

  • 최상욱;최상택;정윤호;장재덕
    • 시스템엔지니어링학술지
    • /
    • 제8권1호
    • /
    • pp.21-28
    • /
    • 2012
  • Development of the system that has required performance is the most important figure and that is the key of project succeed. In order to perform that, systems engineering has come to the fore as a solution. In each step of system engineering process, particularly, requirement analysis and derivation, logical solution, architecture design step are known to affect many of the function and efficiency. Of these, this paper focus on architecture design. We introduce methodology for physical architecture design by applying DSM(Design Structure Matrix) methodology which is based on result of logical solution from MBSE methodology.

국방무기체계 개발을 위한 요구사항 개발/관리 프로세스 정립 및 도구 개선 사례 연구 (Case Study on Process Establishment and Tool Improvement of Requirement Development & Management for Defense Weapon System Development)

  • 최상욱;최상택;정윤호;장재덕
    • 한국정보처리학회:학술대회논문집
    • /
    • 한국정보처리학회 2012년도 추계학술발표대회
    • /
    • pp.1482-1485
    • /
    • 2012
  • 연구개발 수명주기의 첫 단계에서 도출되는 요구사항은 수명주기 전 단계 수행의 가이드라인 및 기준이 되는 매우 중요한 요소이다. 본 논문은 표준 및 지침을 분석, 정리하여 국방무기체계 개발을 위한 요구사항의 개발 및 관리 프로세스를 정립하고 이를 제대로 적용할 수 있도록 전산지원도구(DOORS)를 개선한 것을 사례를 들어 보여준다.

Scene-based Nonuniformity Correction by Deep Neural Network with Image Roughness-like and Spatial Noise Cost Functions

  • Hong, Yong-hee;Song, Nam-Hun;Kim, Dae-Hyeon;Jun, Chan-Won;Jhee, Ho-Jin
    • 한국컴퓨터정보학회논문지
    • /
    • 제24권6호
    • /
    • pp.11-19
    • /
    • 2019
  • In this paper, a new Scene-based Nonuniformity Correction (SBNUC) method is proposed by applying Image Roughness-like and Spatial Noise cost functions on deep neural network structure. The classic approaches for nonuniformity correction require generally plenty of sequential image data sets to acquire accurate image correction offset coefficients. The proposed method, however, is able to estimate offset from only a couple of images powered by the characteristic of deep neural network scheme. The real world SWIR image set is applied to verify the performance of proposed method and the result shows that image quality improvement of PSNR 70.3dB (maximum) is achieved. This is about 8.0dB more than the improved IRLMS algorithm which preliminarily requires precise image registration process on consecutive image frames.

유도무기 시험평가 방법과 신뢰성 성장을 고려한 시험 수량 산출 (Calculation of Sample Size for Guided Missile Considering Test Method and Reliability Growth)

  • 이연호;김재황;이계신;이종신;이명진;김두현
    • 한국군사과학기술학회지
    • /
    • 제20권6호
    • /
    • pp.844-852
    • /
    • 2017
  • Since guided weapon is high-cost and one-shot device which is non-reusable, it requires a lot of resources to prove required accuracy as a part of reliability demonstration. Once a test for proving accuracy rate of guided missile fails, it causes an additional cost and delay of schedule. This study introduces an equation for proper sample size and plan for guided-missile accuracy rate test in order to minimize the risk of test failure. Proper sample size for the test is derived by considering the reliability growth. Furthermore, each task for accuracy rate test is defined according to the development step. Therefore, this study can contribute to reduce sample size for accuracy rate test in order to meet the reliability requirement and assure transparency in the test process.

Simulation of a Laser Tomography Adaptive Optics with Rayleigh Laser Guide Stars for the Satellite Imaging System

  • Ahn, Kyohoon;Lee, Sung-Hun;Park, In-Kyu;Yang, Hwan-Seok
    • Current Optics and Photonics
    • /
    • 제5권2호
    • /
    • pp.101-113
    • /
    • 2021
  • Adaptive optics (AO) systems are becoming more complex to improve their optical performance and enlarge their field of view, so it is a hard and time consuming process to set up and optimize the components of AO systems with actual implementation. However, simulations allow AO scientists and engineers to experiment with different optical layouts and components without needing to obtain and prepare them physically. In this paper, we introduce a new AO simulation named the Korea Adaptive Optics Simulation (KAOS), independently developed by LIG Nex1. We verified the performance of KAOS by comparing with other AO simulation tools. In the comparison simulation, we confirmed the results from KAOS and other AO simulation tools were very similar. Also, we proposed a laser tomography AO system with five Rayleigh laser guide stars (LGSs) optimized by using KAOS to overcome the disadvantages of the AO system with a single sodium LGS for the satellite imaging system. We verified the performance of the proposed AO system using KAOS, and the simulation result showed averaged Strehl ratio of 0.37.

실용위성 적용을 위한 GaN 기반 TR모듈 설계 기법 (The Design Method of TR Module Based GaN for Satellite)

  • 양호준;이유리;조성민;유경덕;김종필
    • 한국항공우주학회지
    • /
    • 제50권1호
    • /
    • pp.31-38
    • /
    • 2022
  • 위성용 TR모듈이 지상용 또는 항공용 TR모듈과 다른 점은 높은 레벨의 진동, 충격과 같은 발사환경과 우주방사능, 진공, 온도 변화와 같은 궤도환경 고려가 필요하다는 것이다. 따라서 지상용 또는 항공용 요구사양과는 다른 전기적, 구조적 조건을 요구한다. 이에 따라 서론에서는 발사환경, 궤도환경에 대해서 서술하고, 본문에서는 위성용 TR모듈에서 GaN의 활용과 요구되는 조건을 충족하기 위한 설계 과정을 기술한다. 특히, 부품선정부터 Derating&RF Budget 설계, 공정설계, 열해석까지 각각 단계에서 고려할 설계요소에 대해서 서술한다.