http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Defining a new feature set for content-based image analysis using histogram refinement
Park, Jongan,An, Youngan,Kang, Gwangwon,Rasheed, Waqas,Park, Seungjin,Kwon, Goorak Wiley Subscription Services, Inc., A Wiley Company 2008 INTERNATIONAL JOURNAL OF IMAGING SYSTEMS AND TECHN Vol.18 No.2
<P>The proposed method is based on color histogram. A new set of features are proposed for content-based image retrieval (CBIR) in this article. The selection of the features is based on histogram analysis. Standard histograms, because of their efficiency and insensitivity to small changes, are widely used for CBIR. But the main disadvantage of histograms is that many images of different appearances can have similar histograms because histograms provide coarse characterization of an image. We define an algorithm that utilizes the concept of Histogram Refinement (Pass and Zabih, IEEE Workshop on Applications of Computer Vision (1996), 96–102) and we call it color refinement method. Color refinement method splits the pixels in a given bucket into several classes just like histogram refinement method. The classes are all related to colors and are based on color coherence vectors. After the calculation of clusters using color refinement method, inherent features of each of the cluster is calculated. These inherent features include size, mean, variance, major axis length, minor axis length, and angle between x-axis and major axis of ellipse for various clusters. These inherent features are finally used for image retrieval using Euclidean distance. © 2008 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 18, 86–93, 2008</P>
박종안(Jongan Park),홍철의(Chuleui Hong),김원일(Wonil Kim) 大韓電子工學會 2011 電子工學會論文誌-CI (Computer and Information) Vol.48 No.5
사례기반 추론(Case-Based Reasoning, CBR)은 새로운 문제가 주어질 때 과거의 유사한 문제해결 사례를 기반으로 그 해법을 적절히 변용함으로써 새로운 문제에 적합한 해결책을 효율적으로 도출하고자 하는 문제해결 방법으로 인간이 문제를 해결해 나가는 절차와 매우 유사하여 일상생활 속에 널리 사용되고 있다. 본 연구에서는 이러한 사례기반 추론을 국방 전술 시스템에 적용하여, 전투행위 시 과거의 유사한 사례를 기반으로 현재의 상황에 가장 적절한 전술을 사용할 수 있도록 하는 시스템을 설계하고자 한다. 국방 전술 시스템의 경우, 분대원(Non-Player Character, NPC)들이 모여 분대 규모의 작전을 수행할 때, 분대는 최종 목표에 도달하기 위해 정해진 작전에 따라서 행동하게 된다. 이 과정에서 공격, 매복, 전술적 이동 등의 행위를 위한 전술이 구성되어야 한다. 다시 말해 주변 환경, 엄폐물의 위치, 적의 위치에 따라 상황에 맞는 새로운 전술이 필요하며 이러한 전술은 분대장 혹은 소대장 등이 교범에서 배운 과정과 경험에서 축적된 지식을 토대로 생성된다. 본 연구는 사례기반 추론을 사용하여 각 지휘통제 에이전트를 통해 정보가 전달되면 사례기반 데이터베이스에 저장되어 있는 사례와 유사도를 측정하고 가장 적절한 사례를 선택하여 사용하며 새로운 사례는 사례 데이터베이스에 저장하여 다음 번 사례검색 시 사용될 수 있도록 시스템을 설계한다. Case-based reasoning is an efficient method to find solutions for new problems by using past cases after appropriate changes. It is widely used in everyday life because it resembles the way human acts. In this paper, we propose a military system that generates the most appropriate tactics for CGF (Computer Generated Forces) by utilizing past practices. It indeed applies case-based reasoning at the process of armed conflict. When the CGF squad on a mission, they will be given an action plan to reach the final goal. In the process of executing, tactics for specific action should be organized such as attacks, ambushes, and tactical moves. By using the proposed method, tactics were generated by case-based reasoning. The proposed system successfully receives input through each command and control agent, measures the degree of similarity with the case in case DB, selects the most similar case, modifies, uses, and then stores it for next time.
체인코드와 기하학적 복잡도 지수에 기반한 가중치 형태 검색
박종안(Jongan Park),천종훈(Jonghun Chun),강성관(Sungkwan Kang),안영은(Youngeun An),이용은(Yongeun Lee) 한국정보기술학회 2022 한국정보기술학회논문지 Vol.20 No.10
Since the human visual structure primarily recognizes the objects based on the shape feature of objects, the shape retrieval is widely used in CBIR. The image retrieval algorithm proposed in this study first extracts the object contour of the query image, obtains the chaincode index by contour length versus chaincode change frequency, and obtains the geometric complexity by object area versus contour length. In addition, the chaincode index and the geometric complexity are adaptively weighted to have flexibility. In the simulation, the two retrieval weights were normalized and cyclically applied at intervals of ±0.02 to obtain the optimal weight for the object. For the retrieval results in complex image, when the complexity weight was increased, the recall rate and the accuracy were increased, and for the image retrieval results with low complexity, the recall rate and the accuracy were increased when the chaincode index weight was increased.
목적 지향적 학습을 이용한 적응적 전술 생성 시스템 설계
박종안(Jongan Park),홍철의(Chuleui Hong),김원일(Wonil Kim) 大韓電子工學會 2011 電子工學會論文誌-CI (Computer and Information) Vol.48 No.5
에이전트는 특정 목적을 위해 행동을 하는데 이것은 자율지능형 가상군(Computer Generated Forces, CGF)의 공통된 요소이다. 목적을 달성하기 위해 지정된 스크립트를 따라 행위를 하거나 업무 수행의 계획을 세우는 것을 기본적인 에이전트의 지능이라 볼 수 있는데 이보다 더 발전된 지능 에이전트는 계획을 세우는 것뿐만 아니라 계획했던 수행이 어려울 때 계획을 다시 수정하거나 새로운 계획을 적응적으로 만들어내는 것이다. 계획을 수행 할 때 에이전트가 목적을 위한 적응적 행동을 하려면 목표를 달성할 가능성이 적어질 때 스스로 계획을 수정하고 이러한 방식으로 수정되는 방법을 계속적으로 학습하여 차후 같은 경우에는 학습이 반영된 더 좋은 계획 및 전술을 반영하도록 해야 한다. 즉, 목표와 현재의 상태를 실시간으로 분석하고 측정하여 목표 달성도를 정량적으로 계산하고 측정값이 임계값보다 적으면 수정된 계획을 선택하도록 하는 것이다. 본 논문에서는 위와 같이 에이전트가 목표 달성 가능성이 적어질 때 적응적으로 계획을 새롭게 수정하여 적용하는 방법을 연구한 목적 지향적 행위계획 방법을 제안한다. Agent acts for specification purpose, which is common element of CGF (Computer Generated Forces). When basic agent acts as planned, the advanced intelligence agent can do more than this. It can follow predefined actions along appointed script to achieve purpose or lay another plans when it is difficult to achieve. In other words, it can amend plan again or make new plan in order to achieve goals. When plan fails, agent amends oneself, possibly decreases target level to achieve easily. In doing so, the agent calculates a quantitative value for changing plans in realtime, and choose appropriate alternative plans when the threshold value reaches an limit. In this paper, we propose an military system in which the planned action can be modified according to the level of achievement and alternative plans can be generated accordingly.
Jeong, Sanghwa,Moon, Yongsun,Park, Soobong,Lim, Chunhwan,Park, Seungjin,Park, Jongan CHOSUN UNIVERSITY 1997 Basic Science and Engineering Vol.1 No.2
In this paper, we propose a neural network that has improved convergence performance based on redundancy modification in the multi-layer perceptron. The algorithm adjusts continuously weight in the limited learning condition. And the neural networks is learned by the sign of the modified weight. The simulation results show that the proposed learning method is faster than the general back-error propagation method in convergence speed.
Switching 트란지스터의 最大 電力 減小에 관한 硏究
朴鍾安 조선대학교 동력자원연구소 1979 動力資源硏究所誌 Vol.1 No.1
Several inductors for reducing the peak power of a switching transistor are examined, and a transistor switching circuit using a pararrel combination of a zener diode and an inductor is designed. The zener diode is added to provide a return path for the current in the inductor at turn-off time, and prevent damage to the transistor. It was shown that, for an experimental, the peak power was reduced from 3,905〔w〕to 1.75〔w〕.
자기조직화 학습에 의한 영상부호화에서 훈련벡터와 코드북의 상관관계
박성배,신용길,박종안 조선대학교 생산기술연구소 1995 生産技術硏究 Vol.17 No.1
In storing or transmitting image data, image coding plays a important role. Recently, the vector quantization techniques based on a Kohonen's Self-organizing Feature Map have announced. It is possible to design better codebook algorithms. In this paper, we exploit these properties of the codebook design with spatially organized codewords. Also, the learning feature according to training vector and codebook size are estimated. Blocking and blurring effects are analyzed.