DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Kim, Yeong-Dae | - |
dc.contributor.advisor | 김영대 | - |
dc.contributor.author | Kim, Kun-Woo | - |
dc.contributor.author | 김건우 | - |
dc.date.accessioned | 2011-12-14T04:09:22Z | - |
dc.date.available | 2011-12-14T04:09:22Z | - |
dc.date.issued | 2009 | - |
dc.identifier.uri | http://library.kaist.ac.kr/search/detail/view.do?bibCtrlNo=308673&flag=dissertation | - |
dc.identifier.uri | http://hdl.handle.net/10203/40824 | - |
dc.description | 학위논문(석사) - 한국과학기술원 : 산업및시스템공학과, 2009.2, [ vi, 53 p. ] | - |
dc.description.abstract | In robot intelligence, object recognition is one of the most important processes and it has been developed mainly by three approaches such as template matching, knowledge reasoning and probabilistic approach. Those approaches are very effective for the environment such as industrial places where object location and pose are tightly controlled. However, object recognition under uncertainty still needs to be improved. Under uncertain environment, robot often fails to recognize objects because of occlusion. Thus, this paper proposes an approach for occluded object recognition based on ontology and Bayesian network. The approach is composed of two phases; knowledge base structuring and object recognition. In knowledge base structuring, whole objects in a domain space, part objects, and spatial relationships between whole objects are learned, and ontology base of them is built. Objects are then recognized from a captured image using ontology matching and Bayesian inference in object recognition phase. Ontology matching is applied for recognition based on the complete information while Bayesian inference is for recognition based on partial information and object information recognized in ontology matching. In this way, the occluded objects or objects with partial information can be recognized efficiently. To demonstrate the benefits of our approach, we show a case study in an experimental environment. | eng |
dc.language | eng | - |
dc.publisher | 한국과학기술원 | - |
dc.subject | Ontology | - |
dc.subject | Bayesian network | - |
dc.subject | robot intelligence | - |
dc.subject | object recognition | - |
dc.subject | occluded object | - |
dc.subject | 온톨로지 | - |
dc.subject | 베이지안 네트워크 | - |
dc.subject | 로봇지능 | - |
dc.subject | 물체인식 | - |
dc.subject | 가려진 물체 | - |
dc.subject | Ontology | - |
dc.subject | Bayesian network | - |
dc.subject | robot intelligence | - |
dc.subject | object recognition | - |
dc.subject | occluded object | - |
dc.subject | 온톨로지 | - |
dc.subject | 베이지안 네트워크 | - |
dc.subject | 로봇지능 | - |
dc.subject | 물체인식 | - |
dc.subject | 가려진 물체 | - |
dc.title | Occluded object recognition using ontology and bayesian network | - |
dc.title.alternative | 온톨로지와 베이지안 네트워크를 이용한 가려진 물체의 인식 | - |
dc.type | Thesis(Master) | - |
dc.identifier.CNRN | 308673/325007 | - |
dc.description.department | 한국과학기술원 : 산업및시스템공학과, | - |
dc.identifier.uid | 020073042 | - |
dc.contributor.localauthor | Kim, Yeong-Dae | - |
dc.contributor.localauthor | 김영대 | - |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.