CASIA OpenIR  > 毕业生  > 博士学位论文
动态场景语义理解和分析
其他题名Semantic Understanding and Analysis of Dynamic Scenes
信伦
2008-01-27
学位类型工学博士
中文摘要动态场景语义理解和分析的研究目标是让计算机视觉系统具有和人类类似的视觉感知能力,能够对动态场景进行感知、分析和理解,能够得出人类习惯的语义描述。作为计算机视觉研究的最高目标,随着相关研究的进展和技术的完善,动态场景的语义理解越来越被国际学术研究团体所重视。有益的研究成果不断涌现,其中一些成果已经成功并广泛地应用到了人们的日常生产、生活中,发挥了重要的作用。巨大的实际需求和一定程度上的成功应用极大鼓舞了计算机视觉学术和产业界,使得该领域的研究和产业化成为近些年来的热点。 我们提出了一个比较完整的动态场景语义理解和分析的框架结构,实现了从场景中的底层视觉可感知数据到高层语义分析和理解的功能。整个框架涉及到许多图像处理、计算机视觉和人工智能的基本问题,包括视觉可感知实体的提取、场景的语义化建模、目标的检测和跟踪、目标的分类和识别、信息的结构化表达、时空数据分析、行为和事件的分析及语义理解等等。 总的说来,本文围绕一类特定动态场景(监控场景)深入并系统地研究了底层视觉可感知数据提取、信息结构化表达、时序数据分析,以及最终的高层语义分析和理解等问题,构建了一个完整有效的动态场景语义理解和分析的理论体系框架。本文中的一些有益的成果将对数据挖掘、人工智能等相关领域的研究有一定的借鉴意义。
英文摘要The goal of research in semantic analysis of dynamic scenes is to make thecomputer vision system have the visual and cognitive ability to analyze, understand and represent the environment semantically. As an important issue in computer vision, more and more researchers have paid their ebullient attention on this topic, and there emerge a lot of excellent work. Some of these research work have been adopted in many real applications successfully. And this inspirer situation drive research and industry of computer vision to become a very active area. There are some necessary steps in semantic analysis of dynamic scenes. First, all related visual entities should be extracted. Next, all these extracted data should be labeled semantically and represented in a structured form. Then, temporal structured data sequences should be analyzed. After modeling activity and event, semantic representations will be obtained. Though it is become a very hot research topic, there are still a lot of important and di±cult problems in related theories and practices. In this thesis, we provide a general framework of semantic analysis in dynamic scenes to achieve semantic representation and understanding from low-level visual features. This involves many basic problems in image processing, computer vision, data mining and artificial intelligence, which include visual entity extraction and classi¯cation, semantic modeling of dynamic scenes, object detection and tracking, object classi¯cation and recognition, structured representation of information, spatial-temporal data analysis, activity and event analysis and semantic representation and understanding. The main contributions of this thesis include the following: In a word, in this thesis, we have made fruitful attempts and signifincant progresses on semantic analysis in dynamic scenes for our surveillance system.
关键词动态场景 事件 语义 本体 流形 相似性 Dynamic Scene Event Semantic Ontology Manifold Similarity
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/6046
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
信伦. 动态场景语义理解和分析[D]. 中国科学院自动化研究所. 中国科学院研究生院,2008.
条目包含的文件
条目无相关文件。
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[信伦]的文章
百度学术
百度学术中相似的文章
[信伦]的文章
必应学术
必应学术中相似的文章
[信伦]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。