项目名称: 视觉语义的Web统计模型及理解深化
项目编号: No.61472103
项目类型: 面上项目
立项/批准年度: 2015
项目学科: 计算机科学学科
项目作者: 姚鸿勋
作者单位: 哈尔滨工业大学
项目金额: 80万元
中文摘要: 针对目前对Web知识利用的局限性,建立一套基于Web知识的视觉计算学理论,指引探索和解决大数据背景下视觉语义计算中存在的科学问题。本项目将从Web知识的起源出发,对Web视觉媒体数据的统计分布特性展开探索,建立面向大规模统计分析的视觉媒体特征表达、独立空间概率密度估计和语义计算的基础理论框架,揭示Web媒体数据中视觉语义的形成、关联和传播原理,探索Web视觉媒体在语义特征空间的分布规律,将数据自身的分布特点、语义实体的上下文关系以及人类认知的相关启发式假设有机结合起来,逐步构建基于Web数据的视觉语义挖掘理论以及语义的广义计算模型,实现大数据背景下带有语义约束的Web视觉知识抓取,进而突破基于Web的视觉语义的定位和协同分割、图像补充及超分辨率、图像检索重排等关键技术,建立一套Web视觉语义挖掘及智能化视觉信息分析和处理的视觉计算学理论与技术体系,实现Web视觉语义计算的深化理解和应用。
中文关键词: 互联网智能;统计建模;视觉语义计算;知识挖掘;视觉理解
英文摘要: Considering the limited research and applications of Web knowledge, we propose Web-Oriented visual computing theory to direct exploration of the scientific problems in visual semantic computing research. This project starts with the origin of Web knowledge, and explores the distribution characteristics of Web visual media, trying to construct a theoretical framework for fundamental representation, independent space probability estimation and generic visual semantic computing, to reveal the creation, connection and propagation patterns of Web semantics. By exploring the distribution patterns in the feature space of visual media, and integrating such patterns with the context of semantic entities and related hypotheses of human cognition, we could gradually build the theoretical foundation of Web knowledge mining and also construct a more generic computational sematic model, and finally achieve the goal of Web knowledge extraction from Big Data. The research of this project would also make a series of technical breakthroughs such as Web-based semantic localization and co-segmentation, image re-ranking, image completion and super-resolution etc. The final achievement of this project will be a unified system of Web driven visual computing theories and techniques including Web-based visual knowledge mining, intelligent visual information analysis and processing etc., leading to deep understanding and wide applications for visual semantic computing of the Web.
英文关键词: Web Intelligence;statistical modeling;visual semantic computing;knowledge mining;visual understanding