基于跨媒体可视分析的三维对象关联检索方法研究

项目名称： 基于跨媒体可视分析的三维对象关联检索方法研究

项目编号： No.61503164

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 自动化技术、计算机技术

项目作者： 邹宽胜

作者单位： 江苏师范大学

项目金额： 21万元

中文摘要： 随着三维扫描、三维建模、三维打印技术及计算机图形学与可视化技术的迅猛发展，网络上的三维模型已呈井喷式增长。三维模型检索逐渐成为研究的热点问题。目前，尽管已经提出多种三维模型检索方法，当前的检索方法及系统仍存在以下问题：检索性能与检索时间难以兼顾；测试用的模型与实际三维对象存在较大差异；不能支持更为丰富的查询需求；不能呈现更为丰富的检索分析。对三维对象检索方法做深入研究以解决上述问题并推进到各个领域的应用中仍有很大意义。本课题拟在如下方面开展研究：第一、跨数据源三维对象关联分析及特征融合方法；第二、基于复杂网络的三维对象特征提取及特征匹配方法；第三、基于可是分析的三维对象离线学习、相关反馈及检索分析方法研究；第四、三维对象检索与三维扫描、三维打印结合的跨空间应用研究。本课题拟通过上述内容的研究，提出新的检索方法，研发实用性检索系统，并投入到实际应用中。

中文关键词： 基于内容的多媒体检索；复杂网络理论；跨媒体分析；特质提取；可视分析

英文摘要： With the rapid development of three-dimensional (3D) scanning, 3D modeling, 3D printing, computer graphics and visualization technologies, 3D models had shown a growth spurt on the internet. 3D model retrieval has becomming a research hotspot during the recent years. Although a variety of 3D model retrieval methods and searching engines have been proposed, the current retrieval methods and systems are still have the following problems: It is very difficult to balance the retrieval performance and searching time; there is a big difference with the test 3D model and the 3D actual object; the retrieval interface of the current 3D searching system does not support the multi-modal query; the presentation of the current search results inconsistent with expectations. Thus, there are still a lot of sense to do an in-depth research on 3D model retrieval methods to solve the above problems and to apply the proposed methods in a various fields. In this project, the research is carried out in the following areas: firstly, The feature association analysis and feature fusion for 3D models which are achieved from various different data source; secondly, 3D model feature extraction and feature matching methods based on the complex network theory; thirdly, the offline processing of 3D objects, relevance feedback and analysis of retrieval results based on visual analysis. Finally, a cross space application by combing the proposed 3D model retrieval methods, 3D scanning, and 3D printing technologies. In this project, we intend to address these existed four issues through the above research, and tried to apply the proposed methods in the smart dressing areas at first, and then we will gradually use the proposed technolgies to many other areas.

英文关键词： content-based multimedia retrieval;Complex network theory;cross-media analysis ;feature extraction;visual analysis