一种三维模型最优视图的分类选择方法  

A classification-based approach for best view selection of 3D models

作　　者：李晨曦[1] 孙正兴[1] 宋沫飞[1] 章菲倩[1] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,江苏南京210093

出　　处：《智能系统学报》2014年第1期12-18,共7页CAAI Transactions on Intelligent Systems

基　　金：国家"863"计划资助项目(2007AA01Z334);国家自然科学基金资助项目(61272219;61100110;61021062);教育部新世纪优秀人才计划资助项目(NCET-04-0460);江苏省科技计划资助项目(BE2010072;BE2011058;BY2012190)

摘　　要：为了三维模型数据集的可视化展示和缩略图的自动化生成,提出了一种三维模型最优视图选择方法,将最优视图选择归结为评价视图分类性能的问题。首先提取三维模型每个视图的特征,然后在预先分好类的三维模型集上,对每个视图训练分类器评价视图分类性能,选择具有最佳分类性能的视图作为最优视图。实验结果表明,选择出的视图符合人的视觉习惯,而且相同类别的模型具有一致的最优视图,证明了所提方法对于三维模型最优视图选择的有效性。For automatic generation of the visualization and thumbnails of the 3D model set, an approach for best view selection of the 3D models is proposed , and this approach formulates the best view selection as a problem of e-valuating views＆#39;classification performance .Initially each of the 3D models＆#39;view features is extracted, then for each view, we train the classifier that evaluates the view＆#39;s classification performance on a pre-classified set of 3D models. The view with the best classification performance is chosen as the best option .Finally, the selected views in the ex-periments are in accordance with the visual characteristics of humans , and the 3D models of the same class have the consistent best views .The results demonstrate the effectiveness of the approach for the best view selection .

关 键 词：三维模型 可视化 最优视图选择 评价视图 分类 分类选择方法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]