摘要
随着三维应用的普及,三维模型大量产生并广泛传播。由于三维模型广泛应用于计算机辅助设计、三维游戏、电影特效制作等诸多领域,已经形成了大量的三维模型数据库。三维模型语义标注的目的是给出描述其语义的标注词,是三维模型管理和基于文本的三维检索的关键技术。针对互联网大量存在的弱标签三维模型现状,提出一种基于弱标签的三维模型语义标注方法LPMLL,首先,采用半监督学习方法进行标签传播,得到标注词置信度,达到提升训练集的目的。然后,采用一种基于最大后验概率准则的方法进行多标签学习,得到最终标注词。实验数据表明了该方法的有效性。
With the increasing popularity of 3D applications, a lot of 3D geometry models are being created The purpose of annotation for 3D model is that it can automatically list the best suitable labels to describe the 3D models; it is an important part of the text-based 3D model retrieval. A novel method LPMLLfor 3D models multiple semantic annotation was proposed based on weak label First, a graph-based method was proposed to expand the labeled data set. Then, a multi-label lazy learning approach was proposed based on its k nearest neighbors, and maximum a posteriori (MAP) principle was utilized to determine the label set for the unseen 3D model. Experimental evaluation of the method shows that the proposed method is effective in autotagging 3D models.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2012年第9期1873-1876,1881,共5页
Journal of System Simulation
基金
国家高技术研究发展计划(2009AA012103)
国家自然科学基金(60533070)
黑龙江省教育厅基金(12511011
12521055)
关键词
三维模型自动标注
语义标注
半监督学习
弱标签
3D model automatic annotation
semantic annotation
semi-supervised learning
weak label