基于内容检索的三维模型语义标注方法研究  

Research on Semantic Annotation of 3D Model Based on Content Retrieval

作　　者：田枫[1] 孙宁[1] 刘贤梅[1] TIAN Feng;SUN Ning;LIU Xianmei(School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318)

机构地区：[1]东北石油大学计算机与信息技术学院

出　　处：《微型电脑应用》2019年第12期1-4,9,共5页Microcomputer Applications

基　　金：国家自然科学基金项目(61502094);黑龙江省自然科学基金项目(F2016002);黑龙江省教育科学规划重点课题(GJB1215019)

摘　　要：随着三维模型数量的迅速增加,三维模型语义检索技术研究的重要性日益凸现。语义标注是语义检索的基础,为了降低人工标注的人力物力成本,提出一种基于内容检索的三维模型语义标注方法。该方法利用基于内容检索的方式在三维模型底层特征与高层语义之间架起了桥梁。首先提取和比对三维模型特征,然后计算特征间的EMD距离、构建相似近邻空间,最后通过近邻投票的方式来传播三维模型标签并最终实现自动标注。对Princeton Shape Benchmark的实验表明,该方法在三维模型的检索和标注等方面均取得了较好的效果。With rapid increasing of 3D models,research on the semantic-based 3D model retrieval technology holds more important status,and semantic annotation is the basis of semantic retrieval.Because of the cost of manual annotation in large scale 3D model library,a semantic annotation method for 3D model based on content retrieval is introduced in this paper.This method builds a bridge between the low-level features of the 3D model and the high-level semantics by content-based retrieval.Firstly,the 3D model features are extracted and compared,then the EMD distance among the features is calculated,and the nearest similar neighbor space is constructed.Finally,the 3D model label is propagated by the nearest neighbor voting,and the automatic tagging is realized.The experiments on Princeton shape benchmark show that the proposed method is effective in 3D model retrieval and annotation.

关 键 词：三维模型 语义标注 内容检索 相似度计算 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]