基于主动式图学习的3维模型检索方法  

3D model retrieval based on active graphic learning

作　　者：任加慧 叶巍[1] 赵向军[1] 

机构地区：[1]江苏师范大学智慧教育学院(计算机科学与技术学院),江苏徐州221116

出　　处：《江苏师范大学学报（自然科学版）》2017年第2期53-57,共5页Journal of Jiangsu Normal University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61272297);江苏省普通高校研究生实践创新计划项目(SJLX15-0710)

摘　　要：针对传统检索方法无法克服3维模型姿态与分辨率变化对精度的影响,基于主动式图学习方法,提出了一种全新的3维模型检索方法.以3维模型侧影轮廓的概率密度分布作为3维模型相似度度量,借助图学习模型完成3维模型的流形构建和检索,并利用主动学习原理增加模型的适用范围,提高检索准确率.实验结果表明,该算法有更高的检索准确率.Aimed at the lack of precision with the attitude and resolution of 3D model in traditional methods,a new 3D model retrieval method is provided based on active graphic learning.It makes use of silhouettes' probability density as similarity measure,and then accomplishes manifold reconstruction and retrieval of 3D models with the help of graphic learning model while extending the scope of application of models by active learning principle so as to increase the accuracy of retrieval.The experimental results present a higher accuracy of our algorithm.

关 键 词：3维模型检索 侧影轮廓 概率密度 流形学习 主动学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]