基于混合光场描述符的三维目标检索研究

3D object retrieval based on lightfield hybrid descriptor

针对三维目标(3Dobject)检索问题,提出了一种基于新型描述符的3D目标检索方法。首先,在分析现行基于视图的3D模型描述符在描述方法上不充分的基础上,提出了混合描述符HD的总体思路。进而讨论了HD总体框架,即在光场图像阵列自适应的基础上,实现了直方图颜色描述符HCD,shock图形状描述符HSD及贝叶斯网络(Bayesian Network,BN)纹理描述符HTD的优化组合。其次,讨论了HD各部分的具体实现及度量机制,最后,对HD检索性能进行了实验分析,结果表明提出的方法是优于其他基于视图的检索方法。

A novel 3D object retrieval algorithm is proposed for meeting 3D object retrieval.Firstly,based on the shortcomings analysis of existing view-based 3D model descriptors,many 3D model descriptors can not adequately describe 3D object information.So the concept of lightfield hybrid descriptor（denoted as HD） is developed.The HD includes 3 sub-descriptors,i.e.color descriptor HCD based on color histogram,texture descriptor HTD based on Bayesian network learning and shape descriptor HSD based on shock graph.Then overall framework of HD is discussed,adaptive clustering algorithm is used to gain Characteristic View （CV）,and then the method that how to generate HD and the corresponding metric mechanism are described.In the end,some experiments are done to analyze retrieval performance based on HD.The results show that the proposed method is superior to other view-based methods.