A general framework for 3D model co-alignment. (September 2017)

• Record Type: Journal Article
• Authors: Xie, Xuanmeng; Luo, Shan; Feng, Jieqing
• Is Part Of: Computer aided design. Volume 90(2017)
• Page Start: 59
• Page End: 70
• Publication Date: 2017-09
• Subjects: Model alignment -- Co-alignment -- Feature evaluation -- Feature selection -- Shape analysis; Computer-aided design -- Periodicals; Engineering design -- Data processing -- Periodicals; Computer graphics -- Periodicals; Conception technique -- Informatique -- Périodiques; Infographie -- Périodiques; Computer graphics; Engineering design -- Data processing; Periodicals; Electronic journals; 620.00420285
• DOI: 10.1016/j.cad.2017.05.001
• Languages: English
• ISSNs: 0010-4485

Abstract

Abstract: 3D model alignment is a fundamental step in many shape analysis processes, and various algorithms have been proposed to solve this problem. However, to the best of our knowledge, they are effective only on specific categories of models. Therefore, we present a novel framework that can align general categories by combining different features together. In order to align given groups of models, multiple features are evaluated first in this framework, according to three types of quantified characteristics, i.e., the intensity, the uniqueness and the consistency. Then, the quantified characteristics are combined into scores and a data-driven model is learned to predict the alignment errors according to the scores. Finally, the features with the minimum predicted alignment errors are selected to align the given groups. Experimental results show that our framework can generate consistent alignments on general categories, which are much better than those generated using single features. Graphical abstract: Highlights: Multiple features can be combined for 3D model co-alignment in the framework. Intensity, uniqueness and consistency are proposed to evaluate the features. Scores, angular errors and energies are defined for feature selection. Curves are fitted to predict angular errors.