Manifold Learning for 3D Shape Description and Classification

摘要

Periodically, the US Army conducts detailed measurement surveys of its soldiers as a way to understand the impact that changes in soldier body size have for the design, fit and sizing of virtually every piece of clothing and equipment that Soldiers wear and use in combat. Recently finished US Army Anthropometric Survey (ANSUR II) has collected 3D body scan data of soldiers at the Natick Solider Center (NSC), as shown in Figure 1. By applying new techniques for shape analysis and classification to these 3D body scan data will help designers of clothing and personal protection equipment to understand and fit Army population. The overall research goal of this proposal is to create a new manifold learning framework for large-scale graph decomposition and approximation problems by low-rank approximation and guarantee computable, stable and fast optimizations for 3D shape description and classification. The PI's group has published (or accepted for publication) 1 book through Springer and 13 scientific papers partially supported by this grant. In particular, these papers are in top journals and conference proceedings such as TPAMI, IJCV, TCSVT, ICCV, AAAI, SDM, ACM MM, etc. One paper, 1 out of 384, receives the Best Paper Award in SDM 2014. The PI, Dr. Y. Raymond Fu has received the 2014 INNS Young Investigator Award, from International Neural Networks Society (INNS), 2014. Leveraged by this grant, the PI has been granted an ARO Young Investigator Program (YIP) Award and a Defense University Research Instrumentation Program (DURIP) award.