DocumentCode :
3463921
Title :
HINDSITE: A user-interactive framework for fragment assembly
Author :
Kimia, Benjamin ; Aras, H. Can
Author_Institution :
LEMS, Brown Univ., Providence, RI, USA
fYear :
2010
fDate :
13-18 June 2010
Firstpage :
62
Lastpage :
69
Abstract :
The assembly of fragments into vessels is a significant task in the analysis of archaeological finds. The current method of reconstruction which relies on experts is time-consuming and laborious, and leads only to a fraction of reconstructions possible. Automated tools have been able to assemble at most two or three dozen fragments, while in practice, archaeologists deal with hundreds and thousands of fragments, in the context of missing fragments and fragments mixed in from different vessels. This limitation is mainly due to the exponential increase in the number of possible assembly configurations as the number of fragments increases, and as the path to successful assembly is impeded by missing fragments. This combinatorial explosion can partially be addressed by improving fragment pair matching, but not all archaeological cues can be captured in an automated system. We propose a interactive, expert-guided framework for fragment assembly, HINDSITE, built on a backbone of a web-based database system, whose architecture is independent of the particular fragment pair matching scheme used. The relational database stores fragments, fragment pairs, and partial assemblies as well as algorithms for pairwise fragment matching and for constructing assemblies, incorporates user interaction at each stage, and allows for automated experiments to optimize assembly parameters. We show that HINDSITE can reconstruct a few dozens of vessel fragments even in the presence of extra and missing pieces.
Keywords :
archaeology; relational databases; user interfaces; HINDSITE framework; Web-based database system; archaeological finds; fragment assembly; fragment pair matching; reconstruction method; relational database; user interaction; Assembly systems; Database systems; Explosions; Forensics; Impedance; Relational databases; Shape measurement; Spine; Surgery; Tiles;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Vision and Pattern Recognition Workshops (CVPRW), 2010 IEEE Computer Society Conference on
Conference_Location :
San Francisco, CA
ISSN :
2160-7508
Print_ISBN :
978-1-4244-7029-7
Type :
conf
DOI :
10.1109/CVPRW.2010.5543532
Filename :
5543532
Link To Document :
بازگشت