Boston University | Center for Computational Science
HomeNews and EventsResearchEducationPeopleSeminarsFacilitiesContact Us

CCS Seminar
Richard Ivey (ACES trainee)
Boston University
Center for Computational Science/Cognitive & Neural Systems
Friday - April 16, 2010
12:00 noon
Physics Research Building - Room 595
3 Cummington Street

"Multi-hypothesis kinematic tracking and feature-aided tracking
algorithms and approaches"


Multiple hypothesis tracking (MHT) algorithms solve the report-to-track association problem by aggregating likelihood values over multiple frames of kinematic evidence. MHT technologies have long been applied to such problems as machine vision and surveillance/security applications. I will review the basic MHT algorithm/framework and present MHT tracker results for various scenarios. MHTs that rely on kinematic information alone in a multi-target scenario are error-prone due to sensor limitations, obscuration of the targets, or target/confuser proximity.

Therefore, feature-aided tracking (FAT) extensions to MHT were developed to leverage feature information (e.g. shape or color) to disambiguate kinematically confusing events. In practice, the typical joint solution to FAT can be computationally expensive or infeasible due to the increased number of nodes in the hypothesis tree. I will describe a new track fusion algorithm, the Long-Term Hypothesis Tree (LTHT) that solves the FAT problem using a partitioned solution rather than a joint solution, reducing computational burden and providing additional system flexibility.

 

 


Home

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

copyright © 2006, Center for Computational Science | Boston University , MA, 02215