A Beginning-to-End System for Efficiently Gathering Tracking Data on Multiple Targets

1 online resource (61 pages) : PDF

2016

University of North Carolina at Charlotte

Multi-object tracking in video has the potential for broad applicability, from analyzing division-of-labor within biological networks such as insects, to surveillance and traffic monitoring. Despite any potential the field of multiple object tracking holds, it often remains passed over in favor of manual annotation methods due to difficulties in reliably obtaining accurate results. Existing tools for gathering tracking data are either very time consuming, inaccessible due to non-intuitive interfaces, or simply produce too many tracking failures with no means of correcting them. In this thesis, we address the multi-object tracking task that is ill served by existing tracking tools. Specifically, we discuss intuitive means of parameter tuning, training various subroutines of the tracking algorithm, and correcting tracking failures. To this end, we developed a new desktop application, ABCTracker, which provides a sequence of efficient steps the direct the user from beginning (the video) to the end (accurate tracking information).

masters theses

Computer science

M.S.

Abctracker
Computer Vision
Multiple Object Tracking
Object Tracking
Software
Tracking

Computer Science

Shin, Minchul

Souvenir, Richard
Zhang, Shaoting

Thesis (M.S.)--University of North Carolina at Charlotte, 2016.

This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.

Copyright is held by the author unless otherwise indicated.

Rice_uncc_0694N_11197

http://hdl.handle.net/20.500.13093/etd:1172