An investigation into viewpoint invariant pedestrian recognition for surveillance systems

This is a study of viewpoint invariant appearance models for pedestrian recognition. The study investigates basic models, color histograms and region based histograms, to gain intuition about the appearance models as well as create new baseline results for evaluating future appearance models. This insight is then used to create an appearance model based on large-margin nearest-neighbor classication. This model significantly outperforms the current state of the art. Finally, the IFP Image Processor, a basic framework for implementation of surveillance systems algorithms, is introduced.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2010

Author dc:creator

• Stieber, Brian M.

Contributors dc:contributor

• Huang, Thomas S.

• viewpoint invariant
• pedestrian recognition
• non-overlapping cameras
• large margin nearest neighbor

Statement dc:rights

• Copyright 2010 Brian M. Stieber

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/16499

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/16499