| Descript |
108 p. : col. ill. |
| Note |
"August 2010" |
|
"UMI Number: 3436141"--T.p. verso |
|
Thesis advisor: Robert T. Collins |
|
Thesis (Ph.D.)--Pennsylvania State University, 2010 |
|
Includes bibliographical references (p. [99]-108) |
|
Crowd analysis is of increasing interest. Sociologists, for example, are interested instudying social influence within and between small groups of individuals travelingin a crowd. However, current empirical studies conducted by human observersare very time-consuming. We propose an automatic, vision-based system that de-tects and tracks all individuals in video of crowds of varying nature and densitywhile discovering small social groups. Individual components of the system tacklecomputer vision problems that are challenging in their own right, namely, 1) toreliably detect individual people under reasonable crowd density and from differ-ent viewpoints, 2) to robustly track individuals through crowds where inter-personocclusion is frequent, and 3) to infer at a distance which small groups of people areinteracting or traveling together. Results from our automated crowd analysis sys-tem reveal interesting pedestrian group walking patterns that complement currentresearch in crowd dynamics. These discoveries also may provide helpful insightsfor evacuation planning and for real-time situation awareness during emergencyresponse to public disturbances |
|
Photocopy. Ann Arbor, Mich. : University Microfilms International, 2010. 23 cm |
| Subject |
Pennsylvania State University -- Dissertations
|
| Alt Author |
Collins, Robert
|
|
University Microfilms International
|
|
