Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this thesis through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2017

Degree Type

Thesis

Degree Name

M.S.

Department

Electrical and Computer Engineering

First Advisor

Nabil Sarhan

Abstract

The automatic control of Pan/Tilt/Zoom (PTZ) cameras has been a major research problem. We consider the control of PTZ cameras in a manner that optimizes the overall recognition accuracy. The camera control solution operates into two alternating phases: pre-recording and recording. In the first phase, the processing architecture performs the necessary algorithmic calculations to determine the optimal PTZ camera setting. However, in the second phase, the PTZ cameras apply the desired settings, capture the videos, and stream these videos to the proxy station for analysis.

We enhance the overall recognition accuracy by developing a parallel PTZ control algorithm, which reduces the time spent on pre-recording and thus increases the fraction of time dedicated to capturing the actual videos of the surveillance site. Additionally, we propose a dynamic approach for determining the pre-recording time and thus allowing the system to extract the best benefits of the parallel algorithm. As the parallel algorithm leads to early completion of the pre-recording tasks, the dynamic approach empowers

the system to benefit from the unused remaining time in the pre-recording phase and subsequently to place more dedication to the actual recording.

We analyze the effectiveness of the proposed solutions through extensive simulation, considering the impacts of major parameters, including the subject arrival rate, surveillance area, and the number of cameras. To make the simulations as realistic as possible, we incorporate an inclusive speed model to constantly update and maintain the speed values for related subjects while they are crossing throughout the surveillance site. This speed model considers many factors, including the social tendencies and density of the people

present in the surveillance site. Our overall solution assumes realistic 3D environments and not just 2D scenes.

We demonstrate that the proposed parallel algorithm substantially reduces the pre-recording time. We also show that the combination of the proposed parallel algorithm and dynamic approach greatly enhances the overall face recognition accuracy.

Off-campus Download

Share

COinS