Indoor Surveillance System using Stereo Vision

Participants: 

Silas Waltzer
Eric Tang

Advisor: 
Prof. Kristin Dana



Abstract

The accurate tracking of people is an ongoing problem in computer vision. It has applications in security, as well as business, manufacturing, and entertainment. Rather than monitoring people with a simple video feed, a computer vision system can track activity automatically and provide useful data. This stereo vision system allows a computer to locate multiple people simultaneously in 3-D space and
track their movements.

• Tracking people accurately is an ongoing computer vision problem
• Possible applications in security as well as many other fields
• System can provideuseful location and movement data
• Stereo cameras can detect object locations in 3-D space

Key Points

The two cameras are fully calibrated, and the images are rectified.

A running average background is subtracted from each image, leaving just the foreground.

Points are triangulated and displayed in a top-down view.