Nicolas Alt, M.Sc.

• Master Student, thesis finshed in January 2009
• Now in a Hiwi position
• Groups: Computer Vision

Master's Thesis

Title:
Optimal Template Selection for Visual Tracking
Abstract:
Visual tracking refers to estimating the pose of an object in a video or set of images with respect to the camera. It is a widely used technique in the field of computer vision. Versatile systems must not rely on specially attached markers on the object, but need to work with any given texture on the object (markerless tracking). Small templates are selected on the object's surface, which are detected throughout the video. This work investigates how to identify optimal templates that may be recovered reliably and in realtime. A set of simple template features is identified, which can be used to predict tracking quality. The findings are applied to LEOPAR, a tracking-by-detection system, increasing detection success rate by up 25%. Additionally, studies are conducted on the performance of tracking algorithms and interest point detectors in a simulative environment.