Description:
Abstract
USC researchers have developed an algorithm capable of tracking objects and humans in unconstrained environments in the presence of occlusions, abrupt motion, motion blur, and frame-cuts. This technology outperforms current state-of-the-art methods by automatically exploiting and tracking all regions with similar appearance to the target. The algorithm allows accurate and reliable tracking of targets, even when the targets move in and out of the field of view.
Benefit
-
More effective than state-of-the-art technologies
-
Real-time tracking
-
Target reacquired even when leaving field of view
-
Excels in real-world video sequences
Market Application
Visual tracking in unconstrained environments is challenging due to variances in lighting conditions, cluttered backgrounds, and changes in observations. Other issues arise with targets leaving the field of view and regions with similar appearances as the targets being tracked instead of reacquiring the right target. Previous attempts of tracking targets in unconstrained environments involve frequent tracking failures and are complex and expensive. A method of tracking an object leaving the field of view or while completely occluded while in an unconstrained environment is critical for many applications, including video surveillance and checkout-free shopping.
Publications
Context tracker: exploring supporters and distracters in unconstrained environments, IEEE 2011
Other
Status: Software available