Engineers at the University of Pennsylvania have developed an algorithm that works in conjunction with high speed cameras to enable feature tracking on extremely fast moving objects. This innovative algorithm could enable rapid object tracking, inform robotic positioning, and improve image stabilization.

The speed with which standard cameras can capture motion is inherently limited by their collection frame rates. As a result, novel “event-based” cameras that can take images with a very high time resolution (1 micro-second) have recently been developed. These cameras have the ability to capture the movement of very fast-moving objects.

However, current methods to process the videos taken by the “event-based” cameras are unable to accurately track features or objects captured in the videos. Thus, there is a need for an algorithm that can track the movement of fast-moving features and objects recorded in the videos of “event-based” cameras.

To address this gap in technology, members of the Daniilidis Lab developed an image processing algorithm that is capable of tracking features on extremely fast moving objects imaged by the “event-based” cameras. Their algorithm relies upon probabilistic data associations to accurately track features over time. In addition to enabling rapid object tracking with relevant uses in automation and military defense, this technology could be used to inform robotic positioning and improve image stabilization.

• Enables feature tracking on extremely fast-moving objects
• Can be used in conjunction with high speed “event-based” cameras
• More accurate than standard feature-tracking methods used with frame-based cameras

• Fast object tracking (with uses in automation and military defense)
• Image stabilization
• Robotic positioning
• 3D image reconstruction