The pose tracking especially for high-speed motions is important for object manipulation. Frame-based methods have already achieved convincing tracking results, but they are not suitable for objects with high-speed motions. Considering event-based cameras can provide information on brightness changes asynchronously with low latencies, they have ability to capture fast motion information. Combine frames and events information is our strategy for high-speed tracking. It uses abundant object information, such as color, texture, and shape, as well as motion information encoded in events. Therefore, the target is to design a high-speed tracking algorithm through information fusion for frames and events.
In our method, frame-events information is fused for pose tracking. The tracking result is high frequency and smoothed which is important for robot manipulation. Frame-based tracking will fail with high-speed motions because general frame cameras are unable to capture clear high-speed motion information. And it is difficult to track only rely on events information, because the information is very abstract. Our tracking method takes advantage of abundant frame information, and leverages abstract but high frequency events information, therefore, high-speed object motions can be tracked.
Methods & Tools
frame-events based tracking, information fusion
People: Zhichao Li, Franco Di Pietro
L. Natale, N.A. Piga HSP group