A Motion Blur Resilient Fiducial for Quadcopter Imaging

Fiducials are commonly placed in environments to provide a uniquely identifiable object in the scene. In quad copter applications, these fiducials are often used to evaluate planning algorithms given that ground truth positions can be detected from the quad copter's camera. Low cost quad copters, however, are subject to quick and unstable motions that can cause significant motion blur that severely affects the detection rate of existing fiducials. This problem motivated us to design a fiducial that is robust to motion blur. Our proposed design uses concentric circles with the observation that the direction perpendicular to the motion blur direction will be relatively unaffected by the blur. As a result, an appropriate fiducial code orthogonal to the blur direction can be recognized. Since the direction of motion blur is unknown, the circular design is good for all motion blur directions. We describe the design of binary fiducials, and also a detection algorithm. We show that our marker can significantly outperform existing fiducials in scenes captured with a quad copter.