MOTH: Moving Object Tracking via Head-Mounted Displays
Keywords: Moving Object Tracking, Head-mounted Displays, HMD, Augmented Reality, AR
TL;DR: We review moving object tracking algorithms and adapt them for use on a HoloLens 2 HMD
Abstract: Modern head-mounted displays (HMDs) carry a plethora of sensors, comparable to autonomous systems such as robots or self-driving cars. We can thus leverage algorithms and methods developed for those systems to support and help users wearing HMDs. Using the HoloLens 2 as our system of choice, we explore different methods for tracking moving objects. An exhaustive review of current methods and their applicability to HMDs is conducted. Two methods are selected as basis for implementation - a voxel grid based and cluster based approach. We modify the two approaches to be suitable for use on HMDs. After qualitative analysis, they were compared to a ground truth obtained via a marker tracking system. The overall accuracy of tracking moving humans was shown to be between 15 and 25 cm. Though the primary motivation was to facilitate human-robot collaboration in indoor industrial facilities via augmented reality glasses, the research can easily be applied to any Mixed Reality (MR) device possessing appropriate sensors. Likewise, such systems may be of use outdoors as MR slowly exits into the open.
Submission Number: 10
Loading