JanusVF: accurate navigation using SCAAT and virtual fiducials.

Area of Science:

• Computer Science
• Human-Computer Interaction
• Virtual Reality

Background:

• Current virtual reality (VR) tracking systems often have hardware limitations that can obstruct user vision.
• Existing solutions like Hedgehog offer good performance but may introduce visual artifacts.

Purpose of the Study:

• To design a VR tracking solution that matches commercial performance without visual interference.
• To develop an integrated hardware-software system for unobtrusive head pose tracking.

Main Methods:

• The JanusVF system utilizes a backward-facing camera on the user's head.
• Virtual fiducial markers are rendered within the VR scene, visible only to the camera.
• ARToolKitPlus and a single-constraint-at-a-time (SCAAT) filter algorithm track markers to update head pose.

Main Results:

• Experiments in a CAVE-like system demonstrated comparable accuracy, precision, and latency to commercial tracking solutions.
• The JanusVF system successfully tracked head pose without introducing visual obstructions.

Conclusions:

• JanusVF provides a viable alternative for VR tracking, offering high performance with an unobstructed user experience.
• This approach integrates rendering and sensor data for effective, non-intrusive head tracking in VR.