Robust detection and tracking of annotations for outdoor augmented reality browsing.

Area of Science:

• Computer Vision
• Augmented Reality
• Mobile Computing

Background:

• Outdoor augmented reality (AR) systems require stable registration of virtual content to real-world anchor points.
• Existing methods struggle with diverse outdoor viewing conditions and temporal variations.
• Accurate tracking of anchor points is crucial for immersive AR experiences on mobile devices.

Purpose of the Study:

• To present an enhanced approach for registering and tracking real-world anchor points in outdoor AR applications.
• To improve the robustness and accuracy of anchor point re-detection and tracking under various environmental conditions.
• To enable stable AR experiences on current mobile phones, even with user rotation.

Main Methods:

• On-the-fly panoramic image generation for stable orientation tracking during rotational movements.
• Extended dynamic range panoramic images for improved re-detection under varying illumination.
• Integration of internal orientation sensors to prime active search for anchor points.
• Statistical estimation for global consistency and minimization of anchor point position errors.
• Hybrid orientation tracking combining vision, inertial, and magnetic sensors for 3-degree-of-freedom tracking.

Main Results:

• Achieved a re-detection rate of almost 90%, a twofold improvement over previous methods.
• Demonstrated effectiveness across a wide variety of test cases and viewing conditions.
• Maintained interactive frame rates suitable for real-time AR applications.
• Successfully tested using an AR campus guide application on an off-the-shelf smartphone.

Conclusions:

• The proposed approach significantly enhances the reliability of anchor point registration and tracking in outdoor AR.
• The system is practical for current mobile devices and robust to real-world environmental challenges.
• This advancement facilitates more stable and immersive outdoor augmented reality experiences.