Augmented reality registration algorithm based on T-AKAZE features.

Area of Science:

• Computer Vision
• Augmented Reality

Background:

• Traditional 3D registration methods in AR systems often suffer from low accuracy and robustness.
• Accurate 3D registration is crucial for seamless integration of virtual objects in AR.

Purpose of the Study:

• To propose a novel and robust 3D registration method for augmented reality (AR) systems.
• To improve the accuracy and stability of 3D registration under challenging conditions.

Main Methods:

• Feature point extraction and matching using the AKAZE algorithm combined with Tanimoto similarity measurement.
• Camera pose estimation based on the constraint relationship of feature points.
• Real-time 3D registration and virtual object tracking using the Lucas-Kanade (LK) optical flow algorithm.

Main Results:

• The proposed method demonstrates high registration accuracy and low time consumption.
• The method maintains accurate registration even when markers are rotated or partially blocked.
• Robust performance is observed under varying environmental conditions, including changes in light intensity and parallax.

Conclusions:

• The novel AKAZE and Tanimoto-based 3D registration method offers superior accuracy and robustness for AR applications.
• This approach provides stable and precise tracking of virtual objects, overcoming limitations of traditional methods.