Direct and Indirect vSLAM Fusion for Augmented Reality

Area of Science:

• Computer Vision
• Robotics
• Human-Computer Interaction

Background:

• Augmented reality (AR) adoption is limited by expensive hardware.
• Visual simultaneous localization and mapping (vSLAM) offers a cost-effective solution using cameras.
• Current vSLAM methods (direct and indirect) excel in specific environments but lack broad applicability.

Purpose of the Study:

• To develop a novel vSLAM method fusing direct and indirect approaches.
• To enhance AR system robustness and enable seamless transitions across diverse scenes.
• To improve the accessibility of AR technology through cost-effective solutions.

Main Methods:

• A fusion algorithm combining direct and indirect vSLAM methods was developed.
• The proposed method was evaluated on three datasets.
• Performance was benchmarked against state-of-the-art algorithms (LSD-SLAM, LCSD, ORBSLAM2) in trajectory planning and AR scenarios.

Main Results:

• The fusion algorithm demonstrated comparable efficiency to leading methods in trajectory accuracy.
• The method achieved high robustness and stability in AR augmentation quality.
• Results indicate successful integration of direct and indirect vSLAM strengths.

Conclusions:

• The proposed fusion algorithm offers a robust and versatile vSLAM solution for AR.
• This approach enhances AR performance across various scene types.
• The method contributes to making AR technology more accessible and widely usable.