Non-linearity analysis of depth and angular indexes for optimal stereo SLAM

Area of Science:

• Computer Vision
• Robotics
• Human-Computer Interaction

Background:

• Visually impaired individuals often face navigation challenges in indoor environments.
• Existing navigation aids may lack sufficient real-time position and orientation accuracy.
• Development of robust egomotion estimation is crucial for assistive navigation technologies.

Purpose of the Study:

• To present a real-time 6DoF egomotion estimation system for indoor environments.
• To develop a vision-based system to aid the navigation of the visually impaired.
• To explore a novel feature parametrization combining inverse depth and 3D points.

Main Methods:

• Utilized a wide-angle stereo camera as the sole sensor for egomotion estimation.
• Implemented a hybrid feature parametrization using inverse depth and 3D points based on a dynamic depth threshold.
• Employed non-linearity analysis to compute the depth threshold and patch warping with normal vector information for improved tracking.

Main Results:

• Demonstrated a real-time 6DoF egomotion estimation system effective in indoor settings.
• Showcased the system's capability to provide accurate orientation and depth information.
• Validated the approach for potential application in assistive navigation for the visually impaired.

Conclusions:

• The proposed system offers a viable solution for real-time egomotion estimation in indoor environments.
• The combined feature parametrization effectively balances orientation and depth information.
• This technology holds promise for enhancing the independence and safety of visually impaired individuals through advanced navigation assistance.