Visual Positioning Indoors: Human Eyes vs. Smartphone Cameras

Area of Science:

• Computer Vision
• Indoor Positioning Systems
• Artificial Intelligence

Background:

• Artificial Intelligence (AI) applications are rapidly advancing, with indoor positioning being a critical enabling technology.
• Humans spend approximately 80% of their time indoors, necessitating accurate indoor localization.
• Current indoor positioning methods often lack the simplicity and accessibility of human visual perception.

Purpose of the Study:

• To develop a visual positioning solution utilizing a single smartphone camera image.
• To enable smartphones to perform relative localization against well-defined indoor objects.
• To assess the feasibility of camera-based indoor positioning mimicking human visual capabilities.

Main Methods:

• A novel visual positioning algorithm was designed, leveraging a single image from a smartphone camera.
• The system simulates human visual observation for relative positioning against distinct objects.
• Experiments were conducted across diverse indoor environments (meeting room, library, reading room) using five different smartphone models.

Main Results:

• The developed visual positioning solution achieved an average accuracy of 30.6 cm across multiple smartphones and settings.
• This accuracy significantly surpasses human visual observation, which yielded an average accuracy of 73.1 cm over 300 samples.
• The solution demonstrates robust performance in various indoor environments.

Conclusions:

• Smartphone camera-based visual positioning is a viable and accurate method for indoor navigation.
• This technology offers a more precise alternative to human visual estimation for indoor localization.
• The developed system holds potential for enhancing AI-driven applications reliant on indoor positioning.