Accurate Smartphone Indoor Visual Positioning Based on a High-Precision 3D Photorealistic Map

Area of Science:

• Computer Vision
• Robotics
• Geomatics

Background:

• Indoor positioning is crucial for numerous applications.
• Visual positioning in indoor environments presents significant challenges.
• Existing methods often lack the required accuracy and reliability.

Purpose of the Study:

• To propose an accurate and reliable visual positioning method for smartphones in indoor environments.
• To leverage high-precision 3D maps for enhanced localization.
• To improve upon established baseline methods in terms of accuracy and robustness.

Main Methods:

• Generation of a high-precision 3D photorealistic map using a mobile mapping system.
• Feature matching (SIFT) and multi-view forward intersection for point cloud calculation.
• Hamming embedding for image retrieval, followed by feature matching and pose voting for accurate correspondence.
• P3P and Perspective-n-point (PnP) algorithms with random sample consensus (RANSAC) for intrinsic and extrinsic parameter estimation and smartphone positioning.

Main Results:

• The proposed method demonstrates superior accuracy and reliability compared to baseline methods.
• Experimental results show 70% of images achieve a location error below 0.9 meters in a 10m x 15.8m room.
• The system successfully estimates intrinsic and extrinsic camera parameters for precise localization.

Conclusions:

• The developed visual positioning method offers a significant advancement for smartphone localization in indoor settings.
• The approach provides a robust solution for accurate positioning by integrating 3D mapping and advanced computer vision techniques.
• Further improvements and broader applications of the method are anticipated.