Large-Scale Indoor Camera Positioning Using Fiducial Markers

Area of Science:

• Computer Vision
• Robotics
• Photogrammetry

Background:

• Accurate indoor camera pose estimation is crucial for augmented reality, autonomous navigation, and surveillance.
• Existing methods struggle with reliance on distinct features, large overlapping views, and specific environmental conditions.

Purpose of the Study:

• To develop a robust and scalable method for estimating the poses of multiple fixed indoor cameras.
• To overcome limitations of current techniques by utilizing simple fiducial markers.

Main Methods:

• A novel approach using a small set of fiducial markers on paper to establish pairwise camera relationships.
• Markers are strategically placed and moved to build a connected graph of all cameras.
• Final pose optimization minimizes reprojection errors and enforces physical constraints like coplanarity.

Main Results:

• The proposed method demonstrates superior performance compared to state-of-the-art techniques.
• Experiments on artificial and real datasets show increased effectiveness in practical scenarios.
• Validated approach with varying complexity levels.

Conclusions:

• The fiducial marker-based method offers a practical and effective solution for large-scale indoor camera pose estimation.
• The approach enhances accuracy and applicability in real-world systems.
• Open-sourced code and tutorials are provided to facilitate community adoption.