An Indoor Positioning Method Based on UWB and Visual Fusion.

Area of Science:

• Robotics and Automation
• Computer Vision
• Signal Processing

Background:

• Accurate multi-pedestrian tracking in large indoor spaces is crucial for security and location services.
• Ultra-wide band (UWB) offers high-precision positioning but suffers from Non-Line-of-Sight (NLOS) errors.
• Existing single-sensor systems struggle to meet stringent accuracy requirements due to indoor environmental challenges.

Purpose of the Study:

• To develop a high-precision and stable fusion positioning system integrating UWB and visual data.
• To address the limitations of UWB positioning, particularly NLOS errors, by incorporating visual sensor data.
• To enhance indoor tracking accuracy and reliability for applications like security and emergency evacuation.

Main Methods:

• A fusion system combining Ultra-wide band (UWB) and vision sensors was designed.
• The Hungarian algorithm was employed for identity matching between UWB and visual localization results.
• Federated Kalman filtering was utilized for fusion localization, incorporating a novel Principal Component Analysis (PCA)-based Kalman filter to handle colored noise.

Main Results:

• The UWB and visual fusion method demonstrated a 25.3% improvement in average accuracy compared to UWB alone.
• The proposed system effectively mitigates NLOS error influence through the stability and continuity of visual positioning.
• The PCA-Kalman filtering algorithm reduced the mean square error by 31.8% compared to traditional Kalman filtering, effectively reducing colored noise.

Conclusions:

• The UWB and vision fusion system offers a robust solution for high-precision indoor pedestrian tracking.
• The integration of PCA-Kalman filtering enhances noise reduction and improves the accuracy of state estimation.
• This research provides a significant advancement in indoor positioning technology, overcoming key limitations of existing methods.