An Improved Unscented Particle Filter Approach for Multi-Sensor Fusion Target Tracking

Area of Science:

• Sensor Fusion
• Multi-Target Tracking
• Signal Processing

Background:

• Accurate tracking of maneuvering targets is crucial in various applications.
• Existing multi-sensor fusion algorithms face challenges in real-time performance and accuracy.
• Integrating diverse sensors like radar and infrared requires advanced fusion models.

Purpose of the Study:

• To develop an improved unscented particle filter (IUPF) for enhanced multi-sensor fusion.
• To propose a novel multi-sensor distributed fusion model integrating radar and infrared data.
• To create a multi-target tracking algorithm combining joint probabilistic data association (JPDA) with IUPF.

Main Methods:

• Utilized minimum skew simplex and scaled unscented transforms to reduce UPF computational load.
• Implemented a self-adaptive gain modification coefficient to address sigma point reduction inaccuracies.
• Modified particle weight calculation to mitigate particle degradation.
• Developed a new distributed fusion architecture for radar and infrared sensors.

Main Results:

• The improved unscented particle filter (IUPF) significantly enhances real-time performance.
• Tracking accuracy is maintained and improved compared to existing algorithms.
• The novel fusion architecture effectively leverages radar and infrared sensor data.
• The combined JPDA and IUPF algorithm shows superior multi-target tracking capabilities.

Conclusions:

• The proposed IUPF algorithm offers improved real-time processing for maneuvering target tracking.
• The novel multi-sensor fusion model enhances data integration and tracking precision.
• This research provides a more effective solution for multi-sensor fusion and multi-target tracking applications.