Multi-Target Tracking with Collaborative Roadside Units Under Foggy Conditions

Area of Science:

• Transportation Engineering
• Computer Vision
• Sensor Fusion

Background:

• Intelligent Road Side Units (RSUs) are vital for Intelligent Transportation Systems (ITSs).
• Roadside LiDAR sensors offer high precision but are challenged by fog, which degrades performance.
• Fog-induced scattering and attenuation of LiDAR beams hinder multi-target tracking and compromise ITS safety.

Purpose of the Study:

• To develop an enhanced collaborative RSU method for accurate multi-target tracking in foggy environments.
• To improve the reliability and safety of ITSs by addressing LiDAR performance degradation in fog.
• To integrate denoising and tracking capabilities for robust RSU-based perception.

Main Methods:

• A modified bilateral filter dynamically adjusts kernel scale for effective point cloud denoising.
• A multi-RSU cooperative tracking framework utilizes a particle Probability Hypothesis Density (PHD) filter for measurement fusion.
• Implementation of a multi-target tracking system on an intelligent roadside platform for real-world testing.

Main Results:

• The proposed method significantly improves target detection accuracy by 8% (thin fog) and 29% (thick fog) compared to statistical filtering after fog noise removal.
• The system demonstrates strong performance in tracking multi-class targets.
• Superiority over state-of-the-art methods is shown, particularly in high-order metrics like HOTA, MOTA, and IDs.

Conclusions:

• The developed collaborative RSU method effectively enhances multi-target tracking accuracy and reliability in foggy conditions.
• This approach contributes to improved safety and performance of Intelligent Transportation Systems.
• The algorithm shows excellent real-time performance and robustness for practical ITS applications.