Back to AI Research

AI Research

No Pedestrian Left Behind: Real-Time Detection and... | AI Research

Key Takeaways

  • No Pedestrian Left Behind: Real-Time Detection and Tracking of Vulnerable Road Users for Adaptive Traffic Signal Control Current traffic signals often rely o...
  • We introduce No Pedestrian Left Behind (NPLB), a real-time adaptive traffic signal system that monitors VRUs in crosswalks and automatically extends signal timing when needed.
  • We evaluated five state-of-the-art object detection models on the BGVP dataset, with YOLOv12 achieving the highest mean Average Precision at 50% ([email protected]) of 0.756.
  • NPLB integrates our fine-tuned YOLOv12 with ByteTrack multi-object tracking and an adaptive controller that extends pedestrian phases when remaining time falls below a critical threshold.
  • Through 10,000 Monte Carlo simulations, we demonstrate that NPLB improves VRU safety by 71.4%, reducing stranding rates from 9.10% to 2.60%, while requiring signal extensions in only 12.1% of crossing cycles.
Paper AbstractExpand

Current pedestrian crossing signals operate on fixed timing without adjustment to pedestrian behavior, which can leave vulnerable road users (VRUs) such as the elderly, disabled, or distracted pedestrians stranded when the light changes. We introduce No Pedestrian Left Behind (NPLB), a real-time adaptive traffic signal system that monitors VRUs in crosswalks and automatically extends signal timing when needed. We evaluated five state-of-the-art object detection models on the BGVP dataset, with YOLOv12 achieving the highest mean Average Precision at 50% ([email protected]) of 0.756. NPLB integrates our fine-tuned YOLOv12 with ByteTrack multi-object tracking and an adaptive controller that extends pedestrian phases when remaining time falls below a critical threshold. Through 10,000 Monte Carlo simulations, we demonstrate that NPLB improves VRU safety by 71.4%, reducing stranding rates from 9.10% to 2.60%, while requiring signal extensions in only 12.1% of crossing cycles.

No Pedestrian Left Behind: Real-Time Detection and Tracking of Vulnerable Road Users for Adaptive Traffic Signal Control

Current traffic signals often rely on fixed timing, which assumes that all pedestrians walk at the same speed. This creates a significant safety risk for vulnerable road users (VRUs)—such as the elderly, individuals with disabilities, or children—who may be unable to clear a crosswalk before the light changes. The No Pedestrian Left Behind (NPLB) system addresses this by using computer vision to monitor crosswalks in real time and automatically extending the signal phase if a VRU is still in the street.

How the System Works

The NPLB system functions through three distinct layers:

  • Perception: A fine-tuned YOLOv12 model analyzes video feeds to detect VRUs and generate bounding boxes around them.

  • Detection: A tracking algorithm called ByteTrack follows these individuals across multiple frames, maintaining unique IDs for each person. This layer also includes a "timeout" feature that ignores temporary gaps in detection to prevent errors.

  • Control: If a VRU is detected and the remaining signal time is less than 4 seconds, the system automatically adds 3 seconds to the pedestrian phase. This ensures that those who have already started crossing have enough time to reach the other side safely.

Evaluating Detection Performance

To ensure the system is accurate, the researchers benchmarked five different object detection models using the BG Vulnerable Pedestrian (BGVP) dataset. This dataset is unique because it specifically includes labels for elderly pedestrians, wheelchair users, and children. Among the models tested, YOLOv12 achieved the highest performance with a mean Average Precision ([email protected]) of 0.756, making it the most effective choice for the NPLB system's real-time requirements.

Key Results

The researchers tested the NPLB system through 10,000 Monte Carlo simulations to see how it would perform in real-world conditions, including accounting for potential detection failures. The results showed that NPLB significantly improved safety:

  • Reduced Stranding: The rate of pedestrians being left in the crosswalk when the signal expired dropped from 9.10% to 2.60%.

  • Overall Safety: The system improved VRU safety by 71.4%.

  • Efficiency: The system only needed to extend the signal in 12.1% of crossing cycles, demonstrating that it can protect vulnerable users without causing unnecessary traffic delays.

Important Considerations

While NPLB offers a promising solution for urban safety, it is designed with specific operational boundaries. The system assumes that pedestrians are crossing legally at designated crosswalks during the appropriate signal phase. It also requires adequate lighting to function correctly. The current implementation does not address jaywalking or illegal crossing scenarios, as these would require different detection and response strategies. The system’s extension parameters, such as the 3-second boost, are intended to be calibrated based on the specific width and traffic patterns of a local intersection.

Comments (0)

No comments yet

Be the first to share your thoughts!