Traffic Sign Detection System for Locating Road Intersections and Roundabouts: The Chilean Case

Area of Science:

• Computer Vision
• Robotics and Automation
• Intelligent Transportation Systems

Background:

• Accurate detection of traffic signs like stop and yield signs is crucial for driver assistance systems, particularly near intersections and roundabouts.
• Existing methods often struggle with variations in lighting, scale, and occlusion, impacting reliability in complex road scenarios.
• Robust detection systems are needed to improve vehicle safety and navigation accuracy.

Purpose of the Study:

• To develop and evaluate a novel traffic sign detection method utilizing statistical color templates.
• To specifically address the detection of stop and yield signs in challenging environments like road intersections.
• To analyze the method's detection rate performance as a function of distance and compare it against established algorithms.

Main Methods:

• Employs a segmentation approach using the RGB-normalized (ErEgEb) color space and a chromaticity filter to generate Regions of Interest (ROIs).
• Utilizes statistical templates, considering mean and standard deviation of normalized color, applied at 10 scales for robust sign identification.
• Incorporates a classification stage using YCbCr and ErEgEb color spaces with background removal via a probability function based on chromaticity.

Main Results:

• The proposed method achieves high detection rates: 87.5% for yield signs and 95.4% for stop signs at distances under 48 meters.
• Achieves 100% detection accuracy for both yield and stop signs at distances under 30 meters.
• Significantly outperforms the Viola-Jones method, which shows detection rates below 20% at distances between 30-48 meters and up to 60% between 20-30 meters.

Conclusions:

• The statistical color-based template method offers a robust and effective alternative for traffic sign detection, particularly at intersections.
• The approach demonstrates superior performance compared to shape-based methods like Viola-Jones, especially in close-range scenarios.
• Validated through real-world video data from Santiago, Chile, the method shows promise for enhancing driver assistance systems.