Infrastructure-Based Vehicle Localization through Camera Calibration for I2V Communication Warning.

Area of Science:

• Computer Vision
• Robotics
• Transportation Engineering

Background:

• Object detection and tracking are crucial for Advanced Driving Assistance Systems (ADASs) and Connected Autonomous Vehicles (CAVs).
• Road intersections, particularly in urban areas, are high-risk zones for traffic accidents.
• Existing systems require improvement for accurate vehicle monitoring at intersections.

Purpose of the Study:

• To develop and evaluate an Intersection-to-Vehicle (I2V) warning system for enhanced road safety.
• To detect and track vehicles at intersections, identifying potential obstacles for other road users.
• To provide accurate vehicle localization using roadside camera data.

Main Methods:

• A single camera installed on a Roadside Unit (RSU) performs image detection and tracking.
• A localization algorithm utilizes linear transformations with intrinsic camera parameters, height, and pitch angle.
• Vehicle position is determined in the global reference frame (latitude and longitude).

Main Results:

• The localization algorithm achieved an average error of 0.62 meters.
• Detection reliability demonstrated a low false positive rate of 1.9%.
• Missed detection rate was recorded at 3.6%.

Conclusions:

• The proposed I2V system effectively detects and tracks vehicles at intersections.
• The localization algorithm provides accurate positioning for safety warnings.
• This technology can significantly contribute to reducing intersection-related accidents.