Energy-Efficient Pedestrian Marker Light Based on Pose-Probability for Computer Vision

The pedestrian marker lights of modern matrix headlights improve road safety at night. However, existing solutions illuminate a wide field around the pedestrian. A more precise illumination of pedestrian contours has the potential to reduce energy consumption and increase pedestrian visibility. The contribution at hand presents a novel approach for developing a luminous intensity distribution (LID) that is resized to the bounding box of pedestrians once they are detected. This LID is based on a superposition of hundreds of images of human poses, which are analyzed for the probability of each pose. Simulation-based evaluations of the new approach show a reduction in energy consumption of up to 29.7 % compared to full illumination of the pedestrian while maintaining comparable detection quality to computer vision.