Pedestrian Object Distance Prediction on Robot Car Using YOLOv4 and Bounding Box Analysis

This study focuses on enhancing safety in robot car navigation by developing a pedestrian distance prediction system using the YOLOv4 algorithm and bounding box analysis. YOLOv4, known for its speed and accuracy, efficiently detects pedestrians and estimates their distance from robot cars. The research involves creating datasets with distance variations from 1.5 to 5 meters and utilizing a 40th-order polynomial interpolation model for distance estimation. Results indicate that YOLOv4 offers precise detection, and the interpolation model predicts distances with an average error of 2.69% and a correlation coefficient of 0.9223, improving navigation safety and efficiency in dynamic environments.