Vehicle Localization Based on Phase Information and Edge Projection

摘要

A novel vehicle localization algorithm based on phase information and contour projection is proposed. Firstly, HSV color model and RGB color model are combined, and contours of the vehicles in image are detected based on phase congruency. Secondly, contour projection method is utilized to get candidate vehicle regions. Finally vehicle region distinguishing algorithm are applied to the candidates, as a result the exact positions of vehicles are obtained. Experiment results show that the new algorithm has strong robustness and suffers small from the luminance, contrast of image, noise and shadows. Furthermore, it is suitable for images including more than one vehicles and has higher accuracy compared with common algorithms.