Obstacle and Free Space Detection Based on U-V-Disparity

Abstract

This paper presents a novel method for obstacle and free space detection for intelligent vehicles by using U-V-disparity of stereovision. The U-V-disparity map is generated by calculating the histogram of the disparities with regards to the row and column. The obstacles and road surface appear as various straight lines in the U-V-disparity maps. By detecting these lines and reconstructing them back to the initial disparity map, obstacles can be located and measured. The free space is extracted from the subtraction of the obstacle disparity map obtained from the U-disparity from the initial dense disparity map. The experimental results show that the proposed methods can efficiently detect obstacle and free space.