Complex ground plane detection based on V-disparity map in off-road environment

The work in the paper is to apply image pairs gathered by vehicle-mounted stereo cameras to detect the position of ground. The research focus is to detect complex ground planes for vehicle from image pairs based on investigating of the V-disparity map. This paper describes an enhancement of traditional V-disparity algorithm for off-road environment especially. The enhanced method can acquire the parameters of ground plane such as slope and pit. Experimental results with real data from stereo cameras mounted on a vehicle moving in off-road environment are presented. According to the simulation results, by comparison with traditional V-disparity algorithm, the average of recognition rate for ground using the enhanced V-disparity algorithm increases from 37.68% to 86.67%. The enhanced method can minimize errors of ground representation, and it is an efficient way to extract more details of ground structure than traditional V-disparity algorithm. The process is without dealing with any explicit structure such as road edges or lane marking. So it can be used for autonomous vehicle driving in off-road environment in the future.