Estimation of Road Surface Shape and Object Height Focusing on the Division Scale in Disparity Image Using Fisheye Stereo Camera

In this paper, we propose a novel algorithm to estimate road surface shape and object height using a fisheye stereo camera. Environment recognition is an important task for Advanced Driver-Assistance Systems (ADAS). However, previous studies use sensors with a narrow measurement range. They also use the constraint that the road is flat and cannot deal with changes in the road slope. We use a fisheye stereo camera, which can measure wide and dense 3D information, and design a novel algorithm focusing on the scale of division on the disparity image to overcome these defects. Experiments show that our method can detect an object. Furthermore it is shown that measurement accuracy of obstacle height and distance to obstacles strongly depends on the 3D measurement accuracy.