RANGE-VIDEO FUSION AND COMPARISON OF INVERSE PERSPECTIVE ALGORITHMS IN STATIC IMAGES

Creating three-dimensional (3-D) descriptions of road boundaries from single two-dimensional images (2-D) of segmented road is a central problem for the autonomous land vehicle (ALV) road-following task. In the past, the computational cost of this 2-D to 3-D inverse perspective problem has been minimized with simplifying assumptions to allow increased vehicle speeds. The paper in part describes how heuristic assumptions may be avoided if actively scanned laser range data are used to determine the 3-space location of features in a 2-D image. An “epipolar plane” approach, commonly used to restrict search spaces in stereo correspondence problems, is used to combine data gathered from the ALVs video camera and an ERIM laser range scanner for accurate 3-D descriptions of roads. An additional objective of this work is to compare various heuristic inverse perspective algorithms with each other and the video/range scanner “fusion” approach, in terms of accuracy, failure situations, and estimated computational speed. The simplest, and earliest used, “flat-earth” algorithm made the assumption that the vehicle was at all times resting in a level attitude on an infinite flat plane. Image points were projected onto this plane. A second approach, the “hill-and-dale” algorithm, permits the plane’s elevation to vary such that image pairs of left and right road edge points (derived in a simple way after road segmentation) define a road of constant width. The “zero-bank” algorithm imposes a constant road width constraint to heuristically extracted image pairs of road edge points. The algorithm seeks image pairs of edge points that correspond to true “crossroad” road edges on curving roads. Graphical output describing the performance of the various algorithms on real world scenes will be presented and discussed. © 1990 IEEE