Key points are not available for this paper at this time.
Obstacle detection (OD) is one of the main components of the control system of autonomous vehicles. In the case of indoor/urban navigation, obstacles are typically defined as surface points that are higher than the ground plane, but in cross-country and unstructured environments the notion of "ground plane" is often not meaningful. We introduce a fast, fully 3D OD technique that overcomes such a problem, reducing the risk of false-negatives while keeping the same rate of false-positives. A simple addition to our algorithm allows one to segment obstacle points into clusters, where each cluster identifies an isolated obstacle in 3D space. Obstacle segmentation corresponds to finding the connected components of a suitable graph, an operation that can be performed at a minimal additional cost during the computation of obstacle points. Rule-based classification using 3D geometrical measures derived for each segmented obstacle is then used to reject false-obstacles (for example, objects that are small in volume, or of low height). Results for a number of scenes of natural terrain are presented, and compared with a pre-existing obstacle detection algorithm.
Talukder et al. (Wed,) studied this question.