Noise-Tolerant Range Analysis for Autonomous Navigation

Engineering

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 5

Vision and Signal Understanding

Techniques for detecting horizontal regions, obstacles, ditches, and shoulders along a road from range data are described. The noise level in each scan line of the range image is computed and an adaptive threshold is used for noise compensation. The sources of noise and the scanning geometry for a time-of-flight range sensor are discussed and experimental results of applying these techniques to ERIM range images are presented.