Abstract: In this paper, an ellipsoid detection and extraction algorithm is proposed. The
algorithm is designed and implemented based on the RANSAC (RANdom SAmpling
Consensus) framework. Firstly, several minimum sets, one of which can define an ellipsoid, are
selected from the point clouds by using random sampling strategy. Then, candidate set is formed
from the effective ellipsoids, which come from the minimum sets and pass through the
verification. Finally, the best approximate ellipsoids are extracted after the evaluation step using
the score function. Experiments demonstrate that the algorithm is robust since it can effectively
extract the ellipsoids from either manually synthesized point clouds or the raw data acquired by
the scanners.

Key words: point clouds, primitive extraction, ellipsoid, RANSAC