A great effort has been done during last years to improve range image segmentation results. The efficacy of the algorithms is affected by the parameters tuning.

In this work two well-known search techniques have been applied to this task: genetic algorithms and simulated annealing. These techniques are adopted in cascade: the former to obtain a rough seed point set and the latter to have a more precise refinement of suitable solutions.

We addressed our efforts towards the range segmenter proposed by the University of Bern, that seems to be the best in term of versatility, being able to segment planar and curved surfaces, and in term of speed and quality of the performed segmentations.