Registration of 3D data, in order to obtain 3D models, has been an area where much research has been done. Several different approaches and algorithms have been proposed during the years but the Iterative Closest Point algorithm is still, even though it has been almost 15 years since it was presented, one of the most popular approaches for solving registration problems. Most of the applications of the ICP algorithm have assumed static scenes as statues, toys, buildings and cultural heritage. An interesting scenario, that not has been as frequently treated, is how the ICP algorithm deals with nonstatic scenes containing mostly forest and vegetation where some parts in the scene might move, due to wind, between different views. The main purpose with this thesis is to evaluate how the ICP algorithm deals with data representing noisy natural scenes. Some extensions of the ICP algorithm are implemented as well as a keypoint extraction approach. The keypoint extraction aims to find static points, which are reliable for registration, in the non-static scenes. A brief discussion about change detection, and whether the ICP algorithm can be used for change detection purposes, will also be held. In this thesis it is shown that the ICP algorithm can produce a decent registration of partially overlapping views representing non-static scenes. Unfortunately, the average alignment error in the registered views is larger than the laser range scanners measurement accuracy.