Three dimensional environment mapping plays an increasingly important role in mobile robotics. 3D sensors can be used to create very precise point clouds of the mapped scenes. But even with high point density, these devices only deliver a sampling of the present surfaces, not a continuous representation. Besides that, dealing with point clouds requires a lot of memory. One approach to get a continuous environment model is to calculate a polygonal mesh based on the sampled data. This thesis presents a method to automatically produce compressed polygonal meshes of arbitrary environments using a modified Marching Cubes algorithm. The initially created models are optimized using several filters to create an optimal mesh that can be used in robotic applications. The practical usability of the created maps is shown on several examples.