 We developed an efficient and accurate method for point-wise semantic classification of 3D point clouds. Our approach handles unstructured and inhomogeneous point clouds, such as those derived from terrestrial LiDAR or photogrammetry, and is able to process point clouds with millions of points in just minutes. To reduce computational complexity, we defined a multi-scale neighborhood structure which allowed us to create a compact and expressive feature set. This feature set was then used to classify each point into one of several categories, including buildings, roads, vegetation, etc. We evaluated our method on two data sets, one from a mobile mapping platform and another from terrestrial laser scanning, and found that it outperformed existing methods in terms of accuracy and speed. This article was authored by T. Hackel, J. D. Wagner and K. Schindler.