This paper presents a method for the real-time genera- tion of grids from 3D line scan data for in-line scan pre- views and the evaluation of large point clouds derived from different 3D-scanners. By exploiting the underlying mea- suring principles, we generate regular grids for each scan operation even if the sensor movement is non-linear. These grids are finally used for NURBS patch approximations, which enable the determination of higher order features, e.g. curvature and quality evaluations. Experimental re- sults at the example of different point clouds illustrate the effectiveness of our methods in practice.