Analyzing of large data and visual information is becoming more and more crucial in today's world. Faster and more accurate analysis for visual data is vital for industries, which depend on visual information, such as robotics and automated drive systems. No matter how the information is obtained, it must be analyzed in order to draw conclusions and create hypothesis. One of these factors, which is to be analyzed are critical points. With the help of critical point analysis, borders of objects in a scene can be found. Here robots could calculate grasping points of the object and use this information for further tasks. In this thesis a pipeline for computing critical area curvature on point clouds is created. The entire information of the processing background will be explained and it will be shown how these interact to create the pipeline. Computations of different normal estimations and curvatures are going to be introduced and compared. Later these will be implemented to a pipeline, which computes critical areas of the scanned point clouds. The thesis concludes with the comparison for real and synthetic point clouds.