Automatic Detection of Planes and Cylinders from Point Clouds and Calculation of Their Parameters

Point-based models have been analyzed for different applications and depending on the demand, features are extracted by various algorithms. Surfacerecognition from point clouds is most applicable problem in machine vision and reverse engineering. In this research for estimating design parameters, planes and cylinders in point clouds are investigated and new algorithm for detecting these surfaces and calculate their parameters is proposed. According to proposed method, at first surface normal vector by using nearest neighbors must be estimated. By using surface normal vector homogeneous neighborhood is selected. Umbrella curvature is then calculated using homogeneous neighbors. The points are grouping by umbrella curvature as a criterion. Planes and Cylinders are detected by analyzing umbrella curvature values and pattern of homogeneous neighbors. A new method for calculation of planes and cylinders parameters is proposed. This method used fordetermining parameters of eachgroup of points. For evaluation that the proposed algorithm, the calculated parameters are compared with parameters are measured by coordinate measuring machine (CMM). Analyzing results shows that the proposed algorithm performs well and has appropriate abilities on automatic detection of cylinders and planes and calculation their parameters.