Classification of 3D shape deviation using feature recognition operating on parameterization control points

A numerically efficient expert system for evaluation of 3D shape based on features extracted from the parameterization control points data-set is developed. Reference shapes are stored and periodically compared with current shapes at the level of windowed control-point data grids for the purpose of detection of 3D shape deviation. Classification heuristics for the respective types of deviations which operate on the control point sets rather than the original raw point clouds are developed based on operations of windowing, coordinate transformation, filtering and singular value decomposition (SVD)/principal component analysis (PCA). The methodology is demonstrated with the cases of detecting and computationally recognizing local impact damage and cavities, narrow gaps or fatigue cracks and wear-based surface deterioration on a wind turbine blade. (C) 2014 Elsevier B.V. All rights reserved.