Accurate Correction of Measurement Errors - A series of algorithms to accurately remove measurement noises
in this paper we investigate the spatial distribution of measurement noise at sharp edges. A local cylindrical coordinate system is used to decompose the measurement noise at sharp edges into to two components, magnitude and angle. The spatial distributions of these two components are analyzed at sharp edges for both touch probe and laser sensor. A new concept, statistical binary tree, is proposed for representing the spatial noise distribution on a sharp edge, and a synthesis simulator is proposed for accurate reproduction of measurement noise at sharp edges.
In this paper we propose a new algorithm for accurate correction of surface noises of polygonal meshes. It consists of three basic components: (a) feature-preserving pre-smoothing; (b) partitioning of feature and non-feature regions; © second-order predictor for non-feature regions and median filter for feature regions. The unique contributions of our approach include (a) an idea of partitioning an input surface into feature and non-feature regions so that different smoothing algorithms, which are best suited for either feature or non-feature regions can be, respectively, applied; (b) a second-order predictor that provides higher smoothing accuracy and better convergence on smoothly curved surfaces. In comparison with several existing algorithms, our algorithm is evaluated quantitatively in terms of surface normal and vertex distance error metrics. Numerical experiments indicate the effectiveness of our approach in the aspects of convergence and accuracy. Copyright © 2005 John Wiley & Sons, Ltd.