Direct Measurement from Scan Data with Adaptive Moving Least-Squares Surfaces under Controlled Spatial Dependency

点击翻译按钮获取中文摘要

The research objective of this award is to apply the method of moving least-squares to obtain a faithful representation of as-measured surfaces, from which dimensional characteristics of both engineered artifacts and natural objects can be accurately extracted. A moving least-squares surface defines a continuous surface directly from a set of points, a canonical representation of data output from various three-dimensional scanning systems. The proposed approach is to adaptively vary kernel widths of Gaussian functions and to control spatial dependency of deviations between points and the moving least-squares surface. If successful, this research would lead to a paradigm shift in computational metrology, from current intermediate surface reconstruction with accuracy lost in translation to direct measurement from scan data. The research will positively impact a host of industries where three-dimensional scanning is used such as aerospace, automobile, mass customization and biomedical applications. This research would shorten inspection cycles, reduce defect rates, and improve product quality, thus broaden the usage of three-dimensional scanning technologies in manufacturing systems, and thereby strengthen domestic manufacturer's competitiveness. The enhanced accuracy in scan data processing would also benefit biomedical industry and lead to enhanced biomedical imaging and diagnosis. Through its integrated research, education and outreach activities, this project will provide advanced knowledge in computational metrology for students from high schools to graduate schools and will increase domestic students' interest in science and engineering and therefore strengthen our competitiveness in the global workforce.