Researcher(s)
- Bryton Burrows, Computer Engineering, University of Delaware
Faculty Mentor(s)
- Mark Mirotznik, Department of Electrical and Computer Engineering, University of Delaware
- Rachel Ciganik, Department of Electrical and Computer Engineering, University of Delaware
Abstract
Accurate tracking of surface deformation during composite forming is critical for predicting part performance and engineering predistortion patterns. This work details the development of an automated 3D optical metrology and feature-tracking pipeline designed to quantify marker displacements between flat and formed composite geometries.
Feature detection methodologies evolved significantly across scanner generations. Initial geometric segmentation on a Keyence VL-500 relied on Gaussian curvature and local surface normal vectors to detect physical ink relief. While functional, this approach proved computationally slow and sensitive to substrate surface roughness. An intermediate hybrid approach mapped top-down 2D PNG images onto 3D point clouds to enable color filtering but remained labor-intensive. Upgrading to a Keyence VL-800 unlocked full OBJ/MTL texture exports, enabling streamlined HSV and RGB color-space segmentation directly on textured 3D meshes for robust marker classification.
To quantify geometric strain, a custom MATLAB pipeline extracts nominal marker positions directly from CAD STL geometry. Detected surface markers from flat and formed scans are aligned to the nominal grid using projection, iterative closest point (ICP) registration, and KD-Tree nearest-neighbor matching algorithms. By calculating 2D in-plane displacement vectors between matched markers across both states, the framework uses scattered interpolation to construct continuous, full-field predistortion correction maps.
This integrated workflow eliminates manual alignment bottlenecks, enhances spatial registration accuracy, and delivers an actionable predistortion model to compensate for geometric distortion in advanced composite manufacturing.