| This thesis describes novel methods for processing three-dimensional laser-scanner data for automotive styling design. We introduce methods for smoothing, surface fitting, and surface reconstruction, each one addressing distinct but associated problems with using scanned data for industrial design. Our smoothing operator removes geometric noise from the data over varying scales while preserving sharp edges and corners. Our surface fitting method automatically segments the scanned data into distinct regions approximated by parametric surface patches. Finally, our surface reconstruction method takes approximating surface patches and uses a spherical cover and an implicit partition of unity to create a piecewise-smooth reconstruction of the scanned points. All of our approaches take into consideration the strict engineering and aesthetic constraints of automotive design, and we demonstrate their effectiveness on real-world data. |
