Many applications of computer graphics use real world 3D object models. It is a well known technic to use a 3D scanner to provide this virtual 3D data. Different types of 3D scanners have been developed for different types of applications. My task was to develop a triangulation based 3D scanner system, which uses laser light to illuminate the target object. Such scanners are used to scan objects of size of several ten centimeters. The laser light illuminates a slice of the object surface. With the assistance of a camera and a computer, the 3D model of the target object can be reconstructed stripe by stripe.
To identify the stripes of the laser light, a stripe detector algorithm is required. The accuracy of the scanned model depends highly on the accuracy of the stripe detector. I examined some peak detectors which have subpixel accuracy. The algorithms are presented here.
One of my goals with the development of this system is to create the opportunity to examine the behavior of these stripe detection methods in a real 3D scanning environment. The workflow of the scan consists of the hardware control, the stripe detection, and mesh generation. The camera must be calibrated before scanning, which is performed by a 2D homography based calibration method.
Regression is sused to measure the accuracy of the stripe detectors, and some screenshots show some 3D models of scanned objects.