Effectiveness of rough initial scan for high-precision automatic 3D scanning

Abstract

Herein, we present an efficient method for the high-precision automatic 3D scanning of unknown objects for mechanical parts. Our method comprises two phases, namely a rough initial scan and a precision scan. The goal of the rough initial scan is to scan the rough shape rapidly and to provide scan data for the precision scan, thereby reducing the duration of the entire process. Researchers have attempted to provide rough information regarding an object before precision scanning, e.g. by building a rough three-dimensional (3D) model using 2D images or capturing the shape in advance using a low-accuracy scanner with a larger view frustum. However, our two-phase scanning method uses a single type of high-precision scanner for scanning the rough shape and also for the precision scan, which comes afterwards. In the rough initial scan phase, the next scanning view is determined based on the scan data captured by the latest view, eventually forming helical shape movement. In this study, we apply the two-phase scan method to 18 types of models in a virtual 3D scanning environment. For diverse configurations, the model size is adjusted from small to large relative to the virtual scanner's view frustum, and the number of next best views calculated per iteration during the precision scan phase is adjusted. We demonstrate that the rough initial scan provides a certain amount of scan data rapidly regardless of the model size and shape. Furthermore, we demonstrate that the two scan methods complement each other, thereby reducing the overall process time and workload. ？ 2021 The Author(s). Published by Oxford University Press on behalf of the Society for Computational Design and Engineering.