Quality-driven Poisson-guided Autoscanning
2014-08-25
Scanning real physical objects is becoming more and more commonplace. The scanned 3D model is crucial for reverse engineering, 3D printing, industrial design and etc. However, high quality scanning of physical objects remains a meticulous and tedious task. Only a few auto-scanning systems are emerging on the market, which are expensive and unable to scan complex objects.
Researchers of Shenzhen Institutes of Advanced Technology (SIAT) presented a novel view planning method for robot-guided auto-scanning system. Unlike previous scan planning methods, they did not aim to minimize the number of scans that needed to cover the object’s surface, but rather to ensure the high quality scanning of the model. They showed that by placing the scanner at strategically selected by Next-Best-View locations, the geometric details of the scanned object were progressively captured toward high fidelity. A patent related to this research has been applied both in China and US, which is expected to benefit the 3D modeling industry.
The research results have been accepted by SIGGRAPH Asia 2014 and will be published also on ACM Transactions on Graphics. This work is supported in part by NSFC, CAS Technology Program, Shenzhen Innovation Program, Natural Science and Engineering Research Council of Canada and the Israel Science Foundation. The research group, led by Prof. HUANG Hui at SIAT, specializes in computer graphics, computer vision and visualization.