Abstract: “Smart phones and tablet computers are typically equipped with a camera and a powerful processor, and together with the right computer vision software they can support engineers and other professionals in a broadrange of tasks. Creating 3D models of real world objects is one job that can be very time consuming for humans if carried out manually, whereas with an adequate image processor the task can be completed very efficiently. In this paper we focus on the recognition of lattice-like objects made up of elements from a certain set of basic components. We propose a procedure incorporating a 2D detection of object parts, a 3D reconstruction, and a 3D model composition. The efficiency of the procedure is demonstrated by means of a prototype in the form of an Android app, accomplishing a 3D recognition and rendering of molecular models.”

Keywords: computer vision, pattern recognition, 3D modeling, bundle adjustment