Automatic alignment of piping system components and generation of CAD models of industrial site plants

Abstract

The ability to (semi-)automatically obtain CAD models from physical installations has two important benefits: (i) it can be used to identify, as soon as possible during a con struction process, any deviations from the original designs; and (ii) it can be used to document complex installations for which CAD representations are outdated or inexis tent. Both scenarios have important practical and economic value. An ongoing project in our research group aims to reconstruct CAD representations from point clouds of in dustrial sites. However, pose estimation of pipes and piping system components is not perfect, resulting in misalignments in the reconstructed scene, which is unacceptable for a CAD model. For this undergraduate thesis, I propose to use optimization techniques to fix these misalignments. I also propose to convert the detected pipes and piping system components into actual CAD model representations for a popular commercial CAD soft ware, namely AutoCAD Plant 3D.