3D reconstruction system for semi-automatic estimation of objects length and area by means of stereo vision

Abstract

It is mandatory to characterize dimensionally the manufactured industrial pieces for quality control purposes. As it is not possible to touch some pieces when trying to retrieve dimensional information, then non-invasive techniques are required to do so. Stereo vision is a passive technology which is both robust and accurate for non-invasive applications. For this reason, in this work we describe the design and implementation of a 3D reconstruction system for the estimation of the length and area of certain objects. This tool allows to easily incorporate new image correspondence techniques to its main execution pipeline. We carry some experiments and show certain benefits when selecting an accurate image correspondence technique for the estimation of the length and area.