Three-dimensional volume and position recovering using a virtual reference box.

Abstract

This paper proposes a procedure to determine the position and volumetric information of a 3-D object, based on two previously calibrated camera images. The positioning and volumetric information is achieved by inscribing the object in a virtual bounding box, created from known vertices and vanishing points. The images of a starting virtual box are adjusted to fit a region of interest around each projection of the solid object. The volumes and positions defined by the adjusted boxes are approximations for the dimensions and location of the solid. The work displays the application of such a technique in an automated manufacturing process.