Reconstrução 3D de Blocos de Rochas Ornamentais Utilizando Visão Estéreo e SfM

Abstract

Traffic accidents involving the transport of ornamental stones are largely due to non-compliance with legislation. In addition, there is little inspection, which contributes to an increase in disobedience. However, recently, there has been a great advance in the field of computer vision, especially in the area of semantic analysis of objects. In this sense, this work proposes an alternative approach to the inspection of truck loading carrying ornamental stones. The approach involves a pipeline of 3D reconstruction from images, a semantic segmentation network, and a cuboid fitting algorithm. To reduce processing time, a keyframe selector algorithm is proposed. The presented pipeline has shown practical potential, although changes are still needed to bring it closer to a real-time operation.