Diseño e implementación de un sistema de reconstrucción volumétrica virtual de objetos en un entorno indoor mediante cámaras estereoscópicas y procesamiento digital de imágenes.

Abstract

The objective of this work was to design and implement a virtual volumetric reconstruction system of objects in an indoor environment using stereoscopic cameras and digital image processing, for which images were obtained from a square arrangement of four stereo cameras located in the corners of the arrangement. The images were obtained using MYNT EYE S1030 brand stereo cameras, through the cascade methodology a reconstruction algorithm was developed that made use of rectified and calibrated images, the depth search was performed using the semi-block matching method. global (SGBM), therefore, the reconstruction used the triangulation method from the depth map found; To form a complete object from individual point clouds obtained from each view in the array, the algorithm adjusted each point cloud relative to its previous neighboring view using rotation and translation matrices, such that all views formed a single view. point cloud, the system was evaluated through the statistical analysis of three characteristic performance parameters for volumetric scanning systems: response speed, precision and reliability. The reconstruction system presented a mean reconstruction time of 12.84 seconds, for precision it was compared with the COLMAP software, which by means of a t-student test of 30 paired samples showed that there is no significant difference between the means of these samples. In terms of reliability, the system showed that it is not accurate, resulting in an error of 2 mm on average. It is concluded that the system is not precise as in other reconstruction methods due to the stereo technique, but it can be used for applications that do not require a high level of precision. The use of other active techniques such as laser triangulation is recommended.