Diseño de un sistema de reconstrucción ambiental 3D indoor que permita determinar la trayectoria de un robot móvil mediante análisis y procesamiento de imágenes.

Abstract

The present Curricular Integration Work had the objective of designing and implementing a 3D reconstruction system of interior environments, which allows defining the trajectory of a teleoperated mobile robot, for which the mobile robot that contains a MYNT EYE S1030 camera was implemented to capture images that, through the stereo vision technique, to be able to reconstruct an indoor scenario, the operation of the system was determined by means of a randomness analysis. The reconstruction of environments and subsequent calculation of the trajectory was defined by stages, which were: Analysis, Design, Implementation, Evaluation and Maintenance of the system, where substages are defined for the reconstruction algorithm programmed in MATLAB. For the randomness and comparative analysis, it was decided to apply different statistical tests such as the Shapiro test, the F test and the Student's T test to evaluate the system, from which it was obtained that the system differs in measurements compared to the commercial tool COLMAP; In the analysis of accuracy with the real environment, it is found that the system differs, but on a smaller scale, making an absolute error calculation, giving an average error of 3.27 meters with respect to the real value, and finally, a reconstruction of the scenario was made. indoor by varying its parameters. It is concluded that the system gives acceptable and not highly reliable results due to the camera and the quality of the captured images, since the creation of the point clouds generates noise, so the ICP algorithm for joining the point clouds overlaps the point clouds that gradually lose the real dimensions of the stage. It is recommended to complement the stereo vision technique with depth sensors to obtain values close to the real ones and thus improve the reconstruction of environments.