Avaliação da digitalização 3D obtida com o Kinect One

Abstract

This work evaluates the measurement in a way made possible by Kinect One, a sensor based on time of flight technology marketed by Microsoft Corporation originally to be used in conjunction with the Xbox One console. Five different geometric objects were subjected to 3D scanning using this sensor. In the 3D scanning procedure and analysis of the results produced, different pieces of software were used: the MeshLab application for the reconstruction, the presentation of the 3D models and obtaining measurements; the ARUCO coded target library for creating references for the clipping of points clouds; the PCL (Point Cloud Library) library for segmenting and cutting areas of interest of the point clouds and general manipulation of the point clouds; as well as specific applications built for the job. The final point clouds of each scanned object were compared to reference models visually and through objective metrics. Among the measures used are area, volume of 3D models, as well as variants of Hausdorff distance. The results indicate that the accuracy of the point clouds obtained by Kinect One is higher in the minimum recommended working distance range (0.5 m). The error values obtained in this range vary from 2.8 mm to 6.8 mm in the mean error and with a standard deviation range varying from 3.0 mm to 7.2 mm.