Comparative study of methods for metric scale estimative through the fusion of inertial data and visual motion

Abstract

The purpose of this research project is to obtain the metric scale VSLAM algorithms (Visual Simultaneous Localization and Mapping) using a commercial MAV (Micro Aerial Vehicle) equipped with monocular camera to obtain 3D maps and metrics locations. Three methods are compared, the former uses the extended Kalman filter to carry out the fusion between the VSLAM data with inertial movement of the aircraft data to estimate the scale parameter. The second and third comparing the displacement estimated by VSLAM with the displacement estimated by the inertial movement of the vehicle by the perspective of a least squares problem or the maximum likelihood method, respectively. The proposed approaches will initially be developed in MATLAB with real data previously acquired. And once validated, they will be implemented as a ROS (Robot Operating System) package. Later, the development of the experiments will be done with the Parrot Bebop 2.0, a quadrotor kind MAV of with monocular camera, and the open source library VSLAM ORB-SLAM2. The results obtained will also be validated through the Vicon motion tracking cameras. (AU)