Monocular Vision Based Navigation and Localisation in Indoor Environments

Abstract

This paper describes a set of tools and algorithms to enable vision based navigation for a Micro Aerial Vehicle (MAV) flying in an indoor environment. An optical flow balancing strategy is used to avoid obstacles in the vehicle's path. An Extended Kalman Filter (EKF) based visual odometry system is described which can estimate the trajectory of the vehicle since low cost MEMS based Inertial Measurement Units (IMU) are not capable of providing drift free attitude and position information. To extract depth information using a single camera, a simple triangulation based method using a laser pointer is presented. This enables the MAV to detect oncoming obstacles directly in front of it and execute a turn to avoid collision in case the optical flow balancing algorithm is unable to detect an obstruction.