• dlt
• ottimizzazione
• stereo camere
• visual odometry

MAVs are naturally unstable platforms exhibiting great agility and they thus require a trained pilot to operate them. Industrial inspection with MAVs is a challenging task, and the ability to hover with high accuracy and stability is a key issue. In addition, payload on micro aerial vehicles is very limited, and the use of lightweight sensing equipment directly results in longer flight duration or the ability to carry additional payload.

This motivates the use of vision sensors for control. In this thesis, a modular framework that allows the user to test different components of a stereo vision based pose estimation algorithm has been created. Using this framework, different approaches and different implementations for the steps of the algorithm have been tested.

As a result, two approaches using SURF and SIFT features and descriptors based matchers are compared and a solution for the inspection of large industrial building.