Embeded Visual Servoing for aerial drones
I am member of the project FUI ADOPIC which aims at developping tools to help diagnose problems for large structures using conventional imagery (such as bridges, barrages, ... ). The I3S studies the problem of stabalising the DRONE from Infotron with respect to a set of target images given by an operator. The work carried out in this project aims at developping new algorithms for localisation and servoing by vision. This requires developping and integrating these algorithms within an embedded system and coupling them with existing systems onboard the drone. The measure of the velocity is essential for all control laws, however, obtaining this measurement is far from being straight forward. We have recently shown that a planar target is rich in information and efficient to compute on embedded hardware whilst remaining robust to varying natural conditions. The extracted measure isnt the velocity of the drone by this homography measure can be exploted to develop stabilisation techniques for the position when combined with inertial sensors.
The platform we have developped is based on the Kontron ComExpress technology.
Associated publications : ICRA 2011, "Real-time Direct Tracking of Color Images in the Presence of Illumination Variation",