Vision Only Pose Estimation and Scene Reconstruction on Airborne Platforms Michael Warren, David McKinnon, Toby Gifford, He Hu, Michael Shiel, Dawid Preller, & Ben Upcroft [email protected] RoboVis School of Mechanical and Mining Engineering We present unaided visual 3D pose estimation and map reconstruction using bundle adjustment on stereo data from ground and airborne platforms. We outline some of the issues specific to airborne platforms and how they have been overcome. We have generated consistent pose estimates for publicly available ground- based stereo datasets and those generated from our own ground and airborne platforms. [1] N. Cornelis, B. Leibe, K. Cornelis, L. Van Gool, 3D Urban Scene Modelling Integrating Recognition and Reconstruction, International Journal of Computer Vision, 78:121-141, 2008 Airborne/ground pose estimate results Snapshot of real-time depth map generation from airborne images Typical planes of interesting features in a) ground based scenarios and b) aircraft based scenarios Experimental platform and equipment. Pose estimates from ground platform [1]