Vision only pose estimation and\ud scene reconstruction on airborne platforms

We aim to demonstrate unaided visual 3D pose estimation\ud and map reconstruction using both monocular and stereo vision\ud techniques. To date, our work has focused on collecting data\ud from Unmanned Aerial Vehicles, which generates a number\ud of significant issues specific to the application. Such issues\ud include scene reconstruction degeneracy from planar data, poor\ud structure initialisation for monocular schemes and difficult 3D\ud reconstruction due to high feature covariance.\ud Most modern Visual Odometry (VO) and related SLAM\ud systems make use of a number of sensors to inform pose and\ud map generation, including laser range-finders, radar, inertial\ud units and vision [1]. By fusing sensor inputs, the advantages and\ud deficiencies of each sensor type can be handled in an efficient\ud manner. However, many of these sensors are costly and each\ud adds to the complexity of such robotic systems. With continual\ud advances in the abilities, small size, passivity and low cost of\ud visual sensors along with the dense, information rich data that\ud they provide our research focuses on the use of unaided vision\ud to generate pose estimates and maps from robotic platforms. We\ud propose that highly accurate (�5cm) dense 3D reconstructions\ud of large scale environments can be obtained in addition to the\ud localisation of the platform described in other work [2].\ud Using images taken from cameras, our algorithm simultaneously\ud generates an initial visual odometry estimate and scene\ud reconstruction from visible features, then passes this estimate\ud to a bundle-adjustment routine to optimise the solution. From\ud this optimised scene structure and the original images, we aim\ud to create a detailed, textured reconstruction of the scene.\ud By applying such techniques to a unique airborne scenario,\ud we hope to expose new robotic applications of SLAM techniques.\ud The ability to obtain highly accurate 3D measurements\ud of an environment at a low cost is critical in a number of\ud agricultural and urban monitoring situations. We focus on cameras\ud as such sensors are small, cheap and light-weight and can\ud therefore be deployed in smaller aerial vehicles. This, coupled\ud with the ability of small aerial vehicles to fly near to the ground\ud in a controlled fashion, will assist in increasing the effective\ud resolution of the reconstructed maps