Rover navigation using stereo ego-motion

An Experimental Study of Aerial Stereo Visual Odometry

“…Also unlike ground robots, the camera pointing angle on a helicopter is not restricted by proximity to the terrain surface -camera angles from horizontal (zero degrees down) to nadir view (90 degrees down) are possible. Our work complements the study in (Olson et al, 2003) of the optimal camera field of view for planetary rover visual odometry.…”

“…Our visual odometry algorithm is based on the approach presented in (Olson et al, 2003) and originally described in (Matthies, 1989). We track point landmarks through sequential stereo image pairs, triangulating the 3D positions of the landmarks at each time step from their projections into the left and right camera images.…”

A neural model of how the brain computes heading from optic flow in realistic scenes

“…Most use feature tracking, sometimes with stereo vision, in concert with a global positioning system (GPS) and inertial sensor-based data (Agrawal, Konolige, & Bolles, 2007;Alenyà, Martinez, & Torras, 2004;Olson, et al, 2003). Autonomous robots are assessed in terms of their ability to navigate from some start point to a goal while avoiding obstacles.…”

Long‐range rover localization by matching LIDAR scans to orbital elevation maps

“…VO (Olson et al, 2003) tracks features across pairs of images to measure changes in pose. Features are typically extracted using image processing techniques such as Scale Invariant Feature Transforms (SIFT) (Lowe, 1999;Barfoot, 2005) or Speeded Up Robust Features (SURF) (Bay et al, 2006).…”

“…Since each estimate depends on the previous, error may continue to grow without bound. Olson et al (2003) reduced this error growth to linear rates by including absolute measurements of orientation in the VO algorithm. In field experiments, Konolige et al (2007) showed a similar VO algorithm to yield less than 0.1% error over a 9km traverse in rough-terrain.…”