eduzhai > Applied Sciences > Engineering >

Lunar Terrain Relative Navigation Using a Convolutional Neural Network for Visual Crater Detection

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 6 pages

Abstract: Terrain relative navigation can improve the precision of a spacecraft sposition estimate by detecting global features that act as supplementarymeasurements to correct for drift in the inertial navigation system. This paperpresents a system that uses a convolutional neural network (CNN) and imageprocessing methods to track the location of a simulated spacecraft with anextended Kalman filter (EKF). The CNN, called LunaNet, visually detects cratersin the simulated camera frame and those detections are matched to known lunarcraters in the region of the current estimated spacecraft position. Thesematched craters are treated as features that are tracked using the EKF. LunaNetenables more reliable position tracking over a simulated trajectory due to itsgreater robustness to changes in image brightness and more repeatable craterdetections from frame to frame throughout a trajectory. LunaNet combined withan EKF produces a decrease of 60 in the average final position estimationerror and a decrease of 25 in average final velocity estimation error comparedto an EKF using an image processing-based crater detection method when testedon trajectories using images of standard brightness.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×