A Vision-Based Algorithm for UAV State Estimation During Vehicle Recovery

Abstract

A computer vision-based algorithm for Unmanned Aerial Vehicle state estimation during vehicle recovery is presented. The algorithm is intended to be used to augment or back up Global Positioning System as the primary means of navigation during vehicle recovery for UAVs. The method requires a clearly visible recovery target with markers placed on the corners in addition to known target geometry. The algorithm uses clustering techniques to identify the markers, a Canny Edge detector and a Hough Transform to verify these markers actually lie on the recovery target, an optimizer to match the detected markers with coordinates in three-space, a non-linear transformation and projection solver to observe the position and orientation of the camera, and an Extended Kalman Filter (EKF) to improve the tracking of the state estimate. While it must be acknowledged that the resolution of the test images used is much higher than the resolution of images used in previous algorithms and that the images used to test this algorithm are either synthetic or taken in static conditions, the algorithm presented does give much better state estimates than previously-developed vision systems.