(Georgia Institute of Technology, 2023-08)
DeVries, Carl; Christian, John A.
Feature matching techniques often encode a brightness pattern from an image
without knowledge of the underlying scene. This makes feature matching difficult
when illumination conditions change significantly between images or are not consistent
with onboard maps. The brightness patterns in images of the lunar surface
can be partially described by a surface reflectance model which can often be parameterized
by quantities known onboard from a sun sensor and calibrated camera.
Unfortunately, these models can be intractable due to their complexity. This work
develops a reduced-order lunar reflectance model for future illumination-informed
feature descriptor development.
(Georgia Institute of Technology, 2022-10)
Mancini, Michela; DeVries, Carl; Thrasher, Ava C.; Christian, John A.
Science images of the Moon and Mars are
often captured with pushbroom cameras. Craters with elliptical
rims are common objects of interest within the the
images produced by such sensors. This work provides a
framework to analyze the appearance of crater rims in
pushbroom images. With knowledge of only common ellipse
parameters describing the crater rim, explicit formulations
are developed and shown to be convenient for drawing
the apparent crater in pushbroom images. Implicit
forms are also developed and indicate the orbital conditions
under which craters form conics in images. Several
numerical examples are provided which demonstrate how
different ellipse formulations can be interpreted and used
in practice.
(Georgia Institute of Technology, 2022-08)
DeVries, Carl; Christian, John A.; Hansen, Michael; Crain, Tim
Future missions to the lunar surface are expected to make use of LIDAR sensors
for navigation during landing. This is especially true when the lunar landing
must be accomplished under lighting conditions that are undesirable for camera based
navigation. Moreover, when local maps of the lunar terrain are also poor or
unavailable to the lander in real-time, concepts from LIDAR odometry (LO) are
highly relevant. This work develops an algorithmic framework for LO suitable for
supporting future lunar exploration missions.