Progresses on Integration of Orbital, Descent and Ground Imagery for Topographic Capability Analysis

1
Progresses on Integration of Orbital, Descent and
Ground Imagery for Topographic Capability
Analysis in Mars Landed Missions

• Ron Li, Kaichang Di, Ju Won Hwangbo and Yunhang
  Chen
• Mapping and GIS Laboratory, CEEGS, The Ohio
  State University
• May 7, 2008

2
Contents

• Rigorous photogrammetric modeling of HiRISE
  stereo images
• HiRISE topographic mapping at Mars Exploration
  Rover (MER) landing sites
• MER mission support
• Preliminary results of landmark extraction and
  matching from orbital and ground images

3
HiRISE Imaging Geometry

• HiRISE is a push-broom imaging sensor with 14
  CCDs (10 red, 2 blue-green and 2 NIR).
• Each CCD consists of 2048 pixels in the
  across-track direction and 128 pixels in the
  along-track direction.
• After excluding overlapping pixels, HiRISE can
  generate images with up to 20,264 cross-track
  observation pixels
• The ground resolution is 30 cm/pixel at 300 km
  altitude.
• Using HiRISE instrument kernel, pixel positions
  with respective to individual CCD centers are
  converted to pixel positions with respect to
  HiRISE optical axis (HiRISE frame ).

AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
4
Camera Position and Pointing
The MBF frame is described as the
following  Zmbf Mars spin axis, pointing
toward Martian North Pole. Xmbf Vector lies in
the Mars equatorial plane and intersects the
prime meridian. Ymbf Vector lies in the Mars
equatorial plane and completes a right handed
coordinate system
Camera position and pointing information for each
scan line can be retrieved from SPICE kernel.
5
Mathematical Model for Bundle Adjustment

• Collinearity Equation

2nd order polynomial for modeling the exterior
orientation parameters of HiRISE scan lines
6
HiRISE Stereo Images of Spirit Landing Site

• Study area
• 14.6 S latitude
• 175.5E longitude
• Columbus Hills
• Entire Spirit traverse so far
• TRA_001777_1650 (left)
• Dec 12, 2006
• 40,000 rows
• 26.3 cm/pixel
• TRA_001513_1655 (right)
• Nov 22, 2006
• 40,000 rows cropped from 80,000 rows
• 27.1 cm/pixel

AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
7
Workflow of the Hierarchical Stereo Matching
Process
AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
8
Matching Result Verification with Manually
Matched Check Points
AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
9
Bundle Adjustment with no GCP Point
Distribution
Tie Points Red, 460 Check Points Blue, 460
10
Bundle Adjustment with no GCPError Vectors
Before and After BA
Error vectors(50 times exaggeration) Along
track direction Mean residual 1.5 pixels Cross
track direction Mean residual -0.007
pixel Along track direction standard deviation
4.4 pixels Cross track direction standard
deviation 0.03 pixel
Error vectors (200 times exaggeration) Along
track direction Mean residual 0 pixel Cross
track direction Mean residual 0 pixel Along
track direction standard deviation 0.29
pixel Cross track direction standard deviation
0.0006 pixel
11
Orbital-Ground Correspondence for BA with Ground
Control
5000 x 2048 pixels
Husband Hill
Home Plate
Red circle orbital ground tie points (GCP) Red
dot orbital tie points Blue dot orbital check
points.
12
Difference in GCP Coordinates from Orbital and
Ground Imagery
From left to right, it is GCP 1, 2, 3, 4
respectively
13
Orbital Jitter

• Small motions of the spacecraft around its
  nominal pointing are called jitter.
• Jitter can be filtered out by subtracting the
  best-fitting polynomial from the original
  telemetry HiRISE pointing angle data.

Jitter extracted from HiRISE Image
(PSP_001513_1655)
14
Topographic Effect of Orbital Jitter
15
Topographic Effect of Orbital Jitter

• Compare the two footprints in both along
  track and cross track direction.

m
m
Line
Line
Difference in along track direction (lt1m)
Difference in cross track direction (lt2.5m)
16
Image Distortion Correction

• A ray coming from a ground feature to the camera
  perspective center intersects the ideal focal
  plane at the position of the image tie point
  (x1,y1) and intersects the actual (with jitter
  effect) focal plane at (x2,y2).
• Knowing the jitter, a correction is made to
  correct (x2,y2) to (x1,y1).
• This means (x1,y1) should have been the image
  point if the trajectory yielded to the best
  fitting polynomial.

17
Bundle Adjustment and Final EO

• BA is done based on the undistorted image
  points, so that the polynomial model fits.
• Final EO jitter term refined polynomial term
• Evaluation by back projection residuals using
  final EO and measured check points.

Without Jitter Incorporated
With Jitter Incorporated
18
HiRISE Topographic Mapping at MER Landing Sites
3D view of digital elevation model (DEM)
The mapped area (red box) is 4000m X 3500m. (X,
Y) coordinates are in MER landing site local
coordinate system, while Z is referenced to a
radius of 3392 km.
19
Seamless DEM Generation
In addition to tie points between left and right
images, tie points between adjacent CCDs of the
same image (stitch points) are also incorporated
in the BA. Compared to the previous DEM,
inconsistencies between different CCDs are
removed.
20
Application in MER Operation for Finding Winter
Heaven for Spirit Rover
21
3D Points from Hard Baseline and Wide Baseline
Ground Images
Mapped Area 150m from rover Hard baseline Sol
1348 (AVLF, Pancam red stereo pairs, L2,
R2) Wide baseline (8m) Sol 1348 hard baseline
plus Sol 1350 (AVMA, 16 single images, L2)
22
Von Braun 3D north-facing slopes generated from
hard baseline and wide baseline ground images
Application in MER Operation for Finding Winter
Heaven for Spirit Rover
23
Application in MER Operation for Finding Winter
Heaven for Spirit Rover
North-facing Slope Map (1.4 m window)
24
Topographic Mapping at Duck Bay (Opportunity
Site)
Distribution of 3D points from ground images

• Mapping area
• 83.54 m X 117.31m
• Measured points
• 29699
• Measured points include
• Hard baseline
• Sol 953 (76EV 26 stereo pairs)
• Hard baseline
• Sol 1204 (85HE 12 stereo pairs)
• Wide baseline (5m)
• Sol 1204 (85HE 12 left images)
• Sol 1210 (85JW 12 left images)

25
3D Surface of Duck Bay
26
Duck Bay Slope Map (1.4m window)
27
Pancam Orthophoto Draped on 3D Surface of Duck
Bay
28
Opportunity Traverse Map (Sol 1511)
29
Rock Detection from Orbital Imagery
AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
30
Rock Detection from Ground Imagery
From stereo-based 3-D points
AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
31
Rock Matching from Orbital and Ground Images
Orbital Ground

• Find candidate rock pairs within 2 m range
• Calculate Affine transformation parameters for
  every combination of three orbital-ground pairs
• Apply Affine transformation to ground rocks
• Constraint 0.9lt Scale lt 1.1, ? lt 10º
• See if transformed position is within 0.3 meters
  from orbital rocks
• Select Max( rocks)gt Min (average distance)

AISR Investigator Workshop, May 5-7, 2008,
Adelphi, MD.
32
Orbital and Ground Rock Matching Result
33
Future Research

• Extraction, modeling and matching of landmarks
  from orbital and ground images
• Software development for integrated bundle
  adjustment of orbital, descent and ground imagery