Edward A. Guinness
Department of Earth and Planetary Sciences
McDonnell Center for the Space Sciences
Washington University
St. Louis, Missouri 63130
This document contains a description of the geometric
properties of the Viking Lander cameras. Figures associated with
this document are located in the GEOM directory and are stored in
GIF format. The topics covered in this document include: A)
Lander locations; B) Lander orientations; C) Coordinate systems
and camera locations; D) Camera coordinate system
transformations; and E) Stereo-ranging measurements.
A complete description of the Viking Lander camera system is
found in the VOLINFO.TXT
file located in the DOCUMENT directory.
The Viking Lander cameras are facsimile scanning systems with
each lander having two identical cameras. Looking from a lander
toward the front, the camera to the left hand side is camera 1
and the camera to the right hand side is camera 2
(Figure 1).
Each camera has a mirror that rotates about a horizontal axis by
about 100 degrees in the elevation direction. Light enters the
camera through 1 or 2 windows and is reflected off the mirror and
onto the photosensor array. The entire camera assembly also
rotates about a vertical axis to scan nearly 360 degrees in the
azimuth direction. In acquiring an image, the camera first moves
to the commanded starting azimuth position. The mirror rotates
in 512 steps in elevation to scan a vertical line of the image.
In-between each scan line, the camera moves a step in azimuth to
build up the image. The process repeats until the commanded stop
azimuth is reached. A detailed discussion of how the actual
cameras are different from this simple, idealized system is found
in Wolf [1981]. Such differences include, for example, the
elevation rotation axis being behind the mirror and offset from
the azimuth rotation axis; diffraction of light rays passing
through the windows; and photodiodes being offset from the
optical axis.
The locations of the two Viking landers on Mars are given in
the table below. Values are given in planetographic coordinates
[Kieffer et al., 1992; Michael, 1979].
The orientations of the landers with respect to the local
gravity vector and the direction of north are needed for
planning, the preparation of topographic maps, and establishing
viewing conditions in the scene at various times of the sol (Mars
day) and season. Lander direction is defined as the compass
direction of the line perpendicular to the intercamera baseline
(perpendicular to front of lander). This direction can be
thought of as looking straight out in front of the lander. The
lander directions are 141.91 deg east of north for Lander 1
[Tucker, 1978; Mutch et al., 1976a] and 29.13 deg east of north
for Lander 2 [Mutch et al., 1976b]. Each lander is also tilted
relative to a local horizontal surface. Tilt is defined by a
magnitude and direction. Lander 1 is tilted 2.99 deg downward in
the direction of 285.17 deg east of north [Mutch et al., 1976a].
Lander 2 is tilted 8.2 deg downward in the direction of 277.7 deg
east of north [Mutch et al., 1976b]. One effect of the tilt is
that the horizon seen in EDR images is a sine wave pattern, which
is much more noticeable in Lander 2 images because of the larger
tilt magnitude. Figure 1
has schematic diagrams of the two
landers showing the orientations and directions of tilt.
Several coordinate systems are associated with the Viking
Landers and cameras. In this section, four coordinate systems
are described; two are centered on the cameras and two are
centered on the landers. The next section describes how to
convert from a given coordinate system to another. Several
coordinate systems are in use to serve specific functions. For
example, the Camera Aligned Camera Coordinate System (CACCS) is
used for camera commands and the Lander Aligned Coordinate System
(LACS) is used for the locations of lander components. There are
additional lander coordinate systems that are not described here,
such as the Lander Science Coordinate System and the Sampler
Aligned Coordinate System. See Moore et al. [1987] for a
description of these other coordinate systems.
In this document, image coordinates are defined as lines and
samples. The line coordinate is aligned with camera elevation,
whereas the sample coordinate is aligned with camera azimuth.
Lines start with a value of 1 at the top of the image and
increase toward the bottom. Viking Lander EDR images always have
512 lines. Samples begin with a value of 1 at the left side of
the image and increase toward the right. Because the Viking
Lander cameras could be commanded to scan from a minimum of 2.5
degrees to a maximum of about 340 degrees in azimuth, the number
of samples can vary from one Viking Lander image to another.
The coordinate system used to command the start and stop
azimuths and center elevation of images was the Camera Aligned
Camera Coordinate System (CACCS). Values of elevation and
azimuth in the PDS labels and index files of this archive are
also given in the CACCS. This coordinate system is a spherical
system that is unique to each camera [Tucker, 1978].
The origin of the CACCS is the nominal intersection of the
elevation and azimuth axes of a given camera
(Figure 2). The
elevation reference direction (0 deg elevation) in the CACCS is a
plane containing the CACCS origin and perpendicular to the camera
azimuth rotation axis. Positive values of elevation generally
point toward the sky, whereas negative values generally point
toward the ground. The azimuth reference direction (0 deg
azimuth) is unique for each camera. For camera 1, the azimuth
reference direction points toward camera 2 and is nominally at
9.5 degrees clockwise from the intercamera baseline when viewed
from above. The intercamera baseline is the line connecting the
origins of the CACCS of each camera. For camera 2, the azimuth
reference direction is generally toward camera 1 and is nominally
at 5.5 degrees clockwise from the intercamera baseline. In other
words, the azimuth reference for camera 1 points slightly behind
camera 2, whereas for camera 2 it points slightly in front of
camera 1 (Figure 2).
Differences from these nominal values are
known as bolt-down errors, which are listed in the next section.
Azimuths for both cameras increase in a clockwise direction.
A second spherical coordinate system centered on each camera
is the Lander Aligned Camera Coordinate System (LACCS). This
system is used on some mission produced photoproducts and
catalogs. The LACCS is similar to the CACCS in that both have
the same origin and elevation definition. The difference in the
two systems is in the azimuth reference direction. In the LACCS
the azimuth reference direction for both cameras is perpendicular
to the intercamera baseline and points toward the back of the
lander. Azimuths increase in a clockwise direction when viewed
from above (Figure 2) [Tucker, 1978].
The Lander Aligned Coordinate System (LACS) was a coordinate
system used for some engineering aspects of the landers. The
LACS is a Cartesian system, defined relative to the lander
(Figure 3).
However, the origin of the LACS is outside the
lander and below the surface [Liebes, 1982]. The Z-Y plane of
the LACS is parallel to the upper deck of the lander, but lies
about 1.1 m below the upper deck (Figure 3).
Note that both
landers are tilted, so that the Y-Z plane is not necessarily
parallel to the ground surface. The Z-X plane is perpendicular
to the upper deck of the lander body and passes through the
center of the upper deck and midway between the two cameras. The
Y-X plane passes through the center of the lander upper deck and
is perpendicular to the Z-Y and Z-X planes. The Z-axis is formed
by the intersection of the Z-Y and Z-X planes and is positive
toward the front of the lander. The X-axis is formed by the
intersection of the Y-X and Z-X planes and is positive downward.
The Y-axis is formed by the intersection of the Y-X and Z-Y
planes and is positive to the left. Thus, the LACS is a right-
handed coordinate system. The LACS origin is defined relative to
the geometric center of the lander deck with the center point of
the lander deck being along the -X axis and about 1.1 m above the
origin [Moore et al., 1987]. As stated earlier, this means that
the origin of the LACS is located below the surface of the
nominal landing plane (Figure 3).
The Local Mars System (LMS) is a Cartesian system that takes
Lander orientation and tilt into account. The origin of the LMS
is the same as the LACS. The +Z axis in the LMS is parallel with
the local zenith vector (i.e., parallel to the local gravity
normal vector). The +Y axis is pointed toward the north and the
+X axis is pointed toward the east [Liebes, 1982].
Camera locations are defined here as the locations of the
CACCS and LACCS origins (i.e., the intersection of the camera
elevation and azimuth rotation axes). The coordinates listed
below are given in LACS coordinates. These values are nominal
values. Wolf [1981] discusses deviations from these nominal
values.
Given these values, the length of the intercamera baseline
is the difference in Y coordinates of the two cameras, or
nominally 0.822 m. The origins of the CACCS and LACCS are 0.487
m above the lander deck and about 1.3 m above a nominal
horizontal landing plane (Figure 3) [Liebes, 1982].
1. Introduction
2. Lander Locations
Lander
Latitude
Longitude
1
22.480 deg N
47.968 deg W
2
47.967 deg N
225.737 deg W
3. Lander Orientations
Lander
Front Direction
Tilt Magnitude
Tilt Direction
1
141.91 deg
2.99 deg
285.17 deg
2
29.13 deg
8.2 deg
277.7 deg
4. Coordinate Systems and Camera Locations
a. Image Coordinates
b. Camera Aligned Camera Coordinate System
c. Lander Aligned Camera Coordinate System
d. Lander Aligned Coordinate System
e. Local Mars System
f. Camera Locations
Camera 1
Camera 2
X
-1.583 m
-1.583 m
Y
0.411 m
-0.411 m
Z
0.472 m
0.472 m
El = Elc + S * (256.5 - ln) + Be
| (1) |
where Elc
is the image center elevation, S
is the image sampling rate (either 0.04 or 0.12 degrees/pixel),
ln
is the image line number of the point, and
Be
is the elevation bolt-down correction. The center
elevation and sampling rate (as sampling_parameter_interval) are found
in the PDS image label and index file of this archive. The bolt-down
term in equation 1 corrects for the effect of the camera not being
mounted in a true vertical position. Values of Be
are
constants unique to each camera. Values of Be
in degrees
are listed in the table below and are from an unpublished Lander
Imaging Team memo.
Lander 1 | Lander 2 | |||
---|---|---|---|---|
camera 1 | -0.18 | -0.08 | ||
camera 2 | -0.07 | -0.17 |
Several Viking Lander images were acquired where the diode and
sampling rate did not match in resolution, e.g., high resolution color
images. In such cases an additional term (D
) is needed
in equation 1 due to the camera electronics:
El = Elc + S * (256.5 -ln) + Be + D
| (2) |
where D
is -5.6 deg if the sampling rate was 0.04 deg
and the diode was one of blue, green, red, IR1, IR2, IR3, survey, or
sun. The value of D
is +5.6 deg if the sampling rate
was 0.12 deg and the diode was one of BB1, BB2, BB3, or BB4. In other
words, the image is shifted down in elevation for high resolution
sampling with a low resolution diode and shifted up in elevation for
low resolution sampling with a high resolution diode.
CACCS azimuth (Az
) is computed from image sample number by:
Az = Azs + S * (sm - 1) + Ba + C
| (3) |
where Azs
is the start azimuth, S
is the
sampling rate, sm
is the image sample number of the
point, Ba
is the azimuth bolt-down correction, and
C
is the coning angle correction. Again start azimuth
and sampling rate are found in the PDS image label or the index file
of this archive. The azimuth bolt-down correction compensates for the
difference in the actual and nominal azimuth reference direction. The
values of Ba
are given below in degrees and are from an
unpublished Lander Imaging Team memo.
Lander 1 | Lander 2 | |||
---|---|---|---|---|
camera 1 | -0.79 | -0.87 | ||
camera 2 | -0.20 | -0.10 |
The coning angle correction is due to the diodes being offset from the optical axis by +/- 0.48 degrees. This effect causes the scan lines to scribe curved lines in the scene instead of straight lines. The magnitude of the coning error is a function of elevation and the sign depends on which side of the photosensor array the diode is on. The coning correction is computed as:
C = +/- arctan (tan(a)/cos(El)) - a
| (4) |
where a
is 0.48 degrees and El
is the
elevation angle. The sign is positive for diodes BB2, BB4, blue,
green, red, and sun; and negative for diodes BB1, BB3, IR1, IR2, IR3,
and survey.
The conversion from CACCS to LACCS is a simple change in the azimuth reference direction (0 deg azimuth). Elevation values in CACCS and LACCS are the same. The conversion for camera 1 is:
Azl = Azc - 80.5
| (5) |
where Azl
is azimuth in LACCS and Azc
is
azimuth in CACCS. Likewise, the conversion for camera 2 is:
Azl = Azc + 95.5
| (6) |
All values of azimuth are in degrees.
Using the definitions of the LACCS (or CACCS) provides only the direction to a point in the scene. To convert to LACS coordinates, the range to the point is needed along with the direction. Both direction and range can only be determined from stereo images (i.e., viewing the same point with two cameras separated by some distance). Methods for determining true direction and range to points are discussed in the next section.
Crude estimates of range to a point can be made from a single
image by assuming the surface is a horizontal plane 1.3 m below the
cameras (Figure 3). Such estimates are useful for
areas of the lander site that can only be viewed by one camera, i.e.,
areas to either side of the landers, like the large boulder named Big
Joe. An estimate of slant range (R
) between the camera
and a point in the scene is computed using a right triangle as:
R = 1.3 / sin(El)
| (7) |
where El
is the camera elevation of the point and
R
is in meters. An estimate of the horizontal distance
(H
) from the camera to the point is:
H = 1.3 / tan(El)
| (8) |
The size of an object (Sz
) can also be approximated from
a single image from its angular size (i.e., difference in azimuth at
constant elevation) as:
Sz = Da * R
| (9) |
where Da
is the difference in azimuth in radians and
R
is the slant range from the camera (from equation 7).
The conversion from LACS to LMS removes lander tilt and aligns the axes with north and east. The transformation involves a matrix rotation that is unique to each lander:
[Vm] = [Rm] * [Vl]
| (10) |
where [Vm]
is a three element vector of LMS x, y, and z
coordinates, [Rm]
is a 3x3 rotation matrix, and
[Vl]
is a three element vector of LACS x, y, and z
coordinates. The elements of [Rm]
for the two Viking
landers are given in the tables below [Liebes, 1982].
0.0503457 | 0.7858010 | 0.6164240 | ||
-0.0136545 | 0.6176890 | -0.7862990 | ||
-0.9986370 | 0.0311701 | 0.0418279 |
0.1414660 | -0.8623340 | 0.4861690 | ||
-0.0191174 | 0.4886360 | 0.8722730 | ||
-0.9897580 | -0.1326930 | 0.0526407 |
Viking Lander stereo imaging techniques can measure the
position of a point by using azimuth and elevation pointing
information from both cameras. Because of the camera separation,
a point in the scene is viewed by each camera with a different
perspective. The direction from each camera to the point defines
two vectors, with the intersection of the vectors giving the
3-dimensional position of the point. Stereo measurements are
largely confined to areas in front of the landers due to limits
in the azimuth range and due to obscuration by lander parts.
Some areas behind the landers can be seen stereoscopically.
During the Viking Mission a custom hardware and software
system existed for stereo analysis mainly to support mission
activities, such as collecting samples [Liebes and Schwartz,
1977]. This system included real-time stereo viewing and
ray-trace algorithms for geometric correction of non-ideal imaging
system effects [Wolf, 1981]. Accurate ranging measurements can
still be made by a less sophisticated method of simply measuring
image coordinates of a point in both camera 1 and 2 images and
computing the LACS coordinates using the equations listed below
[Moore et al., 1987]. This simple method works best with pairs
of images acquired with the same diode and with similar lighting
conditions so that features look similar in both images.
However, the equations given below will give reasonable solutions
for any diode combination. Also, there is evidence that Viking
Lander 2 moved a small amount during the mission, so using images
acquired close in time will eliminate errors due to lander
movement [Moore et al., 1987]. Ranging errors increase rapidly
with distance from the lander, when the azimuth of the two
cameras are similar, and when the azimuths approach the direction
of the intercamera baseline [Liebes, 1982; Liebes and Schwartz,
1977].
Figure 4 shows the geometry of the vectors and
angles used to derive the ranging equations. For convenience, LACCS
azimuths are adjusted by subtracting 90 degrees:
6. Stereo Measurements
Camera 1: A = Az2 - 90
| (11) |
Camera 2: B = Az1 - 90
| (12) |
where Az1
and Az2
are the LACCS azimuths for
camera 1 and 2, respectively. These new angles (A
and
B
) are relative to the +Y axis of the LACS. The
equations used to compute the coordinates of a point in LACS are:
f = I * sin(A) / sin(B-A)
| (13) |
x = -f * tan(E1) - 1.583
| (14) |
y = f * cos(B) + 0.411
| (15) |
z = f * sin(B) + 0.472
| (16) |
where f
is the horizontal distance between camera 1 and
the point, I
is the intercamera baseline length (0.822
m), E1
is the camera 1 elevation angle of the point, and
x, y
, and z
are LACS coordinates in meters.
Figure 1: Diagram showing the lander orientation and tilt
directions, along with the relative locations of the two cameras.
The dashed line between the cameras represents the intercamera
baseline. After Tucker [1978].
Figure 2: Schematic showing the azimuth reference
directions (0 deg. azimuth) for the CACCS and LACCS. After
Tucker [1978].
Figure 3: Schematic showing the origin and orientation of
axes for the LACS. The diagram also shows the position of the
cameras relative to the LACS origin. Diagram is not to scale.
After Liebes [1982].
Figure 4: Schematic showing the vectors and angles used to
derive the equations for stereo ranging to point P. The Y and Z
axes are for the LACS. The side view is a plane parallel to the
LACS X axis and intersects both camera 1 and the point P.
7. Figures