Археологическая разведка Луны: результаты проекта SAAM
|
Archaeological Reconnaissance of the Moon:
Results of SAAM Project
A.V. Arkhipov
(rai@ira.kharkov.ua)
Institute of Radio Astronomy, Nat. Acad. Sci. of Ukraine
(Материалы конференции "SETI-XXI")
Our Moon is a
potential indicator of a possible alien presence near the Earth at
some time during the past 4 billion years. To ascertain the presence
of alien artifacts, a survey for ruin-like formations on the Moon
has been carried out as a precursor to lunar archaeology. Computer
algorithms for semi-automatic, archaeological photo-reconnaissance
are discussed. About 80,000 Clementine lunar orbital images have
been processed, and a number of quasi-rectangular patterns found.
Morphological analysis of these patterns leads to possible
reconstructions of their evolution in terms of erosion. Two
scenarios are considered: 1) the collapse of subsurface
quasi-rectangular systems of caverns, and 2) the erosion of hills
with quasi-rectangular lattices of lineaments. We also note the
presence of embankment-like, quadrangular, hollow hills with
rectangular depressions nearby. Tectonic (geologic) interpretations
of these features are considered. The similarity of these patterns
to terrestrial archaeological sites and proposed lunar base concepts
suggest the need for further study and future in situ exploration.
1. Introduction
The idea of
lunar archaeology was discussed long before space flight. In the
1930s, J.Wyndham (alias J.Beynon) wrote "The Last Lunarians" - a
fictional report about an archaeological mission to the Moon
[1]. In writing
about the discovery of an ancient lunar artifact in the short story,
The Sentinel
, Arthur C. Clarke
said: "There are times when a scientist must not be afraid to make a
fool of himself" [2]. Today, the
idea of exploring the Moon for non-human artifacts is not a popular
one among selenologists. Yet, because we know so little about the
Moon, the investigation of unusual surface features can only add to
our knowledge. When we return to the Moon, it is possible that lunar
archaeological studies may someday follow.
It has been
argued [3], [4] that the Moon
could be used as an indicator of extraterrestrial visits to our
solar system. Unfortunately, the detection of ET artifacts on the
Moon is outside the interest of most selenologists due to their
orientation towards natural formations and processes. It is also not
of interest to mainstream archaeologists, as archaeology tends to
adhere to a pre-Copernican geocentric point-of-view.
In 1992, the
Search for Alien Artifacts on the Moon (SAAM) - the first
privately-organized archaeological reconnaissance of the Moon - was
initiated. The justifications of lunar SETI, the wording of specific
principles of lunar archaeology, and the search for promising areas
on the Moon were the first stage of the project (1992-95).
Preliminary results of lunar exploration [5] show that the
search for alien artifacts on the Moon is a promising SETI-strategy,
especially in the context of lunar colonization plans. The aim of
the second stage of SAAM (1996-2001) was the search for promising
targets of lunar archaeological study. The goals of this second
stage involved 1) developing new algorithms for space archaeological
reconnaissance, 2) using these algorithms to detect possible
archaeological sites on the Moon, and 3) examining the reaction of
mainstream scientists to these results.
2. Methodology
It is
generally accepted that the search for alien artifacts on the Moon
is not necessary because there are none. Circular logic leads to a
deadlock: no finds, hence no searches, hence no finds, etc. Given
the success in using terrestrial remote sensing to find
archaeological sites on Earth, can similar techniques be used to
find possible artificial constructions on the Moon and other
planets? Hardly, if planetologist think only in terms of natural
formations. For example, the ancient Khorezmian fortress
Koy-Krylgan-kala in Uzbekistan, constructed between the 4th century
BC to the first century AD, appeared as an impact crater before
excavation in 1956 (Fig. 1). On the Moon, Koy-Krylgan-kala would not
be perceived among all of the impact craters.
Fig. 1. The ancient Khorezmian fortress Koy-Krylgan-kala appeared as an impact crater on the air photo (left); its artificiality is obvious after the excavations in 1956 (right) [6]. |
Instead of
the current presumption that all surface features are natural, an
alternative search strategy is to be open to the possible existence
of artifacts. If we are open to this possibility, then one can
extend Carl Sagan's search criteria for detecting signs of life on
Earth [7] to other
planets:
"Let us first imagine a photographic
reconnaissance by orbiter spacecraft of the Earth in reflected
visible light. We imagine we are geologically competent but have no
prior knowledge of the habitability of the Earth. Photography of the
Earth at a range of surface resolutions down to 1 km reveals a great
deal that is of geological and meteorological interest, but nothing
whatever of biological interest. At 1 km resolution, even with very
high contrast, there is no sign of life, intelligent or otherwise,
in Washington, London, Paris, Moscow, or Peking. We have examined
many thousands of photographs of the Earth at this resolution with
negative results. However when the resolution is improved to about
100 m, a few hundred photographs of say 10 km x 10 km coverage are
adequate to uncover terrestrial civilization. The patterns revealed
at 100 m resolution are the agricultural and urban reworking of the
Earth's surface in rectangular
arrays... These patterns would be extremely
difficult to understand on geological grounds even on a highly
faulted planet. Such rectangular arrays are clearly not a
thermodynamic or mechanical equilibrium configuration of a planetary
surface. And it is precisely the departure from thermodynamic
equilibrium which draws our attention to such photographs."
In 1962 Sagan
spoke on the possibility of discovering alien artifacts on the Moon
stating that "Forthcoming photographic reconnaissance of the moon
from space vehicles - particularly of the back - might bear these
possibilities in mind." [8]
Rectangular patterns on air-space photos are recognized as signs of human
culture in the remote sensing of the Earth and air archaeology
[9]. It seems
reasonable then to search for rectangular patterns on the Moon. For
example, assume that the equivalent of proposed modern lunar bases
were built long ago (e.g., 1-4 billion years ago) on the Moon. Such
structures would have been built under the surface for protection
from ionizing radiation and meteorites. Today these ancient
structures might appear as eroded systems of low ridges and
depressions, covered by regolith and craters (Fig. 2).
Fig. 2. Simulation of probable HIRES view of ancient settlement on the Moon (left). The erosion wipes off the surface tracks of construction (center), but the SAAM processing could reveal the rectangular anomaly (right). |
A wealth of
lunar imagery collected by the Clementine probe are available in
digital form [10]. Initial
SETI studies [11] used images
from the ultraviolet-visible (UVVIS) camera. The resolution of UVVIS
images is ~200 m. According to Sagan's detection criteria, this
resolution would not be sufficient even to detect the presence of
our own civilization on Earth. Studies of the Moon at this
resolution would probably not reveal any convincing evidence of the
existence of artificial structures. On the other hand, Clementine's
high-resolution (HIRES) camera produced images of adequate
resolution (9-27 m), but they are much more numerous (~ 600,000
images total) and they are thus largely unstudied. The next section
discusses algorithms for automatically scanning large numbers of
HIRES images for potential artifacts.
3. Algorithms
3.1. Preliminary Fractal Test
As a rule,
the structure of natural landscapes is self-similar over a range of
spatial scale. For example, lunar craters between 10-1 m
to 10 4 m in
size appear similar in structure. In contrast to the self-similar
structure of natural features, the structure of artificial objects
is expressed over a narrower range of scale. Hence, possible
artifacts in an image might be recognized as anomalies in the
distribution of spatial detail as a function of scale. The search
for such anomalies is the essence of the fractal method proposed by
M.C. Stein and M.J. Carlotto [12], [13].
Unfortunately their method is too computationally-intensive to
process all of the candidate HIRES images (~80,000).
An
alternative algorithm that is simpler and faster was used for the
same purpose. Let M(r) be the probability distribution of the
distances between local minima in brightness along horizontal lines
in an image. M(r) thus provides a measure of the size distribution
of image detail. At long scales, this function can be approximated
by the fractal power law:
(1) |
As artificial
objects have some typical size, their presence should increase the
squared residuals of linear regression:
(2) |
where C is a
constant. According to empirical results, M(r) of the HIRES images
can be approximated by a power law at r > 4 pixels. The regression
is calculated from 4 < r < 31 pixels
(i.e., over a scale range from 50 to 900m).
Images are
divided into K=12, 96x96 pixel regions. In each region the best
model parameters are calculated by least squares, and the average of
the squared residuals determined:
(3) |
where k is
the number of the test square, gk compensates for gain variations across the sensor, and N is
the number of scales. The average dispersion is estimated from these
regional squared residuals.
An analysis
of 733 HIRES images using the 0.75 micrometer filter, from orbits
112-115 (up to 75 deg. latitude) shows the distribution of residuals
to be Gaussian in form. According to the Student's criterion for
K=12 estimates, if the inequality
(4) |
is true in
any test square, this area could be considered as statistically
anomalous with a probability of 0.95.
3.2. Detailed Fractal Test
A modified
version of Stein's fractal method was used as a more detailed test.
First, the range of HIRES image brightness was increased linearly up
to 256 gradations. Then the image could be considered as an
intensity surface in a 3-D rectangular frame of coordinates (x and y
are the pixel coordinates, and z the brightness). Stein's method can
be thought of as enclosing the image intensity surface in volume
elements. These volume elements are cubes with a side of 2r, where r
is the scale in terms of pixel coordinates or brightness. Let V(r)
be the average minimal volume of such
elements enclosing an image intensity surface at some point. Then
the surface area is A(r) = V(r)/2r.
As a function of scale, A(r) characterizes the size distribution of
image details. The fractal linear relation between log A(r) and log
r is a good approximation for natural landscapes. However, fractals
do not approximate artificial objects as a rule. This is why Stein
used the average of the squared residuals of the linear regression
(5) |
as a measure
of artificiality. Unfortunately, the value of the squared residuals
depends on the number of pixels in an image. Therefore, it is
difficult to compare images with different sizes. Moreover, shadows
increase the residuals and generate false alarms. These problems can
be resolved by the non-linear regression:
(6) |
where the
'artificiality parameter' "alpha" is independent of the
image size.
Fig. 3 plots
alpha of a random set of images representing the natural
lunar background (crosses), and the set of images containing
anomalous objects (squares). The shadows lead to values of
alpha greater than zero, but anomalous objects have values
less than zero. At any Solar zenith angle, Zsun the anomalous
formations have systematically lower alpha than the random
set of HIRES images. The average linear regression relating
alpha of the random set and Zsun is shown as a dashed line where the standard deviation of the
crosses from this regression is 0.0113. A deviation of 3 sigma
(solid line) is adopted as a formal criterion for the final
selection of candidate objects.
Fig. 3. Selection of lunar features based on 'artificiality parameter' alpha |
3.3. Rectangle Test
The rectangle
test reveals rectangular patterns of lineaments on the lunar
surface. For each pixel of the image, a second pixel at a distance
of 6 pixels and a given position angle is selected. Let N be the
total number of pixel pairs, and n be the number of pairs where the
pixel brightnesses are equal. The function
(7) |
characterizes
the anisotropy of the image in terms of position angle. To correct
for camera effects it is normalized by its average over many images.
The anisotropy is smoothed and position angle maxima are found. The
maxima are the orientations of lineament groups. If there are 90
deg. ± 10 deg. differences between maxima, the image is classified
as interesting.
3.4. SAAM Transformation
To aid in
false alarm rejection, the SAAM transformation (Fig. 2) of the image
was used to enhance subtle details of the lunar surface. This
transformation involves smoothing the image over a sliding circular
window of radius R, and subtracting the result from the initial
image. Pixel that are brighter than the smoothed level (difference
greater than zero) are labeled as 'white'; the others are 'black'.
Clipping helps us to see details of both low and high contrast.
Moreover, large details (greater than R in size) are de-emphasized
and so do not interfere with smaller-sized features.
3.5. SCHEME Algorithm
The SCHEME
algorithm searches for local extremities of lunar relief. It does so
by detecting peaks in the image intensity surface in the direction
of the sun. An example of the SCHEME algorithm is shown in Fig. 4.
Fig. 4. The image LHD0331A.062 and a map of relief extremities found by the SCHEME algorithm. |
3.6. Geological Test
J. Fiebag has
suggested that when parallelism exists between a structure and the
lineaments of its surroundings, it is likely to be natural [14]. Although
human activities do sometimes correlate with geological lineaments
(e.g. rivers), the conservative Fiebag test was applied to the lunar
finds.
The lineament orientation of surroundings was estimated by the rectangle test technique applied to the ultraviolet-visible (UVVIS) camera. The UVVIS image covers 196 times the HIRES area wi