Download Mining Knowledge from Spatial Data - SwRI-Boulder`s

SOLICITED
NASA
PROPOSAL
APPLICATION
SOLICITED NASA
PROPOSAL
APPLICATION
LEAVE BLANKLEAVE BLANK
IN RESPONSE TO ANNOUNCEMENT NRA2-38169
NUMBER
REVIEW GROUP
DATE RECEIVED
PLEASE FOLLOW INSTRUCTIONS CAREFULLY
1a. COMPLETE TITLE OF PROJECT:
1b. PROPOSAL TOPIC:
2. PRINCIPAL INVESTIGATOR/PROGRAM DIRECTOR (First, middle, and last name; degrees; position)
3. COMPLETE MAILING ADDRESS
Congressional District:
4. TELEPHONE NUMBER (area code, number, extension)
FAX NUMBER:
5.
TAXPAYER IDENTIFICATION NO. (TIN)
6.
NAIS CODE
CAGE CODE
E-MAIL ADDRESS:
7. HAS THIS PROPOSAL (OR SIMILAR REQUEST) BEEN SUBMITTED TO ANY OTHER AGENCY?
No
Yes IF YES, SPECIFY AGENCY AND YEAR SUBMITTED:
8. WILL HUMAN SUBJECTS BE USED IN THE PROPOSED RESEARCH:
9. CO-INVESTIGATORS (First, middle, and last name; degrees)
No
Yes
10. CO-INVESTIGATOR'S ORGANIZATION
11. OTHER PARTICIPATING ORGANIZATIONS (e.g., NASA Field Centers or Other
Institutions):
12. DATES OF ENTIRE
PROPOSED PROJECT PERIOD
From:
Through:
13. FY2004 BUDGET
Total Cost:
NASA
Contribution:
14. FY2005 BUDGET 15. FY2006 BUDGET
Total Cost:
Total Cost:
NASA
NASA
Contribution:
Contribution:
16. TOTAL BUDGET
Total Cost:
NASA
Contribution:
17. APPLICANT ORGANIZATION (Organization Name)
18. TYPE OF ORGANIZATION (U.S. ONLY)
For Profit (General)
For Profit (Small Business)
College or University (General)
For Profit (Woman-owned Business)
College or University (Historically Black or other Minority)
19. ORGANIZATION OFFICIAL TO BE NOTIFIED IF AN AWARD IS
MADE (Name, title, address, and telephone number)
For Profit (Minority-owned Business)
Public, Specify:
Federal
Non Profit
State
Local
20. OFFICIAL SIGNING FOR APPLICANT
ORGANIZATION (Name, title, and telephone number)
21. PRINCIPAL INVESTIGATOR/PROGRAM DIRECTOR ASSURANCE:
I agree to accept responsibility for the technical conduct of the project and to provide the
required progress reports if a grant is awarded as a result of this application. Willful provision
of false information is a criminal offense (U.S. Code, Title 18, Section 1001).
22. CERTIFICATION AND ACCEPTANCE:
By submitting the proposal identified in this Cover Sheet/Proposal Summary in response
to NRA2-38169, the Authorizing Official of the proposing institution (or the individual proposer
if there is no proposing institution): 1) certifies that the statements made in this proposal are
true and complete to the best of his/her knowledge; 2) agrees to accept the obligations to comply with
the sponsoring agency award terms and conditions if an award is made as a result of this proposal; and
3) if the applicant organization is an entity of the United States of America, confirms compliance with all
provisions, rules, and stipulations set forth in the three Certifications contained in this NRA (namely, i)
Certification Regarding Debarment, Suspension, and Other Responsibility Matters -- Primary Covered
Transactions, ii) Certification Regarding Lobbying, and iii) Certification of Compliance with the National
Aeronautics and Space Administration Regulations Pursuant to Nondiscrimination in Federally Assisted
Programs). Willful provision of false information in this proposal and/or its supporting documents, or in
reports required under an ensuing award, is a criminal offense (U.S. Code, Title 18, Section 1001).
SIGNATURE OF PERSON NAMED IN BLOCK 2
(In ink; "Per" signature not acceptable.)
________________________________________________ Date:_____________
SIGNATURE OF PERSON NAMED IN BLOCK 20
(or person named in 2, if there is no proposing institution)
(In ink; "Per" signature not acceptable.)
__________________________________________________________
Date:_______________________________________
[Autonomous Recognition and Discovery …]
Contents
Cover Page ................................................................................................................. i
Table of Contents ...................................................................................................... ii
Abstract ...................................................................................................................... 1
1 NASA Relevance .................................................................................................... 2
1.1 Mission Impact.......................................................................................... 2
1.2 Technology Transition Plan ...................................................................... 3
2 Technical Plan ........................................................................................................ 4
2.1 Objectives ................................................................................................. 4
2.2 Scientific Relevance.................................................................................. 4
2.3 Technical Approach .................................................................................. 5
2.3.1 Overview ....................................................................................
2.3.2 Related Work .............................................................................
2.4 Expected Results ....................................................................................... 13
2.5 Evaluation Methodology ........................................................................... 13
2.6 Synergy with Other Activities ................................................................... 13
2.7 Facilities and Equipment........................................................................... 14
2.8 References ................................................................................................. 15
3 Management Plan .................................................................................................. 16
3.1 Personnel Commitments .......................................................................... 16
3.2 Contribution by the PI and each Co-I ....................................................... 16
4 Cost Plan ................................................................................................................. 18
4.1 Budget Summary ...................................................................................... 18
4.1.1 Year 1 .........................................................................................
4.1.2 Year 2 .........................................................................................
4.1.3 Year 3 .........................................................................................
4.1.4 Total Budget...............................................................................
4.1.5 Budget Details ............................................................................
4.2 Current and Pending Support ................................................................... 25
5 Qualifications ......................................................................................................... 31
iii
Mining Knowledge from Spatial Data
ABSTRACT
We propose to develop new techniques for mining knowledge from spatial and spatio-temporal
what is “spatio-temporal”? datasets. The physical processes that govern the formation/evolution
of landforms on a planetary surface often have localized spatial and temporal support, which
leads to significant correlations and inter-relations between nearby objects. Proper interpretation
of the observations can be achieved only by taking into account the context. For geologic
landforms, this means understanding and reasoning about the surroundings and the relationship
of the feature to nearby geologic units. Although some functionality for accessing and visualizing
spatial data is currently available in Geographic Information Systems (GIS) and Spatial
Databases, the capabilities targeted in this proposal go much deeper with the aim of enabling
scientists and engineers to efficiently harvest knowledge from large spatial datasets. As a testbed
for initial development, we will focus on mining knowledge from the cratering record on
planetary, satellite, and asteroidal surfaces. Craters are a particularly worthy prototype landform
to study because they are ubiquitous throughout the solar system and they form with a welldefined and known initial shape. Subsequent geological processes alter a crater’s morphology so
that understanding the spatial distribution of crater morphologies allows significant conclusions
to be made about the processes that have acted on a surface. The techniques developed in this
pilot study are expected to be broad-ranging and applicable to a variety of spatial data mining
problems including study of other landforms and monitoring of distributed sensor networks.
Teaming technologists with practicing scientists will insure that this research is both innovative
and relevant. In previous joint work, our team has made significant advances in automated
analysis of imagery collected by spacecraft (e.g., development of adaptive recognizers for shield
volcanoes [6] and impact craters [7,11,12,14], an autonomous satellite search capability [8,13],
and the core perception algorithms in the New Millennium Program’s ST6 Autonomous
Sciencecraft Experiment [15]).
1 NASA Relevance
1.1 Mission Impact
Due to limited science analysis funds and the time-consuming nature of analysis of craters and
other geomorphic features in planetary spacecraft imaging, only a very tiny fraction of important
images have been analyzed from past and ongoing missions. NASA’s Planetary Data System
houses an enormous library of images from past missions to the Moon, Mars, asteroids, and outer
planet satellites awaiting analysis. Development of state-of-the-art tools for automated detection
and characterization of craters and other geomorphic features, which could operate either
independently or as an assistant to a human analyst, will vastly extend the science analysis yield
from past and ongoing missions. Among the important imaging datasets of solid surfaces that
could be exploited in this way (for “NASA Science Measurements” are Clementine, Galileo, and
Mars Global Surveyor. In addition, imaging data showing craters, rocks, and other features that
have been recently collected by ongoing missions range from the stunning visible and infrared
imaging data from the THEMIS instrument on Mars Global Surveyor to the much smaller but
unique images of the nucleus of Comet Wild 2 obtained in January 2004 by the Stardust
spacecraft.
Two other ongoing missions, already launched, have begun to collect (or soon will) additional
imaging datasets: Cassini has both narrow- and wide-angle cameras that will image the moons of
Saturn. And the successfully deployed Mars Express Orbiter has a high-resolution camera that
will image the entire globe of Mars to 10 m resolution and selected areas to 2 m. Other NASA
and ESA missions currently in development have ambitious imaging plans and will produce
datasets – rich with geomorphic features – that will dwarf those already obtained. These include
the Deep Impact mission to a comet (with high- and medium-resolution instruments); the Mars
Reconnaissance Orbiter, to be launched in 2005, with several imaging instruments including the
amazing HiRISE instrument capable of resolving 30-60 cm pixels on the Martian surface; the
MESSENGER orbiter mission to heavily cratered Mercury; the New Horizons mission to Pluto,
the Rosetta mission to a comet (with its Osirus imager); and the DAWN mission to large mainbelt asteroids. Farther in the future, there are preliminary plans for planetary missions with
orbital imaging systems, including the Jupiter Icy Moons Orbiter, Mars orbiters in the timeframe
beyond 2009, the Phoenix Mars lander with its MARDI descent imager, comet and lunar sample
returns, and a Europa geophysical explorer. Still other missions, mainly to Mars, propose
extensive landed imaging experiments that may be assisted – either in design or in subsequent
data analysis – by the techniques we propose to develop.
It must be remembered that impact craters on the surfaces of solar system bodies are ubiquitous
and fundamental to planetary science. Since impact craters are the basis for some of the most
fundamental analyses in planetary geology (including determination of stratigraphic age
relations), there will always be a need to analyze primary crater populations in images from
future missions. Beyond primary impact craters, numerous other kinds of geological features
exist in great abundance and their spatial distributions hold important clues to the fundamental
processes that shape planetary surfaces. These include large scale tectonic features (e.g. faults
and graben), depositional features (e.g. flows, mass wasting deposits, dunes, lake beds),
constructional features (e.g. volcanic domes), secondary impact features (e.g. secondary craters
and rays), and erosional features (e.g. gullies and channels). At high resolutions, many planetary
1
surfaces (e.g. Eros, Mars) are seen to have innumerable rocks and boulders. Past (and, of course,
future) imaging data sets are replete with never-analyzed data concerning all of these kinds of
features. Thus a reliable, automated approach to unbiased recognition and classification of such
features, in order to determine their global distributions, would be an invaluable asset to
planetary researchers of the future.
1.2 Technology Transition Plan
By teaming computer scientists with domain experts having significant mission experience, we
expect to produce solutions that cut directly to the critical issues and provide scientifically
meaningful results. Our team has an excellent track record for bringing information technology
research to higher technology readiness levels (TRLs) and into actual deployment. Two highprofile examples include: (i) an autonomous satellite-search capability that we developed under
NASA CETDP funding, which was used to aid in the search for satellites of Eros in approach
mosaics that were down-linked from NEAR as it entered orbit around asteroid Eros [?] and (ii)
the core perception algorithms in the New Millennium Program’s ST6 Autonomous ScienceCraft
Experiment [?]. From an institutional standpoint, CU Boulder, JPL, and SwRI are all heavily
involved in actual spacecraft missions and in the analysis of data from those missions. This
culture provides both a large body of expertise, as well as a path of lower resistance for migrating
successful new technology into deployment.
Once the feasibility of automatically mining knowledge from spatial data has been established in
the IS research program, we will pursue follow-on funding from higher TRL programs in order to
make the algorithms broadly available to the scientific community (“tool building”). The recently
restored Applied Information Systems (AIS) program, under which several of our team members
previously led projects, is a good potential program to pick up this activity.
2
2 TECHNICAL PLAN
2.1 Objectives
The goal of our proposed work is to develop automated techniques to assist in the laborious task
of intpreting spatial data. Our primary focus will be on craters since they are a ubiquitous
landform whose spatial distribution conveys valuable scientific information. The techniques
developed in this pilot study will be readily applicable to other landforms --- water and wind
erosional or depositional features (e.g., river valleys or sand dunes), tectonic features (e.g., faults,
graben), and volcanic features (e.g. flow boundaries, rilles, collapse features) and potentially to
distributed senso networks .
2.2 Scientific Relevance
Close interaction between computer scientists and planetary scientists will help insure that the
research conducted is both innovative and beneficial to the scientific community. The Southwest
Research Institute (SwRI) team will provide scientific guidance to the project based on their
expertise in the analysis of planetary features.
The ever-growing volume of image data being returned by NASA spacecraft makes necessary the
development of new, automated techniques for data visualization, reduction, and analysis. The
need is not only to keep up with the data volume, but also to allow maximal extraction of useful
information from the data, acquired at high cost to the public and to competing science programs.
There are many useful data sets now sitting in archives that have not been (or cannot be, due to
lack of automated assistance techniques) fully or even partially analyzed and interpreted with
regard to the geologic understanding. And yet, even now, image data from Mars Odyssey and
Mars Global Surveyor continues to pour in. Future NASA missions, like Mars Reconnaissance
Orbiter (expected to return more than 35 Terabytes of data), will not realize their full scientific
potential if we do not develop intelligent methods for data understanding.
Craters are a particularly worthy landform for prototyping spatial data mining techniques.
Craters are ubiquitous, forming randomly over all parts of exposed surfaces in the solar system.
Even Europa, which has the reputation of being nearly devoid of craters, in fact has many --- for
his PhD thesis, which we directed, Beau Bierhaus, detected and measured 25,000+ craters over
the small fraction of Europa's surface [1] that was imaged by the Galileo spacecraft. Craters
form with a well-defined initial shape in an explosion process in which the main relevant
variable is the energy of the impactor. Changes from the initial shape over millions of years help
scientists understand the processes that have acted on the body. Adjacent geological features that
may affect interpretations of crater morphologies include sand dunes (affecting a crater’s
morphology by sand deposition), local evidence of sapping (collapse due to removal, e.g., by
melting of subsurface ice, possibly causing pits that could be misinterpreted as craters if the
image resolution is insufficient), clusters of small craters near a large crater (suggesting they are
secondary craters formed from the ejecta of the large impactor), the nearby presence of river
valleys or channels (increasing the likelihood that water erosion has acted on the crater), or the
spatial relationship to tectonic features that may shift or alter the crater morphology. Studies of
3
the nature of endogenic processes on planetary surfaces (e.g., volcanic flooding, filling by dust,
tectonic processes) that degrade topography, such as the controversial process(es) that erased
most of the craters on Venus that are older than 500 million years, can be made from statistics of
craters of different morphological classes. Knowledge of planetary stratigraphy and the
interplanetary correlation of geologic time, can also be gained from these cratering studies.
Some of the major insights, as well as major continuing controversies concerning planetary
evolution have been based on cratering studies. The central problem in understanding the nature
of our sister planet, Venus --- whether or not it is subjected to catastrophic resurfacing --- was
first recognized from cratering studies. There continues to be disagreement over the ages of
some of the surfaces of the Galilean satellites (largest 4 moons of Jupiter). In particular,
estimated ages for the surface of Europa have ranged from a mere several million years to several
hundred million years, a significant fraction of the entire time span of the solar system. While
the controversy is not likely to be solved solely by more efficient and reliable crater
morphological measurement and analysis of spatial interrelations with other landforms, age
estimates will depend heavily on these aspects of the data. The implications of these ages can be
profound. If the age of Europa's surface is only a few tens of millions of years old, then it can be
considered currently active. Whether the surface is active or inactive has a bearing on whether
there is likely to be a liquid-water ocean beneath the crust and hence whether there is a potential
for life. This has further implications for the proposed NASA mission to Jupiter's moons, the
JIMO mission (Jupiter Icy Moons Orbiter), and the potential experiments and instrument design.
The spatial context of surrounding terrains and the morphological variations of craters on a
regional or global scale help shape our understanding of the data and the geology. Taken out of
context, the image of a single crater (or any other geologic feature) by itself has limited scientific
value. In short, geologic data simply must be interpreted in the context of its surroundings and its
relationship to other geologic units. To date, most geologic interpretation of spatial processes has
involved human-intensive scrutiny of whatever set of images could be analyzed in the available
time. New images, taken at higher resolution and numbering in the hundreds of thousands now
make this a daunting, if not impossible, task without machine assistance. The goal of our
proposed work is to develop automated techniques to assist in this laborious task and potentially
call the scientist’s attention to trends or relationships that are not immediately evident. A
scientist, who can bring years of experience and conceptual understanding to bear on interpreting
the data, must still do the final detailed geologic interpretation.
While our focus in this study will be on craters as a prototype landform, the process of detecting
features, measuring morphology, and making spatial associations can be (and necessarily will be
in some cases) extended to other landforms --- water and wind erosional or depositional features
(e.g., river valleys or sand dunes), tectonic features (e.g., faults, graben), and volcanic features
(e.g. flow boundaries, rilles, collapse features). In fact, many of these features will have to be
recognized and interpreted automatically to aid the scientist.
2.3 Technical Approach
Spatial datasets consist of a set of spatial locations and associated measurements (attributes). (In
4
the case of spatio-temporal data, the “location” component also includes a time index, but in this
proposal we will focus exclusively on spatial data.). Image data can be viewed as a specialized
type of spatial data in which the pixel coordinates provide the location component. The attributes
in a spatial dataset may be very low-level (e.g., the brightness value of a pixel, measurements of
temperature and pressure) or high-level (e.g., the name of an object and a list of its properties).
There are a number of interesting spatial data mining problems and applications; for surveys and
background information please refer to [5,6].
As a spatial data mining problem, the study of crater morphology from image data is especially
challenging, but, as outlined above in the science section, it is potentially a very rewarding task to
automate. The basic steps involved include: (1) detecting and sizing the craters, (2) making
detailed measurements of the crater morphology, (3) examining spatial properties of crater
morphology, and (4) studying the relationships between craters and other features.
2.3.1 Detecting and Sizing Craters in a Region of Interest:
We have already developed algorithms that can automatically detect and measure sizes of craters
in image data using a combination of techniques including continuously scalable template
models, Hough transforms, and specialized annular kernels [2,8]. By combining the detections of
these various methods with high confidence thresholds required for each, we are able to detect
and accurately size the majority of real craters in moderately complex landscapes, while
minimizing false detections. At a minimum, this provides an intelligent assistant, which will
allow a human labeler to quickly edit any mistakes made by the algorithms. Also, since in this
proposal we are mainly interested in the spatial variations in morphology rather than in crater
detection per se, we may be able to utilize the human-generated catalogs of crater locations that
exist for some surfaces,such as the Bierhaus survey of Europa and the Nadine Barlow catalog of
large Martian craters. Such catalogs would enable one to skip ahead to making the detailed
measurements of crater morphology at each cataloged location.
2.3.2 Detailed Measurements of Crater Morphology
Once the craters have been accurately located, the next step is to perform detailed measurements
of their morphology. We must go further than mere categorization and classify the craters
according to the effects of specific physical processes that may be affecting them. For example,
we need to know whether the crater walls are slumped, the walls breached, the crater floor
infilled by lava, or there are sand dunes (indicating the crater is in a region of material deposition
by wind). Possibly the walls are falling apart in place (disintegrating), such as occurs with craters
on Jupiter's moon Callisto. Here, the crater walls, formed from a combination of ice and rock
initially, disintegrate with time as the ice sublimes away on this cold, airless world.
Do we have any examples of these morphology features that could be put into a figure so
reviewer could hopefully see that measuring such features automatically is plausible?.
To measure the crater morphology and assess processes that may be acting on the craters, we will
pursue a multi-pronged approach. For quickly obtaining short-term results, we will try to directly
incorporate as much of the scientist’s background knowledge into the measurement process as
possible by developing specific feature detectors/measurement functions to assess quantities that
5
are known to be of interest, e.g., the rim height, the depth of the floor, the presence of a central
peak, or the integrity of the rim.
For the mid-term, we will investigate techniques by which the desired mappings from pixel
values to attributes can be learned with feedback from the scientist. Note that mappings to both
continuous-valued attributes (rim height) and discrete-valued attributes (central peak present) are
needed. The scientist will manually measure or judge the various attribute values on a number of
craters to provide training data to the learning algorithm. Classification and function
approximation (regression) methods such as feed-forward neural networks will be used to map
from the raw pixel-value image patches containing the feature to the desired attribute value.
For the longer term, we will look at active learning methods and incorporating advice or hints
from the scientist into the learning process to make the most of the scientist’s time. Active
learning methods will ask the scientist to label additional crater examples that are likely to be
most helpful in pinning down the proper mapping. Features that are more complex such as edges
and filter outputs will also be considered as inputs to the learning process.
I’m not sure how this next line was intended to work. Maybe Becky has some idea.
Relevance feedback methods to determine the most useful input features (or weightings over
input features) may enable quick discovery of the desired mapping.
2.3.3 Examining Spatial Properties of the Morphology of Craters and Other Features
An interesting problem here is to use unsupervised clustering of morphological attributes to
determine equivalence classes (“types of craters”) and then create summary statistics that capture
the spatial distribution of different morphologies, e.g., size-frequency distributions, aggregated
over various scale regions. We will also investigate whether there are spatial trends in various
morphological attributes (e.g., we may quantify the way morphology changes with latitude or
altitude, spatial parameters with which aeolian modification processes vary). Many geologic
features are distributed very non-randomly on planetary surfaces. Analysis of their spatial
distributions often yields vital insights about processes. For example, the non-random
distribution of boulders on Eros points to their origin in the impact that formed Shoemaker crater
Another issue will be to determine if there are statistically significant differences between the
populations of craters in two different regions of the surface. Initially, we may use stratigraphic
units as previously defined through traditional geologic mapping; later, we will explore the
possibility for actually defining the boundaries of geologic units from analysis of the relative
constancy of statistical properties of morphologic parameters. We will also be able to look for
localized subpopulations of craters and craters that differ significantly from their neighbors
(spatial clusters/spatial outliers). [To the degree I understand the above section, I think it is not in
good accord with principles of planetary geology. Different crater morphologies are likely to
vary much more with size than with any spatial variable. Thus, to study spatial variability of
morphology, one must first take into account the variability with size. I would address such
issues by forming size-frequency distributions of different morphological classes (what you call
“types of craters,” I guess) and studying how those distributions change at different locations on
the planetary surface. Size-frequency distributions are not “morphologies”, so I don’t understand
6
that usage. In general, geologic processes do not vary over large spatial scales, so a generalized
decrease in craters as sand dunes are approached is not plausible. Instead, geologists habitually
define sharp-boundaried “units” (presumably of the same “age” and within which the same
processes affected the landscape), within which statistical properties of geomorphological
variables tend to be pretty constant. Therefore, a more useful approach to this topic (“examining
spatial properties of crater morphology”) might be an automated determination of unit
boundaries, having the above rather constant properties.]
2.3.4 Studying the Relationship Between Craters and Other Features
This process will explore the relationship between craters and other geologic processes. In
effect, we will try to determine co-occurrence and anti-co-occurrence relationships. Hypotheses
will come from the scientist and the algorithm will search for evidence to support or refute the
hypothesis. For example, the algorithm might note that there is an unusual size-frequency
distribution in region X. The scientist could then postulate that this is caused by volcanic activity
in the area. The algorithm could then collect information to present to the scientist to support or
reject this hypothesis. In its most general form, this type of activity would require detectors for
other types of features, e.g., volcanic vents. [So do we propose to develop such detectors for
other types of features? That would be very ambitious and the funding probably isn’t in our
budget to do that. If we can’t do that, is there an alternative way to pose Activity 4?] .
Secondary craters bear spatial relationships to the primary craters from which the secondarycrater-forming ejecta came. The unexpected poleward-facing preference for Martian gullies
inspired much theoretical analysis to understand how features apparently formed by water
(requiring warmth) seemed to prefer colder locations; later analyses have questioned whether
these features really do prefer the colder places, which is an example of why an unbiased, more
efficient, automated gully detector analyzed in terms of localized slopes having various
directions might have focused scientists’ attention on the correct facts.
If guess the answer to this depends on what type and how many other types of features might be
of interest. Some things like “linear features” I expect would not be so hard to detect at some
level of reliability. Certain kinds of textured regions might not be so hard. If we’re talking about
more semantic-level objects that have names, say relationship between craters and dunes, then
we’re starting to get into a lot more work to do automatically.
2.3.5 Perceptual Grouping for Object Segmentation and Subsequent Spatial Data Mining:
Making spatial associations directly from image data without first passing to object labels could
greatly simplify the process of exploring co-occurrences and broaden the applicability of the
methods. A bottom-up grouping process may enable relevant objects to be segmented from the
background. For example, in an image with many craters, connected regions of pixels with
similar brightness values might be grouped into higher-level tokens. The consistent (or common)
organization of these tokens in adjacency relationships with related sizes provides a statistical
regularity that could be recognized and used to group parts into a larger token. Similarly, a
connected linear feature consisting of bright features (e.g., from a rille) might be grouped into a
single object. This process potentially gives a varied set of objects without having to explicitly
build a detector for each one. The spatial data mining algorithms could then explore possible cooccurrence or other spatial relations between these objects. For example, on Earth one might
7
find a relationship between unhealthy trees and roads by observing that dark spots (dead trees)
occur more frequently in green textured areas near black linear features (highways) than in green
textures areas that do not have nearby black linear features. [I don’t really understand this
paragraph. I understand the dead-tree/highway example, but what would be a plausible planetary
example? What, exactly, are we proposing to DO here?]
I am not sure about planetary example. Would investigating the relationship between
craters and the linear features on Europa be something of interest? Would a dearth of
craters in close proximity to the linear features be an interesting discovery? Would a
dearth of craters in close proximity to particular linear features suggest that those might
have been more recent fractures?
I guess what I was trying to get at was way to explore spatial relationships between
“objects” without having to determine a semantic label for the object, i.e., without having
to build a detector for all types of objects or a classifier. Instead I am suggesting that we
could label patches of the image with some sort of cluster label (e.g., if we applied the kmeans clustering algorithm to all 8x8 pixel blocks and then assigned each block the id
number from the closest cluster center (in terms of similarity now not spatial distance),
then we could investigate things like spatial association rules, which might say that the
probability a tile of type 64 is to the right of a tile of type 36 is quite high.
Going beyond tiles or blocks, I was suggesting a spatial grouping process to link nearby
structures together into bigger units and these structures could be given some sort of label
based on the building blocks in the structure and their arrangement.
2.4 Evaluation Methodology
How will we objectively measure whether the algorithms are doing something
useful or whether one algorithm is better than another?
2.5 Expected Significance
This proposal directly addresses the IS/IDU goal of ``future capabilities that streamline
investigations by automating tasks that are best performed by machines while freeing scientists
and engineers to focus on the creative process of hypothesis generation and knowledge synthesis.
Geologic interpretation and understanding of image data is largely one of interpreting the spatial
correlations and interrelations of the various types of landforms. Our work here will develop the
techniques necessary for scientists to make effective use of the large existing and forthcoming
spatial data sets. In the process, the scientist is freed to perform the highest-level task of data
understanding and interpretation for which he/she is better suited, while the machine can perform
the data organization, correlations, trend recognition, and first-order synthesis.
2.6 Facilities and Equipment
Need to update this.
CU Boulder: The University of Colorado Intelligent Systems Group has several modern Linux
workstations, each with a fast CPU (Pentium 4 at 2.5Ghz or better), ample RAM (1-2 GB) and
8
disk space (>100GB), and 100Mbps Ethernet connection. Additionally, we have requested funds
for FY04 for 10 1U slave nodes (dual CPU with 2GB of fast RAM and fast disks) to include in a
rack mount compute server. The PI will provide the rack, head node, and network switch from
separate funds. The slave nodes are needed to enable rapid experimentation and evaluation of the
algorithms over sufficiently large datasets.
JPL: The JPL Machine Learning Systems Group currently has at its disposal approximately 15
dual CPU, SUN Ultra60 workstations with 1-2GB RAM each, disk space on the order of 0.2TB,
as well as a number of recently purchased Pentium-based PCs. Standard software development
tools (compilers, debuggers, etc.), database products (ObjectStore, PostgreSQL), 100Mbps
network connections, and prototyping/data visualization software (e.g., Matlab) are readily
available.
In addition, the group has two Beowulf-style clusters. One cluster has 64 Pentium III processors
at 933 MHz with 0.5GB RAM per processor, 0.65TB total disk space, and UPS power to ensure
continuous operation. A second cluster has 32 Pentium III Processors at 1.6GHz with 1GB RAM
per processor, ? total disk space, and UPS power. These systems will enable rapid testing of
algorithms at a variety of parameter settings and over large image collections. As an illustration,
with this new system, the JARtool algorithm developed in earlier work for finding small
volcanoes in the Magellan imagery of Venus could process the entire 30,000 image Magellan
collection in a single day, as compared with several months for a single processor system. The
entire processing power of this cluster is under the direct control of the Machine Learning
Systems Group, i.e., it is a dedicated resource, not time-shared across groups.
SwRI: The offices of SwRI are well equipped with workspace, computer resources, a T1 NSI
connection, and administrative and computer support to meet the goals of the program. We will
have, for our immediate use, two SUN Ultra2 workstations, 3 SUN Ultra1 workstations, a Dell
7500 Inspiron laptop, a pair of 933MHz Linux PCs dedicated to running algorithm tests, and 100
GB of hard-disk storage. New equipment, should it be needed, can be acquired through an
internal capital equipment fund.
2.7 Selected Technical References
[1] Bierhaus, E.B., Merline, W.J., Chapman, C.R., Burl, M.C. "Characterization of Secondary
Craters Using Machine Vision", Proc. 6th International Symposium on Artificial Intelligence,
Robotics, and Automation in Space, 2001 (Montreal), CD-ROM, paper AM114 (2001).
[2] Burl, M.C., Stough, T., Colwell, W., Bierhaus, E.B., Merline, W.J., Chapman, C.R.
"Automated Detection of Craters and Other Geological Features", Proc. 6th International
Symposium on Artificial Intelligence, Robotics, and Automation in Space, 2001 (Montreal), CDROM, paper AM118 (2001).
[3] M.C. Burl, ``Mining Large Image Collections'' , Chapter in Data Mining for Scientific and
Engineering Applications, Eds. R. Grossman, C. Kamath, V. Kumar, and R. Namburu, Kluwer,
(2001).
[4] M.C. Burl, L. Asker, P. Smyth, U. Fayyad, P. Perona, J. Aubele, and L. Crumpler, ``Learning
to Recognize Volcanoes on Venus'', Machine Learning, (April 1998).
[5] K. Koperski,J. Adhikary, J. Han, ``Spatial Data Mining: Progress and Challenges Survey
9
Paper'', In Proc. ACM SIGMOD Workshop on Research Issues on Data Mining and Knowledge
Discovery, Montreal, Canada, (1996).
[6] S. Shekhar and S. Chawla, Spatial Databases: A Tour, Chapter 7, Prentice Hall, (2003).
[7] X. Song, Y. Abu-Mostafa, J. Sill, H. Kasdan, ``Incorporating Contextual Information into
White Blood Cell Image Recognition'', Advances in Nueral Information Processing Systems,
MIT Press, (1997).
[8] Magee, M., Chapman, C.R., Dellenback, S.W., Enke, B., Merline, W.J., Rigney, M.P.
“Automated Identification of Martian Craters using Image Processing”, Proc. Lunar & Planetary
Science Conf. XXXIV, 1756 (2003).
[9] S. Chien, R. Sherwood, M. Burl, R. Knight, G. Rabideau, B. Engelhardt, A. Davies, P.
Zetocha, R. Wainright, P. Klupar, P. Cappelaere, D. Surka, B. Williams, R. Greeley, V. Baker, J.
Doan, "The Techsat-21 Autonomous Sciencecraft Constellation Demonstration", Proc. of iSAIRAS 2001, Montreal, Canada, June 2001.
[10] R. Castano, T. Mann, E. Mjolsness, “Texture Analysis for Mars Rover Images”, in
Applications of Digital Image Processing XXII, Proc. of SPIE Vol. 3808, Denver, CO, pp. 162173, (July 1999).
[11] E. Mjolsness and D. DeCoste, “Machine Learning for Science: State of the Art and Future
Prospects”, Science 293, pp. 2051-2055, 14 (Sept 2001).
10
3
MANAGEMENT PLAN
3.1 Personnel Commitments
FY04
0.16
0.05
0.10
0.05
Prof. Michael Burl, CU Boulder (PI)
Dr. William Merline, SwRI (Co-I)
Dr. Clark Chapman, SwRI (Co-I)
Dr. Rebecca Castano, JPL
(Co-I)
FY05
0.16
0.05
0.10
0.05
FY06
0.16
0.05
0.10
0.05
The time commitment by the PI includes one full summer month and one additional month
during the academic year that is not charged to the project.
In addition to the time commitments by the above named investigators, the work plan includes:
 Two full-time University of Colorado graduate students.
 One Research Analyst at SwRI at average FTE level 0.25 per year.
 Programming support at JPL at average FTE level 0.5 per year.
3.2 Contribution by the PI and each Co-I
Prof. Burl is the PI and will manage all aspects of the project including technical direction,
budgets, and schedules, including timely completion of milestones. He will directly oversee the
quality of deliverables and other technical output (e.g., reports, presentations). Prof. Burl has
long-standing research ties with both the JPL Machine Learning Group and with the SwRI
science team. He will work closely with both groups to insure a concerted effort toward the
project objectives. In addition, he will coordinate the activities of two graduate students, who
will conduct research relevant to the project as part of a program leading to a doctoral degree.
SwRI: FIX! The SwRI team brings an impressive body of expertise in planetary science to the
project, including substantial mission experience. The SwRI team will (1) provide scientific
guidance to the project, (2) establish scientifically meaningful performance objectives, (3)
evaluate ground-based or other proof-of-concept prototypes on relevant datasets to measure
progress against objectives, (4) provide timely feedback to the PI to facilitate continued
algorithmic improvements. Specific roles are aligned with the expertise of the individual
scientists. Dr. Merline will lead and coordinate the SwRI team’s efforts to guide the
development and testing, including oversight of SwRI support personnel (Research Analyst). Dr.
Chapman will serve as the primary adviser for applications involving geological feature
detection, classification, and cataloging.
JPL: The JPL team has significant expertise in computer vision and machine learning techniques
particularly as related to clustering. Dr. Castano will serve as the primary interface between the
PI and the JPL team. The JPL team will focus on measurement of morphological features of
craters from image data.
11
12
4 COST PLAN
4.1 Budget Summary
Year 1 (03/01/04 - 09/30/04)
PROPOSAL TITLE:
Knowledge Discovery from Spatial Data, NRA2-38169
PROGRAM:
Research in Intelligent Systems
($K)
|
A
B
1 Direct Labor (salaries, wages, and fringe benefits)
$
2 Other Direct Costs:
a. Subcontracts
$
b. Services
$
c. Equipment
$
d. Supplies
$
e. Travel
$
f. Other (MPS & ADC)
$
3 Facilities and Administrative Costs
$
4 Other Applicable Costs: (Award Fee)
$
5
$
SUBTOTAL—Estimated Costs
NASA USE ONLY
C
6 Less Proposed Cost Sharing (if any)
7 Total Estimated Costs
$
XXXXXXX
8 Other NASA Center Costs
9 APPROVED BUDGET
$
13
XXXXXXX
|
Year 2 (10/01/04 - 09/30/05)
PROPOSAL TITLE:
Knowledge Discovery from Spatial Data, NRA2-38169
PROGRAM:
Research in Intelligent Systems
($K)
|
A
B
1 Direct Labor (salaries, wages, and fringe benefits)
$
2 Other Direct Costs:
a. Subcontracts
$
b. Services
$
c. Equipment
$
d. Supplies
$
e. Travel
$
f. Other (MPS & ADC)
$
3 Facilities and Administrative Costs
$
4 Other Applicable Costs: (Award Fee)
$
5
$
SUBTOTAL--Estimated Costs
NASA USE ONLY
C
6 Less Proposed Cost Sharing (if any)
7 Total Estimated Costs
$
XXXXXXX
8 Other NASA Center Costs
9 APPROVED BUDGET
$
14
XXXXXXX
|
Year 3 (10/01/05 – 09/30/06)
PROPOSAL TITLE:
Knowledge Discovery from Spatial Data, NRA2-38169
PROGRAM:
Research in Intelligent Systems
($K)
|
A
B
1 Direct Labor (salaries, wages, and fringe benefits)
$
2 Other Direct Costs:
a. Subcontracts
$
b. Services
$
c. Equipment
$
d. Supplies
$
e. Travel
$
f. Other (MPS & ADC)
$
3 Facilities and Administrative Costs
$
4 Other Applicable Costs: (Award Fee)
$
5
$
SUBTOTAL--Estimated Costs
NASA USE ONLY
C
6 Less Proposed Cost Sharing (if any)
7 Total Estimated Costs
$
XXXXXXX
8 Other NASA Center Costs
9 APPROVED BUDGET
$
15
XXXXXXX
|
Total Budget (03/01/03 – 09/30/06)
PROPOSAL TITLE:
Knowledge Discovery from Spatial Data, NRA2-38169
PROGRAM:
Research in Intelligent Systems
|
A
B
1 Direct Labor (salaries, wages, and fringe benefits)
$
2 Other Direct Costs:
a. Subcontracts
$
b. Services
$
c. Equipment
$
d. Supplies
$
e. Travel
$
f. Other (MPS & ADC)
$
3 Facilities and Administrative Costs
$
4 Other Applicable Costs: (Award Fee)
$
5
$
SUBTOTAL--Estimated Costs
6 Less Proposed Cost Sharing (if any)
$
7 Total Estimated Costs
$
8 Other NASA Center Costs
$
9 APPROVED BUDGET
$
16
NASA USE ONLY
C
XXXXXXX
XXXXXXX
|
4.2 Current and Pending Support (PI)
Project
Sponsor
C/P
Budget
($K)
PERIOD
Time
FTE
Visual Recognition and Discovery
for Onboard Science Analysis and
Data Exploration (this proposal)
NASA, IS
(J. Coughlan)
P
500 03/04-09/06
0.16
Multi-Objective Simulation-driven
Discovery
NASA, IS
(J. Coughlan)
P
200 03/04-09/06
0.08
Mining Knowledge from Spatial
Data
NASA, IS
(J. Coughlan)
P
500 03/04-09/06
0.16
Visual Tracking of Multiple People
LLNL
(C. Kamath)
P
60 05/03-10/03
0.08
17
5 QUALIFICATIONS:
Dr. Michael C. Burl (PI)
Department of Computer Science
University of Colorado, Boulder
Boulder, CO 80309
http://www.cs.colorado.edu/~mburl/
Phone: (303) 492-2913
Fax:
(303) 492-2844
[email protected]
Research Interests
Computer vision, machine learning, data mining, and robotics. Specific areas include:
 Detection and recognition of objects in images and other data sources
 Smart sensors and systems for autonomous exploration and data gathering
 Video surveillance and tracking

Mining scientific and engineering datasets
Education
Ph.D.
M.S.
B.S.
Electrical Engineering, Caltech, 1997
(Advisor: Pietro Perona)
Electrical Engineering, Caltech, 1992
Electrical Engineering and Applied Mathematics, Caltech, 1987 (with Honors)
Professional Experience
08/02 – present
07/01 – 08/02
11/96 – 07/01
07/87 – 08/91
University of Colorado at Boulder, Assistant Professor, Dept. of Computer Science
Evolution Robotics, Inc. , Senior Research Scientist
NASA JPL, Technical Group Leader/Sr. Staff, Machine Learning Group
MIT Lincoln Laboratory, Associate Staff, Battlefield Surveillance Group
Selected Honors and Awards




NASA Group Achievement Awards (1996,1999)
National Science Foundation, Graduate Student Fellowship, 1995-1996.
National Merit Scholarship, 1983-1987.
Michigan Mathematics Prize Competition Scholarship, 1982-1983.
Professional Activities




Lead Organizer of a series of SIAM Workshops on Mining Scientific and Engineering Datasets
Program Committees for Computer Vision and Pattern Recognition Conference (CVPR), Workshop
on Data Mining in Resource-Constrained Environments, SIAM Int. Conference on Data Mining, SPIE
Data Mining and Knowledge Discovery (AeroSense99, AeroSense00, AeroSense01), and Knowledge
Discovery in Databases (KDD98).
Reviewer for various federal programs and agencies (NSF, NASA, JPL, LLNL). Also, books, journals,
and leading technical conferences (CVPR, ICCV, ECCV, KDD, SDM, AAAI).
Invited Panelist: NASA Industry Day Forum; SPIE Data Mining and Knowledge Discovery Conf.
Selected Publications and Presentations
P. Wetzler, R. Honda, B. Enke, W.J. Merline, C.R. Chapman, M.C.Burl, “Learning to Detect Small
Craters”, (in review)
18
M.C. Burl and A. Lakshmikumar, “Joint Appearance and Trajectory-based Data Association for Multiobject Tracking” (in review)
M.C. Burl, “Mining Patterns of Activity from Video Data”, SIAM Int. Conf. on Data Mining, (May 2004)
S. Chien, et al, “Autonomous Science on the EO-1 Mission”, iSAIRAS, Nara, Japan (May 2003).
M.C. Burl, T. Vaid, B. Sisk, and N.S. Lewis, “Classification Performance of Carbon Black-Polymer
Composite Vapor Detector Arrays as a Function of Array Size and Detector Composition”, Sensors and
Actuators B, (2002).
M.C. Burl, chapter “Mining Large Image Collections”, In Data Mining for Scientific and Engineering
Applications, R. Grossman, C. Kamath, V. Kumar, and R. Namburu (Eds), Kluwer Academic, (2001).
D. DeCoste, M.C. Burl, A. Hopkins, N.S. Lewis, “Support Vector Machines and Kernel Fisher
Discriminants: A Case Study using Electronic Nose Data”, Fourth Workshop on Mining Scientific Datasets,
(2001).
M.C. Burl, W.J. Merline, W. Colwell, E.B. Bierhaus, C.R. Chapman, “Automated Detection of Craters and
Other Geological Features”, Int. Symposium on AI, Robotics, and Automation for Space (i-SAIRAS), (Jun
2001)
W.J. Merline, W. Colwell, V. Gor, C.R. Chapman, E.B. Bierhaus, M.C. Burl, P. Tamblyn, “Automated
Search for Moons of Eros in NEAR Approach Mosaics”, iSAIRAS, Montreal, Canada (Jun 2001).
D. DeCoste and M.C. Burl, “Achieving Distortion-invariant recognition via Jittered Queries”, Proc.
Computer Vision and Pattern Recognition Conf. (CVPR00), (Jun 2000)
M.C. Burl and D. Lucchetti, “Autonomous Visual Discovery”, SPIE Aerosense Conference: Data Mining
and Knowledge Discovery, Orlando, FL (Apr 2000).
M.C. Burl, C. Fowlkes, J. Roden, A. Stechert, and S. Mukhtar, “Diamond Eye: A Distributed Architecture
for Image Data Mining”, In SPIE Aerosense Conference: Data Mining and Knowledge Discovery, Orlando,
FL, (Apr 1999).
M.C. Burl, L. Asker, P. Smyth, U.M. Fayyad, P. Perona, L. Crumpler, and J. Aubele, “Learning to
Recognize Volcanoes on Venus”, Machine Learning Journal, Vol. 30, No. 2/3, pp. 165-194, (Feb/Mar
1998).
T.K. Leung, M.C. Burl, and P. Perona, “Probabilistic Affine Invariants for Recognition”, Computer Vision
and Pattern Recognition Conf. (CVPR98), Santa Barbara, CA (Jun 1998).
M.C. Burl, M. Weber, and P. Perona, “A Probabilistic Approach to Object Recognition Using Local
Photometry and Global Geometry”, Fifth European Conf. on Computer Vision, Freiburg, Germany, (Jun
1998).
M.C. Burl, “Recognition of Visual Object Classes”, Ph.D. Dissertation, Dept. of Electrical Engineering,
California Institute of Technology, Pasadena, CA (Jun 1997).
M.C. Burl, T.K. Leung, and P. Perona, “Recognition of Planar Object Classes”, Computer Vision and
Pattern Recognition Conf. (CVPR96), San Francisco, CA (1996).
S. Chien, J. Gratch, and M. Burl, “On the efficient allocation of resources for hypothesis evaluation: a
statistical approach”, IEEE Trans. Pattern Analysis and Machine Intelligence, 17(7):652-665, 1995.
M.C. Burl, T.K. Leung, and P. Perona, “Face Localization via Shape Statistics”, In Intl Workshop on
Automatic Face and Gesture Recognition, Zurich, Switzerland, (1995).
L.M. Novak, M.C. Burl, and W.W. Irving, “Optimal polarimetric processing for enhanced target detection”,
IEEE Trans on AES, 29(1):234-244, (Jan 1993).
L.M. Novak and M.C. Burl, “Optimal speckle reduction in polarimetric SAR imagery”, IEEE Trans. On AES,
26 (2):293-305, (Mar 1990).
19
Dr. William Merline (Co-I)
EDUCATION
Ph.D.
B.S.
Planetary Sciences, University of Arizona, 1995
Ph.D. minor: Physics
Dissertation title: Observations of Small-Amplitude Oscillations in the Radial Velocity of Arcturus
Dissertation director: Dr. Robert S. McMillan
Physics and Astronomy, University of Wisconsin - Madison, 1978
Mathematics, University of Wisconsin - Madison, 1978
BRIEF BIOGRAPHY
Dr. Merline is Principal Investigator of programs, funded by NASA and NSF, to perform a ground-based
search for asteroidal satellites. The programs employ the relatively new technology of adaptive optics to
reduce the blurring caused by the Earth’s atmosphere. This work is carried out at the Keck and CFHT
telescopes atop Mauna Kea, Hawaii. Dr. Merline led this international team in the first-ever discovery of
an asteroidal satellite from the Earth (ground-based or HST), that of 45 Eugenia in 1998. In 2000, the
team announced discovery of another moon, that of 762 Pulcova, and the first double asteroid to be
imaged, that of 90 Antiope.
He also has two funded programs under NASA’s Applied Information Systems Program to develop
artificial-intelligence tools for crater detection and analysis, and also to develop web-based science
analysis tools. He is also managing the SwRI-side of a project, with JPL, to produce an on-board software
technology-demonstration, using artificial intelligence, to automate, among other tasks, a search for
asteroidal satellites on future missions. The method is already being applied to images of Eros from
NEAR.
He is an Associate Member of the Imaging & Near-Infrared Spectrometer Team of the NEAR (Near-Earth
Asteroid Rendezvous) spacecraft mission. He had the prime responsibility for the search for satellites of
asteroid Mathilde using NEAR images, and during the 1999 flyby and 2000 orbital tour of asteroid Eros.
Dr. Merline's Ph.D. dissertation was entitled "Observations of Small-Amplitude Oscillations in the Radial
Velocity of Arcturus". During this work, he helped design, build, test, calibrate, and operate a spectrometer
designed for extreme sensitivity to small changes in the radial velocities of stars. The dissertation work
led directly to the discovery of what will likely be a new class of variable stars (K giants), previously thought
to be stable. His work has shown that the oscillations in Arcturus are complex and analogous to solar
acoustic oscillations. He also produced a computational/theoretical study showing methods for optimizing
the amount of information obtained in radial velocity observations. Astronomical interests include the
detection of extra-solar planetary systems, velocity variations in the Sun and stars, the solar/stellar
connection, and optimization of radial velocity determination and of photometric observations using CCD
detectors.
He is associated with the imaging team of the Galileo spacecraft mission. He (with Clark Chapman) had
the prime responsibility for analysis of the Galileo imaging data of the Shoemaker-Levy 9 collision with
Jupiter, the only direct images of the actual impacts. He is currently involved with analysis of the Galileo
images of Jupiter's satellites Europa, Ganymede, and Callisto, attempting to determine surface histories
from the cratering record.
He has played a major role in the analysis of Galileo imaging data of Gaspra, Ida, and Dactyl. He assisted
in the determination of the orbit of the satellite of Ida, Dactyl, and in the associated estimates of Ida's bulk
density. He had the prime responsibility for the search for additional satellites of Ida and also performed a
cursory search for satellites of Gaspra. He was heavily involved in the post-discovery search for Dactyl
using HST images, in support of the orbit-determination effort on Galileo. He is Principal Investigator for a
program to study cratering on the Galilean satellites under NASA’s Jovian System Data Analysis Program.
RECENT POSITIONS
1998-present Senior Research Scientist, Southwest Research Institute, Boulder, CO
1996-1998 Postdoctoral Researcher, Southwest Research Institute, Boulder, CO
1989-1996 Research Associate, Planetary Science Institute, Tucson, AZ
PROFESSIONAL ACTIVITIES, HONORS, & AWARDS
Member: Intl. Astronomical Union (IAU), Amer. Astronomical Society (AAS), Div. for Planetary Sciences (DPS) of
the AAS, Solar Physics Div. of the AAS, Amer. Geophysical Union, Meteoritical Society, Planetary Society
Principal Investigator on 6 NSF/NASA grants, Co-I on 3 NSF/NASA grants
Asteroid 7607 Billmerline named in his honor
NASA Certificate of Recognition, 1999, “Technically Significant Software: Satellite Detector”
NASA Certificate of Recognition, 1998, “Technically Significant Software: Onboard UV Spectral Analyzer”
RECENT PUBLICATIONS
Merline, W.J., Close, L.M., Dumas, C., Shelton, J.C., Menard, F., Chapman, C.R., Slater, D.C. “Discovery of
Companions to Asteroids 762 Pulcova and 90 Antiope by Direct Imaging”, BAAS 32, 1309 (2000).
Veverka, J., Robinson, M., Thomas, P., Murchie, S., Bell III, J.F., Izenberg, N., Chapman, C., Harch, A., Bell,
M., Carcich, B., Cheng, A., Clark, B., Domingue, D., Dunham, D., Farquhar, R., Gaffey, M.J., Hawkins, E.,
Joseph, J., Kirk, R., Li, H., Lucey, P., Malin, M., Martin, P., McFadden, L., Merline, W.J., Millier, J.K.,
Owen, Jr., W.M., Peterson, C., Prockter, L, Warren, J., Wellnitz, D., Williams, B.G., Yeomans, D.K. “NEAR
at Eros: imaging and spectral results”, Science 289, 2088-2097 (2000).
Merline, WJ., Close, L.M., Dumas, C., Chapman, C.R., Roddier, F., Menard, F., Slater, D.C., Duvert, G.,
Shelton,C, Morgan,T “Discovery of a moon orbiting the asteroid 45 Eugenia”, Nature 401, 565.(1999)
Merline, W.J., “Precise Velocity Observations of K-Giants: Evidence for Solar-like Oscillations in Arcturus”,
Proc. of IAU Colloquium 170, Vol. 185, Precise Stellar Radial Velocities (J.B. Hearnshaw & C.D. Scarfe,
eds.), p. 187 (1999).
Merline, W.J., Close, L.M., Dumas, C., Chapman, C.R., Roddier, F., Menard, F., Colwell, W., Slater, D.C.,
Duvert, G., Shelton, C., Morgan, T. “Discovery of a Satellite of (45) Eugenia”, abstract for invited talk,
Asteroids, Comets, Meteors ’99 (1999).
Merline, W.J., Close, L.M., Dumas, C., Chapman, C.R., Roddier, F., Menard, F., Slater, D., Duvert, G., Shelton,
C. Morgan, T., Dunham, D.W. “S/1998(45)1”, IAU Circular 7129 (1999).
Merline, W.J., Chapman, C.R., Colwell, W.B., Veverka, J., Harch, A., Bell, M., Bell, J.F. III, Thomas, P., Clark,
B.E., Martin, P., Murchie, S., Cheng, A., Domingue, D., Izenberg, N., Robinson, M., McFadden, L.,
Wellnitz, D., Malin, M., Owen, W., Miller, J. “Search for Satellites Around Asteroid 433 Eros from NEAR
Flyby Imaging”, Proc. Lunar & Planetary Sci. Conf. 30, 2055 (1999).
Veverka, J., Thomas, P., Harch, A., Clark, B., Bell, J.F., Carcich, B., Joseph, J., Murchie, S., Izenberg, N.,
Chapman, C., Merline, W., Malin, M., McFadden, L., Robinson, M., “NEAR Encounter with Asteroid 253
Mathilde: Overview”, Icarus 140, 3 (1999).
Murchie, S., Robinson, M., Hawkins, S.E., Harch, A., Helfenstein, P., Thomas, P., Peacock, K., Owen, W.,
Heyler, G., Murphy, P., Darlington, E.H., Keeney, A., Gold, R., Clark, B., Izenberg, N., Bell, J.F., Merline,
W., Veverka, J. “Inflight Calibration of the NEAR Multispectral Imager”, Icarus 140, 66 (1999).
Chapman, C.R., Merline, W.J., Thomas, P. “Cratering on Mathilde”, Icarus 140, 28 (1999).
Thomas, P.C., Veverka, J., Bell, J.F. Clark, B.E., Carcich, B., Joseph, J., Robinson, M. McFadden, L.A., Malin,
M.C., Chapman, C.R., Merline, W., Murchie, S. “Mathilde: Size, Shape, and Geology”, Icarus 140, 17
(1999).
Merline, W.J. “Evidence for Solar-Like Oscillations in Arcturus”, Proc. ASP Conference Series, Vol. 135, A
Half-Century of Stellar Pulsation Interpretation, (P. Bradley & J. Guzik, eds.), p 208 (1998).
1
Dr. Clark R. Chapman (Co-I)
Current Position
Previous Position
Education
Institute Scientist, Dept. of Space Studies, Southwest Research Institute, Boulder, CO (1996-)
Senior Scientist, Planetary Science Institute (SJI; SAIC), Tucson, AZ (1971-1996)
A.B., Astronomy, Harvard College (1967)
M.S., Meteorology, Massachusetts Institute of Technology (1968)
Ph.D., Planetary Science, Massachusetts Institute of Technology (1972)
Relevant Experience
Galileo Imaging Science Team; Near Earth Asteroid Rendezvous (NEAR) Mission MSI/NIS Imaging/Spectroscopy
Team, MESSENGER Science Team; telescopic, interpretation, modelling, and theoretical studies of asteroids,
comets, planetary surfaces, early solar system processes; writing and public outreach.
Professional Societies
American Astronomical Society, Division for Planetary Sciences (Chairman 1982-1983; Member DPS Committee
1978-1981; Chairman, Program Committee 1976-1977; Chairman, Nominating Committee 1977; Chairman,
Kuiper/Urey Prize Subcommittee 1983-1984; Sagan Medallist 1999)
American Geophysical Union, Planetology Section (Editor, Journal of Geophysical Research-Planets 1991-1994)
American Association for the Advancement of Science (Chairman, Astronomy Nom. Comm. 1981; Elected Fellow
2000)
Meteoritical Society (Chairman, Leonard Medal Comm. 1983-84, Fellow Elected 1988, Council Member 1992-96)
International Astronomical Union Commission 15 (President 1982-1985; Member Org. Comm. 1976-1985)
Scientific Committee Work and Consulting
Member
URSA Lunar and Planetary Science Council of the Lunar and Planetary Institute
Member
Lunar Science Review Panel
Member
NASA Terrestrial Bodies Science Working Group
Member
Committee on Planetary Exploration (COMPLEX)
Space Science Board, National Academy of Sciences
Member
Management Operations Working Group for NASA Infrared Telescope Facility
Member
Small Bodies Working Group, NASA Solar System Exploration Committee (SSEC)
Member
Primitive Bodies Study Team, NAS/ESF Joint Working Group on
Cooperation in Planetary Exploration
Member
NASA Solar System Exploration Management Council
Head
Education Subcommittee
Member
Comet Rendezvous Asteroid Flyby Science Working Group
Member
Near Earth Asteroid Rendezvous Study Group
Member/Chairman
MOWGroup for NASA Planetary Astronomy Program
Member
NASA Planetary Astronomy Committee
Member
M.I.T. Corp. Visiting Committee to Dept. of Earth, Atmospheric, and Planetary Sciences
Member
Organizing Committee, Asteroids II meeting
Chairman
Organizing Committee, International Conference on Near-Earth Asteroids
Member
NASA International Near-Earth-Object Detection Workshop
Consultant
NASA Near-Earth Objects Survey Working Group
Member
Organizing Committee, IAU Colloq. Mercury and the Moon
Member
Workshop on Prediction in the Earth Sciences: Use and Misuse in
Policy Making (Geol. Soc. Am. Inst. for Environmental Education)
Member
Task Group on Sample Return from Small Solar System Bodies, National Research
Council
1997-1998
Member/Assoc. Editor
Editorial Board, Icarus
2
Relevant Publications (selected from hundreds)
 J. Veverka, M. Robinson, P. Thomas, S. Murchie, J.F. Bell III, N. Izenberg, C. Chapman, A. Harch, M. Bell, B.
Carcich, A. Cheng, B. Clark, D. Domingue, D. Dunham, R. Farquhar, M.J. Gaffey, E. Hawkins, J. Joseph, R. Kirk,
H. Li, P. Lucey, M. Malin, P. Martin, L. McFadden, W.J. Merline, J.K. Miller, W. M. Owen Jr., C. Peterson, L.
Prockter, J. Warren, D. Wellnitz, B.G. Williams, D.K. Yeomans 2000. NEAR at Eros: imaging and spectral results.
Science 289, 2088-2097.
 C.R. Chapman 2000. The asteroid/comet impact hazard:homo sapiens as dinosaur? In Prediction: Science,
Decision Making, and the Future of Nature (ed. D. Sarewitz, R.A. Pielke, Jr. & R. Byerly; Island Press, Washington
D.C.), pp. 107-134.
 W.J. Merline, L.M. Close, C. Dumas, C.R. Chapman, F. Roddier, F. Menard, D.C. Slater, G. Duvert, C. Shelton &
T. Morgan 1999. Discovery of a moon orbiting the asteroid 45 Eugenia. Nature 401, 565-568.
 L. Orgel, M. A'Hearn, J. Bada, J. Baross, C. Chapman, M. Drake, J. Kerridge, M. Race, M. Sogin, & S. Squyres
2000. Sample return from small solar system bodies. Adv. Space Res.25 (2), 239-248.
 J.M. Moore, E. Asphaug, D. Morrison, J.R. Spencer, C.R. Chapman, B. Bierhaus, R.J. Sullivan, F.C. Chuang, J.E.
Klemaszewski, R. Greeley, K.C. Bender, P.E. Geissler, P. Helfenstein & C.B. Pilcher 1999. Mass movement and
landform degradation on the icy Galilean satellites: Results of the Galileo nominal mission. Icarus 140, 294-312.
 R.T. Pappalardo, M.J.S. Belton, H.H. Breneman, M.H. Carr, C.R. Chapman, G.C. Collins, T. Denk, S. Fagents, P.E.
Geissler, B. Giese, R. Greeley, R. Greenberg, J.W. Head, P. Helfenstein, G. Hoppa, S.D. Kadel, K.P. Klaasen, J.E.
Klemaszewski, J. Magee, A.S. McEwen, J.M. Moore, W.B. Moore, G. Neukum, C.B. Phillips, L.M. Prockter, G.
Schubert, D.A. Senske, R.J. Sullivan, B.R. Tufts, E.P. Turtle, R. Wagner & K.K. Williams 1999. Does Europa have
a subsurface ocean? Evaluation of the geological evidence. J. Geophys. Res.--Planets 104, 24015-24055.
 J. Veverka, P.C. Thomas, J.F. Bell III, M. Bell, B. Carcich, B. Clark, A. Harch, J. Joseph, P. Martin, M. Robinson,
S. Murchie, N. Izenberg, E. Hawkins, J. Warren, R. Farquhar, A. Cheng, D. Dunham, C. Chapman, W.J. Merline, L.
McFadden, D. Wellnitz, M. Malin, W.M. Owen Jr., J.K. Miller, B.G. Williams & D.K. Yeomans 1999. Imaging of
asteroid 433 Eros during NEAR's flyby reconnaissance. Science 285 562- 564.
 C.R. Chapman, W.J. Merline & P. Thomas 1999. Cratering on Mathilde. Icarus 140, 28-33.
 M.H. Carr, M.J.S. Belton, C.R. Chapman, M.E. Davies, P. Geissler, R. Greenberg, A.S. McEwen, B.R. Tufts, R.
Greeley, R. Sullivan, J.W. Head, R.T. Pappalardo, K.P. Klaasen, T.V. Johnson, J. Kaufman, D. Senske, J. Moore, G.
Neukum, G. Schubert, J.A. Burns, P. Thomas, J. Veverka 1998. Evidence for a subsurface ocean on Europa. Nature
391, 363.
 J. Veverka, P. Thomas, A. Harch, B. Clark, J. Bell, B. Carcich, J. Joseph, C. Chapman, W. Merline, M. Robinson,
M. Malin, L.A. McFadden, S. Murchie, S. E. Hawkins III, R. Farquhar, N. Izenberg, A. Cheng 1997. NEAR's flyby
of 253 Mathilde: Images of a C asteroid. Science 278, 2109-2114.
 C.R. Chapman 1997. Gaspra and Ida: implications of spacecraft reconnaissance for NEO issues. In Near-Earth
Objects: The United Nations Conference (ed. J.L. Remo, Annals NY Acad. Sci. 822) 227- 235.
 C.R. Chapman 1996. The risk to civilization from extraterrestrial objects and implications of the Shoemaker-Levy 9
comet crash. Abhandlungen der Geologischen Bundeanstalt (Wien) 53, 51-54.
 C.R. Chapman 1996. S-type asteroids, ordinary chondrites, and space weathering: the evidence from Galileo's Flybys of Gaspra and Ida. (Invited review). Meteoritics & Planet. Sci., 31, 699-726.
 C.R. Chapman 1996. Galileo observations of the impacts. The Collision of Comet Shoemaker-Levy 9 and Jupiter
(eds. K.S. Noll, H.A. Weaver, P.D. Feldman; Cambridge Univ. Press), 121-132.
 C.R. Chapman, E.V. Ryan, W.J. Merline, G. Neukum, R. Wagner, P.C. Thomas, J. Veverka, R.J. Sullivan 1996.
Cratering on Ida. Icarus 120, 77-86.
 C.R. Chapman, J. Veverka, M.J.S. Belton, G. Neukum, and D. Morrison 1996. Cratering on Gaspra. Icarus 120,
231-245.
 C.R. Chapman, J. Veverka, P.C. Thomas, K. Klaasen, M.J.S. Belton, A. Harch, A. McEwen, T.V. Johnson, P.
Helfenstein, M.E. Davies, W.J. Merline, T. Denk 1995. Discovery and physical properties of Dactyl, a satellite of
asteroid 243 Ida. Nature 374, 783-785.
 C.R. Chapman and D.P. Cruikshank 1995. Prelude to exploration and the Voyager mission to Neptune. In Neptune
and Triton (Ed. D.P. Cruikshank, Univ. of Arizona Press, Tucson), 3-14.
 C.R. Chapman, W.J. Merline, K. Klaasen, T.V. Johnson, C. Heffernan, M.J.S. Belton, A.P. Ingersoll, and the Galileo
Imaging Team 1995. Preliminary results of Galileo direct imaging of S-L 9 impacts. Geophys. Res. Lett. 22, 15611564.
3
Dr. Rebecca Castano (Co-I)
Jet Propulsion Laboratory
4800 Oak Grove Drive, MS 126-347
Pasadena, CA 91109
Tel: (818) 393-5344
Fax: (818) 393-5244
Email: [email protected]
Education
Ph.D., Electrical and Computer Engineering., University of Illinois, Urbana, IL, 1998.
Dissertation: Yield Estimation for Multichip Module Ceramic Substrates using Visual Inspection
M.S., Electrical and Computer Engineering, University of Illinois, Urbana, IL
Thesis: A Probabilistic Framework for Grouping Image Features
B.S., (high honors) Electrical and Computer Engineering, University of Iowa, Iowa City, IA
Professional Experience
2002-present
1998 -2002
1995 -1995
1994 - 1998
1991 – 1992
1991 – 1991
1989 - 1990
Supervisor, Machine Learning Systems Group – Jet Propulsion Laboratory
Electrical Engineer - Senior, Jet Propulsion Laboratory, Caltech
Summer Technical, Oak Ridge National Laboratory, Oak Ridge, TN
Research Assistant, Beckman Institute, University of Illinois, Urbana, IL
Teaching Assistant, Dept. of Elec. & Comp. Eng., U. of Illinois, Urbana, IL
Member Technical Staff, Center for I. T. and the Dept. of Mechanical Engineering,
U. of Wyoming, Laramie, WY
Summer Intern, Center for IT and ME, U. Wyoming, Laramie, WY
Honors and Awards
National Science Foundation Graduate Fellowship, 1991-1994
NCAA Postgraduate Scholarship Recipient, 1991
Professional Activities
Reviewer: TPAMI, IJCV, ICRA, ICPR
Selected Publications
R. Castaño, R. C. Anderson, T. Estlin, D. DeCoste, F. Fisher, D. Gaines, D. Mazzoni, M. Judd, "Rover Traverse
Science for Increased Mission Science Return," Proc. IEEE Aerospace Conf., Big Sky, Montana, March 2003.
M. Gilmore, R. Castano, T. Mann, R. C. Anderson, E. Mjolsness, R. Manduchi, and R. S. Saunders, “Strategies for
autonomous rovers at Mars”, in J. of Geophysical Res., Vol. 105, No. E12, Dec. 2000 pp. 29223-29237.
J. Fox, R. Castano, and R. C. Anderson, “Onboard autonomous rock shape analysis for Mars rovers,” IEEE
Aerospace Conference, Big Sky, Montana, Mar. 2002.
R. Castano, R.C. Anderson, J. Fox, J.M. Dohm, A.F.C. Haldemann, and W. Fink, “Automating Shape Analysis of
Rocks on Mars,” Lunar and Planetary Science Conference, March 2002.
R. Castano, R. Manduchi, and J. Fox, “"Classification experiments on real world texture", Third Workshop on
Empirical Evaluation Methods in Computer Vision, Dec. 2001.
V. Gor, R. Castano, R. Manduchi, R. C. Anderson and E. Mjolsness, “Autonomous rock detection for Mars terrain,”
American Institute of Aeronautics and Astronautics, Aug. 2001.
R. C. Anderson, R. Castano, E. Mjolsness, A. Davies, J. Fox, T. Stough, and M. Gilmore, “Autonomous Rock
Identification Using Visual Texture”, GSA Abstracts with Program Vol. 32, No. 7, Nov 2000.
R. L. Castano, T. Mann, and E. Mjolsness, “Texture analysis for Mars rover images,” Proc. Applications of Digital
Image Processing XXII, SPIE Vol. 3808, July, 1999.
M. S. Gilmore, R. Castano, B. Ebel, E. Guiness, T. Mann, E. Mjolsness, T. Roush and R. S. Saunders,
“Spectroscopic measurements at Silver Lake, CA, testbed for the FIDO rover, “ AGU, Dec. 1998.
4