DISTRIBUTED COMPUTING IN GROUND PROCESSING
Belay T. Beshah *, John Welter " Kristian Morin ^
* Leica Geosystems GIS & Mapping, LLC, 10840 Thornmint Rd. Suite 100, San Diego CA 92127, USA -
belay.beshah@gis.leica-geosystems.com
? North West Geomatics Inc., Suite 212, 5438 11st NE, Calgary, Alberta,
CANADA T2E 7E9 - (jwelter, kmorin)@nwgeo.com
Commission III, WG III/8
KEY WORDS: Distributed, Automation, Processing, Digital, Sensor, Rectification, Matching, Performance
ABSTRACT:
In the recent years, airborne digital imaging sensors have gained acceptance in the photogrammetric workflow. However, the
processing and management of data acquired by these sensors requires an enormous computational effort, which is often too high for
single processor architectures. This demand for processing power stems from the large amount of data being generated and the high
rate of automation possible in ground processing. Distributed computing, the method of dividing large processing problems into
smaller tasks that can run on individual systems, has emerged as a key enabling technology for digital photogrammetric workflows.
Using networks of workstations in distributed computing to solve large problems has become popular owing to the proliferation of
inexpensive, powerful workstations. Clusters offer a cost-effective alternative to batch processing and an easy entry into parallel
computing. The main advantage is the potential for future performance enhancement that results from the high rate of advances seen
in computer and network hardware, scalability, fault tolerance and rapid development of applications. This paper summarizes a
range of distributed computation technologies surveyed, design criteria used for choosing a solution and the results obtained in the
ground processing workflow of the Leica Airborne Digital Sensor, especially in rectification and automated point matching. We
conclude by presenting the results from real applications indicating timesaving and benefits of the distributed computing model in a
photogrammetric production department.
1. INTRODUCTION
The digital sensor revolution happening right now will result in
an information explosion. The processing and management of
data acquired by digital sensors requires an enormous
computational effort. The Aerial Digital Sensor (ADS40) from
Leica Geosystems GIS & Mapping (LGGM) is already ahead of
the curve in generating tremendous amount of data. Pretty soon
other digital sensors will also be creating large quantity of data
and face the same problems. This demand for large processing
power stems from the large amount of data being generated and
the high rate of automation possible in the ground processing.
Using single computers one could not exploit the full
automation possibility of digital acquisitions.
Leica’s Airborne Digital Sensor is a high performance digital
sensor, capable of delivering photogrammetric accuracy with
multi-spectral remote sensing quality imagery. The ADS40 is a
three-line-scanner that captures imagery looking forwards, nadir
and backwards from the aircraft. Every portion of the ground
surface is imaged multiple times. With its simultaneous capture
of data from three panchromatic and four multi-spectral bands,
the ADS40 provides unparalleled qualities of image and
position data. Digital workflow from flight planning through
acquisition to product generation is one of the major advantages -
of the sensor. The design principles and advantages compared
to film cameras are discussed in detail in (Sandau, 2000).
The ADS40 generates around 100GB of raw data per hour if all
the CCD lines are active. This data passes through different
levels of processing before products are generated; these
processes are well documented in (Tempelmann, 2000). The
first step in the workflow is downloading the data and adding
the geo-positioning information using data supplied by a
Position and Orientation System from Applanix Corporation.
The geo-referenced images are then rectified to a plane to create
stereo-viewable and geometrically corrected Level 1 images.
The Level 1 images are also useful for the next process which is
automatic point matching. Based on the accuracy requirements,
the images are then triangulated using ORIMA, LGGM’s
triangulation package. Finally, ortho-photo images using
existing DTM or DTM generated from the triangulated images
are created. Due to the large amount of data being processed
this workflow could take an extended time. GPro is the ground
processing software for the ADS40 that controls the workflow
described. It provides the full functionality to download,
generate geo-referenced and ortho-rectified images from the
recorded imagery and positioning data. It consists of highly
optimized and threaded applications that are designed to handle
large data sets. In addition, GPro also allows the user to perform
various automated process and data management tasks.
The last two years have shown that digital sensors, especially
the ADS40, are able to meet the accuracy and radiometry
required for large-scale ortho-photo generation. However, the
quick turnaround time these sensors provide has been a
challenge to realize to its full potential using a single computer.
High volume production businesses like North West Geomatics
(NWG) have already risen the bar for flight to product
turnaround times using the ADS40. This problem can only be
Inter
addr
Perfc
used
signi
amot
fast 1
solve
com[
in cc
super
This
with
senso
how
proce
indic:
plans
North
ADS“
their |
more
Leica
distril
select
proble
@
e
e
e
The n
results
didn't
config
highly
only w
using
workfl
major
have a
to dep
our me
works
require
As in ¢
for pro
critical
volume
content
allow €
toleran:
enviror
develoy
to be n
tools ir
implem
Lastly,
used fc
