PERFORMANCE ANALYSIS OF DISTRIBUTED ORTHO GENERATION USING ERDAS
ORTHO ACCELERATOR (LOA)
Yasemin Kuzu-Sinram
ERDAS, Inc. 5051 Peachtree Comers Circle, Norcross, Georgia 30092-2500 - yasemin.sinram@erdas.com
Commission IV, WG IV/3
KEY WORDS: Distributed, Automation, Processing, Digital, Orthorectification, Performance
ABSTRACT:
The demand for digital ortho imagery is rapidly increasing with the growth of Internet and GIS applications. Orthophotos are a
valuable source of information for government agencies, engineers, planners, land managers and individuals. Orthophotos are
required to be updated periodically and the delivery times are decreasing rapidly. The size and the number of the images increase
with the demand for more detailed and accurate imagery. Since ortho rectification projects are becoming very data heavy, ortho
production is a bottleneck for photogrammetric shops due to huge processing times which is often too high for single processor
architectures.In this paper we are examining the benefits of the ERDAS Ortho Accelerator by analyzing the performance of the
distributed processing. In this experiment, we have chosen large orthorectification jobs to show the time savings of distributed
processing. Distributed orthorectification processing performance metrics are analyzed with respect to the number of processing
nodes applied. The results are also compared to stand-alone processing performance. This paper shows distributed processing is
necessary with large photogrammetric projects as processing time can decrease significantly with the number of nodes used.
1. INTRODUCTION
Adoption rates of distributed processing in the photogrammetry
industry have not been universal. Photogrammetric processing
is largely workstation-based. With the advances in computer
and network hardware, current inexpensive computer systems
can handle larger amounts of data and computer prices are
getting relatively cheaper. In order to generate orthos in reduced
delivery times demanded by the market, photogrammetric
workflows can use automated procedures such as distributed
computing. Distributed computing can be defined as the ability
to divide large processing jobs into smaller tasks and running
them on multiple processors or multiple machines. Distributed
processing harnesses the processing power of several CPU
nodes to increase throughput. It results in greater throughput
and frees up operators to perform other tasks. Also, purchasing
nodes instead of softcopy workstations will result in cost
savings.
ERDAS Ortho Accelerator (LOA) is developed under a
strategic development agreement between Leica Geosystems
Geospatial Imaging (now ERDAS) and GeoCue Corporation
(formerly NIIRS10). LOA was developed to speed up the rate
with which digital orthophotos are produced. LOA integrates
ERDAS photogrammetric processing components into the
GeoCue geospatial process management framework to create a
very efficient, enterprise-enabled orthophoto production
environment. LOA is an extension (CuePac) to GeoCue. A
CuePac is a collection of GeoCue menus, checklist and
auxiliary programs that implement a “canned” set of workflows.
GeoCue is not workstation centric. The command dispatch
system allows distributing ortho jobs to remote nodes. With
orthorectification and mosaicking capabilities, ERDAS Ortho
Accelerator takes advantage of the distributed and scheduled
workflow processing capabilities as well as process
management tools provided by GeoCue.
LOA provides below solutions to several market problems
which exist in orthophoto production processes:
• Orthorectification jobs are distributed to individual
machines (ortho nodes).
• Dispatched tasks can be run on a machine other than
the one on which it was launched
• Orthorectification and mosaicing jobs can be
scheduled to run at user defined times
• Multi user access to the same project from any
workstation in the network.
• Accurate and real-time monitoring of scheduled tasks.
• Each manipulation of the data is incrementally saved
via transaction processing against the database.
• The system is protected from unauthorized access to
data
• User-defined metadata can be associated with the
imagery
LOA consists of 3 modules which are: Project Importer, Ortho
CuePac and Mosaic CuePac. LOA is built on LPS and ERDAS
mosaicing components.
Project Importer enables introducing a photogrammetric
project into the GeoCue managed workflow. LOA supports LPS
block files, BAE SOCET SET® projects, INPHO MATCH-AT
projects and Intergraph Z/I Imaging® and ImageStation®
Automatic Triangulation (ISAT) projects. When a
photogrammetric project is imported, graphical objects are
created in GeoCue with geographically correct footprints and
locations in the Map View. Also, metadata is attached to each
entity that includes fields such as interior and exterior
orientation, image size, camera information and so fort.
Ortho CuePac is an enterprise ortho production system that is