The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi XXXVII. Part B4. Beijing 2008
207
4. GRADED PRODUCTS
The image service can be dynamically updated. Imagery can be
added, and various processes and parameters can be changed
without stopping the published image services. This is
important for the use of the system in time-critical applications,
and also enables the creation of a graded product. In this
method, a preliminary image service is created based on the
best available parameters and when the parameters and models
are refined, they are updated within the image service definition.
This provides a preliminary product as soon as the imagery is
available on a server. For example, an image service can be
created based on imagery from a digital camera as soon as the
data is available. The initial orientation for orthorectification
may come from preliminary orientation parameters available
from the GPS/IMU. The initial radiometric processing could be
default values. Such image services can be created as soon as an
aircraft has landed and made available to users—when the value
of the imagery is highest for applications, such as emergency
response. As different steps in the orthoimage production flow
are performed, the processes and the parameters in the image
service definition are updated. For example, after performing an
aerial triangulation the orientation parameters may be updated
increasing accuracy. After updating the DTM the accuracy
further increases. Similarly, the radiometric processing or pan-
sharpening parameters can be updated as improved values
become available. The initial product may not use seamlines
and after the system has generated the seamlines this can be
used to create a more seamless product. Thereby the graded
product improves over time.
5. QUALITY CONTROL
The methodology of ArcGIS Image Server not only enables use
of the imagery sooner which increases the value of the imagery,
but it is also used for efficient quality assurance processes and
is an integral part of Quality Control (QC).
In most production workflows, QC becomes a major component
of the total workflow. Some of the QC can be performed in an
automated manner, for example, in analyzing the histograms of
the output images, while other steps must be viewed by
operators to check the accuracy of control points or the
junctions of seamlines. If these QC processes require the final
product to be produced, then the processing must be repeated if
any errors are found. Since the processing with ArcGIS Image
Sever is performed on the fly, the QC steps can be performed
prior to the extensive processing. Corrections can be updated in
the database the corrections quickly reviewed. The integration
of ArcGIS Image Server into multiple GIS client applications
further enables the implementation of geospatially enabled QC
processes. With the ability to serve both the imagery and QC
tasks on the web, QC can be checked by the end user of the
product prior to production and delivery.
Such dynamic image services with on-the-fly processing have
significant advantages over image servers that only serve pre
generated and mosaicked static imagery. Dynamic imagery
servers that are driven from such databases and generate
products directly from the base imagery can be expected to
quickly augment the traditional static image servers that require
the pre-generation of large images or lots of individual tiles.
ArcGIS Image Server can easily be scaled to serve hundreds of
simultaneous users. To scale to very large numbers of
simultaneous users map caches are used. Web applications that
provide interactive display capabilities make requests to servers
using predefined tiles that are assumed to be static. Such static
tiles can be stored on the web servers as small static image files.
Such static tiles can be easily distributed on the Web and are
utilized by edge servers to reduce server requests, data transfers
and speed so up applications. Many web applications also cache
tiles locally to further reduce requests to the server and provide
faster interaction in areas that the user has already visited. Such
static tiles can be generated in advance and stored on the Web
server, but for large project would require extensive processing
that can take a long time. With ArcGIS Server 9.3, such tiles
need not be pre-generated. Instead they can be generated on
demand when a user first accesses an area at a specified
resolution. As the majority of high resolution requests to servers
repeatedly cover only a small percentage of the full area, the
high performance is maintained without the need to pre
generate high resolution caches for the complete area. Such
caching can be driven directly from ArcGIS Image Server,
where the cache is generated directly from the original image
data. Caches therefore have significant advantages over
solutions that utilize large static files that can not utilize web
caching infrastructures and need to be pre-generated.
7. ORTHOIMAGE PRODUCTION
Although dynamic serving of imagery directly from the source
and using on-demand caches have many advantages, there is
currently a demand to create orthoimages as tiles that conform
to a specific grid or map sheet layout. Many large orthoimage
mapping projects currently specify the creation of image
products as tiles in a range of formats including TIFF, NITF,
and JPEG 2000. It can be expected that the requirement to
deliver such pre-generated tiled products will continue as this
delivery mechanism is required by many organizations and
provides a clearly defined product that conforms to the pre
defined standards. Therefore, the need will remain for
companies to produce large sets of tiled products. With the size
and number of these projects increasing, production companies
are being challenged to create very large numbers of such tiled
images.
With ArcGIS Image Server, the actual processing of the pixels
is performed on demand as required for a user’s specified area
of interest, resolution, and projection. This processing can be
performed in near-real-time for screen sized requests as a user
zooms into a specific area. Large image requests, that define the
extent of a map sheet or large format plot at high resolution, can
also be made. This feature can be used to process and extract
large tiles of imagery from the server. The output can be written
to a number of formats such as GeoTIFF or JPEG 2000. As
with the smaller requests, the important image level metadata
can also be derived as output. This metadata is extensive and
includes the mosaic polygons that define which part of the
output comes from which input raster, as well as the processing
performed and metadata from the original imagery.
8. PRODUCTION SCALING
An important scalability feature of ArcGIS Image Server is its
ability to process multiple requests simultaneously and spread
the processing load over multiple service providers. The server