International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
3. IMPLEMENTATION
3.1 The Development of Geospatial Web Services
To build the prototype system, we need to have the geospatial
data and services be available over the Web. With the OGC
and W3C standards that define the interfaces for accessing the
geospatial data and the method for constructing chainable
geospatial Web services, we have implemented a limited
number of interoperable services, including WCSPortral
service, reprojection, visualization, classifications, reformatting,
subsetting, and georectification. Those services are
implemented with standard interfaces defined by OGC for
taking coverage data from coverage servers or from the output
of other services as the inputs.
WCSPortal service has been built first because the OGC Web
coverage service (WCS) [OGC, 2003] provides a way to access
and fetch geospatial data online [ ]. This WCSPortal service is a
wrapper of WCS. It uses the GET method of WCS to get a
coverage, save it locally, and return a URL for next step.
Some very useful geospatial Web services such as reformatting,
re-projection, sub-setting, geo-rectification are also being
implemented. We can envision that more and more services and
data will be available on-line in the near future. Those services
will be stand-alone applications that can be used independently
as well as the components for chaining. And the availability of
services on line will also save many data users significan time
in preprocessing the data for their own use.
Some image classification services are also implemented for
demonstration, such as “WICS Grass Unsupervised” , a
SOAP based Web service. It uses Geographic Resources
Analysis Support System (GRASS), an open source GIS
system, to do an unsupervised image classification. We use one
of our image classification services as an example to represent
our service model and structure in Figure 2. The bottom part is
the Image Classification Core Class. The basic classification
algorithm is written in C (or another computer language). A
Java class — ImageClassifier — provides Java access to those
methods through the Java Native Interface .The Image
Classification Core Class provide the functions to be used in
classification services, such as supervised classification,
unsupervised classification, and training. This Core Class will
be called by the Web Image Classification Service.
The Web Image Classification Service consists of two different
implementations, the HTTP GET & POST version and the
SOAP version. Both versions provide four operations,
GetCapability, ^ DescribeClassifier, TrainClassifier, and
GetClassifier. Both versions accept the same parameters. The
only difference is the protocol of request and response.
The WICS Web interface is a Web application, including a set
of html pages and a Java servlet. To the Web
Classification Service, this Web interface is a client. A user can
use these Web pages to submit a WICS request to a Web Image
Classification Service (HTTP version) and specify whether he
wants the raw data back or a picture result. The Web interface
will generate pictures for the user if requested. Also the user
can use other client tools to connect to both versions of the
service.
Image
220
The details for those services and the WSDL documents
describing those services can be accessed by visiting our
Website http://ws.laits.gmu.edu.
p^ User i
| |
ARLE EER E = 3
WICS web interface | Client
(A Wek Application halt wath HTML page & Java Servier) | Software
| |
| i
* Ed. v
Web Image Classification Service
HTTP GET & POST
(eva Serviet)
|
|
|
i
Image Classification Core Class |
e i
|
|
ImageClassifier (Java Class)
Java Native Interface
Figure 1. The Web Image Classifcation Service
Figure 2. The Web Image Classification Service Structure
3.2 The Development of the Prototype System
The real power of a geospatial Web service system relies on its
ability to chain individual standards-based services. A service
chain is a service composite. Built on XML, SOAP, and
WSDL, and originated from XLANG and WSFL, the Business
Process Execution Language for Web Service (BPELAWS)
specification is positioned to become the Web service standard
composition language over all others. It supports two distinct
usage scenarios: one is used to describe an abstract process
which is non-executable and the other is to define an executable
process. We adopt BPELAWS in our geospatial Web service
system. We use BPELAWS to define a new Web service by
composing a set of existing services ( a service chain). The
interface of the composite service is described as a collection of
WSDL portTypes, just like other Web services. The chain is
executed with the BPEL engine. Figure 3 is the current
prototype system structure implementation flowchart.
i
|
—m
|
AllsIdaw
Figure 3. The Implementation of the Prototype System
Internatio
DG iiia
The core
compone
Catalogue
process
construct
will prod
the servic
through !
document
are coup
discovery
publishes
services (
theBPEL
service of
for users |
interests.
3.3 Hlus
For demc
system w.
Francisco
ways. Foi
service cl
the availa
data for 1
user sear
service to
unsupervi
services |
image (Fi
files are
WCSPort
WICS G
services r
as a new
define the
engine ta
result of
invisible
without tl
the existi:
know, on
service in
a chain |
testing an
{i GEOSPATIAL Wi
# wms_Bosto
3E WCS Rasta
35 WCSPortalS
x WICS Grass
X5 Reprojectior