collaborative analysis on medical image. Considering successful 
applications of CSCW in the area of medicine, it makes sense to 
introduce CSCW for collaborative interpretation of geospatial 
data (e.g. remote sensing images). (Baraghimian and Young, 
2001) proposed a virtual collaborative software environment 
called GeoSpace™ by combining improved InfoWorkSpace™ 
and GIS analysis functionalities. It mainly provides interactive 
analysis and visualization of geospatial information for 
collaborators. (Liu, 2004) develops a group-based image 
interpretation system for remote sensing applications. In this 
system, Windows Netmeeting tool is employed to implement 
information communication and application sharing among 
interpreters. (Convertino et al., 2005) integrates CSCW and 
Multiple View Visualizations to present a collaborative 
visualization framework for distributed and synchronous 
teamwork. (Xu, 2005) develop Socket messages of fixed format 
to pass cooperative perception among interpreters. 
(Austerschulte and KeBler, 2010) integrates several tools 
(ArcPad, GPS, etc.) to build a remote collaborative system that 
supports information sharing among the teams participating in 
geological data gathering. (Di Ciaccio et al., 2011) adopts web- 
services and other open standards and libraries to provide 
collaboration environment for disaster incident management. 
(Xu et al., 2011) gives redefinition and conceptual framework 
of collaborative virtual geographic environment to support geo- 
collaboration. All researches above are related work in 
collaborative image interpretation. Different technologies have 
been adopted to implement collaboration. However some 
shortcomings exist. For example, web service needs plenty of 
network resource and the server bottleneck exists. Specific 
Socket messages are hard to be extended. Windows Netmeeting 
tool is inefficient to share application in the WAN. Besides, all 
existed systems are hard to manage and visualize massive large- 
size images and referential information. In this paper, a KML- 
based approach is proposed for distributed collaborative 
interpretation of remote sensing images in the geo-browser. 
KML, as an OGC standard, is an xml-based file format used to 
display geographic data in a geo-browser. It supplies advances 
in extensibility and less data quantity. This article employs 
KML to share interpretation results and operations among 
interpreters. Meanwhile the geo-browser such as Google Earth 
has advantages in effective management, analysis and 
visualization of massive spatial data. It is appropriate to provide 
collaborators with unify interface and analysis tools of image 
browsing, operation and interpretation. 
2. DESIGN OVERVIEW 
2.1 Architecture of Collaborative Image Interpretation 
Firstly, this article will present the architecture of collaborative 
image interpretation. In order to provide collaborative 
interpretation environment in the geo-browser, some constraints 
are required. Firstly, data transmitting of large-scale image 
among collaboration clients should be avoided. It promises no 
limitation of network bandwidth. Secondly, it is prefer to 
conduct image processing in server-side. Moreover, 
geographically distributed interpreters can participate 
collaboration interpretation anytime and anywhere with a geo- 
browser. Finally, all interpreters can interactively communicate 
and share interpretation results with each other. In this paper, 
the architecture of collaborative image interpretation is 
composed of three tiers: collaboration client tier, spatial data 
server tier and collaboration sever tier (Figure 1). Collaboration 
client tier contains four main modules. There are spatial data 
visualization module, vector interpretation module, 
communication module and collaboration module. Spatial data 
visualization module is designed to display image, vector, 
annotation, DEM (digital elevation model) and other spatial 
data. Task images and results of vector interpretation are 
visualized in this module. Vector interpretation module 
provides various vector tools (point icon, annotation, polyline, 
polygon and compound geometry) to indicate targets in the 
image. Communication module is responsible for Socket 
messages communication among interpreters. All Socket 
messages are delivered to specified clients via collaboration 
server. Collaboration module implements cooperative work and 
collaboration awareness. Spatial data server tier is composed of 
one catalog server and several geospatial data servers. 
Geospatial data server provides tiles service of various spatial 
data, such as image, vector DEM, annotation, etc. All 
geospatial data servers register published tile services in the 
catalog server. Catalog server responses to spatial data requests 
and invokes registered tile services to obtain spatial data for 
clients. Collaboration server mainly implements three 
collaboration related services that are collaboration launch 
service, collaboration perceiving service and communication 
service. Collaboration launch service enables any collaboration 
client to create a collaborative image interpretation task. Task 
images and cooperative interpreters’ information are uploaded 
from the client. Collaboration perceiving service is called to 
share clients’ interpretation operations and results in the form of 
KML files. Communication service delivers Socket messages 
among collaboration interpreters. 
Spatial Data Server | Collaboration Client 
  
T Annotation Tiles service Geo-Browser 
Geospatial - = | 
Data Server 1 | DEM Tiles service [4H invoke i Ty 
Geospatial : WAL 
Server Visualization Module 
Data Server 2 erve { 1 ISustzation v ocu 
  
Image Tiles service | Vector Interpretation 
; : Catal requests Spatial Data 
Vector Tiles service og. reque p ala 
| 
  
Ttt em um UU mT ry med Module 
Ran les (4 Péceivins lon 
ervic ml/kmz: 1 
; Collaboration Module 
Task Image Launch Collaboration 
Files Service i Server - 
x invoke | [Socket Communication 
Collaboration Communication Service J { messages Module 
Interpreters List ; 
Collaboration Sever! 
Figure 1. Architecture of collaborative image interpretation 
2.2 Collaboration Mode 
In this section, the work mode of collaborative image 
interpretation will be detailed. Different from existed methods 
(Liu, 2004; Xu, 2005), this article adopts KML-based 
collaboration mode (Figure 2). In KML-based collaboration 
mode, any client could launch a collaboration interpretation task. 
The sponsor firstly selects remote sensing images and 
appropriate online interpreters. All these initial information are 
uploaded into collaboration server via collaboration launch 
service. Collaboration server responses to the service and 
performs image pre-processing, such as projection 
transformation, invalid pixel value elimination, image pyramid 
creation, etc. After that, collaboration server delivers task 
description and image configuration to specific clients. A 
uniform interpretation interface will be prepared for 
collaboration clients. Each client employs provided vector tools 
to accomplish image interpretation according to professional 
knowledge and personal experience. When ambiguity exists, 
interpreters could communicate each other with Socket 
messages (Maybe third-party voice software will provide better 
communication). Meanwhile collaboration awareness service 
208 
hel 
am 
con 
che 
and 
elei 
sen 
Int 
resi 
intc 
des 
usil 
3.1 
In 
col 
Co 
sen 
inf 
anc 
des 
des 
inf 
adc 
bo) 
use 
fou 
pul 
«P 
lea 
obj 
virt 
use 
<L 
coc 
«C 
inti 
des 
«n