rating 
apidly 
1990s 
] with 
raphic 
uction 
uction 
hybrid 
odetic 
buted 
ement 
Action. 
nming 
s, and 
The 
stem 
and 
DPS 
DPS 
is the 
gment 
DMA 
| data 
ollects 
to a 
on for 
kages 
. The 
ions, 
red to 
dcopy 
lation 
igery). 
errain 
lata is 
ise to 
in the 
ration 
, and 
y and 
cision 
stereo 
data 
| and 
roduct 
| DPS 
3 from 
signed 
nance 
n the 
s and 
esign, 
most 
ction 
ct with 
3 KBS 
action 
for DMA. It has been designed to place cartographic 
knowledge as well as, systematic search and reasoning 
skills at the finger tips of all DE/S cartographers in the 
production data extraction environment. With the 
objectives being improved production quality and 
throughput. 
DATA EXTRACTION SEGMENT 
The Data Extraction Segment provides the capabilities 
in the DPS environment to extract terrain elevation and 
feature data from digital stereo imagery using softcopy 
photogrammetric equipment; to edit and merge newly 
extracted data with existing MC&G data; to perform 
quality checks to ensure that the extraction process has 
conformed to production assignment specifications, and 
to store the resulting digital elevation data and feature 
data files in the MC&G Data Base for later use in 
production processes such as product finishing. 
In order to achieve DPS production requirements, the 
DE/S production environment consists of production and 
support activities. The DE/S production activities are 
segment control, data extraction, data integration, quality 
control and quality assurance. The support activities are 
training and maintenance. 
Segment control includes the functions of data access; 
internal and external communication; resource 
allocation and utilization; production monitoring; and 
production control. 
Data extraction is performed on the softcopy (digital) 
Terrain/Feature Extraction (TE/FE) workstation. Data 
extraction can be thought of as the most significant 
production activity in the DE/S production work flow. 
To support the collection of terrain data, DE/S 
functionality has Automated Terrain Extraction (ATE) 
and Interactive Terrain Extraction (ITE) subsystems. 
ATE provides the capability to automatically collect and 
output elevation data at fixed and variable grid spacings 
without the aid of the cartographer. ATE is the first in a 
sequence of processes performed for data extraction. 
ITE follows and utilizes the output generated from the 
ATE process. The ITE provides the capabilities to collect 
geomorphic feature data and fill adverse areas where 
the ATE was unable to extract elevation data. 
To support the collection of feature data, the DE/S 
provides the cartographer with an Automated Feature 
Extraction (AFE) subsystem. This subsystem contains a 
KBS and computer vision (CV) tools. The KBS supports 
the AFE process in the identification, attribution, and 
consistency checking of delineated feature data. The 
computer vision tools are designed to assist the 
cartographer in the automatic delineation, metric 
measurement, smoothing and thinning of feature data. 
The DE/S provides the cartographer with an option to 
perform automated delineation using CV tools or 
interactive delineation using graphic/editing tools. Once 
a delineation has been completed and accepted by the 
cartographer, the KBS is automatically invoked for the 
sequential execution of automated feature identification, 
attribution, and consistency checking of the delineation. 
The cartographer must respectively review all candidate 
identification, attribution, and consistency check queues. 
947 
During the review process the cartographer has the 
opportunity to override or accept a KBS suggestion. 
Data integration is performed on a Global Edit 
workstation. This workstation is a TE/FE workstation 
configured with additional storage space to 
accommodate feature and terrain data from multiple 
assignment areas. Data integration provides for the 
merging of terrain and feature data between adjacent 
data sets and for interactive editing of data. 
Quality assurance provides for inspection and 
verification of extracted data before release of the data 
to the MC&G Data Base. 
As for DE/S support activities, training utilizes DE/S 
subsystems as required. This training includes a Self- 
Paced Autotutorial System (SPAS) and conventional 
training courses for DE/S managers, operators, users, 
and programmers. SPAS training is computer-based 
instruction that is interactively done on the TE/FE 
workstation. 
DE/S maintenance consists of a collection of functions 
which aid in the general upkeep of DE/S hardware and 
software resources. These functions include system 
startup procedures, fault detection, software backups, 
corrective maintenance, preventative maintenance, 
equipment status reports, KBS maintenance, hardware 
upgrades, and software upgrades. 
The maintenance subfunction of primary interest to this 
paper is KBS maintenance. KBS maintenance provides 
a means for correcting problems with the KBS, for 
responding to extraction and product specification 
changes, and for enhancing the performance of the 
The DE/S KBS maintenance function is supported by 
the following resources: DE/S cartographers; Summary 
Reports; a Knowledge Engineer (KE) workstation; and a 
TE/FE workstation. DE/S cartographers are utilized as 
"domain experts" for knowledge acquisition activities in 
support of KBS rule development and performance 
analysis. The Summary Report is the primary source of 
information for assessing KBS performance. 
Essentially, it is an audit trail of KBS performance for an 
extraction assignment. Information collected includes 
statistical information about extracted data, cartographer 
comments, and the number of times KBS suggestions 
were overridden by the cartographer. The KE 
workstation is used for Knowledge Base rule editing and 
verification, data base management, and report 
management. The TE/FE workstation is used during 
Production Center scheduled maintenance sessions to 
test candidate changes and new rules for the KBS. 
KNOWLEDGE BASE SYSTEM DESIGN 
Much ingenuity and effort has been expended tackling 
the design and integration of the DE/S KBS. Located in 
the AFE subsystem, the underlying KBS software 
architecture is written in Ada and uses VAX relational 
data base software to store the specification files from 
which the AFE Knowledge Base is built. The 
operational concept of the KBS is to support automated 
feature extraction by serving as a decision aid in the