se 1201 10 
Pd 
DATA INTEGRATION 
The use of digital ancillary data helps to increase the speed 
and accuracy of satellite image analysis. However, much of 
the detailed digital information which would be considered 
helpful is typically stored in a vector format. Since satellite 
data is stored in a raster format, there has been the problem of 
how to store two completely different data sources on one 
system. Recently, driven by the need to transfer data from 
one format to another, many software companies have 
implemented software which allows the user to purchase data 
in one file structure and convert it to a different software 
environment. This implementation has created an open 
market for agencies to purchase ready made data bases to 
update rather than re-create. 
Software which allows the complete integration of vector and 
raster image data also allows for the creation of more 
innovative approaches to satellite analysis. Computer firms 
such as ERDAS and ESRI, Inc., (Redlands, California) have 
linked forces to create recent software advances which allow 
vector and raster data bases to reside within the same 
software environment. This capability allows raster and 
vector coverages to be used interchangeably in a way that is 
transparent to the user. Advances include the ability to 
display spatial vector data on top of spatial/spectral image 
satellite data. Once integrated, these data can be used 
interchangeably; vector data can be used to enhance raster 
data for image processing, or raster data can be used to 
update and enhance vector data to reflect current land 
conditions. 
PROCESSING SATELLITE IMAGERY USING 
ANCILLARY INFORMATION 
The value of the information derived from satellite data is 
directly related to the spatial/spectral quality of that data and 
the scientific methodology used to extract the information. 
Regardless of the quality of the data, if the methodology used 
to analyze the data is weak, the information derived from the 
data will be weak. Therefore, it is very important that the 
image analyst use every resource available to extract 
information and then later verify it in accuracy assessments. 
In working with large satellite coverage study areas, the 
methodology for image processing, classification, or data 
base development must be well thought out and focused on 
the project rationale. A project map or progress chart should 
be created by the project manager and image processing team. 
Once the final objectives of the image product have been 
established, a list of potential ancillary products and their 
potential uses should be listed. Each set of ancillary data 
should be discussed as to the relevancy of its use and how 
much additional work may be needed to bring the data into 
the system environment. Issues of concern include: How 
much will the data cost? How much time is needed to convert 
the data? Will it be necessary to purchase new software to 
complete the conversion? What level of personnel experience 
will it take to complete the data conversion? Each of these 
issues should be considered and factored into the overall time 
needed to complete the image analysis. 
The following are two image processing case studies 
completed by ERDAS Production Services. These two 
studies were generated under completely different 
circumstances and required completely different types of 
ancillary data. The first case study was a typical classification 
study using standard types of ancillary information 
commonly used by today's image analyst. However, the 
ability to convert information from raster to vector formats 
speeded up the image processing stage. The second image 
processing study was unique in that 50% or more of the 
classification was completed using ancillary data, the time 
needed to complete the project was cut in half, and no 
ground truthing was completed. 
105 
Case Study One - Wetlands Classification of Georgia 
The Georgia Department of Natural Resources (DNR) 
contracted with Production Services to produce a full land 
cover classification of the state for the purpose of identifying 
wetlands. The maps resulting from the classification effort 
were to be used as a reference tool in development offices 
throughout the state to identify and protect ecologically fragile 
areas. 
Due to the size of the area to be mapped it became apparent 
that satellite imagery would allow the quickest, most cost- 
effective method of attaining wetland information. The study 
area consisted of 11 full and two one-quarter scenes of 
EOSAT Thematic Mapper data, which stretched across 10 
distinct physiographic regions. Because limited aerial 
photography was available for training sample information or 
accuracy assessment, the Georgia DNR agreed to provide this 
information in the form of field investigation. Because of the 
size of the area it was decided that the following ancillary data 
would be used: 
« 1:24,000 USGS Topographic Maps (1016 total) 
» USGS DEM Data 
* NHAP Aerial Photography (ordered for major 
waterways) 
* LORAN Helicopter Coordinates 
* Physiographic Map of Georgia, 1976 
+ ARC/INFOTM Map Grid 
The state of Georgia is unique in that it is covered by several 
very distinct physiographic regions. These region 
differences, combined with the difference between each 
scene’s date and path complicated the classification task 
further. As a result, it was decided that the state would be 
divided into regions based on a composite of the major 
physiographic regions and the Landsat satellite path. 
Division according to physiographic regions was done to 
simplify the classification by isolating some of the endemic 
geomorphic and vegetative differences which occur between 
regions. 
To complete this task, scenes along the same path were 
mosaicked and then masked according to the region boundary 
lines. Each region was classified separately as a unique study 
area and assessed for accuracy by region. At the conclusion 
of the project the state of Georgia was completed in 14 
separate classifications. Breaking the data into unique logical 
components which could be analyzed separately put the study 
in perspective for the project team. It also limited the 
overwhelming amount of image processing to be done and set 
up a cycle for work progression. The project was 
completed at an overall confidence level of greater then 85% 
for the entire state, and several regions reached accuracy 
levels of 90%. 
Other ancillary data was used throughout the project. The 
mountainous areas of North Georgia required extensive use 
of topographic maps, GIS techniques to extract classification 
information from shadowed areas, and a limited amount DEM 
data. Final classification accuracy assessments were 
completed using GIS coordinates converted to LORAN 
helicopter coordinates. 
Services to the Georgia DNR included the full image 
classification and production of raster GIS files of the 7 1/2 
minute quad areas of Georgia (1,016 total), color electrostatic