International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV,
2. STUDY SITE AND DATA SETS
The study site is located southeast of Jacksonville, North
Caroline, USA. It presents one of the largest US Marines sites
for which an extensive amount of ground truth, GIS, and remote
sensing data is available (fig. 1)
CMAPSVEST. a ^ff Enfield dau d$ 40mi
tes) Greensboro Gel A03 TH 7] Elizabe0E—— km
| Durham
X ‚Rocky | Mount - Cl columbia
i PAS ;
qme 23 As J frame 5
i20 ms “Wilson. 3S BE
Sanford d (35 QU | Goldsboro | ‘Washington ix Engelhard
és S SA 4 Dobef Gull Rok
ip E faÿetteviie Pagi m Bern ; _ocratoke
- Hope, Mills gana, ort
ve par kichiahds_ i
érection A © Jacksonville,/
% Lumberton: Toni ipe e.
; in Ne m elarkton, pug Morehead City
Dillon | 4
S fà eure” * Wilmington
Eiören« e Y
Florance, s : +6,
m Gor NC Makina Bolivia
3 Marion (i57 Lon Boach
Türbeville ME 17 9
foe ave Beach
| Garden city 2
“io : Georgetown
Moncks Corner
o
e 2004 MapQ S est com, Inc, 4
Figure 1. Camp Lejeune study site indicated by a red star
The study was conducted at the Center for Remote Sensing and
Mapping Science where the corresponding author stayed for his
sabbatical in 2003. The datasets consisted of Landsat, SPOT,
IKONOS and Qickbird images as well as GIS landuse/
landcover data in shape format (figure 2).
Figure 2. IKONOS multispectral image (2048 x 2048 subset)
of the Camp Lejeune study site overlaid with vector GIS
information
3. METHODOLOGY
Selected stretching especially for regions of low contrast is
nothing new in the analysis of remotely sensed data (see, for
example, Jensen, 1996). Usually, this is done interactively by
the analyst either by selecting a box or digitizing a certain area
398
Part B4. Istanbul 2004
of interest in the image. This area is then enhanced using
standard image processing techniques (e.g, histogram
equalization or linear contrast stretch). The subset is then
displayed separately to highlight certain features that would
have been impossible to discern in a global enhancement mode.
The goal of this study was to develop automated procedures for
feature based image enhancement techniques for rapid display
purposes, especially of high resolution remote sensing images
(Ehlers, 2004). Feature based enhancement means that different
feature classes in the image require different procedures for
optimum display. The procedures do not only encompass
locally varying enhancement techniques such as histogram
equalization or contrast stretch but also the selection of
different spectral bands. The image class water, for example,
may be best displayed in a true color mode whereas for the
feature class vegetation a false color infrared display is more
appropriate. It is envisioned that this technique could be
implemented in a near-realtime environment making use of a
priori information.
There are two main sources for this kind of information: (a)
storage of a priori knowledge in a GIS, and (b) context based
image information that can be extracted through a segmentation
process. Both techniques can also be applied for optimum
feature class selection.
For many areas in the world, there exists a wealth of a priori
information in existing spatial databases, digital maps or
previous analyses of remotely sensed data. Usually, this type of
information is stored in a raster or vector based GIS. With the
progress in the integration of remote sensing and GIS software,
many commercial systems allow the simultaneous display and
use of GIS and image layers. For a joint analysis, however,
usually GIS vector layers have to be converted to raster data.
GIS Remote Sensing Image
[mage Lp! Segmen-
tation
Yv v
Feature
: Feature
Selection and. | | ; 4
: Masking n
Encoding S
v
Band Selection
Image
Enhancement
Y
Merging of
Feature Based
Image Masks
Figure 3. Concept for GIS and context based image
enhancement
The case study for our research was conducted in an integrated
ArcGIS/ERDAS environment. The developed procedure,
however, is system independent and can work in any integrated
Internati
GIS/rem
steps in
masking
3.1 GIS
First, al
common
displaye
remotely
and sem
2).
In a sec
for local
classes.
all non-
GIS mas
separate
GIS lay
inspectic
(Figure -
Figure
‘Wat
The thir
based o
layers fi
| = are
indepen
is multi
multispe
values ¢
created
classes.
location
image
location
image s
the sepa
