Full text: Fusion of sensor data, knowledge sources and algorithms for extraction and classification of topographic objects

International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999 
image representing different spectral regions and (ii) to get 
information on Fe 2 /Fe 3 minerals by using the signal in the blue 
band (Kaufmann et al., 1995). 
The MOMS-02 imaging geometry is shown in Figure 1. Due to 
the higher orbit of the MIR station (350 - 410 km) compared to 
the Space Shuttle flight of D2 (290-310 km), the MOMS-2P 
datasets have lower geometric resolution and larger swath width 
(Table 1). 
1.2. Scientific background 
In remote sensing, the classification of surfaces is done by 
analysing reflectance. Targets of the natural environment show 
significant reflectance differences, dependent on wavelength, 
irradiation conditions, look direction and look-angle differences 
(Kriebel, 1978; Kimes, 1983; Kadro, 1981; Deering et al., 
1992; Ammer et al., 1991 etc.). Five generically different 
"signature" types are known: spatial, spectral, temporal, angular 
and polarisation signatures (Gerstl, 1990). 
Due to practical reasons, mainly the spectral characteristics have 
been used for the classification of spacebome data. Spatial 
(structure, texture) and temporal (multi-temporal, multi- 
seasonal) characteristics mainly serve to improve the results 
retrieved from spectral data analysis. The use of angular and 
polarisation signatures in optical remote sensing is still in an 
experimental stage. 
The "multispectral" approach is based on the different 
reflectance behaviour of nonvegetated and vegetated surfaces 
(Figure 2). Nonvegetated surfaces like soils, bare rocks, etc., 
show a nearly linear, slow increase in reflectance from the blue 
to the shortwave infrared (SWIR) region. In contrary, 
vegetation’s reflectance is characterised by typical absorption 
features in the blue and red region as well as by a steep increase, 
the so-called "red edge" towards the near infrared (NIR). 
The "anisotropy" classification approach under investigation 
takes advantage of the fact that soils and nonvegetated surfaces 
show in general a backward orientated backscatter characteristic 
(Irons et al., 1989; Irons et al., 1992), while vegetated surfaces 
show a more forward orientated one (Deering et al., 1992) (see 
Figure 3). A significant exception is the so-called "hot spot" 
effect, which gives a sharp maximum of the backscatter 
reflection (Kuusk, 1991) around the incidence point, caused by 
the lack of shadow combined with the fraction of specular re 
flectance. Gerstl et al. (1986) proposed to use this effect for 
crop identification. 
In case of the angular signatures evaluated for the anisotropy 
approach, the signal consists of a basic „spectral minimum 
albedo“ fraction, which is equal for all observation directions, 
and a superimposed fraction which differs depending on look- 
angle (see Figure 3b and c). This means that a wet soil is under 
each observation angle darker than a dry soil, while the 
difference between forward and backward looking signal may 
be the same for both. 
Anisotropy effects influence multispectral data analysis, 
especially in datasets of coarse resolution satellites (NOAA, 
AVHRR, WIFS, etc.) and high resolution airborne spectral 
scanner data (DAEDALUS ATM, AVIRIS, DAIS, HyMap, 
etc.), both with an unfavourable swath width to flight altitude 
ratio, but also in high resolution satellite data of operational 
sensors (Landsat TM, SPOT, IRS-1C, etc.). 
The ongoing investigations on the angular signatures are 
mainly concentrating on the elimination of the anisotropy 
effects. Recent research activities to approximate the 
bidirectional reflection distribution function (BRDF) of natural 
surfaces are mainly driven by the necessity to normalise across- 
track effects on imaging spectroscopy data, a precondition for 
the quantification of the bio- and chemo-physical parameters of 
the surfaces of interest (Martonchik, 1994; Sandmeier and Itten, 
1999; Solheim, 1998). 
Fig. 3. Backscatter characteristics of natural surfaces. 
Signal correction. For both multispectral and anisotropy 
approaches, sources that influence the image information have 
to be considered before relating the signal to bio- and chemo- 
physical parameters of the surface. The three main signal 
alteration sources are related to: (i) transmissivity of the 
atmosphere, (ii) sensor performance and calibration, and (iii) 
terrain relief effects. 
2.1. Test site 
The main test site covers the area of the research farm 
„Dürnast“ of the Agricultural Faculty of the TU of Munich in 
Freising /Weihenstephan with an average altitude of about 450 
m above sea level. It has an area of about 7 x 2.5 km and lies 
between 11°37’30.4” E and 11°43’55.4” E longitude and 
48°22’39.4” and 48°24’59.4” N latitude. The area is used 
mainly for agriculture and is including a part of the city of 
Freising and some villages as well. Moreover, there is a big area 
covered by coniferous and broad-leaved forest. 
The main part of the test area belongs to tertiary aged hills, 
formed from clastic sediments of the upper molassic series. 
Common soils in this area are sandy loams and loam with a 
score from 34 to 67 on a scale from 0 (for unproductive soil) to 
100 (very productive soil). The flat areas in the SE belongs to 
the so-called „Munich gravel plane“, a tectonic basin, filled 
after the last glacial period with mostly carbonatic sediments of

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.