II., A SCHEME FOR COLLATION OF MULTI-STAGE MSS DATA
AND ITS APPLICATION IN FOREST COVER CLASSIFICATION
nson
by
A. KANDYA, W. MAUSER, P. LOSCHE and G. HILDEBRANDT
University of Freiburg
Federal Republic of Germany
Abs tenet
Since the launch of Landsat 1 in July 1971, number of rescarch
studios have been conducted using the satellite collected multi-
spectral scanner data and encouraging results have buen obtained.
However, il was cmphasized that the satellite dala , when analysed
in conjuction wilh airborne data and ground observations might
yield better results as cach successive data source might provide
more detailed information over smaller geographical areas. This
Concept ol multi-stage Romoto Sensing was proposed by a fow €
authors (Colwel! 1968, Langley 1969, Langley, Addrich, Heller
19002) for iuventories of natural resources. Langley doscribed a
mulbti-slageo Forest sampling study which combined Apollo 9 sipaco
photographs wilh aerial photography at scales of 1:60,000,
|: 12,000U and 1:2,000, Since the advent of LANDSAT sone otbor
authors described/discussed forest resources surveys for large
arcas based on Landsat data, areial photography and minimized
ground truth (Aldred 1972, Colwell 1973, Hildebrandt 1973 and
Schade 1979,1930). for :
In what follows is an attempt in the similar direction using
the scanner data. It is tried to establish correlation in the
spectral behaviour of a particular class in Lhe airborne and
salcilile collected MSS data. Homogeneous windows for the various
forest cover types were selected in the airborne data and werc
mapped into satellite data. Signatures for these windows in the
satellite data were generated and the maximum likelihood and
the minimum angle classification were performed. To allow com-
parision w:th the toposheets, lineprinter outputs of these re-
sults have been created to the scale of 1:25,000 and tho per-
formance of ihe two types of classifications was compared. Since
the min:mum angle classification uses the normalized data which
means the ratio of reflactance values in the different bands
and not the reflactance itself, it is possible to develop similar
ratios from the 1! channel Bendix Scanner data, Making use of
the sensiivity curves of the Landsat channels, these ratios
we odified to generate a new set of signatures for once again
performing tho minimum angle classification of the Landsat data.
The area for this study has been selected from the forested
rocion in Gottenheim,(Freiburg), Fed. Republic of Germany for
which the 11 channcl airborne data at the altitudes of 1000 and
2200 meters and the satellite data (in the same,season but in
the preceding year) were available. The high resolution camera
CIR photographs and field checks have been used for confirming
the cover types in the airborne data.
377
eee ey t oim NN a