111
Symposium on Remote Sensing for Resources Development and Environmental Management / Enschede / August 1986
The determination of optimum parameters for identification
of agricultural crops with airborne SLAR data
P.Binnenkade
National Aerospace Laboratory NLR, Amsterdam, Netherlands
ABSTRACT: The SLAR campaigns of the Dutch ROVE-Team "Crop Identification" in the years 1983-1985 are
discussed. From data—acquisition through classification the determining parameters are elaborated. The 1985
campaign which used steeper incidence angles in order to simulate satellite data, and is thus different from
previous ROVE-campaigns, is discussed.
1 INTRODUCTION
Microwave remote sensing has played an important
role in The Netherlands Remote Sensing community
since the late sixties. Especially the work on the
radar backscatter of vegetation canopies has always
been given great emphasis. The ROVE-team (ROVE
stands for Radar Observation of VEgetation), which
has been involved in all land-applications of micro-
wave remote sensing in The Netherlands, then and
today consists of representative scientists and
technicians from the following institutes:
- Agricultural University of Wageningen and its
associated institutes
- Delft University of Technology, Microwave Group
- Physics and Electronics Laboratory TNO
- National Aerospace Laboratory NLR.
Starting with a pulse-type X-band measuring radar
operating from a TV-tower in 1968, the ROVE-group
nowadays makes use of a calibrated digital SLAR
(Fig. 1) mounted in the NLR Metro II Laboratory air
craft and a multiband airborne scatterometer
(DUTSCAT) installed in the NLR Queen Air laboratory
aircraft in combination with supporting ground truth
instrumentation.
One of the principal aims of the ROVE-group is the
identification and classification of agricultural
crops from microwave data.
Such data (obtained from airborne or spaceborne
sensors) contain only one type of information: radar
backscatter as a function of position. This radar
backscatter is the condensed result of a great
number of contributing factors. Even when confined
to only one frequency the radar return is influenced
by:
- polarization effects
- illumination function
- incidence (grazing) angle
- crop height
- crop coverage
- vegetation-structure parameters
- soil roughness
- moisture content.
Some of these factors may be considered fixed and
measurable whereas others can be influenced and
changed by setting the radar and/or antenna and by
flight-geometry modifications.
Nevertheless, due to the complexity of the above
parameters, work on theoretical models became indis-
pendable (Ulaby, Attema, Hoekman). At an early stage
it was recognized that this modelling work had to be
supported by ground-based measurements on crops and
soils. Special emphasis was given to the need for
simplicity in the models to limit the number of un
known parameters (Ref. 1).
2 CROP IDENTIFICATION
One of the specialized sub-groups of ROVE is the
workinggroup "Crop Identification" in which the
following institutes participate:
- Centre for Agrobiological Research CABO
- Delft University of Technology, Information Theory
group
- Physics and Electronics Laboratory TNO
- National Aerospace Laboratory NLR.
The workinggroup has chosen to undertake two
parallel paths of (applied) research:
1. Research into (hierarchical) classification
methods for radar data.
2. Initiation and execution of semi-operational
application programmes.
As for the latter, upon request from The Nether
lands agricultural authorities an application pro
gramme has been carried out in 1983 and 1984 to
monitor crop rotation, especially potatoes, in a
number of (ecologically-different) regions within
The Netherlands. Crop disease control authorities
allow potatoes to be grown only once every three
years on any one field (if no other protective
measures are taken); thus it has become desirable
to monitor adherence to this practice.
The group was asked to devise a method to uniquely
discriminate potatoes from all other crops, hence
time-consuming ground investigations be reduced.
Three test-sites with an area each of 20*4 km
were chosen in the Flevo-polders, Groningen and
Brabant province respectively.
3 THE DUTCH DIGITAL SLAR
Sideways-looking airborne radar (SLAR) is an in
strument which allows for data to be acquired on a
line- to-line basis. The calibrated digital SLAR
(X-band) uses a digital recording system where
each line is formed on a point-to-point basis
(Table 1). The received radar backscatter varies
widely in a stochastic way when natural surfaces
and -objects are observed; good estimates of the
backscatter coefficient can only be made by com
bining a large number of independent measurements.
The Netherlands SLAR has been designed to do so
(Table 1).