Assessment of crop discrimination using polarimetric C-band SAR data
18 ave. Edouard Belin,
(2) University of Sheffield, Hicks Building, Sheffield S3 7RH, United Kingdom
(3) Wageningen Agricultural University, Nieuwe Kanaal 11, 6709 PA Wageningen, Netherlands
(4) Technical University of Denmark, Oersteds Plads, Building 348, DK-2800 Kgs. Lyngby, Denmark
In the frame of the POLSAR project, multi-temporal C-band polarimetric SAR data over Flevoland (NL) have been analysed, aiming at selecting optimum parameters for crop classification.
The dataset, held in the European ERA-ORA database, contains a large amount of measurements derived from AIRSAR data, acquired on June 15, and July 3, 12 and 28, 1991.
The data analysis shows very strong temporal variation in the intensity and polarimetric measurements during June and July. Relatively invariant features for different dates in July appear to be the RR/RL ratio to discriminate crops with vertical from those with random structure, the correlation between HH and VV to separate structure in sugarbeet and potatoes, HV to separate crops with different biomass levels, and the HH/HV ratio to separate vegetation from bare soil (or harvested fields).
Different classification schemes have been applied to the filtered image data, in particular hierarchical classification using the optimum features, ISODATA clustering followed by classification, and WISHART classification. The results are compared and interpreted, and conclusions are given on the most effective use of polarimetric SAR data for crop classification.