Classification of full polarimetric SAR-data using artificial neural networks and fuzzy algorithms

Within recent years several investigations have reported the use of polarimetric data to map Earth terrain types and land covers. For the operational application some demands, besides the accuracy requirements, must be fulfilled. In order to make the handling of the classification easy for the common user, the algorithm has to be data set independent and the handling must be possible without a priori knowledge. The authors outline a classification based on the entropy (H)-/spl alpha/ decomposition theorem extended by the use of the first eigenvalue of the coherency matrix. Fuzzy algorithms as well as artificial neural network (ANN) strategies are applied to improve the classification accuracy and to enhance the handling. The algorithms are applied to a L-band data set of the test site Oberpfaffenhofen, Germany, acquired with the DLR's airborne Experimental SAR (E-SAR) in April 1997. The classification results are discussed and compared to reference data i.e. topographic maps in the scale 1:25000 (maps 7833 and 7933) and airborne optical data acquired with a ZEISS RMK A30/23 Camera in August 1997.

[1]  Raúl Rojas,et al.  Theorie der neuronalen Netze , 1993 .

[2]  Eric Pottier,et al.  An entropy based classification scheme for land applications of polarimetric SAR , 1997, IEEE Trans. Geosci. Remote. Sens..

[3]  Lotfi A. Zadeh,et al.  Fuzzy Sets , 1996, Inf. Control..

[4]  Raúl Rojas Der Backpropagation-Algorithmus , 1993 .