Abstract
The advent of fully polarimetric systems has led to an increased amount of information acquired by synthetic aperture radar (SAR) sensors but also to an increased complexity of the data to be analysed and interpreted. In particular, the choice of several representations of the data, in terms of different parameters with peculiar characteristics and physical meaning, has been offered. With this work, we intend to address their systematic investigation with a twofold goal: (1) to provide a brief review of the polarimetric representations under consideration; and (2) to characterize and compare them with respect to their usefulness for classification purposes.
The analysis procedure consists of the accuracy estimation of classification tests performed on different parameters derived from L‐band polarimetric SAR data. In order to ensure a common basis for their comparison, a neural network classifier, the Multi‐Layer Perceptron trained by the Back‐Propagation learning rule, was used which permits us to operate on the data without making any a priori assumption on their statistics. In this way, the considered polarimetric parameters, in general characterized by different statistical distributions, may undergo the same classification process and the results compared.
Our results indicate that the overall classification performance varies depending on the polarimetric parameters used. However, these variations are relatively limited and do not permit us, at this stage, to define an ‘absolute’ best representation to identify the classes under investigation in an optimal way.
Acknowledgements
This work was also supported by the European Union in the frame of the Research Training Network (RTN) ‘Application of Multiparameter Polarimetry in Environmental Remote sensing’ (AMPER) and partly conducted at the Danish Defence Research Establishment (DDRE), Copenhagen, Denmark, and at the Microwaves and Radar Institute (HR) of the German Aerospace Centre (DLR), Oberpfaffenhofen, Germany; this last institution kindly provided the experimental data.
Some of the SAR data processing was performed using the free RAT software developed at the Department of Computer Vision and Remote Sensing of the Technical University of Berlin, Germany; the authors would like to acknowledge the excellent service rendered by the RAT team. Thanks also to Andreas Danklmayer (DLR/HR) for collaborating to the implementation of the PCA algorithm.
Finally, we would like to thank Madhu Chandra, Gerd Wanielik (Technical University of Chemnitz, Germany) and Ernst Krogager (DDRE) for the helpful discussions and Nicole van Barneveld for proof‐reading the text.
Notes
1. However, we could show in previous works (Alberga, Citation2007; Alberga et al., Citation2006) that also ‘rough’ statistical approximations or simpler classification algorithms do provide reliable results.
2. One should note the difference in the use of the word ‘symmetry’ when referred to scattering matrices and to targets. Indeed, according to the Cameron decomposition, scattering matrices which are symmetric due to the reciprocity constraint may describe targets which are geometrically more or less symmetric in the plane orthogonal to the radar line‐of‐sight (in the case of a helix, a symmetric scattering matrix represents a target which is not geometrically symmetric).