next up previous contents
Next: Pauli Decomposition Up: How to do a Previous: Before you start   Contents

The data sets

check your data!!! make sure you got everything you need in the quality you need. The guys who process the data are usually experts and know there job. But still to err is human and sometimes even experts make errors. So always double check your data before you start, otherwise - garbage in - garbage out. You need a a fully polarimetric data set. That means 4 channels of SLC (single look complex) data sets. 'Real' complex data ;o) with Amplitude and phase (or real and imaginary part). Simple magnitude data sets without phase information are not sufficient for polarimetric techniques. In some cases you will only get 3 channels since after the calibration the cross-polar channels are usually equal (the reciprocity theorem - you remember?). Make sure you know the data format and how to read the data. Each and everyone has his own preferred data format, so make sure you know how to read your data. In some case (e.g. for the Danish EMISAR data you first have to transform the data into a standard format. At http://epsilon.nought.de you can find a routine which reads some commonly used data formats.

O.k. make sure that all channels have the same size (if not something is fishy) and that they really cover the same area. Once you are sure that the data sets are 'formally o.k. you should perform a quick quality check. Calibration is real hard core stuff and I recommend to leave that part to the experts, but you should at least check if your data sets are reliable. Some basic checks involve the analysis of a/some corner reflectors (if you have some in the scene) trihedral corner reflectors are widely used as calibration targets. The trihedral structure results in a triple bounce (odd-bounce) scattering behavior. Therefore the cross-polar backscatter should be much much lower than the co-polar backscatter. The amplitudes of the HH and the VV channel should be similar and the phase difference close to zero. If you have no trihedral corner reflectors in your scene you can use a good reflecting surface (make sure that you get enough backscatter). For an overall analysis you can also use a Pauli or Sphere/diplane/Helix decomposition. Just look at the result of the decomposition and ponder if the result is reasonable. Actually I suggest to make the Pauli or Sphere/diplane/Helix decomposition in any case and to try to interpret what you see. You can learn a lot just from fooling around with the data. Once you think you know your data and the data quality is o.k. you can start the real work

In the following I wrote down some simple step by step recipes for the decomposition approaches we discussed before.


next up previous contents
Next: Pauli Decomposition Up: How to do a Previous: Before you start   Contents