On the behavior of information theoretic criteria for model order selection of InSAR signals corrupted by multiplicative noise

摘要

This paper deals with the estimation of the number of components in a multibaseline interferometric synthetic aperture radar (InSAR) signal corrupted by multiplicative noise, in the presence of the layover phenomenon. The appearance of multiplicative noise, termed speckle in the radar jargon, makes this problem very atypical. In fact, all the approaches proposed in literature have been applied to constant amplitude sinusoidal signals. In particular, the information theoretic criteria (ITC) have been conceived to estimate the number of signal components embedded in additive white noise. In this case, the problem of model order selection is equivalent to the estimation of the multiplicity of the smallest eigenvalues of the data covariance matrix. In presence of multiplicative noise, the signal eigenvalues spectrum changes. Consequently, the classical ITC methods operates under model mismatch. Nevertheless, before to look for other ad hoc methods, which could be difficult to derive or too heavy to implement, it is reasonable from an engineering point of view to investigate how the classical ITC methods are robust to the presence of multiplicative noise. Performance of the various ITC methods are analysed and compared under different operational scenarios. The relationship between performance and the eigenvalues distribution of the true covariance matrix is also investigated.