Unsupervised segmentation of polarimetric SAR data using the covariance matrix

E Rignot,P Dubois,R Chellappa

doi:10.1109/36.158863

Abstract

A method for unsupervised segmentation of polarimetric synthetic aperture radar (SAR) data into classes of homogeneous microwave polarimetric backscatter characteristics is presented. Classes of polarimetric backscatter are selected on the basis of a multidimensional fuzzy clustering of the logarithm of the parameters composing the polarimetric covariance matrix. The clustering procedure uses both polarimetric amplitude and phase information, is adapted to the presence of image speckle, and does not require an arbitrary weighting of the different polarimetric channels; it also provides a partitioning of each data sample used for clustering into multiple clusters. Given the classes of polarimetric backscatter, the entire image is classified using a maximum a posteriori polarimetric classifier. Four-look polarimetric SAR complex data of lava flows and of sea ice acquired by the NASA/JPL airborne polarimetric radar (AIRSAR) are segmented using this technique.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>

Highlights

Anumber of polarimetric synthetic aperture radar (SAR) analysis techniques have been reported in the literature to measure and characterize the polarization response of natural targets [l], to maximize the contrast between regions based on polarimetric filtering [2], or to classify data using Bayes’ classifier [3]-[7]M. ost of these techniques are supervised and require selection of training areas for each class of terrain cover
Fig. l(a) shows a color overlay at L-band frequency ( X = 23.98 cm) and three different polarizations (HH is red, VV is green, and HV is blue) of a four-look acquired by the NASNJPL airborne polarimetric radar (AIRSAR) image of the Pisgah lava flows in the Mojave Desert, CA [21]
The scene contains various geological surfaces [13], [21], [22], which can be divided into six classes of terrain cover: phase I lava, phase I1 lava, phase I11 lava, cobble, alluvial surface, and dry lake bed

Summary

INTRODUCTION

Anumber of polarimetric SAR analysis techniques have been reported in the literature to measure and characterize the polarization response of natural targets [l], to maximize the contrast between regions based on polarimetric filtering [2], or to classify data using Bayes’ classifier [3]-[7]M. ost of these techniques are supervised and require selection of training areas for each class of terrain cover. An accurate and detailed knowledge of the scene contents is required to select the appropriate classes, and training areas should be homogeneous and contain enough samples to estimate the polarimetric backscatter characteristics of each class with good accuracy. This is not always possible, e.g., in the case of sea ice studies where ground truth data are often sparse and time-limited (due to the cost of collecting extensive ground truth information and the rapid evolution of the ice), and sea ice features are difficult to characterize and extract (e.g., ridges, broken-up ice floes, and open leads).

11. SELECTION OF THE IMAGE CLASSES

Images of Lava Flows

Images of Sea Ice

CONCLUSION