UNSUPERVISED FINE LAND CLASSIFICATION USING QUATERNION AUTOENCODER-BASED POLARIZATION FEATURE EXTRACTION AND SELF-ORGANIZING MAPPING

Abstract

We propose an unsupervised polarimetric synthetic aperture radar (PolSAR) land classification system consisting of a series of two unsupervised neural networks, namely, a quaternion autoencoder and a quaternion self-organizing map (SOM). Most of the existing PolSAR land classification systems use a set of feature information that humans designed beforehand. However, such methods will face limitations in the near future when we expect classification into a large number of land categories recognizable to humans. By using a quaternion autoencoder, our proposed system extracts feature information based on the natural distribution of PolSAR features. In this paper, we confirm that the information necessary for land classification is extracted as the features while noise is filtered. Then, we show that the extracted features are classified by the quaternion SOM in an unsupervised manner. As a result, we can discover even new and more detailed land categories. For example, town areas are divided into residential areas and factory sites, and grass areas are subcategorized into furrowed farmlands and flat grass areas. We also examine the realization of topographic mapping of the features in the SOM space.

Existing System

Support vector machine (SVM), fuzzy logic and neural networks

Proposed System

The proposed system consists of three functions, that is, preprocess of Poincare-parameter construction, feature extraction, and classification. Poincare-parameter construction is a function to describe more abundant polarization scattering properties than a simple 2 × 2 complex scattering matrix. In this system, Poincare parameters assuming polarized waves in six states are used as a polarization descriptor. Microwave received by the PolSAR system contains not only backscattering information due to the land structure but also various external disturbances. To improve classification performance, it is very important to eliminate unnecessary influences as much as possible to collect main backscattering information. Feature extraction is a function to select the features suitable for land classification from the Poincare parameters to generate a new set of feature vectors. We employ a quaternion autoencoder here. Then, we classify the generated feature vectors into a number of groups based on their distribution in the feature vector space. For this purpose, we use a quaternion SOM as a classifier. The autoencoder and the SOM used for feature extraction and classification, respectively, are widely used unsupervised neural networks. However, we develop a quaternion networks to deal with polarization features in Poincare space. With these two networks, the proposed system can classify the land adaptively without human-predefined categories.

Architecture

BLOCK DIAGRAM