Classification of Multi-Frequency Polarimetric SAR Images Based on Multi-Linear Subspace Learning of Tensor Objects

摘要: One key problem for the classification of multi-frequency polarimetric SAR images is to extract target features simultaneously in aspects frequency, polarization and spatial texture. This paper proposes a new method data based on tensor representation multi-linear subspace learning (MLS). Firstly, each cell represented by third-order domains, with order corresponding one domain. Then, two main MLS methods, i.e., principal component analysis (MPCA) extension linear discriminant (MLDA), are used learn tensors. MPCA analyze MLDA applied improve discrimination between different land covers. Finally, lower dimension subtensor extracted algorithms classified neural network (NN) classifier. The scheme accessed using multi-band (C-, L- P-band) acquired Airborne Synthetic Aperture Radar (AIRSAR) sensor Jet Propulsion Laboratory (JPL) over Flevoland area. Experimental results demonstrate that proposed has good performance comparison classic Wishart overall accuracy close 99%, even when number training samples small.