PROJECT TITLE :

POL-SAR Image Classification Based on Wishart DBN and Local Spatial Information

ABSTRACT:

Inspired by a common deep neural network, i.e., deep belief network (DBN), a novel method for polarimetric artificial aperture radar (POL-SAR) image classification is proposed during this paper. For the particularity of POL-SAR information, a new type of restricted Boltzmann machine (RBM) is specially outlined, which we tend to name the Wishart–Bernoulli RBM (WBRBM), and is employed to create a deep network named as Wishart DBN (W-DBN). Various unlabeled POL-SAR pixels are made full use of in the modeling of POL-SAR pixels by W-DBN. As well, the coherency matrix is used directly to represent a POL-SAR pixel without any manual Feature Extraction, which is straightforward and time saving. Local spatial data, along with the confusion matrix, is utilized in this paper to scrub the preliminary classification result obtained by the strategy based on W-DBN. Creating full use of the prior information of POL-SAR knowledge and local spatial information, the proposed technique overcomes shortcomings of ancient ways, in which they're sensitive to extracted features and slow to execute. The experiments, tested on 3 POL-SAR information sets, show that the proposed technique produces better results and is much faster than ancient ways.