Complex Valued Convolutional Neural Network for TerraSAR-X patch categorization

Conference: EUSAR 2018 - 12th European Conference on Synthetic Aperture Radar
06/04/2018 - 06/07/2018 at Aachen, Germany

Proceedings: EUSAR 2018

Pages: 4Language: englishTyp: PDF

Personal VDE Members are entitled to a 10% discount on this title

Gleich, Dusan; Sipos, Danijel (University of Maribor, Faculty of Electrical Engineering and Computer Science, Slovenia)

Over the past few years image categorization using deep learning became very popular, because it can handle large databases and has shown good recognition results. This paper presents the complex valued convolutional network (CV-CNN) for Synthetic Aperture Radar (SAR) patch categorization of TerraSAR-X data sets. The CV-CNN consist in general of a real or complex valued input layer, output layer and one or more hidden layers. Hidden layers represent any combination of convolutional layers, pooling layers, activation functions, and are fully defined within complex valued domain. The custom database of patches was designed using 3 classes and parameters of CV-CNN were observed in order to achieve the best accuracy results. The experimental results showed that accuracy of 86% can be achieved using a complex valued SAR patches applied to the complex valued CNN.