Home // International Journal On Advances in Telecommunications, volume 5, numbers 3 and 4, 2012 // View article

Information Content of Very High Resolution SAR Images: Study of Dependency of SAR Image Structure Descriptors with Incidence Angle

Authors:
Corneliu Octavian Dumitru
Mihai Datcu

Keywords: classes; features; inicdence angle; orbit direction; query; SAR iamges

Abstract:
This paper provide systematic results of the influence of the Synthetic Aperture Radar image structure descriptors with incidence angle and orbit direction. The evaluation is done on TerraSAR-X data and the interpretation is done semi-automatically. In the first part, we study and assess the behavior of the primitive feature extracted methods for images of the same scene with 2 look angles covering the min-max range of the sensor. After that the influence of the orbit looking is shortly discuss. The tests are done on TerraSAR-X products High Resolution Spotlight mode at 3 m resolution and two sites covering the Berlin and Ottawa area are found to be suitable for this investigation. To identify the best features and appropriate incidence angle for them the Support Vector Machine and as a measure of the classification accuracy the precision–recall were considered. The recall shows that the optimal value of the incidence angle in order to have a higher classification is obtained for a value of the incidence angle closer to upper bound of the sensor range. In the second part of the paper a list of queries that can be asked by Earth Observation users are presented and proposed to be implemented in the next generation of our system. The first contribution of this paper is the evaluation of four primitive features that are very known (gray level co-occurrence matrix, Gabor filter, quadrature mirror filter, and non-linear short time Fourier transform) but not used and compared for SAR images. After the best primitive feature is identified the second contribution of this paper lies in the fact that the parameters of the data namely, incidence angle and orbit direction are systematically investigated in order to find the dependency between these parameters and the accuracy of the retrieved classes.

Pages: 239 to 251

Copyright: Copyright (c) to authors, 2012. Used with permission.

Publication date: December 31, 2012

Published in: journal

ISSN: 1942-2601