Title | Incorporating texture information into polarimetric radar classification using neural networks |
Publication Type | Conference Paper |
Year of Publication | 2004 |
Authors | Ersahin, K., B. Scheuchl, and I. Cumming |
Conference Name | Geoscience and Remote Sensing Symposium, 2004. IGARSS '04. Proceedings. 2004 IEEE International |
Pagination | 563 |
Date Published | sep. |
Keywords | covariance matrices, covariance matrix representation, feature space separability, Fisher criterion, Flevoland data set, geophysical signal processing, GLCP, gray level cooccurrence probability, image texture, multiple channel, neural nets, neural network, observation dimensionality, pixel-based techniques, polarimetric SAR classification, radar polarimetry, self-organising feature maps, self-organizing map, SOM neural network, synthetic aperture radar, texture feature extraction, texture information potential, unsupervised scheme, Wishart classifier |
Abstract | Most of the recent research on polarimetric SAR classification focused on pixel-based techniques using the covariance matrix representation. Since multiple channels are inherently provided in polarimetric data, conventional techniques for increasing the dimensionality of the observation, such as texture feature extraction, were ignored. In this paper, we have demonstrated the potential of texture classification through gray level cooccurrence probabilities (GLCP), and proposed an unsupervised scheme using the self-organizing map (SOM) neural network. The increase in separability of the feature space is shown via the Fisher criterion and also verified by increased classification performance. Compared to the Wishart classifier, promising classification results are obtained from the Flevoland data set. |
URL | http://dx.doi.org/10.1109/IGARSS.2004.1369088 |
DOI | 10.1109/IGARSS.2004.1369088 |