A SVMS-based hyperspectral data classification algorithm in a similarity space

• Published in: 2009 First Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing pp. 1 - 4
• Main Authors: Hosseini, R.S., Homayouni, S.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2009
• Subjects: Classification algorithms; Data engineering; Educational institutions; Hyperspectral data; Hyperspectral imaging; Hyperspectral sensors; Remote sensing; Similarity Space; Spatial resolution; Supervised Classification; Support vector machine classification; Support vector machines; Training data
• ISBN: 9781424446865; 1424446864
• ISSN: 2158-6268
• DOI: 10.1109/WHISPERS.2009.5288980

In this paper, a multi-steps algorithm based on support vectors machines (SVMs) in similarity space is proposed. The SVMs is used as a recent classification method and separation boundary estimation technique for high dimensional data. It benefits of limited number of data for training of supervised classification, which is a key challenge in hyperspectral data analysis. SVMs based classifier is applied in a similarity space. This space can be seen as a new feature space with relatively low dimension. In other words, a similarity projection is used to reduce the number of spectral bands in a sagacious manner. In deed, the hyperspectral data is projected to the similarity space of a specific class-of-interest using the spectral similarity measures such as spectral angle, distance, etc. This algorithm was applied to two sets of remotely sensed data; first is a Hyperion imagery set contains 242 bands with 30 m of spatial resolution. The second is a CASI (compact airborne spectrographic imager) imagery set having 9 spectral bands at 4 m of spatial resolution. Both of image sets cover the natural land areas. The trusty ground truth data are available for these images. So, the evaluation study is done to assess the accuracy of classification and role of different parameter setting and similarity measures. The results demonstrate the efficiency and the reliability of this algorithm.