|
|
|
Abstract :
Hyperspectral images provide abundant information about objects. The high dimensionality of such images causes various problems like curse of dimensionality and large hypothesis space.
There are two methods to achieve high dimensionality problem, band selection and feature extraction. In this paper we present a feature extraction method based on angular criterion, this method define as minimize angle between mean vector and samples with in each class and maximize the angle of between mean classes. This method explores other aspect of pattern in feature space and tries to discriminate classes with respect of geometric parameters. Angular criterion method has been employed for feature extraction and spectral angle mapper (SAM) classifier for classification. The results demonstrate that this method can improve the discrimination of objects in feature space and improve classification accuracy of SAM classifier
|
|
|