Abstract
Abstract. In this study, correlation clustering is introduced to hyperspectral imagery for unsupervised classification. The main advantage of correlation clustering lies in its ability to simultaneously perform feature reduction and clustering. This algorithm also allows selection of different sets of features for different clusters. This framework provides an effective way to address the issues associated with the high dimensionality of the data. ORCLUS, a correlation clustering algorithm, is implemented and enhanced by making use of segmented principal component analysis (SPCA) instead of principal component analysis (PCA). Further, original implementation of ORCLUS makes use of eigenvectors corresponding to smallest eigenvalues whereas in this study eigenvectors corresponding to maximum eigenvalues are used, as traditionally done when PCA is used as feature reduction tool. Experiments are conducted on three real hyperspectral images. Preliminary analysis of algorithms on real hyperspectral imagery shows ORCLUS is able to produce acceptable results.
Highlights
Hyperspectral imaging has become an important tool for information extraction, especially in the remote sensing community (Villa et al, 2013)
Generally supervised classification methods are more successful in providing higher classification accuracy as compared to unsupervised methods (Paoli et al, 2009), in reality, collection of high quality training sample is very expensive and time consuming procedure
The objective of this study is to investigate the performance of one correlation clustering approach for high dimensional data on hyperspectral imagery
Summary
Hyperspectral imaging has become an important tool for information extraction, especially in the remote sensing community (Villa et al, 2013). Hyperspectral imagery, provides detailed spectral information but it leads to certain challenges It has very high dimensionality which gives rise to several problems, collectively addressed as “curse of dimensionality”. In clustering four major problems emerge (Kriegel et al, 2009) These are – (i) in high dimensionality, concept of distance or neighbourhood becomes less meaningful (Beyer et al, 1999), (ii) for a pixel, among various observed dimensions/bands, some of the dimensions will be irrelevant, which in turn will affect the distance computation, (iii) subset of some dimensions/bands may be relevant to one cluster and subset of some other dimensions may be relevant to other cluster, and so on. It may be difficult for global feature reduction methods (e.g. principal component analysis) to identify one common subspace in which all the cluster will be discernible, and (iv) for high dimensional data many dimensions may be correlated
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
