Impact of Dimensionality Reduction Techniques on Endmember Identification in Hyperspectral Imagery
The image derived hyperspectral endmembers fulfil the requirement of ground truth information required for supervised classification and spectral unmixing of the hyperspectral scenes. In hyperspectral images endmembers are very difficult to identify by visual inspection, since their population within the scene is significantly low. There are several algorithms (PPI, ATGP, NFINDR, CCA, VCA and SGA) developed for endmember identification. Most of these algorithms begins with MNF transformation based data dimensionality reduction, even though there other dimensionality reduction algorithms are available in the literature. This paper critically evaluates the comparative performances of NFINDR and ATGP endmember finding algorithms using original hyperspectral scene and PCA, MNF and ICA transformed data sets separately. The experimental outcomes are evaluated using two important parameters. First parameters compares the execution time of EM identification algorithms. Second parameter compares the spatial coordinates and their corresponding spectral signatures of NFINDR and ATGP identified endmembers. The comparative experimental analysis showcase that the execution time of NFINDR and ATGP algorithms is significantly improved with ICA transformed principle components.
KeywordsHyperspectral image Dimensionality reduction Endmember extraction PCA MNF ICA NFINDR ATGP
The Authors acknowledge to DST, GOI, for financial support under major research project (No. BDID/01/23/2014-HSRS/35 (ALG-V)) and for providing AVIRIS-NG data. The authors also extend sincere thanks to UGC SAP for providing lab facilities to the Department of Comp. Science and IT, Dr. B. A. M. University, Aurangabad-(MS), India.
- 2.Nandibewoor, A., Hegadi, R.: A novel SMLR-PSO model to estimate the chlorophyll content in the crops using hyperspectral satellite images. Cluster Comput., 1–8 (2018)Google Scholar
- 3.Solankar, M.M., Gite, H.R., Dhumal, R.K., Surase, R.R., Nalawade, D., Kale, K.V.: Recent advances and challenges in automatic hyperspectral endmember extraction. In: Krishna, C.R., Dutta, M., Kumar, R. (eds.) Proceedings of 2nd International Conference on Communication, Computing and Networking. LNNS, vol. 46, pp. 445–455. Springer, Singapore (2019). https://doi.org/10.1007/978-981-13-1217-5_44CrossRefGoogle Scholar
- 4.Boardman, J.W., Kruse, F.A., Green, R.O.: Mapping target signatures via partial unmixing of AVIRIS data (1995)Google Scholar
- 6.Winter, M.E.: N-FINDR: an algorithm for fast autonomous spectral end-member determination in hyperspectral data. In: SPIE’s International Symposium on Optical Science, Engineering, and Instrumentation. International Society for Optics and Photonics, pp. 266–275 (1999)Google Scholar
- 11.Phillips, R.D., Watson, L.T., Blinn, C.E., Wynne, R.H.: An adaptive noise reduction technique for improving the utility of hyperspectral data. In: Proceedings of the 17th William T. Pecora Memorial Remote Sensing Symposium, pp. 16–20 (2008)Google Scholar
- 12.Rodarmel, C., Shan, J.: Principal component analysis for hyperspectral image classification. Surv. Land Inf. Sci. 62(2), 115 (2002)Google Scholar
- 15.Green, R., Landeen, S., McCubbin, I., Thompson, D., Bue, B.: Airborne visible/infrared imaging spectrometer next generation (AVIRIS-NG), 1st edn. [PDF] JPL, California Institute of Technology (2015). http://vedas.sac.gov.in:8080/aviris/pdf/20150726_AVRISINGDataGuide_v4.pdf. Accessed 1 July 2017