Faculty Publications

Now showing 1 - 2 of 2

Subtractive clustering and phase correlation similarity measure for endmember extraction
(Elsevier B.V., 2020) Palla, P.Y.; Shetty, A.; Raghavendra, B.S.; Narasimhadhan, A.V.
Target identification using Remote Sensing techniques saves time, cost and reduces difficulties in field investigation. The endmember is a reference spectral response of a pure pixel in the hyperspectral image and is used for object identification/classification from hyperspectral data. Quality of endmembers selected influences classification accuracy. Though there have been several algorithms proposed for endmember extraction, choosing a benchmark algorithm requires further investigation. To the best of our knowledge, similarity measures have not been explored much in the extraction of spectrally distinct signatures called endmembers. In this paper, we propose a similarity measures based subtractive clustering algorithm (SM-SCA) for endmember extraction. The objective of this paper is to explore the applicability of a SM-SCA and effectiveness of different similarity measures in endmember extraction and to compare it's performance with classical endmember extraction algorithms. Implementation on airborne hyperspectral (Samson data and AVIRIS data over Cuprite region) and synthetic data proves that SM-SCA is capable of extracting endmembers of all the materials identified in the data, with appropriate similarity measure. Experimental results show that (i) the similarity measures are potential not only to discriminate but also in extraction of different endmember signatures and (ii) the proposed SM-SCA with phase correlation similarity measure perform comparable to the classical endmember extraction algorithms in identifying endmembers. © 2020 Elsevier B.V.
Knowledge distillation: A novel approach for deep feature selection
(Elsevier B.V., 2023) C, D.; Shetty, A.; Narasimhadhan, A.V.
High dimensional data in hyperspectral remote sensing leads to computational, analytical, and storage complexities. Dimensionality reduction serves as an efficient tool to remove redundant, irrelevant, and highly correlated features. Recently, deep learning approaches have received remarkable progress in hyperspectral data analysis. In this paper, a new end-to-end deep learning framework based on a teacher-student network inspired by knowledge distillation is proposed for deep feature selection. Initially, a complicated teacher deep neural network is employed on complex high dimensional data to learn its corresponding best low dimensional representation. Then, the knowledge from the network is transferred to a simple student network that performs feature selection. Hence, it eventually leads to deep neural network compression which is of prime concern in hyperspectral remote sensing. Limited studies have been carried out to explore the benefits of knowledge distillation on hyperspectral data. The proposed method could be employed to choose deep features for both supervised and unsupervised tasks. Experimental results reveal the performance of the proposed scheme using limited features. In comparison to 1D and simple autoencoder models, the 2D model based on convolutional autoencoder delivers greater classification accuracies, with a classification accuracy value of 96.15% for the Indian Pines dataset and 97.82% for the Pavia University dataset. A similar trend is reported with unsupervised learning as well. Furthermore, the proposed model has a low degree of sensitivity to parameter selection. © 2022 National Authority of Remote Sensing & Space Science

Faculty Publications

Browse

Filters

Settings

Sort By

Results per page

Search Results