Conference Papers
Permanent URI for this collectionhttps://idr.nitk.ac.in/handle/123456789/28506
Browse
6 results
Search Results
Item Comparison of OMP and SOMP in the reconstruction of compressively sensed hyperspectral images(2011) Aravind, N.V.; Krishnan, K.; Acharya, V.V.; Sumam David, S.In this paper, we present a novel method for the acquisition and compression of hyperspectral images based on two concepts - distributed source coding and compressive sensing. Compressive sensing (CS) is a signal acquisition method that samples at sub Nyquist rates which is possible for signals that are sparse in some transform domain. Distributed source coding (DSC) is a method to encode correlated sources separately and decode them together in an attempt to shift complexity from the encoder to the decoder. Distributed compressive sensing (DCS) is a new framework suggested for jointly sparse signals which we apply to the correlated bands of hyperspectral images. We compressively sense each band of the hyperspectral image individually and can then recover the bands separately or using a joint recovery method. We use the Orthogonal Matching Pursuit (OMP) for individual recovery and Simultaneous Orthogonal Matching Pursuit (SOMP) for joint decoding and compare the two methods. The latter is shown to perform consistently better showing that the Distributed Compressive Sensing method that exploits the joint sparsity of the hyperspectral image is much better than individual recovery. © 2011 IEEE.Item Optimal Band Selection Using Generalized Covering-Based Rough Sets on Hyperspectral Remote Sensing Big Data(Springer Verlag service@springer.de, 2019) Kelam, H.; Venkatesan, M.Hyperspectral remote sensing has been gaining attention from the past few decades. Due to the diverse and high dimensionality nature of the remote sensing data, it is called as remote sensing Big Data. Hyperspectral images have high dimensionality due to number of spectral bands and pixels having continuous spectrum. These images provide us with more details than other images but still, it suffers from ‘curse of dimensionality’. Band selection is the conventional method to reduce the dimensionality and remove the redundant bands. Many methods have been developed in the past years to find the optimal set of bands. Generalized covering-based rough set is an extended method of rough sets in which indiscernibility relations of rough sets are replaced by coverings. Recently, this method is used for attribute reduction in pattern recognition and data mining. In this paper, we will discuss the implementation of covering-based rough sets for optimal band selection of hyperspectral images and compare these results with the existing methods like PCA, SVD and rough sets. © 2019, Springer Nature Singapore Pte Ltd.Item Spatial data-based prediction models for crop yield analysis: A systematic review(Springer, 2020) Mohan, A.; Venkatesan, M.Agriculture plays a vital role in the global economy. WHO states that there are three pillars of food security: availability, access, and usage. Among these three pillars, availability is the most important one. Ensuring food for the entire population of a country is achieved only through an increase in crop production. Accurate and timely forecasting of the weather can help to increase the yield production. Early prediction of crop yield has a vital role in food availability measure. Researchers monitor different parameters that affect the crop yield regularly. Yield prediction did through either statistical data or spatial data. Crop monitoring through remote sensing can cover a vast land area. Therefore, spatial data-based prediction is widespread in recent decades. Satellite images such as multispectral, hyperspectral, and radar images were used to calculate crop area, soil moisture, field greenness, etc. Among these imaging modalities, hyperspectral images give more accurate results, but its higher dimensionality is a challenging issue. Optimal band selection from hyperspectral images helps to reduce this curse of dimensionality problem. Crop area is one of the essential parameters for yield prediction. The exact crop area measure can be achieved only through the best crop discrimination methods. This paper provides a comprehensive review of crop yield prediction using hyperspectral images. Besides, we explore the research challenges and open issues in this area. © Springer Nature Singapore Pte Ltd 2020.Item Spatiospectral feature extraction and classification of hyperspectral images using 3d-cnn + convlstm model(Springer, 2020) Mohan, A.; Venkatesan, M.Hyperspectral images (HSIs) are contiguous bands captured beyond the visible spectrum. The evolution of deep learning techniques places a massive impact on hyperspectral image classification. Curse of dimensionality is one of the significant issues of hyperspectral image analysis. Therefore, most of the existing classification models perform principal component analysis (PCA) as the dimensionality reduction (DR) technique. Since hyperspectral images are nonlinear, linear DR techniques fail to reserve the nonlinear features. The usage of both spatial and spectral features together improves the classification accuracy of the model. 3D-convolutional neural networks (CNN) extract the spatiospectral features for classification, whereas it is not considering the dependencies in features. This research work proposes a new model for HSI classification using 3D-CNN and convolutional long short-term memory (ConvLSTM). The optimal band extraction is performed by a hybrid DR technique, which is the combination of Gaussian random projection (GRP) and Kernel PCA (KPCA). The proposed deep learning model extracts spatiospectral features using 3D-CNN and dependent spatial features using 2D-ConvLSTM in parallel. Combination of extracted features is fed into a fully connected network for classification. The experiment is performed on three widely used datasets, and the proposed model is compared against the various state-of-the-art techniques and found better classification accuracy. © Springer Nature Singapore Pte Ltd 2020.Item Hybrid dimensionality reduction technique for hyperspectral images using random projection and manifold learning(Springer Science and Business Media Deutschland GmbH info@springer-sbm.com, 2020) Mohan, A.; Venkatesan, M.Hyperspectral images (HSI) are contiguous band images having hundreds of bands. However, most of the bands are redundant and irrelevant. Curse of dimensionality is a significant problem in hyperspectral image analysis. The band extraction technique is one of the dimensionality reduction (DR) method applicable in HSI. Linear dimensionality reduction techniques fail for hyperspectral images due to its nonlinearity nature. Nonlinear reduction techniques are computationally complex. Therefore this paper introduces a hybrid dimensionality reduction technique for band extraction in hyperspectral images. It is a combination of linear random projection (RP) and nonlinear technique. The random projection method reduces the dimensionality of hyperspectral images linearly using either Gaussian or Sparse distribution matrix. Sparse random projection (SRP) is computationally less complex. This reduced image is fed into a nonlinear technique and performs band extraction in minimal computational time and maximum classification accuracy. For experimental analysis of the proposed method, the hybrid technique is compared with Kernel PCA (KPCA) using different random matrix and found a promising improvement in results for their hybrid models in minimum computation time than classic nonlinear technique. © Springer Nature Switzerland AG 2020.Item Salient Object Detection in Hyperspectral Images Using Felzenswalb’s Segmentation Algorithm(Springer Science and Business Media Deutschland GmbH, 2024) Lone, Z.A.; Pais, A.R.Salient object detection has been explored extensively in low dimensional images like RGB, grayscale, etc., but have been explored very little in high dimensional images like Hyperspectral images (HSI) etc. In HSI, few studies have used low-level features to perform salient object detection. In this paper, we propose a high-level feature-based salient object detection algorithm. The manifold ranking is applied on the self-supervised CNN features learned by an unsupervised segmentation task. The training of the model continues until the clustering loss or saliency map converges to a defined error. We found out that the proposed algorithm performed better than state-of-the-art in terms of precision. © Springer Nature Switzerland AG 2024.