DIMENSIONALITY REDUCTION VIA AN ORTHOGONAL AUTOENCODER APPROACH FOR HYPERSPECTRAL IMAGE CLASSIFICATION

Abstract. Nowadays, the increasing amount of information provided by hyperspectral sensors requires optimal solutions to ease the subsequent analysis of the produced data. A common issue in this matter relates to the hyperspectral data representation for classification tasks. Existing approaches address the data representation problem by performing a dimensionality reduction over the original data. However, mining complementary features that reduce the redundancy from the multiple levels of hyperspectral images remains challenging. Thus, exploiting the representation power of neural networks based techniques becomes an attractive alternative in this matter. In this work, we propose a novel dimensionality reduction implementation for hyperspectral imaging based on autoencoders, ensuring the orthogonality among features to reduce the redundancy in hyperspectral data. The experiments conducted on the Pavia University, the Kennedy Space Center, and Botswana hyperspectral datasets evidence such representation power of our approach, leading to better classification performances compared to traditional hyperspectral dimensionality reduction algorithms.

Download Full-text

Dimensionality Reduction and Classification of Hyperspectral Images using Genetic Algorithm

Indonesian Journal of Electrical Engineering and Computer Science ◽

10.11591/ijeecs.v3.i3.pp503-511 ◽

2016 ◽

Vol 3 (3) ◽

pp. 503 ◽

Cited By ~ 2

Author(s):

R. Kiran Kumar ◽

B. Saichandana ◽

K. Srinivas

Keyword(s):

Genetic Algorithm ◽

Dimensionality Reduction ◽

Hyperspectral Image ◽

Image Data ◽

Unsupervised Classification ◽

Hyperspectral Data ◽

Band Selection ◽

Data Set ◽

Mode Decomposition ◽

Fused Image

This paper presents genetic algorithm based band selection and classification on hyperspectral image data set. Hyperspectral remote sensors collect image data for a large number of narrow, adjacent spectral bands. Every pixel in hyperspectral image involves a continuous spectrum that is used to classify the objects with great detail and precision. In this paper, first filtering based on 2-D Empirical mode decomposition method is used to remove any noisy components in each band of the hyperspectral data. After filtering, band selection is done using genetic algorithm in-order to remove bands that convey less information. This dimensionality reduction minimizes many requirements such as storage space, computational load, communication bandwidth etc which is imposed on the unsupervised classification algorithms. Next image fusion is performed on the selected hyperspectral bands to selectively merge the maximum possible features from the selected images to form a single image. This fused image is classified using genetic algorithm. Three different indices, such as K-means Index (KMI) and Jm measure are used as objective functions. This method increases classification accuracy and performance of hyperspectral image than without dimensionality reduction.

Download Full-text

Dimensionality reduction and hierarchical clustering in framework for hyperspectral image segmentation

Bulletin of Electrical Engineering and Informatics ◽

10.11591/eei.v8i3.1451 ◽

2019 ◽

Vol 8 (3) ◽

pp. 1081-1087

Author(s):

K. Mallikharjuna Rao ◽

B. Srinivasa Rao ◽

B. Sai Chandana ◽

J. Harikiran

Keyword(s):

Dimensionality Reduction ◽

Hierarchical Clustering ◽

Clustering Algorithm ◽

Hyperspectral Image ◽

Hyperspectral Data ◽

Multispectral Images ◽

Pixel Intensity ◽

Fused Image ◽

Histogram Similarity ◽

Spectral Channels

The hyperspectral data contains hundreds of narrows bands representing the same scene on earth, with each pixel has a continuous reflectance spectrum. The first attempts to analysehyperspectral images were based on techniques that were developed for multispectral images by randomly selecting few spectral channels, usually less than seven. This random selection of bands degrades the performance of segmentation algorithm on hyperspectraldatain terms of accuracies. In this paper, a new framework is designed for the analysis of hyperspectral image by taking the information from all the data channels with dimensionality reduction method using subset selection and hierarchical clustering. A methodology based on subset construction is used for selecting k informative bands from d bands dataset. In this selection, similarity metrics such as Average Pixel Intensity [API], Histogram Similarity [HS], Mutual Information [MI] and Correlation Similarity [CS] are used to create k distinct subsets and from each subset, a single band is selected. The informative bands which are selected are merged into a single image using hierarchical fusion technique. After getting fused image, Hierarchical clustering algorithm is used for segmentation of image. The qualitative and quantitative analysis shows that CS similarity metric in dimensionality reduction algorithm gets high quality segmented image.

Download Full-text

Robust Graph Dimensionality Reduction

Proceedings of the Twenty-Seventh International Joint Conference on Artificial Intelligence ◽

10.24963/ijcai.2018/452 ◽

2018 ◽

Cited By ~ 1

Author(s):

Xiaofeng Zhu ◽

Cong Lei ◽

Hao Yu ◽

Yonggang Li ◽

Jiangzhang Gan ◽

...

Keyword(s):

Dimensionality Reduction ◽

Optimization Algorithm ◽

Graph Embedding ◽

Original Data ◽

Transformation Matrix ◽

High Dimensional ◽

Local Similarity ◽

Robust Estimators ◽

Classification Tasks ◽

Low Dimensional

In this paper, we propose conducting Robust Graph Dimensionality Reduction (RGDR) by learning a transformation matrix to map original high-dimensional data into their low-dimensional intrinsic space without the influence of outliers. To do this, we propose simultaneously 1) adaptively learning three variables, \ie a reverse graph embedding of original data, a transformation matrix, and a graph matrix preserving the local similarity of original data in their low-dimensional intrinsic space; and 2) employing robust estimators to avoid outliers involving the processes of optimizing these three matrices. As a result, original data are cleaned by two strategies, \ie a prediction of original data based on three resulting variables and robust estimators, so that the transformation matrix can be learnt from accurately estimated intrinsic space with the helping of the reverse graph embedding and the graph matrix. Moreover, we propose a new optimization algorithm to the resulting objective function as well as theoretically prove the convergence of our optimization algorithm. Experimental results indicated that our proposed method outperformed all the comparison methods in terms of different classification tasks.

Download Full-text

Hyperspectral Dimensionality Reduction Based on Inter-Band Redundancy Analysis and Greedy Spectral Selection

Remote Sensing ◽

10.3390/rs13183649 ◽

2021 ◽

Vol 13 (18) ◽

pp. 3649

Author(s):

Giorgio Morales ◽

John W. Sheppard ◽

Riley D. Logan ◽

Joseph A. Shaw

Keyword(s):

Feature Extraction ◽

Dimensionality Reduction ◽

Redundancy Analysis ◽

Hyperspectral Image ◽

Hyperspectral Data ◽

Data Cube ◽

Computational Time ◽

Redundant Data ◽

Spectral Bands ◽

Spectral Selection

Hyperspectral imaging systems are becoming widely used due to their increasing accessibility and their ability to provide detailed spectral responses based on hundreds of spectral bands. However, the resulting hyperspectral images (HSIs) come at the cost of increased storage requirements, increased computational time to process, and highly redundant data. Thus, dimensionality reduction techniques are necessary to decrease the number of spectral bands while retaining the most useful information. Our contribution is two-fold: First, we propose a filter-based method called interband redundancy analysis (IBRA) based on a collinearity analysis between a band and its neighbors. This analysis helps to remove redundant bands and dramatically reduces the search space. Second, we apply a wrapper-based approach called greedy spectral selection (GSS) to the results of IBRA to select bands based on their information entropy values and train a compact convolutional neural network to evaluate the performance of the current selection. We also propose a feature extraction framework that consists of two main steps: first, it reduces the total number of bands using IBRA; then, it can use any feature extraction method to obtain the desired number of feature channels. We present classification results obtained from our methods and compare them to other dimensionality reduction methods on three hyperspectral image datasets. Additionally, we used the original hyperspectral data cube to simulate the process of using actual filters in a multispectral imager.

Download Full-text

NON-LINEAR AUTOENCODER BASED ALGORITHM FOR DIMENSIONALITY REDUCTION OF AIRBORNE HYPERSPECTRAL DATA

ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ◽

10.5194/isprs-archives-xlii-3-w6-593-2019 ◽

2019 ◽

Vol XLII-3/W6 ◽

pp. 593-598

Author(s):

S. Priya ◽

R. Ghosh ◽

B. K. Bhattacharya

Keyword(s):

Remote Sensing ◽

Dimensionality Reduction ◽

Infrared Imaging ◽

Hyperspectral Image ◽

Signal To Noise Ratio ◽

Hyperspectral Data ◽

Correlation Technique ◽

Spectral Correlation ◽

Spectral Angle Mapper ◽

Non Linear

Abstract. Hyperspectral remote sensing is an advanced remote sensing technology that enhances the ability of accurate classification due to presence of narrow contiguous bands. The large number of continuous bands present in hyperspectral data introduces the problem of computational complexity due to presence of redundant information. There is a need for dimensionality reduction to enhance the ability of users for better characterization of features. Due to presence of high spectral correlation in the hyperspectral datasets, optimum de-correlation technique is required which transforms the hyperspectral data to lower dimensions without compromising with the desirable information present in the data. In this paper, focus has been to reduce the spectral dimensionality problem. So, this research aimed to develop computationally efficient non-linear autoencoder algorithm taking the advantage of non-linear properties of hyperspectral data. The proposed algorithm was applied on airborne hyperspectral image of Airborne Visible Infrared Imaging Spectrometer - Next Generation (AVIRIS-NG) over Anand region of Gujarat and the performance of the algorithm was evaluated. The Signal-to-Noise Ratio (SNR) increased from 22.78 dB to 48.48 dB with increase in number of nodes in bottleneck layer for reconstruction of image. Spectral distortion was also measured using Spectral Angle Mapper Algorithm (SAM), which reduced from 0.38 to 0.05 with increase in number of nodes in bottleneck layer up to 10. So, this algorithm was able to give good reconstruction of original image from the nodes present in the bottleneck layer.

Download Full-text

Dimensionality Reduction using Bi-Dimensional Empirical Mode Decomposition Method for Hyperspectral Image Segmentation

International Journal of Recent Technology and Engineering - 2 ◽

10.35940/ijrte.d9566.118419 ◽

2019 ◽

Vol 8 (4) ◽

pp. 11300-11304

Keyword(s):

Dimensionality Reduction ◽

Empirical Mode Decomposition ◽

Spatial Information ◽

Hyperspectral Image ◽

Clustering Algorithms ◽

Hyperspectral Data ◽

Processing Load ◽

Data Set ◽

Mode Decomposition ◽

Empirical Mode Decomposition Method

This paper presents a dimensionality reduction of hyperspectral dataset using bi-dimensional empirical mode decomposition (BEMD). This reduction method is used in a process for segmentation of hyperspectral data. Hyperspectral data contains multiple narrow bands conveying both spectral and spatial information of a scene. Analysis of this kind of data is done in three sequential stages, dimensionality reduction, fusion and segmentation. The method presented in this paper mainly focus on the dimensionality reduction step using BEMD, fusion is carried out using hierarchical fusion method and the segmentation is carried out using Clustering algorithms. This dimensionality reduction removes less informative bands in the data set, decreasing the storage and processing load in further steps in analysis of data. The qualitative and quantitative analysis shows that best informative bands are selected using proposed method which gets high quality segmented image using FCM.

Download Full-text

Locality Preserving Projection Based on Endmember Extraction for Hyperspectral Image Dimensionality Reduction and Target Detection

Applied Spectroscopy ◽

10.1177/0003702816665992 ◽

2016 ◽

Vol 70 (9) ◽

pp. 1573-1581 ◽

Cited By ~ 4

Author(s):

Yiting Wang ◽

Shiqi Huang ◽

Zhigang Liu ◽

Hongxia Wang ◽

Daizhi Liu

Keyword(s):

Dimensionality Reduction ◽

Target Detection ◽

Hyperspectral Image ◽

Image Data ◽

Hyperspectral Data ◽

Detection Accuracy ◽

Endmember Extraction ◽

Locality Preserving Projection ◽

Locality Preserving ◽

Weighted Matrix

In order to reduce the effect of spectral variability on calculation precision for the weighted matrix in the locality preserving projection (LPP) algorithm, an improved dimensionality reduction method named endmember extraction-based locality preserving projection (EE-LPP) is proposed in this paper. The method primarily uses the vertex component analysis (VCA) method to extract endmember spectra from hyperspectral imagery. It then calculates the similarity between pixel spectra and the endmember spectra by using the spectral angle distance, and uses it as the basis for selecting neighboring pixels in the image and constructs a weighted matrix between pixels. Finally, based on the weighted matrix, the idea of the LPP algorithm is applied to reduce the dimensions of hyperspectral image data. Experimental results of real hyperspectral data demonstrate that the low-dimensional features acquired by the proposed methods can fully reflect the characteristics of the original image and further improve target detection accuracy.

Download Full-text

Comparison and Evaluation of Dimensionality Reduction Techniques for Hyperspectral Data Analysis

Proceedings ◽

10.3390/iecg2019-06209 ◽

2019 ◽

Vol 24 (1) ◽

pp. 6 ◽

Cited By ~ 1

Author(s):

K Nivedita Priyadarshini ◽

V Sivashankari ◽

Sulochana Shekhar ◽

K Balasubramani

Keyword(s):

Dimensionality Reduction ◽

Infrared Imaging ◽

Principal Component ◽

Original Data ◽

Hyperspectral Data ◽

Imaging Spectrometer ◽

Reduction Techniques ◽

Spectral Bands ◽

Dimensionality Reduction Techniques ◽

Highly Correlated

Hyperspectral datasets provide explicit ground covers with hundreds of bands. Filtering contiguous hyperspectral datasets potentially discriminates surface features. Therefore, in this study, a number of spectral bands are minimized without losing original information through a process known as dimensionality reduction (DR). Redundant bands portray the fact that neighboring bands are highly correlated, sharing similar information. The benefits of utilizing dimensionality reduction include the ability to slacken the complexity of data during processing and transform original data to remove the correlation among bands. In this paper, two DR methods, principal component analysis (PCA) and minimum noise fraction (MNF), are applied to the Airborne Visible/Infrared Imaging Spectrometer-Next Generation (AVIRIS-NG) dataset of Kalaburagi for discussion.

Download Full-text