Hyperspectral Dimensionality Reduction Research Articles

Hyperspectral dimensionality reduction based on SAE-1DCNN feature selection approach

Hyperspectral dimensionality reduction based on SAE-1DCNN feature selection approach

Spectral characteristics analysis of typical ground objects in Poyang Lake wetland based on “Zhuhai-1” hyperspectral data

Hyperspectral image data provides new ideas and methods for the fine classification of wetland. Research on wetland spectral characteristics lays the foundation for wetland classification and improves classification accuracy. Based on “Zhuhai-1” hyperspectral image data, combined with the Pure Pixel Index (PPI), this paper extracted the average spectral curves of 11 typical ground objects in Poyang Lake wetland. Two mathematical transformation methods, namely, the continuum removal and the first derivative, were used to analyze the spectral characteristics of the ground objects before and after transformation, in combination with spectral characteristic parameters. The method of combining the mean distance of ground objects with the range of data error was used to select characteristic wavelengths before and after spectral transformation. Finally, the Mahalanobis distance method was used to verify the recognition effect. The results demonstrate that: (1) Spectral transformation can effectively amplify the differences between ground objects. The continuum removal led to an increase in the characteristic bands of four types of vegetation and the emergence of characteristic bands of four types of other ground objects. The first derivative method extended the characteristic bands of four vegetation from red-edge to near-infrared. (2) The method of combining the mean distance and the data error range can effectively screened out the spectral characteristic bands of typical ground objects in Poyang Lake wetland, resulting in hyperspectral dimensionality reduction. (3) The Mahalanobis distance method validates the recognition effect of the characteristic bands. In the screened characteristic bands, the Mahalanobis distance between different ground objects is larger than that between the same ground object, except between shoaly land and road. This study offers valuable insights into spectral recognition and classification of freshwater lake wetlands, and also demonstrates the potential of “Zhuhai-1” hyperspectral image data in wetland research.

Two New Indices for Unsupervised Dimensionality Reduction of Hyperspectral Data

Hyperspectral imaging, with hundreds of spectral bands provide a rich data source for thematic mapping. However, the high dimensionality of these images is a challenge which is commonly treated by dimensionality reduction techniques. In this paper, a pixel-based hyperspectral dimensionality reduction technique is proposed in which the spectral signature curve (SSC) is divided into segments with equal lengths. Afterwards two new mathematical indices, which are based on integral and L2-norm, are proposed to transform each SSC segment into a new feature of the reduced feature vector. The proposed method considers the ordinance of SSC and, unlike PCA, can be applied individually to all pixels independently and simultaneously which means the method is applicable in parallel processing. The proposed data reduction technique was applied to two well-known agricultural hyperspectral scenes and was compared to wavelet and PCA data reduction techniques. The obtained results proved the efficiency of the proposed method from both classification accuracy and processing time aspects.

Dimensionality reduction strategies for land use land cover classification based on airborne hyperspectral imagery: a survey.

Hyperspectral image (HSI) contains hundreds of adjacent spectral bands, which can effectively differentiate the region of interest. Nevertheless, many irrelevant and highly correlated spectral bands lead to the Hughes phenomenon. Consequently, hyperspectral image dimensionality reduction is necessary to select the most informative and significant spectral band and eliminate the redundant spectral band. To this end, this paper represents an extensive and systematic survey of hyperspectral dimensionality reduction approaches for land use land cover (LULC) classification. Moreover, this paper reviewed the following important points: (1) hyperspectral imaging data acquisition methods, (2) the difference between hyperspectral and multispectral images, (3) hyperspectral image dimensionality reduction based on machine learning (ML) and deep learning (DL) techniques, (4) the popular benchmark hyperspectral datasets with the performance metrics for LULC classification, and (5) the significant challenges with the future trends for hyperspectral dimensionality reduction.

Band selection for heterogeneity classification of hyperspectral transmission images based on multi-criteria ranking

Hyperspectral images contain hundreds of continuous spectral bands, which can provide rich information for target detection and image classification. However, high-dimensional data are easy to cause the “Hughes phenomenon”, which affects the classification accuracy of images. In this paper, a band selection method based on multi-criteria ranking is proposed for hyperspectral dimensionality reduction. First, the hyperspectral image acquisition system is built to collect the hyperspectral transmission image datasets containing various heterogeneities. Second, the original data are preprocessed and the band selection method proposed is used for data dimensionality reduction. That is, (1) the whole band is divided into subintervals according to the correlation criteria, and then the bands with large information amount and good inter-class separability are selected according to other criteria functions, (2) the bands of each subinterval are combined, and the combined bands are evaluated according to the optimum index factor and Bhattacharyya distance, and (3) the combined bands are ranked and the top-ranked combined bands are selected as the representative bands for image classification. Finally, the support vector machine is used to verify the effectiveness of the representative bands selected by this method. The experimental results show that the proposed method achieves data dimensionality reduction, and the overall classification accuracy can reach 100% on some selected subsets of bands, which has better classification performance than other related band selection methods. In addition, the computational complexity of high-dimensional data can be effectively reduced by band selection, and the average running time is 95.19% less than that of the full bands while ensuring classification accuracy.

Hyper-Embedder: Learning a Deep Embedder for Self-Supervised Hyperspectral Dimensionality Reduction

Hyperspectral imaging has attracted growing interest among the researchers from the geoscience and remote sensing fields owing to its very rich spectral information. However, the high spectral dimensionality of hyperspectral images tends to suffer from information redundancy. Manifold embedding is a mainstream strategy of nonlinear hyperspectral dimensionality reduction. The sensitivity to the noise and the inflexibility to the out-of-sample problem (i.e., new samples) are the main drawbacks of the manifold embedding based methods. To this end, we propose to learn a deep embedder in a self-supervised fashion for hyperspectral dimensionality reduction, called Hyper-Embedder. Hyper-Embedder effectively reduces the computational complexity and storage-costing compared to conventional embedding models, and improves the robustness against various noises, e.g., spectral variabilities. More significantly, Hyper-Embedder is capable of learning an explicit nonlinear mapping to make a one-to-one match between each original pixel (spectral signature) in the hyperspectral image and its dimension-reduced representation. These low-dimensional representations can be generated and given by existing and classic nonlinear manifold embedding methods. In this paper, we attempt to learn the correspondence or mapping by optimizing a deep regression network. The to-be-developed network can not only capture the local topological knowledge graph of all spectral signatures of hyperspectral data but be applicable to fast prediction and inference of samples from other hyperspectral scenes. The proposed Hyper-Embedder outperforms existing state-of-the-art hyperspectral dimensionality reduction algorithms on two commonly-used hyperspectral datasets, i.e., Indian Pines and Augsburg scenes.

SSMD: Dimensionality Reduction and Classification of Hyperspectral Images Based on Spatial–Spectral Manifold Distance Metric Learning

Metric learning, which aims to obtain a metric matrix M such that samples from the same class are close to one another and samples of different classes are far from one another, is widely used in the field of hyperspectral dimensionality reduction (DR) and classification. Traditional metric learning is based on the Mahalanobis distance, which measures the similarity between samples via point-to-point distance, ignoring the structural features of the hyperspectral images (HSIs). To solve the above problem, we proposed clustered multiple manifold metric learning (CM3L), which obtains a manifold distance (MD), aimed at improving discrimination by introducing structural features of the HSIs and achieving good results. However, this manifold distance still has certain shortcomings in specific application situations. MD only considers the labeled data in the construction of the manifold and ignores the unlabeled data, resulting in the destruction of the manifold. Therefore, this article proposes a new spatial–spectral manifold distance (SSMD) to improve the performance of metric learning in hyperspectral DR and classification by maintaining the integrity of the constructed manifolds. The SSMD selects suitable neighboring points in the labeled and unlabeled data through the spectral–spatial information in order to participate in the construction of the manifold. Then, the distance between the manifolds is calculated to replace the traditional Mahalanobis distance. The results of seven sets of comparison experiments on three real HSI datasets demonstrate the effectiveness of SSMD in improving the classification results of HSIs.

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

Joint and Progressive Subspace Analysis (JPSA) With Spatial-Spectral Manifold Alignment for Semisupervised Hyperspectral Dimensionality Reduction.

Conventional nonlinear subspace learning techniques (e.g., manifold learning) usually introduce some drawbacks in explainability (explicit mapping) and cost effectiveness (linearization), generalization capability (out-of-sample), and representability (spatial-spectral discrimination). To overcome these shortcomings, a novel linearized subspace analysis technique with spatial-spectral manifold alignment is developed for a semisupervised hyperspectral dimensionality reduction (HDR), called joint and progressive subspace analysis (JPSA). The JPSA learns a high-level, semantically meaningful, joint spatial-spectral feature representation from hyperspectral (HS) data by: 1) jointly learning latent subspaces and a linear classifier to find an effective projection direction favorable for classification; 2) progressively searching several intermediate states of subspaces to approach an optimal mapping from the original space to a potential more discriminative subspace; and 3) spatially and spectrally aligning a manifold structure in each learned latent subspace in order to preserve the same or similar topological property between the compressed data and the original data. A simple but effective classifier, that is, nearest neighbor (NN), is explored as a potential application for validating the algorithm performance of different HDR approaches. Extensive experiments are conducted to demonstrate the superiority and effectiveness of the proposed JPSA on two widely used HS datasets: 1) Indian Pines (92.98%) and 2) the University of Houston (86.09%) in comparison with previous state-of-the-art HDR methods. The demo of this basic work (i.e., ECCV2018) is openly available at https://github.com/danfenghong/ECCV2018_J-Play.

A Study of Nitrogen Deficiency Inversion in Rice Leaves Based on the Hyperspectral Reflectance Differential.

To achieve rapid, accurate, and non-destructive diagnoses of nitrogen deficiency in cold land japonica rice, hyperspectral data were collected from field experiments to investigate the relationship between the nitrogen (N) content and the difference in the spectral reflectance relationship and to establish the hyperspectral reflectance difference inversion model of differences in the N content of rice. In this study, the hyperspectral reflectance difference was used to invert the nitrogen deficiency of rice and provide a method for the implementation of precision fertilization without reducing the yield of chemical fertilizer. For the purpose of constructing the standard N content and standard spectral reflectance the principle of minimum fertilizer application at maximum yield was used as a reference standard, and the acquired rice leaf nitrogen content and leaf spectral reflectance were differenced from the standard N content and standard spectral reflectance to obtain N content. The difference and spectral reflectance differential were then subjected to discrete wavelet multiscale decomposition, successive projections algorithm, principal component analysis, and iteratively retaining informative variables (IRIVs); the results were treated as partial least squares (PLSR), extreme learning machine (ELM), and genetic algorithm-extreme learning machine (GA-ELM). The results of hyperspectral dimensionality reduction were used as input to establish the inverse model of N content differential in japonica rice. The results showed that the GA-ELM inversion model established by discrete wavelet multi-scale decomposition obtained the optimal results in data set modeling and training. Both the R2 of the training data set and the validation data set were above 0.68, and the root mean square errors (RMSEs) were <0.6 mg/g and were more predictive, stable, and generalizable than the PLSR and ELM predictive models.

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.

Band Selection Using Combined Divergence–Correlation Index and Sparse Loadings Representation for Hyperspectral Image Classification

Band selection and feature extraction are the two paradigms of hyperspectral dimensionality reduction. While feature extraction methods have many desirable properties, band selection methods keep the actual reflectances intact, which results in better interpretation of the behavior of the materials. Keeping this in mind, several feature extraction-based band selection methods have been proposed in the past, which can bridge the gap between these two approaches and combine their salient advantages. In this article, a feature extraction-based, clustering-ranking type band selection method is proposed in which the band selection is performed in two stages. In the clustering stage, a new and improved representation of bands using their component loadings for the top principal components is introduced, which is employed to cluster similar bands into groups using sparse subspace clustering. In the ranking stage, a new metric called combined divergence–correlation index is introduced to elect the most discriminative as well as least correlated bands as cluster representatives. Experimental results indicate that the proposed method can proficiently select a set of distinct and discriminative bands, which can help in effective hyperspectral classification. Additionally, it was also shown that the selected bands show rich information content, low interband correlation and high-class separabilities.

Learning to propagate labels on graphs: An iterative multitask regression framework for semi-supervised hyperspectral dimensionality reduction

Hyperspectral dimensionality reduction (HDR), an important preprocessing step prior to high-level data analysis, has been garnering growing attention in the remote sensing community. Although a variety of methods, both unsupervised and supervised models, have been proposed for this task, yet the discriminative ability in feature representation still remains limited due to the lack of a powerful tool that effectively exploits the labeled and unlabeled data in the HDR process. A semi-supervised HDR approach, called iterative multitask regression (IMR), is proposed in this paper to address this need. IMR aims at learning a low-dimensional subspace by jointly considering the labeled and unlabeled data, and also bridging the learned subspace with two regression tasks: labels and pseudo-labels initialized by a given classifier. More significantly, IMR dynamically propagates the labels on a learnable graph and progressively refines pseudo-labels, yielding a well-conditioned feedback system. Experiments conducted on three widely-used hyperspectral image datasets demonstrate that the dimension-reduced features learned by the proposed IMR framework with respect to classification or recognition accuracy are superior to those of related state-of-the-art HDR approaches.

Tensor Based Multiscale Low Rank Decomposition for Hyperspectral Images Dimensionality Reduction

Dimensionality reduction is an essential and important issue in hyperspectral image processing. With the advantages of preserving the spatial neighborhood information and the global structure information, tensor analysis and low rank representation have been widely considered in this field and yielded satisfactory performance. In available tensor- and low rank-based methods, how to construct appropriate tensor samples and determine the optimal rank of hyperspectral images along each mode are still challenging issues. To address these drawbacks, an unsupervised tensor-based multiscale low rank decomposition (T-MLRD) method for hyperspectral images dimensionality reduction is proposed in this paper. By regarding the raw cube hyperspectral image as the only tensor sample, T-MLRD needs no labeled samples and avoids the processing of constructing tensor samples. In addition, a novel multiscale low rank estimating method is proposed to obtain the optimal rank along each mode of hyperspectral image which avoids the complicated rank computing. Finally, the multiscale low rank feature representation is fused to achieve dimensionality reduction. Experimental results on real hyperspectral datasets demonstrate the superiority of the proposed method over several state-of-the-art approaches.

Learning a local manifold representation based on improved neighborhood rough set and LLE for hyperspectral dimensionality reduction

Hyperspectral data with high dimensionality always needs more storage space and increases the computational consumption, manifold learning based dimensionality reduction method is a popular way to meet these requirements. Local manifold learning consists of two vital steps: neighbors selection and weights computation, and the former is critical to the accuracy of the result. In this paper, an improved neighborhood rough set (INRS) and local linear embedding (LLE) are proposed for local neighbors selection, thus INRSLLE is proposed. Firstly, neighbor candidates are selected using Euclidean distance and spectral distance, then a parameter undirected graph is constructed from original data to improve the anti-noise ability. INRS improved by this parameter undirected graph can select enough neighbors from these candidates, which increases the contribution of neighbors to corresponding samples. Finally, LLE is used to learn local manifold representation. Results of the proposed method and comparison methods like NMF and LP-KSVD have been classified by four different classifiers respectively. Experimental results performed over two real-world hyperspectral datasets indicate the proposed INRSLLE not only considers the spectral-spatial information of hyperspectral data, but also selects more suitable neighbors on local manifolds and increases the anti-noise ability.

Implementation of the Principal Component Analysis onto High-Performance Computer Facilities for Hyperspectral Dimensionality Reduction: Results and Comparisons

Dimensionality reduction represents a critical preprocessing step in order to increase the efficiency and the performance of many hyperspectral imaging algorithms. However, dimensionality reduction algorithms, such as the Principal Component Analysis (PCA), suffer from their computationally demanding nature, becoming advisable for their implementation onto high-performance computer architectures for applications under strict latency constraints. This work presents the implementation of the PCA algorithm onto two different high-performance devices, namely, an NVIDIA Graphics Processing Unit (GPU) and a Kalray manycore, uncovering a highly valuable set of tips and tricks in order to take full advantage of the inherent parallelism of these high-performance computing platforms, and hence, reducing the time that is required to process a given hyperspectral image. Moreover, the achieved results obtained with different hyperspectral images have been compared with the ones that were obtained with a field programmable gate array (FPGA)-based implementation of the PCA algorithm that has been recently published, providing, for the first time in the literature, a comprehensive analysis in order to highlight the pros and cons of each option.

BAS-ELM based UAV hyperspectral remote sensing inversion modeling of rice canopy nitrogen content

Abstract: The rapid, nondestructive, and accurate estimation of nitrogen content in rice can help to obtain the growth condition of rice in time, which is of great significance for guiding rice field management. In order to improve the accuracy of high-spectral inversion of rice canopy nitrogen content, the discrete wavelet multiscale decomposition (DWMD) was used to downscale the high-spectral information in the range of 400 nm to 1000 nm by using the UAV hyperspectral image data and the synchronously measured rice canopy nitrogen content as the data source, from which the hyperspectral characteristic variables for rice nitrogen content inversion modeling were extracted. And using the dimensionality reduction variables as the data base, three neural network inversion methods, including extreme learning machines (ELM), particle swarm optimization for extreme learning machines (PSO-ELM), and beetle antennae search algorithm for extreme learning machines (BAS-ELM), were used to establish the rice nitrogen content drone hyperspectral remote sensing inversion model, and the results showed that: (1) The hyperspectral range from 400 nm to 1000 nm was dimensionally reduced by DWMD, and finally the continuous hyperspectral reflectance information was dimensionally reduced to sixteen discrete hyperspectral features for subsequent inversion modelling of rice nitrogen content.(2) In the model adopted in this study, BAS-ELM has the highest accuracy, where the R 2 of training data, the R 2 of test data, the RMSE of training data, and the RMSE of test data is were 0.864, 0.863, 0.247, 0.254. Keywords: UAV, Nitrogen inversion, rice, Hyperspectral dimensionality reduction, ELM DOI:  10.33440/j.ijpaa.20200303.105  Citation: Yu F H, Feng S,Yao W X, Wang D K, Xing S M, Xu T Y.  BAS-ELM based UAV hyperspectral remote sensing inversion modeling of rice canopy nitrogen content.  Int J Precis Agric Aviat, 2020; 3(3): 59–64.

Fast Spatial Spectral Schroedinger Eigenmaps algorithm for hyperspectral feature extraction

Based on the Laplacian Eigenmaps (LE) algorithm and a potential matrix, the Spatial Spectral Schroedinger Eigenmaps (SSSE) technique has proved a great yield during the hyperspectral dimensionality reduction process. Experimentally, SSSE is in deficiency of high computing time which may hinder its contribution in the remote sensing field. In this paper, a fast variant of the SSSE approach, called Fast SSSE, was proposed. The new suggested method substitutes the quadratic constraint employed during the optimization problem, by a linear constraint. This overhaul preserves the data properties in analogous way to the SSSE technique, but with a fast implementation. Two real hyperspectral data sets were adopted during the experimental process. Experiment analysis exhibited good classification accuracy with a reduced computational effort, compared with the original SSSE approach.

Learning a Robust Local Manifold Representation for Hyperspectral Dimensionality Reduction

Local manifold learning has been successfully applied to hyperspectral dimensionality reduction in order to embed nonlinear and nonconvex manifolds in the data. Local manifold learning is mainly characterized by affinity matrix construction, which is composed of two steps: neighbor selection and computation of affinity weights. There is a challenge in each step: First, the neighbor selection is sensitive to complex spectral variability due to nonuniform data distribution, illumination variations, and sensor noise; second, the computation of affinity weights is challenging due to highly correlated spectral signatures in the neighborhood. To address the two issues, in this paper, a novel manifold learning methodology based on locally linear embedding is proposed through learning a robust local manifold representation. More specifically, a hierarchical neighbor selection is designed to progressively eliminate the effects of complex spectral variability using joint normalization and to robustly compute affinity (or reconstruction) weights reducing multicollinearity via the refined neighbor selection. Additionally, an idea that combines spatial–spectral information is introduced into the proposed manifold learning methodology to further improve the robustness of affinity calculations. Classification is explored as a potential application for validating the proposed algorithm. The classification accuracy in the use of different dimensionality reduction methods is evaluated and compared, while two kinds of strategies are applied in selecting the training and test samples: random sampling and region-based sampling. Experimental results show the classification accuracy obtained by the proposed method is superior to those state-of-the-art dimensionality reduction methods.

Hyperspectral Dimensionality Reduction by Tensor Sparse and Low-Rank Graph-Based Discriminant Analysis

Recently, sparse and low-rank graph-based discriminant analysis (SLGDA) has yielded satisfactory results in hyperspectral image (HSI) dimensionality reduction (DR), for which sparsity and low-rankness are simultaneously imposed to capture both local and global structure of hyperspectral data. However, SLGDA fails to exploit the spatial information. To address this problem, a tensor sparse and low-rank graph-based discriminant analysis (TSLGDA) is proposed in this paper. By regarding the hyperspectral data cube as a third-order tensor, small local patches centered at the training samples are extracted for the TSLGDA framework to maintain the structural information, resulting in a more discriminative graph. Subsequently, dimensionality reduction is performed on the tensorial training and testing samples to reduce data redundancy. Experimental results of three real-world hyperspectral datasets demonstrate that the proposed TSLGDA algorithm greatly improves the classification performance in the low-dimensional space when compared to state-of-the-art DR methods.