Comparative Analysis of Haar and Daubechies Wavelet for Hyper Spectral Image Classification

Abstract. Hyperspectral image enhancement has been a concern for the remote sensing society for detailed end member detection. Hyperspectral remote sensor collects images in hundreds of narrow, continuous spectral channels, whereas multispectral remote sensor collects images in relatively broader wavelength bands. However, the spatial resolution of the hyperspectral sensor image is comparatively lower than that of the multispectral. As a result, spectral signatures from different end members originate within a pixel, known as mixed pixels. This paper presents an approach for obtaining an image which has the spatial resolution of the multispectral image and spectral resolution of the hyperspectral image, by fusion of hyperspectral and multispectral image. The proposed methodology also addresses the band remapping problem, which arises due to different regions of spectral coverage by multispectral and hyperspectral images. Therefore we apply algorithms to restore the spatial information of the hyperspectral image by fusing hyperspectral bands with only those bands which come under each multispectral band range. The proposed methodology is applied over Henry Island, of the Sunderban eco-geographic province. The data is collected by the Hyperion hyperspectral sensor and LISS IV multispectral sensor.

In recent studies, superpixel segmentation has been integrated into hyperspectral (HS) image classification methods. However, the existing superpixel-based classification methods usually suffer from two serious problems. First, the accuracy and efficiency of current superpixel segmentation approaches cannot meet the demands of practical applications for HS images; second, conventional superpixel-based classification methods generally consider each generated superpixel as a unit for the image classification, which may help to reduce the computing time but result in a significant decrease of the classification accuracy. To solve the problems, we propose a fast region growing based superpixel segmentation (FRGSS) algorithm and a novel texture-adaptive superpixel integration strategy (TASIS) for the HS image classification. Experimental results on real Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) HS images demonstrate that the proposed FRGSS outperforms the state-of-the-art superpixel algorithm. In addition, the superiority of the TASIS is verified compared to the pixel-wise and the conventional superpixel-based classification methods.

Despite the importance of data reduction as part of the processing of reflection-based classifications, this study represents one of the first in which the effects of both spatial and spectral data reductions on classification accuracies are quantified. Furthermore, the effects of approaches to data reduction were quantified for two separate classification methods, linear discriminant analysis (LDA) and support vector machine (SVM). As the model dataset, reflection data were acquired using a hyperspectral camera in 230 spectral channels from 401 to 879 nm (spectral resolution of 2.1 nm) from field pea (Pisum sativum) samples with and without internal pea weevil (Bruchus pisorum) infestation. We deployed five levels of spatial data reduction (binning) and eight levels of spectral data reduction (40 datasets). Forward stepwise LDA was used to select and include only spectral channels contributing the most to the separation of pixels from non-infested and infested field peas. Classification accuracies obtained with LDA and SVM were based on the classification of independent validation datasets. Overall, SVMs had significantly higher classification accuracies than LDAs (P < 0.01). There was a negative association between pixel resolution and classification accuracy, while spectral binning equivalent to up to 98% data reduction had negligible effect on classification accuracies. This study supports the potential use of reflection-based technologies in the quality control of food products with internal defects, and it highlights that spatial and spectral data reductions can (1) improve classification accuracies, (2) vastly decrease computer constraints, and (3) reduce analytical concerns associated with classifications of large and high-dimensional datasets.

In this paper, we propose a lightweight machine learning (ML) framework based on unsupervised spectral domain discretization for hyperspectral image (HSI) classification. Firstly, the high-dimensional HSI data is mapped into a discretized image by unsupervised learning method, and then the histogram statistics of discrete features are performed to align feature vectors. Finally, supervised ML method is used for classification, thus achieving a lightweight ML method of high-dimensional HSIs. Practical applications and comparative studies on three publicly available HSI datasets show that the framework approaches and surpasses deep learning models in classification accuracy while significantly compressing computational time consumption. The performance of six unsupervised clustering methods in HSI spectral domain discretization is compared in the study. Among them, K-means and GMM are superior in terms of classification accuracy. And SOM provides high classification accuracy while its discretization results are better interpretable due to better maintenance of topology during discretization.

ABSTRACTClassification is an important problem in a large variety of applications, which makes it an open-ended forum for researchers in various disciplines. In this paper, the proposed two approaches are mostly focused on Nearest Regularized Subspace (NRS) classifier. The proposed pair of variants are: (1) reducing the computational complexity of NRS classifier termed as Conditional Fast Nearest Regularized Subspace (CFNRS) classifier and (2) incorporation of dissimilarity feature termed as Conditional Dissimilarity-based Nearest Regularized Subspace (CDNRS) classifier. Regarding the first approach, the simple k-NN classifier result is used as a condition to evaluate NRS classifier. Regarding the second approach, an intra-feature dissimilarity measure is considered to create a pair of dictionary which contains a conditional binary matrix and a dissimilarity feature matrix. Each conditional binary matrix is a collection of distinct sub-spaces representing their respective classes. The incoming data is classified with the help of a dictionary and classification performance is validated over other state-of-art approaches. As Hyper-Spectral (HS) data have a strong spectral correlation in terms of spectral signatures, various such HS data are included for validation of our approach. The experimental study reveals the pros and cons of different classifiers along with the effectiveness of our proposed approach. The proposed conditional classifier is found to be able to compete with other, standard classifiers in terms of accuracy and computational complexity, mostly where features possess some sort of sequence information. The suitability of the proposed classifier is also verified for both binary and multi-class classification problems. Along with the experimental validation, the statistical significance of the proposed conditional measure is assessed over other standard state-of-art methods, and the proposed one outperformed the others. The experimental results reveal the importance of dissimilarity features for classification and also suggest to incorporate a simple classifier such as k-NN, prior to any other classifier. This is mainly meant to reduce the computational complexity of the selected classifier without compromising the accuracy.

Suaeda salsa (S. salsa) has a significant protective effect on salt marshes in coastal wetlands. In this study, the abilities of airborne multispectral images, spaceborne hyperspectral images, and LiDAR data in spatial distribution extraction and aboveground biomass (AB) estimation of S. salsa were explored for mapping the spatial distribution of S. salsa AB. Results showed that the increasing spectral and structural features were conducive to improving the classification accuracy of wetland vegetation and the AB estimation accuracy of S. salsa. The fusion of hyperspectral and LiDAR data provided the highest accuracies for wetlands classification and AB estimation of S. salsa in the study. Multispectral images alone provided relatively high user's and producer's accuracies of S. salsa classification (87.04% and 88.28%, respectively). Compared to multispectral images, hyperspectral data with more spectral features slightly improved the Kappa coefficient and overall accuracy. The AB estimation reached a relatively reliable accuracy based only on hyperspectral data (R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.812, root-mean-square error of 0.295, estimation error of 24.56%, residual predictive deviation of 2.033, and the sums of squares ratio of 1.049). The addition of LiDAR data produced a limited improvement in the process of extraction and AB estimation of S. salsa. The spatial distribution of mapped S. salsa AB was consistent with the field survey results. This study provided an important reference for the effective information extraction and AB estimation of wetland vegetation S. salsa.

Abstract. Hyperspectral images (HSIs) contain hundreds of spectral bands, providing high-resolution spectral information pertaining to the Earth’s surface. Additionally, abundant spatial contextual information can also be obtained simultaneously from a HSI. To characterize the properties of ground objects, classification is the most widely-used technology in the field of remote sensing, where each pixel in a HSI is assigned to a pre-defined class. Over the past decade, deep learning has attracted increasing attention in the machine-learning and computer-vision domains, due to its favourable performances for various types of tasks, and it has been successfully introduced to the remote-sensing community. Instead of utilizing the shallow features within in a given image, which is the approach that is generally adopted in other conventional classification methods, deep-learning algorithms can extract hierarchical features from raw HSI data. Within the deep-learning framework, recurrent neural networks (RNNs), which are able to encode sequential features, have exhibited promising capabilities and have achieved encouraging performances, especially for the natural-language processing and speech-recognition communities. As multi-temporal remote-sensing images can be readily obtained from increasing numbers of satellite and unmanned aircraft systems, and since analysis of such multi-temporal data comprises a critical issue within numerous research subfields, including land-cover and land-change analyses, and land-resource management, RNNs have been applied in recent studies in order to extract temporal sequential features from multi-temporal remote-sensing images for the purpose of image classification. Apart from using multi-temporal image datasets, RNNs can also be utilized on a single image, where the spectral feature/band of each individual pixel can be taken as a sequential feature for the input layer of RNNs. However, the application of such sequential feature extraction that relies on a single image still needs to be further investigated since applying RNNs to spectral bands will directly introduce more parameters that need to be optimized, consequently increasing the total training time.In this study, we propose a novel RNN-based HSI classification framework. In this framework, unlabelled pixels obtained from a single image are considered when constructing sequential features. Two spatial similarity measurements, referred to as pixel-matching and block-matching, respectively, are employed to extract pixels that are “similar” to the target pixel. Then, the sequential feature of the target pixel is constructed by exploiting several of the most “similar” pixels and ordering them based on their similarities to the target pixel. The aforementioned two schemes are advantageous, as unlabelled pixels within the given HSI are taken into consideration for similarity measurement and sequential feature construction for the RNN model. Moreover, the block-matching scheme also takes advantage of spatial contextual information, which has been widely utilized in spatial-spectral-based HSI classification methods. To evaluate the proposed methods, two benchmark HSIs are used, including a HSI collected over Pavia University, Italy by the airborne Reflective Optics System Imaging Spectrometer (ROSIS) sensor, and an image acquired over the Salinas Valley, California, USA via the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensor. Spatio-temporally coincident ground-reference data accompanies each of these respective HSIs. In addition, the proposed methods are compared with three state-of-the-art algorithms, including support vector machine (SVM), the 1-dimensional convolutional neural network (1DCNN), and the 1-dimensional RNN (1DRNN).Experimental results indicate that our proposed methods achieve markedly better classification performance compared with the baseline algorithms on both datasets. For example, for the Pavia University image, the block-matching based RNN achieves the highest overall classification accuracy, with 94.32% accuracy, which is 9.87% higher than the next most accurate algorithm of the aforementioned three baseline methods, which in this case is the 1DCNN, with 84.45% overall accuracy. More specifically, the block-matching method performs better than the pixel-matching method in terms of both quantitative and qualitative assessments. Based on visual assessment/interpretation of the classification maps, it is apparent that “salt-and-pepper” noise is markedly alleviated; with block-matching, smoother classified images are generated compared with pixel-matching-based methods and the three baseline algorithms. Such results demonstrate the effectiveness of utilizing spatial contextual information in the similarity measurement.

Compressive sensing and adaptive direct sampling in hyperspectral imaging

For many remote sensing applications it is preferable to have images with both high spectral and spatial resolutions. On this regards, hyperspectral and multispectral images have complementary characteristics in terms of spectral and spatial resolutions. In this paper we propose an approach for the fusion of low spatial resolution hyperspectral images with high spatial resolution multispectral images in order to obtain superresolution (spatial and spectral) hyperspectral images. The proposed approach is based on the assumption that, since both hyperspectral and multispectral images acquired on the same scene, the corresponding endmembers should be the same. On a first step the hyperspectral image is spectrally down-sampled in order to match the multispectral one. Then an endmember extraction algorithm is performed on the down-sampled hyperspectral image and the successive abundance estimation is performed on the multispectral one. Finally, the extracted endmembers are up-sampled back to the original hyperspectral space and then used to reconstruct the superresolution hyperspectral image according to the abundances obtained from the multispectral image.

Recent growth in unmanned aerial vehicle (UAV) technology have promoted the detailed mapping of individual tree species. However, the in-depth mining and comprehending of the significance of features derived from high-resolution UAV data for tree species discrimination remains a difficult task. In this study, a state-of-the-art approach combining UAV-borne light detection and ranging (LiDAR) and hyperspectral was used to classify 11 common tree species in a typical natural secondary forest in Northeast China. First, comprehensive relevant structural and spectral features were extracted. Then, the most valuable feature sets were selected by using a hybrid approach combining correlation-based feature selection with the optimized recursive feature elimination algorithm. The random forest algorithm was used to assess feature importance and perform the classification. Finally, the robustness of features derived from point clouds with different structures and hyperspectral images with different spatial resolutions was tested. Our results showed that the best classification accuracy was obtained by combining LiDAR and hyperspectral data (75.7%) compared to that based on LiDAR (60.0%) and hyperspectral (64.8%) data alone. The mean intensity of single returns and the visible atmospherically resistant index for red-edge band were the most influential LiDAR and hyperspectral derived features, respectively. The selected features were robust in point clouds with a density not lower than 5% (~5 pts/m2) and a resolution not lower than 0.3 m in hyperspectral data. Although canopy surface features were slightly different from original LiDAR features, canopy surface information was also important for tree species classification. This study proved the capabilities of UAV-borne LiDAR and hyperspectral data in natural secondary forest tree species discrimination and the potential for this approach to be transferable to other study areas.

In recent years, there has been a notable surge in the development of methods dedicated to classifying hyper spectral image samples. However, the inherent high dimensionality of hyper spectral data presents significant challenges for traditional supervised classification algorithms. Prior studies have attempted to address these challenges by incorporating active learning methods for hyper spectral data segmentation. Nevertheless, a crucial drawback of active learning lies in its dependence on labeled data samples, leading to less improved classification results for unlabeled data samples. In response to this limitation, our work introduces Supervised Active Learning (SAL) techniques that harness both spectral and spatial information. The key innovation of our research lies in the introduction of a novel hyper spectral image segmentation approach based on the k-dependence Bayesian (KDB) model. We estimate the maximizer of the posterior marginal using Symmetric Iterative Proportional Fitting (SIPF) within this KDB model. By implementing the KDB model for hyper spectral image segmentation, our goal is to enhance the classification accuracy of our SAL classifiers. The SAL framework integrates shape prior probability estimation, class density results from Improved Multi-Level Logistic (IMLL) methods, and the maximization of the posterior marginal estimated through Symmetric Iterative Proportional Fitting (SIPF). We evaluate the accuracy of our proposed MPM-SIPF-SAL system using the real ROSIS Hyper Spectral Dataset, demonstrating that our system outperforms existing methods in terms of classification accuracy.

The hyperspectral and multispectral image fusion process, commonly known as the multisharpening process (also known as the hypersharpening process), results in the creation of unobservable high spectral and spatial resolution hyperspectral remote sensing data, by combining (or fusing) a hyperspectral image, characterized by a high spectral resolution, with, especially, a high spatial resolution multispectral one. There are several approaches in the literature performing such a process, in particular, those based on the spectral unmixing concept. The latter fusion methods aim to extract pure-material spectra from the high spectral resolution hyperspectral image and combine them with their corresponding proportions extracted from the high spatial resolution multispectral image. However, most of these methods, which mainly consider the standard linear mixing model, do not address the spectral variability phenomenon, precisely due to the considered data mixing model, which does not handle variations in illumination and atmospheric conditions or material composition inducing spectral variability. Neglecting such a phenomenon may lead to errors in the fusion process, thus decreasing the spatial and spectral qualities of sharpened products. In this paper, a new hyperspectral and multispectral data fusion method, based on the linear spectral unmixing concept that deals with spectral variability, is proposed. This approach, based on a multiplicative Nonnegative Matrix Factorization (NMF) technique, substantially extends the multiplicative Coupled NMF (CNMF) algorithm by considering a linear data mixing model that addresses the considered phenomenon. Indeed, the introduced approach minimizes new cost functions involving matrices with specific structures in order to deal with the spectral variability phenomenon. The proposed method is applied to synthetic and real data, and its effectiveness is evaluated with established performance criteria. Experimental results show that the designed algorithm globally yields sharpened hyperspectral data with fine spectral and spatial fidelities. Also, these results demonstrate that the proposed technique globally outperforms the tested literature approaches.

A significant challenge in hyperspectral remote sensing image analysis is the presence of noise, which has a negative impact on various data analysis methods such as image classification, target detection, unmixing etc. In order to address this issue, hyperspectral image denoising is used as a preprocessing step prior to classification. This paper presents an effective, fast and reliable method for denoising hyperspectral images followed by classification based on sparse representation of hyperspectral data. The use of Legendre-Fenchel transform for denoising is an effective spatial preprocessing step to improve the classification accuracy. The main advantage of Legendre-Fenchel transform is that it removes the noise in the image while preserving the sharp edges. The sparsity based algorithm namely, Orthogonal Matching Pursuit (OMP) is used for classification. The experiment is done on Indian Pines data set acquired by Airborne Visible Infrared Imaging Spectrometer (AVIRIS) sensor. It is inferred that the denoising of hyperspectral images before classification improves the Overall Accuracy of classification. The effect of preprocessing using Legendre Fenchel transformation is shown by comparing the classification results with Total Variation (TV) based denoising. A statistical comparison of the accuracies obtained on standard hyperspectral data before and after denoising is also analysed to show the effectiveness of the proposed method. The experimental result analysis shows that for 10\(\%\) training set the proposed method leads to the improvement in Overall Accuracy from 83.18\(\%\) to 91.06\(\%\), Average Accuracy from 86.17\(\%\) to 92.78\(\%\) and Kappa coefficient from 0.8079 to 0.8981.

Convolutional neural networks (CNNs) have recently achieved outstanding performance for hyperspectral (HS) and multispectral (MS) image fusion. However, CNNs cannot explore the long-range dependence for HS and MS image fusion because of their local receptive fields. To overcome this limitation, a transformer is proposed to leverage the long-range dependence from the network inputs. Because of the ability of long-range modeling, the transformer overcomes the sole CNN on many tasks, whereas its use for HS and MS image fusion is still unexplored. In this article, we propose a spectral-spatial transformer (SST) to show the potentiality of transformers for HS and MS image fusion. We devise first two branches to extract spectral and spatial features in the HS and MS images by SST blocks, which can explore the spectral and spatial long-range dependence, respectively. Afterward, spectral and spatial features are fused feeding the result back to spectral and spatial branches for information interaction. Finally, the high-resolution (HR) HS image is reconstructed by dense links from all the fused features to make full use of them. The experimental analysis demonstrates the high performance of the proposed approach compared with some state-of-the-art (SOTA) methods.

Encoding of multispectral and hyperspectral image data using wavelet transform and gain shape vector quantization