Spectral Classification Methods Research Articles

Hyperspectral Image Classification Based on Mutually Guided Image Filtering

Hyperspectral remote sensing images (HSIs) have both spectral and spatial characteristics. The adept exploitation of these attributes is central to enhancing the classification accuracy of HSIs. In order to effectively utilize spatial and spectral features to classify HSIs, this paper proposes a method for the spatial feature extraction of HSIs based on a mutually guided image filter (muGIF) and combined with the band-distance-grouped principal component. Firstly, aiming at the problem that previously guided image filtering cannot effectively deal with the inconsistent information structure between the guided and target information, a method for extracting spatial features using muGIF is proposed. Then, aiming at the problem of the information loss caused by a single principal component as a guided image in the traditional GIF-based spatial–spectral classification, a spatial feature-extraction framework based on the band-distance-grouped principal component is proposed. The method groups the bands according to the band distance and extracts the principal components of each set of band subsets as the guide map of the current band subset to filter the HSIs. A deep convolutional neural network model and a generative adversarial network model for the filtered HSIs are constructed and then trained using samples for HSIs’ spatial–spectral classification. Experiments show that compared with the traditional methods and several popular spatial–spectral HSI classification methods based on a filter, the proposed methods based on muGIF can effectively extract the spatial–spectral features and improve the classification accuracy of HSIs.

DRC-Net Method for Two-dimensional Spectral Classification

Spectral classification plays a crucial role in the analysis of astronomical data. Currently, stellar spectral classification primarily relies on one-dimensional (1D) spectra and necessitates a sufficient signal-to-noise ratio (S/N). However, in cases where the S/N is low, obtaining valuable information becomes impractical. In this paper, we propose a novel model called DRC-Net (Double-branch celestial spectral classification network based on residual mechanisms) for stellar classification, which operates solely on two-dimensional (2D) spectra. The model consists of two branches that use 1D convolutions to reduce the dimensionality of the 2D spectral composed of both blue and red arms. In the following, the features extracted from both branches are fused, and the fused result undergoes further feature extraction before being fed into the classifier for final output generation. The data set is from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope, comprising 15,680 spectra of F, G, and K types. The preprocessing process includes normalization and the early stopping mechanism. The experimental results demonstrate that the proposed DRC-Net achieved remarkable classification precision of 93.0%, 83.5%, and 86.9% for F, G, and K types, respectively, surpassing the performance of 1D spectral classification methods. Furthermore, different S/N intervals are tested to judge the classification ability of DRC-Net. The results reveal that DRC-Net, as a 2D spectral classification model, can deliver superior classification outcomes for the spectra with low S/Ns. These experimental findings not only validate the efficiency of DRC-Net but also confirm the enhanced noise resistance ability exhibited by 2D spectra.

Spectral classification by generative adversarial linear discriminant analysis

Generative adversarial linear discriminant analysis (GALDA) is formulated as a broadly applicable tool for increasing classification accuracy and reducing overfitting in spectrochemical analysis. Although inspired by the successes of generative adversarial neural networks (GANs) for minimizing overfitting artifacts in artificial neural networks, GALDA was built around an independent linear algebra framework distinct from those in GANs. In contrast to feature extraction and data reduction approaches for minimizing overfitting, GALDA performs data augmentation by identifying and adversarially excluding the regions in spectral space in which genuine data do not reside. Relative to non-adversarial analogs, loading plots for dimension reduction showed significant smoothing and more prominent features aligned with spectral peaks following generative adversarial optimization. Classification accuracy was evaluated for GALDA together with other commonly available supervised and unsupervised methods for dimension reduction in simulated spectra generated using an open-source Raman database (Romanian Database of Raman Spectroscopy, RDRS). Spectral analysis was then performed for microscopy measurements of microsphereroids of the blood thinner clopidogrel bisulfate and in THz Raman imaging of common constituents in aspirin tablets. From these collective results, the potential scope of use for GALDA is critically evaluated relative to alternative established spectral dimension reduction and classification methods.

Dual Classification Approach for the Rapid Discrimination of Metabolic Syndrome by FTIR

Metabolic syndrome is a complex of interrelated risk factors for cardiovascular disease and diabetes. Thus, new point-of-care diagnostic tools are essential for unambiguously distinguishing MetS patients, providing results in rapid time. Herein, we evaluated the potential of Fourier transform infrared spectroscopy combined with chemometric tools to detect spectra markers indicative of metabolic syndrome. Around 105 plasma samples were collected and divided into two groups according to the presence of at least three of the five clinical parameters used for MetS diagnosis. A dual classification approach was studied based on selecting the most important spectral variable and classification methods, linear discriminant analysis (LDA) and SIMCA class modelling, respectively. The same classification methods were applied to measured clinical parameters at our disposal. Thus, the classification's performance on reduced spectra fingerprints and measured clinical parameters were compared. Both approaches achieved excellent discrimination results among groups, providing almost 100% accuracy. Nevertheless, SIMCA class modelling showed higher classification performance between MetS and no MetS for IR-reduced variables compared to clinical variables. We finally discuss the potential of this method to be used as a supportive diagnostic or screening tool in clinical routines.

Two-Branch Pure Transformer for Hyperspectral Image Classification

Owing to its capability of building long-range dependencies and global context connections, Transformer has been used for hyperspectral image (HSI) classification. However, most of the existing Transformer-based HSI spatial–spectral classification methods consist of a convolutional neural network (CNN) and Transformer, which are used to extract the local and global information, respectively. In this study, to fully explore the potential of Transformer, a pure Transformer is investigated for HSI classification. First, a spatial Transformer (Spa-TR) is designed for HSI spatial classification, which learns the spatial features locally and globally by adopting the window partition and shifted window schemes. Especially, the self-attention computations are limited within the local windows and cross-windows. Second, to fully use the abundant spectral information in HSIs, a two-branch pure Transformer (i.e., Spa-Spe-TR) is proposed, which includes a spectral Transformer (Spe-TR) and a Spa-TR. The spectral sequence features learned by Spe-TR and the spatial features generated by Spa-TR are effectively fused with a branch fusion strategy, which explicitly and automatically measures the importance between the joint spatial–spectral features and improves the discriminability of the joint features. Experimental results on the two widely used HSI datasets (i.e., Pavia and Indian Pines) demonstrate the efficacy of the proposed methods in comparison with other state-of-the-art approaches.

Efficiently Mapping Large Areas of Olive Trees Using Drones in Extremadura, Spain

Remotely sensed multispectral (MS) images have been widely used in different areas of research in recent years, allowing their exploitation in many domains like precision agriculture. This study uses high-resolution RGB data, collected by an unmanned aerial vehicle (UAV/drone), equipped with a camera sensor and flown over an olive tree field in Extremadura (Spain), for land-cover classification purposes. Drone flights over large agricultural areas, collecting high spatial resolution images such as the scenario studied, can be tedious to process, requiring a large amount of random-access memory (RAM). To solve this issue, a new olive tree cover mapping approach is proposed by applying a spatial partitioning methodology that consists of splitting the image automatically, before processing. Each part (known as the raster window) is classified by using machine learning methods. Finally, a classification map is obtained after processing all windows. Among different supervised spectral classification methods, the following classifiers are considered: random forest (RF), support vector machine (SVM), multilayer perceptron (MLP), and k-nearest neighbor (kNN), with the ultimate goal of detecting olive trees and distinguishing them from other land-cover types available in the scene, such as soil, shadows, and weeds. In addition, our experimental results demonstrate that these classifiers can be used to avoid time-consuming fieldwork tasks, facilitating the analysis of agricultural fields with “a la carte” UAV flights resulting in high classification accuracy.

Spatial Density Peak Clustering for Hyperspectral Image Classification With Noisy Labels

The “ noisy label ” problem is one of the major challenges in hyperspectral image (HSI) classification. In order to address this problem, a spatial density peak (SDP) clustering-based method is proposed to detect mislabeled samples in the training set. Specifically, the proposed methods consist of the following steps: first, the correlation coefficients among the training samples in each class are estimated. In this step, instead of measuring the correlation coefficients by considering individual samples, all neighbor samples or $K$ representative neighbor samples in a local window surrounding each training sample are considered. By this way, the spatial contextual information could be used, and two versions of the proposed method, i.e., measuring the correlation coefficients using all neighbor samples or $K$ representative samples, are referred as SDP and $K$ -SDP, respectively. Second, with the correlation coefficients calculated above, the local density of each training sample can be obtained by the DP clustering algorithm. Finally, those mislabeled samples which usually have lower local densities in each class are able to be identified by a defined decision function. The effectiveness of the proposed detection method is evaluated using a series of spectral and spectral-spatial classification methods on several real hyperspectral data sets.

Smartphone-based multispectral imaging and machine-learning based analysis for discrimination between seborrheic dermatitis and psoriasis on the scalp.

For appropriate treatment, accurate discrimination between seborrheic dermatitis and psoriasis in a timely manner is crucial to avoid complications. However, when they occur on the scalp, differential diagnosis can be challenging using conventional dermascopes. Thus, we employed smartphone-based multispectral imaging and analysis to discriminate between them with high accuracy. A smartphone-based multispectral imaging system, suited for scalp disease diagnosis, was redesigned. We compared the outcomes obtained using machine learning-based and conventional spectral classification methods to achieve better discrimination. The results demonstrated that smartphone-based multispectral imaging and analysis has great potential for discriminating between these diseases.

Semisupervised Hyperspectral Image Classification Using Spatial-Spectral Information and Landscape Features

In hyperspectral image classification, the foremost task is that: how can we apply limited labeled samples to achieve good classification results? Spatial–spectral classification methods, which assign a label to each pixel regarding both spatial and spectral information, are effective to improve classification performance. Moreover, semisupervised learning (SSL) focuses on the scenario that the number of labeled data is rather small while a large number of unlabeled data are available. To complement spatial–spectral classification methods and semisupervised learning for each other, we propose a novel learning landscape features semisupervised framework (LLFSF) based on M-training algorithm and weighted spatial-spectral double layer SVM classifiers module (WSS-DSVM). In this novel framework, we first propose a SLIC (simple linear iterative clustering) based non-local superpixel segmentation algorithm to initially learn landscape feature and spatial composition. Then, we apply WSS-DSVM module to obtain initial classification maps. To better characterize complex scenes of hyperspectral images, we quantizes both the landscape diversity and separability from the initial classification map, which increase availability of spatial details and structural information of objects. Finally, we put some patches with lower accuracy into Multiple-training algorithm for further classification. In order to achieve an unbiased evaluation, we have evaluated the performance of LLFSF on three different scene hyperspectral data sets and compare it with that of three state-of-the-art hyperspectral image classification methods. The experimental results confirm the efficacy of the proposed framework.

Su Kalitesinin Sınıflandırılmasında Spektral Sınıflandırma Yöntemlerinin Karşılaştırılması

Today, water quality and water pollution can be detected using remote sensing and its terrestrial components. Remote sensing does not only provide a quick solution to detect water quality and pollution, but it could also be low cost. Within the scope of the study, the water quality of the İmranlı area of the Kızılırmak River, one of the most important water resources of the Sivas region and the İmranlı dam on the river, was investigated by spectral classification methods. Water samples were taken from various points on the river and dam and their chemical oxygen demands were determined in the laboratory. In addition, the reflectance values of the water samples taken by the local spectral measurements were examined in order to use as end members for spectral classification. CHRIS Proba is used as satellite image. Match filtering (MF), spectral angle mapping (SAM) and spectral information divergence (SID) methods have been used as the spectral classification methods and it has been examined which method gives better results in determining water quality. According to the results, it is understood that SAM method provides better classification accuracy than other methods.

Sub-pixel vs. super-pixel-based greenspace mapping along the urban–rural gradient using high spatial resolution Gaofen-2 satellite imagery: a case study of Haidian District, Beijing, China

ABSTRACTGreenspace in urban areas is closely related to urban ecosystems, economy, culture, and society. Recently, rapid urban development and expansion are always dominated by a series of human–environment interactions, which can lead to various spatial patterns of urban greenspace especially along the urban–rural gradient. Urban–rural greenspace mapping is therefore of great importance to provide a comprehensive insight for urban planners and managers. In our study, we adopted both the sub-pixel and super-pixel strategies to map the greenspace in Haidian District, Beijing, China. Specifically, the fully constrained linear spectral unmixing and object-based classification methods were implemented as the representatives of sub-pixel and super-pixel strategies, respectively. The high spatial resolution Gaofen-2 multispectral imagery collected in September, 2015 was used in this study. The results showed that the overall accuracies of greenspace mapping by the super-pixel method were higher than those by the sub-pixel method in the selected dense urban, sub-urban, and rural subsets. Obviously, the super-pixel method was more advantageous for mapping greenspace from the high spatial resolution imagery, especially for patches of greenspace in rural and mountain areas. When further comparing these two methods using the medium spatial resolution Landsat-8 imagery, we concluded that the sub-pixel method failed to keep the same levels of greenspace mapping accuracies as those using the high spatial resolution Gaofen-2 imagery but outperformed the super-pixel method especially in the dense urban and sub-urban subsets due to their high degrees of greenspace fragmentation. Furthermore, the sub-pixel method also demonstrated its merits in terms of automation and operability compared to the super-pixel method.

Advanced Investigation of Remote Sensing to Geological Mapping of Zefreh Region in Central Iran

This study has tried to prove the ability of remote sensing techniques to extract information necessary for preparation of geological mapping of the earth’s surface using multi-spectral satellite images which are rich sources of Earth’s surface information. In this study, the surface geological mappings of Zefreh region have been investigated through ASTER, OLI, and IRS-PAN remote sensing data. To prepare the geological map, preprocessing steps and reducing noises from data using MNF algorithm were firstly carried out. Then a set of processing algorithms and image classification methods are included; the band rationing, color composite and pixel classification based on maximum likelihood, spectral and sub-pixel classification methods of spectral angle mapper (SAM), spectral feature fitting (SFF), linear spectral differentiation (LSU), hill-shade images and automatic lineament extraction were used. Confusion matrix was formed for all classified images through control points were randomly selected from 1:25,000 map of the region to determine the accuracy of obtained results, which indicated the maximum accuracy (up to 90%) of output images. Comparing the results obtained from these methods with the map prepared by ground operations confirmed accuracy results. Finally, the surface geology and fault map of Zafreh region was produced by combining detected geological formations and tectonic lineaments.

Mapping iron oxides with Landsat-8/OLI and EO-1/Hyperion imagery from the Serra Norte iron deposits in the Carajás Mineral Province, Brazil

ABSTRACT: Mapping methods for iron oxides and clay minerals, using Landsat-8/Operational Land Imager (OLI) and Earth Observing 1 (EO-1)/Hyperion imagery integrated with airborne geophysical data, were applied in the N4, N5, and N4WS iron deposits, Serra Norte, Carajás, Brazil. Band ratios were achieved on Landsat-8/OLI imagery, allowing the recognition of the main minerals from iron deposits. The Landsat-8/OLI imagery showed a robust performance for iron oxide exploration, even in vegetated shrub areas. Feature extraction and Spectral Angle Mapper hyperspectral classification methods were carried out on EO-1/Hyperion imagery with good results for mapping high-grade iron ore, the hematite-goethite ratio, and clay minerals from regolith. The EO-1/Hyperion imagery proved an excellent tool for fast remote mineral mapping in open-pit areas, as well as mapping waste and tailing disposal facilities. An unsupervised classification was carried out on a data set consisting of EO-1/Hyperion visible near-infrared 74 bands, Landsat-8/OLI-derived Normalized Difference Vegetation Index, Laser Imaging Detection and Ranging-derived Digital Terrain Model, and high-resolution airborne geophysical data (gamma ray spectrometry, Tzz component of gradiometric gravimetry data). This multisource classification proved to be an adequate alternative for mapping iron oxides in vegetated shrub areas and to enhance the geology of the regolith and mineralized areas.

Characterization of the relative contributions from systemic physiological noise to whole-brain resting-state functional near-infrared spectroscopy data using single-channel independent component analysis.

Functional near-infrared spectroscopy (fNIRS) is a noninvasive neuroimaging technique used to measure changes in oxygenated hemoglobin (oxy-Hb) and deoxygenated hemoglobin (deoxy-Hb) in the brain. In this study, we present a decomposition approach based on single-channel independent component analysis (scICA) to investigate the contribution of physiological noise to fNIRS signals during rest. Single-channel ICA is an underdetermined decomposition method, which separates a single time series into components containing nonredundant spectral information. Using scICA, fNIRS signals from a total of 17 subjects were decomposed into the constituent physiological components. The percentage contribution of the classes of physiology to the fNIRS signals including low-frequency (LF) fluctuations, respiration, and cardiac oscillations was estimated using spectral domain classification methods. Our results show that LF oscillations accounted for 40% to 55% of total power of both the oxy-Hb and deoxy-Hb signals. Respiration and its harmonics accounted for 10% to 30% of the power, and cardiac pulsations and cardio-respiratory components accounted for 10% to 30%. We describe this scICA method for decomposing fNIRS signals, which unlike other approaches to spatial covariance reduction is applicable to both single- or multiple-channel fNIRS signals and discuss how this approach allows functionally distinct sources of noise with disjoint spectral support to be separated from obscuring systemic physiology.

Detecting of Lithological Units by Using Terrestrial Spectral Data and Remote Sensing Image

The objective of the study was to carry out an automatic classification of the lithological units of interest using the integration of remote sensing image, in which various objects are spread on, and terrestrial spectral measurement data. Only endmembers of interest are classified using spectral classification methods such as Spectral Angle Mapper. Following the identification of the types of rock and minerals, integration of remote sensing images and spectral measurement data enable spectral classification. In this study, Short Wave Infrared detector images of Advanced Spaceborne Thermal Emission and Reflection Radiometer satellite and spectroradiometer measurements were used. The study area, Golova with its geological diversity is located in the Kelkit Valley section of the North Anatolian Fault Zone in Northeast of Turkey. Seventeen rock samples were collected and their coordinates were recorded. The samples were categorized via spectral measurements on their thin sections through petrographic analyses. Marble and Meta lava with different lithological were selected as endmembers. SAM was used as the classification method that enables the analysis of the endmember with the threshold value of 0.009 radian for marble and 0.010 radian for metalava. SAM analysis was compared by visual analysis to principle component analysis, decorrelation stretch, band ratio (R: 4/7, G: 4/1, B (2/3) x (4/3)) and band combination analysis (R: 9, G: 4 and B: 5). This study demonstrates that the SAM method can be successfully used in both the integration of remote sensing image and terrestrial spectral measurement data in lithological classification. Both the endmembers of metalava and marbles were detected in the SAM results at the GPS coordinates noted down whilst collecting the rock samples for accuracy assessment.

Stellar Spectral Classification with Minimum Within-Class and Maximum Between-Class Scatter Support Vector Machine

Support Vector Machine (SVM) is one of the important stellar spectral classification methods, and it is widely used in practice. But its classification efficiencies cannot be greatly improved because it does not take the class distribution into consideration. In view of this, a modified SVM named Minimum within-class and Maximum between-class scatter Support Vector Machine (MMSVM) is constructed to deal with the above problem. MMSVM merges the advantages of Fisher’s Discriminant Analysis (FDA) and SVM, and the comparative experiments on the Sloan Digital Sky Survey (SDSS) show that MMSVM performs better than SVM.

Broadband active acoustics for synoptic studies of marine ecosystem

Marine ecosystems host a wide variety of organisms spanning many trophic levels, ranging from the smallest planktonic organisms that comprise the base of the marine food web, to large apex predator fishes, marine mammals, such as whales, and sea birds. Active acoustical systems that span a wide range of frequencies are required to quantitatively characterize these organisms, their distribution and abundance. In this presentation, we discuss the development and use of broadband acoustic systems that are optimized for studies of fish and zooplankton through use of spectral classification methods. For swimbladder-bearing fish, a low-frequency (1–6 kHz) broadband system is described that classifies fish according to the resonance of its swimbladder. For zooplankton, a high frequency (25–600 kHz) broadband system is described that classifies zooplankton either according the resonance of their gas (gas-bearing zooplankton only) or according to the transition between the Rayleigh and geometric scattering region (non-gaseous zooplankton). Because of the resolving power of the broadband signals, the echoes also tend to be non-Rayleigh, which has additional classification information. Applications of the spectral and statistical broadband acoustics methods to ecosystem research are given. [Work supported by the Office of Naval Research and the Woods Hole Oceanographic Institution.]

A Hyper spectral Images Classification Method Based on Maximum Scatter Discriminant Analysis

To overcome “small sample size problem” problem faced by some hyper spectral classification methods, the Maximum Scatter Discriminant criterion is used to analyzed hyperspectral data. Maximum Scatter Discriminant analysis searches for the project axes by maximizing the difference of between-class scatter and within-class scatter matrices, which avoid to calculate the inverse of matrices. Experiment results on Indian Pines HSI data set show that the proposed method outperforms the other methods in terms of recognition accuracy. The proposed method is an effective and feasible method for hyper pectral data classification.

Spectral mapping of the Paraíba do Sul River plume (Brazil) using multitemporal Landsat images

Coastal zones are influenced by oceanic, atmospheric, and continental forces, which make them highly vulnerable to climate and anthropogenic changes. The Parado Sul River (PSR) Estuary (Brazil), is especially affected by intensive industrial, urban, and rural activities, along its catchment area. Few works, though, have been done concerning the impacts of these changes. Remote sensing is, thus, an important and unique tool to assess the past scenes for a temporal analysis. The present work aimed to analyze spatial-temporal trends of the PSR plume from 1985 to 2009, using Landsat 5 TM images. Two spectral classification methods were used to map the river plume: maximum likelihood and spectral linear mixture analysis (SLMA). The images corresponded to the months of greatest river discharge, totalizing 11 cloud-free images. Geographical, radiometric, and atmospheric corrections were applied to the five spectral bands used for the classification. Both methods showed good results, however the SLMA provided more information of the water constituent's distribution. The sediment river plume and inner shelf phytoplankton dominated waters showed a negative trend associated with a diminishing of the river discharge. Further works concern in situ validation of the classifications, bio- optical modeling, and more investigations of climate and anthropogenic changes on the PSR.

PRINCIPAL COMPONENT ANALYSIS OF SLOAN DIGITAL SKY SURVEY STELLAR SPECTRA

We apply Principal Component Analysis (PCA) to ~100,000 stellar spectra obtained by the Sloan Digital Sky Survey (SDSS). In order to avoid strong non-linear variation of spectra with effective temperature, the sample is binned into 0.02 mag wide intervals of the g-r color (-0.20<g-r<0.90, roughly corresponding to MK spectral types A3 to K3), and PCA is applied independently for each bin. In each color bin, the first four eigenspectra are sufficient to describe the observed spectra within the measurement noise. We discuss correlations of eigencoefficients with metallicity and gravity estimated by the Sloan Extension for Galactic Understanding and Exploration (SEGUE) Stellar Parameters Pipeline. The resulting high signal-to-noise mean spectra and the other three eigenspectra are made publicly available. These data can be used to generate high quality spectra for an arbitrary combination of effective temperature, metallicity, and gravity within the parameter space probed by the SDSS. The SDSS stellar spectroscopic database and the PCA results presented here offer a convenient method to classify new spectra, to search for unusual spectra, to train various spectral classification methods, and to synthesize accurate colors in arbitrary optical bandpasses.