SEBASS hyperspectral thermal infrared data: surface emissivity measurement and mineral mapping

This study focuses on mapping surface minerals using a new hyperspectral thermal infrared (TIR) sensor: the spatially enhanced broadband array spectrograph system (SEBASS). SEBASS measures radiance in 128 contiguous spectral channels in the 7.5- to 13.5-μm region with a ground spatial resolution of 2 m. In September 1999, three SEBASS flight lines were acquired over Virginia City and Steamboat Springs, Nevada. At-sensor data were corrected for atmospheric effects using an empirical method that derives the atmospheric characteristics from the scene itself, rather than relying on a predicted model. The apparent surface radiance data were reduced to surface emissivity using an emissivity normalization technique to remove the effects of temperature. Mineral maps were created with a pixel classification routine based on matching instrument- and laboratory-measured emissivity spectra, similar to methods used for other hyperspectral data sets (e.g. AVIRIS). Linear mixtures of library spectra match SEBASS spectra reasonably well, and silicate and sulfate minerals mapped remotely, agree with the dominant minerals identified with laboratory X-ray powder diffraction and spectroscopic analyses of field samples. Though improvements in instrument calibration, atmospheric correction, and information extraction would improve the ability to map more pixels, these hyperspectral TIR data nevertheless show significant advancement over multispectral thermal imaging by mapping surface materials and lithologic units with subtle spectral differences in mineralogy.     

Spectral unmixing of airborne hyperspectral data for baseline mapping of mine tailings areas

The Kam Kotia mine tailings areas near Timmins in Ontario, Canada have been generating and discharging acidic mine drainage (AMD) into the surrounding areas for more than 35 years, killing large areas of forest and polluting the local water system. This paper presents results from the remote sensing monitoring programme in the Kam Kotia mine. Hyperspectral TRW (Thompson Ramo Wooldridge Inc.) Imaging Spectrometer III data were acquired over the Kam Kotia mine and tailings areas. This paper describes (1) the data pre‐processing (noise removal, atmospheric correction, spectral smile correction, scene‐based calibration) needed to radiometrically calibrate the images and (2) a novel procedure which combines constrained spectral mixture analysis and threshold‐based classification. With this developed procedure one can retrieve fraction maps of major mine tailings‐related surface materials and hence generate a surface map separating green vegetation, transition zones, dead vegetation, and oxidized tailings, and calculate the extent (surficial area) of each of the zones. The four zones are correlated with the extent and degree of vegetation cover affected by tailings material and are interpreted to span respectively from very low to medium, high, and very high AMD pollution. This procedure can be used to monitor changes in the course of the boundary between affected zones and finally quantify the rehabilitation process in mine tailings areas with high vegetation cover.     

Nontidal wetland mapping in South Carolina using airborne multispectral scanner data

High resolution multispectral scanner (MSS) imagery of Savannah River nontidal wetlands were analyzed to identify 1) useful spectral bands for discriminating among National Wetland Inventory classes, 2) where the classes cluster in n-dimensional feature space, and 3) what wetland classification accuracies can be expected. Spectral measurements in the green (0.55–0.60 μm), red (0.65–0.70 μm), and near-infrared wavelengths (0.70–0.79 μm and 0.92–1.10 μm) provided the most useful information. Emergent marsh (both persistent and nonpersistent), scrub-shrub, mixed deciduous swamp forest, and mixed deciduous upland forest were found to cluster in somewhat predictable regions of 2- and 3-dimensional feature space. The overall classification accuracy of the Steel Creek delta study area was 83% and was assessed by comparing the remote sensing derived thematic map with 1325 linear meters of transects sampled in situ. These results suggest that high resolution aircraft MSS data can provide detailed vegetation type information for mapping both thermally affected and rejuvenating nontidal wetland in the South Carolina Savannah River Swamp System.     

Integrating visible, near-infrared and short-wave infrared hyperspectral and multispectral thermal imagery for geological mapping at Cuprite, Nevada

This study investigated the potential value of integrating hyperspectral visible, near-infrared, and short-wave infrared imagery with multispectral thermal data for geological mapping. Two coregistered aerial data sets of Cuprite, Nevada were used: Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) hyperspectral data, and MODIS/ASTER Airborne Simulator (MASTER) multispectral thermal data. Four classification methods were each applied to AVIRIS, MASTER, and a combined set. Confusion matrices were used to assess the classification accuracy. The assessment showed, in terms of kappa coefficient, that most classification methods applied to the combined data achieved a marked improvement compared to the results using either AVIRIS or MASTER thermal infrared (TIR) data alone. Spectral angle mapper (SAM) showed the best overall classification performance. Minimum distance classification had the second best accuracy, followed by spectral feature fitting (SFF) and maximum likelihood classification. The results of the study showed that SFF applied to the combination of AVIRIS with MASTER TIR data are especially valuable for identification of silicified alteration and quartzite, both of which exhibit distinctive features in the TIR region. SAM showed some advantages over SFF in dealing with multispectral TIR data, obtaining higher accuracy in discriminating low albedo volcanic rocks and limestone which do not have unique, distinguishing features in the TIR region.     

Interpretation of irrigation methods using airborne thermal infrared imagery

In areas of limited water resources for agriculture, management decisions could be enhanced by the availability of additional data about the amount and type of irrigation systems employed. Imagery produced using an HRB Singer AN/A AS 14 optical-electronic thermal infrared scanning system was found to be of great use for the interpretation of irrigation methods. Comparison of the interpreted irrigation method with ground information show agreement within 18 per cent.     

Deep feature learning versus shallow feature learning systems for joint use of airborne thermal hyperspectral and visible remote sensing data

Recently, the development of various remote sensing sensors has provided more reliable information and data for identification of different ground classes. Accordingly, multisensory fusion techniques are applied to enhance the process of information extraction from complementary airborne and spaceborne remote sensing data. Most of previous research in the literature has focused on the extraction of shallow features from a specific sensor and on classification of the resulted feature space using decision fusion systems. In recent years, Deep Learning (DL) algorithms have drawn a lot of attention in the machine learning area and have had different remote sensing applications, especially on data fusion. This study presents two different feature-learning strategies for the fusion of hyperspectral thermal infrared (HTIR) and visible remote sensing data. First, a Deep Convolutional Neural Network (DCNN)-Support Vector Machine (SVM) was utilized on the features of two datasets to provide the class labels. To validate the results with other learning strategies, a shallow feature model was used, as well. This model was based on feature fusion and decision fusion that classified and fused the two datasets. A co-registered thermal infrared hyperspectral (HTIR) and Fine Resolution Visible (Vis) RGB imagery was available from Quebec of Canada to examine the effectiveness of the proposed method. Experimental results showed that, except for the computational time, the proposed deep learning model outperformed shallow feature-based strategies in the classification performance that was based on its accuracy.     

Applying image transformation and classification techniques to airborne hyperspectral imagery for mapping Ashe juniper infestations

Ashe juniper (Juniperus ashei Buchholz) in excessive coverage reduces forage production, interferes with livestock management, and degrades watersheds and wildlife habitat on infested rangelands. The objective of this study was to apply minimum noise fraction (MNF) transformation and different classification techniques to airborne hyperspectral imagery for mapping Ashe juniper infestations. Hyperspectral imagery with 98 usable bands covering a spectral range of 475–845 nm was acquired from two Ashe juniper infested sites in central Texas. MNF transformation was applied to the hyperspectral imagery and the transformed imagery with the first 10 and 20 MNF bands was classified using four hard classifiers: minimum distance, Mahalanobis distance, maximum likelihood and spectral angle mapper (SAM). For comparison, the 10‐ and 20‐band MNF imagery was inversely transformed to noise‐reduced 98‐band imagery in the original data space, which was also classified using the four classifiers. Accuracy assessment showed that the first 10 MNF bands were sufficient for distinguishing Ashe juniper from associated plant species (mixed woody species and mixed herbaceous species) and other cover types (bare soil and water). Although the 20‐band MNF imagery provided better results for some classifications, the increase in overall accuracy was not statistically significant. Overall accuracy on the 10‐band MNF imagery varied from 88% for SAM to 93% for minimum distance for site 1 and from 84% for SAM to 94% for maximum likelihood for site 2. The 98‐band imagery derived from the 10‐band MNF imagery resulted in overall accuracy ranging from 91% for both SAM and Mahalanobis distance to 97% for maximum likelihood for site 1 and from 87% for SAM to 93% for minimum distance for site 2. Although both approaches produced comparable classification results, the MNF imagery required smaller storage space and less computing time. These results indicate that airborne hyperspectral imagery incorporated with image transformation and classification techniques can be a useful tool for mapping Ashe juniper infestations.     

Optimising the use of hyperspectral and LiDAR data for mapping reedbed habitats

Reedbeds are important habitats for supporting biodiversity and delivering a range of ecosystem services, yet reedbeds in the UK are under threat from intensified agriculture, changing land use and pollution. To develop appropriate conservation strategies, information on the distribution of reedbeds is required. Field surveys of these wetland environments are difficult, time consuming and expensive to execute for large areas. Remote sensing has the potential to replace or complement such field surveys, yet the specific application to reedbed habitats has not been fully investigated. In the present study, airborne hyperspectral and LiDAR imagery were acquired for two sites in Cumbria, UK. The research aimed to determine the most effective means of analysing hyperspectral data covering the visible, near infrared (NIR) and shortwave infrared (SWIR) regions for mapping reedbeds and to investigate the effects of incorporating image textural information and LiDAR-derived measures of canopy structure on the accuracy of reedbed delineation. Due to the high dimensionality of the hyperspectral data, three image compression algorithms were evaluated: principal component analysis (PCA), spectrally segmented PCA (SSPCA) and minimum noise fraction (MNF). The LiDAR-derived measures tested were the canopy height model (CHM), digital surface model (DSM) and the DSM-derived slope map. The SSPCA-compressed data produced the highest reedbed accuracy and processing efficiency. The optimal SSPCA dataset incorporated 12 PCs comprised of the first 3 PCs derived from each of the spectral segments: visible (392-700 nm), NIR (701-972 nm), SWIR-1 (973-1366 nm) and SWIR-2 (1530-2240 nm). Incorporating image textural measures produced a significant improvement in the classification accuracy when using MNF-compressed data, but had no impact when using the SSPCA-compressed imagery. A significant improvement (+ 11%) in the accuracy of reedbed delineation was achieved when a mask generated by applying a 3 m threshold to the LiDAR-derived CHM was used to filter the reedbed map derived from the optimal SSPCA dataset. This paper demonstrates the value in combining appropriately compressed hyperspectral imagery with LiDAR data for the effective mapping of reedbed habitats.     

A multi-scale wavelet-based temperature and emissivity separation algorithm for hyperspectral thermal infrared data

ABSTRACT

Land surface temperature and emissivity are essential variables in numerous environmental studies. This article proposes a multi-scale wavelet-based temperature and emissivity separation (MSWTES) algorithm. MSWTES is based on the fact that the high frequencies of ground-leaving radiance and derived emissivity spectra using inaccurate temperature are both closely correlated with the atmospheric downward radiance spectrum. First, surface emissivity can be decomposed by multi-scale wavelet into an optimal level that can be derived from correlation between reconstructed high frequency of ground-leaving radiance and atmospheric downward radiance. Then the ratio of high-frequency energy to low-frequency energy of surface emissivity spectrum is used to measure the degree of atmospheric downward radiance residue in the calculated emissivity spectrum as well as the disparity between the initial surface temperature and the true value. Finally, we can derive the optimal estimate of surface temperature and calculate the surface emissivity spectrum accordingly with this criterion. The MSWTES is first tested by simulation data. When a noise-equivalent spectral error of 2.5 × 10^–9 W cm^?2 sr^?1 cm is considered, the average temperature bias is 0.027 K and the root mean square error (RMSE) of emissivity is less than 0.003, except at the low and high ends of the 750–1250 cm^?1 spectral region. Then, the MSWTES is applied to field measurements. As a whole, the MSWTES achieves an RMSE of 0.01 for emissivity retrieval under most conditions, but its accuracy degrades when sample emissivity is extremely low. Meanwhile, the MSWTES is compared to the iterative spectrally smooth temperature and emissivity separation (ISSTES) algorithm. The performance of the MSWTES is better than that of the ISSTES, which demonstrates the good performance of the MSWTES.     

Ronda peridotite massif: Methodology for its geological mapping and lithological discrimination from airborne hyperspectral data

The Ronda peridotite massif in the Sierra Bermeja, southern Spain, was imaged in July 1991 by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) during the Europe 1991 Multispectral Airborne Campaign. Principal component analysis was used (i) to extract the most spectrally extreme pixels for empirical line calibration; (ii) to physically interpret the image spectral information; and (iii) to define an optimal endmember selection based on both spatial and spectroscopic characteristics. Two successive spectral mixture analyses that allow one to focus on subtle spectral variations related to bedrock and soil lithology were applied. The first spectral mixture analysis was used to identify the major surface constituents in the image and extract the geological target to be investigated, i.e. the peridotite massif; and the second one was used to model the spectral variability within the designated target. Although mineralogical variations observed in the rocks were at a sub-pixel scale for the airborne survey, spatially organized units could be identified within the major outcrops of the peridotite massif from their spectral variations. A mineralogical interpretation of these spectral variations is proposed in relation with the field observations, in terms of relative abundance variations in the pyroxene/olivine ratio among the pixels. This work demonstrates that the proposed methodology makes possible the spectral distinction of lithological units within an ultramafic body, despite the occurrence of partial vegetation cover and multiple sub-pixel mixtures. Furthermore, it shows that hyperspectral data can provide such information, resulting in a very cost-effective method of petrologic mapping.     

Multi-year sea ice mapping by thermal infrared radiometry

Marine operations in polar and subpolar regions rely on accurate sea ice information for operational planning purposes. Before venturing into operations, however, mapping of the prevailing sea ice conditions are important to feasibility analyses and planning of the operation. Multi-year sea ice information is often derived from passive microwave radiometers such as the Special Sensor Microwave Imager (SSM/I) on board the U.S. Defense Meteorological Satellite Program (DMSP) satellites, or the Scanning Multichannel Microwave Radiometer (SMMR) on board the Nimbus-7 satellite. These sources provide wide aerial coverage and all weather capability, but offer only low spatial resolution, 30 km. In contrast, the thermal infrared channel of the Advanced Very High Resolution Radiometer (AVHRR) on board the NOAA satellites, provides a 1 1km spatial resolution at nadir with a reasonable cost. A technique for extraction of multi-year sea ice information from thermal infrared AVHRR data was thus created. It relies on surface temperature differences between first-year and multi-year sea ice. Adjustments of the absolute concentration levels were made based on a regression relation between AVHRR and SMMR based multi-year sea ice concentrations. The technique is inapplicable during periods of dark, cloud cover, or melting conditions.     

Integrating visible,near-infrared and short-wave infrared hyperspectral and multispectral thermal imagery for geological mapping at Cuprite,Nevada: a rule-based system

Previous research has shown that integrating hyperspectral visible and near-infrared (VNIR) / short-wave infrared (SWIR) with multispectral thermal infrared (TIR) data can lead to improved mineral and rock identification. However, inconsistent results were found regarding the relative accuracies of different classification methods for dealing with the integrated data set. In this study, a rule-based system was developed for integration of VNIR/SWIR hyperspectral data with TIR multispectral data and evaluated using a case study of Cuprite, Nevada. Previous geological mapping, supplemented by field work and sample spectral measurements, was used to develop a generalized knowledge base for analysis of both spectral reflectance and spectral emissivity. The characteristic absorption features, albedo and the location of the spectral emissivity minimum were used to construct the decision rules. A continuum removal algorithm was used to identify absorption features from VNIR/SWIR hyperspectral data only; spectral angle mapper (SAM) and spectral feature fitting (SFF) algorithms were used to estimate the most likely rock type. The rule-based system was found to achieve a notably higher performance than the SAM, SFF, minimum distance and maximum likelihood classification methods on their own.     

Airborne thermal mapping for winter highway maintenance using the Barr and Stroud IR18 thermal video frame scanner

Abstract

The use of a portable video-based airborne thermal sensor for thermal mapping of roads is described. The design and operational characteristics of the Barr and Stroud IRI8 thermal video frame scanner are discussed together with image processing requirements for a project of this nature. Finally, a summary of the results obtained for a project carried out in the Lothian region of Scotland is presented.     

Airborne hyperspectral mapping of trees in an urban area

The management of trees in urban areas requires accurate maps, which are difficult to build in the dense patchwork of numerous material properties. Remote sensing is a useful technique that measures the response of all vegetation occurrences, including trees, when high spatial resolution is available. The continuous narrow spectral bands of hyperspectral images enable the detection of the oxygen and water content, which ensures a perfect correction of the atmospheric effect. When calibrated, sunlight reflectance images can be used to map surface chemical compositions by the detection of diagnostic and sharp absorption features. In the visible and near- infrared, the vegetation is detected by chlorophyll-a absorption features that are characteristic of the pigment content. The reflectance intensities due to the texture of leaves occur between 450 and 920 nm while the water content imprint is detectable beyond 920 nm. The sharp spectral feature intensities of the main associated pigments, not only chlorophylls, are well quantified by indices measuring a normalized difference of reflectance in a spectral interval between two bounding wavelengths. A regression line calculated on all bands within that interval ensures a low sensitivity of the indices to the smaller variations in reflectance intensity. Such unbiased indices may be combined, using successive index thresholds deduced from a training spectral library, to divide the spectra into subsets, minimizing the confusion between the numerous vegetation types with almost identical compositions. Therefore, for each subset of the spectra, a classic spectral angle mapping (SAM) method can be used on the corresponding sub-selection of the spectral library to measure angles at full spectral resolution and map tree types with great accuracy, grouped according to their spectral similarity. In this study, chemical and physical information is carefully separated. The tree crown physical properties are studied by comparing the local juxtaposition of pixel sets to a characteristic texture identifiable by image segmentation into objects. Instead of looking for objects in the reflectance image or any statistical compression of its information, a 25 channel co-image, built from 11 information layers of chemical sharp spectral feature indices and 14 information layers of SAM indices matching a spectral library of reference vegetation groups, was used. Tree canopies also present wide internal variations due to (i) a complex mixture with a background in the case of sparse foliage, or (ii) pigment content adaptation to light exposure intensity from one side to another. Both effects are minimized by using the mean spectrum of each object, assuming that less significant spectra, being at plus or minus one or two standard deviations from an object mean spectrum, would be less affected by anomalous pixel data. Thus, two overlapping hierarchic layers at the pixel scale and the object scale are available to describe the main chemistry or pigment content that identifies the vegetation types. The final classification is given by the upper layer at the object scale but in such an organization, the pixel scale layers can be used to analyse the data further and reorder them to obtain other parameters potentially useful for management purposes.     

Detection and correction of adjacency effects in hyperspectral airborne data of coastal and inland waters: the use of the near infrared similarity spectrum

A method for the detection and correction of water pixels affected by adjacency effects is presented. The approach is based on the comparison of spectra with the near infrared (NIR) similarity spectrum. Pixels affected by adjacency effects have a water-leaving reflectance spectrum with a different shape to the reference spectrum. This deviation from the similarity spectrum is used as a measure for the adjacency effect. Secondly, the correspondence with the NIR similarity spectrum is used to quantify and to correct for the contribution of the background radiance during atmospheric correction. The advantage of the approach is that it requires no a priori assumptions on the sediment load or related reflectance values in the NIR and can therefore be applied to turbid waters. The approach is tested on hyperspectral airborne data (Compact Airborne Spectrographic Imager (CASI), Airborne Hyperspectral Scanner (AHS)) acquired above coastal and inland waters at different flight altitudes and under varying atmospheric conditions. As the NIR similarity spectrum forms the basis of the approach, the method will fail for water bodies for which this similarity spectrum is no longer valid.     

Evaluation of Random Forest and Adaboost tree-based ensemble classification and spectral band selection for ecotope mapping using airborne hyperspectral imagery

Detailed land use/land cover classification at ecotope level is important for environmental evaluation. In this study, we investigate the possibility of using airborne hyperspectral imagery for the classification of ecotopes. In particular, we assess two tree-based ensemble classification algorithms: Adaboost and Random Forest, based on standard classification accuracy, training time and classification stability. Our results show that Adaboost and Random Forest attain almost the same overall accuracy (close to 70%) with less than 1% difference, and both outperform a neural network classifier (63.7%). Random Forest, however, is faster in training and more stable. Both ensemble classifiers are considered effective in dealing with hyperspectral data. Furthermore, two feature selection methods, the out-of-bag strategy and a wrapper approach feature subset selection using the best-first search method are applied. A majority of bands chosen by both methods concentrate between 1.4 and 1.8 μm at the early shortwave infrared region. Our band subset analyses also include the 22 optimal bands between 0.4 and 2.5 μm suggested in Thenkabail et al. [Thenkabail, P.S., Enclona, E.A., Ashton, M.S., and Van Der Meer, B. (2004). Accuracy assessments of hyperspectral waveband performance for vegetation analysis applications. Remote Sensing of Environment, 91, 354-376.] due to similarity of the target classes. All of the three band subsets considered in this study work well with both classifiers as in most cases the overall accuracy dropped only by less than 1%. A subset of 53 bands is created by combining all feature subsets and comparing to using the entire set the overall accuracy is the same with Adaboost, and with Random Forest, a 0.2% improvement. The strategy to use a basket of band selection methods works better. Ecotopes belonging to the tree classes are in general classified better than the grass classes. Small adaptations of the classification scheme are recommended to improve the applicability of remote sensing method for detailed ecotope mapping.     

Optimum thermal infrared bands for mapping general rock type and temperature from space

A study was carried out to determine quantitatively the number and location of spectral bands required to perform general rock type discrimination from spaceborne imaging sensors using only thermal infrared measurements. Beginning with laboratory spectra collected under idealized conditions from relatively well-characterized homogeneous samples, a radiative transfer model was used to transform ground exitance values into the corresponding spectral radiance at the top of the atmosphere. Taking sensor noise into account, analysis of these data revealed that three 1 μ wide spectral bands would permit independent estimations of rock type and sample temperature from a satellite infrared multispectral scanner. This study, which ignores the mixing of terrain elements within the instantaneous field of view of a satellite scanner, indicates that the location of three spectral bands at 8.1–9.1, 9.5–10.5, and 11.0–12.0 μ, and the employment of appropriate preprocessing to minimize atmospheric effects makes it possible to predict general rock type and temperature for a variety of atmospheric states and temperatures.     

Assessing the utility of airborne hyperspectral and LiDAR data for species distribution mapping in the coastal Pacific Northwest, Canada

To effectively manage forested ecosystems an accurate characterization of species distribution is required. In this study we assess the utility of hyperspectral Airborne Imaging Spectrometer for Applications (AISA) imagery and small footprint discrete return Light Detection and Ranging (LiDAR) data for mapping 11 tree species in and around the Gulf Islands National Park Reserve, in coastal South-western Canada. Using hyperspectral imagery yielded producer's and user's accuracies for most species ranging from > 52-95.4 and > 63-87.8%, respectively. For species dominated by definable growth stages, pixel-level fusion of hyperspectral imagery with LiDAR-derived height and volumetric canopy profile data increased both producer's (+ 5.1-11.6%) and user's (+ 8.4-18.8%) accuracies. McNemar's tests confirmed that improvements in overall accuracies associated with the inclusion of LiDAR-derived structural information were statistically significant (p < 0.05). This methodology establishes a specific framework for mapping key species with greater detail and accuracy then is possible using conventional approaches (i.e., aerial photograph interpretation), or either technology on its own. Furthermore, in the study area, acquisition and processing costs were lower than a conventional aerial photograph interpretation campaign, making hyperspectral/LiDAR fusion a viable replacement technology.     

Identification and mapping of minerals in drill core using hyperspectral image analysis of infrared reflectance spectra

A Portable Infrared Mineral Analyzer II (PIMA II) field spectrometer was used to measure infrared reflectance spectra (1·3-2·5 μm) of split drill core at 1 cm intervals in both the along-core and cross-core directions. These data were formatted into an image cube similar to that acquired by an imaging spectrometer with 600 spectral channels, and multi-spectral and hyperspectral analysis techniques were used for analysis. Colour images and enhancements provided visual displays of the spectral information, while real-time digital extraction of individual spectra allowed identification of minerals. Absorption band-depth mapping and spectral classification were used to map the spatial distribution of specific minerals in the core. Linear spectral unmixing provided estimated mineral abundances. Analysis results demonstrate that multi-spectral and hyperspectral image analysis methods can be used to produce detailed mineralogical maps of drill core. They suggest that the concepts and analytical techniques developed for analysis of hyperspectral image data can be applied to field and laboratory spectra in a variety of disciplines, and raise the question of the use of hyperspectral scanners in the laboratory.