Field-Based High-Quality Emissivity Spectra Measurement Using a Fourier Transform Thermal Infrared Hyperspectral Imager

Emissivity information derived from thermal infrared (TIR) hyperspectral imagery has the advantages of both high spatial and spectral resolutions, which facilitate the detection and identification of the subtle spectral features of ground targets. Despite the emergence of several different TIR hyperspectral imagers, there are still no universal spectral emissivity measurement standards for TIR hyperspectral imagers in the field. In this paper, we address the problems encountered when measuring emissivity spectra in the field and propose a practical data acquisition and processing framework for a Fourier transform (FT) TIR hyperspectral imager—the Hyper-Cam LW—to obtain high-quality emissivity spectra in the field. This framework consists of three main parts. (1) The performance of the Hyper-Cam LW sensor was evaluated in terms of the radiometric calibration and measurement noise, and a data acquisition procedure was carried out to obtain the useful TIR hyperspectral imagery in the field. (2) The data quality of the original TIR hyperspectral imagery was improved through preprocessing operations, including band selection, denoising, and background radiance correction. A spatial denoising method was also introduced to preserve the atmospheric radiance features in the spectra. (3) Three representative temperature-emissivity separation (TES) algorithms were evaluated and compared based on the Hyper-Cam LW TIR hyperspectral imagery, and the optimal TES algorithm was adopted to determine the final spectral emissivity. These algorithms are the iterative spectrally smooth temperature and emissivity separation (ISSTES) algorithm, the improved Advanced Spaceborne Thermal Emission and Reflection Radiometer temperature and emissivity separation (ASTER-TES) algorithm, and the Fast Line-of-sight Atmospheric Analysis of Hypercubes-IR (FLAASH-IR) algorithm. The emissivity results from these different methods were compared to the reference spectra measured by a Model 102F spectrometer. The experimental results indicated that the retrieved emissivity spectra from the ISSTES algorithm were more accurate than the spectra retrieved by the other methods on the same Hyper-Cam LW field data and had close consistency with the reference spectra obtained from the Model 102F spectrometer. The root-mean-square error (RMSE) between the retrieved emissivity and the standard spectra was 0.0086, and the spectral angle error was 0.0093.

On the Applicability of Laboratory Thermal Infrared Emissivity Spectra for Deconvolving Satellite Data of Opaque Volcanic Ash Plumes

The ASTER Volcanic Ash Library (AVAL) is presented, developed using quantitative laboratory thermal infrared (TIR) emission spectroscopic methods, spanning the 2000–400 cm−1 (5–25 μm wavelength) range, including the Earth’s TIR atmospheric window (8–12 μm). Each spectral suite is unique owing to the chemical composition and proportion of glass to crystals per sample and is divided into six size fractions. AVAL, used with an appropriate spectral mixture model applied to orbital multispectral TIR data, provides a unique ability to study active volcanic ash plumes. We present the first example of this application to an ash plume produced by the Sakurajima Volcano in Japan. The emissivity variations measured in ash plumes using an ever-expanding ash spectral library will provide future quantitative inputs for both atmospheric models, where the ash composition is unknown or estimated, as well as compositional probes into ongoing eruptions.

COMPARATIVE ANALYSIS OF SVM, ANN AND CNN FOR CLASSIFYING VEGETATION SPECIES USING HYPERSPECTRAL THERMAL INFRARED DATA

<p><strong>Abstract.</strong> Vegetation includes a significant class of terrestrial ecosystem. Information on tree species categorization is important for environmentalists, foresters, agriculturist, urban managers, landscape architects and biodiversity conservationist. The traditional methods of measuring and identifying tree species (i.e., through field-based survey) are time taking, laborious and costly. Remote sensing data provides an opportunity to identify and classify vegetation species over a large spatial extent. Hyperspectral remote sensing can detect the sublet spectral details among species classes and thus make it possible to differentiate vegetation species based on these subtle variations. This research examines the thermal infrared (2.5 to 14.0&amp;thinsp;&amp;mu;m) hyperspectral emissivity spectra (comprised of 3456 spectral bands) for the classification of thirteen different plant species. The use of thermal infrared hyperspectral emissivity spectra for the identification of vegetation species is very rare. Three different machine learning methods including support vector machine (SVM), artificial neural network (ANN) and convolutional neural network (CNN) are used to classify thirteen vegetation species and their performance is assessed based on their overall accuracy. The accuracy obtained by CNN, ANN and SVM is 99%, 94% and 91%, respectively. Each classifier was also tested for the advantage associated with increase in training samples or object segmentation size. Increase in the training samples improved the performance of SVM. In a nutshell, all comparative machine learning methods provide very high classification accuracy and CNN outperformed the comparative methods. This study concludes that thermal infrared hyperspectral emissivity data has the potential to discern vegetation species using state of the art machine learning and deep learning methods.</p>

Emittance Spectroscopy and Broadband Thermal Remote Sensing Applied to Phosphorite and Its Utility in Geoexploration: A Study in the Parts of Rajasthan, India

The contrast in the emissivity spectra of phosphorite and associated carbonate rock can be used as a guide to delineate phosphorite within dolomite. The thermal emissivity spectrum of phosphorite is characterized by a strong doublet emissivity feature with their absorption minima at 9 µm and 9.5 µm; whereas, host rock dolomite has relatively subdued emissivity minima at ~9 µm. Using the contrast in the emissivity spectra of phosphorite and dolomite, data obtained by the thermal bands of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) sensor were processed to delineate phosphorite within dolomite. A decorrelation stretched ASTER radiance composite could not enhance phosphorite rich zones within the dolomite host rock. However, a decorrelation stretched image composite of selected emissivity bands derived using the emissivity normalization method was suitable to enhance large surface exposures of phosphorite. We have found that the depth of the emissivity minima of phosphorite gradually has increased from dolomite to high-grade phosphorite, while low-grade phosphate has an intermediate emissivity value and the emissivity feature can be studied using three thermal bands of ASTER. In this context, we also propose a relative band depth (RBD) image using selected emissivity bands (bands 11, 12, and 13) to delineate phosphorite from the host rock. We also propose that the RBD image can be used as a proxy to estimate the relative grades of phosphorites, provided the surface exposures of phosphorite are large enough to subdue the role of intrapixel spectral mixing, which can also influence the depth of the diagnostic feature along with the grade. We have validated the phosphorite pixels of the RBD image in the field by carrying out colorimetric analysis to confirm the presence of phosphorite. The result of the study indicates the utility of the proposed relative band depth image derived using ASTER TIR bands for delineating Proterozoic carbonate-hosted phosphorite.