Search NASA⌕ Search

SEARCH · Search NASA

Results for “Spectral feature extraction”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 181 records · Page 10

Development of Multimodal Few-Shot Analytics for Electron Micrographs

Recent advances in materials data analytics have provided new avenues for determining process-structure-property (PSP) linkages in a variety of materials. Machine learning techniques including few-shot learning have increased the efficiency of classifying microscopy images for the purposes of material characterization. Attempts at creating a multimodal approach can provide further improvements to current models and help extract more salient features from data. In this vein, raw spectrum data was taken to provide an additional modality to our current pyCHIP classifier. Modifications in segmentation also show potential in improving the accuracy of the pyCHIP classifier. Classifier output was analyzed using network graphs and unsupervised clustering algorithms such as spectral clustering to detect better segmentation methods than the current “chipping” approach. We suggest that the chip selection process can be automated in the future using a combination of these techniques to enable high-throughput analyses.

36 MATERIALS SCIENCE↗

Image-Based Verification Algorithms for Arms Control

PNNL is developing and evaluating radiographic image analysis techniques (active and passive) for verifying sensitive objects in a material control or warhead counting regime in which sensitive information may be acquired and processed behind an information barrier. Since sensitive image information cannot be present outside the information barrier, these techniques are necessary to extract features from the full images and reduce them to relevant parameters (attributes) of the inspected items. This evaluation can be done behind the information barrier, allowing for reporting and storage of non-sensitive attributes only. Several advances have been made to radiographic object verification algorithms, in the areas of spectral imaging for passive detectors and estimation of material density in active radiographic images. Both of these advances are pertinent in an arms control context. While passive radiographic images produced by previous work may be evaluated for the presence of emissive objects, approaches which leverage the spectroscopic potential of the detectors allow a much greater discrimination of SNM from background and other sources. Spectral passive imaging approaches to warhead discrimination and counting include specific materials and geometric arrangement localization, as well as “spectral difference” metrics which group regions with similar spectra together. These approaches may improve resolution for discrimination between materials in addition to locating SNM within surrounding shielding and/or structural elements. Previous work by our group has developed the capability to discern material density and composition in radiographic images by examining the edge transition characteristics of objects. The material construction of an object can be investigated in this way. In a weapons counting or discrimination context, unknown occultation of objects of interest, as well as additional elements of warhead construction, construction materials of varying geometry and makeup and various angles of radiograph are expected to impact algorithm performance. Advances in material discrimination algorithms are presented as a mechanism to help make these approaches robust to these sorts of variations.

Robinson, Sean M.↗

CASSIS: The Cornell atlas of Spitzer/infrared spectrograph sources. II. HIGH-Resolution observations

The Infrared Spectrograph (IRS) on board the Spitzer Space Telescope observed about 15,000 objects during the cryogenic mission lifetime. Observations provided low-resolution (R=λ/Δλ≈60−127) spectra over ≈5−38 μm and high-resolution (R≈600) spectra over 10–37 μm. The Cornell Atlas of Spitzer/IRS Sources (CASSIS) was created to provide publishable quality spectra to the community. Low-resolution spectra have been available in CASSIS since 2011, and here we present the addition of the high-resolution spectra. The high-resolution observations represent approximately one-third of all staring observations performed with the IRS instrument. While low-resolution observations are adapted to faint objects and/or broad spectral features (e.g., dust continuum, molecular bands), high-resolution observations allow more accurate measurements of narrow features (e.g., ionic emission lines) as well as a better sampling of the spectral profile of various features. Given the narrow aperture of the two high-resolution modules, cosmic ray hits and spurious features usually plague the spectra. Our pipeline is designed to minimize these effects through various improvements. A super-sampled point-spread function was created in order to enable the optimal extraction in addition to the full aperture extraction. The pipeline selects the best extraction method based on the spatial extent of the object. For unresolved sources, the optimal extraction provides a significant improvement in signal-to-noise ratio over a full aperture extraction. We have developed several techniques for optimal extraction, including a differential method that eliminates low-level rogue pixels (even when no dedicated background observation was performed). The updated CASSIS repository now includes all the spectra ever taken by the IRS, with the exception of mapping observations.

79 ASTRONOMY AND ASTROPHYSICS↗

Hierarchical classifier design in high-dimensional, numerous class cases

As progress in new sensor technology continues, increasingly high spectral resolution sensors are being developed. These sensors give more detailed and complex data for each picture element and greatly increase the dimensionality of data over past systems. Three methods for designing a decision tree classifier are discussed; a top down approach, a bottom up approach, and a hybrid approach. Three feature extraction techniques are implemented. Canonical and extended canonical techniques are mainly dependent on the mean difference between two classes. An autocorrelation technique is dependent on the correlation differences. The mathematical relationship among sample size, dimensionality, and risk value is derived.

Kim, Byungyong↗

Iterative self-organizing SCEne-LEvel sampling (ISOSCELES) for large-scale building extraction

Convolutional neural networks (CNN) provide state-of-the-art performance in many computer vision tasks, including those related to remote-sensing image analysis. Successfully training a CNN to generalize well to unseen data, however, requires training on samples that represent the full distribution of variation of both the target classes and their surrounding contexts. With remote sensing data, acquiring a sufficiently representative training set is a challenge due to both the inherent multi-modal variability of satellite or aerial imagery and the general high cost of labeling data. To address this challenge, we have developed ISOSCELES, an Iterative Self-Organizing SCEne LEvel Sampling method for hierarchical sampling of large image sets. Using affinity propagation, ISOSCELES automates the selection of highly representative training images. Compared to random sampling or using available reference data, the distribution of the training is principally data driven, reducing the chance of oversampling uninformative areas or undersampling informative ones. In comparison to manual sample selection by an analyst, ISOSCELES exploits descriptive features, spectral and/or textural, and eliminates human bias in sample selection. Using a hierarchical sampling approach, ISOSCELES can obtain a training set that reflects both between-scene variability, such as in viewing angle and time of day, and within-scene variability at the level of individual training samples. We verify the method by demonstrating its superiority to stratified random sampling in the challenging task of adapting a pre-trained model to a new image and spatial domain for country-scale building extraction. Using a pair of hand-labeled training sets comprising 1,987 sample image chips, a total of 496,000,000 individually labeled pixels, we show, across three distinct model architectures, an increase in accuracy, as measured by F1-score, of 2.2–4.2%.

42 ENGINEERING↗

A METHOD FOR INDIVIDUAL SOURCE BRIGHTNESS ESTIMATION IN SINGLE- AND MULTI-BAND DATA

We present a method of reliable extraction of the flux of individual sources from sky maps in the presence of noise and a source population in which number counts are a steeply falling function of flux. The method is an extension of a standard Bayesian procedure in the millimeter/submillimeter literature. As in the standard method, the prior applied to source flux measurements is derived from an estimate of the source counts as a function of flux, dN/dS. The key feature of the new method is that it enables reliable extraction of properties of individual sources, which previous methods in the literature do not. We first present the method for extracting individual source fluxes from data in a single observing band, then we extend the method to multiple bands, including prior information about the spectral behavior of the source population(s). The multi-band estimation technique is particularly relevant for classifying individual sources into populations according to their spectral behavior. We find that proper treatment of the correlated prior information between observing bands is key to avoiding significant biases in estimations of multi-band fluxes and spectral behavior, biases which lead to significant numbers of misclassified sources. We test the single- and multi-band versions of the method using simulated observations with observing parameters similar to that of the South Pole Telescope data used in Vieira et al.

79 ASTRONOMY AND ASTROPHYSICS↗

Identifying Dark Matter Annihilation Products In The Diffuse Gamma Ray Background

Annihilation of cosmologically distributed dark matter is predicted to produce a potentially observable flux of high energy photons. Neglecting the contribution from local structure, this signal is predicted to be virtually uniform on the sky and, in order to be identified, must compete with various extragalactic backgrounds. We focus here on unresolved blazars and discuss several techniques for separating the dark matter signal from this background. First, the spectral shapes of the signal and background are expected to differ, a feature which can be exploited with the Fisher Matrix formalism. Second, in any given angular pixel, the number of photons from blazars is drawn from a distribution which is far from Poisson. A knowledge of this distribution enhances one's ability to extract the dark matter signal, while ignorance of it can lead to the introduction of a large systematic error.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

HostSub_GP: Precise Galaxy Background Subtraction in Transient Long-slit Spectroscopy with Gaussian Processes

We present a novel host galaxy subtraction technique in long-slit spectroscopy for extragalactic transients. Unlike classic methods which generally estimate the background using simple interpolation of local galaxy flux in the 2D spectrum, our approach leverages multi-band archival images of the host galaxies to model the background emission from the galaxy in the 2D spectrum. Such imaging encodes the wavelength-dependent galaxy profile along the slit, and is readily accessible through wide-field imaging surveys. We construct a smooth prior for the 2D galaxy profile with a Gaussian process (GP) based on these reference images, and use another GP to model the correlated deviations from the prior in the observed spectrum. This enables accurate inference of the galaxy flux blended with the transient. On synthetic long-slit data of a spiral galaxy extracted from a Multi Unit Spectroscopic Explorer hyper-spectral cube, the GP method remains robust as long as the host galaxy is spatially resolved and consistently outperforms classic methods. We apply the method to archival Keck spectra of two real transients, SN 2019eix and AT 2019qiz, to further demonstrate how the method uniquely recovers weak spectral features amid strong galaxy contamination, enabling refined constraints on the properties of both transients. We have released the software implementation, HostSub_GP, a scalable toolkit that leverages JAX, with an MIT license.

79 ASTRONOMY AND ASTROPHYSICS↗

Vision-aided Monitoring and Control of Thermal Spray, Spray Forming, and Welding Processes

Vision is one of the most powerful forms of non-contact sensing for monitoring and control of manufacturing processes. However, processes involving an arc plasma or flame such as welding or thermal spraying pose particularly challenging problems to conventional vision sensing and processing techniques. The arc or plasma is not typically limited to a single spectral region and thus cannot be easily filtered out optically. This paper presents an innovative vision sensing system that uses intense stroboscopic illumination to overpower the arc light and produce a video image that is free of arc light or glare and dedicated image processing and analysis schemes that can enhance the video images or extract features of interest and produce quantitative process measures which can be used for process monitoring and control. Results of two SBIR programs sponsored by NASA and DOE and focusing on the application of this innovative vision sensing and processing technology to thermal spraying and welding process monitoring and control are discussed.

Agapakis, John E.↗

A new generic method for the semi-automatic extraction of river and road networks in low and mid-resolution satellite images

This paper addresses the problem of semi-automatic extraction of road or hydrographic networks in satellite images. For that purpose, we propose an approach combining concepts arising from mathematical morphology and hydrology. The method exploits both geometrical and topological characteristics of rivers/roads and their tributaries in order to reconstruct the complete networks. It assumes that the images satisfy the following two general assumptions, which are the minimum conditions for a road/river network to be identifiable and are usually verified in low- to mid-resolution satellite images: (i) visual constraint: most pixels composing the network have similar spectral signature that is distinguishable from most of the surrounding areas; (ii) geometric constraint: a line is a region that is relatively long and narrow, compared with other objects in the image. While this approach fully exploits local (roads/rivers are modeled as elongated regions with a smooth spectral signature in the image and a maximum width) and global (they are structured like a tree) characteristics of the networks, further directional information about the image structures is incorporated. Namely, an appropriate anisotropic metric is designed by using both the characteristic features of the target network and the eigen-decomposition of the gradient structure tensor of the image. Following, the geodesic propagation from a given network seed with this metric is combined with hydrological operators for overland flow simulation to extract the paths which contain most line evidence and identify them with the target network.

47 OTHER INSTRUMENTATION↗

SPECTRAL LAGS AND THE LAG-LUMINOSITY RELATION: AN INVESTIGATION WITH SWIFT BAT GAMMA-RAY BURSTS

Spectral lag, the time difference between the arrival of high-energy and low-energy photons, is a common feature in gamma-ray bursts (GRBs). Norris et al. reported a correlation between the spectral lag and the isotropic peak luminosity of GRBs based on a limited sample. More recently, a number of authors have provided further support for this correlation using arbitrary energy bands of various instruments. In this paper, we report on a systematic extraction of spectral lags based on the largest Swift sample to date of 31 GRBs with measured redshifts. We extracted the spectral lags for all combinations of the standard Swift hard X-ray energy bands: 15-25 keV, 25-50 keV, 50-100 keV, and 100-200 keV and plotted the time dilation corrected lag as a function of isotropic peak luminosity. The mean value of the correlation coefficient for various channel combinations is -0.68 with a chance probability of {approx}0.7 x 10{sup -3}. In addition, the mean value of the power-law index is 1.4 +- 0.3. Hence, our study lends support to the existence of a lag-luminosity correlation, albeit with large scatter.

79 ASTRONOMY AND ASTROPHYSICS↗

Optical properties of TiN thin films close to the superconductor-insulator transition.

We present the intrinsic optical properties over a broad spectral range of TiN thin films deposited on an Si/SiO{sub 2} substrate. We analyze the measured reflectivity spectra of the film-substrate multilayer structure within a well-establish procedure based on the Fresnel equation and extract the real part of the optical conductivity of TiN. We identify the metallic contribution as well as the finite energy excitations and disentangle the spectral weight distribution among them. The absorption spectrum of TiN bears some similarities with the electrodynamic response observed in the normal state of the high-temperature superconductors. Particularly, a mid-infrared feature in the optical conductivity is quite reminiscent of a pseudogap-like excitation.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

On the Investigation of Phase Fault Classification in Power Grid Signals: A Case Study for Support Vector Machines, Decision Tree and Random Forest

In monitoring the power grid, an ability to differentiate between fault types is essential to ensuring electrical safety. Accordingly, this study introduces a fault detection and classification method by considering different machine learning (ML) and feature extraction (FE) methods combinations. Specifically, the proposed method is established in two classification layers; the first layer determines the fault, and the second layer distinguishes the type of fault. Based on the proposed system model, this study seeks to determine the influential data attributes in a power grid signal using FE methods, including fast Fourier transform, power spectral density (PSD), auto-correlation, and wavelet transform (WT). A cross-comparison of the effectiveness of the Support Vector Machine (SVM), Decision Tree (DT), and Random Forest (RF) is also performed to accomplish the classification layers of the proposed method. The designed algorithm is analyzed under the various combinations of FE and ML methods, and outcomes are presented by considering the trade-off between computational complexity and prediction accuracy. The results reveal that the RF-based ML algorithm shows the most accurate classification performance with PSD, and the most time-saving of the models is the DT WT. Also, SVM emerges superior on a subsequent test of the simulated models on real-world signals.

Galbraith, Kelli↗

Exploring the SDSS Data Set with Linked Scatter Plots. I. EMP, CEMP, and CV Stars

We present the results of a search for extremely metal-poor (EMP), carbon-enhanced metal-poor (CEMP), and cataclysmic variable (CV) stars using a new exploration tool based on linked scatter plots (LSPs). Our approach is especially designed to work with very large spectrum data sets such as the SDSS, LAMOST, RAVE, and Gaia data sets, and it can be applied to stellar, galaxy, and quasar spectra. As a demonstration, we conduct our search using the SDSS DR10 data set. We first created a 3326-dimensional phase space containing nearly 2 billion measures of the strengths of over 1600 spectral features in 569,738 SDSS stars. These measures capture essentially all the stellar atomic and molecular species visible at the resolution of SDSS spectra. We show how LSPs can be used to quickly isolate and examine interesting portions of this phase space. To illustrate, we use LSPs coupled with cuts in selected portions of phase space to extract EMP stars, CEMP stars, and CV stars. We present identifications for 59 previously unrecognized candidate EMP stars and 11 previously unrecognized candidate CEMP stars. We also call attention to 2 candidate He ii emission CV stars found by the LSP approach that have not yet been discussed in the literature.

79 ASTRONOMY AND ASTROPHYSICS↗

Be/X-Ray Pulsar Binary Science with LOFT

Accretion disks are ubiquitous in astronomical sources. Accretion powered pulsars are a good test bed for accretion disk physics, because unlike for other objects, the spin of the neutron star is directly observable allowing us to see the effects of angular momentum transfer onto the pulsar. The combination of a sensitive wide-field monitor and the large area detector on LOFT will enable new detailed studies of accretion powered pulsars which I will review. RXTE observations have shown an unusually high number of Be/X-ray pulsar binaries in the SMC. Unlike binaries in the Milky Way, these systems are all at the same distance, allowing detailed population studies using the sensitive LOFT WFM, potentially providing connections to star formation episodes. For Galactic accreting pulsar systems, LOFT will allow measurement of spectral variations within individual pulses, mapping the accretion column in detail for the first time. LOFT will also provide better constraints on magnetic fields in accreting pulsars, allowing measurements of cyclotron features, observations of transitions into the centrifugal inhibition regime, and monitoring of spin-up rate vs flux correlations. Coordinated multi-wavelength observations are crucial to extracting the best science from LOFT from these and numerous other objects.

Wilson-Hodge, Colleen A.↗

Laser-Induced Recoverable Fluorescence Quenching of Perovskite Films at a Microscopic Grain Scale

Understanding the fundamental properties of metal-halide perovskite materials is driving the development of novel optoelectronic applications. Here, in this paper, we report the observation of a recoverable laser-induced fluorescence quenching phenomenon in perovskite films with a microscopic grain-scale restriction, accompanied by spectral variations. This fluorescence quenching depends on the laser intensity and the dwell time under Auger recombination dominated conditions. These features indicate that the perovskite lattice deformation may take the main responsibility for the transient and show a new aspect to understand halide perovskite photo-stability. We further modulate this phenomenon by adjusting the charge carrier recombination and extraction, revealing that efficient carrier transfer can improve the bleaching resistance of perovskite grains. Our results provide future opportunities to attain high-performance devices by tuning the perovskite lattice disorder and harvesting the energetic carriers.

36 MATERIALS SCIENCE↗

Infrared spectroscopy of electronic bands in bilayer graphene

We present infrared spectra (0.1-1 eV) of electrostatically gated bilayer graphene as a function of doping and compare it with tight-binding calculations. All major spectral features corresponding to the expected interband transitions are identified in the spectra: a strong peak due to transitions between parallel split-off bands and two onset-like features due to transitions between valence and conduction bands. A strong gate voltage dependence of these structures and a significant electron-hole asymmetry are observed that we use to extract several band parameters. The structures related to the gate-induced band gap are less pronounced in the experiment than predicted by the tight-binding model that uses parameters obtained from previous experiments on graphite and recent self-consistent band-gap calculations.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Machine learning of Kondo physics using variational autoencoders and symbolic regression

In this work, we employ variational autoencoders to extract physical insight from a dataset of one-particle Anderson impurity model spectral functions. Autoencoders are trained to find a low-dimensional, latent space representation that faithfully characterizes each element of the training set, as measured by a reconstruction error. Variational autoencoders, a probabilistic generalization of standard autoencoders, further condition the learned latent space to promote highly interpretable features. In our study, we find that the learned latent variables strongly correlate with well known, but nontrivial, parameters that characterize emergent behaviors in the Anderson impurity model. In particular, one latent variable correlates with particle-hole asymmetry, while another is in near one-to-one correspondence with the Kondo temperature, a dynamically generated low-energy scale in the impurity model. Using symbolic regression, we model this variable as a function of the known bare physical input parameters and "rediscover" the nonperturbative formula for the Kondo temperature. The machine learning pipeline we develop suggests a general purpose approach, which opens opportunities to discover new domain knowledge in other physical systems.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗