Hyperspectral Image Target Detection Using Deep Ensembles for Robust Uncertainty Quantification.
Abstract not provided.
Sahay, Rajeev↗
SEARCH · Search NASA
Results for “hyperspectral imaging”
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.
HyperKube: A Kubernetes Based System for the Automation of Processing and Analysis of Hyperspectral Data Obtained from Multiple Hyperspectral Imaging Systems
Hyperspectral imagery is an emerging field of technology that has enormous potential for remote and proximal sensing in numerous areas of research. The plant phenotyping community is applying this technology to advance the throughput and accuracy of plant phenotypes based on airborne and lab-based hyperspectral imaging technology. Here we report an automated processing and analysis pipeline for four different hyperspectral imaging platforms, discuss the data issues involved, and present a strategy for computing and data architecture to handle hyperspectral data.
47 OTHER INSTRUMENTATION↗
Applications Of Machine Learning to Gas Plume Analysis In Longwave Infrared Hyperspectral Images
Longwave infrared hyperspectral images can be used for gas plume analysis, as many gases exhibit distinct absorption features in this portion of the electromagnetic spectrum. In practice, accurately identifying weak gas signatures is difficult because the observed radiance is dominated by background radiance, which varies with material, temperature, and viewing conditions. Many gas plume analysis pipelines operate on single images, limiting the ability to leverage spatial and multi-view information that could enhance the analysis. The goal of this dissertation is to explore how machine learning and deep learning methods can complement classical approaches to improve gas plume identification in longwave infrared hyperspectral imagery, and to investigate the use of neural radiance fields for hyperspectral scene reconstruction.
3D Scene Reconstruction↗
Application of automated iterative target detection for standoff hyperspectral imaging
The utility of hyperspectral imaging (HSI) has been well established for a wide array of applications but has generated a need for automated screening of high volumes of large HSI cubes. We report two important automated algorithms for more efficient standoff processing: atmospheric correction and target detection. The atmospheric correction method is based on a fast asymmetric least squares approach that is applied on a pixel-by-pixel basis. Here, the correction can be applied to entire images without manually identifying regions of interest and utilizes only in-scene information, no ancillary modeling of the atmosphere is required. An iterative target detection approach is also introduced which demonstrates faster speeds relative to moving window approaches. The target detection algorithm classifies each pixel as true target detections, near target detections, clutter, and no-calls. The algorithms were tested on forty images of twenty-two solid mineral targets placed at a 14-meter standoff distance allowing general observations on expected detection performance for a variety of minerals. In addition to identifying anomalous pixels, the inclusion of “no-calls” reduced the number of false detections significantly.
47 OTHER INSTRUMENTATION↗
APPL Hyperspectral_Imaging_Dataset_for_Heritability_Analysis_in_Populus_trichocarpa
This dataset contains hyperspectral imaging data collected at the Advanced Plant Phenotyping Laboratory (APPL) at Oak Ridge National Laboratory. Natural variants of Populus trichocarpa were imaged using a high-throughput hyperspectral phenotyping pipeline to quantify spectral reflectance traits for downstream quantitative genetics analyses. The dataset includes hyperspectral image files and derived reflectance data products suitable for extracting spectral features across the measured wavelength range (e.g., VNIR and/or SWIR, depending on instrument configuration), along with associated sample metadata (e.g., genotype identifiers, experimental design factors, and imaging run identifiers). These data were generated to support analyses of broad-sense heritability of hyperspectral traits and their relationships with biochemical phenotypes (including lignin traits from Py-MBMS).
APPL↗
Toward implementing autonomous adaptive data acquisition for scanning hyperspectral imaging of biological systems
Autonomous experimentation is an emerging area of research, primarily related to autonomous vehicles, scientific combinatorial discovery approaches in materials science and drug discovery, and iterative research loops of planning, experimentation, and analysis. However, autonomous approaches developed in these contexts are difficult to apply to high-dimensional mapping technologies, such as scanning hyperspectral imaging of biological systems, due to sample complexity and heterogeneity. We briefly cover the history of adaptive sampling algorithms and surrogate modeling in order to define autonomous adaptive data acquisition as an objective-based, flexible building block for future biological imaging experimentation driven by intelligent infrastructure. We subsequently summarize the recent implementations of autonomous adaptive data acquisition (AADA) for scanning hyperspectral imaging, assess how these address the difficulties of autonomous approaches in hyperspectral imaging, and highlight the AADA design variation from a goal-oriented perspective. Finally, we present a modular AADA architecture that embeds AADA-driven flexible building blocks to address the challenge of time resolution for high-dimensional scanning hyperspectral imaging of nonequilibrium dynamical systems. In our example research-driven experimental design case, we propose an AADA infrastructure for time-resolved, noninvasive, and label-free scanning hyperspectral imaging of living biological systems. This AADA infrastructure can accurately target the correct state of the system for experimental workflows that utilize subsequent expensive, high-information-content analytical techniques.
59 BASIC BIOLOGICAL SCIENCES↗
Deep learning-driven super-resolution in Raman hyperspectral imaging: Efficient high-resolution reconstruction from low-resolution data
Deep learning (DL) has become an indispensable tool in hyperspectral data analysis, automatically extracting valuable features from complex, high-dimensional datasets. Super-resolution reconstruction, an essential aspect of hyperspectral data, involves enhancing spatial resolution, particularly relevant to low-resolution hyperspectral data. Yet, the pursuit of super-resolution in hyperspectral analysis is fraught with challenges, including acquiring ground truth high-resolution data for training, generalization, and scalability. The pressing issue of extended spectral acquisition times, notably for high-resolution scans, is a significant roadblock in hyperspectral imaging. Super-resolution methods offer a promising solution by providing higher spatial resolution data to expedite data collection and yield more efficient outcomes. This paper delves into a practical application of these concepts using Raman imaging, where spectral acquisition times can be prohibitively long. In this context, DL-based super-resolution models demonstrate their efficacy by predicting and reconstructing high-resolution Raman data from low-resolution input, eliminating the need for resource-intensive high-resolution scans. While previous work often relied on substantial high-resolution datasets, this study showcases the ability to achieve similar outcomes even with limited data, presenting a more practical and cost-effective approach. In conclusion, the results offer a glimpse into the transformative potential of this technology to streamline hyperspectral imaging applications by saving valuable time and resources through the successful generation of high-resolution data from low-resolution inputs.
42 ENGINEERING↗
New application of hyperspectral imaging to steady-state plasma observations
A new application of hyperspectral imaging (HSI) to steady-state plasma emission observations is proposed because of its prominent feature: an HSI camera records a two-dimensional image, and each spatial pixel contains spectral data typically with more than a hundred bands, while conventional digital cameras have only three bands. The characterization of an HSI camera (Specim IQ) has been performed during steady-state plasma–material interaction experiments using the linear plasma device PISCES-A. By easily subtracting the background/continuum emission in contrast to conventional filter cameras, two-dimensional images of multiple emission lines at different wavelengths are simultaneously obtained during a single measurement, demonstrating the advantage in plasma emission observations.
47 OTHER INSTRUMENTATION↗
Nearly degenerate two-color impulsive coherent Raman hyperspectral imaging
Impulsive stimulated Raman scattering (ISRS) is a robust technique for studying low frequency (<300 cm –1 ) Raman vibrational modes, but ISRS has faced difficulty in translation to an imaging modality. A primary challenge is the separation of the pump and probe pulses. Here we introduce and demonstrate a simple strategy for ISRS spectroscopy and hyperspectral imaging that uses complementary steep edge spectral filters to separate the probe beam detection from the pump and enables simple ISRS microscopy with a single-color ultrafast laser source. ISRS spectra are obtained that span from the fingerprint region down to <50 cm –1 vibrational modes. Furthermore, hyperspectral imaging and polarization-dependent Raman spectra are also demonstrated.
47 OTHER INSTRUMENTATION↗
Towards operational atmospheric correction of airborne hyperspectral imaging spectroscopy: Algorithm evaluation, key parameter analysis, and machine learning emulators
Atmospheric correction of airborne hyperspectral imaging spectroscopy (AHIS) to obtain high-quality surface reflectance is the prerequisite for remote sensing applications. Over the last decades, different atmospheric correction methods have been developed based on radiative transfer models (RTMs), however, the relative performances of different algorithms are unclear. Automated operational atmospheric correction methods to process large-volume AHIS data in a high-accurate and high-throughput manner are still lacking. Therefore, this study proposed an operational atmospheric correction pipeline for deriving surface reflectance from AHIS data. To ensure the accuracy and efficiency of the pipeline, we focused on three specific aspects: (1) selecting a suitable RTM for the development of atmospheric lookup tables (LUTs) by comparing the commercial MODerate resolution atmospheric TRANsmission (MODTRAN) and open-sourced Library for Radiative TRANsfer (LibRadTRAN) models, where the widely-used software, Atmospheric/Topographic Correction for Airborne Imagery (ATCOR), was used as benchmarks; (2) identifying key atmospheric correction parameters and determining suitable sources for parameter retrievals including AHIS, Moderate Resolution Imaging Spectroradiometer (MODIS), and AErosol RObotic NETwork (AERONET); and (3) testing the performance of using machine learning emulators to speed up the RTM-based atmospheric correction. Results indicate that (1) atmospheric correction based on MODTRAN LUTs can produce surface reflectance accurately with mean absolute errors < 0.05 and cosine similarities > 0.98 compared to field measurements, which is comparable to the software ATCOR and slightly outperforms the LibRadTRAN LUTs; (2) sobol global sensitivity analysis demonstrates that in the atmospheric correction, visibility and water vapor are two key parameters that can be accurately derived from AHIS in contrast to MODIS or AERONET data; and (3) Random Forest emulators can produce accurate estimations of surface reflectance with mean absolute errors < 0.03 and cosine similarities > 0.98 for higher processing efficiency and determine a suitable set of wavelengths for retrieving atmospheric visibility and water vapor. In conclusion, the proposed atmospheric correction pipeline also improved the four-stream radiative transfer theory for airborne applications by considering adjacent effects from airborne surrounding pixels and can also be applied for atmospheric correction of hyperspectral data from spaceborne missions.
47 OTHER INSTRUMENTATION↗
Hyperspectral imaging for real-time waste materials characterization and recovery using endmember extraction and abundance detection
Hyperspectral imaging, combined with advanced spectral unmixing techniques and artificial intelligence, offers a powerful solution for improving material identification and classification. Here, this study evaluates the effectiveness of the pixel purity index and the sequential maximum angle convex cone algorithms in extracting and validating spectral signatures from pure samples of paper components (cellulose and lignin) and plastic (polypropylene). Principal-component analysis showed that both algorithms captured nearly all relevant variance for the tested materials. Spectral signatures were compared using the spectral angle mapper, revealing high similarity in the short-wave infrared region and greater variability in the visible near-infrared range. The methodology was then applied to a disposable coffee cup to detect and quantify mixed materials, accurately estimating material abundance and object area with less than 1% error. This approach enhances material classification, supporting product verification, quality control, and automated sorting for sustainable waste management and resource recovery.
36 MATERIALS SCIENCE↗
Towards non-contact pollution monitoring in sewers with hyperspectral imaging
Monitoring water quality in sewers is challenging, particularly because state-of-the-art technologies require contact with the raw wastewater. The presence of fat, oil, grease, and solids makes automated grab sampling difficult and causes sensor fouling. To overcome these limitations, non-contact methods based on light reflectance, such as hyperspectral imaging (HSI), are gaining attention. However, HSI has never been tested for raw wastewater. To assess its accuracy for measuring pollution, we developed a laboratory setup and performed targeted experiments with a combination of raw and diluted wastewater, as well as synthetic turbidity stock solutions. We measured seven pollution variables: chemical oxygen demand, turbidity, dissolved organic compounds, ammonium, total nitrogen, phosphate, and sulphates. We used automated pixel selection and partial least squares regression to retrieve pollution information from the hyperspectral images. Our results, based on 144 samples, suggest that HSI can estimate pollution levels with a precision in the range of state-of-the-art absorbance spectrophotometric methods. Additionally, we found that the combination of pixel and wavelength selection, enabled by the hyperspectral data structure, significantly influences the performance of partial least square modelling. Overall, our findings indicate that HSI is a promising technology for non-contact monitoring of water quality in raw wastewater.
54 ENVIRONMENTAL SCIENCES↗
Optical recognition of constructs using hyperspectral imaging and detection (ORCHID)
Abstract Challenges to deep sample imaging have necessitated the development of special techniques such as spatially offset optical spectroscopy to collect signals that have travelled through several layers of tissue. However, these techniques provide only spectral information in one dimension (i.e., depth). Here, we describe a general and practical method, referred to as Optical Recognition of Constructs Using Hyperspectral Imaging and Detection (ORCHID). The sensing strategy integrates (1) the spatial offset detection concept by computationally binning 2D optical data associated with digital offsets based on selected radial pixel distances from the excitation source; (2) hyperspectral imaging using tunable filter; and (3) digital image binding and collation. ORCHID is a versatile modality that is designed to collect optical signals deep inside samples across three spatial (X, Y, Z) as well as spectral dimensions. The ORCHID method is applicable to various optical techniques that exhibit narrow-band structures, from Raman scattering to quantum dot luminescence. Samples containing surface-enhanced Raman scattering (SERS)-active gold nanostar probes and quantum dots embedded in gel were used to show a proof of principle for the ORCHID concept. The resulting hyperspectral data cube is shown to spatially locate target emitting nanoparticle volumes and provide spectral information for in-depth 3D imaging.
47 OTHER INSTRUMENTATION↗
NACHOS, a CubeSat-based high-resolution UV-Visible hyperspectral imager for remote sensing of trace gases: system overview and science objectives
Here, the Nano-satellite Atmospheric Chemistry Hyperspectral Observation System (NACHOS) is a high-throughput (f/2.9), high spectral resolution (~1.3 nm optical resolution, 0.6 nm sampling) Offner-design hyperspectral imager operating in the 300-500 nm spectral region. The 1.5U instrument payload (1U optical system, 0.5U electronics module) is hosted by a 1.5U LANL-designed CubeSat bus to comprise a 3U complete satellite. Spectroscopically similar to NASA’s Ozone Monitoring Instrument (OMI), which provides wide-field global mapping of ozone and other gases at coarse spatial resolution, NACHOS fills the complementary niche of targeted measurements at much higher spatial resolution. With 350 across-track spatial pixels and a 15-degree across-track field of view, NACHOS will provide spectral imaging at roughly 0.4 km per pixel from 500 km altitude. NACHOS incorporates highly streamlined gas-retrieval algorithms for rapid onboard processing, alleviating the need to routinely downlink massive hyperspectral data cubes. We will discuss the instrument design, challenges in achieving mechanical robustness to launch vibration in such a compact instrument, the onboard calibration system, and gas-retrieval data downlink strategy. We will also discuss potential science missions, including monitoring of NO 2 as an easily detected proxy for anthropogenic fossil-fuel greenhouse gases, monitoring lowlevel SO 2 degassing at pre-eruptive volcanoes, H 2 CO from wildfires, and characterization of aerosols. The long-term vision is for a many-satellite constellation that could provide both high spatial resolution and frequent revisits for selected targets of interest. As an initial technology demonstration of this vision, the NACHOS project is currently slated to launch two CubeSats in early 2022.
46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
Airborne hyperspectral imaging of nitrogen deficiency on crop traits and yield of maize by machine learning and radiative transfer modeling
Nitrogen is an essential nutrient that directly affects plant photosynthesis, crop yield, and biomass production for bioenergy crops, but excessive application of nitrogen fertilizers can cause environmental degradation. To achieve sustainable nitrogen fertilizer management for precision agriculture, there is an urgent need for nondestructive and high spatial resolution monitoring of crop nitrogen and its allocation to photosynthetic proteins as that changes over time. Here, we used visible to shortwave infrared (400–2400 nm) airborne hyperspectral imaging with high spatial (0.5 m) and spectral (3–5 nm) resolutions to accurately estimate critical crop traits, i.e., nitrogen, chlorophyll, and photosynthetic capacity (CO 2 -saturated photosynthesis rate, V max,27 ), at leaf and canopy scales, and to assess nitrogen deficiency on crop yield. We conducted three airborne campaigns over a maize (Zea mays L.) field during the growing season of 2019. Physically based soil-canopy Radiative Transfer Modeling (RTM) and data-driven approaches i.e. Partial-Least Squares Regression (PLSR) were used to retrieve crop traits from hyperspectral reflectance, with ground truth of leaf nitrogen, chlorophyll, V max,27 , Leaf Area Index (LAI), and harvested grain yield. To improve computational efficiency of RTMs, Random Forest (RF) was used to mimic RTM simulations to generate machine learning surrogate models RTM-RF. The results show that prior knowledge of soil background and leaf angle distribution can significantly reduce the ill-posed RTM retrieval. RTM-RF achieved a high accuracy to predict leaf chlorophyll content (R 2 = 0.73) and LAI (R 2 = 0.75). Meanwhile, PLSR exhibited better accuracy to predict leaf chlorophyll content (R 2 = 0.79), nitrogen concentration (R 2 = 0.83), nitrogen content (R 2 = 0.77), and V max,27 (R 2 = 0.69) but required measured traits for model training. We also found that canopy structure signals can enhance the use of spectral data to predict nitrogen related photosynthetic traits, as combining RTM-RF LAI and PLSR leaf traits well predicted canopy-level traits (leaf traits × LAI) including canopy chlorophyll (R 2 = 0.80), nitrogen (R 2 = 0.85) and V max,27 (R 2 = 0.82). Compared to leaf traits, we further found that canopy-level photosynthetic traits, particularly canopy V max,27 , have higher correlation with maize grain yield. This study highlights the potential for synergistic use of process-based and data-driven approaches of hyperspectral imaging to quantify crop traits that facilitate precision agricultural management to secure food and bioenergy production.
54 ENVIRONMENTAL SCIENCES↗
Detection of Foreign Materials on Broiler Breast Meat Using a Fusion of Visible Near-Infrared and Short-Wave Infrared Hyperspectral Imaging
Foreign material (FM) found on a poultry product lowers the quality and safety of the product. We developed a fusion method combining two hyperspectral imaging (HSI) modalities in the visible-near infrared (VNIR) range of 400–1000 nm and the short-wave infrared (SWIR) range of 1000–2500 nm for the detection of FMs on the surface of fresh raw broiler breast fillets. Thirty different types of FMs that could be commonly found in poultry processing plants were used as samples and prepared in two different sizes (5 × 5 mm 2 and 2 × 2 mm 2 ). The accuracies of the developed Fusion model for detecting 2 × 2 mm 2 pieces of polymer, wood, and metal were 95%, 95%, and 81%, respectively, while the detection accuracies of the Fusion model for detecting 5 × 5 mm 2 pieces of polymer, wood, and metal were all 100%. The performance of the Fusion model was higher than the VNIR- and SWIR-based detection models by 18% and 5%, respectively, when F1 scores were compared, and by 38% and 5%, when average detection rates were compared. The study results suggested that the fusion of two HSI modalities could detect FMs more effectively than a single HSI modality.
36 MATERIALS SCIENCE↗
Cross-scale sensing of field-level crop residue cover: Integrating field photos, airborne hyperspectral imaging, and satellite data
Conservation tillage practices can bring benefits to agricultural sustainability. Accurate spatial and temporal resolved information of field-scale crop residue cover, which reflects tillage intensity, is highly valuable for evaluating the outcomes of government conservation programs and voluntary ecosystem service markets, as well as facilitating agroecosystem modeling to quantify cropland biogeochemical processes. Remote sensing has the potential to cost-effectively detect crop residue cover, however, existing regional-scale studies were limited by insufficient ground truth data, scale mismatch between coarse satellite pixels and ground data, and the lack of key spectral data for detecting crop residues. Therefore, this study developed an innovative cross-sensing framework to integrate proximal sensing, airborne hyperspectral imaging, and satellite Earth Observation through deep learning to quantify field-level crop residue cover fractions at the regional scale. Specifically, we have collected intensive ground orthographic photos and conducted airborne hyperspectral surveys at corn and soybean fields of Champaign and nearby counties in Illinois, the heartland of the U.S. Corn Belt. Through semi-automatic labeling aided by ResNet-50 and superpixel image segmentation, we obtained 6719 records of ground residue fractions. With these ground data, we developed the 1-dimensional convolution neural network (CNN) model using airborne hyperspectral reflectance, which has 0.5m spatial resolution and 3–5 nm spectral resolution from 400 to 2400 nm, to predict residue fractions. By applying the CNN model to airborne pixels, we augmented “ground truth” data of crop residues and further combined them with Harmonized Landsat and Sentinel-2 (HLS) satellite data to quantify regional residue fractions at 30 m resolution. Results show that airborne hyperspectral imagery with CNN can accurately detect residue fractions (R 2 = 0.82, relative RMSE = 11.73%) to effectively generate quasi “ground truth” data to support satellite upscaling to all fields. With independent ground data for testing, we found that the ground-airborne-satellite integrative framework achieved better predictions in estimating crop residue cover (R 2 = 0.67, relative RMSE = 17.53%) than the conventional ground-satellite upscaling (R 2 = 0.22, relative RMSE = 32.09%). Here we also found that the shortwave infrared wavelengths, particularly 2100–2300 nm, are vital for predicting crop residue cover. Sentinel-2 and Landsat-8 data have a comparable capability to track residue fractions due to similar shortwave infrared wavelengths. This study highlights the high accuracy of hyperspectral imaging to detect agroecosystem tillage management practices and the advantages of cross-scale sensing to cost-effectively integrate multi-source data to quantify field-level agroecosystem variables across scales.
60 APPLIED LIFE SCIENCES↗
Batch Active Learning for Multispectral and Hyperspectral Image Segmentation Using Similarity Graphs
Abstract Graph learning, when used as a semi-supervised learning (SSL) method, performs well for classification tasks with a low label rate. We provide a graph-based batch active learning pipeline for pixel/patch neighborhood multi- or hyperspectral image segmentation. Our batch active learning approach selects a collection of unlabeled pixels that satisfy a graph local maximum constraint for the active learning acquisition function that determines the relative importance of each pixel to the classification. This work builds on recent advances in the design of novel active learning acquisition functions (e.g., the Model Change approach in arXiv:2110.07739) while adding important further developments including patch-neighborhood image analysis and batch active learning methods to further increase the accuracy and greatly increase the computational efficiency of these methods. In addition to improvements in the accuracy, our approach can greatly reduce the number of labeled pixels needed to achieve the same level of the accuracy based on randomly selected labeled pixels.
97 MATHEMATICS AND COMPUTING↗