Search NASA⌕ Search

SEARCH · Search NASA

Results for “image classification”

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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
NASA NTRS2000

Applications of Fractal Analytical Techniques in the Estimation of Operational Scale

The observational scale and the resolution of remotely sensed imagery are essential considerations in the interpretation process. Many atmospheric, hydrologic, and other natural and human-influenced spatial phenomena are inherently scale dependent and are governed by different physical processes at different spatial domains. This spatial and operational heterogeneity constrains the ability to compare interpretations of phenomena and processes observed in higher spatial resolution imagery to similar interpretations obtained from lower resolution imagery. This is a particularly acute problem, since longterm global change investigations will require high spatial resolution Earth Observing System (EOS), Landsat 7, or commercial satellite data to be combined with lower resolution imagery from older sensors such as Landsat TM and MSS. Fractal analysis is a useful technique for identifying the effects of scale changes on remotely sensed imagery. The fractal dimension of an image is a non-integer value between two and three which indicates the degree of complexity in the texture and shapes depicted in the image. A true fractal surface exhibits self-similarity, a property of curves or surfaces where each part is indistinguishable from the whole, or where the form of the curve or surface is invariant with respect to scale. Theoretically, if the digital numbers of a remotely sensed image resemble an ideal fractal surface, then due to the self-similarity property, the fractal dimension of the image will not vary with scale and resolution, and the slope of the fractal dimension-resolution relationship would be zero. Most geographical phenomena, however, are not self-similar at all scales, but they can be modeled by a stochastic fractal in which the scaling properties of the image exhibit patterns that can be described by statistics such as area-perimeter ratios and autocovariances. Stochastic fractal sets relax the self-similarity assumption and measure many scales and resolutions to represent the varying form of a phenomenon as the pixel size is increased in a convolution process. We have observed that for images of homogeneous land covers, the fractal dimension varies linearly with changes in resolution or pixel size over the range of past, current, and planned space-borne sensors. This relationship differs significantly in images of agricultural, urban, and forest land covers, with urban areas retaining the same level of complexity, forested areas growing smoother, and agricultural areas growing more complex as small pixels are aggregated into larger, mixed pixels. Images of scenes having a mixture of land covers have fractal dimensions that exhibit a non-linear, complex relationship to pixel size. Measuring the fractal dimension of a difference image derived from two images of the same area obtained on different dates showed that the fractal dimension increased steadily, then exhibited a sharp decrease at increasing levels of pixel aggregation. This breakpoint of the fractal dimension/resolution plot is related to the spatial domain or operational scale of the phenomenon exhibiting the predominant visible difference between the two images (in this case, mountain snow cover). The degree to which an image departs from a theoretical ideal fractal surface provides clues as to how much information is altered or lost in the processes of rescaling and rectification. The measured fractal dimension of complex, composite land covers such as urban areas also provides a useful textural index that can assist image classification of complex scenes.

Emerson, Charles W.↗
NASA NTRS2000

BOREAS RSS-15 SIR-C and Landsat TM Biomass and Landcover Maps of the NSA

As part of BOREAS, the RSS-15 team conducted an investigation using SIR-C, X-SAR, and Landsat TM data for estimating total above-ground dry biomass for the SSA and NSA modeling grids and component biomass for the SSA. Relationships of backscatter to total biomass and total biomass to foliage, branch, and bole biomass were used to estimate biomass density across the landscape. The procedure involved image classification with SAR and Landsat TM data and development of simple mapping techniques using combinations of SAR channels. For the SSA, the SIR-C data used were acquired on 06-Oct-1994, and the Landsat TM data used were acquired on 02-Sep-1995. The maps of the NSA were developed from SIR-C data acquired on 13-Apr-1994. The data files are available on a CD-ROM (see document number 20010000884), or from the Oak Ridge National Laboratory (ORNL) Distributed Active Archive Center (DAAC).

Hall, Forrest G.↗
NASA NTRS2001

Low Altitude AVIRIS Data for Mapping Land Cover in Yellowstone National Park: Use of Isodata Clustering Techniques

Yellowstone National Park (YNP) contains a diversity of land cover. YNP managers need site-specific land cover maps, which may be produced more effectively using high-resolution hyperspectral imagery. ISODATA clustering techniques have aided operational multispectral image classification and may benefit certain hyperspectral data applications if optimally applied. In response, a study was performed for an area in northeast YNP using 11 select bands of low-altitude AVIRIS data calibrated to ground reflectance. These data were subjected to ISODATA clustering and Maximum Likelihood Classification techniques to produce a moderately detailed land cover map. The latter has good apparent overall agreement with field surveys and aerial photo interpretation.

Spruce, Joe↗
NASA NTRS2004

Image Analysis via Fuzzy-Reasoning Approach: Prototype Applications at NASA

A set of imaging techniques based on Fuzzy Reasoning (FR) approach was built for NASA at Kennedy Space Center (KSC) to perform complex real-time visual-related safety prototype tasks, such as detection and tracking of moving Foreign Objects Debris (FOD) during the NASA Space Shuttle liftoff and visual anomaly detection on slidewires used in the emergency egress system for Space Shuttle at the launch pad. The system has also proved its prospective in enhancing X-ray images used to screen hard-covered items leading to a better visualization. The system capability was used as well during the imaging analysis of the Space Shuttle Columbia accident. These FR-based imaging techniques include novel proprietary adaptive image segmentation, image edge extraction, and image enhancement. Probabilistic Neural Network (PNN) scheme available from NeuroShell(TM) Classifier and optimized via Genetic Algorithm (GA) was also used along with this set of novel imaging techniques to add powerful learning and image classification capabilities. Prototype applications built using these techniques have received NASA Space Awards, including a Board Action Award, and are currently being filed for patents by NASA; they are being offered for commercialization through the Research Triangle Institute (RTI), an internationally recognized corporation in scientific research and technology development. Companies from different fields, including security, medical, text digitalization, and aerospace, are currently in the process of licensing these technologies from NASA.

Dominguez, Jesus A.↗
NASA NTRS2015

Development of a Near-Real Time Hail Damage Swath Identification Algorithm for Vegetation

The Midwest is home to one of the world's largest agricultural growing regions. Between the time period of late May through early September, and with irrigation and seasonal rainfall these crops are able to reach their full maturity. Using moderate to high resolution remote sensors, the monitoring of the vegetation can be achieved using the red and near-infrared wavelengths. These wavelengths allow for the calculation of vegetation indices, such as Normalized Difference Vegetation Index (NDVI). The vegetation growth and greenness, in this region, grows and evolves uniformly as the growing season progresses. However one of the biggest threats to Midwest vegetation during the time period is thunderstorms that bring large hail and damaging winds. Hail and wind damage to crops can be very expensive to crop growers and, damage can be spread over long swaths associated with the tracks of the damaging storms. Damage to the vegetation can be apparent in remotely sensed imagery and is visible from space after storms slightly damage the crops, allowing for changes to occur slowly over time as the crops wilt or more readily apparent if the storms strip material from the crops or destroy them completely. Previous work on identifying these hail damage swaths used manual interpretation by the way of moderate and higher resolution satellite imagery. With the development of an automated and near-real time hail swath damage identification algorithm, detection can be improved, and more damage indicators be created in a faster and more efficient way. The automated detection of hail damage swaths will examine short-term, large changes in the vegetation by differencing near-real time eight day NDVI composites and comparing them to post storm imagery from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard Terra and Aqua and Visible Infrared Imaging Radiometer Suite (VIIRS) aboard Suomi NPP. In addition land surface temperatures from these instruments will be examined as for hail damage swath identification. Initial validation of the automated algorithm is based upon Storm Prediction Center storm reports but also the National Severe Storm Laboratory (NSSL) Maximum Estimated Size Hail (MESH) product. Opportunities for future work are also shown, with focus on expansion of this algorithm with pixel-based image classification techniques for tracking surface changes as a result of severe weather.

Bell, Jordan R.↗
NASA NTRS2018

Advanced Analytics and Big Earth Data

NASA's Earth Science Data Systems process, archive and distribute petabytes of Earth Observation data to a variety of end users. These end users will face dramatically increased data size in the near future, bringing about new challenges and opportunities in analyzing those data. One area of particular ferment currently is Machine Learning. Many Machine Learning methods are black boxes, limiting direct insight into the data's properties. However, they can be used for a variety of data enhancement purposes, such as parameter retrieval, data fusion and image classification and segmentation. The Earth Observing System Data and Information System is also evolving to host large data volumes in the cloud, enabling data proximal analysis. As part of this effort, an Analytics framework is being developed to support and enhance user analysis of the data. By using standards based services in the framework, diverse user communities can be served, while also allowing inter-system collaboration in the analysis process.

Cloud Computing↗
NASA NTRS2019

Image Labeler: A Web Interface to Catalog Earth Science Events

Advances in machine learning (ML) have made it possible to automatically detect Earth science phenomena from satellite imagery. While useful, ML algorithms typically require an extensive dataset containing labeled images for training. Systematic labeling and management of such datasets is quite cumbersome. With this in mind, we present the Image Labeler. Image Labeler is a fast and scalable cloud-based tool that facilitates the rapid development of Earth science event databases, in order to aid automated ML-based image classification.

Case Study↗
NASA NTRS2003

Detecting water hazards for autonomous off-road navigation

Detecting water hazards for autonomous, off-road navigation of unmanned ground vehicles is a largely unexplored problem. In this paper, we catalog environmental variables that affect the difficulty of this problem, including day vs. night operation, whether the water reflects sky or other terrain features, the size of the water body, and other factors. We briefly survey sensors that are applicable to detecting water hazards in each of these conditions. We then present analyses and results for water detection for four specific sensor cases: 1) using color image classification to recognize sky reflections in water during the day, 2) using ladar to detect the presence of water bodies and to measure their depth, 3) using short-wave infrared (SWIR) imagery to detect water bodies, as well as snow and ice, and 4) using mid-wave infrared (MWIR) imagery to recognize water bodies at night. For color imagery, we demonstrate solid results with a classifier that runs at nearly video rate on a 433 MHz processor. For ladar, we present a detailed propagation analysis that shows the limits of water body detection and depth estimation as a function of look ahead distance, water depth, and ladar wavelength. For SWIR and MWIR, we present sample imagery from a variety of data collections that illustrate the potential of these sensors. These results demonstrate significant progress on this problem.

McHenry, Mike↗
NASA NTRSDate not supplied

QuantifyML: How good is my machine learning model?

This paper presents an approach, QuantifyML, which employs model counting to assess the learnability and robustness of machine learning models. Typically the efficacy of machine learning models is determined by computing their accuracy statistically on test data sets. However, this may be misleading, if the test data is not representative of the problem that is being studied. Further, two different models may have the same accuracy on a given data set, measured statistically, but may be very different in their behavior on unseen data. Also, models with high accuracy could have poor adversarial robustness. In QuantifyML, our goal is to precisely quantify the extent to which machine learning models have learned and generalized from the given data. In QuantifyML, a trained model is translated into a C program, which is fed to the CBMC model checking tool to produce a formula in Conjunctive Normal Form (CNF), which in turn is analyzed with state-of-the-art model counters to efficiently obtain precise counts w.r.t different outputs. QuantifyML enables i) evaluating the learnability of models by comparing the counts for the outputs to ground truth, expressed as logical predicates (if available), ii) comparing the performance of different models that may be built with different machine learning algorithms (e.g., decision-trees vs. neural networks), and iii) quantifying the robustness of trained models around given inputs. Our evaluation demonstrates these applications of QuantifyML on decision trees and neural networks trained to learn relational properties of graphs, for which we know the ground truth, and to perform image classification, for which we do not have the ground truth, but we can quantify local robustness.

Deep Neural Networks↗
NASA NTRS2022

Neurosymbolic Hybrid Approach to Driver Collision Warning

There are two main algorithmic approaches to autonomous driving systems: (1) An end-to-end system in which a single deep neural network learns to map sensory input directly into appropriate warning and driving responses. (2) A mediated hybrid recognition system in which a system is created by combining independent modules that detect each semantic feature. While some researchers believe that deep learning can solve any problem, others believe that a more engineered and symbolic approach is needed to cope with complex environments with less data. Deep learning alone has achieved state-of-the-art results in many areas, from complex gameplay to predicting protein structures. In particular, in image classification and recognition, deep learning models have achieved accuracies as high as humans. But sometimes it can be very difficult to debug if the deep learning model doesn't work. Deep learning models can be vulnerable and are very sensitive to changes in data distribution. Generalization can be problematic. It's usually hard to prove why it works or doesn't. Deep learning models can also be vulnerable to adversarial attacks. Here, we combine deep learning-based object recognition and tracking with an adaptive neurosymbolic network agent, called the Non-Axiomatic Reasoning System (NARS), that can adapt to its environment by building concepts based on perceptual sequences. We achieved an improved intersection-over-union (IOU) object recognition performance of 0.65 in the adaptive retraining model compared to IOU 0.31 in the COCO data pre-trained model. We improved the object detection limits using RADAR sensors in a simulated environment, and demonstrated the weaving car detection capability by combining deep learning-based object detection and tracking with a neurosymbolic model.

Wang, Pei↗
NASA NTRSDate not supplied

Advancement of Deep Learning and Geometric Methods for Active Terrain Relative Navigation

To enhance NASA’s precision landing capabilities, in conjunction with the development of a novel active terrain relative navigation (ATRN) and terrain mapping system, denoted SHERIF, this work performed a comparative analysis between both deep-learning (DL) based and geometric approaches to hazard detection (HD) and safe-site-identification (SSI) through hardware-in-the loop testing on the Six degree-of-freedom Tendon Actuated Robot (STAR). The Standalone Hazard Evaluation and Refinement using Instrument Findings (SHERIF) system is capable of ingesting sensor data at an asynchronous rate, stitching successive terrain scans together to yield a high-resolution digital elevation map (DEM), performing absolute and relative localization using novel 3D feature extraction and matching methods, and HD/SSI activities. The DL-based HD/SSI algorithm provides a modular alternative to classical geometric approaches which have performance times that scale with map resolution. As the adoption of AI solutions become more prevalent for autonomous system decision making, it is prudent to explore the utility of such solutions in applications where they traditionally excel, such as image classification. Along with the development of a DL-based HD system, this work performed the first comparative analysis between DL and geometric approaches to HD/SSI using real sensor data from real-time testing in a relevant environment.

Davis Adams↗
NASA NTRS2024

EdgeCortix SAKURA-I Machine-Learning, PCIe Accelerator SEE Heavy Ion Test Report

To enable autonomy in space, machine-learning and computer vision applications become invaluable for sensor processing. However, these algorithms are computationally complex and unfeasible for many embedded central processing units (CPUs) and usually require external coprocessors, such as graphics processing units (GPUs) or accelerators specific to the application, including application specific integrated circuits (ASICs). In power-constrained systems, GPUs tend to consume more power than is acceptable (>40W), so lower-power accelerators have shown promise to provide the performance needed under spacecraft constraints. For radiation engineers, developing methodologies that can properly test CPUs, GPUs, and accelerators, and enable comparisons between them remains a necessary complication to solve as the devices become more complex. The methodology in this test aims to be a start in developing a baseline single-event effect (SEE) test for client-device machine learning accelerators. This category of devices do not host their own operating system. This testing campaign is a continuation of a previous 200 MeV proton test performed in January 2024. This report covers two heavy ion tests of the SAKURA-I card: one in April 2024, and one in June 2024. Additional data was needed after the April test due to ion-range issues experienced at higher linear-energy transfers (LETs). These range issues are described in more detail in Section 8. This experiment characterizes SEEs and data error susceptibility of the EdgeCortix SAKURA-I machine-learning accelerator under heavy ions. The device was monitored for single event upsets (SEUs) and single event functional interrupts (SEFIs) at the Lawrence Berkeley National Laboratory’s 88-inch cyclotron. The SAKURA-I board accelerates machine-learning inference applications on a host computer through a PCIex16 connection. For the purposes of devising an end to end automated analysis workflow for this experiment, the YOLO-V5 and SSD300 objection-detection models, and the ResNet-50, EfficientNet, and MobileNetV2 image classification models were used as a representative suite of analytical machine-learning models.

Seth S Roffe↗
NASA NTRS1992

Automating sky object classification in astronomical survey images

We describe the application of machine classification techniques to the development of an automated tool for the reduction of a large scientific data set. The 2nd Palomer Observatory Sky Survey is nearly completed. This survey provides comprehensive coverage of the northern celestial hemisphere in the form of photographic plates. The plates are being transformed into digitized images whose quality will probably not be surpassed in the next ten to twenty years. The images are expected to contain on the order of 10(exp 7) galaxies and 10(exp 8) stars. Astronomers wish to determine which of these sky objects belong to various classes of galaxies and stars. The size of this data set precludes manual analysis. Our approach is to develop a software system which integrates the functions of independently developed techniques for image processing and data classification. Digitized sky images are passed through image processing routines to identify sky objects and to extract a set of features for each object. These routines are used to help select a useful set of attributes for classifying sky objects. Then GID3* and O-BTree, two inductive learning techniques, learn classification decision trees from examples. These classifiers will be used to process the rest of the data. This paper gives an overview of the machine learning techniques used, describes the details of our specific application, and reports the initial encouraging results. The results indicate that our approach is well-suited to the problem. The primary benefits of the approach are increased data reduction throughput and consistency of classification. The classification rules which are the product of the inductive learning techniques will form an object, examinable basis for classifying sky objects. A final, not to be underestimated benefit is that astronomers will be freed from the tedium of an intensely visual task to pursue more challenging analysis and interpretation problems based on automatically cataloged data.

Fayyad, Usama M.↗
NASA NTRS1981

Contextual classification of multispectral image data: An unbiased estimator for the context distribution

A key input to a statistical classification algorithm, which exploits the tendency of certain ground cover classes to occur more frequently in some spatial context than in others, is a statistical characterization of the context: the context distribution. An unbiased estimator of the context distribution is discussed which, besides having the advantage of statistical unbiasedness, has the additional advantage over other estimation techniques of being amenable to an adaptive implementation in which the context distribution estimate varies according to local contextual information. Results from applying the unbiased estimator to the contextual classification of three real LANDSAT data sets are presented and contrasted with results from non-contextual classifications and from contextual classifications utilizing other context distribution estimation techniques.

Tilton, J. C.↗
NASA NTRS1995

Morphology classification of galaxies in CL 0939+4713 using a ground-based telescope image

Morphological classification is studied for galaxies in cluster CL 0939+4712 at z = 0.407 using simple photometric parameters obtained from a ground-based telescope image with seeing of 1-2 arcseconds full width at half maximim (FWHM). By ploting the galaxies in a plane of the concentration parameter versus mean surface brightness, we find a good correlation between the location on the plane and galaxy colors, which are known to correlate with morphological types from a recent Hubble Space Telescope (HST) study. Using the present method, we expect a success rate of classification into early and late types of about 70% or possibly more.

Fukugita, M.↗
NASA NTRS1982

Estimation of context for statistical classification of multispectral image data

Recent investigations have demonstrated the effectiveness of a contextual classifier that combines spatial and spectral information employing a general statistical approach. This statistical classification algorithm exploits the tendency of certain ground cover classes to occur more frequently in some spatial contexts than in others. Indeed, a key input to this algorithm is a statistical characterization of the context: the context function. An unbiased estimator of the context function is discussed which, besides having the advantage of statistical unbiasedness, has the additional advantage over other estimation techniques of being amenable to an adaptive implementation in which the context-function estimate varies according to local contextual information. Results from applying the unbiased estimator to the contextual classification of three real Landsat data sets are presented and contrasted with results from noncontextual classifications and from contextual classifications utilizing other context-function estimation techniques.

Tilton, J. C.↗
NASA NTRS1981

Contextual classification of multispectral image data - An unbiased estimator for the context distribution

Recent investigations have demonstrated the effectiveness of a contextual classifier that combines spatial and spectral information employing a general statistical approach. This statistical classification algorithm exploits the tendency of certain ground-cover classes to occur more frequently in some spatial contexts than in others. Indeed, a key input to this algorithm is a statistical characterization of the context: the context distribution. Here a discussion is given of an unbiased estimator of the context distribution which, besides having the advantage of statistical unbiasedness, has the additional advantage over other estimation techniques of being amenable to an adaptive implementation in which the context distribution estimate varies according to local contextual information. Results from applying the unbiased estimator to the contextual classification of three real Landsat data sets are presented and contrasted with results from noncontextual classifications and from contextual classifications utilizing other context distribution estimation techniques.

Tilton, J. C.↗
NASA NTRS1981

Comparison of wheat classification accuracy using different classifiers of the image-100 system

Classification results using single-cell and multi-cell signature acquisition options, a point-by-point Gaussian maximum-likelihood classifier, and K-means clustering of the Image-100 system are presented. Conclusions reached are that: a better indication of correct classification can be provided by using a test area which contains various cover types of the study area; classification accuracy should be evaluated considering both the percentages of correct classification and error of commission; supervised classification approaches are better than K-means clustering; Gaussian distribution maximum likelihood classifier is better than Single-cell and Multi-cell Signature Acquisition Options of the Image-100 system; and in order to obtain a high classification accuracy in a large and heterogeneous crop area, using Gaussian maximum-likelihood classifier, homogeneous spectral subclasses of the study crop should be created to derive training statistics.

Dejesusparada, N.↗