SEARCH · Search NASA
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.
Complete sensor/platform modelling is derived and used for the generation of synthetic data and for rectification studies of satellite scanner data. All satellite position and sensor attitude parameters are recovered. Rectification accuracy improves marginally when using more than 25 control points, and is highly sensitive to errors in image point identification.
The detection of inland water bodies from Synthetic Aperture Radar (SAR) data provides a great advantage over water detection with optical data, since SAR imaging is not impeded by cloud cover. Traditional methods of detecting water from SAR data involves using thresholding methods that can be labor intensive and imprecise. This paper describes Water Across Synthetic Aperture Radar Data (WASARD): a method of water detection from SAR data which automates and simplifies the thresholding process using machine learning on training data created from Geoscience Australia’s WOFS algorithm. Of the machine learning models tested, the Linear Support Vector Machine was determined to be optimal, with the option of training using solely the VH polarization or a combination of the VH and VV polarizations. WASARD was able to identify water in the target area with a correlation of 97% with WOFS. Sentinel-1, Open Data Cube, Earth Observations, Machine Learning, Water Detection 1. INTRODUCTION Water classification is an important function of Earth imaging satellites, as accurate remote classification of land and water can assist in land use analysis, flood prediction, climate change research, as well as a variety of agricultural applications [2]. The ability to identify bodies of water remotely via satellite is immensely cheaper than contracting surveys of the areas in question, meaning that an application that can accurately use satellite data towards this function can make valuable information available to nations which would not be able to afford it otherwise. Highly reliable applications for the remote detection of water currently exist for use with optical satellite data such as that provided by LANDSAT. One such application, Geoscience Australia’s Water Observations from Space (WOFS) has already been ported for use with the Open Data Cube [6]. However, water detection using optical data from Landsat is constrained by its relatively long revisit cycle of 16 days [5], and water detection using any optical data is constrained in that it lacks the ability to make accurate classifications through cloud cover [2]. The alternative solution which solves these problems is water detection using SAR data, which images the Earth using cloud-penetrating microwaves. Because of its advantages over optical data, much research has been done into water detection using SAR data. Traditionally, this has been done using the thresholding method, which involves picking a polarization band and labeling all pixels for which this band’s value is below a certain threshold as containing water. The thresholding method works since water tends to return a much lower backscatter value to the satellite than land [1]. However, this method can be flawed since estimating the proper threshold is often imprecise, complicated, and labor intensive for the end user. Thresholding also tends to use data from only one SAR polarization, when a combination of polarizations can provide insight into whether water is present. [2] In order to alleviate these problems, this paper presents an application for the Open Data Cube to detect water from SAR data using support vector machine (SVM) classification. 2. PLATFORM WASARD is an application for the Open Data Cube, a mechanism which provides a simple yet efficient means of ingesting, storing, and retrieving remote sensing data. Data can be ingested and made analysis ready according to whatever specifications the researcher chooses, and easily resampled to artificially alter a scene’s resolution. Currently WASARD supports water detection on scenes from ESA’s Sentinel-1 and JAXA’s ALOS. When testing WASARD, Sentinel-1 was most commonly used due to its relatively high spatial resolution and its rapid 6 day revisit cycle [5]. With minor alterations to the application's code, however, it could support data from other satellites. 3. METHODOLOGY Using supervised classification, WASARD compares SAR data to a dataset pre-classified by WOFS in order to train an SVM classifier. This classifier is then used to detect water in other SAR scenes outside the training set. Accuracy was measured according to the following metrics: Precision: a measure of what percentage of the points WASARD labels as water are truly water Recall: a measure of what percentage of the total water cover WASARD was able to identify. F1 Score: a harmonic average of the precision and recall scores Both precision and recall are calculated at the end of the training phase, when the trained classifier is compared to a testing dataset. Because the WOFS algorithm’s classifications are used as the truth values when training a WASARD classifier, when precision and recall are mentioned in this paper, they are always with respect to the values produced by WOFS on a similar scene of Landsat data, which themselves have a classification accuracy of 97% [6]. Visual representations of water identified by WASARD in this paper were produced using the function wasard_plot(), which is included in WASARD. 3.1 Algorithm Selection The machine learning model used by WASARD is the Linear Support Vector Machine (SVM). This model uses a supervised learning algorithm to develop a classifier, meaning it creates a vector which can be multiplied by the vector formed by the relevant data bands to determine whether a pixel in a SAR scene contains water. This classifier is trained by comparing data points from selected bands in a SAR scene to their respective labels, which in this case are “water” or “not water” as given by the WOFS algorithm. The SVM was selected over the Random Forest model, which outperformed the SVM in training speed, but had a greater classification time and lower accuracy, and the Multilayer Perceptron Artificial Neural Network, which had a slightly higher average accuracy than the SVM, but much greater training and classification times. Figure 1: Visual representation of the SVM Classifier. Each white point represents a pixel in a SAR scene. In Figure 1, the diagonal line separating pixels determined to be water from those determined not to be water represents the actual classification vector produced by the SVM. It is worth noting that once the model has been trained, classification of pixels is done in a similar manner as in the thresholding method. This is especially true if only one band was used to train the model. 3.1 Feature Selection Sentinel-1 collects data from two bands: the Vertical/Vertical polarization (VV) and the Vertical/Horizontal polarization (VH). When 100 SVM classifiers were created for each polarization individually, and for the combination of the two, the following results were achieved: Figure 2: Accuracy of classifiers trained using different polarization bands. Precision and Recall were measured with respect to the values produced by WOFS. Figure 2 demonstrates that using both the VV and VH bands trades slightly lower recall for significantly greater precision when compared with the VH band alone, and that using the VV band alone is inferior in both metrics. WASARD therefore defaults to using both the VV and VH bands, and includes the option to use solely the VH band. The VV polarization’s lower precision compared to the VH polarization is in contrast to results from previous research and may merit further analysis [4]. 3.2 Training a Classifier The steps in training a classifier with WASARD are 1. Selecting two scenes (one SAR, one optical) with the same spatial extents, and acquired close to each other in time, with a preference that the scenes are taken on the same day. 2. Using the WOFS algorithm to produce an array of the detected water in the scene of optical data, to be used as the labels during supervised learning 3. Data points from the selected bands from the SAR acquisition are bundled together into an array with the corresponding labels gathered from WOFS. A random sample with an equal number of points labeled “Water” and “Not Water” is selected to be partitioned into a training and a testing dataset 4. Using Scikit-Learn’s LinearSVC object, the training dataset is used to produce a classifier, which is then tested against the testing dataset to determine its precision and recall The result is a wasard_classifier object, which has the following attributes: 1. f1, recall, and precision: 3 metrics used to determine the classifier’s accuracy 2. Coefficient: Vector which the SVM uses to make its predictions. The classifier detects water when the dot product of the coefficient and the vector formed by the SAR bands is positive 3. Save(): allows a user to save a classifier to the disk in order to use it without retraining 4. wasard_classify(): Classifies an entire xarray of SAR data using the SVM classifier All of the above steps are performed automatically when the user creates a wasard_classifier object. 3.3 Classifying a Dataset Once the classifier has been created, it can be used to detect water in an xarray of SAR data using wasard_classify(). By taking the dot product of the classifier’s coefficients and the vector formed by the selected bands of SAR data, an array of predictions is constructed. A classifier can effectively be used on the same spatial extents as the ones where it was trained, or on any area with a similar landscape. While
Using an ensemble of classifiers instead of a single classifier has been shown to improve generalization performance in many pattern recognition problems. However, the extent of such improvement depends greatly on the amount of correlation among the errors of the base classifiers. Therefore, reducing those correlations while keeping the classifiers' performance levels high is an important area of research. In this article, we explore input decimation (ID), a method which selects feature subsets for their ability to discriminate among the classes and uses them to decouple the base classifiers. We provide a summary of the theoretical benefits of correlation reduction, along with results of our method on two underwater sonar data sets, three benchmarks from the Probenl/UCI repositories, and two synthetic data sets. The results indicate that input decimated ensembles (IDEs) outperform ensembles whose base classifiers use all the input features; randomly selected subsets of features; and features created using principal components analysis, on a wide range of domains.
Synthetic observations of the FeI line (6173A), which corresponds to the operational wavelength of Helioseismic and Magnetic Imager (HMI) on board of the NASA Solar Dynamics Observatory (SDO). Synthetic observations represent 4D maps for four angular distances from the solar disk center (0deg), 30deg, 45deg, and 60deg at 30degrees latitude in range +/- 0.75 Angstrom from the reference line of 6173.341 Angstrom, number of points across the spectral line is 256, spatial resolution of 100km, and temporal resolution of seconds. Synthetic data set obtained using SPINOR code (Frutiger et al. 2000). Synthetic observations computed from hydrodynamic simulations, which cover 25Mm of upper layers of the convection zone and 1Mm of atmosphere, the horizontal size of the domain of 80x80 Mm. Simulation results obtained the “StellarBox” code (Wray et al., 2015, 2018) with imposed rotation rate at 30deg latitude using F-plane approximation.
A program, designed specifically to study the respective effects of some common data problems on results obtained through stepwise iterative Fourier transformation of synthetic data with known waveform composition, was outlined. Included in this group were the problems of gaps in the data, different time-series lengths, periodic but nonsinusoidal waveforms, and noisy (low signal-to-noise) data. Results on sinusoidal data were also compared with results obtained on narrow band noise with similar characteristics. The findings showed that the analytic procedure under study can reliably reduce data in the nature of (1) sinusoids in noise, (2) asymmetric but periodic waves in noise, and (3) sinusoids in noise with substantial gaps in the data. The program was also able to analyze narrow-band noise well, but with increased interpretational problems. The procedure was shown to be a powerful technique for analysis of periodicities, in comparison with classical spectrum analysis techniques. However, informed use of the stepwise procedure nevertheless requires some background of knowledge concerning characteristics of the biological processes under study.
The purpose of this experiment was to investigate the value of altimetry data in high-precision satellite orbit determination. To accomplish this, software was developed to process laser, C-band, doppler and altimeter data singly or jointly. Initially, orbit determination studies were undertaken using synthetic data to validate the software. As data became available, preliminary experiments were carried out. When all the data became available, an intensive study was made covering a 4-day span in 1976. The results showed that even with sparse altimeter data it was possible to accurately determine the semimajor axis and eccentricity with altimeter data only. When altimeter data was supplemented with (as few as) two C-band passes, high-precision ephemerides were obtained. Using two laser passes to supplement the altimetry data did not achieve that same high precision. This is probably because the geographic location (mid-Atlantic) of the highly accurate laser data were such that they did not ideally complement the available (south Atlantic and Indian Ocean) altimeter data.
Several energy functions for synthesizing neural networks are tested on 2-D synthetic data and on Landsat-4 Thematic Mapper data. These new energy functions, designed specifically for minimizing misclassification error, in some cases yield significant improvements in classification accuracy over the standard least mean squares energy function. In addition to operating on networks with one output unit per class, a new energy function is tested for binary encoded outputs, which result in smaller network sizes. The Thematic Mapper data (four bands were used) is classified on a single pixel basis, to provide a starting benchmark against which further improvements will be measured. Improvements are underway to make use of both subpixel and superpixel (i.e. contextual or neighborhood) information in tile processing. For single pixel classification, the best neural network result is 78.7 percent, compared with 71.7 percent for a classical nearest neighbor classifier. The 78.7 percent result also improves on several earlier neural network results on this data.
Orbital measurements of neutrons by the Lunar Exploring Neutron Detector (LEND) onboard the Lunar Reconnaissance Orbiter are being used to quantify the spatial distribution of near surface hydrogen (H). Inferred H concentration maps have low signal-to-noise (SN) and image restoration (IR) techniques are being studied to enhance results. A single-blind. two-phase study is described in which four teams of researchers independently developed image restoration techniques optimized for LEND data. Synthetic lunar epithermal neutron emission maps were derived from LEND simulations. These data were used as ground truth to determine the relative quantitative performance of the IR methods vs. a default denoising (smoothing) technique. We review and used factors influencing orbital remote sensing of neutrons emitted from the lunar surface to develop a database of synthetic "true" maps for performance evaluation. A prior independent training phase was implemented for each technique to assure methods were optimized before the blind trial. Method performance was determined using several regional root-mean-square error metrics specific to epithermal signals of interest. Results indicate unbiased IR methods realize only small signal gains in most of the tested metrics. This suggests other physically based modeling assumptions are required to produce appreciable signal gains in similar low SN IR applications.
Results of digital processing of airborne X-band synthetic aperture radar (SAR) data acquired over Dade County, Florida, and Acadia Parish, Louisiana are presented. The goal was to investigate the utility of SAR data for land cover mapping and area estimation under the AgRISTARS Domestic Crops and Land Cover Project. In the case of the Acadia Paris study area, LANDSAT multispectral scanner (MSS) data were also used to form a combined SAR and MSS data set. The results of accuracy evaluation for the SAR, MSS, and SAR/MSS data using supervised classification show that the combined SAR/MSS data set results in an improved classification accuracy of the five land cover classes as compared with SAR-only and MSS-only data sets. In the case of the Dade County study area, the results indicate that both HH and VV polarization data are highly responsive to the row orientation of the row crop but not to the specific vegetation which forms the row structure. On the other hand, the HV polarization data are relatively insensitive to the orientation of row crop. Therefore, the HV polarization data may be used to discriminate the specific vegetation that forms the row structure.
The SEASAT Synthetic-Aperture Radar (SAR) system, the data processors, the extent of the image data set, and the means by which a user obtains this data are described and the data quality is evaluated. The user is alerted to some potential problems with the existing volume of SEASAT SAR image data, and allows him to modify his use of that data accordingly. Secondly, the manual focuses on the ultimate focuses on the ultimate capabilities of the raw data set and evaluates the potential of this data for processing into accurately located, amplitude-calibrated imagery of high resolution. This allows the user to decide whether his needs require special-purpose data processing of the SAR raw data.
Human behavior often consists of a series of distinct activities, each characterized by a unique pattern of interaction with the visual environment. This is true even in a restricted domain, such as a piloting an aircraft, where activities with distinct visual signatures might be things like communicating, navigating, and monitoring. We propose a novel analysis method for gaze-tracking data, to perform blind discovery of these hypothetical activities. The method is in some respects similar to recurrence analysis, but here we compare not individual fixations, but groups of fixations aggregated over a fixed time interval. The duration of this interval is a parameter that we will refer to as delta. We assume that the environment has been divided into a set of N different areas-of-interest (AOIs). For a given interval of time of duration delta, we compute the proportion of time spent fixating each AOI, resulting in an N-dimensional vector. These proportions can be converted to integer counts by multiplying by delta divided by the average fixation duration (another parameter that we fix at 280 milliseconds). We compare different intervals by computing the chi-square statistic. The p-value associated with the statistic is the likelihood of observing the data under the hypothesis that the data in the two intervals were generated by a single process with a single set of probabilities governing the fixation of each AOI. The method has been applied to approximately 100 hours of eye movement data collected from pilots in a high-fidelity B747 flight simulator, and the results have been compared to synthetic data in which the each activity is represented as first-order Markov process with random probabilities assigned to the AOIs. Randomly-generated synthetic activities can require thousands of fixations to be discriminated with statistical significance, while the human data can be clustered using averaging windows of some 10's of seconds, suggesting that the actual activities are much more narrowly focused than random Markov models.
Rough-surface inversion has significant applications in interpretation of SAR data obtained over bare soil surfaces and agricultural lands. Due to the sparsity of data and the large pixel size in SAR applications, it is not feasible to carry out inversions based on numerical scattering models. The alternative is to use parameter estimation techniques based on approximate analytical or empirical models. Hence, there are two issues to be addressed, namely, what model to choose and what estimation algorithm to apply. Here, a small perturbation model (SPM) is used to express the backscattering coefficients of the rough surface in terms of three surface parameters. The algorithm used to estimate these parameters is based on a nonlinear least-squares criterion. The least-squares optimization methods are widely used in estimation theory, but the distinguishing factor for SAR applications is incorporating the stochastic nature of both the unknown parameters and the data into formulation, which will be discussed in detail. The algorithm is tested with synthetic data, and several Newton-type least-squares minimization methods are discussed to compare their convergence characteristics. Finally, the algorithm is applied to multifrequency polarimetric SAR data obtained over some bare soil and agricultural fields. Results will be shown and compared to ground-truth measurements obtained from these areas. The strength of this general approach to inversion of SAR data is that it can be easily modified for use with any scattering model without changing any of the inversion steps. Note also that, for the same reason it is not limited to inversion of rough surfaces, and can be applied to any parameterized scattering process.
A general analytical theorem developed by van de Hulst (1946) for inverting the convolution integral is reviewed and illustrated both with synthetic data and with experimental data from time-of-flight measurements. If the undesired influence of an instrument used in an experimental measurement can be represented by the convolution integral, the original undistorted or true distribution may sometimes be recovered in postprocessing the data by means of deconvolution. Analytical deconvolution is achieved by using the coefficients from a power series representation of the distorted output distribution and a set of 'solving polynomials' which may be readily derived from the response function of the instrument.
We introduce a statistical data model and an associated optimization-based clustering algorithm which allows data vectors to belong to zero, one or several "parent" clusters. For each data vector the algorithm makes a discrete decision among these alternatives. Thus, a recursive version of this algorithm would place data clusters in a Directed Acyclic Graph rather than a tree. We test the algorithm with synthetic data generated according to the statistical data model. We also illustrate the algorithm using real data from large-scale gene expression assays.
The rank histogram filter (RHF) and the ensemble Kalman filter (EnKF) are assessed for soil moisture estimation using perfect model (identical twin) synthetic data assimilation experiments. The primary motivation is to gauge the impact on analysis quality attributable to the consideration of non-Gaussian forecast error distributions. Using the NASA Catchment land surface model, the two filters are compared at 18 globally-distributed single-catchment locations for a 10-year experiment period. It is shown that both filters yield adequate estimates of soil moisture, with the RHF having a small but significant performance advantage. Most notably, the RHF systematically increases the normalized information contribution (NIC) score of the mean absolute bias by 0.05 over that of the EnKF for surface, root-zone and profile soil moisture. The RHF also increases the NIC score for the anomaly correlation of surface soil moisture by 0.02 over that of the EnKF (at a 5% significance level). Results also demonstrate that the performance of both filters is somewhat improved when the ensemble priors are adaptively inflated to offset the negative effects of systematic errors.
The rank histogram filter (RHF) and the ensemble Kalman filter (EnKF) are assessed for soil moisture estimation using perfect model (identical twin) synthetic data assimilation experiments. The primary motivation is to gauge the impact on analysis quality attributable to the consideration of non-Gaussian forecast error distributions. Using the NASA Catchment land surface model, the two filters are compared at 18 globally distributed single-catchment locations for a 10-yr experiment period. It is shown that both filters yield adequate estimates of soil moisture, with the RHF having a small but significant performance advantage. Most notably, the RHF consistently increases the normalized information contribution (NIC) score of the mean absolute bias by 0.05 over that of the EnKF for surface, root-zone, and profile soil moisture. The RHF also increases the NIC score for the anomaly correlation of surface soil moisture by 0.02 over that of the EnKF (at a 5% significance level). Results additionally demonstrate that the performance of both filters is somewhat improved when the ensemble priors are adaptively inflated to offset the negative effects of systematic errors.
An algorithm for retrieving snow over oceans from combined cloud radar and millimeter-wave radiometer observations is developed. The algorithm involves the use of physical models to simulate cloud radar and millimeter-wave radiometer observations from basic atmospheric variables such as hydrometeor content, temperature, and relative humidity profiles and is based on an optimal estimation technique to retrieve these variables from actual observations. A high-resolution simulation of a lake-effect snowstorm by a cloud-resolving model is used to test the algorithm. That is, synthetic observations are generated from the output of the cloud numerical model, and the retrieval algorithm is applied to the synthetic data. The algorithm performance is assessed by comparing the retrievals with the reference variables used in synthesizing the observations. The synthetic observation experiment indicates good performance of the retrieval algorithm. The algorithm is also applied to real observations from the Wakasa Bay field experiment that took place over the Sea of Japan in January and February 2003. The application of the retrieval algorithm to data from the field experiment yields snow estimates that are consistent with both the cloud radar and radiometer observations.
We describe the X-ray pulse profile models we use and how we use them to analyze Neutron Star Interior Composition Explorer(NICER)observations of rotation-powered millisecond pulsars to obtain information about the mass–radius relation of neutron stars and the equation of state of the dense matter in their cores. Here we detail our modeling of the observed profile of PSR J0030+0451 that we analyzed in Miller et al. and Riley et al. and describe a cross-verification of computations of the pulse profiles of a star with R/M 3, in case stars this compact need to be considered in future analyses. We also present our early cross-verification efforts of the parameter estimation procedures used by Miller et al. and Riley et al. by analyzing two distinct synthetic data sets. Both codes yielded credible regions in the mass–radius plane that are statistically consistent with one another, and both gave posterior distributions for model parameter values consistent with the values that were used to generate the data. We also summarize the additional tests of the parameter estimation procedure of Miller et al. that used synthetic pulse profiles and the NICER pulse profile of PSR J0030+0451. We then illustrate how the precision of mass and radius estimates depends on the pulsar’s spin rate and the size of its hot spot by analyzing four different synthetic pulse profiles. Finally, we assess possible sources of systematic error in the estimates made using this technique, some of which may warrant further investigation.