SEARCH · Engineering Papers
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.
Classification performance from LANDSAT 4 TM and MSS data is evaluated using the SECHO computer program. The data accuracy is compared using forest, corn, soybeans, bare soil, grass, water, and urban areas as classes for investigation.
Mapping basic vegetation or land cover types is a fairly common problem in remote sensing. Knowledge of the land cover type is a key input to algorithms which estimate geophysical parameters, such as soil moisture, surface roughness, leaf area index or biomass from remotely sensed data. In an earlier paper, an algorithm for fitting a simple three-component scattering model to POLSAR data was presented. The algorithm yielded estimates for surface scatter, double-bounce scatter and volume scatter for each pixel in a POLSAR image data set. In this paper, we show how the relative levels of each of the three components can be used as inputs to simple classifier for vegetation type. Vegetation classes include no vegetation cover (e.g. bare soil or desert), low vegetation cover (e.g. grassland), moderate vegetation cover (e.g. fully developed crops), forest and urban areas. Implementation of the approach requires estimates for the three components from all three frequencies available using the NASA/JPL AIRSAR, i.e. C-, L- and P-bands. The research described in this paper was carried out by the Jet Propulsion Laboratory, California Institute of Technology under a contract with the National Aeronautics and Space Administration.
Mapping basic vegetation or land cover types is a fairly common problem in remote sensing. Knowledge of the land cover type is a key input to algorithms which estimate geophysical parameters, such as soil moisture, surface roughness, leaf area index or biomass from remotely sensed data. In an earlier paper, an algorithm for fitting a simple three-component scattering model to POLSAR data was presented. The algorithm yielded estimates for surface scatter, double-bounce scatter and volume scatter for each pixel in a POLSAR image data set. In this paper, we show how the relative levels of each of the three components can be used as inputs to a simple classifier for vegetation type.
Image Reconstruction has been mostly confined to context free linear processes; the traditional continuum interpretation of digital array data uses a linear interpolator with or without an enhancement filter. Here, anti-aliasing context dependent interpretation techniques are investigated for image reconstruction. Pattern classification is applied to each neighborhood to assign it a context class; a different interpolation/filter is applied to neighborhoods of differing context. It is shown how the context dependent interpolation is computed through ensemble average statistics using high resolution training imagery from which the lower resolution image array data is obtained (simulation). A quadratic least squares (LS) context-free image quality model is described from which the context dependent interpolation coefficients are derived. It is shown how ensembles of high-resolution images can be used to capture the a priori special character of different context classes. As a consequence, a priori information such as the translational invariance of edges along the edge direction, edge discontinuity, and the character of corners is captured and can be used to interpret image array data with greater spatial resolution than would be expected by the Nyquist limit. A Gibb-like artifact associated with this super-resolution is discussed. More realistic context dependent image quality models are needed and a suggestion is made for using a quality model which now is finding application in data compression.
Electroluminescence (EL) imaging of photovoltiac (PV) modules offers high-speed, high-resolution information about device performance, affording opportunities for greater insight and efficiency in module characterization across manufacturing, research and development, and power plant operations and management. Predicting module electrical properties from EL image features is a critical step toward these applications. In this article, we demonstrate quantification of both generalized and performance mechanism-specific EL image features, using pixel intensity-based and machine learning classification algorithms. From EL image features, we build predictive models for PV module power and series resistance, using time-series current-voltage (I-V) and EL data obtained stepwise on five brands of modules spanning three Si cell types through two accelerated exposures: damp heat (DH) (85°C/85% RH) and thermal cycling (TC) (IEC 61215). Overall, 195 pairs of EL images and I-V characteristics were analyzed, yielding 11700 individual PV cell images. A convolutional neural network was built to classify cells by the severity of busbar corrosion with high accuracy (95%). Generalized power predictive models estimated the maximum power of PV modules from EL images with high confidence and an adjusted-R 2 of 0.88, across all module brands and cell types in extended DH and TC exposures. Mechanistic degradation prediction was demonstrated by quantification of busbar corrosion in EL images of three module brands in DH, and subsequent modeling of series resistance using these mechanism-specific EL image features. For modules exhibiting busbar corrosion, we demonstrated series resistance predictive models with adjusted-R 2 of up to 0.73.
Landsat MSS data were evaluated for mapping forest cover types in the northern Lower Peninsula of Michigan. The study examined seasonal variations, interpretation procedures and vegetation composition/distribution and their effect on overall classification accuracy and ability to identify individual pine species. Photographic images were used for visual interpretations while digital analysis was performed using a common (ERDAS) microcomputer image processing system. The classification schemes were evaluated using contingency tables and were ranked using the KAPPA statistic. The various classification schemes were ranked differentially according to study site location. Visual interpretation procedures ranked best, or least accurate, depending on the spatial distribution and complexity of the forest cover. Supervised classification techniques were more accurate than unsupervised clustering over all sites and seasons. Maximum likelihood classification of June data was superior to any digital classification technique of February data. The study indicates that classification accuracy is more dependent on the composition and distribution of forests in the northern lower Peninsula of Michigan than on the selection of a particular classification scheme.
A land cover-vegetation map with a base classification system for remote sensing use in a tropical island environment was produced of the island of Hawaii for the State of Hawaii to evaluate whether or not useful land cover information can be derived from Landsat TM data. In addition, an island-wide change detection mosaic combining a previously created 1977 MSS land classification with the TM-based classification was produced. In order to reach the goal of transferring remote sensing technology to State of Hawaii personnel, a pilot project was conducted while training State of Hawaii personnel in remote sensing technology and classification systems. Spectral characteristics of young island land cover types were compared to determine if there are differences in vegetation types on lava, vegetation types on soils, and barren lava from soils, and if they can be detected remotely, based on differences in pigments detecting plant physiognomic type, health, stress at senescence, heat, moisture level, and biomass. Geographic information systems (GIS) and global positioning systems (GPS) were used to assist in image rectification and classification. GIS was also used to produce large-format color output maps. An interactive GIS program was written to provide on-line access to scanned photos taken at field sites. The pilot project found Landsat TM to be a credible source of land cover information for geologically young islands, and TM data bands are effective in detecting spectral characteristics of different land cover types through remote sensing. Large agriculture field patterns were resolved and mapped successfully from wildland vegetation, but small agriculture field patterns were not. Additional processing was required to work with the four TM scenes from two separate orbits which span three years, including El Nino and drought dates. Results of the project emphasized the need for further land cover and land use processing and research. Change in vegetation composition was noted in the change detection image.
The paper presents the results from a multi-year effort to develop and validate image processing methods forselecting the best physical models based on solar image observations. The approach consists of selecting the physical models based on their agreement with coronal holes extracted from the images. Ultimately, the goal is to use physical models to predict geomagnetic storms. We decompose the problem into three subproblems: (i) coronal hole segmentation based on physical constraints, (ii) matching clusters of coronal holes between different maps, and (iii) physical map classification. For segmenting coronal holes, we develop a multi-modal method that uses segmentation maps from three different methods to initialize a level-set method that evolves the initial coronal hole segmentation to the magnetic boundary. Then, we introducea new method based on Linear Programming for matchingclusters of coronal holes. The final matching is then performedusing Random Forests. The methods were carefully validatedusing consensus maps derived from multiple readers, manualclustering, manual map classification, and method validation for50 maps. The proposed multi-modal segmentation method significantly outperformed SegNet, U-net, Henney-Harvey, and FCNby providing accurate boundary detection. Overall, the methodgave a 95.5% map classification accuracy.
An empirical method was developed to measure how solar and atmospheric conditions affect Landsat images by comparing Landsat MSS data with terrain reflectance. Terrain reflectance was measured with a four-channel radiometer designed to measure radiance in wavelength bands matched to the Landsat MSS data. Other instrumentation included a digital data logger, an irradiance meter, and a video camera with recorder. Reflectance of many terrain elements which could be registered to the Landsat digital data were measured from a low-flying aircraft. Correlations of reflectance with Landsat radiance verified a linear atmospheric model with an additive (path radiance) term and a multiplicative (transmittance) term. Coefficients from this model permit the extension of spectral signatures in computer-aided classification of Landsat images.
Cryogenic transmission electron microscopy (cryo–TEM) combined with single particle analysis (SPA) is an emerging imaging approach for soft materials. However, the accuracy of SPA–reconstructed nanostructures, particularly those formed by synthetic polymers, remains uncertain due to potential packing heterogeneity of the nanostructures. In this study, the combination of molecular dynamics (MD) simulations and image simulations is utilized to validate the accuracy of cryo–TEM 3D reconstructions of self–assembled polypeptoid fibril nanostructures. Using CryoSPARC software, image simulations, 2D classifications, ab initio reconstructions, and homogenous refinements are performed. By comparing the results with atomic models, the recovery of molecular details is assessed, heterogeneous structures are identified, and the influence of extraction location on the reconstructions is evaluated. In conclusion, these findings confirm the fidelity of single particle analysis in accurately resolving complex structural characteristics and heterogeneous structures, exhibiting its potential as a valuable tool for detailed structural analysis of synthetic polymers and soft materials.
Abstract Quantum machine learning (QML) can complement the growing trend of using learned models for a myriad of classification tasks, from image recognition to natural speech processing. There exists the potential for a quantum advantage due to the intractability of quantum operations on a classical computer. Many datasets used in machine learning are crowd sourced or contain some private information, but to the best of our knowledge, no current QML models are equipped with privacy-preserving features. This raises concerns as it is paramount that models do not expose sensitive information. Thus, privacy-preserving algorithms need to be implemented with QML. One solution is to make the machine learning algorithm differentially private, meaning the effect of a single data point on the training dataset is minimized. Differentially private machine learning models have been investigated, but differential privacy has not been thoroughly studied in the context of QML. In this study, we develop a hybrid quantum-classical model that is trained to preserve privacy using differentially private optimization algorithm. This marks the first proof-of-principle demonstration of privacy-preserving QML. The experiments demonstrate that differentially private QML can protect user-sensitive information without signficiantly diminishing model accuracy. Although the quantum model is simulated and tested on a classical computer, it demonstrates potential to be efficiently implemented on near-term quantum devices [noisy intermediate-scale quantum (NISQ)]. The approach’s success is illustrated via the classification of spatially classed two-dimensional datasets and a binary MNIST classification. This implementation of privacy-preserving QML will ensure confidentiality and accurate learning on NISQ technology.
A conceptually simple way to classify images is to directly compare test-set data and training-set data. The accuracy of this approach is limited by the method of comparison used, and by the extent to which the training-set data cover configuration space. Here we show that this coverage can be substantially increased using coarse-graining (replacing groups of images by their centroids) and stochastic sampling (using distinct sets of centroids in combination). We use the MNIST and Fashion-MNIST data sets to show that a principled coarse-graining algorithm can convert training images into fewer image centroids without loss of accuracy of classification of test-set images by nearest-neighbor classification. Distinct batches of centroids can be used in combination as a means of stochastically sampling configuration space, and can classify test-set data more accurately than can the unaltered training set. On the MNIST and Fashion-MNIST data sets this approach converts nearest-neighbor classification from a mid-ranking- to an upper-ranking member of the set of classical machine-learning techniques.
Three methods were tested for collection of the training sets needed to establish the spectral signatures of the land uses/land covers sought due to the difficulties of retrospective collection of representative ground control data. Computer preprocessing techniques applied to the digital images to improve the final classification results were geometric corrections, spectral band or image ratioing and statistical cleaning of the representative training sets. A minimal level of statistical verification was made based upon the comparisons between the airphoto estimates and the classification results. The verifications provided a further support to the selection of MSS band 5 and 7. It also indicated that the maximum likelihood ratioing technique can achieve more agreeable classification results with the airphoto estimates than the stepwise discriminant analysis.
A C-130 aircraft was flown over the Sam Houston National Forest on March 21, 1973 at 10,000 feet altitude to collect multispectral scanner (MSS) data. Existing textural and spatial automatic processing techniques were used to classify the MSS imagery into specified timber categories. Several classification experiments were performed on this data using features selected from the spectral bands and a textural transform band. The results indicate that (1) spatial post-processing a classified image can cut the classification error to 1/2 or 1/3 of its initial value, (2) spatial post-processing the classified image using combined spectral and textural features produces a resulting image with less error than post-processing a classified image using only spectral features and (3) classification without spatial post processing using the combined spectral textural features tends to produce about the same error rate as a classification without spatial post processing using only spectral features.
Particle morphology is an emerging signature that has the potential to identify the processing history of unknown nuclear materials. Using readily available scanning electron microscopes (SEM), the morphology of nearly any solid material can be measured within hours. Coupled with robust image analysis and classification methods, the morphological features can be quantified and support identification of the processing history of unknown nuclear materials. The viability of this signature depends on developing databases of morphological features, coupled with a rapid data analysis and accurate classification process. With developed reference methods, datasets, and throughputs, morphological analysis can be applied within days to (i) interdicted bulk nuclear materials (gram to kilogram quantities), and (ii) trace amounts of nuclear materials detected on swipes or environmental samples. In conclusion, this review aims to develop validated and verified analytical strategies for morphological analysis relevant to nuclear forensics.
Abstract The Earth mover’s distance (EMD) is a useful metric for image recognition and classification, but its usual implementations are not differentiable or too slow to be used as a loss function for training other algorithms via gradient descent. In this paper, we train a convolutional neural network (CNN) to learn a differentiable, fast approximation of the EMD and demonstrate that it can be used as a substitute for computing-intensive EMD implementations. We apply this differentiable approximation in the training of an autoencoder-inspired neural network (encoder NN) for data compression at the high-luminosity LHC at CERN The goal of this encoder NN is to compress the data while preserving the information related to the distribution of energy deposits in particle detectors. We demonstrate that the performance of our encoder NN trained using the differentiable EMD CNN surpasses that of training with loss functions based on mean squared error.
The author has identified the following significant results. A method has been suggested for the forecasting of the lateral eruptions of Mount Etna, through the multispectral analysis of the vegetation behavior. Unknown geological lineaments which seem to be related to deep crustal movements have been discovered using the ERTS-1 imagery. Results in the geological field were obtained in the study of the general structure of the Alpine range. In the field of official vegetation classification, ERTS-1 images were used for a preliminary study of rice fields in northern Italy. Very good experimental results have been obtained using the Skylab multispectral photographs. In the field of hydrogeology and soil type discrimination discoveries of unknown paleoriver beds have been made in the northeastern part of the Po Valley using the multispectral imagery of SL3. The superior resolution of Skylab was a fundamental element for the success of this investigation.
An eight-neighbor filter was developed for the LARSYS program. It is used in cleaning zones and sharpening boundaries during the digitization of hand-painted zone maps, in making computer-based vegetation zones more homogeneous, and in classification of natural images, such as LANDSAT or other multispectral imagery.