Search

The Multi-angle Imaging SpectroRadiometer (MISR) data products now include a scene classification for each 1.1-km pixel that was developed using Support Vector Machines (SVMs), a cutting-edge machine learning technique for supervised classification. Using a combination of spectral, angular, and texture features, each pixel is classified as land, water, cloud, aerosol, or snow/ice, with the aerosol class further divided into smoke, dust, and other aerosols. The classifier was trained by MISR scientists who labeled hundreds of scenes using a custom interactive tool that showed them the results of the training in real time, making the process significantly faster. Preliminary validation shows that the accuracy of the classifier is approximately 81% globally at the 1.1-km pixel level. Applications of this classifier include global studies of cloud and aerosol distribution, as well as data mining applications such as searching for smoke plumes. This is one of the largest and most ambitious operational uses of machine learning techniques for a remote-sensing instrument, and the success of this system will hopefully lead to further use of this approach.

The Radiometric Camera-by-Camera Cloud Mask (RCCM) is archived at the NASA Langley Distributed Active Archive Center as one of the standard products from the Multi-angle Imaging SpectroRadiometer (MISR) mission. The RCCM algorithm applied over land surfaces uses an Automated Threshold Selection Algorithm (ATSA) to derive thresholds that are applied to a cloud masking test to determine whether a given image pixel is clear or contains cloud. In this article, we established a framework for the selection of ATSA and the cloud masking tests, which is not only suitable for the RCCM over land, but cloud detection for other satellite missions. Using this framework, we have undertaken the largest comparison of existing histogram-based ATSAs (16 in total) and applied them to four cloud masking tests that can be constructed from the MISR radiances, namely the red channel bidirectional reflectance function (BRF), the standard deviation (STDV) of the red channel BRF, the normalized difference vegetation index (NDVI), and a parameter D that is constructed by optimizing the information from NDVI and red channel BRF for cloud detection. The cloud masking tests and ATSAs are applied to 35 MISR scenes from six snow-free land surface types. To evaluate their performance, reference cloud masks are constructed for the 35 scenes using interactive, supervised learning, visualization software. Independent of the ATSA and as a single cloud masking test, D performed the best in terms of having the lowest misclassification rate using the best possible threshold, the highest bimodal rate in the shape of the histograms derived from the 35 scenes, and the least sensitivity to errors in the choice of threshold. Of the 16 ATSAs, the methods of Li and Lee [Li, C.H., and Lee, C.K., (1993). Minimum cross-entropy thresholding. Pattern Recognition, 26(4), 617-625.] and Pal and Bhandari [Pal, N. R., and Bhandari, D., (1993). Image thresholding: some new techniques. Signal Processing, 33, 139–158.] performed the best when applied to D , with essentially unbiased performance and a root mean square of 15% when compared to cloud masks using the best possible thresholds. It is recommended that increased performance of the RCCM-land algorithm can be had through an increase in the space–time sampling used to generate histograms of D and the addition of a STDV cloud masking test to improve the detection of small cumulus clouds.

Satellites provide unique perspectives on aerosol global and regional spatial and temporal distributions, and offer compelling evidence that visibility and air quality are affected by particulate matter transported over long distances. The heights at which emissions are injected into the atmosphere are major factors governing downwind dispersal. In order to better understand the environmental factors determining injection heights of smoke plumes from wildfires, we have developed a prototype system for automatically searching through several years of MISR and MODIS data to locate fires and the associated smoke plumes and to retrieve injection heights and other relevant measurements from them. We are refining this system and assembling a statistical database, aimed at understanding how injection height relates to the fire severity and local weather conditions. In this paper we focus on our working proof-of-concept system that demonstrates how machine-learning and data mining methods aid in processing of massive volumes of satellite data. Automated algorithms for distinguishing smoke from clouds and other aerosols, identifying plumes, and extracting height data are described. Preliminary results are presented from application to MISR and MODIS data collected over North America during the summer of 2004.

The Multi-angle Imaging SpectroRadiometer (MISR) instrument aboard Terra acquires imagery at 275-m resolution at nine angles ranging from 0° to 70° off-nadir. This multi-angle capability facilitates the stereoscopic retrieval of smoke heights associated with near-source plumes. A new visualization and analysis program called MISR INteractive eXplorer (MINX) takes advantage of wind-direction information inherent in smoke plumes from active fires to determine plume heights and wind speeds at higher resolution and with greater accuracy than provided by the standard, operational MISR product. Among the software tool's many features are several designed for in-depth study of plumes, including animations of the nine MISR camera images that provide a visual 3-D perspective, and interactive digitization of plumes in order to automatically retrieve heights and winds. Aerosol properties from MISR, and fire power based on infrared brightness temperatures from MODIS (also on Terra) are archived along with the retrieved height and wind data. MINX retrievals have sufficient spatial detail to provide valuable input to studies of plume dynamics as well as large-scale climatological studies. Current efforts are focusing on fires in North America, but application to other areas of the world is also envisioned. Case study examples will be presented to illustrate MINX capabilities.

SciFlo is a system for Scientific Knowledge Creation on the Grid using a Semantically-Enabled Dataflow Execution Environment. SciFlo leverages Simple Object Access Protocol (SOAP) Web Services and the Grid Computing standards (WS-* & Globus Alliance toolkits), and enables scientists to do multi-instrument Earth Science by assembling reusable Web Services and native executables into a distributed computing flow (operator graph). SciFlo’s XML dataflow documents can be a mixture of concrete operators (fully bound operations) and abstract template operators (late binding via semantic lookup). All data objects and operators can be both simply typed (simple and complex types in XML schema) and semantically typed (linked to OWL ontologies). We will demonstrate a version of the SciFlo workflow engine executing a variety of workflow documents. [Note: This document is a quick description of the proposed demo, not a fully referenced and defended paper.]

The Multi-angle Imaging SpectroRadiometer (MISR) currently provides three independently derived cloud mask products at 1.1 km spatial resolution. The Radiometric Camera-by-camera Cloud Mask (RCCM) is terrain-referenced and calculated for each of the nine MISR cameras, the Stereoscopically Derived Cloud Mask (SDCM) is feature-projected and uses radiances from one pair of the MISR cameras, and the Angular Signature Cloud Mask (ASCM) uses a band-differenced angular signature based on the two most oblique cameras viewing forward scattering radiation. While each mask has been extensively validated, each having its own strengths and weaknesses, there has been no effort to combine the strengths of all of the masks to create a single consensus product. We present an algorithm which addresses the problem and produces a so called "consensus cloud mask" of improved performance, and elaborate on further cloud climatology applications.

The Onboard Autonomous Science Investigation System (OASIS) is a technology for increasing science return during rover traverses by prioritizing data onboard, and identifying and reacting to unanticipated science opportunities. Rovers of the future will have the capacity to collect more data than can be downlinked back to Earth. OASIS can increase mission science return by carefully selecting the data with the highest science interest for downlink. These rovers may also be required to traverse long distances with little to no interaction with the science team on Earth. OASIS can act as a geologist's assistant and can autonomously direct the rover to take additional measurements of "interesting" rocks. The importance of characterizing the terrain along these traverses, a study that is now becoming known as traverse science, increases with the distances the rover must travel. This paper provides a brief overview of the entire OASIS system and how it analyzes one type of data - grayscale images taken by the rover for engineering and hazard avoidance purposes. Although the OASIS system can apply the same type of analysis to different data types, such as color images, hyperspectral images or point spectrometer data, we will only focus on grayscale images here. The paper also describes the latest advances in two key aspects of the system: image prioritization and the science alert. In image prioritization, we combine the results from three distinct prioritization methods to arrive at an overall downlink ranking of the images collected during a traverse. The science alert is a capability that enables the rover to identify and react to a pre-specified, and scientifically significant, signature. Once this signature has been detected via the onboard science analysis component, the planning and scheduling module updates the rover command sequence to stop the traverse and signal Earth of the find. If there is sufficient time and onboard resources before the next downlink opportunity, additional data samples of the target may be autonomously collected.

Rover traverse distances are increasing at a faster rate than downlink capacity is increasing. As this trend continues, the quantity of data that can he returned to Earth per meter traversed is reduced. The capacity of the rover to collect data, however, remains high. This circumstance leads to an opportunity to increase mission science return by carefully selecting the data with the highest science interest for downlink. We have developed an onboard science analysis technology for increasing science return from missions. Our technology evaluates the geologic data gather by the rover. This analysis is used to prioritize the data for transmission, so that the data with the highest science value is transmitted to Earth. In addition, the onboard analysis results are used to identify science opportunities. A planning and scheduling component of the system enables the rover to take advantage of the identified science opportunity. Although our tcchniqucs arc applicable to a wide range of data modalities, our initial emphasis has focused on image analysis, since images consume a large percentage of downlink bandwidth. We have further focused our foundational work on rocks. Rocks are among the primary features populating the Martian landscape. Characterization and understanding of rocks on the surface is a first step leading towards more complex in situ regional geological assessments by the rover.

A deadlock-free tree-based multicast routing algorithm is presented for all direct networks, regardless of interconnection topology. The algorithm delivers a message to any number of destinations using only a single start up phase. In contrast to existing tree-based schemes, this algorithm applies to all interconnection topologies, requires only fixed-sized input buffers that are independent of maximum message length, and uses a simple asynchronous flit replication mechanism. The theoretical basis of the technique used here is sufficiently general to develop other tree-based multicasting algorithms for regular and irregular topologies. Simulation results demonstrate that this tree-based algorithm provides a very promising means of achieving very low latency multicast.

A unicast-based multicasting algorithm is presented for arbitrary interconnection networks arising in switch-based networks of workstations. The algorithm is optimal with respect to the number of startups incurred and is provably free from depth contention. Specifically, no two constituent unicasts for the same multicast contend for a common channel, even if some unicasts are delayed due to unpredictable variations in latencies. The algorithm uses an underlying partially adaptive deadlock-free unicast routing algorithm. Simulation results indicate that the algorithm behaves as predicted by its theoretical properties and provides a promising approach to unicast-based multicasting.

Typically, data collected by a spacecraft is downlinked to Earth and preprocessed before any analysis is performed. We have developed classifiers that can be used onboard a spacecraft to identify high priority data for downlink to Earth, providing a method for maximizing the use of a potentially bandwidth limited downlink channel. Onboard analysis can also enable rapid reaction to dynamic events, such as flooding, volcanic eruptions or sea ice break-up.Four classifiers were developed to identify cryosphere events using hyperspectral images. These classifiers include a manually constructed classifier, a Support Vector Machine (SVM), a Decision Tree and a classifier derived by searching over combinations of thresholded band ratios. Each of the classifiers was designed to run in the computationally constrained operating environment of the spacecraft. A set of scenes was hand-labeled to provide training and testing data. Performance results on the test data indicate that the SVM and manual classifiers outperformed the Decision Tree and band-ratio classifiers with the SVM yielding slightly better classifications than the manual classifier.The manual and SVM classifiers have been uploaded to the EO-1 spacecraft and have been running onboard the spacecraft for over a year. Results of the onboard analysis are used by the Autonomous Sciencecraft Experiment (ASE) of NASA's New Millennium Program onboard EO-1 to automatically target the spacecraft to collect follow-on imagery. The software demonstrates the potential for future deep space missions to use onboard decision making to capture short-lived science events.