Coherent Tracking via Keystoning David Zasada, Principal Investigator Problems:
Current radars require enhancements in their ability to reliably track surface moving targets. These radars can experience problems tracking slowly moving and stopping targets due to a variety of environmental conditions. Better tracking performance will enhance our ability to prosecute time critical or otherwise fleeting targets.
Objectives:
1. To develop algorithms that combine short duration multi-channel synthetic aperture radar data with novel interferometric phase and amplitude thresholding, pre-detection tracking, and false alarm mitigation.
2. To exploit collected radar data to quantify algorithmic performance.
3. To develop detection statistics, receiver operating curves, track metrics, and sensor cost functions to facilitate integration of these algorithms into modern radars.
Activities:
Task 1: We will research algorithmic alternatives for range-Doppler-interferometric phase (coherent) tracking. We will research algorithmic alternatives for look-ahead constant false alarm rate maintenance.
Task 2: We will select appropriate radar data sets to exercise and evaluate algorithms developed under Task 1.
Task 3: We will develop detection statistics, receiver operating curves, track metrics, and sensor cost functions.
Impact:
These techniques will be beneficial wherever continuous tracking of high value targets is of paramount importance. This unique marriage of SAR and Surface Moving Target Information techniques supports our technical staff engaged in both imagery and moving target exploitation. It allows MITRE to contribute new, unique, and urgently needed capabilities to support our warfighter and research partners.
Approved for Public Release: 07-1508 Presentation [PDF]
Distance-based Approaches for Classification David Harris, Principal Investigator Problems:
Recently, Support Vector Machines have been the classifier of choice for many applications. However, users must describe similarity in ways that may not fit the application. Also, they could benefit from tools that report confidence values. Moreover, degraded results can occur in problems that involve matching (e.g., a new report will match to only a few entries in a corpus).
Objectives:
Under this MOIE we intend to develop and implement a set of classification tools that will synergistically target the three aforementioned problems. Specifically, these tools will improve accuracy by allowing domain-specific notions of similarity, assign confidence measures to results, and be robust in unbalanced training domains.
Activities:
We will conduct empirical work centered on distance-based and probability-supported classification. Verifying improvements will require answering several questions about the nature of embeddings and their relationship to an underlying notion of distance. Our primary focus will be on the database entity resolution problem -- how do you know when two reports are really about the same person, place, or event?
Impact:
We will measure success by running algorithms on standard datasets and comparing results to achievements to date. Sponsor applications will benefit from improved performance and lower entry cost through the use of more intuitive features. This research will position MITRE to provide high quality guidance to meet sponsors' information access and exploitation needs as they pertain to machine learning.
Approved for Public Release: 08-0095 Presentation [PDF]
Generic Transformational Scalable Modular Affordable RF Transceiver Perry Hamlyn, Principal Investigator Presentation [PDF]
Interactive ISR Data Exploitation and Sensor Operation James Witkoskie, Principal Investigator Problems:
GMTI data exploitation requires both signal processing by computers and contextual reasoning by humans. The human has a more natural understanding of vehicle behaviors in terms of contextual information such as terrain and road networks, while the algorithms have a speed advantage and can better process noisy data (especially from disadvantaged sensors). Optimal tracking performance will require the development of an environment that leverages both the human and machine.
Objectives:
In this MSR we explore the interplay of algorithmic signal processing and human contextual reasoning by developing an interactive exploitation environment. By creating algorithms that recognize problems that a human can correct, providing higher level information than the raw detections (track segments/convoy clusters), and prompting the user for input, this environment will be able to create more reliable information faster than current algorithmic or human operations.
Activities:
1) Develop prototype interactive environment.
2) Develop algorithms to produce higher level information (track segments and convoy clustering).
3) Develop algorithms to identify algorithmic ambiguities and prompt user for corrective action.
4) Allow human to override algorithmic decisions and catalogue these decisions to develop better understanding of algorithms missing contextual information.
Impact:
1) Disadvantaged systems (proposed Space Radar and UAV radars) produce data that cannot be visually inferred as in current GMTI Forensics applications, so the algorithms producing higher level smoothed information will be of great value.
2) LSRS/JSTARS users will be able to track more targets since they can designate several targets to track and then only concentrate on situations where they are prompted to address algorithmic ambiguity.
3) Forensic users will also benefit from ability to track more objects, and using less mouse clicks to vette a track
Approved for Public Release: 08-0434 Presentation [PDF]
Montage: Exploiting UAV Video in Mission Context Dave Anderson, Principal Investigator Problems:
Unmanned aerial vehicle (UAV) sensor data and telemetry are being aggregated in huge archives, yet retrospective analysis of those resources remains difficult. This is partly due to a lack of technical support for that analysis, but also to a lack of any searchable description of the sensor content. Extracting descriptions from the video imagery remains a difficult research problem.
Objectives:
We will demonstrate improved exploitation of historical UAV mission archives by providing a simple search service over the "what" as well as the "where and when," and enabling targeted analysis in external multi-source visualization tools such as ISR Forensics and Google Earth. We will also define a simple taxonomy of known-important entities, events, and attributes, and tag a corpus of augmented mission data for further research in video content analysis.
Activities:
We will collect operator audio speech and analyst IRC (Internet Relay Chat) during UAV missions, and time-correlate them with video and telemetry. We will provide search services based on Audio Hot Spotting and other tools. We will work with multi-source analysts to identify a taxonomy of important semantic tags for mission data, and develop a corpus with human-vetted annotations.
Impact:
Broad, content-oriented access to historical UAV mission data will improve UAV archive exploitation and situation awareness for deployed soldiers. It will also provide a basis for evaluation of research on video content extraction that is grounded in real-world problems.
Approved for Public Release: 07-0179 Presentation [PDF]
Optically Sensed Tags Sherry Olson, Principal Investigator Problems:
The need to detect evidence of chemical residues and effluents is a primary concern for many communities. Often these materials are difficult to detect, because they are present in very low concentrations and exist in complicated environments. It would be advantageous to employ methods that would allow for standoff detection, while also being sensitive, small, inexpensive, and unobtrusive.
Objectives:
The goal is to develop optically sensed tags based on induced chemical interactions and emissions for detecting low concentrations of chemicals and materials over a range of standoff distances. The tagging mechanisms developed will leverage existing related technologies used for chemical-specific sensing, but crafted for optical use and thus for standoff and unobtrusive sensing.
Activities:
We will investigate relevant technologies with the goal of constructing prototypes. The first phase will include additional trade studies, initial prototype constructions, and lab testing. The second phase would start with tests in a "friendly" inside environment and then begin to introduce adverse conditions (e.g., temperature, humidity, contamination). The third phase will test the tag outside and include refinement cycles.
Impact:
Through the use of optically sensed tags we expect to improve standoff and unobtrusive detection and identification of dangerous chemicals and other materials. Progress in tackling these difficult problems has significant implications for environmental monitoring in several contexts.
Approved for Public Release: 06-1465
Pixel Registered, Stacked, Multi-Band Compact Imager Gary Euliss, Principal Investigator Presentation [PDF]
Sensor Layer Prototype David Zasada, Principal Investigator Problems:
Binary formatted sensor reports designed decades ago are insufficiently flexible to accommodate a rapidly evolving net-centric warfighting enterprise. Even if these binary formatted reports are enhanced by modern metadata descriptors, such as XML, they are insufficiently flexible. The Sensor Layer vision calls for unprocessed and semi-processed data to be exposed and made available on an enterprise service bus.
Objectives:
We will extend the Sensor Layer Prototype to make upstream sensor data accessible on demand to the enterprise. Upstream data consists of both unprocessed and semi-processed data that has previously been largely unavailable to the end user community. This will improve the possibilities for innovation within the sensor data processing and utilization communities.
Activities:
Task 1: We will extend the current prototype to increase both usability and functionality. Both external and internal program components will have improved interactivity.
Task 2: We will expand data collections and data formats included within the Sensor Layer server. This includes the addition of the newly defined UPHD standard.
Task 3: We will apply a focus on making I&Q more accessible to the end user. This includes improving I&Q data availability and accessibility.
Impact:
This prototype will demonstrate that sensor systems can be responsive to unanticipated threats, missions and needs by providing the user community with greater flexibility of data transfer. The prototype will provide the user community with all products, so that they can achieve their desired results. This flexibility will foster innovation and allow for new products to be formed as needs arise.
Approved for Public Release: 07-1468 Presentation [PDF]
Structured ISR Fusion Walter Kuklinski, Principal Investigator Problems:
The importance of developing fusion algorithms to improve the warfighting capability derived from Intelligence Surveillance and Reconnaissance (ISR) systems is well documented. Many operational requirements mandate levels of performance that can only be obtained by fusing data from multiple sensors and platforms. However, only limited attempts have been made to develop formal design procedures for data fusion systems.
Objectives:
The objective of this MSR is to use formal mathematical methods such as Random Set Theory and Category Theory as frameworks to design and implement fusion algorithms and to predict fusion system performance. We will develop and implement a widely applicable methodology for fusion system design and analysis.
Activities:
We will develop and evaluate, using both simulated and field data, multi-target tracking algorithms that fuse conventional radar SMTI data with additional data sources including SIGINT, COMINT and ELINT, using both Random Set Theory and Category Theory frameworks.
Impact:
The successful completion of this MSR will both provide a significant advance in the general area of fusion system analysis and design, and yield specific fusion algorithms applicable to the future fusion requirements of many MITRE sponsors.
Approved for Public Release: 05-0178 Presentation [PDF]
Unconventional Optics for Imaging Sensors Ravi Athale, Principal Investigator Problems:
During the past 10 years, focal plane detector array technology has been improving at a very rapid rate driven in part by developments in lithography and materials. Lens design and fabrication have progressed far more slowly if at all during the same period. As a result, size, weight, and form factor of imaging sensors is now dominated by the imaging optics.
Objectives:
The main objective of the project is to identify new designs for front end optics that reduce system volume and weight by an order of magnitude and result in imaging systems with "extreme" form factors (extremely-thin, extremely-narrow). We will also explore novel optical transforms enabled by unconventional optical elements strategically designed to exploit additional information encoded on the optical field.
Activities:
Task 1: Lens design by combining graded index, refractive and diffractive effects in a single element.
Outcome -- preliminary design and analysis.
Task 2: Exploring alternate approaches to imaging optics (coded aperture, photon sieve, negative index).
Outcome -- limitations and scaling analysis.
Task 3: Study unconventional transforms for multi-domain (spatial, spectral) information extraction with post-detection computation.
Outcome -- concepts.
Impact:
Imaging sensors (mmW to X-rays) are being used extensively in defense, intelligence, homeland security, entertainment, biomedical and scientific communities. Any new concepts that lead to dramatic reduction in size and weight of imaging sensors will have significant impact on most of these uses. Extracting useful information from the scene without exponentially increasing associated data volume is another potential impact of this project.
Approved for Public Release: 08-0348 Presentation [PDF]
Last Updated:05/05/2008 | ^TOP | 
