6
Priorities for NGA GEOINT Research

The National Geospatial-Intelligence Agency (NGA) faces an extraordinary challenge in the years ahead in moving from an intelligence environment that remained essentially unchanged during the Cold War, to an era of ubiquitous and real-time geospatial intelligence (GEOINT). The decision to redesign NGA around the concept of GEOINT was sound. Yet the hard research problems faced by NGA in the years ahead will require a concerted effort to devote resources to developing the new approach and to nurturing the revised and expanded collaboration among the intelligence community, government, industry, and academia necessary for achieving this goal.

The committee feels that NGA’s new doctrine and vision statement offers an impressive view of the way forward. Nevertheless, NGA is a large government agency, with a definitive culture and workforce and with capabilities that need to be considered as the doctrine is upheld and the vision implemented. Many challenges in the years ahead will be related to human and organizational infrastructure as much as to technical and methodological architecture.

The committee’s charge was to examine the hard problems in geospatial science that must be addressed to improve geospatial intelligence, and to identify promising methods and tools in geospatial science and related disciplines that can be brought to bear on national security and homeland defense problems. Most of the committee’s recommendations address this charge directly, and Chapter 4 structures the identification of problems and methods around the “top 10” priority list generated by the NGA.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency 6 Priorities for NGA GEOINT Research The National Geospatial-Intelligence Agency (NGA) faces an extraordinary challenge in the years ahead in moving from an intelligence environment that remained essentially unchanged during the Cold War, to an era of ubiquitous and real-time geospatial intelligence (GEOINT). The decision to redesign NGA around the concept of GEOINT was sound. Yet the hard research problems faced by NGA in the years ahead will require a concerted effort to devote resources to developing the new approach and to nurturing the revised and expanded collaboration among the intelligence community, government, industry, and academia necessary for achieving this goal. The committee feels that NGA’s new doctrine and vision statement offers an impressive view of the way forward. Nevertheless, NGA is a large government agency, with a definitive culture and workforce and with capabilities that need to be considered as the doctrine is upheld and the vision implemented. Many challenges in the years ahead will be related to human and organizational infrastructure as much as to technical and methodological architecture. The committee’s charge was to examine the hard problems in geospatial science that must be addressed to improve geospatial intelligence, and to identify promising methods and tools in geospatial science and related disciplines that can be brought to bear on national security and homeland defense problems. Most of the committee’s recommendations address this charge directly, and Chapter 4 structures the identification of problems and methods around the “top 10” priority list generated by the NGA.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency However, it is also useful to put the problems in the context of the actual process of doing GEOINT, to better show how research will support the evolution of the various steps in the GEOINT process. The next section puts forward a framework that describes the GEOINT2 process and information flow and then correlates the hard problems identified in Chapter 4 with the steps in the framework. Whereas the top 10 challenges are focused on the overall process and its outcomes, the framework described below is focused more on the individual steps of the process. Looking at the hard problems in both ways will allow the development of a more organized and robust research agenda and clarify the needed prioritization. These priorities are covered later in this chapter. GEOINT2 PROCESS FLOW The key stages in geospatial information handling are to acquire, identify, integrate, analyze, disseminate, and preserve. In prior eras, these were separate and quite distinct tasks, even compartmentalized in terms of security. Early era space surveillance, for example, included segments of this cycle that took place in different states, at different times, and with different skills and affiliations such that most participants had no concept of the remainder of the cycle or were even aware that there was a cycle. This is the environment in which today’s GEOINT evolved, yet the NGA vision recognizes that the Cold War compartmentalized model is no longer adequate. A cycle that once could take months must now happen in minutes. There is no longer time to rely on fortuitous knowledge synthesis, nor can the system depend on specialists who spend their entire careers on a single problem. The GEOINT2 analysis framework as envisioned by this committee is shown in Figure 6.1. The framework operates within, and is supported by, the existing cyberinfrastructure to sustain on-demand intelligence, to monitor and minimize uncertainties, and to preserve semantics in data and in GEOINT. GEOINT is a circular flow from newly acquired data to archived result. Yet thinking of the framework as a processing cycle with a clear beginning and end is a fallacy. In reality, new data arrive in a never-ending stream from instruments and the Internet. Thus, data input is a network of networks, remote sensing systems, cyberinfrastructure, sensor webs, additional intelligences (INTs), and so forth. Flowing out of the cycle is knowledge, in the form of specific decisions, reports, and actions, but also flowing back from this knowledge are new data. Both preservation and dissemination are outputs to specific communities, the “customers” for intelligence, but they are also sources of new data for future use. In GEOINT2, it should be as easy to acquire existing data with embedded links to the current time period as it is to acquire new data or

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency FIGURE 6.1 GEOINT2 information flow.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency images. Expanding and building on prior and existing knowledge in addition to new data will permit the establishment of a knowledge-based GEOINT2. Much of the necessary research and development for GEOINT2 is required to improve these stages, whereas some issues are overarching and impact all parts of the cycle and their synthesis. The stages of the information cycle, as imagined operating under GEOINT2 and illustrated in this framework, can be expanded as follows. Acquire. Information acquisition incorporates targeted gathering strategies and intelligent preprocessing and filtering, to collect no more and no less information than is required. A current strategy is to collect everything. In GEOINT2, the goal is to collect only what has changed. Identify. This operation starts with specific intelligence tasks and identifies the information resources required to address the problem. This is followed by pre-processing of individual retrieved datasets, to filter, enhance, or segment out of them the required task-specific content. Image processing extracts features and identifies patterns and anomalies. Current classification methods are automated but still require some degree of manual oversight for interpretation. Identification in urban scenes is a special problem due to the rate at which items change and because of the amount of indoor, under-canopy, and under-ground activity. Tasks of identification include embedding data with “hidden” metadata and lineage information (watermarking) and the discovery of pseudosignatures or other image deception. As features are identified or tracked within a dataset, evidence of uncovered content could be automatically stored in the permanent metadata record. As the data move through the rest of the framework, these metadata are progressively augmented. By the time the preservation phase of the framework is reached, a chronological tag of successful (and failed) applications of the data should initiate the transformation of an archive that simply contains data into a self-describing knowledge base. Integrate. Data acquired from multiple sensors carry varying granularity, geometric type, time stamps, and registered footprints. Data fusion rectifies coordinate positions to establish which features have not changed over time in order to focus on what has changed. The fusion, however, involves confronting several hard problems discussed in Chapter 4, including spatial and temporal conflation, dealing with differential accuracy and resolutions, creating the ontologies and architectures necessary for interoperability, and managing uncertainty with metadata.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency Analyze. The dependence of geospatial data on resolution, proximity, and local context creates special problems for conventional forms of spatial analyses. NGA additionally needs to support real-time and near-real-time analysis of battlefield planning scenarios including, where appropriate, spatial optimization problems (e.g., trafficability) and trade-off analysis using distributed computing technologies. Hard problems of relevance include the integration of analytical results into salient displays, integrating high-performance computing (HPC), and communicating uncertainty in the analytical stages. The analysis flow itself is an important part of metadata, and while new capabilities in geographic information systems (GIS), (e.g., Environmental Systems Research Institute’s model-builder) facilitate this, much work remains to be done. Disseminate. Disseminating intelligence as it is prepared forms a major part of NGA activities because it impacts operations planning, distributed and collaborative GEOINT, and augmented reality (the concurrent use of map and/or image data to augment one’s view of a real landscape). The traditional paper map or image photo print and physical distributions and publishing have still to yield completely to digital representations and Internet distribution. As noted above, successful sharing of data and information analysis results will require shared standards describing data format and meaning among the systems used by different agencies and including provenance, workflow, and uncertainty information. Preserve. Preserving geospatial information poses several challenges, particularly in terms of the volume of information that is collected (currently on the order of terabytes per day). Challenges include dealing with the sheer redundancy of much of the information that is collected, the difficulties involved with indexing for efficient cataloging and retrieval, and the security and declassification policies that must be established to protect both discovered and as-yet-undiscovered intelligence. HARD PROBLEMS IN THE CONTEXT OF THE GEOINT2 FRAMEWORK This section links the hard problems (as summarized in the recommendations) identified in Chapter 4 with the steps in the GEOINT2 framework described above. It then gives a priority to each of the recommendations. The process recommendations from Chapter 5 are also prioritized. Priorities are assigned numerical levels 1, 2, and 3. Priority 1 research is considered vital, immediate in terms of support needs, and a prerequisite to higher priorities. Priority 2 needs are problems that require solution if the preconditions for the GEOINT2 transition are to be met. Priority 3

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency involves research that is necessary to complete the full set of GEOINT2 demands. Acquire Three recommendations relate primarily to geospatial data acquisition. Recommendation 10 seeks to widen the scope of data input to include not just the more traditional static targets but also to include moving objects. Moving targets are literally harder to hit because they demand data synthesis (1) across sensors, (2) across scales and resolutions, (3) across the spectrum, and (4) in time. In the 2020 scenario in Box 2.2, the target is not a single point, nor is it a building or object. What is significant is the convergence of multiple sources of information on an action point and time, which is the choice of the person on the ground. No one sensor provides all of the necessary data, nor could it. Data and information fusion, however, are indeed in their infancy. Without this capability, and the theory, methods, and technology to support it, GEOINT2 would not be possible. This recommendation is therefore assigned Priority 2. Recommendation 1 is similar, but recognizes that sensor network research should focus on the impact of sensor networks on the entire knowledge assimilation process. Stated succinctly, this recommendation warns that sensor networks cannot be thought of in the same way as imagery intelligence (IMINT). Sensors can inform the entire cycle and provide useful and essential links at all stages. Dissemination, for example, in the push-pull Internet model, is simply allowing NGA intelligence to be input to other processes in other agencies and services. Here NGA must seek its place within the vast array of public, commercial, and government information environments and technologies, both as information and knowledge supplier and as a consumer through sensor networks that are beyond its control, but essential for its operations. Given that sensor networks will develop independently of NGA’s needs and that NGA will not be the only user or supplier of them, this research need is rated as Priority 3. Recommendation 3 is also of importance for the acquisition phase and recognizes the demands that GEOINT2 will place on database management systems. The recommendation targets research at ensuring that current database architectures scale up to meet the demands of agile geospatial sensor networks. With a vast variety of sensors, remote and otherwise, and with a substantial increase in data volumes from all sensors due to increased spectral, spatial, and temporal resolution, the likelihood of failure of the existing models is high. If NGA participates in supporting basic research in this area, it will help ensure that new architectures are available when NGA needs them. This is a high-priority need and so is assigned Priority 2.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency Identify On the important topic of identification, the committee makes a single recommendation (Recommendation 2): Research should be encouraged on spatiotemporally enabled geospatial data mining and knowledge discovery, on the visualization of time-space, and on the detection and description of time-space patterns such as trajectories. This recommendation formalizes the importance placed by the committee on the need to move away from static GEOINT, such as maps and images, toward data streams linked to features moving against a background of more static, but still changing, reference data. This means not simply being able to update the spatial information framework rapidly (e.g., imagery, vector cultural data, place names, vegetation, digital elevation model), but also being able to identify and recognize patterns, repetitions of prior movements, time-space structures, and so forth. Examples might be tracking a vehicle as it moves across the landscape, perhaps with multiple sensors; recognizing a movement pattern from the past (e.g., distinguishing between a routine and a special boat patrol); or entire development patterns (e.g., systematic movement of troops or material toward a border, or the large-scale planned digging of mass graves). The committee ranks the ability to manage, detect, and encode both small- and large-scale movement patterns as Priority 1. Without these tools, little can be done to migrate to GEOINT2, and similarly, the challenges are evident in existing and near-term systems. Integrate Issues of integration are generally a matter of achieving interoperability of concepts, systems, and data. The committee strongly believes that a common ontology for time-space data is necessary before a next-generation architecture can emerge. This is also the key to data sharing with other agencies and with coalition partners. Five recommendations fall into this stage of the intelligence cycle. Recommendation 7 recognizes the broad need to raise the level of academic scholarship on fusion, especially of geospatial data. There are a host of critical issues that have to be confronted. How are different data sources matched with each other seamlessly? Where do problems of scale compatibility make fusion difficult or impossible? How are data from one spatial sampling system (e.g., points or pixels) matched with data from others (e.g., census tracts, police districts). At the syntactic level, how literally can data formats be meaningfully conflated, given that both have independently measured locations and identifications for what should be the same objects? This recommendation is assigned Priority 3, being more long-term oriented.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency Recommendation 3 impacts the integrate phase as well as the acquire phase as described above. At the syntactic interoperability level, Recommendation 3 targets the ability of current database architectures and data models to scale up to meet the demands of agile geospatial sensor networks. If NGA is to meet its projected needs, this problem is likely to become of concern in the very near future. It is assigned Priority 2 only because the current architectures may be able to meet some future needs without disruptive changes in technology. Nevertheless, this capacity cannot be relied upon in the longer term. Recommendation 8 stresses the importance of toponymy for interoperability. As data find their way from the huge variety of sources, in various forms and formats, it is simple language that holds one of the keys to integration. While NGA’s capabilities around toponymic services are exemplary, to go to the next stage (i.e., to fully integrate the Internet, news feeds, intelligence reports, presidential daily briefing summaries), the walk between places as text and places as coordinates must be a robust transformation. Text-based sources must be up-to-date, reliable, and authoritative. The algorithms that allow them to extract coordinates must also be, and they must function quickly and accurately. At the very least, NGA needs superior capabilities to on-line systems such as MapQuest worldwide that support the use of multiple languages. The committee assigns advanced toponymic services to Priority 2. Recommendation 9 seeks research to promote the reuse and preservation of data. Past preservation paradigms have been based on the tile (e.g., a map sheet or image scene) or on a collection (a revision, a map series, or a whole instrument’s coverage). GEOINT2’s most difficult challenge is that the key operation unit of the GEOINT database should be the feature. Features appear in multiple sources and need to be searchable across sources. A typical image search from the Cold War era was to find all features in all (identical) images that match a specific template, such as a missile silo or a henhouse-shaped radar building. Such tasks could be automated because the target was essentially fixed and invariant. With sensor fusion, each sensor (or intelligence source) has a different pattern for the same object. To a pressure-sensitive smart-dust mote, a particular tank is recognized by its weight, magnetic signal, or sound. To a remotely sensed image, it is a fixed-shape object that is darker than its surroundings. To a video camera on a Predator, it may be a distant exhaust plume or dust cloud. All data should be brought together and used collectively to assert the existence of the object and then compute its position, velocity, off-road capability, time-space trajectory, and so forth. This approach is strengthened when past data can be used for matching too, but only if the signal has been detached from the reference frame in which it was captured and converted into smart data at the feature level. Similarly, features

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency that encode their own descriptions and histories are self-contained and can be stored and searched in innovative ways. The committee assigns this recommendation to Priority 2, recognizing that its requirements will be disruptive to existing software and database systems. Recommendation 11 seeks the creation of a complete descriptive schema for geospatial objects of importance to GEOINT—that is, a fully formalized GEOINT ontology. As a top-down exercise, perhaps led by analysts, and concentrating on analysis tasks rather than the particulars of sensor systems, this recommendation sets the scene both for the development of the database architecture for GEOINT2, and for a more effective GEOINT. What are the objects of interest to interpreters and analysts? How are they described? How do they move? What are their relations (1) to the geographical environment and (2) to each other? On the surface this is an abstract exercise but a critical one. The committee assigns this task Priority 3. Analyze Recommendation 4, while directed at all stages of the intelligence cycle, is probably most applicable to the analysis phase. The recommendation is that research should be directed toward the determination of what processes are most suitable for automated processing, which favor human cognition, and which need a combination of human-machine assistance. With the demands for intelligence at their peak when the analyst does his or her work, this recommendation seeks to create a research task out of determining how the job is conducted, what information is needed during which tasks, and how the computer can either do work, assist, or just get out of the way. When, for example, is too much information a detriment rather than help? How does an analyst shut out information while filtering or scanning? How can salient information be brought to an analyst’s attention without distraction? What happens when the analyst is also the field operative, conducting the analysis in real time on the battlefield? The foundation in human spatial behavior and cognition can be of help in this task. The committee assigns this recommendation Priority 2. Given the importance to NGA of visualization of GEOINT data, Recommendation 5 suggests that research should be supported that investigates new methods of data representation that facilitate visualization. Again, although visualization is likely to be of most use in analysis and identification, there are few parts of the intelligence cycle in which visualization is not useful. New methods such as latent semantic indexing, the semantic web, and self-organizing maps are able to take nonspatial data and use spatial visualization methods to seek pattern. Visualization research has other funders and disciplines for support, and that research

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency is robust. Nevertheless, the committee assigns this recommendation Priority 2. Disseminate Recommendation 12 directs NGA research toward the particular needs of geospatial data for protection with multilevel security and the processing of data so that they resemble public domain or commercially produced structures and formats. Anecdotally, the committee heard that this problem was “eating our lunch” and of the highest priority. While cryptography is well supported in terms of research, the lack of attention to geospatial data and the need for short- and long-term solutions lead the committee to assign this recommendation to Priority 1. Preserve Recommendation 9, the reuse and preservation of data, also impacts the integrate phase and is described above in that section. The preserve step is also impacted by Recommendation 6. The goal of self-describing feature-level data is ambitious. Few commercial database systems are designed to handle these data and even fewer in large volumes. The benefits of Recommendation 6, while also generic to all phases of the GEOINT cycle (i.e., NGA should ensure that the special needs of geospatial data are met by high-performance grid computing), are most likely to be pertinent to the archiving and retrieval of data after the fact. Given the unprecedented data volumes, and the relatively slow speed with which GIScience (geographic infromation science) is interacting with HPC and grid computing, plus the availability of the national laboratories, much could be done to advance the research frontier now to NGA’s advantage. This recommendation is assigned Priority 2. Table 6.1 summarizes the hard problems by framework step. The Research Process Some of the recommendations refer not to the framework, but to the overall process of doing research. Recommendation 13 states the committee’s strong support for the peer review process in NGA research. Other federal agencies, such as the National Science Foundation, are exemplars in this respect. Few NGA research funding mechanisms would need to be outside the peer review process, and innovative mechanisms could be used to ensure their quality. The NGA Academic Research Program (NARP) is off to a solid start in this regard, but it could benefit from

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency TABLE 6.1 Hard Problems in the Context of GEOINT2 Framework GEOINT Framework Step Hard Research Problems Recommendation Acquire Assimilation of new, numerous, and disparate sensor networks within the tasking, processing, exploitation, and dissemination (TPED) process 1 Detection of moving objects from multiple heterogeneous intelligence sources 10 Spatiotemporal database management systems 3 Identify Spatiotemporal data mining and knowledge discovery from heterogeneous sensor data streams 2 Integrate Image data fusion across space, time, spectrum, and scale 7 Spatiotemporal database management systems 3 Role of text and place name search in data integration 8 Reuse and preservation of data 9 GEOINT ontology 11 Analyze Process automation versus human cognition 4 Visualization 5 Disseminate Multilevel security 12 Preserve High-performance grid computing for geospatial data 6 Reuse and preservation of data 9 additional attention to the peer review process, including feedback to its collaborators. This recommendation is assigned Priority 1. The casual reader of Chapter 2 would probably become confused about the number and diversity of means by which NGA research brings new knowledge and ideas forward. Recommendation 14 calls on NGA to define clear roles, responsibilities, and relationships for the various types of organizations that conduct and disseminate results from research and development (R&D) in NGA’s priority areas. There appears to be some room for more synthesis and perhaps consolidation that could benefit

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency NGA’s programs and directorates. This recommendation is assigned Priority 3. Recommendation 15 suggests that NGA define and publicly articulate the current and future geospatial information architecture, at a level of detail sufficient for the integration of future prototype systems and components. The goal of this task is to align and interoperate current and near-future systems. This task is not difficult; indeed, it is development rather than research. By choosing a set or suite of components, or even a set of rules and specifications, immediate orders-of-magnitude improvements in interoperability will be possible, at least at the data exchange level. It will also help NARP to move more projects into years four and five, where actual products can result from research. Too few demonstration prototypes find their way into operational systems and, thus, never get the chance to have an impact. A common architecture, or even a loose set of desirable traits, may be sufficient. This recommendation is assigned Priority 1. Recommendation 16 asks that NGA be explicit about how the results of R&D projects can be incorporated into current and future architectures. This recommendation seeks closer collaboration among NGA’s disparate research community, a goal that may not be a “hard problem,” and so is assigned Priority 3. Recommendation 17 is a response to NGA’s own goal of working more closely with the geospatial science and technology R&D community from coalition countries. This not only ensures access to human capital and deals partially with the shortage of people trained in GIScience, but also builds future collaborations that could have lasting value. Naturally, this recommendation depends highly on the assurance of new means of horizontal integration and so is assigned Priority 3. Table 6.2 summarizes all of the recommendations by priority. CONCLUSIONS All told, the series of recommendations made in this report are the result of a considerable amount of research, reflection, and discussion among the members of the committee. With the broad representation reflected by the committee’s membership, there is consensus that these recommendations represent considerable wisdom, and not a small amount of “intelligence.” The committee urges NGA to consider them carefully. A next stage in consideration of these hard problems would be for NGA to work with its partners both within and outside the intelligence community to create a research agenda. Publication of this agenda would have the dual benefits of informing NGA’s research partners of the sense of priorities for research at NGA and making the problems tangible; it

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency TABLE 6.2 Prioritization of Hard Problems and Recommendations Priority Hard Problem, Recommendation Recommendation 1 Spatiotemporal data mining and knowledge discovery from heterogeneous sensor data streams 2 Multilevel security 12 Increase use of peer review 13 Communicate current and future architecture to researchers 15 2 Spatiotemporal database management systems 3 Process automation versus human cognition 4 Visualization 5 High-performance grid computing for geospatial data 6 Role of text and place name search in data integration 8 Reuse and preservation of data 9 Detection of moving objects from multiple heterogeneous intelligence sources 10 3 Assimilation of new, numerous, and disparate sensor networks within the tasking, processing, exploitation, and dissemination (TPED) process 1 Image data fusion across space, time, spectrum, and scale 7 GEOINT ontology 11 Define roles of various research participants 14 Define how projects fit into architecture 16 Collaborate with coalition countries on geospatial R&D 17 NOTE: Bold indicates hard problem; no bold indicates process recommendation. could also serve as a guidance document for future Broad Area Announcements and directed research requests. When linked to the more abstract but critical doctrine guidance provided by NGA’s leadership, the agency will be well on the way to the design and implementation of GEOINT2, the geospatial intelligence infrastructure for the twenty-first century.

OCR for page 73
Priorities for Geoint Research at the National Geospatial-Intelligence Agency This page intentionally left blank.