Print This Post

Registries for GeoINT

Using geospatial information for effective and actionable intelligence may depend as much on semantics as is does on geography. Knowing a threat location is of interest, but that information is enormously more valuable if intelligence sources can easily classify the threat in terms of type, immediacy, certainty, and verifiability. The information is more valuable still if it is easy to relate the threat to other threats, events, persons of interest, and geographic features.

A registry represents a new approach to the management and utilization of geographic information that combines semantics and geography. It readily supports information integration and makes it easy to trace from a decision to supporting source data.

Recently evolving registry technology provides the intelligence community with a practical geo-semantic platform that provides these and many other features. Registries are the subject of a new work item at the ISO TC211, namely ISO 19164.

Consider the issue of a “single source of truth,” something that is critical for intelligence decision making. Intelligence information may be distributed across many technologies (e.g. SIGINT, GeoINT, ELINT, MASINT) and, consequently, many databases; unfortunately, this makes effective fusion of this information difficult. Here is where registries can be quite helpful. Registries can describe and support any number of data sources and they make it easy to classify and relate these sources to one another or to other data objects. Registries can also provide data virtualization support, which means that local copies of the resources are not necessary for such semantic fusion.

A simple example might serve to illustrate this further. Suppose our intelligence units acquire SAR, FLIR, and Optical Images of a target area. These are first stored in the registry with associated property descriptions (such as resolution, image dimensions, and date of capture) and classification (i.e. image type). The arrival of each image type causes a message to be automatically generated using the registries automatic notification feature and sent to a specific analyst, maybe one who specializes in particular target types in SAR imagery. The analyst interprets the image and creates a “feature hypothesis” assigning it specific properties, such as dimensions measured in the image, location, and so on, and one or more classifications, like “metal object”, and “tank”. Property values may be assigned to each classification to specify the certainty of the classification, or its technical properties.

After the same procedure has been repeated for the other image types, a senior analyst may then query the region for all of the “feature hypotheses” to see if they are directly overlapping (e.g. everyone thinks it is a tank) or if the classifications (e.g. “metal object”, “hot object”, “appears like tank”) support a specific conclusion (“it is a tank”). The senior analyst may then construct a “feature” called “tank” with associations to the supporting evidence (“feature hypotheses”) which, in turn, are linked to underlying imagery. An application can then easily present this information through a web page or desktop display so that another analyst or a decision maker can see it and trace from the primary decision through to all of the underlying decisions and supporting evidence.

Intelligence data is far from static. The passage of time and changes in events may easily alter the status of any piece of information. In a registry, this is tracked by life cycle status. Registries provide a configurable set of life cycle states and allowable state transitions on an object type basis, and any number of life cycle state lists can be created. An intelligence observation, for example, might have states like {submitted, approved, invalid, markedfordelete, deleted}, while a physical asset might have states like {submitted, observed, targeted, destroyed}.

Registries make it easy to create any number of logical collections of objects. These collections might group objects by related events, incidents, observations, or whatever makes sense for a particular purpose. Registry objects can have any number of properties associated with them, including geographic ones. A collection or registry package can easily define a logical folder for a particular case, or intelligence objective.

One of the common failures in intelligence is the recognition of related events. This is an area where registries can deliver significant benefit by making it easy to relate local or remote objects through user-defined associations. Associations support creating a “single source of truth” by providing a way to remove apparent duplicate objects. For example, linking the Tokyo Tower Restaurant with an “isPartOf” association to the Tokyo Tower means that it will not be seen as a separate building. It can also be useful to use associations between objects when expressing relationships which are not geographic, or which are partly geographic and partly related to some logical model of the objects involved. Equally, associations can be used to express qualitative spatial relationships such as “next to”, “above”, etc., which have not yet been supported by quantitative information.