Wide Area Persistent Surveillance (WAPS) Capabilities

Wide area persistent surveillance yields unusually large and complex datasets from different sources, presenting a challenge for storing, retrieving, and integrating information in a meaningful and timely manner. Sophisticated architectures and tools are required to manage these datasets and allow analysts to quickly search massive data from disparate sources.

What do you do with potentially 100's of terabytes of new data every day? Let's put that number in perspective. A laptop can store 120GB total: so ~833 laptops filled with data. The U.S. Library of Congress Web Capture team claims "as of May 2007, the Library has collected more than 70 terabytes (70,000 billion or 70 million million bytes) of data".

Specific project areas of interest include:

• Storage, Handling and Dissemination - developing a system that allows ingest of WAPS data at rates of 10+ Terabytes per day, online storage of recent missions (< 90 days old) and nearline storage of older mission data (90+ days old). The system shall support high speed random access to online data and be capable of retrieving nearline storage on demand within minutes. The architecture must support 100+ simultaneous client workstation connections with imagery playback speeds > 50x real time. The system must provide fault tolerance for network latency and the ability to work in a communications degraded mode. Custom compression schemes and approaches may be incorporated. Cost per byte of data stored and scalability of the solution are key metrics.
• Multisource Integration - developing algorithms, tools and techniques to perform multisource integration. Required capabilities include geospatial and temporal registration, bias adjustments for geospatial and temporal calibration, and rendering information such that analysts can identify correlations and nonobvious relationships within and across datasets.
• Complex Activity Illustration and Reporting - creating tools, techniques and procedures to quickly generate products and presentation materials that illustrate complex interactions between and amongst multiple entities over time for dissemination to customers. Special customized products and presentation materials are required to accurately and concisely convey the spatial and temporal context and details.
• Data Searching and Discovery Services - developing capabilities to process and index massive (multipetabyte) data holdings to enable rapid search and retrieval of data. This includes implementing database architectures and web-service front ends that allow remote client workstations to access archived information and generate filtered data packages to support analysis over standard communications and networking connections. The user interface must include user definable geographic and time window constraints to reduce the volume of image data returned. Independent components of an overall system design include server hardware, database schema, metadata tagging, partitioning, query logic and data filtering solutions.

Search improvements are required in the following areas:

Activity Detection

Develop algorithms, tools and techniques to detect, identify and discriminate human activity including movement of vehicles of any size, motorcycles/bicycles, individual people (dismounts), and animals within a predefined area of interest (AOI) during a specified time period. This challenge represents the operational application of sensors for monitoring a denied or restricted area, a suspect site, curfew monitoring, border security, or facility surveillance. Systems must alert the analyst in real time, near real time, or forensically. Movement can be detected without tracking. AOIs typically range in size from a single commercial or residential lot to a city block or more.

Specific project areas of interest include:

• General Activity Detection - developing the capability to detect and identify all activity within an AOI during a specified time period. The ability to discriminate human activity of interest from background activity is important.
• Start/Stop Detection - developing the capability to detect and identify only the activity originating or ending within an AOI. Persistent background activity which does not represent motion in or out of the AOI should be rejected. Vehicle traffic that passes through but does not stop or start within the AOI should be rejected as well. Tracking the activity outside the AOI is not required.

Track Continuity and Handoff

Develop algorithms, tools, and techniques to maintain track continuity and positive identify within and outside of a predefined specified AOI during a specified time period. Continuous tracking becomes increasingly difficult with changes in entity behavior, varying environments, and cross-over into different sensor coverage areas. Temporal and spatial alignment as well as geolocation and time stamp errors within different data types and mission segments can prohibit efficient and effective data integration. Track handoff is necessary when a track cannot be continued in one data source due to lack of coverage, obscuration, shadow, activity density or other difficulties.

Specific project areas of interest include:

• Entity Tracking - tracking activity beyond the AOI to its initial starting or final stopping location, through different environments including varying levels of traffic density, entity behavior (pauses, extended stops, turns, U-turns), obscuration, and environmental and image conditions. Activity tracks of varying degrees of detection difficulty will be provided as examples. The objective is to continuously track vehicles traveling through a city for 5-10 minutes or longer.
• Dismount Tracking - developing the capability to detect and identify cases where vehicles stop outside the AOI and passengers dismount and walk back into the AOI.
• Cross Sensor Track Handoff - correlating tracks in overlapping sensor coverage areas and performing hand-offs to maintain track continuity and positive identity. Of specific interest is the ability to track entities from EO imagery to GMTI coverage and vice versa. Wide area EO imagery is typically used for urban coverage and GMTI for rural coverage. Tracking vehicles moving from urban to rural areas, and vice versa, requires handoff between the two data types. As such, only proposers with active personnel and facility clearances will be considered to submit Full Proposals for this specific portion of the requirement. Proposers without active personnel and facility clearances are still eligible to receive awards for solutions that meet one or more other specific portions of this requirement.

Relationship and Pattern Identification

Once activity of interest has been detected, tools are required to distinguish between typical activity patterns and anomalous activity, and identify relationships between areas of interest. Tools must be able to reliably identify such activity and relationships over time. Manually created tracks may be used as a basis to test advanced pattern identification solutions.

Specific project areas of interest include:

• Anomalous Activity Identification and Alerting - identifying and alerting analysts of anomalous activity amongst typical background activity (such as traffic) in real time, near real time, or forensically. Anomalous activity can include extreme changes in traffic volume, traffic jams, sudden traffic disruptions due to accidents or attacks, traffic fleeing from a site, rerouting of traffic, and smoke plumes (detected as a moving feature). Additional indicators can include vehicles avoiding specific areas such as checkpoints, Blue Force locations, or suspected hazardous locations.
• Nodal and Relationship Analysis - determining which locations are associated with a given point of interest based on movement patterns, and the relative strength of the association based on movement patterns and frequency. Methods should distinguish between significant nodes (such as meeting houses) and insignificant nodes (such as a post office). Schemes for prioritizing and visualizing nodal relationships, strengths, and confidence of associations should be considered, in addition to methods for extracting, organizing, and storing.
• Patterns of Life - identifying repetitive behaviors, or patterns of life, associated with specific points of interest. Characterize activity including arrival and departure times and locations, number and size of entities, and movement direction to assist analysts in identifying deliveries, security patrols, daily work commutes, and weekly/monthly meetings. By accurately tracking these movements over an extended period of time, routine spatial and temporal patterns and schedules can be established. Tools must be capable of reliably detecting such activity from many days worth of EO imagery and/or GMTI data with varying conditions and quality. Automated methods for establishing normalcy based on content (time of day, day of week, weather) are desired.
• Nefarious/Suspicious Activity - identifying and alerting analysts of human behavior indicative of nefarious or suspicious activity such as coordinated vehicle movements (convoys, group meetings, brush passes, etc.), unusual stops, and surveillance or counter-surveillance tactics. Tools must be able to identify such activity amongst normal traffic with a high probability of detection/identification and a low probability of false alarm.