Layered Sensing Testbed

The LST is a software and data base environment that integrates layers of sensor video feeds in real time onto a precision geo-referenced database for the visualization, analysis, and exploitations of multi-INT video streams approaching terabytes per hour. The key technology innovations include a structured approach for integrating, registering, and referencing multiple layers of massively heterogeneous motion imagery. There are three components to the LST: the integrations and visualization software, the analysis & exploitation software, and the advanced GIS database architecture that turns every pixel of interest into a Wikipedia or PixelPedia. The integration & visualization software integrates and registers real time video streams from multi-INT air and ground sensor networks into a precision geo-referenced Google-Earth like database architectures. The foundational software for this function is the NASA World Wind. Unique application software was developed to give the environment more sophisticated imagery manipulation capability. The analysis and exploitation software focuses on the data processing, formatting, and representation that allows for the extraction and interpretation of patterns of behavior within a city environment, whether it be vehicle or people traffic. The ability to ingest massive video streams across a heterogeneous set of sensor layers allow for an extremely rich range of analysis options to study behavior at the city wide level, the neighborhood wide level, the city block level, down to the specific building level giving the user an unprecedented degree of situational awareness. The advanced GIS database architecture called PixelPedia is crucial to achieving meaningful situational awareness over a complex urban setting. The concept is elegant in its simplicity. The PixelPedia data structure allows for the preloading of apriori knowledge that acts to bootstrap ISR operations. The resulting insights or declarations can then be stored in a geo-referenced data structure that is directly associated with the sensors in operation. This in affect gives the area of regard the ability to remember and to learn. The belief is that this will increase probability of detecting anomalous behavior while also reducing the probability of false alarms, the holy grail of advanced pattern recognition. Benefits There is no comparable existing technology that has demonstrated the ingestion of multiple video streams geo-referenced to a learning data base architecture. The LST will provide a unique capability to ingest massively heterogeneous data streams and efficient extract out understanding within a framework that learns.