Detection and multi-application Algorithms
PSI is focused on the development and implementation of the next generation detection algorithms for identification of signatures and anomalies in highly complex and cluttered backgrounds. Advanced algorithms have been developed for varied applications that exploit knowledge of sensor systems and are dependent and/or independent of PSI sensor capability. PSI has successfully demonstrated the utilization of a real-time integrated processor performing chemical detection and identification in less than 1 second for complex backgrounds (including urban environments) through the use of a highly adaptable ATR algorithm. The algorithm utilizes a spectrally-matched-filter-based approach coupled to a statistical scene background estimation algorithm that does not require a prior knowledge of the background, also enabling ‘on-the-move’ detection.
Application of PSI’s ATR algorithm: a) Detection of R134a burst at Redstone Technical Test Center. b) Detection of NH3 release at Porton Down. c) Over-water detection of TEP release from ~ 3 km standoff d) Simulataneous detection of TEP release from 3 sensor locations at Dugway Proving GroundIn addition to ATR algorithms for use with multispectral imaging, PSI has been developing highly advanced algorithms for sensor placement optimization and sensor fusion applications.
As chemical and biological agent (CBA) sensors become more reliable, more affordable and more environmentally rugged, it becomes increasingly practical to deploy multi-sensor networks to warn of CBA contamination threats to population centers and material assets. While sensor data fusion is an area of active research and considerable effort has been devoted to developing algorithms for optimal fusion of information from individual sensors in sensor networks, comparatively little research has been conducted to develop algorithms for determining the optimum physical placement of the sensors in the network. Determining optimum CBA sensor placement is a particularly challenging problem when the number of sensors is small (<10) and the area to be protected is large (>100 km2). We are developing a sensor placement algorithm and associated software tool to facilitate selection and placement of point and standoff CBA sensors to accomplish a specified asset protection objective. The software tool which incorporates the placement algorithm is being developed with a GIS (geographic information system)-based graphical user interface and processes GIS map information to generate inputs to the placement algorithm.
Optimized sensor placement solutions for the protection of base facilities and forces deployed in the field
An algorithm that is capable of fusing disparate data sources including hyperspectral standoff sensor data products, weather information, and point sensor data generated from a deployed network of sensors is also being developed. The software capability is intended to provide a systematic approach to fusion of disparate sources. The tool will: 1) automate the analysis, data fusion, prediction of cloud tracks, as well as the visualization of the results and associated confidence levels, 2) provide a higher confidence global assessment of the threat. The software architecture is being designed to be open and modular in order to allow for fusion of any type of data and at any location where multiple sensor sources can transmit information.