Knowledge Engineering

The knowledge engineering focus under the Center for Advanced Sensors is to advance knowledge representation and inference algorithms for machine intelligence applications that will advance sensor information utilization through enhanced understanding of sensor data and metadata, sensor fusion and interoperability networks.

Projects

OntoSensor: a prototype sensor knowledge repository compatible with evolving Semantic Web infrastructure.  OntoSensor includes definitions of concepts and properties adopted (in part) from SensorML, extensions to IEEE SUMO and references to ISO 19115.  Simple queries have been developed and tested using Protégé 2000 and Prolog.  Although OntoSensor is in the early development stage, it presents a practical approach to building a sensor knowledge repository.  It is proposed that OntoSensor may serve as a component in comprehensive applications that include more advanced inference mechanisms. Such comprehensive applications will be used for synergistic fusion of heterogeneous data in a network-centric environment. OntoSensor is a work in progress.  The current OWL is available here.

Performance Modeling of Sensors with Image Processing Enhancements

This focus under the Center for Advanced Sensors aims to develop effective performance models for sensors with image processing enhancements.

Projects

Perception Studies: Sensor performance models are the only cost effective tool for trade studies involving new sensors, sensor design modifications and determining the value of image processing. Human observer studies from perception laboratory experiments directly support and validate the performance models. Validated performance models (with appropriate human observer support) also provide limiting performance information for automatic target recognition (ATR) algorithms. This is particularly true in discrimination tasks for recognition and identification, although less so for detection. Therefore, perception studies are a cornerstone for sensor performance models and will be used extensively for incorporating signal processing enhancements into sensor performance models. Image quality metrics will be investigated throughout this effort. Such metrics offer the possibility of developing an effective performance model for situations that have an image as an input to the enhancement algorithm and have an image as an output.

Performance Modeling of Advanced Architecture Systems

Existing performance models are physics based descriptions of the components and subsystems that form images in the non-terahertz regime. These are generally either infrared (IR) or visual optical modeling systems or electromagnetic field (EMF) based modeling systems in the millimeter wave (mmwave) or at radio frequencies (RF). Though the terahertz (THz) regime has many characteristics in common with these neighboring regions, some aspects are significantly different. The relatively large wavelength in the THz region in conjunction with practical noise limitations encourages imaging system designs utilizing complex coupling antenna, lenses, or other impedance matching / mode selection techniques. Further, the image collection optics will not have as large a ratio of diameter to wavelength as is usually assumed by the previous models. The image formation process associated with convergent wave fronts is anticipated to be a function of the total effective reception pattern of the focal plane detection inclusive of antenna pattern and or lens if present. Proper prediction of imaging performance via computer modeling necessitates the inclusive of these aspects in the modeling. Atmospheric conditions and limitations unique to the THz regime influence the performance modeling. The existing optical performance models should be modified to account for these differences and will be enhanced to accommodate anticipated system architectures. Existing Army Research Laboratory (ARL) programs address this in the broad sense. This focus area intends to compliment and enhance this effort by concentrating on selected issues of mutual interest. This will include corrections for deviations from the classical optical performance model descriptions.