SAVig is a joint project between AFRL, DaytaOhio, and Kno.e.sis Center. Under this effort, sensor exploitation algorithms are being developed for a wide variety of modalities including EO, acoustic, IR, HSI, and RF. Key to most exploitation algorithms is some form of imaging to produce a spatially meaningful signal representation and exploitation of advanced knowledge of sensor phenomenology to develop adaptive models for the observed environment. Detection of targets or anomalous events is then a matter of distinguishing deviations from normal scene behavior. In the past, deviation has been relative to neighboring image pixels, but modern signal exploitation algorithms develop normalcy models by integrating information along the time dimension. To mitigate the large number of false alarms a single sensor would produce in an urban environment, multiple sensors of varied modalities (e.g., RF, EO, IR, HIS, acoustic) are employed to provide persistent observations across time, space, and spectrum. However, cross-modal sensor fusion remains a key hurdle in multi-sensor exploitation. Effective automation of sensor data fusion requires ability to translate heterogeneous sensor data into a common format, to register data accurately within a common spatiotemporal reference frame, and to identify and associate cross-modal phenomenal attributes. To accomplish this task, SAVig is extending state-of-the-art sensor data representation languages, SensorML and TransducerML, to integrate semantic associations into sensor models. Semantic descriptions of concepts and associations within sensor data representations will enable entity and event detection and sensor data fusion in multi-level, multi-modal streaming sensor data.

• Sensor Data Management Summary (short)
• Sensor Data Management Survey (long)