Coherent and non-coherent change detection algorithms produced from SAR imagery have enormous intelligence value due to their sensitivity to change. But they are also difficult to interpret because of high false alarm rates and because of the basic statistical nature of SAR imagery which makes automatic filtering algorithms very difficult to robustly implement. This necessitates man-in-the-loop analysis in order to obtain favorable false alarm and detection probability rates. The challenge gets both better and worse as SAR data becomes more prevalent; more information provides an opportunity to improve false alarm rate, but without more sophisticated processing data overload results. Neva Ridge Technologies proposes to improve the interpretability of change detection products. This includes identification and masking of pixels with an elevated false alarm rate due to well understood phenomenological processes. It also includes adaptive processing to perform spatial averaging in a way that avoids edges, thus preserving resolution. We also propose to study the correlation of change signatures in multiple change detection products. This has two important
Benefits: improving performance in persistent surveillance operations, and combining and reducing the amount of imagery which would need to be interpreted by an IA. We also propose concepts for improved display products.
Benefit: This work effort would contribute target detection and recognition algorithms for which there is significant interest in the DoD and intelligence community. This capability could also be used in civilian applications such as land use classification and monitoring, watershed and aquifer health, and hazard management.
Keywords: Sar, Ccd, Nccd, Coherence, Remote Sensing, Region Growing, Persistent Surveillance, Exploitation
