Multi-sensor Registration Tools-ANDRO's Spatial Multi-sensor Autonomous Registration Toolkit (SMART)

The objective of this effort is to develop autonomous technology in the form of a software toolkit for the spatial alignment of image, SIGINT, and GMTI data collected from the battlespace. Some work has been directed toward the spatial alignment of images, SIGINT and GMTI data collected from the battlespace; yet, the understanding of precisely how to extract and use this data for relative and absolute registration remains limited. The focus of this research is to develop novel approaches to geo-registration by first developing a generic concept of geo-reference invariant feature that can be extracted from each data source and second, developing algorithms for correlating/integrating these salient features across the multiple data sources and a fiducial database for relative and absolute registration. Examples of features that cross multiple source data domains are radio/TV transmission towers, command and control centers, road intersections, fixed search radar towers, and cellular phone towers. Trade-off studies will identify correlation/integration algorithms that will tie disparate types of sensor data together with the fiducial database while exploiting the knowledge of physics and phenomenological models (exact or approximate) of sensors and source types that might be available. The research will exploit innovative approaches including knowledge-based (AI/expert) systems, multiresolutional architectures, and evolutionary computation. Most appropriate approaches will be identified for each sensor, source type, and for correlating all the data and available knowledge. The result will be a reliable, flexible system that can coregister multiple sensor data, estimate and account for errors and uncertainties, and identify measures of effectiveness for automatic convergence to minimize human intervention. The algorithms will be implemented as a software toolkit to enable the flexible system integration for multi-source data. This effort will exploit innovative approaches for image pair prescreening and integrated intelligent computational systems. Whereas R&D and testing are currently being performed to exploit some well-known algorithms (mutual information (MI), wavelet, and feature based), this effort seeks to further extend and optimize work on existing registration algorithms for robustness as well as developing new ones and integrating them into a tool suite. The Phase II SMART toolkit concept will employ a real time expert system based approach to automatically "broker" or select the appropriate geo-registration and correlation algorithm(s) for multi-source data domains. The innovation research will focus on the following: (i) use of optimizing algorithms that apply a multi-source data registration/fusion tree methodology and AI decision making to achieve the goals of autonomous registration; (ii) novel techniques for image pair preprocessing using intensity and knowledge based methods that derive the "frequency content" of images and broker the selection of the appropriate registration algorithm; (iii) methods for handling 2-D rigid and elastic body problems using photogrammetry-based methods; and (iv) new methods to provide tractable error/uncertainty handling and registration performance/effectiveness measures. The thrust in Phase II will be on, but not be limited to the use of MI, feature (control point), and wavelet based methods for multi-modality IMINT and for non-IMINT cases. This capability will directly serve the needs of DoD product centers (ASD) and acquisitions (Joint Strike Fighter) concerned with advancing the state of the art in sensors, electronics, and battlespace technologies for achieving the goals of rapid (near real time), onboard precision target geo-location. SMART technologies will be readily useable by a sensor-to-shooter warfighter, battlespace commander, and airborne surveillance operator.