Acquisition of metrically-accurate imagery under field conditions is of growing importance in DoD applications such as 3-D scene reconstruction, but current methods are ad-hoc and costly because no existing products combine metrically accurate sensors with high-precision position and pose data. General Reality Corporation (GRC) and subcontractor Korbin Systems Incorporated (KSI) propose to solve this problem by developing a modular, scalable, COTS architecture and system applicable to a wide range of controlled image acquisition tasks, using a portable 3U CompactPCI running a WindowsNT real-time OS to host multiple calibrated imagers. The proposed system will incorporate a novel vision-assisted approach for obtaining highly accurate position and orientation data even during poor GPS reception in urban canyons. Our approach establishes feature correspondences between image frames, and uses correspondence information in a Kalman filter to correct inertial measurement device drift and to bridge across sparse GPS data. Expected results include geospatial/pose accuracy to in/mrad levels, sufficient to enable a new generation of image-based 3-D reconstruction applications such as automated surveying. The proposed Phase I effort will provide a strong foundation for rapid implementation and field validation during Phase II by delivering a complete requirements analysis using inputs from NIMA, DMSO, and image understanding researchers, system design to the hardware component/software flowchart level, and design, empirical validation, and simulation of the proposed navigation solution.
