Automated Pixel Geo-Registration for Precise Imaging

SimWright Inc. proposes a capability to georegister objects for gimbaled hyperspectral systems in: surveillance, reconnaissance, and tracking modes while enabling targeting. The concept proposes significant georegistration error reduction between features commonly identified in multiple sensor sweeps than traditional systems by combining existing tiepoint correlation algorithms integrated with a Kalman Filter and correlated with periodic DPPDB/ NTM updates. Three trade studies will be performed 1) evaluating feasibility of adapting tie-point algorithms, and KF to perform frame to frame correlation in real-time, 2) evaluating extending the capability for georegistration across full sensor sweeps and 3) accuracy, cost and feasibility tradeoffs in system and subsystem design. The complex nature of subsystem errors in real-time gimbaled sensors requires an accurate knowledge of subsystem error components. Combining tiepoint algorithms, a particular Kalman filter, and a DPPDB/ NTM (when fielded) update capability permits engineers to diagnose system errors to create error models (e.g. platform attitude (yaw), gimbal errors). We propose a cost effective flight test for final validation of the Phase I algorithmic architecture. SimWright has a ten year heritage integrating inertial navigation systems, and developing software solutions for photogrammetry, image processing, analysis, 3D visualization and metadata creation.