The primary function of the Feature Database (FDB) is to contain information about all the ground features seen so far in a sequence of aerial images. This information is organized so that features can be efficiently located and tracked into later images. As each new image is processed, the database grows to include more images and more features. Feature tracking is a basic component of a number of general-purpose computer vision algorithms, with wide applicability in aeronautical and space activities such as terrain modeling, mapping, GPS-denied navigation, and geo-registration. The Feature Database extends feature tracking capabilities, and should enhance its use in these kinds of applications.
The need for such a system became evident in extending the existing feature tracking techniques to a new step/stare camera system. This is a camera attached to a fast-moving gimbal, so that adjacent frames in the image sequence have very little overlap. The complicated motion means that one cannot assume a feature will appear in long sequences of images. More information needs to be stored to enable prediction of which features are likely to occur in which future frames, in order to track them efficiently. The Feature Database organizes all the required information.
The Feature Database has been applied in support of overseas DOD operations involving several planes in daily use for surveillance activities. The system uses images captured in flight to calibrate the pointing angle of the camera system to the inertial navigation system in near-real time, allowing for more accurate geo-registration of imagery. The system is automated to allow non-expert military personnel to perform the calibration on a regular basis.
This work was done by Daniel S. Clouse, Allen S. Parseghian, and Curtis W. Padgett of Caltech for NASA’s Jet Propulsion Laboratory.