Assessment of Monocular Optic Flow Analysis for Ballistic Trajectories

This paper describes a preliminary study to assess the potential optical-flow performance of a simple on-board camera to provide a low cost navigation solution for unmanned systems such as drone aircraft or other types of projectiles. For decades, complex electronics systems such as the Global Positioning Systems (GPS) and sophisticated inertial measurement units equipped with rate-gyroscopes and accelerometers have proven to be very accurate for navigation updates. However, these systems are costly and may be prohibited from proper function under adverse conditions. For this study, 3D scene generation software developed by Diamond Visionics was used to simulate the view of a forward-looking camera system during a ballistic descent. The algorithm tested for this study does not perform geo-registration to match known locations on the ground from reference imagery, but instead relies only on knowledge of the intrinsic parameters of a properly calibrated camera/lens system for measuring orientation updates and has the potential to measure changes in absolute position provided that another source of velocity (such as a priori estimates or additional sensors) is available. Preliminary results show that reasonable estimates for projectile position and angular rates can be made. These results are made even more accurate when using a tuned bundle adjustment algorithm.