Vision-Augmented GNC: Passive Ranging from Image Flow
Matthew David Markel, Juan Lopez, Glenn A. Gebert, Johnny H. Evers · AIAA Guidance, Navigation, and Control Conference and Exhibit · 2002
Knowledge of range and closing velocity is known to enhance missile guidance, improve launch envelope, and reduce miss; however, scenarios exist where the range-to-target is unknown to the weapon. This paper provides the mathematical framework and algorithm for calculating range-to-target from sequential optical (intensity only) images. Such use of additional information from imaging seekers to assist GNC falls into the general category of Vision-Augmented GNC (VA-GNC), and a brief overview of this emerging fleld is provided. It has been shown that through use of the optical ∞ow constraint (OFC), a range expression at each image pixel may be directly calculated from spatial-temporal gradients, known pixel angular location, and ownship motion. However, these optical ∞ow/gradient-based passive ranging algorithms are subject to singularities, noise, and bogus range estimates for certain image content. In this work, a local median flltering approach is applied to the raw range estimates to remove outliers before segmentation. Using this approach provides the additional beneflt of both intensity and range information for segmentation. Once segmented, the range to the target of interest is extracted and available for guidance or range-rate estimation. Finally, simulation performance results are presented using synthetic, analytically derived images for the case of a body-flxed