Analysis of Simulated Imagery for Real-Time Vision-Based Automated Aerial Refueling

Abstract

This Paper presents a three-dimensional graphical simulation (virtual world) that replicates a complete aerial refueling scenario including the tanker, the tanker’s stereo vision system, and an approaching receiver aircraft. This virtual world generates real-time imagery of a geometrically accurate receiver flying a physically realistic aerial refueling approach. The approach is recorded from the perspective of the tanker’s virtual stereo cameras. The stereo imagery is processed via a novel stereo vision pipeline that dynamically computes the position and orientation (pose) of the receiver relative to the tanker in near real time. Additionally, to increase the visual fidelity, emphasis is given to quantifying the negative aliasing effects caused by the rasterization of the virtual world. An antialiasing solution that mitigates these artifacts and allows simulated imagery to better approximate physically captured imagery is proposed. Results show antialiased imagery reduces epipolar error by a factor of 4. This reduction more than doubles the accuracy of our stereo vision algorithm at the refueling contact point. As simulated imagery is increasingly used to test and develop aeronautical vision pipelines, this aliasing must be mitigated to minimize artifacts intrinsic to rasterized synthetic imagery that negatively affect vision processing algorithms.