Novel Airborne Video Sensors. Super-Resolution Multi-Camera Panoramic Imaging System for UAVs

摘要

Application of a camera array as a flexible, reconfigurable, inexpensive high-resolution panoramic motion-imagery sensor for low-altitude reconnaissance aircrafts is investigated. Assuming multiple-view noisy image position measurements of terrain features and known camera projection matrices by calibration, terrain feature localization and UAV positioning are analyzed by computer simulations, with/without supplementary gyro and GPS. How various system parameters impact the achievable precision of panoramic system in 3-D terrain feature localization and UAV motion estimation is determined for the A=0.5-2 KM) FLIGHT ALTITUDE RANGE. Enhancement of estimation accuracy from GPS and gyro is explored. Estimation error variance plots are given as a function of camera resolutions, viewing angles, flight altitudes, GPS and altitude measurement errors, number of views, etc. Selected results, from point correspondences in 4Kpixx4Kpix images and utilizing GPS readings with one-meter error variance at 0.5-2 KM) ALTITUDES, COMPRISE: Estimating 3-D coordinates of ground features tracked in 1-2 dozen images with A/10 baselines at sub-meter accuracy; Determining UAV pose with 0.1-0.3 DEG variance by matching 2-3 dozens of features in two views. The results provide valuable guidelines for the integration of camera-array images into one super resolution panorama, registering multiple panoramas to construct a single composite view, integration of visual survo with onboard sensors, map-based navigation and AUV positioning. Computed performance charts enable the design of optimal high-resolution imaging system based on the UAV size and capability constraints.