Investigation on large batches of UAV image data processing technology based on acceleration matching method

Aiming at the fast processing demands of large batches of UAV data processing in Photogrammetry, this paper studies the GPU acceleration matching method guided by Position and Orientation System (POS) and combined with RANSAC method for excluding mismatching points. Experiments are carried out according to the Songshan Calibration Field's Mapping Project. Firstly, three groups of stereo image pairs with different typical ground features are selected for GPU acceleration and mismatching elimination test. Then using 16179 images obtained by unmanned aerial vehicle (UAV) and combining with high-precision ground control points, 50 square kilometers of high-precision Digital Orthophoto Map (DOM) over Songshan Calibration Field are produced though fast processing. The reliability and accuracy of acceleration GPU matching method guided by POS condition are verified through experiment. Meanwhile, the generated DOM can also be used as reference for further satellites, large aerial cameras and UAV cameras calibration.