数字航空摄影三维重建理论与技术发展综述

As one of the essential branches of photogrammetry, aerial photogrammetry has made significant progresses in the past decades. The emergence of various new working modes,such as oblique aerial photography and UAV photogrammetry, has brought new challenges to traditional aerial photogrammetry and has spawned many new solutions. In addition, new theories and techniques in computer vision technology and deep learning technology in artificial intelligence are continuously integrated into aerial photogrammetry, promoting the development of aerial photogrammetry in the direction of intelligence and automation. Therefore, modern aerial photogrammetry is already the product of the fusion of multiple sensors, the combination of various data acquisition methods, and the intersection of computer vision and machine learning technologies in traditional photogrammetry and artificial intelligence technologies. Three-dimensional reconstruction is the core problem of aerial photogrammetry. This paper expounds on the problems existing in modern aerial photogrammetry, summarizes and discusses the current research status from four aspects:corresponding tie points extraction of aerial images; block adjustment; image dense matching; and singleton modeling. And finally this paper gives the mainstream aerial image photogrammetry processing framework at home and abroad.