Orientation Of Multi-image Based On Vanishing Points

Image Orientation,as the research emphasis in the development of photogrammetry,is the critical process of photogrammetry and also the foundation of the subsequent product of photogrammetry,such as DOM?DEM?DLG and 3D model.The traditional aerial triangulation has solved the high-precision calculating of exterior orientation elements of images,including bundle adjustment of regional network with additional parameters,GPS/IMU supported bundle adjustment and efficient calculation of large sparse matrix.As new photographic platform,such as low-altitude UAV(unmanned aerial vehicles)?hand-held or stent camera,be widely used in photogrammetry,the traditional photogrammetric model was broken.However,since the restrict of these these new photographic platform,traditional shooting condition and requirements can't be satisfied,which leads to the obtained images often present the characteristics of lage tilt angle?lage swing angle?small degree pf overlap and so on.This article does some research in the orientation of close-range image,adding vanishing points of inages in the process of aerial triangulation to improve the reliability and precision of aerial triangulation.This paper involves the research content focused on the following aspects:1.After studing the current classical method of image line extraction(Such as the combination of Canny algorithm with Hough transformation)and their shortcomings,an algorithm of line feature extraction based on image gradient direction,which is called LSD,is proposed.In addition,a judgment method of lines combination considering whether the angle and distance of lines are smaller than threshold,and the mergeing method based on direction and relative length are put forward.2.Adopt method of straight-line segment classification and vanishing point detection of images based on J-Linkage algorithm.Analysis the deficiency of traditional J-Linkage algorithm,and propose an improved consistency measure function between straight-line segment and vanishing point,which takes into account the length?the end points error of straight-line segment and the relative position between straight-line segment and vanishing point.3.Put forward a strategy of vanishing point supported multi-image orientation:First,obtain regional free net by SFM algorithm;Then level the regional free net with the help of nadir point,add the vanishing point information into the regional free net as well;Third,execute absolute orientation for the regional free net with vanishing point information.Last,build mathematical model of multi-image orientation supported by vanishing point and solve the model by overall adjustment.