A Oblique Images Feature Matching Method Based On Parallax Relaxation Transmission Constraints

Aerial oblique images has become the main data source of large-scale urban 3D reconstruction,but there are large pitch angle,roll angle and big difference between oblique images.There are still obvious reliability problems in automatic feature matching of oblique images,so interactive tie points are essential.Existing interactive matching methods usually only depend on epipolar geometrical information,and the constraints are so weak that the success rate of semi-automatic matching is low.This paper presents a kind of oblique images feature matching method based on parallax relaxation transmission constraints.Based on the continuity hypothesis of parallax,a thin plate spline function which can express approximately the object parallax information is fitted by known tie points,and combined with the epipolar information to determine appropriate search area.The pixels in the search area are sorted by distance reliability,and the template matching is carried out in geometric order until the correlation coefficient satisfies the threshold.Experiments show that the matching speed and reliability of the proposed method are significantly improved compared with the traditional feature matching method,which provides a guarantee for subsequent application of oblique images.The main contents of this research are organized as follows:(1)Firstly introduces the background and significance of this paper,and summarizes the situation and the development trend of the oblique images feature matching at home and abroad,and puts forward the research content and organizational structure.(2)Aiming at the aerial oblique images feature matching,introduces the theoretical basis of aerial oblique images feature matching.And discusses the structure and characteristics of aerial oblique image data,analyzing the geometric relationship of oblique photography.And applies the parallax relaxation transmission constraint,the epipolar constraint,the gray similarity constraint and the two-way consistency constraint to the feature matching of the oblique images,optimizing the matching results by the least square method.(3)Based on the theory of oblique images feature matching,implements oblique images feature matching algorithm considering parallax relaxation transmission constraints.It includes data preprocessing method,thin plate spline function fitting method,oblique images feature point search strategy and reliability optimization method.(4)The algorithm is verified by real images data in Shenzhen area.The experiment shows that,compared with the traditional oblique images feature matching methods,this method determines the candidate point search area based on the parallax relaxation transmission constraint and the epipolar constraint,taking into account more geometric information and can solve the image ambiguity problem well.Using the best first matching strategy and introducing local geometric order constraint,gray similarity constraint and two-way consistency constraint,the error matching is reduced significantly,and the reliability and efficiency of the oblique images feature matching are improved.