Research On High Accuracy Image Matching Of Feature Points In Automatic Aerial Triangulation

The development of digital photogrammetry technology has increasingly become an important method of acquiring and updating the national basic geographic information data,providing a rich source of information for the measurement and updating of the national geographic information data.The development of digital photogrammetry technology has increasingly become an important method of acquiring and updating the national basic geographic information data,providing a rich source of information for the measurement and updating of the national geographic information data.At the same time,digital photogrammetry technology plays an important role in national basic economic construction,and is widely used in urban planning,forest and geological survey,development of land resources and other fields.Image matching technology is the key technology of digital photogrammetry implementation.And it is the key issues and core content of common research in fields of digital photogrammetry and computer vision.Image matching technology has a broad application prospects in the fields of photogrammetry,computer vision,urban three-dimensional modeling and military applications et al.In the field of the photogrammetry,image matching is the foundation of automatic aerial triangulation,is also the foundation of subsequent spatial data production,is also the basis of the artificial targets such as the buildings’3D reconstruction.The present studying of the image matching algorithm is more and more,the requirement of matching accuracy also more and more highly,for the aviation images’ automatic aerial triangulation,the requirement for the image matching with high quality and high precision is particularly important.Aviation image acquisition will be suffered from the influence of natural scenes and the changing of the weather,so some feature points matching algorithm would have false matching points,affecting the matching quality,reducing the matching accuracy,eventually leading to lower quality of the automatic aerial triangulation.In this dissertation,the problem of high accuracy image matching of feature points in aerial images’ automatic aerial triangulation will be studied and discussed,focusing on the two aspects of the researching and experiments for the conventional images’ and oblique images’ high accuracy matching algorithm.In order to get the high accuracy for the conventional images matching,we propose a new matching algorithm based on elevation plane constraint,we called it“object space matching method".In the process of matching with the method,we added the point offset in searching,at the same time,the point offset can also providing a reliably and stability initial value for the Least Square Matching.For the problem of the high accuracy of the oblique images matching,we combining the characteristics of the object patch matching,proposing a oblique images high accuracy matching algorithm,which based on the object patch matching.The oblique images matching method computing the H matrix between object patch plane and image plane,and transformation the information of image plane to the object patch plane by the H matrix,and combined with least square matching method to high accuracy matching.In the experiment,in order to verify the correctness of the method,we design the contrast experiments of the MPGC method,the VLL method and the proposed method for the problem of conventional images’ high accuracy matching algorithm,and the experiments’result shows that the proposed method is better than others in matching results and bundle adjustment results.we also design the contrast experiments for the problem of oblique images’high accuracy matching,and the results show that the proposed oblique images’high accuracy matching algorithm based on object patch improving the matching result and the accuracy of bundle adjustment.