Research On Self-adaptive Feature Matching Based On Local Geometric Constraints

The field of cultural relics protection has received extensive and great attention in today's society,and the digitalization of cultural relics is a key protection method derived from the digital age.Through digital technologies such as 3D reconstruction,the cultural relics can be stored in the computer in the form of 3D model data.This not only achieves the purpose of permanent preservation of precious cultural relics,but also makes the cultural relics freed from the restrictions of time and space and appreciated by more people.In this practical application background,this paper focuses on the optimization on two core issues of slow matching and poor robustness of image matching based on feature points in multi-view 3D reconstruction technology,which is expected to make the digitalization of cultural relics faster and more real.This paper first briefly introduces the fields of cultural relic digitization and 3D reconstruction,and then conducts a detailed review and analysis on the current status of feature matching technology at home and abroad.The main research work can be summed up as follows:(1)Aiming at the problem that it takes too long to search in the global scope of feature matching,this paper proposes a grid-based feature matching method for narrowing the search range through local geometric constraints.This method can reduce the time of feature matching without affecting or even improving the quality of feature matching,and effectively increase the number of matches.(2)To solve the problem of unstable feature matching quality caused by different scene types,a self-adaptive robust feature matching method is proposed.This method can take advantage of the characteristics of the actual scene and adaptively select a more suitable matching algorithm so that the feature matching method is capable of handling more complex and changeable scenes.(3)An automatic 3D reconstruction system based on the above two matching optimization method is designed and implemented.Through comparative experiments on the actual scene dataset,the effectiveness of the proposed matching optimization method for improving the speed and quality of 3D reconstruction is verified.In summary,the adaptive feature matching optimization algorithm based on local geometric constraints proposed in this paper can not only improve the speed and number of feature matching without affecting or even improving the quality of feature matching,but also can handle more complex and variable scenes.Furthermore,it can ultimately speed up the 3D reconstruction,and obtain a more complete and accurate 3D model.