Research And Application Of Feature-enhanced Graph Matching Algorithm Based On Regional Estimation

Cultural relics are the witness and carrier of human material civilization and spiritual civilization.With the rapid development of modern science and technology and economy,the social and economic value of cultural relics has been paid more and more attention.But because of natural disasters,man-made damage and material aging and other reasons,precious cultural relics gradually disappear,so it is urgent to protect and repair related technologies.Digital technology is an important way to protect cultural relics and spread culture,in which multiview 3D reconstruction based on feature point matching algorithm can preserve and restore the appearance of cultural relics more vividly.But in reality,there are many factors that interfere with the reconstruction effect.Because of the influence of light and shadow,lack of texture or occlusion,the number of detected feature points is small,the algorithm of feature point matching is not robust,and a lot of mis-matching occurs,so that the accuracy and integrity of the point cloud model can not be guaranteed.Therefore,it is an important problem to study how to obtain enough features,robust feature matching algorithm and eliminate false matching algorithm to improve the application effect of 3D reconstruction technology in cultural relic protection.This paper focuses on the 3D reconstruction effect of multi-view 3D reconstruction technology in the protection of cultural relics,aiming at the key problems such as insufficient features,low accuracy of feature matching,insufficient integrity of reconstruction and high mismatch rate.This paper mainly from the following aspects of work:(1)A graph matching algorithm combining local and global information feature enhancement of graph nodes is proposed,which mainly combines local and global spatial structure information of graph nodes for feature enhancement.By experiments on three datasets: artificial dataset,CMU house dataset and real image dataset,the matching accuracy is effectively improved,which is superior to a large number of advanced graph matching algorithms.(2)A graph matching algorithm based on region estimation feature enhancement is proposed,which mainly integrates spatial structure to enhance feature while filtering feature points,and uses region estimation posterior method to eliminate false matching.Through the feature point matching experiment in the public building image dataset,the matching result with very low error matching rate is obtained.(3)Based on the classical Colmap 3D reconstruction software framework,the application examples of the 3D reconstruction of cultural relics are realized,the paper mainly combines the graph matching algorithm and reconstructs the complete ancient architecture model and the local details of the buildings with clear texture...