Research On Key Technology Of Terrain 3D Reconstruction Based On Sequence Images

The 3D terrain model provides a more intuitive way for size,surface area and volume measuring in land survey,which has been widely used in many fields such as land use regulation,urban development,and major infrastructure construction.Aiming to solve the problems existing in the key links of the 3D reconstruction,some improved algorithms are proposed,and the effectiveness of these algorithms are verified by experiments on an imagebased 3D terrain reconstruction platform which is built in advance.(1)An adaptive DAISY feature extraction algorithm based on image information entropy is proposed.The proposed algorithm uses adaptive SIFT feature point extractor based on image information entropy to solve the problem of uneven distribution of SIFT feature points,and uses DAISY feature descriptors to ensure robustness and distinguishability in extracting nonextreme feature points.Experiments show that the proposed algorithm can extract feature points more evenly than traditional method.(2)A matching image pairs filtering algorithm based on the comparison of regional low frequency information similarity is studied to solve the problem of lack of auxiliary information for predicting overlapping relationship between photos.Firstly,several matching pairs of feature points are acquired using the similarity of local feature points,and the matching area is selected as the comparison area;then the similarity of the low frequency information in the comparison area is evaluated by a perceptual hashing algorithm,and the similarity mean of several comparison areas is counted to determine whether it is a matching image pair or not.Experiments show that the proposed algorithm can effectively improve the efficiency of the feature point matching procedure.(3)A surface mesh reconstruction algorithm based on edge feature self-constraint is proposed.The conventional mesh reconstruction algorithms always ignore the topological information existing in the image itself and rely on mathematical methods to determine connection relationship of discrete points,thereby resulting in mesh distortion.In order to solve the above problem,we uses the edge topology information in the real scene to constrain the connection relationship of scattered points,and uses constrained triangulation algorithm based on local dimensionality reduction to construct the mesh.Experiments show that the proposed algorithm can improve the restoration degree of surface models to real scenes.