| 3D modeling technology has an irreplaceable role in the fields of conservation and digital display of ancient buildings.At present,in the field of ancient architecture conservation and visual display,oblique and 3D laser scanning technology are the most widely used 3D reconstruction techniques.However,the above two technical means have certain defects and shortcomings in the quality of 3D reconstruction models,which limit their further promotion and application in the field of ancient architecture.In order to realize fine 3D modeling of ancient buildings,this thesis carries out the following studies:(1)By using oblique and 3D laser scanning technology,respectively,3D reconstruction of ancient buildings,through the analysis of the accuracy and texture structure of the 3D reconstruction model of ancient buildings,summarized the advantages and defects of the single modeling approach in the modeling of ancient buildings.In response to the advantages and defects of single modeling approach,a fine modeling scheme of ancient buildings that combines two modeling techniques with complementary advantages is proposed.(2)Ancient building point cloud data thinning and alignment method research.Point cloud model data thinning and alignment is the key to ensure the fine modeling of ancient buildings.The advantages and shortcomings of the classical building point cloud extraction method and ICP point cloud alignment method are analyzed,and the curvature-enveloping box extraction method and the improved ICP alignment method with rotation angle constraint are proposed to combine the two classical building point cloud extraction methods.In the refinement modeling stage of multi-source point cloud fusion of ancient buildings,the effect of different point cloud thinning and alignment methods are compared and analyzed through point cloud model thinning and alignment experiments.(3)The quality analysis of ancient building fusion refinement modeling model.The dense matching point cloud model of oblique of ancient buildings and the 3D laser scanning point cloud model are diluted,aligned and redundantly deleted for fusion refinement modeling.The fusion modeling model of ancient buildings is analyzed in comparison with the oblique model and 3D laser scanning model in terms of both model accuracy and texture structure.Finally,the following conclusions are drawn.(1)The curvature-envelope box extraction thinning method and the rotation angle constrained ICP alignment method have better results in the extraction thinning and alignment of the point cloud model of ancient buildings.Multi-source point cloud fusion modeling experiments using curvature-envelope box pumping thinning and ICP point cloud alignment with the addition of rotation angle constraints,the analysis shows that the curvature-enveloping box thinning method can retain the geometric features of ancient architectural point clouds better than other classical building point cloud thinning methods,and it is not easy to form point cloud voids in the non-geometric feature areas of point clouds;Compared with the ICP alignment method,the rotation angle constrained ICP alignment method outperforms the ICP alignment method in the alignment of ancient architectural point clouds with lower overlap rate,using the Euclidean distance of matching points as the criterion,both in the distribution of matching points to Euclidean distance and the mean and variance of Euclidean distance.(2)The multi-source point cloud fusion refinement modeling model has reliable accuracy and clear & complete texture structure.Based on the analysis of the advantages and inherent defects of the single modeling of oblique and 3D laser scanning of ancient buildings,the fusion refinement modeling is carried out and compared with the oblique model and 3D laser scanning model.In terms of model accuracy,the fusion modeling model of ancient buildings has better positional and geometric accuracy than the oblique model and is close to the 3D laser scanning model;in terms of texture structure,the fusion modeling model of ancient buildings has no blurring and distortion,and the details are clearer.The structure is complete,without deformation or missing phenomenon. |