Research On Fine Modeling Of Mengjiao Temple Based On Multi-source Data Fusion Technology

With the rapid development of China’s economic construction,it has become possible to build digital cities and smart cities.At present,the demand for spatial information data in many fields in China has gradually moved from the traditional two-dimensional to three-dimensional,especially in the real-world three-dimensional modeling has reached an unprecedented level of demand,how to obtain three-dimensional data with low cost,high efficiency and high accuracy has become a key issue in the construction of real-world three-dimensional models.The ground3D laser scanning,UAV tilt photogrammetry and airborne Li DAR scanning technology,no matter which technology for 3D model reconstruction have certain limitations,ground 3D laser scanning of large buildings due to scanning dead angle so(8(69)~′get the top of the building information data,resulting in the phenomenon of empty space;UAV tilt photogrammetry although can get the building Although UAV tilt photogrammetry can obtain information data on the top of buildings,due to the scanning angle and occlusion,the reconstruction of the surface of the model is prone to the phenomenon of pulling,for the defects of single data modelling,it is urgent to carry out research on multi-source data fusion.At the same time,different algorithms in alignment fusion have different effects on the accuracy and efficiency of model fusion,so it becomes extremely important to choose the appropriate alignment algorithm for model reconstruction.The paper,the following in-depth research is carried out to address the various issues that arise above.(1)For heterogeneous data,the classical Iterative Closest Point algorithm registration method uses the ergodic method and has high requirements on the initial position of the point cloud.Aiming at the problem of low efficiency of the traditional registration method,this paper proposes an ICP algorithm based on KD-tree nearest neighbor search optimization for point cloud registration,which can greatly improve the accuracy and efficiency of model registration.The point cloud after down sampling is preprocessed based on the human-computer interaction method.The purpose is to delete the obvious stray points and edge noise points far from the target object,and complete the initial registration of the point cloud based on the model feature points.An ICP algorithm based on KD-tree nearest neighbor search optimization is proposed to complete registration.Based on the registration algorithm in this paper,it is compared with the other two registration algorithms.The results show that the registration accuracy and registration efficiency of the model reconstruction are improved by 93.8%and 19.4%respectively.(2)Set up experiments to verify the reliability of the algorithm.A multi-level fuzzy comprehensive evaluation fusion model is constructed to calculate the weight of the factors affecting the fusion modeling.The model effect is evaluated and analyzed by Delphi method to verify the feasibility of the fusion registration algorithm proposed in this paper.In this paper,the voxel grid filtering algorithm and the ICP algorithm based on feature points and KD-tree optimization are used to perform point cloud registration and fusion experiments on the large-scale typical building’Mengjiao Temple’.The error of the inspection points before and after the registration of the feature parts of the comparison model meets the accuracy requirements of the’Digital Aerial Photogrammetry Specification’for Class I products,indicating that the overall accuracy of the fusion model is high,which has certain guiding significance for the reconstruction of other similar large-volume building models.Figure[43]Table[27]Reference[81]...