3D Modeling Of Buildings With Feature Retention Based On Enhanced 3D Point Cloud Model

In recent years,with the rapid development of drone technology and sensor technology,geographic data acquisition efficient,low cost,and rapid modeling.Oblique photogrammetry have become an important technology for digital city construction.However,in the process of constructing the surface model using the point cloud data generated by the image,oblique photogrammetry has the problem of losing sharp features such as edges and corners that represent the contours of building structures.The reconstructed surface model can't fully reflect the original appearance of the building.In this article,the image data obtained during the oblique photography process is used to generate enhanced 3D point cloud model data.The enhanced 3D point cloud model is used to synthesize spatial geometric information and multi-view image color texture information to extract feature boundary lines.The feature boundary lines are used for point cloud segmentation.The point cloud obtained by segmentation reconstruction are reconstructed.Finally,the reconstructed meshes are used for mesh stitching,to achieve a complete reconstruction of highly fidelity surface model.The main research contents and results are as follows:(1)Definition and construction method of enhanced 3D point cloud modelDue to the point cloud data obtained by oblique photogrammetry has the characteristics of a three-dimensional point corresponding to different pixels of multiple images,a new data organization model is proposed,an enhanced 3D point cloud model.Compared with the traditional three-dimensional point cloud model,it adds associated information such as pixel coordinates,color,depth,etc.On this basis,an enhanced 3D point cloud model building algorithm based on open MVG and open MVS is designed to realize the construction of an enhanced 3D point cloud model.At the same time,it is noted that there is correct data in the wrong matching data in the existing algorithm,and this data is classified and recorded.(2)Feature boundary extraction algorithm based on enhanced 3D point cloud modelBased on the idea of comprehensive utilization of multiple data,a feature boundary extraction algorithm based on enhanced 3D point cloud model is proposed.The algorithm uses the edge detection algorithm to extract the image boundary line.Based on the association relationship between the image and the point cloud in the enhanced 3D point cloud model,the algorithm uses the point cloud to assist in removing the erroneous feature boundary obtained by image detection.Based on the the association relationship of different images in the enhanced 3D point cloud model,the algorithm uses different image edge lines to supplement the missing correct feature boundary lines to obtain a better image feature boundary line result.After using the enhanced 3D point cloud model to convert the image feature boundary line into a 3D feature boundary line,the feature boundary line is closed by the correspondence between the 2D line segment and the 3D line segment,thereby obtaining a better feature boundary line.(3)Point cloud surface reconstruction algorithm based on the idea of segmentation and stitchingBased on the idea of segmentation and stitching,Point cloud segmentation is performed using feature boundary lines.Surface reconstruction is performed on the segmented point cloud,and then the surface model is stitched based on the feature boundary lines to achieve surface reconstruction while maintaining features.The reconstruction results show that,compared with direct surface reconstruction,this method can effectively retain the sharp features of the object surface.