Rapid oblique photography building model extraction method

Abstract

The invention provides a rapid oblique photography building model extraction method. The method comprises steps that an LOD model is generated according to the oblique photography mass data, and a depth map is generated according to the LOD model; the depth map is divided into blocks, a designated height value is utilized to carry out binaryzation of each block of depth map to generate a binary depth map; the single block boundary vector data of each block of depth map is extracted; all of the single block boundary vector data together form a set to generate the target building boundary vector data; the target building model data is extracted from the LOD model according to the target building boundary vector data. According to the method, the LOD model is fully utilized, in combination with point cloud and image processing, respective advantages are integrated, so automatic mass building boundary extraction and batch extraction of single models are realized, and efficiency and effect in extracting the building models are improved.

Description

technical field [0001] The invention relates to the technical fields of three-dimensional model construction and editing, and image processing, and in particular, relates to a method for quickly extracting oblique photographic building models. Background technique [0002] In recent years, with the comprehensive development of "digital city" and "smart city", oblique aerial photography technology and 3D visualization have developed rapidly. At present, everyone pays special attention to the different applications of oblique photography 3D reconstruction technology in the industry, especially for the subsequent singulation and attribute coupling of oblique photography 3D modeling. [0003] Since the oblique aerial photogrammetry data generates integrated 3D scene data, in the management and application of Geographic Information System (GIS for short), the data cannot be managed as objects, and it is not convenient for the later 3D system to analyze the physical objects. mana...

AI Technical Summary

Problems solved by technology

[0003] Since the oblique aerial photogrammetry data generates integrated 3D scene data, in the management and application of Geographic Information System (GIS for short), the data cannot be managed as objects, and it is not convenient for the later 3D system to analyze the physical objects. management and analysis, and most of the 3D applications currently on the market simply superimpose the vector surface of the outline of the building, which appears to be a single effect on the surface, but it is not a true separation of monomers. For example, SuperMap of SuperMap in In application, by establishing a 2D and 3D integrated channel, and superimposing a 2D vector layer on the tilted model to realize the expression and operation of individualization, but the monomer in this sense only manages the vector layer and cannot Achieve real entity object management; in addition, in solving the single application, a method of combining the inclined model and the refined model is proposed, and the management of the single building is realized by superimposing the artificially constructed model on the inclined model. At present, the most widely used in the market is the DPModeler of Wuhan Tianjihang. This software uses the method of direct interaction with the original image to carry out fine reconstruction of the integrated 3D scene data, realize the output of ground object elements, and use the reconstructed model for 3D GIS application, the single model obtained by using this method can make up for the deformation caused by resolution and occlusion during automatic reconstruction, but after integrating it into the 3D scene, the constructed single model and integrated model can be compared visually and There are significant differences in the true sense of the monomer

[0004] The method of superimposing the oblique model to the vectorized layer only realizes the application requirements of GIS in a disguised form, and does not realize the single building model in the true sense. , point cloud respectively, did not make full use of the derivative product of the point cloud - the Level of Detail (LOD) model, and combined the advantages of point cloud and image processing for fusion

Images