3D Reconstruction Of Oblique Photogrammetry Enhanced By Mobile Lidar Mapping Systems

The complete 3D model of high-precision building is the digital expression of real building,which is the key research problem and key technology in the fields of photogrammetry and remote sensing,computer vision and virtual reality,and is of great attentions to academia.In view of the three-dimensional reconstruction of buildings,mobile point cloud data has a greater advantage in the building facade reconstruction,but the lack of rooftop data.However,oblique photography not only contains the vertical angle of the rooftop information good for roof reconstruction,but also contains a oblique angle.Due to the oblique problem,the results of facade recosntrution of the building does not meet the people's needs for real scenes.Therefore,based on the advantages and disadvantages of the above two methods,this paper combines the oblique image and the vehicle point cloud to complement each other for the complete fine three-dimensional reconstruction of the building,and carries out the relevant methods and process research.In the process of three-dimensional reconstruction of the combination of oblique images and mobile point cloud data,the following four problems are encountered.1)Complex urban scene based on triangular surface,automatic classification and identification of is unreliable.2)The results of the current oblique image data are mainly expressed by triangular patches,which are different from the models in current geographic information system;3)The inconsistency between the aerial image and the mobile point cloud leads to local or integral deviation;4)the feature of the facade of the building,there is no effective fusion method to improve the quality of modeling.The above problems limit the effective complementarity of the oblique image data and the mobile laser system data to the reconstruction of the complete model in the urban environment and its application in various fields.In view of the above problems,the research of this paper focuses on the related problems of data registration,scene classification and topology reconstruction and fusing aerial image data and vehicle laser scanning data,so as to propose relevant solutions.1)This paper presents a method of urban scene classification based on triangulation model.This paper focuses on the study of the geometric properties of different features expressed by triangles,and on the basis of taking up the spatial topological relation,the paper focuses on the application of the Markov Random Field classification method.At the same time,based on the results of feature classification,the influence of prior knowledge on classification results is studied,and the optimization of classification results is realized.2)This paper presents a method of building roof structure reconstruction based on hierarchical strategy.The division of the roof of the building is the key to the reconstruction of the roof,and the details of the building level are rich,easily lead to over-segmentation or under-segmentation.Therefore,this paper studies a hierarchical segmentation algorithm,through different details to describe the parameters set to achieve different levels of the roof surface division.Based on the results of different levels of roofing,this paper analyzes the characteristics of roofs at each level,and studies the reconstruction of roofs with different roofs for different roofs.3)This paper presents a method of air image and vehicle point cloud data registration.Based on the analysis of the characteristics of the high-resolution aerial image and the ground laser point cloud,the characteristics of the corner point in the cloud intensity and the optical image are studied synthetically.By using the method of robustness and a priori semantic information,an algorithm that automatically handles aerial imagery and vehicle point cloud registration.4)This paper introduces a building facade reconstruction method based on vehicle point cloud.In the real environment,the building facade is complex,often composed of multiple planes and many details.In order to reconstruct the facade of the building,this paper focuses on the automatic segmentation and extraction method of the elevation plane.This paper presents a method to optimize the facade structure of the image,and the method of elevation function based on the energy function method to achieve the three-dimensional reconstruction of the facade of the building.This research will improve the automation of the complete model of high precision building reconstruction by combining the oblique image data with the mobile point clouds data,and promote this combination in digital city,basic surveying and mapping,Urban planning,disaster reduction and other aspects of practicality and engineering process.