Building Extraction And Reconstruction Of Three-Dimensional Geometry Based On LiDAR And Aerial Image

Face to the Intelligence city's goal,the traditional two-dimensional geographical information can not meet the needs of and application.It is an urgent need that combine three-dimensional space information and geographical attributes of elements together in urban management. Digital City mean that a modern urban management integrateing technologies such as geographic information system (GIS), Global Positioning System (GPS), remote sensing system (RS), and other key network and virtual reality,taking Government, business, public as service objects, adopting Virtual reality way to show the spatial distribution,morphology, properties, and change informationof urban ground surface features. Digital City mainly is applicated in urban design, communications industry, land management, transportation, electricity, water, fire and other industries at present. Therefore, the use of multiple data sources to establish three-dimensional digital cities have important practical significance.The key content of construction of three-dimensional digital city is building information extraction and building models of three-dimensional geometric reconstruction. The wide range building three-dimensional reconstruction of the study area based on the LiDAR point cloud, orthophoto data and topographic map.The frist step in this paper focuses on building information extraction.The LiDAR point cloud data is corrected by the geometric method. The Separation of DEM and DSM based on progressive triangulation point cloud filtering and classification algorithm. Then point cloud will be classified by the order from low to high in the second phase, surface features information extraction is completed such as vegetation, buildings, information, water information, road and bridge. Interference information available has been striped, building information and ground information is gotten.The second step in this paper focuses on Three-dimensional reconstruction of the building. The building can be separated into two parts:the roof and walls. Walls are generally perpendicular to the ground elevation. The roof surface is the focus in this study. The shape of roof includs a flat roof, sloping roof, gable roof, mountain shaped roof and curved spherical roof.In this paper, the bezier is used to form the structure of the simple and complex building. Point cloud clustered by method of regional growth based on vector, the formation of different levels of patches is achieved, the roof is divised into beziers. Peace roof and slope roof generate a bezier generally, gable roof generates two beziers, mountain roof generate four beziers. The cross-chip features of the point cloud and initial characteristics of beziers determine the building structure line of the roof. The building structure line is edited under human-assisted taking aerial photos and topographic maps 1:500 as refrence. The Optimizatial edges line of construction roof are splited and merged by the algorithm and topological relations between building model was automatically builded., Finally, the walls framed by the feature point cloud carry the building heigh, the reconstruction of architectural objects is achieved.The modeling techniques such as non-uniform rational B spline curve modeling (NURBS) and subdivision surface etc. in 3DMAX are used to achieve model reconstruction for the ball and patch the roof surface.This building three-dimensional model methods that integrated with the LiDAR and image is under the consideration of he three-dimensional characteristics of the laser scan data and high resolution and contains a wealth of surface features and texture information of image, and digital topographic map including the line edge features. Integrating these three types of data can get more accurate, more complete, more reliable for three-dimensional model of the building. A wider range of regional three-dimensional reconstruction can be realized based on three integrated data.