Urban landscapes modeling benefits planners in drainage system design，street impprovement project selection，disaster management，and other tasks. Since buildings take an important part in urban landscape，reconstruction of building geometry in an urban environment becomes a key component of the modeling task.　　 Here，we consider flood events as an example，in which the rendering of building models is useful. The Federal Emergency Management Agency (FEMA) defined a threshold equivalent to a 100 year flood. Via a virtual three-dimensional (3D) model，then，given the inundation level of the 100 year flood，we can construct a map，that marks the buildings affected,estimates the number of floors to which the water extends，identifies survivable spaces，and charts a search sequence. In particular，virtual evacuation routes can be calculated in real-time，which can help commanders make correct evacuation commands，assist the evacuation process. At the same time，virtual evacuation process can be simulated，and then based on simulation results，the loss of property and life can be assessed.　　 In the fields of visualization and virtual reality，modeling is quite important. Most current modeling is accomplished by using existing modeling software such as 3DMAX，Maya，and AutoCAD. However，these modeling methods rely on onsite surveys accompanied by manual drafting. This is tedious and requires tremendous amounts of effort. When it comes to building reconstruction，there are two main issues-data acquisition and modeling speed. Both have been the focus of much research.　　 Images and LiDAR(Light Detection and Ranging) data are both important sources for buildingreconstruction. Images are easily acquired，and offer object position and texture information for reconstruction；LiDAR are 3D point clouds reflected from object surfaces and include object position and relationship. Usually，LiDAR data cover a large area，which is convenient for building reconstruction over large regions.　　 Different methods have been proposed for building reconstruction from images and LiDAR data，and much achieved. Despite the progress，there is still much room for improvement，especially for low density data. In the context of land-use classification，Brattberg remarked that techniques that work well for one data set and one geographic location fail when either change. One reason for such is the variations in collection densities among studies.　　 In our research，the LiDAR data were obtained from the Louisiana LiDAR Project managed byRegion VI of FEMA in the U.S in 2002. It is pre-Katrina data acquired using the Leica GeosystemsALS40 instrument.　　 Normally，the point spacing of LiDAR data used for Urban Modeling is 1 m(Point spacing is a one-dimensional measurement of points along a line.)，while the one used in this dissertation is 3 meters，which means our data is much sparser than generally applied. Productive application of this data is a considerable improvement. At the same time，it increases the difficulty of the modeling task.　　 LiDAR is also used for line-of-sight analysis，land-use classification ，and others. The latter domains,however,require data density to be high and acquired at additional cost by flying missions at lower altitudes. The densities thus obtainable are 3-12 pts/m2. The vast majority of LiDAR data will continue to be collected for flood plain maps. Current guidelines for DF1RMs (Digital Flood Insurance Rate Map) (for 1"=500，the smallest scale) are 1.2-foot (37 cm) vertical accuracy with 95％ confidence and 19 foot (5.8m) horizontal accuracy with 95％ confidence. Given these relatively easily obtained accuracies，the point cloud density is typically 0.1-0.2 pts/m2.As instruments improve，missions will be flown at higher altitudes to further reduce costs. 　　 For many applications，a lower point density can save time and costs by reducing acquisition time and data processing as well as potential data storage and handling difficulties. A 500 sq mile project area with 3 m point spacing over flat to moderate terrain may cost $200-300 per sq mile. While the same type of project with a 1 m point spacing over an area of 100-500 sq miles may cost $350-450 per sq mile.