ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
82 
3 Procedure 
In this paper, laser-scanning system on the ground is a thing 
that shall be used. In order to map all details of objects, 
fundamental surveying is brought to data-collected process in 
terms of data acquisition. After that all data is stored into the 
proper format, those shall be filtered in order to reduce an error 
occurring during collecting processes (Baltsavias, 1999). 
Scene Modeling is a next step to create 3D model, after pass 
filtering. Triangulation, segmentation, extraction are involved 
also into Scene Modeling process in terms of mathematics 
algorithms (Ackermann, 1999). And then all data should be 
combined as one based on combination modeling to be 
established for this purpose. The combination process, 
consisting of algorithms to be used for merging different kinds 
of data, is conducted to this process. GIS attribute data, aerial 
photograph, existing map and laser data are the primary data 
to be combined. The results of work can be shown on 3D 
modeling as visualization, and also let the user to walk through 
the visual spatial data (Hala, etc., 1999). Applications for this 
modeling can be applied in several kind of work. 
Our goal towards to reconstruct large-scale 3D spatial objects 
byusing the laser range images (airborne or on the ground) and 
existing maps or image. The Principe is shown in Figure 1 and 
Figure 2. Laser range images can be thinking as a kind of high- 
resolution digital terrain models (DTMs) or digital surface 
models (DSMs). The reconstruction of 3D objects from laser 
range images can be simply thought as the 3D raster-vector 
conversions from a noisy DTM to parametric CAD formatted 3D 
vector data. Due to the limited resolutions and noisy, the 
existed low lever processing methods for reconstruction of 3D 
objects from laser range images generally can not fit to real 
application needs. Urban planning maps sever as set of well- 
organized high lever data source interpreted and generalized 
by human operators would be a kind of useful knowledge for 
3D object reconstruction and recognition. Therefore, we set the 
algorithm of 3D reconstruction into below major steps for our 
approach(see the Figure3) 
• Pre-Processing 
• Laser range image filtering and segmentation; 
• Extraction of 3D spatial objects 
• Reconstruction of 3D objects as a 3D GIS needed 
format. 
• Visual Texture mapping 
• Generation of VR environments 
Figure3 The basic procedure of this research