ISPRS Workshop on Service and Application of Spatial Data Infrastructure, XXXVI (4/W6), Oct.14-16, Hangzhou, China 
188 
sequence is nearly level and the others in the sequences with 
oblique photography are oblique. Three flying routes are shown 
in Figure 1. 
Because image pixels outside the near-nadir area cannot be 
accurately correlated with the laser points, traditional film 
cameras are impractical for the collection of imagery for this 
application. Medium format digital camera system is ideal for 
this treatment of LIDAR data. The longer focal length and 
smaller field of view virtually matches the swath width of the 
laser system allowing the proper alignment of laser data and 
pixels. 
3.2 Laser Scanner 
Airborne laser-scanning has become a viable technique for the 
surveying data during the past few years. As an active technique, 
it delivers reliable height data without requirements to surface 
reflectance variations. The inherent 3-D nature of laser 
scanning data saves time consuming and reduce potentially 
erroneous matching techniques and yields a high potential for 
real time application if laser ranger data can be fused with 
GPS/INS data onboard in aircraft. 
Figure 2. LIDAR point cloud with RGB, perspective view 
3.3 Registration Between LIDAR And Images 
Traditional aerial photographs overlapped about 60 percent in a 
strip. While image sequences taken with digital camera has the 
advantages of high overlapping and redundancy of 
corresponding features, which has a well potential for automatic 
3D reconstruction. 
Automatic aerial triangulation technique can be adopted to 
acquire initial values of camera parameters, linear features were 
used for pose estimation. Many conjugate lines are 
automatically generated by extraction of linear primitives from 
images and laser data. The exterior orientation parameters of the 
images are calculated based on the theory and the arithmetic of 
the line photogrammetry , in which the conjugate lines are used 
as the observation values. The coplanar condition was used as 
error equation to resolve the external orientation parameter of 
digital image. 
3.4 Preprocess in Multi-Spectrum 
The precise calibration and alignment of these two subsystems 
make it possible for the software to integrated LIDAR and 
Color / CIR digital camera systems. 
Multispectral pixels to be photogrammetrically associated with 
individual X,Y,Z values. The imagery is not overlapped to the 
surface, rather each laser return is mathematically projected 
through collinearity equations onto its proper position on the 
taking into consideration the camera model, each surface point 
possesses an accurate spectral signature assigned to its location, 
allowing accurate classification of features using conventional 
remote sensing techniques. 
Figure 3. Multi spectrum of integration system: Color , CIR, 
Elevation and Intensity (Data provided by TopoSys) 
In the segmentation, we compute an approximation of the 
topographic surface by mathematical morphological 
filtering .To reduce noise caused by small objects such as 
antennae or chimneys on roofs ,the median filter was used . To 
Separate the building from other blobs ,such as trees , 
classification can be processed by combined multispectral 
information. 
3.5 Multi-View Feature Matching 
Linear features were extracted in the digital camera image . We 
are interested in 3D straight lines because they are prominent in 
most man-made environments, and usually correspond to 
objects of interest in images, such as buildings and road 
segments.3D position could be calculate by intersection of 
homogeneous lines .We utilization of three views to 
compromise between complexity and quality of the results. 
Trifocal constraint are imposed to get exactly right match result. 
Figure 4. Trifocal geometry. 
3.6 Model Construction 
The initial knowledge database is established by the information 
extracted from existing geographic data .Colour cues, expressed