178 
3. EXPERIMENTAL RESULTS 
The presented approach is applied to the test data collected by 
laser scanner “R1EGL Z-420i’\ The point density is about 120 
pts/'nT. Figure 5 shows the distribution of the raw data, in which 
the color is true color based on the external laser-camera 
registration. In segmentation process, the length of the cube is 
set to be 0.5cm. After distributed plane extraction and 
visualization, the results are shown in Figure 6. The points in 
blue indicate that the planes is parallel to the ground, and red 
and green colours represent those planes with normal vectors 
pointing to the east-west and the north-south directions, 
respectively. 
The boundary of the extracted planes set is produced using less 
than five seconds to compute 500 thousand points by our 
proposed algorithm in Intel Duo2 QuadCore 2.83GHz. 
However, the overall extracted boundary is hard for eyes to 
recognize due to the large scale of data set. Therefore, we show 
another result by applying the same procedures as to the 
original data to a new data set as illustrated in figure 6. Some of 
the modem buildings contain non-straight lines or curved 
boundary. In order to illustrate the capability of convex-hull 
projection to deal these issues, we created synthetic 3D laser 
scan points. It is shown that by adopting our methods, the 
boundaries are extracted correctly. 
Figure 5. The distribution of the test data. 
: v 
*1 - 
\V\N\ 
Figure 6. The visualized points.