The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
Block kriging result: grid size 1m‘1m 
Figure 8. Block Kriging with a block size of 1 m. 
Block kriging result grid size 2m*2m 
Figure 9. Block Kriging with a block size of 2 m. 
Figures 9 and 10 show the Kriging results and the standard error 
in raster format, respectively. In Figure 10, each shaded square 
is centred at the intersection of the gird. Each grid is shaded 
with a different grey tone according to the elevation value 
estimated at that location. When the elevation is high, the 
colour is darker. From this figure, we could see Kriging could 
restore the scene when the change in elevation is continuous. 
But at the building edge, the boundary is blurred. That is due to 
the spatial averaging. Figure 11 is the corresponding standard 
error of Figure 10. This map allows evaluating the precision of 
estimation at any part of the region. The whiter squares 
correspond to areas with smaller error. When compared this 
error diagram with the raw data points shown in figure 4, the 
relationship can be easily found. The areas with high point 
density correspond to areas with low estimation errors. That’s 
to say, as the point density increase, the estimation error 
decrease. Big error occurs at the boundary of the whole dataset, 
this was due to the lack of data at the boundary. 
5. CONCLUSION 
This paper provides a new approach to rasterize raw ALS point 
clouds. Block Kriging is used as an interpolation method to 
estimate elevations on a regular grid using irregularly spaced 
ALS point clouds. Firstly, the spatial structure of the data is 
analyzed by considering the spatial autocorrelation between 
data points. The spatial variability of the data is integrated into 
the estimation procedure of the semivariogram. Then, the data 
points in the block and other data points nearby are modelled, 
which leads to increased precision of the estimated elevation. 
Several points of the method proposed in this paper can be 
improved by further studies. For elevations at building edges, 
the change should not be continuous. So the problem about how 
to preserve edges in block Kriging needs further study. 
Moreover, the calculation time for block Kriging is quite long, 
for the dataset used in this study, it takes several minutes to do 
Kriging. A quick algorithm should be figured out for its use in 
large area.