, CA, 9-11 Nov. 1999 
tion (Std. Dev.) on building 
the variation of the Std. Dey. 
industrial buildings at various 
; investigated. Figure 17 shows 
two samples of size 35 and 7 for 
18s respectively at various grid 
ifference between the computed 
and industrial buildings. The 
residential building decreases 
| resolution increases from 2m to 
he surface area of residential 
value of the mean Std. Dev. 
averaging as the grid resolution 
ll buildings, the computed mean 
n 2m to 1.1m as grid resolution 
almost constant beyond a 16m- 
due to the large surface area of 
s a more consistent mean Std. 
lutions. Therefore, if the size of 
nt factor in categorisation, 
ean Std. Dev. at various grid 
iable approach. 
  
10m 12m 14m 16m 18m 20m 
  
0.5 0.2 0 0 0 0 
  
2 1.5 1.4 1.4 1.1 12 
  
  
  
  
  
  
  
  
esidential and industrial building 
d resolutions. 
inct difference between the mean 
lustrial buildings at various grid 
Std. Dev. measure to identify 
estigated for randomly selected 
td. Dev. for residential buildings 
ym 2m to 10m. The Std. Dev. for 
and s) converges to zero at a grid 
| earlier, this is due to the effects 
due to the complexity of the roof 
gs, significant variation in Std. 
petween 2m and 10m. Figure 19 
or the selected industrial building 
ons of 2m to 10m. The computed 
stant in the industrial buildings 
0 10m. This is partly due to the 
uctures as well as the greater 
ings. The behaviour of the Sid. 
pears to reveal properties of the 
  
International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 3W14, La Jolla, CA, 9-11 Nov. 1999 
nature of the roof type and thus assist in discriminating between 
the two building types. 
Std. Dev. (m) 
Building id. 
  
  
Grid Resolution (m) 
Figure 18: Std. Dev. for selected residential buildings (p, q, r 
and s) using 2m to 10m grid resolution LIDAR DSM. 
Std. Dev. (m) 
2.5 
2 ar c Building id. 
15 e^t 
1 —e6—u 
—H— v 
0.5 kW 
  
0 T T T T M 
2 4 6 8 10 
Grid Resolution (m) 
Figure 19: Std. Dev. for selected industrial building (t, u, v and 
w) at 2m to 10m grid resolution LIDAR DSM. 
5 CONCLUSIONS 
Differentiating between residential and industrial building types 
using simple statistics such as RMSE and Std. Dev. is shown to be 
possible. The findings of this study may provide a basis for 
categorising residential and industrial building types in a more 
automated fashion. The main findings of this study are: 
* By examining the effect of RMSE on 3D models at various 
LIDAR DSM grid resolutions, the differences between roof 
structures of residential and industrial buildings can be 
inferred. 
* The complexity of the roof structures of the two building 
types can be examined using the Std. Dev. measure on 
individual buildings at various LIDAR DSM grid 
resolutions. 
® The use of mean height from LIDAR DSM to construct 3D 
models using the integrated methodology results in smaller 
RMSE compared to the use of the maximum height. 
*  Categorising the two building types using RMSE at various 
grid resolutions reveals the vertical roof heights and might 
be useful for certain applications. 
ACKNOWLEDGEMENTS 
The authors wish to thank the National Centre for Environmental 
Data and Surveillance, Environment Agency, Bath, England, who 
provided the LIDAR datasets. The Ordnance Survey of Great 
Britain kindly provided Land Line dataset. University 
Technology MARA, Malaysia, is supporting Mr. Jaafar’s research 
project. Research and computing facilities were made available 
by the School of Geography, The University of Nottingham. 
REFERENCES 
Gruen, A., Kubler, O., and Agouris, P., (eds), 1995. Automatic 
extraction of man-made objects from aerial and space images, 
Birkhauser Verlag, Basel. 
Gruen, A., Baltsavias, E. P., and Henricsson, O., (eds), 1997. 
Automatic extraction of man-made objects from aerial and space 
images (II), Birkhauser Verlag, Basel. 
Haala, N. and Brenner, C., 1999. Generation of virtual city 
models using laser altimetry, 2D GIS, unwrapping of detailed 
surface model. GIM International, Vol. 13, No. 3, pp. 6-8. 
Hug, C., 1996. Combined use of laser scanner geometry and 
reflectance data to identify surface objects. OEEPE-Workshop 
‘3D city models’, Bonn, Germany, 9 — 11 October 1996. 
Jaafar, J., Priestnall, G., Mather, P.M. and Vieira, C.A., 1999a. 
Construction of DEM from LIDAR DSM using morphological 
filtering, conventional statistical approaches and Artificial Neural 
Networks. RSS'99 - Earth Observation, From data to 
information, Proceedings of the 25™ Annual Conference and 
Exhibition of the Remote Sensing Society, Cardiff, 8-10 
September 1999, pp. 299-306. Remote Sensing Society, 
Nottingham. 
Jaafar, J., Priestnall, G. and Mather, P.M., 1999b. Assessing the 
effects of grid resolution in laserscaning datasets towards the 
creation of DSMs, DEMs and 3D models. Presented at the Fourth 
International Airborne Remote Sensing Conference and 
Exhibition/21* Canadian Symposium on Remote Sensing, 
Ottawa, Ontario, Canada, 21-24 June 1999, pp. 606-613. 
Lohr, U., 1998. Laserscanning for DEM generation, In: GIS 
technologies and their environmental applications, Brebbia C. A. 
and Pascolo P. (eds). Computational Mechanics Publications, 
Southampton, pp. 243-249. 
Wehr, A. and Lohr, U., 1999. Airborne laser scanning — an 
introduction and overview. ISPRS Journal of Photogrammetry 
and Remote Sensing, 54, pp. 68-82.