input into a clustering algorithm for this application. Instead 
the value of the surface normal direction needs to be exploited 
on selected portions of the data set where there range of 
directions is limited and can be used to resolve ambiguities 
arising from planar regions in the data set. 
4. REFINED COMPUTATIONAL APPROACH. 
The minimum and maximum curvatures at a point have been 
found to be the only uniformity measures that are suitable for 
direct input into the selected clustering algorithm for this 
application. These values do not introduce ambiguities or 
uncertainties into the point grouping process, however 
considering these two measures in isolation will group points 
according to surface type alone, ie. planar, cylindrical, 
spherical and conical. Further processing is required if points 
on the same surface type but with different orientations are to 
be split, along with points on the same surface type with 
similar orientations but in different locations. 
Thus, the following processing strategy will be adopted: 
- Minimum and Maximum curvatures will be considered in 
isolation: grouping points according to surface type alone. 
A two dimensional version of Jolion’s algorithm will be 
employed. 
- Points on planes of different orientation will be split by 
considering surface normal directions for planar portions of 
the data set in isolation. This, limits the range of normals 
to be considered, and reduces the complexity of the 
ambiguities associated with this measure. 
- The proximity of points will be evaluated to split points 
lying on surfaces of the same type, but in distinctly 
different locations. 
- Edge points (ungrouped in a two dimensional clustering 
algorithm) will be assigned to multiple point groups based 
on the geometric fit of the point to the underlying surface 
of each point group. 
5. CONCLUSION. 
The difficulties encountered in algorithm development detailed 
in this paper are related to the application of the uniformity 
and proximity measures to the generalisation of target fields. 
The problems encountered do not indicate a failing of the 
uniformity and proximity measures in concept, only in 
application. The problems associated with the application of 
these measures will be addressed with the development of the 
refined computational approach, in which surface curvature 
and surface normal data are not considered simultaneously. 
6. BIBLIOGRAPHY. 
Abdelmalek N.N., 1990. Algebraic error analysis for surface 
curvatures and segmentation of 3-D range images. 
Pattern Recognition. Vol. 23, No. 8, pg 807 - 817. 
August 
Besl P.J. , Jain R.C., 1988. Segmentation through variable- 
order surface fitting. IEEE Trans. Pattern Anal. Machine 
Intell., Vol. PAMI - 10, No. 2, Pg. 167 - 192. March. 
Chen D.S., 1989. A data - driven intermediate level feature 
extraction algorithm. IEEE Trans. Patten Anal 
10 
Machine Intell., Vol. PAMI - 11, No. 7, Pg. 749 - 758. 
July. 
Denning P. J., 1986. Towards a Science of Expert Systems. 
IEEE Expert. Vol. 1, No. 2. Pg. 80 - 83. Summer 1986. 
Fan T-J., Medioni G., Nevatia R., 1987. Segmented 
descriptions of 3-D surfaces. IEEE Journal of robotics 
and automation. Vol 3, No. 6, Pg 527 - 538. 
December. 
Flynn P.J, Jain A.K., 1988. Surface classification: Hypothesis 
testing and parameter estimation. Proceedings IEEE 
Computer Society Conference. Computer vision and 
pattern recognition. Pg. 261 - 267. June. 
Fraser C.S., 1984. Network design considerations for non- 
topographic photogrammetry. Photogrammetric 
engineering and remote sensing. Vol. 50, No. 8, Pg. 1115 
- 1126. August. 
Fraser C.S., 1989. Non-topographic photogrammetry (2nd. 
Ed.). U.S.A. Pub. American Society for Photogrammetry 
and Remote Sensing. Chapter 8. - Optimisation of 
networks in non-topographic photogrammetry. Pg. 95 - 
106. 
Fraser. C.S., 1992. Photogrammetric measurement to one part 
in a million. Photogrammetric engineering and remote 
sensing. Vol. 58, No. 3, Pg. 305 - 310. March. 
Fraser C.S, Mallison J.A., 1992. Dimensional characterisation 
of a large aircraft structure by photogrammetry. 
Photogrammetric engineering and remote sensing. Vol. 
58, No. 5, Pg. 539 - 543. May. 
Grafarend E. W., 1974. Optimisation of geodetic networks. 
Bolletino di Geodesia e Science Affini.Vol. 33, No. 4. 
Pg. 351 - 406. 
Hoffman R., Jain A.K., 1987. Segmentation and classification 
of range images. IEEE Pattern Anal. Machine Intell. Vol. 
PAMI - 9, No. 5. pg. 608 - 620. September. 
Jolion J-M., Meer P., Bataouche S., 1991. Robust clustering 
with applications in computer vision. TEEE Trans. 
Pattern Anal. Machine Intell. Vol. PAMI - 13, No. 8, Pg. 
791 - 802. August. 
Krishnapuram R., Munshi A, 1991. Cluster-based 
segmentation of range images using differential- 
geometric features. Optical Engineering. Vol. 30, No. 10. 
Pg. 1468 - 1478. October. 
Mason S., 1994. Conceptual model of the sensor station 
placement task in configuring multi-station convergent 
networks. Presented paper - ISPRS Commission V 
Symposium. Melbourne, Pg. 256 - 264. 1 - 4 March. 
Mason S., Kepuska V., 1991. An  Al-based 
photogrammetric Network Design System. 
Proceedings, First Australian Photogrammetric 
Conference, Sydney, 7 - 9th November. Paper 
36. 
Newman T. S., Flynn P. J., Jain A. K., 1993. Model-based 
classification of quadric surfaces. CVGIP : Image 
understanding. Vol. 58, No. 2, pg 235 - 249, September. 
Quek F., Jain R., Weymouth T.E., 1993. An Abstraction- 
Based Approach To 3-D Pose Determination From 
Range Images. IEEE Trans. Pattern Anal. Machine Intell. 
Vol. PAMI - 15, No. 7, Pg. 722 - 736. July. 
Roth G., Levine M.D., 1993. Extracting Geometric Primitives. 
CVGIP : Image Understanding. Vol. 58. No. 1. Pg. 1 - 
22. July. 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996 
KEY WORD 
ABSTRACT 
LG.M.I surve 
1993 in Via 
obtained. The 
illumination, 
The Italian ! 
National Map 
still situated i 
The reasons 
cannot be con 
The ties wh 
cultural trac 
professional i 
"Institute" si 
vocations: the 
peoples by m 
technical and 
Therefore, it 
has witnessed 
city. 
For this reaso 
towards Flore 
on the night o 
Through sopl 
of the LG.M 
and passion 
documentatio 
models whi 
interactively 
that terrible a 
LG.M.I's te 
quality of “e 
barbaric acti 
but all the pe 
LG.M.I. surv: 
order to docu 
May, 27 199: 
The explosioi 
severely hit F 
building struc 
heavily dama 
heavily hit a 
Georgofili A: 
with the inr 
40,000 volui 
scientific vah