International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
to 20cm RMS), whilst significantly reducing capture 
cost (3096 reduction over test areas), 
e A perturbation on z-values slightly degrades quality: a 
perturbation with a 2m standard deviation introduces a 
loss of 15 to 45cm in RMS, 
e A perturbation on xy-values can introduce local gross 
errors, 
e The contour simplification implies a 5 to 10cm loss in 
accuracy whilst reducing capture costs from 15 to 
20% on examples. 
  
  
  
  
  
  
  
  
  
Cost Semi-automatic DSM accuracy 
Input | NPts | NPts | Avg | Std | RMS | Emax | Pet Pct 
Poly Used | Ref +-1 | +/-2 
REF | 6131 | 690 | 0.53 | 0.85 | 1.00 | 4.47 | 75.94 |94.64 
Zmed | 6066 | 690 | 0.55 | 0.81 | 0.98 | 4.13 | 76.23 |95.07 
Blocks | 4460 | 690 | 0.42 | 0.95 | 1.04 | 6.44 |80.14 |95.65 
Dzl | 6327 | 688 | 0.40 | 0.89 | 0.98 | 5.67 | 75.73 |05.49 
Dz2 | 6959 | 690 | 0.40 | 1.09 | 1.16 | 5.00 | 71.16 (91.30 
Dz3 {735616907033 1.51 | 1.55) 7.40 | 64.78 (85.94 
Dxyl |5146| 688 | 0.30 | 1.25 | 1.29 |15.47 | 75.00 | 92.59 
Spl3 | 4991; 688 | 0.58 | 0.94 | 1.10 | 4.85 | 73.26 |91.86 
Blocks 
+spl3 | 3464 | 692 | 0.26 | 1.15 | 1.18 | 6.54 | 78.03 92.49 
  
  
  
  
  
  
  
  
  
  
  
Table 5: Semi-automatic DSM cost and accuracy, when input 
planimetry contours vary (Kerlaz) 
4.3.3 DSM production: conclusion 
The semi-automatic process provides DSM close to ground 
truth with an accuracy around Im. It is reasonably robust to 
planimetric and altimetric variations. Capture costs can be 
significantly reduced. For example, using building blocks and 
simplified polygons can reduce capture costs by 40%, whilst 
providing a DSM with a RMS below 1.50m and more than 
75% reliable points. 
5. CONCLUSION 
Two methods for producing urban DTM and DSM from acrial 
images have been compared: 
e A traditional approach based on the manual capture of 
3D vectors, 
e A semi-automatic approach using automatic image 
matching. 
Regarding DTM production, the manual approach provides a 
very good accuracy (better than 85cm) but at an important cost. 
The semi-automatic approach is a good alternative solution for 
reducing capture cost, whilst keeping a Im accuracy. 
As for building description, the data produced with the manual 
approach are good quality but limited by the single elevation 
roof model. The semi-automatic approach provides a 
representation closer to ground truth, including oblique roofs 
and superstructures, even when capture cost is significantly 
reduced. The roof accuracy then lies between 1m and 1.50m. 
These two approaches are complementary and correspond to 
different needs. The manual approach is relevant when a full 
vector description and very accurate data are necessary. The 
semi-automatic approach is more appropriate when cost 
reduction is a priority, or when roof shape is required by the 
user. 
This study finally shows it is possible to compute high quality 
DTM and DSM from 2 aerial images, as long as appropriate 
external data are available. The accuracy could be improved 
using multiple views, or pre-defined roof models. To fulfil this 
study, complementary tests should focus on the planimetric 
accuracy of the produced data and on the specific behaviour of 
vegetation. 
REFERENCES 
Baillard, C., 2003. Production of urban DSMs combining 3D 
vector data and stereo aerial imagery. In Photogrammetric 
Image Analysis, IAPRS Vol.34, Munich, pp. 19-24, sep 2003. 
Brenner, C., 2001. City models — Automation in research and 
practice. In F. Spiller (ed.), Photogrammetric Week 01, pp. 149- 
158. 
Cord, M., Jordan, M., Cocquerez, J.-P., Paparoditis, N., 1999, 
Automatic extraction and modelling of urban buildings from 
high resolution aerial images. In Automatic Extraction of GIS 
Objects from Digital Imagery, IAPRS Vol. 32, pp 187-192. 
Flamanc, D., Maillet, G., Jibrini, H., 2003. 3D city models : an 
operational approach using aerial images and cadastral maps. In 
Photogrammetric Image Analysis, IAPRS Vol.34, Munich, pp. 
53-58, sep 2003. 
Fraser, C.S., Baltsavias, E., Gruen, A., 2002. Processing of 
Ikonos imagery for submetre 3D positioning and building 
extraction. /SPRS j. of Photogrammetry & Remote Sensing, 
56(3), pp. 177-194. 
Jibrini, H., Pierrot Deseilligny, M., Paparoditis, N., Maitre, H., 
2000. Automatic building reconstruction from very high 
resolution aerial stereopairs using cadastral ground plans. In 
xp? Congress ISPRS, IAPRS Vol.33, Amsterdam (NL), July 
2000. 
Jung, F., Paparoditis, N., 2003. Extracting 3D free-form surface 
boundaries of man-made objects from multiple calibrated 
images: a robust, accurate and high resolving power edgel 
matching and chaining approach. In Photogrammetric Image 
Analysis, IAPRS Vol.34, Munich, pp. 39-44, sep 2003. 
Maas, H.G., 2001. The suitability of airborne laser scanner data 
for automatic 3D object reconstruction, 2001. In Automatic 
Extraction of Man-Made Objects from Aerial and Space images 
(III), pp. 291-296. 
Paparoditis, N., Maillet, G., 2001. Improving aerial image 
matching techniques in urban areas using a new true multi- 
image approach guided from object space. In Proc. 3rd 
International Workshop on mobile mapping technology, Cairo, 
January 2001. 
Roux, M., Maitre, H., 2001. Some more steps towards 3D 
reconstruction of urban areas from multiple views. In Automatic 
Extraction of Man-Made Objects from Aerial and Space images 
(IN), pp. 135-147. 
Vosselman, G., Suveg, I, 2001. Map based building 
reconstruction from laser data and images. In Automatic 
Extraction of Man-Made Objects from Aerial and Space images 
(IH), pp. 231-239. 
Zinger, S., Nikolova, M., Roux, M. Maitre, H., 2002. 3D 
resampling for airborne Laser data of urban areas. In JAPRS 
Vol.34, n° 3A, pp. 418-423. 
     
  
  
International Archiv 
ET GMT ren 
  
Fig