The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
Figure 11. LIDAR DSM and 3D line features of the test area 
150 
100 
o 
100 
Figure 12. Aerial image of the test area 
The following cases demonstrate the advantages of the 
proposed scheme: 
Compensation: Figure 15(a) versus (b) and (c) versus (d) are 
the examples of compensating the fully missing and deficient 
3D line features out of LIDAR data by adding the image 
measurements utilizing the user interface. 
(c) (d) 
Figure 15. Compensating the missing and deficient parts 
Update: If there exists change between LIDAR and imagery 
(assumed to be the latest data) data sets, the result of building 
roof reconstruction from LIDAR data can be updated by fusing 
aerial images. Figure 16 (a) versus (b) depict the update effect. 
(a) (b) 
Figure 16. Updating the building roof model 
Enhancement of Reliability: Sometimes due to the inaccuracy 
of LIDAR data, the result reconstructed by fusing LIDAR data 
with aerial images may seem unsatisfied, though rendering high 
precision revealed from the adjustment report. A user interface 
would provide a visual inspection environment for determining 
the most proper employment of line features. 
Improvement of Theoretical Accuracy: According to the 
theory of error propagation, the precision of building roof 
comers will be improved by fusing different data sets as 
compared to the situation when only single data is available. 
Table 1 provides the precision of roof comers before and after 
the CSR approach. Notice that there are two comers obtained 
from LIDAR data and four comers from aerial images. In this 
case, the precisions of comers 1 and 2 are improved and comers 
3 and 4 are updated. 
200 
Figure 13. Initial building roofs from LIDAR system 
Figure 14. Refined roof model 
711