6A-4-5 
Figure 6. Topographic surface of Hagerstown, MD test 
area. 
Figure 7. Object features of the Hagerstown, MD test 
area, NE quadrant. 
Even by a simple visual comparison, a striking difference 
between the stereo image and LIDAR created DEMs can 
be observed. As expected, the image-based extraction is 
unable to effectively cope with buildings, and it tends to 
create a smoothed out, draped surface over the natural and 
man-made objects. In contrast, the direct and independent 
observations from LIDAR give a rather good sampling 
and thus an excellent representation of the surface— 
independently from the underlying object contents. A 
rigorous comparison of the surfaces is a rather difficult 
task, and no attempt was made to analytically evaluate the 
differences among the different surface data sets. 
To see the local characteristics of the LIDAR and stereo 
image-created surface points, an image patch with 
elevation overlaid laser point locations is shown in Figure 
8. The size of the cross-mark around the LIDAR elevation 
spots shows the horizontal positional uncertainty. The 
ground control point shown in Figure 2 is marked by a 
square in Figure 8. The surface extraction-created points 
show a closer match to real object points than the rather 
loose connection of the LIDAR points. Obviously, both 
are very different from operator-measured locations. 
- +4-4 
Figure 8. LIDAR surface elevation point locations. 
To better illustrate the vertical behavior of the spot 
elevations, a vertical profile is shown in Figure 9. 
Figure 9. Elevation profile of LIDAR and stereo image- 
created surface elevation spots. 
The selected profile is the west to east LIDAR scan line 
closest to the marked control point. Circles mark the 
topographical surface, while LIDAR spots and stereo 
image-derived elevations are marked by the x and + 
symbols, respectively. The LIDAR data show an excellent 
match for the flat areas. The first peak of the LIDAR 
profile from the left represents a building, while the 
smaller peak in the center is most likely a car. The stereo 
image-created spots exhibit the typical smoothed out 
pattern; surface discontinuities are smeared. The modest 
performance of this technique is mainly due to the coarse 
image resolution, about 25 cm GSD. 
The measurable footprint of the LIDAR spots results in an 
average elevation value, which is very noticeable around 
discontinuities. Of course, breaklines and the like are a 
problem for the stereo image-extracted surface too, but for 
the non-complex image scenes, their point localization 
usually shows a much better performance.