2km. The data quality requirement was a better than 
1m absolute accuracy and better than 0.5m relative 
local accuracy (the key issue is the relative position 
of the antenna). Aerial photos of the two areas were 
scanned on a Leica scanner, and the entire 
orientation and measurement process was carried 
out on a Leica/Helava DPW 770 softcopy 
workstation. 
2. EXPERIENCES 
Two sites were selected for this project: the 
southern tip of Manhattan in New York City, and 
Silver Spring, Maryland. Both sites measured 
approximately one square mile. The first site is 
typical of a very dense urban area with high-rise 
buildings and tightly packed city blocks. The second 
site included downtown and suburban areas, with a 
less dense building distribution. Panchromatic 
imagery was obtained from an aerial camera with a 
nominal focal length (150mm) and scale of 1:24,000 
for Manhattan, and 1:14,400 for Silver Spring, 
furnished by Air Photographics, Inc., West-V irginia. 
Preparations 
The interior orientation for the digital images was 
performed during the scanning process. The exterior 
orientation. of three consecutive images on the 
softcopy workstation was accomplished by visual 
3-D observation of ground control points, both 
horizontal and vertical. The built-in bundle block 
adjustment of the DPW 770 system delivered the 
final orientation data. Some check points were used. 
The RMS error for control points was in the 10cm 
range, while check points produced a 60-70cm RMS 
error. Since this orientation process is different from 
the usual two-step procedure which includes 
separate relative and absolute orientations, the 
models were set up on an analytical plotter for 
comparison. Due to the different model systems, 
direct comparison of the exterior orientation 
parameters was not possible; thus, check points 
were used to relate the quality of the orientations. 
Comparison of the readings on the Zeiss Pl 
analytical plotter with the softcopy data showed a 
reassuring correspondence (Schenk and Toth, 1989). 
Figure 1 shows the footprint and control point 
distribution for the block of the Manhattan project. 
Prior to data extraction, epipolar resampling of 
imagery was done to avoid visual distortion in stereo 
viewing and to facilitate the use of the automated 
DTM extraction. The DPW 770 softcopy system 
    
864 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
offers two sets of tools for automatically creating 
and editing DTMs and building features, 
respectively. Thus, the data acquisition process 
consists of two basic phases: DTM extraction and 
feature measurements. Starting the sequence with 
the DTM extraction and continuing with the features 
or vice versa is a user option. In our investigations, 
both approaches were tested. 
  
Figure 1. Footprint of the Manhattan site 
In the first data extraction phase for the Manhattan 
site, building features were digitized from digital 
imagery by using DPW 770 "feature extraction" 
tools. These tools allowed us to follow a rooftop 
outline or building block in a number of operating 
modes, including: automatic rooftop, square, static, 
spline, and mixed (Socet Set, 1995). All of these 
digitization modes created the actual building 
volume; that is, the sides of a building were created 
automatically. This became possible by digitizing 
the elevation at ground level next to the building 
footprint. Once the building extraction was 
complete, the automated DTM extraction and 
editing were performed. 
The same set of DTM and feature extraction tools 
was used for the Silver Spring site. The major 
difference, however, was in the building volume 
creation. Since DTM extraction had been performed 
before the features were extracted, the elevation of 
the nearest grid point from the DTM was used for 
building volume computation, and operators’ input 
of elevation near the building footprint was not 
necessary. This sped up the digitization process but 
may have compromised the vertical accuracy. 
Manhattan project 
The project site targeted a square block on the east 
side of the southern tip of Manhattan. Considering 
the required large number of elevation posts, it was 
necessary to divide the entire area into eleven