The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008
748
and installation. The physical structure of stairways can range
from typical straight-run examples, to quite complex, curved,
multi-level structures that can even double or triple back on
themselves. All stairways present a geometrical challenge for
photogrammetry, due to the narrow fields of view and obvious
physical constraints that are placed on possible imaging
locations.
Traditional automated photogrammetric techniques using white
retro targets and highly controlled image exposure are generally
not suitable for stairway measurement. Firstly, it is near
impossible to place targets exactly on the points and features of
interest, such as straight line edges of stair intersections; risers,
which are the vertical portions of each step; and winders, which
are steps narrower on one side to facilitate a change of direction
in a stairway. Hybrid measurement helps by allowing automatic
network orientation via the use of coded targets, and then the
subsequent manual and semi-automatic measurement of
important stair features in the normally exposed imagery. The
stairway will still be clearly visible in each photograph,
allowing the user to see and therefore measure the necessary
elements. This is illustrated in Figure 3. In the left image, the
photogrammetric network is shown, with approximately 30
images and 40 coded targets, measured fully automatically. The
right image shows examples of the manual measurement that
has taken place, including points at the intersection of each step
and the walls either side, and also some semi-automatic
measurement, including scale bars, lines to represent each step,
and ‘stair-plates’ which are linked to a user-defined database for
ease of implementation.
In a hybrid measurement context, verifiable accuracies of about
1:50,000 have been achieved for the automatic component of
stairway measurement, using a 12 megapixel Nikon D200 with
fixed-focus 18mm lens. Achievable accuracy levels for
manually measured untargeted features are roughly an order of
magnitude lower, yet certainly fall within the required tolerance
for stairlift design and installation.
Figure 3. Hybrid measurement of a stairway for lift installation.
5.2 Vehicle Crush Measurement
Another application particularly suited to hybrid measurement
is the accurate reconstruction of traffic accidents, and especially
the measurement of damaged vehicles. Normally exposed
images are required so that the operator can manually digitize
unsignalized features, which typically constitute the primary
features of interest, such as skid marks; debris on the road;
surrounding buildings; traffic signs; road markings, and so on.
Operator-assisted surface extraction may also assist in vehicle
‘crush’ analysis. The number of image stations illustrated in
Figures 4a highlights the need for fully automatic network
orientation, while the abundance of natural features to be
measured, including the curves shown in Figure 4b, illustrates
the need for well exposed imagery. This is not generally
conducive to automatic image scanning and multi-image
orientation as it places an extra burden on these processes in
terms of complexity and required computation time.
->u ■ ■ mb a > b
(a)
Figure 4. Vehicle crush measurement, illustrating: (a) automatic
orientation and (b) manual feature dimensioning.
6. CONCLUDING REMARKS
There are a growing number of photogrammetric measurement
tasks which can substantially benefit from fully automatic
network orientation and 3D target point determination, coupled
with the ability for subsequent semi-automatic and manual
feature extraction from normally exposed imagery. The authors
have dubbed this all-encompassing approach ‘hybrid
measurement’. Potential users include law enforcement
agencies performing traffic accident reconstruction and forensic
or crime scene measurement, architects interested in modelling
buildings for which no CAD models exist, and ship building
companies.
The practical developments of hybrid measurement reported in
this paper, namely the adoption of red retro-reflective targeting;
the new system of coded-targets, fully automatic network
orientation and image point correspondence determination, as