International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
applicable at any scale:
e allowing creation of regular grid-based digital elevation
models (DEMs) of consistent precision and at very high
sampling rates, thus able to record very detailed
morphology:
e commercial software available at competitive rates;
e software runs on relatively cheap UNIX workstations and
PCs;
e user-friendly interface of software makes techniques
available to non-specialists.
2.1 Theory
Theoretical principles underlying photogrammetric methods
were established over a century ago (Finsterwalder, 1897) and
developed for map production in the 1930s (Wolf & Dewitt,
2000; Slama, 1980). The concept of collinearity is critical,
whereby a point on the object, centre of lens and resultant
image point lie on a single line in three-dimensional space.
Based on this principle the three-dimensional object space
coordinates can be extracted from a stereopair of photographs,
provided that interior and exterior orientation of the camera at
the moment of exposure are known. Exterior orientation
parameters of the camera can easily be determined with the help
of a minimum of three ground control points (GCPs) for each
image (Wolf & Dewitt, 2000).
When using archival photographs or a non-metric camera, the
interior parameters of the camera are often unknown. However,
these can be determined by using a self-calibrating bundle
adjustment. In this numerical procedure the interior and exterior
parameters of all frames are simultaneously estimated by using a
least squares estimation, which minimizes and distributes the
errors (Kenefick et al, 1972; Granshaw, 1980; Chandler &
Cooper, 1989). Redundant photo control is recommended, as
this will provide stochastical properties of the solution and
checks on data consistency.
2.2 ADAPT - A Digital Archival Photogrammetric
Technique
Chandler & Brunsden (1995) describe an analytical technique
for extracting spatial data from historical photographs and prove
its value in studies of land-form change. Their procedure is
known as the archival photogrammetric technique and
overcomes the main problems associated with archival
photographs, the lack of camera calibration and control points,
by using a self-calibrating bundle adjustment. The great
advantage of the method is that it can be applied universally to
all photographic formats: both vertical and oblique images,
from small to large-scale. Additionally, all image epochs can be
referenced to the same coordinate system, which is crucial for
quantitative analyses. The approach described in this paper was
developed from this method, but using the latest technologies in
digital processing, which enables the method to be used more
easily, and hence become more universally applicable.
First step is the acquisition of aerial photographs, using the
following considerations (in order of importance):
ground coverage; the area of interest should be completely
covered by the stereoscopic overlap area of the images.
* scale; the scale of the photograph determines with what
precision photo-coordinates can be measured and what
feature sizes can be discerned.
e geometry; the parallax and hence heighting precision is
affected by flying height, airbase, and focal length of the
camera.
e. format; best results are obtained when using high-
resolution scans (15-20u) of contact diapositives from the
original negatives, using a photogrammetric quality
scanner. However, as these are not always available, use of
scans from contact prints may have to be considered.
Obtaining historical aerial photographs is time-consuming. In
the UK the imagery is distributed over numerous archives and
libraries held by a range of institutions, among them the
National Monuments Record, the collection of Cambridge
University, commercial mapping companies and various local
authorities. Some of these organisations have standardized their
search and request systems, which makes the archives easily
accessible to public. However, in some cases this
standardization makes it more difficult to deal with specialist
demands, for example high-resolution photogrammetric scans.
Sometimes only contact prints are available, or access to a
photogrammetric scanner is lacking.
Once imagery is acquired, ground control points must be
identified and measured. Suitable control points are well.
defined natural features, clearly identifiable on the photographs.
For photoscales of 1/4,000-1/50,000 the use of differential GPS
is recommended (Chandler, 1999). Geodetic receivers can
achieve a precision of +10mm +1ppm in plan over distances up
to 15km; precision in height is 2-3 times more. A source of less-
accurate coordinate data are large scale topographic maps,
containing planimetric positions, spot heights and bench marks.
A minimum of two planimetric and three height points is
needed to define a datum, but more control points are desirable
as redundancy provides appropriate checks. The control points
should be evenly distributed over the images to gain a strong
geometry. Ideal positions are points tying frames together and
surrounding the volume of interest (Wolf & Dewitt, 2000).
During photogrammetric processing the relationship between
photo and ground coordinates is established and the interior and
exterior orientation of the camera determined. The processing
involves a self-calibrating bundle adjustment, which
implements rigorously the collinearity equations in a least-
squares estimation procedure (Chandler & Brunsden, 1995). In
order to transform the image coordinates to photo coordinates,
it is necessary to measure the image positions of the fiducial
marks (if present) and the control points.
Finally, it is possible to extract coordinates of points anywhere
on the images, by using the estimated interior and exterior
parameters. A variety of products may be derived automatically
including DEMs, orthophotographs and parameters indicating
the accuracy of these products. The following products can be
used to visualize the landslide movements:
e DEMs-of-difference (elevation differences
epochs);
* cross-section profiles;
* displacement vectors;
* animations.
between
2.3 Quality assessment
The quality of the photogrammetric solution is controlled by
measurement errors associated with elements used in the
functional model to relate image to object. Data quality can be
described by three terms with respect to three types of error
476
Internat
(Cooper
reliabili
using Sc
precisio
e 3e
28 cal
a Im
fide
ae blu
The ap
capable
This prc
Data re
redunda
evaluate
coordin:
rigorous
control
deforma
(Chandl
(e.g. ler
often |
paramet
by u
photogr
values (
3.1 St
Mam T
Castleto
SK1358
is domii
former
construc
consequ
mass. T
(Mam
Shales).
the heac
affected
the slid
drops fi
12°. Ma
al., 198‘
The ini
(Skempl
