×

You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Title
Close-range imaging, long-range vision

3cavi

presented at the

A SOLUTION FOR IMAGE ORIENTATION OF CLOSE RANGE BLOCKS
M. Scaioni, G. Forlani
Università degli Studi di Parma, Dipartimento di Ingegneria Civile, viale Parco Area delle Scienze, 43100 Parma, Italy
(marco.scaioni, gianfranco.forlani)@unipr.it
Commission V
KEY WORDS: Cultural Heritage, Algorithms, Automation, Direct.
ABSTRACT:
Image attitude in architectural photogrammetry blocks may be very far from the normal and standard configurations. Close range
triangulation is therefore a non trivial task to be performed also by a skilled operator. This paper describes the solution developed at
the University of Parma to deal with the orientation of close range blocks without requiring approximate values for the bundle
adjustment parameters. The system implements algorithms for space resection (which needs at least 4 GCP per image), space
intersection and bundle block adjustment, integrated in a management module which oversees the sequence of computations and the
whole orientation procedure. In short, we provide approximate values for the exterior orientation of every image by space resection
and use a closed formula to compute the approximate ground coordinates of tie points. To provide the resection module with at least
4 object points for every image, we apply an incremental process where, starting from a minimal set of oriented images, we add new
images and compute preliminary ground coordinates of tie point as soon as this is feasible. Simulated and real blocks oriented with
the procedure, which seems to be considerably robust, will be presented discussing the performance of the algorithms.
INTRODUCTION
Image orientation is a fundamental step in the workflow of any
photogrammetric task: it takes time and skill to be completed
and its accuracy and reliability will affect the remaining phases.
Efforts to try to make it faster and accurate went along two
different paths:
- direct image orientation: systems like the photo-theodolite
in close range or the solar periscope in aerial
photogrammetry were applied to directly measure the
exterior orientation (EO) parameters;
- indirect image orientation techniques based on tie and
ground control points, capable of reducing as much as
possible the amount of topographic work while still
controlling block deformations were developed.
Over the years, technological developments in computing
capabilities seemed to give the prevalence to indirect methods,
with Aerial Triangulation becoming more and more
sophisticated with the introduction of additional parameters.
Digital photogrammetry gave further momentum to this
approach, by making tie point selection and transfer automatic
to a large extent with Automatic Aerial Triangulation.
In aerial photogrammetry the lead may now be possibly
returning to direct sensor orientation methods, thanks to
IMU/GPS integrated systems. While such systems promise to
get rid of any ground control, apart from calibration and check
purposes, there is nothing like in sight in close range
applications so indirect methods are not challenged.
Roughly speaking, we may say that here two different scenarios
apply. In controlled environments and in industrial applications
strong network geometry and the use of (possibly coded) retro-
reflective targets is commonplace and synonymous of high
accuracy, largely automated measurement techniques. Image
orientation is achieved by a bundle adjustment and approximate
values are obtained mostly alternating between resection and
intersection. In architectural applications of photogrammetry,
possibly within extensive projects such as building high
resolution 3D city models or surveying large historical or
monumental sites, surveying costs must be kept to a minimum
to be viable. Here targeting may often be impractical or too
costly to install and remove. The extent and complexity of the
environment, that can be very high, often demands for still
significant topographic work and for relatively large blocks.
Besides, image attitude may be very far from the normal case
while standard configurations for tie points (i.e. *von Gruber"
positions) and optimal ground control points distribution are
difficult to achieve: this makes network design complex. Initial
approximations for the orientation parameters, though mostly
not required, may be difficult to guess; besides, outliers in the
data set, whose number may be high when using automatic
procedure for image point measurement, may affect the
estimates for those initial values. Close-range triangulation is
therefore usually a complex task that has to be performed by a
skilled operator, so restricting the community of people which
can make use of photogrammetric techniques. Despite this,
photogrammetry in architectural applications is in our country
increasingly the domain of professionals with little or not proper
background in photogrammetry. In an attempt to reduce costs,
specialists in monument restoration also execute the survey,
opting for a “do it yourself" approach. Very often many ground
control points are measured, so as to register each stereo-model
individually, thanks also to the wide diffusion of total stations
capable to measure distances without reflector. In such cases,
space resection may be used to compute the approximate
orientation of each image, followed by a bundle adjustment of
the stereo-pair to refine the parameter values, rather than
proceeding with the relative orientation and later with the
absolute orientation.
Inexpensive monoscopic plotting programs or more expensive
ones with stereo capabilitites are being used. Despite online
help and good tutorials, image orientation in large blocks is still
often a problem and may get stuck trying to get a correct
solution. In most cases this is the results of poor block design or
poor block geometry and self-diagnosis may help to overcome
the problem; sometimes this has to do with the capabilities of
the algorithm used.
In the last years several methods have been proposed in order to
make easier the computation of the orientation of close-range
blocks and nowadays several software packages are available.
Now also many low-cost commercial packages feature
—453-