A future of photogrammetric research
Motivation
by prof, dr.-ing. W. Förstner, Institut für Photogrammetrie, Universitat Bonn.
SAMENVATTING
Een toekomst voor fotogrammetrisch onderzoek
De toepassing van computers voor de verwerking van beeldgegevens vereist een nieuwe definitie van de
fotogrammetrische taken. De geometrische modellering bij aero-triangulatie en de fysische modellering in
remote sensing moet worden ingebed in een semantische modellering van objecten die worden verkregen
uit foto- en satellietbeelden. Dit artikel houdt een pleidooi voor een nieuwe theoretische basis en stipt enke
le zaken aan bij de beeldinterpretatie die de kern vormt voor een wetenschappelijke evolutie in de foto-
grammetrie.
SUMMARY
The computer's access to the image data requires a new definition
of photogrammetric tasks. The geometric modelling in aerial tri-
angulation and the physical modelling in remote sensing has to be
embedded in a semantic modelling of the objects to be extracted
from aerial and satellite images. This article wants to stress the
urgent need for the development of a new theoretical basis and
sketch some research issues in image interpretation being the key
issue for a scientific evolution of photogrammetry.
Photogrammetry is perceived as mensuration from
photos. This holds both, for the internal perception, from
surveying and geodesy, and for the external perception,
from the various neighbouring disciplines. Photogram
metry and remote sensing cover two aspects of the same
technology, photogrammetry being responsible for the
geometric, remote sensing for the thematic part of infor
mation extraction. The inability to separate the used
techniques in detail leads to a joint framework in research
and in practice for using images of any kind for mensu
ration and mapping. Therefore the future perception of
photogrammetry will replace the photo by a photon
sensor, covering the classical analog photo, the scanned
image, the video image but also other imaging sensors
such as radar or laser range finders. The mensuration,
YQapeip - gramein, in photogrammetry is referring to, will
be replaced by information extraction, covering the clas
sical geometric information extraction in the form of
object location and object reconstruction, but also object
detection and the task of image interpretation. This gen
eralization of the inherent tasks of photogrammetry and
remote sensing is caused by the necessary technical
tools, especially from computer science, pattern recog
nition or artificial intelligence, making photogrammetry
and remote sensing a part of image understanding.
Now photogrammetry and remote sensing are mainly
application driven disciplines. Mapping still is the central
1) Further processes are the selection of the parts of the inter
pretation in a GIS, the generalization in order to be able to
adequately visualize the interpretation and the planning based on
the interpretation leading to an action changing the object, the
sensor (active vision, conform e.g. [3]) or other parts of the envi
ronment. These processes also need to be modeled and descri
bed, thus completing the loop in fig. 1We will not discuss these
aspects here.
task. In spite of the digital techniques pushed by the
availability of CCD-cameras, which make close range
photogrammetry a booming area, this is valid due to the
increasingly available image data, especially from
satellites.
The information sources to be dealt with, when making
maps, are manyfold: images, existing maps, data con
tained in Geographic Information Systems (GISs), object
models etc. Taking the fascinating revolution in computer
technology into account, the critical question therefore
arises: 'If we had infinite computer resources at no costs,
would we know how to make maps at no costs'? Cer
tainly: 'No'. The reason is the tack of a mapping theory
[79]. There are quite some indicators for this unfavour
able situation:
multi-spectral classification, being the only working
automatic interpretation technique, still shows a too
low accuracy;
none of the commercially available Digital Photo
grammetric Systems (DPS's) contains tools for semi
automatic or even automatic cartographic feature
extraction;
there is no commonly accepted definition of 'a map',
'the task of a DPS', or 'the functionality of a GIS'.
oil
o to gram me t r y
lysia
mote Sensing
selection
generalization
Fig. 1. The role of photogrammetry and remote sensing within
surveying and mapping.
This article wants to give a framework for describing the
questions to be answered when approaching the central
tasks of photogrammetry and remote sensing. These may
lead to a research program guiding the future photo-
372
NGT GEODESIA 93 - 8