1323
THE SEMANTIC INFORMATION OF IMAGES ACQUIRED BY AERIAL DIGITAL
SENSORS IN CARTOGRAPHIC AND ENVIRONMENTAL APPLICATIONS
A.Lingua a *, F.Nex 3
a Dept. of Land, Environment and Georesource (DITAG), Politecnico di Torino, Corso Duca degli Abruzzi 24, Torino,
Italy - (andrea.lingua, francesco.nex)@polito.it
Commission IV, WG VI/9
KEY WORDS: Digital, Cartography, Mapping, Analysis, Information , Accuracy
ABSTRACT:
Aerial digital cameras were first presented on the occasion of the ISPRS International Symposium in 2000. Since then, several
papers that analyse the geometrical accuracy of digital camera have been presented. These works have underlined the fact that the
geometric accuracy obtainable with the digital sensor is significantly higher than the accuracy achieved with scanned analogue
cameras. Nevertheless, this accuracy is usually obtained considering predefined markers which allow higher geometric precision
than the other points required in the Technical Specifications at a certain map scale, whereas precision (during tests) in map
productions is focused on common map entities required in Technical Specifications. Furthermore, good geometric precision in
triangulation does not guarantee easy stereoplotting of all the map entities with the required precision. However, it is obviously
wrong (and simplistic) to consider, without any logical proof, that digital camera products are comparable, from a semantic point of
view, with traditional photogrammetric camera products acquired approximately at the same nominal scale.
The semantic information of digital images is accurately analyzed in this paper, and the geometrical aspect is neglected. In particular,
the semantic information is considered both from a qualitative and quantitative point of view and the image quality and information
content of several digital sensors (ADS40 Leica Geosystems 1st and 2nd Generation, DMC Intergraph Z/I, UltraCamD Vexcel Corp.,
3-Das-1 Wehrli & Associates), which are commonly employed in the map production at different scales, are evaluated. In this
analysis, particular attention is paid to the handiness of the interpretation of the entities that are requested in modem technical
specifications (at several scales) for map production.
1. INTRODUCTION
Aerial digital cameras were first presented on the occasion of
the ISPRS International Symposium in 2000. Since then, these
cameras have become ever more popular and will eventually
replace analogue film cameras. Traditional cameras are destined
to drop out of the market in a few years, as some of largest film
manufacturers in the world have discontinued production and
the majority of camera producers have decided to only develop
digital sensors (Casella, 2007).
During these years several papers that analyse digital camera
geometric accuracy have been presented. These works have
underlined the fact that the geometric accuracy obtainable with
the digital sensor is significantly higher than the accuracy that
can be achieved with analogue cameras. Nevertheless, this
precision is usually obtained considering predefined markers
which allow geometric precision from 3 to 5 times higher than
the other points (Kraus, 1997) required in the Technical
Specifications at a certain map scale, whereas precision (during
tests) in map productions is focused on the common entities that
are required by the Technical Specifications. Furthermore, good
geometric precision in triangulation does not guarantee easy
stereoplotting of all the map entities with the required
precision.
However, it is obviously wrong (and simplistic) to consider,
without any logical proof, that digital camera products are
comparable, from a semantic point of view, with traditional
photogrammetric camera products acquired approximately at
the same nominal scale.
As a consequence, a comparison of the semantic content of
(scanned) analogue and digital images must be performed. In
this analysis, particular attention must be given to the ease of
interpretation of map entities which are required in modem
technical specifications (at different scales) for map production.
Some papers have already described these differences,
analysing the geometric accuracy and the noise effects through
testing and measurements (Becker et al. 2006; Kolbl, 2005;
Casella et al., 2004; Cramer, 2004; Leberl, et al., 2003; Reulke,
2003). Other papers have detected in the Ground Sample
Distance (GSD), the fundamental parameter in the flight
specifications for digital cameras (Casella, 2006). One
[Jacobsen, 2007] has already compared images acquired by two
frame cameras (DMC and UltraCamD) with scanned analogue
images. A multiplicative factor of 1.5 between the GSD of
digital images and scanned analogue photos was proposed in
this work, in order to obtain the same object detail; this factor,
however, only referred to one particular map scale and there
was no clear reference to technical specifications.
In this present paper, this comparison between scanned
analogue and digital cameras is performed in a more systematic
way; these differences are in particular analysed only from a
semantic point of view, disregarding the geometrical aspect and
a comparison between different cameras. This evaluation is
performed considering the image quality and information
content of several digital sensors which are produced by leaders
in this field such as Intergraph Z/I (DMC), Leica Geosystems
(ADS40 1 st and 2 nd Generation), Vexcel Corp. (UltraCamD) and
Wehrli & Associates (3-DAS-1), which represent over 90% of
the world’s photogrammetric digital sensors. In addition, as is