An Artificial Perceptual Model Approach - Artificial Human Sensors: Science and Applications - page 109

In Bajcsy (1985), two methods are described concerning the concept of an
active perception system, the top-down and bottom-up principle. In the top-down
methodology, the environment is totally unknown to the system and the system
has to use accessible sensor information to perceive the surroundings and comple-
ment with existing knowledge. The system will then literally “open its eyes and
report what it sees”. On the other hand, in the bottom-up principle, the system
will search or follow a certain given and pre-set goal in the environment.
The active perception process is mainly used for focussing the attention on spe-
cific events and field of view. In the literature, active artificial perception processes
can be found, e.g., in Biel (2002a), in specific vision system, Biel (2000), odour
systems Lindquist (2004), tactile systems, Robertsson (2006), and combinations
thereof, Bergstrom (1998).
The other aspect of multi-sensor fusion is in its non-selective approach similar
in data response and provides a co-operation of measurement in the same field
of measurements, that is the sensors work co-operatively by observing the same
properties of sensing interest, Lundh (2009), as illustrated in Fig. 5.3.
The proposed sensor fusion concept in Fig. 5.4 can be considered as a design
structure in three dimensions, aiming to achieve different strategies in the infor-
mation processing. The fusion process may of course vary between fusion ap-
plications and expected performance. The proposed organisation structure below
indicates three general fusion levels in a hierarchical design. The data fusion, fea-
ture fusion and information processes can be executed at a low, medium and upper
level of the hierarchical design
Figure 5.3. A non-selective odour identification system, illustrating that many sensors'
output parameters are overlapping, exhibiting highly redundant results.