Page 88 - Dynamic Vision for Perception and Control of Motion

3 Subjects and Subject Classes
72

3.3.3 Knowledge Base for Perception Including Vision

At least as important for high-performance vision systems as the bottom-up capa-
bilities for sensor data acquisition and processing is the knowledge that can be
made readily available to the interpretation process for integration of information.
Deeper understanding of complex situations and the case decisions necessary for
proper reaction can be achieved only if relevant knowledge is available to a suffi-
cient degree. This will be discussed in later chapters to some extent because of its
importance, after the notions of subjects and situations have been fully introduced
in this and the next chapter. This broad topic is considered a major area of devel-
opment for intelligent systems with the sense of vision.

3.4 Behavioral Capabilities for Locomotion

The behavior most easily detectable by a vision system is motion of other objects
or subjects. Therefore, this will be treated here ahead of decision making, even
though decisions have to precede egomotion internally after signals from sensors
have been received.
Motion capabilities depend very much on the basic shape of the body and on the
means for locomotion of the subject. Legged motion widely spread in biological
systems is hardly found in technical systems. On the other hand, the “axle-and-
wheels” solution for locomotion abundant in technical systems cannot be found in
biological systems because nature has not been able to solve the maintenance prob-
lems of these devices with soft tissue and blood vessels. Also, special preparation
of the natural environment needed for using wheels efficiently could not be pro-
vided; humans solved this problem by road building, one of the outstanding
achievements of human civilization. Tracked vehicles for going cross-country also
use wheels, but have a special device for smoothing the surface these wheels roll
on (tracks with articulated chain members).
Birds are able to walk on two legs and to hop, and, in addition, most species
have the capability to fly by flapping their wings which they can fold to the body
and unfold for flying. On the contrary, human technology again uses the “axle-and-
wheels” solution with blades mounted to the wheel for generating propulsion (in
propellers and partly in jet engines) or lift (in helicopters). Both principles are di-
rectly reflected in the visual appearance of these subjects. In the realm of insects,
many more locomotion solutions can be found. Snakes solved their locomotion
problem by typical wave-like sliding motion. For locomotion in vertical structures
like trees, many-legged solutions may be of advantage.
The most highly developed creatures in biology with four limbs have developed
special skills with their backward “legs” for running on almost flat ground. As
soon as vertical structures have to be dealt with, the forward “arms” may support
locomotion by grasping and swinging. This multiple use of extremities in connec-
tion with the wide variety of image processing needed for this purpose (including
evaluation of data from inertial sensor of their body) may have led to the develop-