Page 34 - Dynamic Vision for Perception and Control of Motion

18 1 Introduction

tencies were accepted, in general, if satisfying explanations could be found. Pro-
gress toward more consistent overall models of “the world” was slow and took mil-
lennia for humankind.
The natural sciences as a specific endeavor of individuals in different cultural
communities looking for a consistent description of “the world” and trying to avoid
biases imposed by their specific cultures have come up with a set of “world mod-
els”, which yield very good predictions. Especially over the last three centuries af-
ter the discovery of differential calculus by Leibnitz and Newton and most promi-
nently over the last five decades after electronic computers became available for
solving the resulting sets of equations in their most general form, these prediction
capabilities soared.
In front of this background, it seems reasonable to equip complex technical sys-
tems with a similarly advanced sensor suite as humans have, with an interpretation
background on the latest state of development in the natural sciences and in engi-
neering. It should encompass a (for all practical purposes) correct description of
the phenomena directly observable with its sensor systems. This includes the light-
ing conditions through sun and moon, the weather conditions as encountered over
time and over different locations on the globe, and basic physical effects dominat-
ing locomotion such as Earth–gravity, dry and fluid friction, as well as sources for
power and information. With respect to the latter ones, technical systems do have
the advantage of being able to directly measure their position on the globe through
the “Global Positioning System” (GPS). This is a late achievement of human tech-
nology only less than two decades of age, which is based on a collection of human-
made Earth satellites revolving in properly selected orbits.
With this information and with digital maps of the continents, technical
autonomous systems will have global navigation capabilities far exceeding those of
biological systems. Adding all-weather capable imaging sensors in the millimeter
wave range will make these systems truly global with respect to space and time in
the future.

1.8 Structuring of Material Covered

Chapters 1 to 4 give a general introduction to dynamic vision and provide the basic
knowledge representation schemes underlying the approach developed. Active sub-
jects with capabilities for perception and control of behaviors are at the core of this
unconventional approach.
Chapter 2 will deal with methods for describing models of objects and processes
in the real world. Homogeneous coordinates as the basic tool for representing 3-D
space and perspective mapping will be discussed first. Perspective mapping and its
inversion are discussed next. Then, spatiotemporal embedding for circumnaviga-
tion of the inversion problems is treated. Dynamic models and integration of in-
formation over time are discussed as a general tool for representing the evolution
of processes observed. A distinction between objects and subjects is made for
forming (super-) classes. The former (treated in Chapter 2) are stationary, or obey
relatively simple motion laws, in general. Subjects (treated in Chapter 3) have the
capability of sensing information about the environment and of initiating motion