| The goal of computer vision research is to provide computers with humanlike
perception capabilities so that they can sense the environment, understand
the sensed data, take appropriate actions, and learn from this experience in
order to enhance future performance. The field has evolved from the application
of classical pattern recognition and image processing methods to advanced
techniques in image understanding like model-based and knowledge-based vision.
In recent years, there has been an increased demand for computer vision systems
to address “real-world” problems. However, much of our current models
and methodologies do not seem to scale out of limited “toy” domains. Therefore,
the current state-of-the-art in computer vision needs significant advancements
to deal with real-world applications, such as navigation, target recognition,
manufacturing, photo interpretation, remote sensing, etc. It is widely understood
that many of these applications require vision algorithms and systems
to work under partial occlusion, possibly under high clutter, low contrast, and
changing environmental conditions. This requires that the vision techniques
should be robust and flexible to optimize performance in a given scenario.
The field of machine learning is driven by the idea that computer algorithms
and systems can improve their own performance with time. Machine learning
has evolved from the relatively “knowledge-free” general purpose learning system,
the “perceptron” [Rosenblatt, 1958], and decision-theoretic approaches
for learning [Blockeel and De Raedt, 1998], to symbolic learning of high-level
knowledge [Michalski et al., 1986], artificial neural networks [Rowley et al.,
1998a], and genetic algorithms [DeJong, 1988]. With the recent advances in
hardware and software, a variety of practical applications of the machine learning
research is emerging [Segre, 1992]. |
