ABSTRACT
The concept of feature selection in visual servoing is different from that in object recognition. The rest of this chapter will focus on feature selection in the context of visual servoing. The features selected will substantially affect vision robustness and hence visual servoing performance [Janabi-Sharifi and Wilson, 1997]. Therefore, feature selection constitutes an important aspect of visual servoing, and its goal is to find the best features that provide speed, accuracy, and reliability of image measurements and characteristics of vision robustness. Examples of task failures with poor feature selection can be found in Smith and Papanikolopoulos [1996]. Two fundamental issues in feature selection are the
number
and
quality
of the features [Feddema et al., 1991]. Because processing of a large feature set is computationally infeasible, the focus will be on the selection of a minimum but information-intensive and reliable feature set.
