Page 195 - Introduction to Autonomous Mobile Robots

Chapter 4
180
sequences generated using the above algorithm. The top string should match Place 1, but
note that there are deletions and insertions between the two strings.
The technique used in the fingerprinting approach for string differencing is known as a
minimum energy algorithm. Taken from the stereo vision community, this optimization-
based algorithm will find the minimum energy required to “transform” one sequence into
another sequence. The result is a distance metric that is relatively insensitive to the addition
or subtraction of individual local features while still able to robustly identify the correct
matching string in a variety of circumstances.
It should be clear from the previous two sections that whole-image feature extraction is
straightforward with vision-based perception and can be applicable to mobile robot local-
ization. But it is spatially localized features that continue to play a dominant role because
of their immediate application to the more urgent need for real-time obstacle avoidance.