Human imitation framework for a machine vision system based on soft computing tools
N.E. Gough, Qasim Mehdi, I. Griffiths, Michael John Allen · 2002
A prototype machine vision system inspired by the human vision system is described. Miniature cameras are actuated via push-pull flexible links and an adaptive controller initiates rapid movements to fixate a target of interest in a sub-region of the image. Object identification is achieved using a neural network that classifies on the basis of colour and is relatively tolerant to changing lighting conditions. Locations of sub-regions of interest are decided by a fuzzy-tuned stochastic transition method that is capable of maintaining interest in several objects, whilst efficiently scanning other regions where new objects might appear. Typical experimental results are presented to demonstrate how the system can maintain attention on a moving object whilst focussing on a static object.