METHODS AND ALGORITHMS FOR OBJECT RECOGNITION

Abstract

The paper  deals with  problems  of  computer  image  processing  with a  view to applications  in  the  area  of  industrial  robotics.  Besides  classical  methods  for  object recognition especially nontraditional methods and  algorithms (the Multi Layer Perceptron neural  network  algorithm  and  the  Levenshtein  algorithm)  were  tested  in  simulation environment. Implemented algorithms were tested on the simulated objects of technological scene  given.  The  results  of  the  simulation  experiments  show  that  the  Multi  Layer Perceptron  neural  network  with  the  Back-Propagation  algorithm  and  the  Levenshtein algorithm are  very promising for object recognition of technological scenes for the use of industrial robots control.