MODEL-based Recognition in Robot Vision 1986 Chin Dyer

Nov 01, 2012 08:58 PM

Roland Chin and Charles Dyer  1986  from  Univ of Wisc Madison.

from Computing Surveys   March 1986...

his  paper  presents  a  comparative  study  and  survey  of  model-based  object-recognition

algorithms  for  robot  vision.  The  goal  of  these  algorithms  is  to  recognize  the  identity,

position,  and  orientation  of  randomly  oriented  industrial  parts.  In  one  form  this  is

commonly  referred  to  as  the  “bin-picking”  problem,  in  which  the  parts  to  be  recognized

are  presented  in  a jumbled  bin.  The  paper  is  organized  according  to  2-D, 2.5d,  and  3-D

object  representations,  which  are  used  as  the  basis  for  the  recognition  algorithms.  Three

central  issues  common  to  each  category,  namely,  feature  extraction,  modeling,  and

matching,  are  examined  in  detail.  An  evaluation  and  comparison  of  existing  industrial

part-recognition  systems  and  algorithms  is  given,  providing  insights  for  progress  toward

future  robot  vision  systems.

need to be real time

goal is to precompile a descrip of each of known set of industrial parts,

then use these object models ro recognize in an image each instance of an object

and to specify its position and orientation relative to the viewer.

Attachments