COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	{I50,0W0,1260,0,1900JUFAJC} SUMMARY OF MY RECENT AND FURTURE WORK IN 
C00007 ENDMK
C⊗;
{I50,0;W0,1260,0,1900;JUFA;JC} SUMMARY OF MY RECENT AND FURTURE WORK IN 

{FDJC} COMPUTER VISION AND ROBOTICS {FA}

{FAJC} BRUCE G. BAUMGART



{JC} ~OVERALL~

	I  have been  and  am  continuing  to do  vision  programming
directed  towards  making  a computer system  that  can  manipulate and
navigate in the real mundane world.  My approach may be characterized
as ad hoc,  numerical and experimental
Overall, I work under the assumption that vision
consists of  three modes: DESCRIPTION,  VERIFICATION and  RECOGNITION
performed at  three levels of abstraction 2-D  IMAGES, 3-D MODELS and
TASK oriented procedures.


{JC} ~RECENT WORK~

	I have recently completed a thesis titled "Geometric Modeling
for   Computer  Vision"  which   included  a   demonstation  of  pure
~DESCRIPTIVE VISION~ in which  3-D polyhderal models of objects  were
computed  from television  views of  the objects  being rotated  on a
turntable.   In computer  vision geometric models provide  a goal for
descriptive  image  anaylsis,   an  origin  for  verification   image
synthesis  and a  context for  spatial problem  solving. Some  of these
design ideas have been implemented in two programs called GEOMED  and
CRE. 


{JC} ~CURRENT WORK~

	I am currently working on ~VERIFICATION  VISION~ under an NSF
grant  for  research  related  to  automating  the  task  of  factory
assembly.   I am  just now  adding simple  ad  hoc verification vision
routines to the existing manipulation programs of Lou Paul so that an
expected  hole can  be visually  located and  such that  an automatic
screw fetch can be visually verified. At present these two simple tasks
are done tactily. 


{JC} ~FUTURE WORK~

	I intend to develope  better 3-D models such as  to represent
the  occupancy of  objects in  space and  curved objects;  better 2-D
image analysis programs such as  to convert sequences of images  into
accurate photometric,  geometric and topological data  concerning the
world  being  viewed; and  I  intend  to develope  goal  seeking task
performing "creature"  simulations which  I believe to  be the
simple essense of intelligence. On such a foundation, the final
goal of ~RECOGNITION VISION~ approaching human ability can be built.