13. Vision Inspection 2012

Embed Size (px)

Citation preview

  • 8/20/2019 13. Vision Inspection 2012

    1/24

    Vision and Visual Search

  • 8/20/2019 13. Vision Inspection 2012

    2/24

    Why is understanding visual search

    performance it important?• Designing:

    • Inspection for quality assurance

    • Inspection for maintenance

    • Inspection for safety

    • Inspection, inspection, inspection…and –• Human-environment interaction

    • Human-product interaction

  • 8/20/2019 13. Vision Inspection 2012

    3/24

    Poor Lighting

    • Can result in eye fatigue– Low light levels

    – Glare

    • Impairs tasks thatrequire vision– Inspection

    – Quality

    – Service

  • 8/20/2019 13. Vision Inspection 2012

    4/24

    Visual Search and Detection

    • The goal of the HFE engineer is usually to

    maximize accuracy, minimize search times,

    and minimize effort

    • Sometimes these occur together and

    sometimes do not

    • Different factors affect each differently

  • 8/20/2019 13. Vision Inspection 2012

    5/24

    Visual Search & Detection

    • Search– Eye Movements through the visual field (various

    strategies)

    • Detection

    – Targets and non-targets (noise)

  • 8/20/2019 13. Vision Inspection 2012

    6/24

    Find the Letter “V”

    Raise your hand when you do!

  • 8/20/2019 13. Vision Inspection 2012

    7/24

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    v   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

  • 8/20/2019 13. Vision Inspection 2012

    8/24

    x

    xxx

    xx

    x

    x

    xx

    xxx   xxx x

    x

    x   xxx

    x

    x

    x

    x

    x

    x

    xxx   x

    xx

    xx

    x

    x

    xx xx

    xxxx xx

    x   xxx x

    xx

      xxx

    x

    xx

    xxxx

    x

    x   xx

    x

    xx

    v

    xxx xx

    xx

    x

    x  x

    xx xxx

    xxx

    x

    x

    x

    x

    xx

    xxx

    xx

    x

    x

    x

    x

    xx

    x   xx

    x

    x

    x

    x xxx xx

    x   xx

    x xx

    x xxxx

    xx

    x

  • 8/20/2019 13. Vision Inspection 2012

    9/24

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xvx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

  • 8/20/2019 13. Vision Inspection 2012

    10/24

    x   xxx xx

    x

      xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xxx   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxv xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

    x   xxx xx

  • 8/20/2019 13. Vision Inspection 2012

    11/24

    Search Strategies

    Serial Search Model

    • Targets & Non-targets

    • Items inspected serially (one at a time) todetermine target/non-target

    • Search patterns affected by organization of space(e.g., left right/top down if organized; more

    random and less exhaustive if no pattern)• For targets that are not readily visible, the increase

    in p(detect) diminishes with time

  • 8/20/2019 13. Vision Inspection 2012

    12/24

    Visual Search & Detection

    • Influences– expectancy – where a target might be

    • use to structure display so know where to start thevisual search

    – conspicuity leads to its detection (“pops out”)of field – search is effectively parallel, or all at

    once rather than serial• good if pop out is desirable, but will distract if not a

    target

  • 8/20/2019 13. Vision Inspection 2012

    13/24

    Detection

    • Once possible target has been detected

    through search, must decide if it is in fact a

    target

    • Can use models of signal detection

  • 8/20/2019 13. Vision Inspection 2012

    14/24

    Signal Detection Theory

    • Model of how humans separate

     “noise” = all stimuli, whether relevant or not from

    “signal” = the relevant, important information

    To make decisions aout, e!g!,

     "  #art defect vs! not

     " $hich icon to click on

  • 8/20/2019 13. Vision Inspection 2012

    15/24

    %pplications of Signal Detection

    Theory to &nspection• 'ollect oservation (through sensory systems),

    then determine if there is enough evidence toconstitute a signal "  'rack is large enough

     "  *egion on +ray is cloudy enough

     "  -ever is high enough (here, not necessarily throughsenses, ut could e)

  • 8/20/2019 13. Vision Inspection 2012

    16/24

    Signal Detection Theory

    Hit False Alarm

    Miss   CorrectRejection

    Signal

    Say

    Noise

    Noise

    SaySignal

    Type 1Error 

    Type 2

    Error 

    State of the World

       O  p  e  r  a   t  o  r   B  e   h  a  v   i  o

      r

  • 8/20/2019 13. Vision Inspection 2012

    17/24

    Signal Detection Theory

    • Sensitivity (d’)- how good an operator is at

    telling signal from noise

    – “Bottom Up Influencers”: an be influenced bysignal to noise ratio, redundancy of signal, etc.

    – “Top Down Influencers” – personal

    characteristics

    – These can be influenced by a number of things 

    E.g., payoffs, incentives, experience,

  • 8/20/2019 13. Vision Inspection 2012

    18/24

    'ase Study

    • Manufacturing #lant for a #harmaceutical 'ompany• .isual &nspection Task to identify defective pills (color an

    important cue)

    • /utput (efore intervention)0 123 ppm

     "  'ommon #rolems0• -atigue

    • 4igh Turnover 

    • $astage 4igh (false alarms)

    • Management willing to look at a new manufacturing process

     "  5oals0 More output, less personnel cost, reduce wastage

    • 4ow can information aout visual search ailities and Signal

    Detection Theory help6

  • 8/20/2019 13. Vision Inspection 2012

    19/24

    Task0 To inspect pill color 

    Manufacturing #rocess0

    7atches (12 parts8atch) on a

    'onveyor 

  • 8/20/2019 13. Vision Inspection 2012

    20/24

    .isual Search and &nspection• -astpaced

    • 9ncertainty of pill in terms of color 

    • .ariaility in response (accept or re:ect)

    • *ules help, ut do ;/T rela< :udgment and have costsassociated with time

  • 8/20/2019 13. Vision Inspection 2012

    21/24

     ;ot &deal &deal ;ot &deal

    .isual *ules (posted in front of the inspector)

    Suggested inspection time: 3 sec/batch = .214 sec per

    pill

  • 8/20/2019 13. Vision Inspection 2012

    22/24

      %dditional Suggestions0 filtering

  • 8/20/2019 13. Vision Inspection 2012

    23/24

    %dditional Suggestions0 separate into rows

  • 8/20/2019 13. Vision Inspection 2012

    24/24

      Redced !n"ormation

    F    l    o  #    n  o  t    i    n  t   e  

    r   r       p  t   e  d     $   

     e  

    x    p  e  c  t   a  

    n  c   y   

     4-8> &ntervention01! /ptimal presentation of material

    ?! -ast rate of parts (@ A3B)

    C! Saved 1! million8Euarter  reduced material wastage

    increased output

    reduced fatigue F MSD