應用機器視覺於IC包裝條管製程之線上檢測

(1)

ᑕϡ፟ጡෛᛇٺ IC Β྅୧გᄦ඀̝ቢ˯ᑭീ

ౘߌฯ അྈᝋ

̂ཧ̂ጯ፟ୠᄃҋજ̼̍඀ጯր

ၡ! ࢋ

ώ͛͹ࢋ੫၆ IC Β྅୧გᄦ඀൴ण˘ቢ˯፟ୠෛᛇҋજᑭീր௚Ă֭ࡁ

൴ᇆည఍நԫఙΒ߁ᇆညؠҜăᇆညͧ၆ăᇆည̶ᙷ͎̈́̇ณീඈĂͽᑭീ IC टळ̝̙։ݡĂ྿זቢ˯ҋજ̼Бᑭ̝ϫ۞ĄIC Β྅୧გ̝৵Մࠎͯېሤ๬

ّՄफ़ĂϤٺՄफ़ώ̝֗ሤණҽᒺٕΐሤ̙πӮĂົౄјटळܑࢬΌౝٕΎ੓

ଐԛ൴ϠĄ̙։ݡᑭീ̶׌ล߱ซҖĂௐ˘ล߱Ӏϡᇆညͧ၆͞ёĂགྷϤޞീ

ۏᄃᇾ໤ۏઇᇆည࠹ഴĂͽᑭീ IC टळܑࢬ̝༂ஷݡĂ֭ͽ௚ࢍ̝͞ёපפ ѩ࠹ഴᇆည̝পᇈࣃĂࢫҲᇆည఍ந̝ኑᗔޘĂ੫၆ѩ࠹ഴᇆညώ͛೩΍௚ࢍ

ᄦ඀ăFuzzy C-mean(FCM)׶ RBF ᙷৠགྷშྮˬ჌͞ڱֽᕩᙷѩ࠹ഴᇆညࠎ։

ݡٕ̙։ݡĂ֭ͧྵѩˬ჌͞ڱ۞ᐹКćௐ˟ล߱੫၆ IC टळ̝͎̇үณീĂ ͽᑭീ͎̇࿅ٕ̂࿅̝̙̈։ݡĂӀϡୗޘڱವᙝĂᅃͽᕭگΝੵ੼ᗔੈĂԱ

΍׌ᙝࠧ౵̈෼ᗓĂͽҿؠߏӎ෹࿅ٙనؠ̝ᇾ໤͎̇ቑಛĂ֭ྻϡྤफ़ऱࡔ ᐂ։ݡ̙̈́։ݡ۞ᄦ඀ᑭീྤफ़Ą౵ޢĂ૟ώ͛ٙ൴ण΍̝ᇆညᑭീԫఙ၁ᅫ ᑕϡٺ IC Β྅୧გ̝ϠயቢĂͽរᙋٙ೩͞ڱ̝ѣड़ّĂᏰᙊதΞ྿ 99%ͽ

˯Ą

ᙯᔣෟĈ፟ጡෛᛇăᇆည఍நăቢ˯ᑭീăIC Β྅୧გĄ

ON LINE INSPECTION TECHNIQUES OF THE IC PACKAGE CARRIAGE PROCESSING USING MACHINE VISION

Chaio-Shiung Chen Eau-Jang Chi

Department of Mechanical and Automatic Engineering Da-Yeh University

Chang-Hwa, Taiwan 514, R.O.C.

Key Words: machine vision, image processing, on line inspection, IC package carriage.

ABSTRACT

This paper develops an on line inspection technique for the IC package carriage processing using machine vision. Image-processing techniques are used to identify the various defects of the IC package carriage, including image position, image matching, and image classification. The IC package carriage is thermal plastic and thus easy to deform when heating is not equal. At the first stage, through subtracting the standard image from the measurement image, we employed image matching to detect surface defects. The features of the subtraction images are extracted by statistical techniques. For the subtraction image, we simultaneously use statistical

(2)

corresponding edges, dimension defects can be found. A database is built to record the inspection results. Finally, we apply the developed techniques to practical IC package carriage manufacturing processes to confirm the validity of the proposed method.

˘ă݈! ֏

IC Β྅୧გࠎᑕϡٺ IC ᄦ඀ޢბ IC ދ྅ޢ̝ IC ј ݡΒ྅Ăࠎፋ॑୧ې̝๬ቱͯՄͷ׎ܑࢬѣΌې̝ IC टळ Ξֻ IC ཉٸΒ྅ĂдΗጱវຽ੼Ϡயฉഇ̝˭Ăଂ೿๪̷

౷זދ྅ౌߏкயҋજ̼Ă༊ IC Β྅୧გ˯̝ IC टळѣ

༂ஷॡĂIC ૟̙ਕึӀٸˢटळ̰Ă̙ҭፋ॑ IC Β྅ε ΝԆፋّĂ˵ֹ IC јݡΒ྅ҋજ̼߹඀̙ึၰĂా૲ֹፋ

࣎ IC ҋજ̼ᄦ඀צז̒ᕘćࠎ˞ឰ IC ҋજ̼Β྅߹඀ึ

ၰĂдፋ̝॑ IC Β྅୧გ˯ࡗѣೀ˼࣎ IC टळĂυื̙

ਕѣ˘࣎༂ஷݡĂЯѩд IC Β྅୧გ̝ᄦౄ࿅඀̚Ă၆ٺ

׎˯̝Տ࣎ IC टळυืБᑭĂ̖ਕ྿זՏ࣎ IC टळ࿬༂

ஷ̝ϫ۞Ăҭݫٺᑭߤిޘ̈́ˠ˧Ăனѣ̝ݡኳᑭീ࠰Ϥ ˠ̍ٺᗓቢ˭٩ᇹԆјĂ൑ڱܲᙋՏ࣎ IC टळ࿬༂ஷĂЯ ѩĂࡁ൴˘इҋજ̼ቢ˯̝ IC टळᑭീր௚Ăͽפ΃ˠ̍

ᑭീߏࠎ̷࢝ᅮࢋ۞ćѩᑭീր௚Ξͽ၆ IC टळ̝ԛې༂

ஷซҖᑭീͷેҖిޘࢋૉԣͽ྿זቢ˯Бᑭ̝ϫ۞Ąன

̫ĂෛᛇᇆညԫఙజјΑᇃھᑕϡٺ IC ᄦ඀̝ݡኳᑭീ

[1-3]Ăෛᛇᇆညᛳٺܧତᛈёᑭീ͞ёĂ౅࿅ᇆညΪ́

CCD פညĂΞԣి૟ޞീۏ 2-D ԛېᖼј࿪ཝΞநྋ̝ᇴ Ҝྤफ़Ăତ඾౅࿅൴णᇴҜᇆည఍நԫఙͽԆј̙։ݡᑭ

ീ̝ϫ۞ĂϤٺᇆည఍நԫఙ̝ซՎ̈́࿪ཝેҖిޘ̝೩

੼Ăෛᛇᇆညϡٺ IC टळ̙։ݡ̝ᑭീࠎ˘ΞҖ̝ᏴፄĄ னѣ IC ༂ஷݡᏰᙊ̝ᇆည఍நԫఙĂࡁտ̝ኢ͛ܧ૱кĂ

̂࡭Ξͽᕩᙷј˭Еણ჌౉शĂௐ˘ᙷࠎ利用機率統計技術，т Bayesian ׶ Maximum likelihood ඈ̝็௚ᕩᙷ͞ڱ [4-6]Ăѩ͞ڱᅮࢋᄓะޝкޞീۏณീྤफ़ֻ௚ࢍ̶ژĂ ͽՐ଀ྤफ़̝፟த̶ҶĂ఍ந˯ྵ൑ड़தĂᏰᙊத˵ྵҲć ௐ˟ᙷࠎӀϡሀቘᕩᙷ͞ڱĂѩ͞ڱА૟ᇆည̝পᇈྤफ़ ሀቘ̼Ă൒ޢГ൴ण੊ቚڱ݋఍நᕩᙷ఺ֱሀቘྤफ़ͽ଀

זᏰᙊड़ڍт Kohonen ጯ௫ڱ݋[7]ăHyperbox ͞ڱ[8]׶

Fuzzy C-mean[9]ඈĂѩ͞ڱΞޝटٽ૟૞छ̝གྷរΐˢሀ ቘᏰᙊߛၹ̚Ăͷሀቘ̼̝পᇈྤफ़ͧྵܕҬৌ၁̝ᇆည পᇈೡࢗĂЯѩΞ೩੼ᏰᙊதĂ1997 ѐ Khunkay ඈˠ[10]

Аͽ Kohonen ̝ҋ௡ᖐপᇈߍड(SOM)͞ڱҋજපפᇆည

̝ሀቘপᇈࣃĂ൒ޢГͽ Fuzzy C-mean ͞ڱವԱ఺ֱሀቘ পᇈࣃ౵ָ̼̝ᕩᙷڱ݋ĂјΑԆјОה࿪ྮڕ̝̙։ݡ ᑭീćௐˬᙷࠎӀϡᙷৠགྷშྮր௚Ăᙷৠགྷშྮր௚Я

׍౯ጯ௫ᄃܧቢّᙝ̶ࠧᙷਕ˧Ăߏ౵ᇃھజᑕϡٺ IC ᄦ඀ᑭീ̝͞ڱĂѩ͞ڱΞͽ̶ࠎႾ༛ёᄃܧႾ༛ёጯ௫

ր௚׌჌ĂႾ༛ёጯ௫ڱ౅࿅੊ቚྤफ़Ξͽटٽጯ௫૞छ

̝གྷរĂҭ༊̶ᙷᙝࠧኑᗔॡĂӀϡкᆸ̝ᙷৠགྷშྮќ ᑦిޘតјߏ࣎યᗟĂ2002 ѐ SU ඈˠ[11]ĂТॡ̶ҾӀ ϡ backpropagation (BP) ăradial basis function (RBF)׶

learning vector quantization (LVQ)ˬ჌ᙷৠགྷშྮٺ IC ೿

̷̮౷ޢ̝̙։ݡᑭീĂ౅࿅Լត̙ТᙷৠགྷᏮˢಏ̮ᇴ ϫ̈́ᔳᖟᆸ࣎ᇴĂͽͧྵ఺ˬ჌͞ڱ̝ќᑦిޘ̈́Ᏸᙊ தĂ൴ன఺ˬ჌͞ڱд̙Т୧І˭ѣ̙Т̝ќᑦड़ڍĄᇆ ညͧ၆(pattern matching)ࠎᏰᙊ։ݡᄃ̙։ݡ౵ۡତᄃᖎ ಏ̝͞ڱĂ2002 ѐ Sakurai ඈˠ[12]ᑕϡᇆညͧ၆̝ԫఙٺ IC ̝ Hot-Al-Cu ጱቢᆸ༂ஷݡᑭീĂ౅࿅አፋ፶Ѝ໰ځૻ

ޘĂͽ̶ᗓᗔੈᄃ̙։ݡᇆညͧ၆ޢ̝ѷล̶ҶĂଂ̙Т ၁រ̚Ա΍౵ָ̝ܝᕣࣃĂҭ઄тᗔੈᄃᇆညͧ၆̝ѷล

̙टٽ̶ᗓĂ݋ᏰᙊதົࢫҲĄ

னѣ̝ࡁտ̚ᑕϡٺྋՙ IC Β྅୧გ̝ IC टळ̙։

ݡᑭീ͛ᚥ;̝ᗒтĂ҃ IC Β྅୧გ̝৵ՄࠎܑࢬЍ໣̝

โҒͯې๬ቱĂໂटٽͅЍĂͷࠎሤ๬ّՄफ़Ăдགྷΐሤă ሀ׍ᑟ٪јݭ̈́ฟሀ࿅඀̚ĂϤٺՄफ़ώ̝֗ሤණҽᒺă ΐሤ̙πӮٕ௲ሀॡม͉ԣඈЯ৵ĂटٽౄјटळܑࢬΌ

ౝٕΎ੓۞ଐԛ൴ϠĂͷटळܜăᆵ׶Όౝஎޘᐌ̙Т IC ఢॾ҃ؠĂᄦౄిޘࡗՏࡋ 2 Ҍ 6 ሀĂᐌ IC ̂̈҃ؠĂᑭ

ീჟޘࡗ 0.1mmĂЯѩٙ൴णҋજ̼̝ቢ˯ᇆညᑭീր௚

υืਕዋϡٺ̙Тఢॾ̝ IC टळͷᑭീిޘࢋૉԣĂ̙ਕ ഴၙ IC Β྅୧გ̝ᄦ඀Ăώ͛૟ޙϲ˘ෛᛇᇆညᑭീր

௚Ă੨Ъ IC Β྅୧გᄦ඀၆ IC टळҋજפညĂ֭൴णᇆ ည఍நԫఙΒ߁ᇆညؠҜăᇆညͧ၆̈́ᇆည͎̇ณീඈĂ ͽᑭീ IC टळ̝̙։ݡĂ྿זቢ˯ҋજБᑭ̝ϫ۞Ăԫఙ ᆸࢬ͹ࢋ̶ࠎα࣎ొ̶Ĉ࿪ཝ̬ࢬăᑭീؠҜăᇆည఍ந

ࢍზ׶ྤफ़ऱޙϲĄ

˟ăIC Β྅୧გᄦ඀ߛၹ

IC Β྅୧გࠎሤ๬ّͯې๬Մགྷሀ׍ΐሤᑟᑅј ԛĂՏ୧ IC Β྅୧გ˯ࡗѣೀ˼࣎ IC टळֻ IC Β྅ĂIC टळԛېົॲፂٙࢋΒ྅̝ IC ԛې҃ؠтဦ 1 Ҍဦ 4 ٙ ϯĂ͎̇ଂ 10mm Ҍ 30mmĂώϠயቢ IC ୧გ̝჌ᙷࡗ 50 ೀ჌Ąдፋ୧ IC Β྅୧გ˯Ă̙टధѣ˘࣎ IC टळࠎ

༂ஷݡĂ̙൒ົᇆᜩҋજΒ྅߹඀̝ึၰĄώ͛ IC Β྅୧ გ̝ᄦ඀тဦ 5 ٙϯĂ׎ᄦ඀ࠎдΐ݈̍॑ې৵Մజߛд ଡજёਖ਼फ़፟˯ĂᖣϤՎซ੺྿੨Ъᓲ୛ͽ׽ؠҖ඀͔ٛ

(3)

ဦ1 ݭཱི16-PT̝ICटळ ဦ3 ݭཱིTO-263-1̝ICटळ

ဦ2 ݭཱིSM-5X̝ICटळ ဦ4 ݭཱིT-2424-01̝ICटळ

CCD

ဦ 5 IC Β྅୧გቢ˯ᇆညҋજᑭീր௚ᄦ඀ဦ

ဦ 6 IC टळܑࢬតԛ

ਖ਼फ़ĂՏҖ඀ࡗ 96mmĂͯې৵Մдᑟᑅјԛ݈ᅮАΐሤ

ֹ׎హ̼Ă൒ޢͽሀ׍ᑅᑟ΍ਕट IC ̝टळĂՏѨᑟᑅ̝

ሀᇴࠎ 4ă6 ٕ 12 ሀĂֶፂ IC ͎̇̂̈҃ؠĂՏѨਖ਼फ़ؠ Ҝ̈́ᑟᑅјݭॡมࡗ 1.2secĂͷѩ IC टळ׌઎͕̈́̚ᅮ՞

΍ؠҜ͋Ă҃ᇆညෛᛇր௚྅ཉд IC टүјݭޢბĂΞᖣ Ϥ CCD ၆ IC टळפညĂ֭ᑕϡᇆည఍நԫఙેҖ̙։ݡ ᑭീĂͷࠎቢ˯БᑭĂ౵ޢ IC Β྅୧გజ෗ગј׽ؠ۞ܜ ޘĄIC टळ̝̙։ݡ̂ࡗΞᕩৼࠎܑࢬតԛтဦ 6ăᙝࠧ

តԛтဦ 7 ͎̈́̇តԛтဦ 8 ඈˬ჌Ăԧࣇ૟൴णᇆညᑭ

ീԫఙĂ੫၆ѩˬ჌༂ஷݡซҖ IC टळ̝ቢ˯ᑭീĄώ͛

̝ᇆညෛᛇր௚ർវߛၹтဦ 9 ٙϯĂ׎̚࿪ཝֹϡ PentiumIII 733MHz ̝࣎ˠ࿪ཝĂᇆညᕜפΙ݋ֹϡ Matrox

̳Φ۞ Meteor standardǶݭᇆညᕜפΙĂᇆည఍நహវࠎ

ဦ 7 IC टळᙝࠧតԛ

ဦ 8 IC टळ͎̇តԛ

A.

B.

C.

A.

B.

C.

Sony 1/2"

640×640

Meteor Standard 2 ADIO-113

CCD PLC

ဦ 9 ᇆညҋજᑭീր௚ߛၹဦ

MIL v6.1ĂCCD ࠎ Sony ̳Φ۞ 1/2”โϨᛷᇆ፟Ăྋژޘ ࠎ 640 Ű480Ă໰ځ፶Ѝֹϡາֲ߷̳Φᒖݭ LED ̝ᖡҒ

፶ЍĂ̬ࢬΙࠎ౶ܷ̳Φ۞ ADIO-113 ̬ࢬΙĂд໛఼̬

ࢬֹϡ ISA ̬ࢬĂᖣϤ ADIO-113 ̬ࢬΙତќ PLC ਖ਼΍

̝ؠҜੈཱིĂͽઇࠎᇆညր௚ୁજפည۞ੈཱིĂ҃༊ᇆ ညր௚ѣᑭീז̙։ݡॡĂГགྷϤ̬ࢬΙਖ਼΍ᛋϯੈཱི

Ҍ PLCĄ

ˬăᇆညᑭീ߹඀

ᇆညᑭീ߹඀Ξ̶ࠎᗓቢనؠ̈́ቢ˯ᑭീ׌჌тဦ 10 ٙϯĂЯࠎ IC Β྅୧გఢॾ჌ᙷிкĂдᑭീ݈ᅮޙ ϲᑭീᇾ໤ͯĂͽઇࠎᑭീֶ̝ፂćᗓቢనؠॡࢵАᕜפ

։ݡ̝ IC टळ༊үᇾ໤ͯĂ൒ޢдѩᇾ໤ͯ˯॰פࢋͧ၆

̝ IC टळĂՏ࣎ࢋᑭീ̝ IC टळΞ॰פк௡ࢋณീ͎̇

̝࠹၆ᑕᙝࠧ͞॰ĂՏ௡͎̇ณീ͞॰υᅮనؠ͎̝̇ᇾ

໤ࣃ̳̈́मቑಛĂͽүࠎ̶Ᏸ͎̙̇։ݡֶ̝ፂĂ҃ᇆည

(4)

Y

FCM RBBF

N

N N

Y

ဦ 10 ᇆညᑭീր௚߹඀ဦ

̚Տ࣎ည৵ٙ΃ܑ̝͎̇ྋژޘࣃĂϤᇾ໤̝ͯ၁ᅫ͎̇

ࣃᄃٙҫည৵࣎ᇴઇᖼೱĂ၁ᅫ̝ᗓቢనؠ൪ࢬтဦ 11

ٙϯĄ

ቢ˯ᑭീΒ߁ᇆညͧ၆ᄃ͎̇ณീ׌჌͞ёĂͽֹਕ ЧҾ̶Ᏸ΍ IC टळܑ̝ࢬ༂ஷᄃ͎̇༂ஷ̝̙։ݡĂдቢ

˯ᑭീॡĂ၆ٺགྷפည̝ޞീ IC टळᇆᇹυืАซҖᇆည

݈఍நᄃᇆညؠҜĂᇆᇹ݈఍நࠎֹϡӮ̼ᕭگጡ(smooth filter)Νੵᇆည̝੼ᗔੈĂ׎ 3Ű3 ̝ዌཊтဦ 12 ٙϯĂ҃

ࠎΝੵޞീͯᄃᇾ໤ͯᇆညЍܪޘ̙Т۞ᇆᜩĂͽπӮѷ ลࣃ۞៍هֽٛܕЍܪᕇ۞मளĂࢵАზ΍ᇾ໤ͯᄃޞീ

ͯᇆညπӮѷลࣃ̝मளࣃ_{li _}_imgт˭ёĈ

∑ ∑ −

= ×

= = N y

M

x fst x y fin x y N

img M li

1 1

)]

, ( ) , ( 1 [

_ (1)

׎̚ f_st(x,y)׶ f_in(x,y)̶Ҿࠎᇾ໤ᇆည׶ޞീᇆညд )

,

(x y Ҝཉ۞ѷลࣃĂГ૟ޞീۏᇆည̝Տ࣎ည৵ѷลࣃ ΐ˯(1)ёზ΍̝मளࣃli _imgĂͽΝੵᇾ໤ͯᄃޞീͯᇆ ညܪޘ̝मளĂ҃ޞീͯᇆည̝າ۞ѷลࣃNin(x,y)т˭

ёĈ

ဦ 11 ᗓቢనؠ̝Ѩ൪ࢬనؠ

1 2 1

2 4 2

1 2 1

ဦ 12 3 Ű3 ዌཊ̝Ӯ̼ᕭگጡ

img li y x f y x

Nin( , )= _in( , )+ _ (2)

˘ਠՏ࣎ IC टळ̚δѣ˘࣎๪͋ĂԧࣇΞͽԱ΍ѩ๪̝͋

͕̚үࠎᇆညͧ၆ॡؠҜ̝ϡĂࠎ˞ਕૉԱ΍๪͋ᙝࠧĂ А ֹ ϡ Sobel ۞ ዌ ཊ ఍ ந Ă ͽ ܮ Ա ΍ ๪ ͋ ۞ ᙝ ࠧ ᕇ

N n y

x_n, _n), 1, ,

( = L Ă൒ޢͽ౵̈π͞ڱӀϡ఺ֱᙝࠧᕇՐ

΍̈๪̝͋๪͕Ă઄т๪̝͋͞඀ёࠎĈ

2 2

2 (x c_x) (y c_y)

r = − + − (3)

׎̚ r ࠎ๪Ηश׶(c_x,c_y)ࠎ๪͕ळᇾĂనؠт˭̝ᆊࣃב ᇴĈ

[ ]

²

1

2 2

2 0

0 ( ) ( )

2 ) 1 , ,

( = ∑ − − − −

= N

n n x n y

y

x c r x c y c

c r

L (4)

੫၆r ă₀ c ׶_x c ઐ຋ͷ΄׎ඈٺ࿬т˭ёĈ _y

[ ] ^{[ ]}

[ ] ^[ ^]

[ ] [ ]

∑ − − − − − =

∂ =

∂

∑ − − − − − =

∂ =

∂

∑ − − − − =

∂ =

∂

=

N

n n x n y n y

y N

n n x n y n x

x N

n n x n y

c y c y c x c r

L

c x c y c x c r

L

r c y c x r r

L

1

2 2 2

0 1

2 2 2

0 1 0

2 2

2 0 0

0 ) ( 2 ) ( ) (

0 2 ) ( ) (

(5)

Г૟(5)ёொีፋநޢΞՐ଀౵ָ̼̝๪͕т˭Ĉ

c AB c

y

x =













(6)

(5)

1

2

1 2

1 1 1

1

2

1 2

2 2

−

= =

= = =

= =























 ∑ 



 



∑

 −

 

 ∑ −∑ ∑





∑ −∑ ∑











 ∑ 



 



∑

−

=

N n

N n n n N

n

N n n N n n n n

N n

N n n N n n n n N

n N n n n

y y N y x x y N

x y x y N x

x N

A (7)













∑ 



 



∑ +∑

−



 



∑ +∑

∑ 



 



∑ +∑

−



 



∑ +∑

=

= = =

= =

= = =

= =

N n

N n n n n N

n

N n n n n

N n

N n n n n N

n N n n n n

y x y y y x N

y x x x y x N B

1 1 1

2 2

1 1

3 2

1 1 1

2 2

1 1

2 3

(8)

ზ΍ѩ͕̚๪̝͋๪͕૟ޞീۏؠҜޢĂಶΞซҖᇾ໤ͯ

ᄃޞീ̝ͯᇆညͧ၆ĂͽԱ΍ IC टळܑࢬ̝༂ஷݡĄ

αăᇆညͧ၆

ᇆညͧ၆౵ᖎಏ̝͞ڱࠎᇆည࠹ഴĂ૟ᇾ໤ͯᇆညᄃ གྷؠҜ̝ޞീͯᇆညઇᇆည࠹ഴĂ׎ᇆည࠹ഴޢ̝ည৵ѷ ลࣃ f_sub(x,y)ࢍზт˭Ĉ

) , ( ) , ( ) ,

(x y f x y Ninx y

f_sub = _st − (9)

੫၆ѩ࠹ഴޢ̝ᇆညĂԧࣇ૟̶ҾӀϡ௚ࢍᄦ඀ăFuzzy C-mean ׶ RBF ᙷৠགྷშྮඈˬ჌͞ёֽҿҾѩ࠹ഴޢ̝

ᇆညࠎ։ݡٕ̙։ݡĂ֭ͧྵ׎ᐹКĂ̶Ҿୃࢗт˭Ĉ 1. ௚ࢍᄦ඀͞ڱ

༊ޞീͯᄃᇾ໤ͯܧ૱࠹Ҭॡ݋࠹ഴᇆည̝πӮѷ ลࣃ૟ᔌܕٺ࿬Ă΃ܑޞീͯࠎ։ݡć̝ͅĂтڍ׌۰म ள෸̂Ă݋࠹ഴᇆည̝πӮѷลࣃ෸̂Ă΃ܑޞീͯࠎ̙

։ݡĂ҃ҿҾ։ݡᄃ̙։ݡπӮѷลࣃ̝ܝᕣࣃĂώ͛Ӏ ϡ௚ࢍ͞ڱĂפ 200 ࣎։ݡ̝ IC टळᇆညᄃᇾ໤ͯᇆည࠹

ഴĂГࢍზ΍ѩ 200 ࣎࠹ഴᇆည۞πӮѷลࣃ X ̈́ᇾ໤म σĂགྷ၁រ൴னᄦ඀გטቢࡗдᇾ໤म 2.1 ࢺॡ౵рĂӈ

༊ޞീ̝ͯ࠹ഴᇆညπӮѷลࣃ̈ٺX+ 12. ×σ ࠎ։ݡĂ

̂ٺѩࣃࠎ̙։ݡĄ

дֹϡ Fuzzy C-mean ׶ RBF ᙷৠགྷშྮ͞ڱ̝݈Ă Տ࣎࠹ഴᇆညυืАපפ΍পᇈࣃĂͽഴ͌ᇆညͧ၆̝ॡ ม׶ኑᗔޘĂώ͛ؠཌྷ΍α࣎পᇈࣃт˭Ĉ

(˘) ࠹ഴᇆညπӮѷลࣃ( f ) ₁

࠹ഴᇆညٙѣည৵̝πӮѷลࣃ X ࢍზт˭Ĉ

× ∑ ∑

= = = N y

M x fsub x y N

X M

1 1

) , 1 (

(10)

(˟) ࠹ഴ̙ࠎ࿬ᕇည৵࣎ᇴ( f ) ₂

࠹ഴᇆည۞ѷลࣃд10< f_sub(x,y)<250ቑಛ̰̝ည

৵ᕇ࣎ᇴĄ (ˬ) ඗၆मࣃ( f ) ₃

࠹ഴᇆညѷลࣃд10< f_sub(x,y)<250ቑಛ̰Ăѷล ࣃ౵݈̣̂࣎πӮࣃA_maxᄃѷลࣃ౵̈ޢ̣࣎πӮ ࣃA_min̝඗၆मࣃĂࢍზт˭Ĉ

min max A A

A= − (11)

(α) ፋវᇾ໤म(f ) ₄

࠹ഴᇆညٙѣည৵̝ᇾ໤मࣃσĂࢍზт˭Ĉ

∑ ∑ −

−

= ×

= = N y

M

x fsub x y X

N

M 1

2 1

2 ( ( , ) )

1

σ 1 (12)

˫ϤٺՏ࣎পᇈࣃ۞̂̈ડાቑಛ̙ТĂ߇ื੫၆Տ

࣎পᇈࣃઇϒఢ̼т˭Ĉ

) max( _i

i i

f

x = f Ăi=1,2,3,4 (13)

ЯѩĂՏ࣎ޞീ̝ͯ࠹ഴᇆညౌΞපפ˘পᇈШณ x T

x, , ] [ ₁L ₄

=

X ĂΞͽӀϡѩপᇈШณ૟ޞീͯᕩᙷ

ј։ݡٕ̙։ݡĄ 2. Fuzzy C-mean ͞ڱ

Fuzzy C-mean 是˘჌ӀϡሀቘទᏭ۞ᕩᙷ͞ڱĂ౅࿅

੊ ቚ ವ Ա ཏ ะ ̚ ͕ ᕇ ྿ ז ᕩ ᙷ ϫ ۞ Ă ώ ͛ ᑕ ϡ Fuzzy C-mean ۞ᕩᙷ͞ڱĂ૟ޞീͯᕩᙷј։ݡᄃ̙։ݡ׌ᙷĂ

ࢵАᅮפ଀੊ቚྤफ़Ăѩ੊ቚྤफ़ΒӣЧ჌̙Т։ݡᄃ̙

։ݡ̝ݭၗĂᔘΒӣЧ჌̙Тఢॾ̝ IC टळĂགྷᄃᇾ໤ͯ

̝ IC टळᇆည࠹ഴĂϤ˯ᇴ̝͞ڱපפপᇈШณĄ઄тග ؠ n ࣎੊ቚྤफ़X Ă_k k=1,L,nͷX_k∈R⁴д఺ֱ n ࣎੊

ቚྤफ़̚Ăԓ୕ᕩᙷјc=2ᙷĂТॡ౵̼̈т˭̝ᆊࣃב ᇴĈ

∑ ∑ −

= = = n k

c

i m k i

J ki 1 1

V 2

µ X (14)

׎̚ • ࠎለೀ֧଀(Euclidean)෼ᗓבᇴăV_i,i=1, 2̶Ҿ ࠎ։ݡᄃ̙։ݡপᇈШณཏะ̝͕̚ࣃăµ_kiࠎௐ k ඊ੊

ቚྤफ़ᕩᛳٺௐ i ̶ᙷ۞ᕩᛳבᇴࣃ׶m>0ࠎ޽ᇴᝋࢦ ࣃĄ̶Ҿ၆µ_ki׶V ઇઐ຋གྷፋநޢΞ଀זт˭׌౵ָᝑ_i

΃ёĈ[8]

( )

∑ − − − −

= =

− c

j

m j

k i

k

ki t t t

1

) 1 /(

) 2

1 ( /

) 1 ( /

1 )

( X V X V

µ (15)

∑ ∑

= = =

n k

m ki k m ki

i t t t

1 1

)]

( [ / )]

( [ )

( µ X µ

V (16)

׎̚k=1,L,n׶i=1,2ĄЯѩĂFuzzy C-mean ̝੊ቚՎូ

т˭Ĉ

Վូ˘Ĉనؠ։ݡᄃ̙։ݡপᇈШณ͕̚ᕇ̝ܐؕࣃĂ )

0 ( ), 0 ( ₂

1 V

V Ą

Վូ˟Ĉͽ(15)ёࢍზՏ࣎੊ቚྤफ़ᕩᛳٺ։ݡᄃ̙։ݡ

׌ᙷ̝࣎Ҿᕩᛳࣃµ_ki(t)Ą

ՎូˬĈͽ(16)ёࢍზͷՀາ։ݡᄃ̙։ݡপᇈШณ͕̚

ᕇࣃV₁(t V), ₂(t)Ą

ՎូαĈࢦኑՎូ˟̈́Վូˬז႕֖т˭̝ᄱमҿҾё ॡĂઃͤՀາĄ

(6)

x₄ w₀

m

ဦ 13! RBF ᙷৠགྷშྮ

 ≤ε





 − − + − −

= ₁() ₁( 1)² ₂() ₂( 1)² ¹^/² )

(t t t t t

E V V V V (17)

當訓練出良品與不良品特徵向量之最佳群集中心值

*

V ᄃ1 V ĂಶΞͽӀϡѩ׌͕̚ࣃֽᕩᙷޞീͯࠎ։ݡٕ₂^*

̙։ݡć઄నޞീͯ࠹ഴᇆည̝পᇈШณࠎX_inĂӀϡለ

ೀ֧଀(Euclidean)෼ᗓבᇴёࢍზѩপᇈШณ̶ҾᄃV₁^*

̈́V ̝෼ᗓࣃт˭Ĉ ₂^*

* 1 1= X_in−V

d Ă d₂= X_in−V₂^* (18)

當d 值較小時，代表₁ X_in෼ᗓV ྵܕĂЯѩޞീͯ૟జ₁^* ᕩᙷࠎ։ݡĂ̝ͅĂࡶd ࣃྵ̈Ă΃ܑ₂ X_in෼ᗓV ྵܕĂ₂^* Яѩޞീͯ૟జᕩᙷࠎ̙։ݡĄ

3. RBF ᙷৠགྷშྮ͞ڱ

RBF 類神經網路如圖 13 所示，其中網路輸入為待測片 4 個特徵值X=[x₁,L,x₄]^T，y( X)ࠎშྮಏ࣎Ꮾ΍Ăᔳ ᖟ ᆸ ѣ m ࣎ ৠ གྷ ̮ ̈́ ٙ ࠹ ၆ ᑕ ̝ m ࣎ श Ш ב ᇴ

)

1(X

φ , φ₂(X) ,…, )φ_m( X Ă ׶ ᄃ Ꮾ ΍ ࠹ ా ̝ ᝋ ࢦ ω1,ω₂,…, ω_mĂѩᙷৠགྷშྮ̝Ꮾ΍т˭Ĉ

w X g X

X) ( ) ( ) (

1

0 m T

i i i

y = +∑ =

=ωφ

ω (19)

其中 w=[ω₀,ω₁,L ,ω_m]^T ׶ g(X)=[1,φ₁(X),φ₂(X),L ,

T m( X)]

φ ĂᏴፄφ_i( X)ࠎ੼೻בᇴт˭Ĉ

) / exp(

)

( _i ² _i²

i σ

φ X = −X−t , i=1,L,m (20)

׎̚t ࠎ͕̚ШณĂ_i σ_iࠎ੼೻בᇴᆵޘĄ༊ৠགྷ̮ᇴפ۞

࿅͌ॡĂــົѣྵҲ̝໤ቁّĂ࠹ͅ۞тڍৠགྷ̮ᇴפ

۞࿅кॡĂົᆧΐྻზ۞ኑᗔͷΞਕѣ overfitting યᗟĄ

઄тගؠ n ࣎੊ቚྤफ़(X_k,d_k)Ăk=1,L,nĂ׎̚ࠎ˞̶

Ᏸ੊ቚྤफ़ࠎ։ݡٕ̙։ݡĂ༊ௐ i ඊ੊ቚྤफ़ࠎ։ݡॡ నd_k =10Ă̝ͅĂࠎ̙։ݡॡనd_k =0Ă҃੊ቚϫ۞дٺ Աז౵ָᝋࢦങ w ͷ౵̼̈т˭̝ᆊࣃבᇴĈ

( )

∑ −

= = n

k k

T

dk

J

1

) 2

2 ( ) 1

(w g X w (21)

) ( )

(

1 ₁

1

k k

k k T

k ⁻ +g X ×p₋ ×g X

] ) ( [ )

( ₁

1 −

− + × × −

= _k _k _k _k ^T _k _k

k w p g X d g X w

w (23)

其中p ࠎᏮˢ۞Вតள৏ੱ(Covariance Matrix)ĄЯѩĂ_k RBF ᙷৠགྷშྮ̝੊ቚՎូт˭Ĉ

Վូ˘ĈᏴؠᔳᖟᆸ m ࣎ৠགྷಏ̮φ_i( X)Ăi=1,L,mĂ

׎̚φ_i(X)ࠎ੼೻בᇴт(20)ёĂৠགྷ̮͕̚t ࠎ_i ੫၆ЧᏮˢx д׎̶ొቑಛ̰ͽॾ̄ёπӮ̶_i

౷̶ёᏴפĂ҃੼೻̶οᆵޘσ_iࠎᏴפዋ༊̝ࣃ

̏உᄏপᇈШณ̶̝ҶቑಛĂనؠܐؕᝋࢦࣃ w ׶Вតள৏ੱ0 p₀=αIĂα>0Ą

Վូ˟ĈᏮˢ˘ඊາ۞੊ቚྤफ़ĂӀϡ(19)ёზ΍ᙷৠགྷ შྮ̝Ꮾ΍Ą

ՎូˬĈֹϡᅍਫ਼౵̈π͞ڱĂͽ(22)ёზ΍າ۞Вតள

৏ੱp ׶ͽ(23)ёზ΍າ۞ᝋࢦࣃ_k w Ą _k ՎូαĈࢦኑࢍზՎូ˟׶ՎូˬۡזᄱमࣃE(k)≤0.01Ă

׎̚E(k)ࢍზт˭Ĉ

E(k)= w_k₋₁−w_k (24) ౵ޢΞზ΍౵ָ۞ᝋࢦШณw Ą ^*

੊ቚԆјޢĂԧࣇΞӀϡѩ RBF ᙷৠགྷშྮ̶Ᏸ։ݡ ᄃ̙։ݡĂ༊ޞീͯ࠹ഴᇆည̝পᇈШณࠎX_inॡĂ૟ѩ পᇈࣃᏮˢᙷৠགྷშྮĂͽ(19)ёზ΍ᙷৠགྷშྮᏮ΍ࣃ

) ( _in

y X Ă༊Ꮾ΍ࣃy X( _in)≥5ॡࠎ։ݡĂ̝ͅĂ݋ࠎ̙։

ݡĄ

̣ă͎̇ณീᄃྤफ़ऱ

༊ޞീͯགྷᄃᇾ໤ͯᇆညͧ၆ᄮؠࠎ։ݡޢĂ΃ܑѩ ޞീͯ൑ܑࢬ༂ஷĂତ˭ֽυᅮᑭീѩޞീͯߏӎѣ͎̇

̙։༂ஷĄࢵАĂυᅮವԱޞീͯٙࢋณ͎̇࠹၆ᑕ̝ᙝ

ࠧĂӀϡ౵̂ୗޘ͞ёವᙝĂ x ׶ y ̶Ш̝ୗޘࣃࢍზт

˭Ĉ

x ̶Ш̝ୗޘࣃĈ

) , ( ) , 1 ( ) ,

(x y Ninx y Nin x y

diff_x = + − (25)

y ̶Ш̝ୗޘࣃĈ

) , ( ) 1 , ( ) ,

(x y Nin x y Ninx y

diff_y = + − (26)

ٙԱವ΍۞ᙝࠧቢҜཉ̚Ă૱૱Яࠎᗔੈ۞ᇆᜩົ൴Ϡѣ ಏᕇҜཉត̼ள૱۞ଐԛ൴ϠĂ˵ಶߏٙᏜ۞؝ϲᕇĂࠎ

˞ᔖҺѩଐڶ൴ϠĂ༊ѣᑭീז؝ϲᕇॡĂಶͽ؝ϲᕇ׌

य़۞ҜཉᕇҜཉ࠹ΐπӮޢ༊үາ۞ҜཉᙝࠧᕇĂ఍நԆ ޢĂޞീ͎ͯ̇ಶͽ׌ᙝ۞Ч࣎࠹၆ᑕ౵̂ត̼ณҜཉֽ

(7)

ܑ˘! FCM ̙Т m ࣃࢍზ΍̝V ̈́₁^* V ₂^* m

পᇈࣃ

=2 m

*

V 1 V ₂^*

=5 m

*

V 1 V ₂^*

=10 m

*

V 1 V ₂^* x 1 0.43 0.81 0.45 0.77 0.46 0.75 x 0.59 0.87 0.59 0.85 0.59 0.83 2

x 3 0.44 0.76 0.44 0.73 0.45 0.72 x 0.44 0.77 0.43 0.79 0.44 0.79 4

ܑ˟! ௚ࢍᄦ඀ăFCM ׶ RBF ᙷৠགྷშྮ̝Ᏸᙊத

͞ڱ ݭཱི

௚ࢍᄦ඀

Ᏸᙊத(%)

FCM Ᏸᙊத(%)

RBF ᙷৠགྷ Ᏸᙊத(%) I-14 92.5 97.5 99.5 16-PT 94.0 98.0 99.0 SM-5X 93.0 99.0 99.5 TO-263-1 95.0 98.0 100 T-2424-01 92.5 99.0 99.0

༊ү෼ᗓĂ༊ޞീͯܜᄃᆵ۞౵̈෼ᗓ෹΍ٙన۞̳म ॡĂಶࠎ̙։ݡ֭൴΍ᛋϯੈཱི̈́ࡔᐂ᏾ᄱᇆညĂ֭ޙϲ

͎̇ᑭീ̙̈́։ݡ׌ྤफ़ऱĄ

̱ă၁រᄃඕڍ

૟ώ͛ٙ൴ण͞ڱᑕϡٺ၁ᅫЧ჌̙Тఢॾ̝ IC ट ळĂ׎̚ CCD ᕜפᇆညቑಛ۞ᆵޘᄃ੼ޘࡗࠎ 4.9cmŰ 4.3cmĂည৵۞ྋژޘࡗࠎ 0.0845mmĂᇆညͧ၆ॡפ 200 ඊ੊ቚྤफ़Ăѩ੊ቚྤफ़ΒӣЧ჌̙Т։ݡᄃ̙։ݡ̝ᇆ ညĂᔘΒӣЧ჌̙Тݭཱི̝ IC टळĂͽീྏ̈́ͧྵ௚ࢍᄦ

඀ăFCM ׶ RBF ᙷৠགྷშྮඈˬ჌͞ڱ̝ᏰᙊதĂͽ FCM

͞ڱ੊ቚĂ༊Լត̙Т޽ᇴᝋࢦࣃm=2ă5ă10 ॡĂᄱम ࣃќᑦଐԛтဦ 14 ٙϯĂ҃ٙზ΍̝։ݡᄃ̙։ݡ౵ָপ ᇈШณཏ௡͕̚ᕇV ̈́₁^* V тܑ˘ٙϯĄଂဦ 14 ᄃܑ˘Ξ₂^*

࠻΍༊Լតᝋࢦࣃ m ॡ၆੊ቚ۞ќᑦిޘᄃќᑦࣃV ̈́₁^*

*

V ᇆᜩ̙̂Ąд RBF ᙷৠགྷშྮ੊ቚĂᏴፄ 81 ࣎ᔳᖟᆸ2

ৠགྷ̮Ăͽᅍਫ਼౵̈π͞ڱ੊ቚĂᄱमࣃќᑦଐԛтဦ 15

ٙϯĂଂဦ̚Ξ࠻΍ᝋࢦШณќᑦిޘܧ૱ԣĄ౵ޢĂ̶

Ҿ૟௚ࢍᄦ඀ăFCM ׶ RBF ᙷৠགྷშྮඈˬ჌͞ڱᑕϡ ٺݭཱི I-14ă16-PăSM-5XăTO-263-1 ׶ T-2424-01 ඈ 5

჌ᙷݭ IC टळĂീྏඊᇴЧפ 200 ඊĂീྏඕڍ̶̝Ᏸத тܑ˟ٙϯĂଂܑ˟̚Ξ࠻΍ RBF ᙷৠགྷშྮᏰᙊड़ڍ౵

ָĂᏰᙊதΞ྿ 99%ͽ˯ĂFCM ͞ڱ݋Ѩ̝ĂᏰᙊதࡗ 98%Ă҃௚ࢍᄦ඀Ᏸᙊத౵मࡗ 94%Ą௚ࢍᄦ඀͞ڱΪͽ ޞീͯ࠹ഴᇆည̝πӮѷลࣃ༊үᄦ඀გטቢĂटٽצಏ

͞ࢬ۞ࢨטĂٙͽड़ڍྵमć҃ FCM ׶ RBF ᙷৠགྷშྮ

͞ڱĂଂޞീͯ࠹ഴᇆည̚පפ 4 ჌̙Тপᇈࣃ༊ઇᏰᙊ

ֶፂĂЯѩᏰᙊड़ڍྵрĂ͍׎ߏ RBF ᙷৠགྷშྮ׍౯ܧ ቢّᙝ̶̝ࠧᙷਕ˧Ă߇ѣՀΐ̝Ᏸᙊड़ڍĄ౵ޢĂ૟ώ

͛ٙ൴ण̝ᇆညෛᛇր௚ᑕϡٺ၁ᅫΒ྅୧გᄦ඀Ϡயቢ

1.5

1

0.5

0

5 10 15 20 25 30 35 40 45 50 55 60 m=2

m=5 m=10

ဦ 14! FCM ޽ᇴᝋࢦࣃm=2ă5ă10 ॡ੊ቚќᑦଐԛ

0.35 0.3 0.25 0.2 0.15 0.1 0.05 0

0 20 40 60 80 100 120

ဦ 15! RBF ᙷৠགྷშྮ੊ቚќᑦଐԛ

ဦ 16! IC ୧გᄦ඀Ϡயቢϒࢬ၁វဦ

тဦ 16 ٙϯĂඕڍរᙋώ͛ٙ೩͞ڱ̝ѣड़ّĄ

˛ăඕ ኢ

ώ͛൴ण˘፟ୠෛᛇԫఙͷ၁ᅫᑕϡٺ IC Β྅୧გ

̝̙։ݡቢ˯ᑭീĂࡁտ͞ڱͽ̶߱ᇆည͞ёĂ੫၆ IC टळٺᄦ඀̚ΌౝٕΎ੓۞ԛېপᇈઇᑭീĂΩγࠎ˞ᔖ Һ IC Β྅୧გкᙷݭ۞តّ̼Ăͽᑭീ॰۞͞ёᏴפ୬ᑭ

(8)

߱ซҖĂௐ˘ล߱Ӏϡᇆညͧ၆͞ёĂགྷϤޞീۏᄃᇾ໤

ۏઇᇆည࠹ഴĂͽᑭ測 IC 容ळܑࢬ̝༂ஷݡĂώ͛೩΍௚

ࢍᄦ඀ăFuzzy C-mean 和 RBF ᙷৠགྷშྮˬ჌͞ڱĂֽᕩ ᙷѩ࠹ഴᇆညࠎ։ݡٕ̙։ݡĂ֭ͧྵѩˬ჌͞ڱ۞ᐹ КĂ၁រᙋ၁ RBF 類神經შྮѣྵָ̝Ᏸᙊड़ڍćௐ˟ล

߱੫၆ IC 容ळ̝পঅ͎̇үณീĂͽᑭീ͎̇࿅ٕ̂࿅̈

̝̙։ݡĂӀϡୗޘڱವᙝĂԱ΍׌ᙝࠧ౵̈෼ᗓĂͽҿ ؠߏӎ෹࿅ٙనؠ̝ᇾ໤͎̇ቑಛĄ౵ޢώࡁտ၁ᅫᑕϡ ٺ IC Β྅୧გᄦ඀၁ᅫϠயቢ̝ቢ˯ᑭീĂඕڍរᙋώ͛

ٙ೩͞ڱΞҖّĂ̙։ݡ̝ᏰᙊதΞ྿ 99%ͽ˯Ą

ᄫ ᔁ

ώࡁտຏᔁ઼ࡊົ૞ᗟࡁտࢍ൪ 92-2212-E-212-009

̝གྷ෱ᙒӄĄ

௑ཱི৶͔

img

li _ ᇾ໤ͯᄃޞീͯᇆညπӮѷลࣃ̝मளࣃ )

, (x y

f_st ᇾ໤ᇆညд(x,y)Ҝཉ۞ѷลࣃ )

, (x y

f_in ޞീᇆညд(x,y)Ҝཉ۞ѷลࣃ )

, (x y

Nin ޞീͯᇆည७ϒܪޘ̝ޢາ۞ѷลࣃ )

,

(c_x c_y ̚δ๪̝͋๪͕ळᇾ f 1 ࠹ഴᇆညπӮѷลࣃ

f 2 ࠹ഴᇆည࠹ഴ̙ࠎ࿬ᕇည৵࣎ᇴ f 3 ࠹ഴᇆည඗၆मࣃ

f 4 ࠹ഴᇆညፋវᇾ໤म x T

x, , ] [ ₁L ₄

=

X ࠹ഴᇆညපפ̝ϒఢ̼পᇈШณ 2

1 ,i ,

i =

V FCM։ݡᄃ̙։ݡপᇈШณཏะ̝͕̚ࣃ µki FCM ௐ k ඊ੊ቚྤफ़ᕩᛳٺௐ i ̶ᙷ۞ᕩ

ᛳבᇴࣃ

m FCM ̝޽ᇴᝋࢦࣃ

*

V ,1 V FCM ։ݡᄃ̙։ݡপᇈШณ̝౵ָཏะ₂^*

͕̚ࣃ )

1(X

φ ,φ₂(X),…, )φ_m( X ᔳᖟᆸ m ࣎ৠགྷ̮

T m] , , [ω₀ ω₁ ,ω

w= L RBF ᙷৠགྷშྮᏮ΍̝ᝋࢦࣃ t i ৠགྷ̮φ_i( X)̝͕̚Шณ

σi ৠགྷ̮φ_i( X)̝ᆵޘ p k Ꮾˢ۞Вតள৏ੱ

ણ҂͛ᚥ

1. Zhang, J. M., Lin, R. M., and Wang, M. J., “The Development of an Automatic Post-Sawing Inspection System Using Computer Vision Techniques,” Computers

Manufacturing, Vol. 13, pp. 366-373 (2000).

3. Sreenivasan, K. K., Srinath, M., and Khotanzad, A.,

“Automated Vision System for Inspection of IC Pads and Bonds,” IEEE Transactions on Components, Hybrids, Manufacturing Technology, Vol. 16, pp. 333-338 (1993).

4. Bartlett, S. L., Besl, P. J., Jian, R., Mukherjee, D., and Skifstad, K. D., “Automatic Solder Joint Inspection,”

IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 10, pp. 31-43 (1988).

5. Ikeuchhi, K., “Determining Surface Orientations of Specular Surfaces by Using the Photometric Stereo Method,” IEEE Trans. Pattern Anal. Machine Intell., vol.3, pp.661~669 (1981).

6. Nayar, S. K., Sandreson, A. C., Weiss, L. E., and Simson, D. D., “Specular Surface Inspection Using Structured Highlight and Gaussian Images,” IEEE Transactions on Robotics and Automation, Vol. 6, pp. 108-218 (1990).

7. Paithoon, K. and Khunkoey, S., “Image Classification by Kohonen Fuzzy C-mean,” Proceedings of the RESTECS’96, Tailand, pp. C70-75 (1996).

8. Hoppner, F., Klawonn, F., Kruse, R., and Runkler, T., Fuzzy Cluster Analysis. Wiley, New York, USA (2001).

9. Mukherjee, D. P., Pal, P., and Das, J., “Sodar Image Segmentation by Fuzzy C-Means,” Signal Processing, Vol.

54, pp. 295-301 (1996).

10. Khunkay, S., and Paith, K., “Image Segmentation by Fuzzy Rule and Kohonen-Constraint Satisfaction Fuzzy C-mean,” ICICS’97 Singapore, pp. 713-717 (1997).

11. Su, C. T., Yang, T., and Ke, C. M., “A Neural Network Approach for Semiconductor Wafer Post Sawing Inspec- tion,” IEEE Transactions on Semiconductor Manufacturing, Vol. 15, No. 2, pp. 260-266 (2002).

12. Sakurai, K., Onoyama, A., Fujii, T., Yamanishi, K. I., Fujii, S., and Morita, H., “Solution of Pattern Maching Inspection Problem for Grainy Metal Layers,” IEEE Transactions on Semiconductor Manufacturing, Vol. 15, No. 1, pp. 118-126 (2002).

13. Goodwin, G. C., and Sin, K. S., Adaptive Filtering Prediction and Control, Prentice-Hall, Engelwood Cliffs, NJ, USA (1984).

2003 ѐ 07 ͡ 30 ͟! ќቇ 2003 ѐ 08 ͡ 15 ͟! ܐᆶ 2003 ѐ 09 ͡ 24 ͟! ኑᆶ 2003 ѐ 10 ͡ 03 ͟! ତצ