JP3581040B2 - Wiring pattern inspection method - Google Patents

Description

【００３８】
である。図１（ｈ）は図１（ｂ）における濃淡画像信号１０を適切に設定された飛散検出用閾値ｔｈＨで２値化して得られる２値パターンＡ_Ｈを示すものであって、
【００３９】
【数２】

【００４０】
である。図１（ｊ）は図１（ｂ）における濃淡画像信号１０を適切に設定されたピンホール検出用閾値ｔｈＬで２値化して得られる２値パターンＡ_Ｌを示すものであって、
【００４１】
【数３】

【００４２】
である。
【００４３】
また、図１（ｉ）は図１（ｃ）における２次微分信号１１を適切に設定されたヘアラインショート検出用閾値ｔｈ２で２値化して得られる２値パターンＡ_２を示すものであって、２次微分値をＩ”_ａ，Ｉ”_ｂ，Ｉ”_ｃ，Ｉ”_ｄとして、
【００４４】
【数４】

【００４５】
である。ここで、２次微分値Ｉ”_ａは図１（ｂ）に示した濃淡画像信号１０に図２（ａ）に示す画素単位で５×５の２次微分オペレータを作用させることによって得られるものであり、以下、２次微分値をＩ”_ｂ，Ｉ”_ｃ，Ｉ”_ｄは夫々、濃淡画像信号１０に図２（ｂ），（ｃ），（ｄ）に示す画素単位で５×５の２次微分オペレータを作用させることによって得られるものである。図１（ｋ）は図１（ｃ）における２次微分信号１１を適切に設定されたヘアラインショート検出用閾値ｔｈ３で２値化して得られる２値パターンＡ_３を示すものであって、
【００４６】
【数５】

【００４７】
である。
【００４８】
なお、図２（ａ）に示す２次微分オペレータは水平方向に沿うパターンエッジを顕在化するものであり、図２（ｂ）に示す２次微分オペレータは右上がり方向に沿うパターンエッジを顕在化するものであり、図２（ｃ）に示す２次微分オペレータは垂直方向に沿うパターンエッジを顕在化するものであり、図２（ｄ）に示す２次微分オペレータは右下がり方向に沿うパターンエッジを顕在化するものである。即ち、これら２次微分オペレータは入力される濃淡画像のエッジ部分を顕在化するものであり、このエッジ部が急峻なほど２次微分値Ｉ”_ａ，Ｉ”_ｂ，Ｉ”_ｃ，Ｉ”_ｄは大きくなる。従って、図１（ｃ）に示すように、ヘアライン断線２やヘアラインショート４の部分の２次微分値はピンホール３や飛散５の部分の２次微分値よりも大きくなる。
【００４９】
ところで、上記数１〜５による２値化を行なうと、基準閾値による２値パターンＡ_１は、図１（ｄ）に示すように、パターンエッジを境として、配線パターン１の部分で“１”，基材部６の部分で“０”となる。つまり、パターンエッジでレベルが反転する信号である。
【００５０】
濃淡画像１０の飛散検出用閾値ｔｈＨによる２値パターンＡ_Ｈは、図１（ｈ）に示すように、基材部６に存在するヘアラインショート４や飛散５といった欠陥に対して“１”となり、飛散検出用閾値ｔｈＨを適切に設定することにより、飛散５に対する信号成分はこの飛散５の幅の実寸法に応じた時間幅の信号として含まれる。しかし、この２値パターンＡ_Ｈには、かかる飛散５に対する信号成分ばかりでなく、ヘアラインショート４に対する信号成分も含まれる場合もあるし、また、配線パターン１の部分では、常時“１”である。
【００５１】
また、濃淡画像１０のピンホール検出用閾値ｔｈＬによる２値パターンＡ_Ｌは、図１（ｊ）に示すように、配線パターン１に存在するヘアライン断線２やピンホール３といった欠陥に対して“０”となり、ピンホール検出用閾値ｔｈＬを適切に設定することにより、ピンホール３に対する信号成分はこのピンホール３の幅の実寸法に応じた時間幅の信号として含まれる。基材部６の部分では、２値パターンＡ_Ｌは常時“０”となる。
【００５２】
２次微分パターン１１のヘアラインショート検出用閾値ｔｈ２による微分２値パターンＡ_２は、図１（ｉ）に示すように、ヘアラインショート検出用閾値ｔｈ２を適切に設定することにより、ヘアラインショート４に対する信号成分がこの欠陥の幅の実寸法に応じた時間幅の“１”の信号として含まれるのであるが、エッジ部分の急峻さに応じて、ピンホール３や飛散５のエッジ部や配線パターン１のエッジ部なども含まれることになる。
【００５３】
２次微分パターン１１のヘアライン断線検出用閾値ｔｈ３による微分２値パターンＡ_２は、図１（ｋ）に示すように、ヘアライン断線検出用閾値ｔｈ３を適切に設定することにより、ヘアライン断線２に対する信号成分がこの欠陥の幅の実寸法に応じた時間幅の“０”の信号として含まれるのであるが、エッジ部分の急峻さに応じて、ピンホール３や配線パターン１などのエッジ部も含まれることになる。
【００５４】
さて、以上のような２値パターンにおいて、２値パターンＡ_Ｌ，Ａ_３を論理積処理することにより、ヘアライン断線２及びピンホール３の夫々の欠陥に対するそれらの実寸法に応じた時間幅の信号成分を含む２値パターン（Ａ_Ｌ・Ａ_３）が得られる。同様にして、２値パターンＡ_Ｈ，Ａ_２を論理積処理することにより、ヘアラインショート４及び飛散５の夫々の欠陥に対するそれらの実寸法に応じた時間幅の信号成分を含む２値パターン（Ａ_Ｈ＋Ａ_２）が得られる。ここで、・は論理積、＋は論理和を表わす。
【００５５】
ところで、上記２値パターン（Ａ_Ｌ・Ａ_３）についてみると、ヘアライン断線２及びピンホール３は配線パターン１の領域に存在する欠陥であるから、これら欠陥の信号成分を得るためには、この２値パターン（Ａ_Ｌ・Ａ_３）のうちの配線パターン１の領域の部分のみ抽出する必要がある。同様にして、上記２値パターン（Ａ_Ｈ＋Ａ_２）についてみると、ヘアラインショート４及び飛散５は基材部６の領域に存在する欠陥であるから、これら欠陥の信号成分を得るためには、この２値パターン（Ａ_Ｈ＋Ａ_２）のうちの基材部６の領域の部分のみ抽出する必要がある。なお、配線パターン１のエッジ部は、上記の各閾値ｔｈ１，ｔｈＨ，ｔｈＬ，ｔｈ２，ｔｈ３による２値化で検出されるが、欠陥ではない。また、閾値ｔｈ１，ｔｈＨ，ｔｈＬ，ｔｈ２，ｔｈ３による２値化では、この配線パターン１のエッジ部は正しく検出されない。即ち、基本閾値ｔｈ１が配線パターン１のエッジ部を検出するためのものであり、この基本閾値ｔｈ１を適切に設定することにより、配線パターン１のエッジ部が正しい位置で検出されることになる。図１（ｄ）に示す２値パターンＡ_１のレベル反転部が配線パターン１のエッジ部を表わしていることになる。
【００５６】
以上のことからして、ヘアライン断線２及びピンホール３は２値パターン（Ａ_Ｌ・Ａ_３）から、ヘアラインショート４及び飛散５は２値パターン（Ａ_Ｈ＋Ａ_２）から、配線パターン１のエッジ部は２値パターンＡ_１から夫々得られるようにする必要がある。
【００５７】
そこで、この実施形態では、さらに、信号の時間軸方向に、エッジ部を除いた配線パターン１の領域とこの配線パターン１のエッジ部を含むエッジ領域と基材部６の領域とに領域分割する。図１（ｅ）は配線パターン１の領域を表わす２値パターンＡ_ＮＬ（なお、この配線パターン１の領域も配線パターン領域Ａ_ＮＬという）を、図１（ｆ）は配線パターン１のエッジ部の領域を表わす２値パターンＡ_ＮＭ（なお、このエッジ部の領域もエッジ領域Ａ_ＮＭという）を、図１（ｇ）は基材部６の領域を表わす２値パターンＡ_ＮＨ（なお、この基材部６の領域も基材部領域Ａ_ＮＨという）を夫々表わしている。かかる領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨは、後述する領域分割オペレータでの着目画素の周囲の“１”の有効画素数をＮ、後述するように判定値をＮ_Ｌ，Ｎ_Ｈとして、次の数６で表わされる。
【００５８】
【数６】

【００５９】
但し、
Ａ_ＮＬ＋Ａ_ＮＭ＋Ａ_ＮＨ＝１
である。
【００６０】
そして、配線パターン領域でヘアライン断線２及びピンホール３の検出を行なうために、２値パターン（Ａ_Ｌ・Ａ_３），Ａ_ＮＬの演算処理、即ち、
Ａ_ＮＬ・（Ａ_Ｌ・Ａ_３）
を行ない、基材部６の領域でヘアラインライン４及び飛散５の検出を行なうために、２値パターン（Ａ_Ｈ＋Ａ_２），Ａ_ＮＨの演算処理、即ち、
Ａ_ＮＨ・（Ａ_Ｈ＋Ａ_２）
を行なう。また、エッジ領域で配線パターン１のエッジ部の検出を行なうために
、 Ａ_ＮＭ・（Ａ_１・Ａ_３＋Ａ_１・Ａ_２）
の処理を行なう。なお、上記演算式において、Ａ_１・Ａ_３はこのエッジ領域でのレベルと配線パターン領域でのＡ_ＮＬ・（Ａ_Ｌ・Ａ_３）のレベルとを揃えるための演算であり、また、Ａ_１・Ａ_２は２値パターンＡ１が図示とはレベル反転しているときのエッジ領域でのレベルと基材部領域でのＡ_ＮＨ・（Ａ_Ｈ＋Ａ_２）のレベルとを揃えるための演算である。
【００６１】
従って、以上の２値化処理によって得られる２値化出力Ａ_Ｒは、
【００６２】
【数７】

【００６３】
となる。この出力Ａ_Ｒを図１（ｌ）に示す。
【００６４】
次に、上記の領域分割について説明する。
【００６５】
ここで、このような配線パターン領域Ａ_ＮＬ，エッジ領域Ａ_ＮＭ，基材部領域Ａ_ＮＨへの分割は、図１（ｄ）に示す基本閾値ｔｈ１による２値パターンＡ_１へ領域分割オペレータを作用させ、２値パターンＡ_１での着目画素がこれら領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨのいずれに属しているかを判定するものである。この領域分割オペレータは、２値パターンＡ_１での画素の２次元配列に作用させるものであって、着目画素が判定基準となる周囲の有効画素にどのような状態で囲まれているかに応じて、この着目画素が領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨのいずれに属しているかを判定するものである。
【００６６】
図３は領域分割オペレータの具体例を示すものである。かかる具体例は、有効画素を着目画素Ｐを中心とする円周上の画素Ｂとするものであって、これら有効画素Ｂが“１”であるか否かを判定して、“１”の有効画素Ｂの個数Ｎをカウントし、このカウント値Ｎに応じてこの着目画素Ｐが領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨのいずれに属しているかを判定するものである。
【００６７】
ここで、周囲画素Ｂが存在する円周の直径が画素単位でｎ画素であるとき、この領域分割オペレータをサイズｎ×ｎのオペレータという。このオペレータサイズは配線パターン１の線幅やこの配線パターンのエッジ幅に応じて設定されるものであって、異なる線幅やエッジ幅の配線パターンのプリント基板を検査する検査装置の場合には、各プリント基板毎に最適なサイズの領域分割オペレータが予め設けられており、検査するプリント基板に応じて使用する領域分割オペレータを異ならせる。
【００６８】
図３には、サイズが７×７，９×９及び１１×１１の領域分割オペレータを示している。サイズ７×７の領域分割オペレータでは、有効画素Ｂは１６画素であり、この画素数に対して２つの判定値Ｎ_Ｌ＝１４，Ｎ_Ｈ＝２が設定されている。また、サイズ９×９の領域分割オペレータでは、有効画素Ｂは２４画素であって、この画素数に対して２つの判定値Ｎ_Ｌ＝２２，Ｎ_Ｈ＝２が設定され、サイズ１１×１１の領域分割オペレータでは、有効画素Ｂは２８画素であり、この画素数に対して２つの判定値Ｎ_Ｌ＝２５，Ｎ_Ｈ＝３が設定されている。
【００６９】
このような領域分割オペレータの判定値Ｎ_Ｌ，Ｎ_Ｈと上記の“１”の有効画素Ｂの個数Ｎとから、上記数６に基づいて着目画素Ｐの属する領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨを判定するものである。
【００７０】
ところで、上記のような欠陥が発生しないとするならば、有効画素Ｂが全て“１”のとき、着目画素Ｐは配線パターンの領域にあると判定でき、また、有効画素Ｂが全て“０”のとき、着目画素Ｐは基材部の領域にあると判定でき、さらに、一部の有効画素Ｂが“１”での凝りの有効画素Ｂが“０”のとき、着目画素Ｐは配線パターンのエッジ部の領域に属すると判定することができる。このため、例えば、サイズ７×７の領域分割オペレータにおいて、Ｎ_Ｌ＝１６，Ｎ_Ｈ＝０とすることも考えられる。
【００７１】
しかし、このようにすると、例えば、サイズ７×７の領域分割オペレータにおいて、配線パターン領域Ａ_ＮＬで有効画素Ｂが並んだ円周を１画素以下の幅のヘアライン断線が横切る場合、２個の有効画素が“０”となって“１”の有効画素Ｂの個数Ｎ＝１４となり、Ｎ＜Ｎ_Ｌであるから、配線パターン領域Ａ_ＮＬであるにもかかわらず、配線パターンのエッジ部の領域Ａ_ＮＭと誤った判定をすることになる。このため、図３では、判定値Ｎ_Ｌを有効画素Ｂの総数よりもヘアライン断線の幅分少ない値としているのである。同様にして、判定値Ｎ_Ｈにも、ヘアラインショートの幅分の値を持たせ、ヘアラインショートがあっても、これにより、基材部の領域Ａ_ＮＨを配線パターンのエッジ部の領域Ａ_ＮＭと誤って判定することがないようにしている。
【００７２】
そして、このように判定値Ｎ_Ｌ，Ｎ_Ｈを設定して正しく配線パターン，そのエッジ部，基材部の各領域を判定できることにより、配線パターンの領域Ａ_ＮＬでのヘアライン断線を、また、基材部の領域Ａ_ＮＨでのヘアラインショートを検出できるようになる。
【００７３】
なお、ヘアライン断線やヘアラインショートの幅は、領域分割オペレータのサイズに応じて異ならせてもよく、その領域分割オペレータを使用する配線パターンでのかかる欠陥の幅がどの程度となるかに応じて決定されるものである。
【００７４】
また、ここでは、有効画素Ｂを着目画素Ｐに対して等方的な配置となる円周上に配置されたものとしたが、これのみに限定されるものではなく、例えば、着目画素Ｐを中心とする円内の全ての画素や着目画素を中心とする正方形内の全ての画素など、被検査物の図形的性質や仕様などに応じて適宜その範囲を変更可能である。
【００７５】
次に、領域分割オペレータと配線パターンの線幅との関係について、図７を用いて説明する。
【００７６】
図７（ａ）は配線パターン１の線幅（ここでは、この線幅を５画素分とする）以下のサイズの領域分割オペレータ（ここでは、オペレータサイズ５×５とする）がこの配線パターン１を、図面上、右方向に横切る動作を示しており、配線パターン１に対する領域分割オペレータの位置が（ｉ），（ｉｉ），（ｉｉｉ）の順に変化していくものとする。
【００７７】
（ｉ）の状態は着目画素Ｐの周りの有効画素Ｂが全て配線パターン１の領域内にあり、このときの有効画素Ｂは全て“１”である。ここで、全有効画素Ｂの個数に対する“１”の有効画素Ｂの個数Ｎの比をαとすると、この（ｉ）の状態では、α＝１２／１２である。次に、配線パターン１に対して領域分割オペレータが２画素分右方に移動した（ｉｉ）の状態では、右側の５個の有効画素Ｂが配線パターン１からはみ出して“０”となり、このときには、比α＝７／１２となる。さらに、配線パターン１に対して領域分割オペレータが１画素分右方に移動した（ｉｉｉ）の状態になると、右側の７個の有効画素Ｂが配線パターン１からはみ出して“０”となり、このときには、比α＝５／１２となる。
【００７８】
このように、オペレータサイズが配線パターン１の線幅よりも小さいときには、領域分割オペレータが配線パターン１内にあるとき、比α＝１となって最大となり、領域分割オペレータが配線パターン１から外れていくにつれてαが小さくなっていく。
【００７９】
これに対し、図７（ｂ）はオペレータサイズが配線パターン１の線幅（ここでも、この線幅を５画素分としている）よりも大きい場合（ここでは、オペレータサイズ７×７とする）を示すものである。この場合も、図７（ａ）の場合と同様、領域分割オペレータがこの配線パターン１を、図面上、右方向に横切る動作を示しており、配線パターン１に対する領域分割オペレータの位置が（イ），（ロ），（ハ）の順に変化していくものとする。
【００８０】
（イ）の状態は着目画素Ｐが配線パターン１の幅方向中心部に位置しているものとしている。この場合には、５画素の線幅に対してオペレータサイズが７×７であるから、左右両側３個ずつの有効画素Ｂが配線パターン１の領域からはみ出して“０”である。従って、この（イ）の状態では、α＝１０／１２である。次に、配線パターン１に対して領域分割オペレータが１画素分右方に移動した（ロ）の状態では、右側の５個の有効画素Ｂのみが配線パターン１からはみ出して“０”となり、このときには、比α＝１１／１２となる。さらに、配線パターン１に対して領域分割オペレータが１画素分右方に移動した（ハ）の状態になると、右側の７個の有効画素Ｂが配線パターン１からはみ出して“０”となり、このときには、比α＝９／１２となる。
【００８１】
このように、オペレータサイズが配線パターン１の線幅よりも大きいときには、着目画素Ｐが配線パターン１の幅方向中心部にあるよりも、領域分割オペレータが配線パターン１から一方に若干片寄った方が比αが大きくなり、しかる後、領域分割オペレータが配線パターン１から外れていくにつれてαが小さくなっていく。
【００８２】
以上のことからして、オペレータサイズが配線パターン１の線幅以下のときには、領域分割オペレータがこの配線パターン１内にあるとき、比α＝１となって最大となり、領域オペレータが配線パターン１から外れていくにつれてαは単調に小さくなっていくが、オペレータサイズが配線パターン１の線幅よりも大きいと、着目画素Ｐが配線パターン１の幅方向中心部にあるよりも、これよりも若干ずれた方がαが大きくなり、領域分割オペレータと配線パターンとの位置関係とαとの関係が不自然なものとなる。
【００８３】
従って、領域分割オペレータのサイズとしては、プリント基板上での配線パターン１の最小線幅以下であることが必要である。これにより、検査対象となるプリント基板上での配線パターン１の最小線幅に応じて、使用可能な領域分割オペレータの最大のサイズが決まることになる。
【００８４】
図８は検査対象となるプリント基板上での配線パターン１に対する使用可能な領域分割オペレータの最小のサイズを説明した図である。
【００８５】
配線パターン１の濃淡画像信号１０のエッジ部は、パターンのにじみにより、図８に示すように、明るさ（濃淡値）が連続して変化していく傾斜した波形の有限の期間を有している。ここで、このエッジ部において、ピンホール検出用閾値ｔｈＬの明るさから飛散検出用閾値ｔｈＨの明るさまでの明るさ範囲に含まれる明るさの画素数を、以下、明かるさ変化画素数という。
【００８６】
図８（ａ）はオペレータサイズ＞明るさ変化画素数の場合のエッジ部近傍での２値パターンを示すものであって、この場合には、領域分割オペレータで決まるエッジ部の領域Ａ_ＮＭはこの濃淡画像信号１０のエッジ部全体を含むため、このこのエッジ部の領域Ａ_ＮＭ 内にのみ基本閾値ｔｈ１で得られた２値パターンＡ_１のエッジＥ_０が得られることになる。
【００８７】
これに対し、図８（ｂ）に示すように、オペレータサイズ＜明るさ変化画素数の場合には、領域分割オペレータで決まるエッジ部の領域Ａ_ＮＭは濃淡画像信号１０のエッジ部の範囲よりも狭くなり、領域分割オペレータで決まる配線パターン１の領域Ａ_ＮＬも基材部６の領域Ａ_ＮＨもこのエッジ部の範囲に入り込んでくる。このため、勿論領域Ａ_ＮＭで基本閾値ｔｈ１で得られた２値パターンＡ_１のエッジＥ_０が得られるが、領域Ａ_ＮＬでも、ピンホール検出用閾値ｔｈＬで得られる２値パターンＡ_Ｌでこのエッジ部に対するエッジＥ_Ｌが、また、領域Ａ_ＮＨでも、飛散検出用閾値ｔｈＨで得られる２値パターンＡ_Ｈでこのエッジ部に対するエッジＥ_Ｈが夫々発生することになる。
【００８８】
このように、オペレータサイズ＜明るさ変化画素数にすると、領域Ａ_ＮＬ，Ａ_ＮＭ，Ａ_ＮＨの全てに配線パターン１での同じエッジ部が誤って検出されるという不自然な事態が発生することになる。従って、領域分割オペレータのサイズは、配線パターン１のエッジ部の幅以上とすることが必要となる。
【００８９】
なお、領域分割オペレータはサイズを７×７，９×９，１１×１１と離散的な値しかとれないが、有効画素数のカウントに対して閾値Ｎ_Ｈ、Ｎ_Ｌを設けることにより、かかる離散的な実際のサイズの間のサイズの領域分割オペレータを実現することができる。即ち、閾値Ｎ_Ｈ，Ｎ_Ｌを近づけると、オペレータサイズが小さくなることと同様な効果が得られるのである。
【００９０】
この実施形態では、上記のようにして、低い閾値ｔｈＬでピンホール欠陥を、高い閾値ｔｈＨで飛散欠陥を夫々顕在化し、また、ヘアライン断線やヘアラインショートは専ら２次微分閾値により検出するものであり、この結果、上記数７で示す２値化出力Ａ_Ｒが得られるのであるが、上記各閾値ｔｈ１，ｔｈＨ，ｔｈＬ，ｔｈ２，ｔｈ３の設定は、図６に示す欠陥検査装置の場合、２値化手段１０１，１０２とこれによって得られる上記数７で表わされる２値化画像を表示する画像表示手段１３１，１３２を用いて行なわれる。
【００９１】
なお、これら２値化手段１０１，１０２では、上記閾値を適宜設定することにより、２値化機能の切り替えを可能となる。即ち、２次微分併用の２値化とするためには、閾値の設定をｔｈ１＝ｔｈＨ＝ｔｈＬとすればよい。これにより、数１，２，３からＡ_１＝Ａ_Ｈ＝Ａ_Ｌとなるので、上記数７に示す２値化出力信号Ａ_Ｒは、

となり、従来の２次微分を併用した２値化を実現できる。但し、この場合の上記数７では、Ａ_ＮＭ・Ａ_１・Ａ_２はＡ_ＮＭ＝１（パターンエッジ領域）かつＡ_１＝０の領域での閾値ｔｈ２による２次微分２値画像を表わすものであり、また、Ａ_ＮＨ・Ａ_３・Ａ_２はＡ_ＮＨ＝１の領域であるが、この領域ではＡ_３＝１であり、Ａ_１＝１と同等である。従って、この場合の数７における（Ａ_ＮＭ・Ａ_１・Ａ_２＋Ａ_ＮＨ・Ａ_３・Ａ_２）はＡ_１・Ａ_２となり、上記の２値化出力信号Ａ_Ｒ が得られることになる。
【００９２】
一方、２次微分２値化を用いないで３閾値２値化として機能させるためには、２次微分画像Ｉ”の閾値をｔｈ２＝∞、ｔｈ３＝−∞とすればよく、上記数４により、Ａ_２＝０、また、上記数５により、Ａ_３＝１となるため、これを上記数７に代入すると、２値化出力信号Ａ_Ｒは、
Ａ_Ｒ＝Ａ_ＮＬ・Ａ_Ｌ＋Ａ_ＮＭ・Ａ_１＋Ａ_ＮＨ・Ａ_Ｈ
となり、３閾値２値化パターンが得られることになる。
【００９３】
さらに、単純２値化（閾値ｔｈ１のみによる２値化）として機能させる場合には、ｔｈ１＝ｔｈＨ＝ｔｈＬ，ｔｈ２＝∞，ｔｈ３＝−∞とすればよい。これにより、Ａ_１＝Ａ_Ｈ＝Ａ_Ｌ、Ａ_２＝０，Ａ_３＝１となり、かつＡ_ＮＬ＋Ａ_ＮＭ＋Ａ_ＮＨ＝１であるから、これを上記数７に代入すると、２値化出力信号Ａ_Ｒは、
Ａ_Ｒ＝Ａ_１
となる。
【００９４】
このように、この実施形態では、各種欠陥の検出に対応することが可能であると共に、柔軟に機能の切り替えを行なうことができる。
【００９５】
次に、パターンを精度良く検出するための２値化閾値ｔｈ１，ｔｈ２，ｔｈ３，ｔｈＨ，ｔｈＬの設定方法について説明する。
【００９６】
この設定のためには、各欠陥モード（即ち、ヘアライン断線やヘアラインショートなどの欠陥の種類）の限界欠陥サンプルを基準として用いる。基準となるサンプルは、実欠陥でも、模擬欠陥でもよい。即ち、検出すべき最小幅を有するヘアライン断線やヘアラインショート，最小導体間隔となる半ショート，最小線幅を有する半断線欠陥，最小検出サイズのピンホールや飛散欠陥などである。
【００９７】
図９は図６の膨張系と収縮系での各２値化閾値の設定基準の一具体例を示すものである。
【００９８】
図６の膨張系での各閾値を設定する場合、パターン検出手段１００から上記のサンプルを持つ濃淡画像を２値化手段１０１に供給し、ここで得られる２値化パターンＡ_Ｒ を画像表示手段１３１で表示させ、閾値を調整することにより、最適な閾値に設定するものである。以下、その設定方法を図９を用いて説明する。
【００９９】
膨張系の場合、パターン検出用の基準閾値ｔｈ１は、検出対象をパターンのエッジ部と半ショート（導体パターン間隔が規定される最小間隔よりも狭くなっている欠陥：図４を参照）とするものであって、この基準閾値ｔｈ１で得られる２値化パターンを、先に説明したように、規定幅だけ膨張したとき、この半ショートが必ずショートするような閾値とするものであり、また、飛散検出用の閾値ｔｈＨを決定する際の基準となるものである。従って、濃淡画像をこの基準閾値ｔｈ１で２値化して得られる２値化パターンの規定される最小導体間隔（欠陥ではない）が実寸通りに検出されるように、この閾値ｔｈ１を設定する必要がある。このことは、半ショートがなければ、導体パターンの間隔が規定の最小の場所でも、膨張したときにショートが生じないことを意味している。このように基準閾値ｔｈ１を設定するためには、ｔｈ１＝ｔｈＨ＝ｔｈＬ，ｔｈ２＝∞，ｔｈ３＝−∞とし、かつ最小導体間隔の濃淡画像をサンプルとして使用して、このサンプルから得られる上記数７の２値化出力信号Ａ_ＲをＡ_Ｒ＝Ａ_１として画像表示手段１３１で表示しながら基準閾値ｔｈ１を調整する。
【０１００】
膨張系の場合、閾値ｔｈ２は、２次微分画像を２値化するものであって、余剰系欠陥としてのヘアラインショートや孤立点を検査対象欠陥とするものであり、検出すべき最小サイズのかかる欠陥が実寸で検出されるように設定されるものであるが、このためには、かかるサイズの欠陥を持つサンプルの濃淡画像を得、これを上記のように基準閾値ｔｈ１で２値化して２値化パターンＡ_１を得るとともに、この濃淡画像の２次微分画像を得、これを閾値ｔｈ２で２値化してかかる最小サイズのかかる欠陥が実寸で検出されるように、この閾値ｔｈ２を調整するものである。この場合、検査対象欠陥が抽出されて画像表示手段１３１で表示されるように、２次微分画像の閾値ｔｈ２による２値画像のうちの２値化パターンＡ_１の“０”の領域だけを抽出して画像表示手段１３１に供給するようにする。このために、ｔｈ１＝ｔｈＨ＝ｔｈＬ，ｔｈ３＝−∞とすることにより、Ａ_１＝Ａ_Ｈ＝Ａ_Ｌ＝０（従って、Ａ_１＝１），かつＡ_３＝１とし、２値化出力信号Ａ_Ｒを、上記数７から、
Ａ_Ｒ＝Ａ_ＮＭ・Ａ_１・Ａ_２＋Ａ_ＮＨ・Ａ_２＝（Ａ_ＮＭ・Ａ_１＋Ａ_ＮＨ）・Ａ_２
とするものである。領域（Ａ_ＮＭ・Ａ_１＋Ａ_ＮＨ）は、図１から明らかなように、Ａ_１＝１の領域である。
【０１０１】
膨張系の場合、閾値ｔｈＨは、濃淡画像を２値化して余剰系欠陥としての飛散を検出するためのものであって、その２値化画像で検出する最小サイズのかかる欠陥が実寸で検出されるように設定される。この場合、検査対象欠陥が抽出されて画像表示手段１３１で表示されるように、濃淡画像の閾値ｔｈＨによる２値画像のうちの２値化パターンＡ_１の“０”の領域だけを抽出して画像表示手段１３１に供給するようにする。このために、ｔｈ１＝ｔｈＬ，ｔｈ２＝∞，ｔｈ３＝−∞とすることにより、Ａ_１＝Ａ_Ｌ＝Ａ_２＝０（従って、Ａ_１＝１），かつＡ_３＝１とし、２値化出力信号Ａ_Ｒを、上記数７から、
Ａ_Ｒ＝Ａ_ＮＨ・Ａ_Ｈ
とするものである。領域Ａ_ＮＨは、図１から明らかなように、パターンエッジ部を除いたＡ_１＝１の領域である。
【０１０２】
閾値ｔｈ３は導体パターン中の不足系欠陥であるヘアライン断線を検出するためのものであり、閾値ｔｈＬは導体パターン中の不足系欠陥であるピンホールを検出するためのものであるから、膨張系では、これら閾値ｔｈ３，ｔｈＬは使用されず、一例として収縮系と同じ値に設定される。
【０１０３】
次に、図６の収縮系での各閾値を設定する場合、パターン検出手段１００からサンプルを持つ濃淡画像を２値化手段１０２に供給し、ここで得られる２値化パターンＡ_Ｒを画像表示手段１３２で表示させ、閾値を調整することにより、最適な閾値に設定するものである。以下、その設定方法を図９を用いて説明する。
【０１０４】
収縮系の場合、パターン検出用の基準閾値ｔｈ１は、検出対象をパターンのエッジ部と半断線（導体パターンの幅が規定される最小幅よりも狭くなっている欠陥：図４を参照）とするものであって、この基準閾値ｔｈ１で得られる２値化パターンを、先に説明したように、規定幅だけ縮小したとき、この半断線が必ず断線するような閾値とするものであり、また、ピンホール検出用の閾値ｔｈＬとを決定する際の基準となるものである。従って、濃淡画像をこの基準閾値ｔｈ１で２値化して得られる２値化パターンの規定される最小導体幅（欠陥ではない）が実寸通りに検出されるように、この閾値ｔｈ１を設定する必要がある。このことは、半断線がなければ、導体パターンの幅が規定の最小の場所でも、収縮したときに断線が生じないことを意味している。このように基準閾値ｔｈ１を設定するためには、ｔｈ１＝ｔｈＨ＝ｔｈＬ，ｔｈ２＝∞，ｔｈ３＝−∞とし、かつ最小導体幅の濃淡画像をサンプルとして使用して、このサンプルから得られる上記数７の２値化出力信号Ａ_ＲをＡ_Ｒ＝Ａ_１として画像表示手段１３１で表示しながら基準閾値ｔｈ１を調整する。
【０１０５】
収縮系の場合、閾値ｔｈ３は、２次微分画像を２値化するものであって、不足系欠陥としてのヘアライン断線を検査対象欠陥とするものであり、検出すべき最小サイズのかかる欠陥が実寸で検出されるように設定されるものであるが、このためには、かかるサイズの欠陥を持つサンプルの濃淡画像を得、これを上記のように基準閾値ｔｈ１で２値化して２値化パターンＡ_１を得るとともに、この濃淡画像の２次微分画像を得、これを閾値ｔｈ３で２値化してかかる最小サイズのかかる欠陥が実寸で検出されるように、この閾値ｔｈ３を調整するものである。この場合、検査対象欠陥が抽出されて画像表示手段１３２で表示されるように、２次微分画像の閾値ｔｈ３による２値画像のうちの２値化パターンＡ_１の“１”の領域だけを抽出して画像表示手段１３２に供給するようにする。このために、ｔｈ１＝ｔｈＨ＝ｔｈＬ，ｔｈ２＝∞とすることにより、Ａ_１＝Ａ_Ｈ＝Ａ_Ｌ＝１（従って、Ａ_１＝０），かつＡ_２＝０とし、２値化出力信号Ａ_Ｒを、上記数７から、
Ａ_Ｒ＝Ａ_ＮＬ・Ａ_３＋Ａ_ＮＭ・Ａ_３＝（Ａ_ＮＬ＋Ａ_ＮＭ）・Ａ_３
（但し、Ａ_ＮＨ＝１の領域でＡ_１＝Ａ_Ｈ＝０であるから、Ａ_ＮＨ・Ａ_３・Ａ_Ｈ＝０）
とするものである。領域（Ａ_ＮＬ＋Ａ_ＮＭ）は、図１から明らかなように、Ａ_１＝１の領域である。
【０１０６】
収縮系の場合、閾値ｔｈＬは、濃淡画像を２値化して不足系欠陥としてのピンホールを検出するためのものであって、その２値化画像で検出する最小サイズのかかる欠陥が実寸で検出されるように設定される。この場合、検査対象欠陥が抽出されて画像表示手段１３２で表示されるように、濃淡画像の閾値ｔｈＬによる２値画像のうちの２値化パターンＡ_１の“１”の領域だけを抽出して画像表示手段１３２に供給するようにする。このために、ｔｈ１＝ｔｈＨ，ｔｈ２＝∞，ｔｈ３＝−∞とすることにより、Ａ_１＝Ａ_Ｈ＝１（従って、Ａ_１＝０），かつＡ_２＝０，Ａ_３＝１とし、２値化出力信号Ａ_Ｒを、上記数７から、
Ａ_Ｒ＝Ａ_ＮＬ・Ａ_Ｌ＋Ａ_ＮＭ・Ａ_１
（但し、Ａ_ＮＨ＝１の領域でＡ_１＝Ａ_Ｈ＝０であるから、Ａ_ＮＨ・Ａ_３・Ａ_Ｈ＝０）
とするものである。領域Ａ_ＮＬは、図１から明らかなように、パターンエッジ部を除いたＡ_１＝１の領域であり、また、分割領域Ａ_ＮＭ（即ち、パターンエッジ部）でのＡ_１＝１の領域である。
【０１０７】
閾値ｔｈ２は導体パターン中の余剰系欠陥であるヘアラインショートを検出するためのものであり、閾値ｔｈＨは導体パターン中の余剰系欠陥である飛散を検出するためのものであるから、収縮系では、これら閾値ｔｈ２，ｔｈＨは使用されず、一例として膨張系と同じ値に設定される。
【０１０８】
以上のようにして、図６の膨張系では、２値化手段１０１と画像表示手段１３１とを用いて、基準閾値ｔｈ１とヘアラインショート／孤立点検出用の閾値ｔｈ２と飛散検出用の閾値ｔｈＨとが設定され、また、図６の収縮系では、２値化手段１０２と画像表示手段１３２とを用いて、基準閾値ｔｈ１とヘアライン断線検出用の閾値ｔｈ２とピンホール検出用の閾値ｔｈＬとが設定される。
【０１０９】
以上の設定に際しては、２値化閾値をパラメータとして、対象欠陥の検出寸法の関係を求めるとよい。図１０にピンホール検出用の閾値ｔｈＬと飛散検出用の閾値ｔｈＨの設定例を示す。
【０１１０】
図１０（ａ）は直径２５μｍのピンホール欠陥に対して閾値ｔｈＬを変化させた場合のこのピンホール欠陥の検出サイズを示したものである。この検出結果から明らかなように、閾値ｔｈＬを高めるにつれて検出サイズが小さくなっていき、図示する関係から、ピンホール欠陥が正確に検出することができるためには、閾値ｔｈＬを５０階調とすればよいことが分かる。ここで、閾値に依存して検出欠陥サイズが変化するのは、濃淡画像におけるパターン波形のモジュレーションに起因するものである。なお、基準閾値ｔｈ１は、上記手順によって最小導体間隔から求めたものを用いる。
【０１１１】
また、図１０（ｂ）は直径２８μｍの飛散欠陥に対する閾値ｔｈＨを変化させた場合のこの飛散欠陥の検出サイズを示したものである。この検出結果から明らかなように、閾値ｔｈＨを高めるにつれて検出サイズが大きくなっていき、図示する関係から、飛散欠陥が正確に検出することができるためには、閾値ｔｈＨを９０〜１１０階調の範囲にすればよく、特に、１００階調が最適であることが分かる。
【０１１２】
以上の２値化閾値の決定に際しては、夫々の２値化画像の表示及び閾値をパラメータとしたときの基準欠陥の検出サイズに関するデータの集計が、図６において、画像表示手段１３１，１３２で行なわれる。これら画像表示手段１３１，１３２は、例えば、パーソナルコンピュータとモニタで構成されたものとすることができる。
【０１１３】
図１１はかかる画像表示手段１３１，１３２での表示画面２００の一具体例を示す図である。
【０１１４】
同図において、画像表示手段１３１，１３２の機能としては、夫々の閾値を変化させたときの２値画像２０１の記憶及び表示画面２００での表示機能、上記５つの閾値ｔｈ１，ｔｈ２，ｔｈ３，ｔｈＬ，ｔｈＨによる２値画像を表示画面２００で選択表示する機能、原画像２０２と２値画像との比較表示が表示画面２００でできるように構成することも可能である。また、表示画面２００で上記２値画像の任意の局所領域２０３を所定の倍率により拡大表示するように構成することも可能である。さらに、上記２値化閾値と欠陥検出サイズに関するデータの集計及び図１０で説明したようなグラフ２０４を表示画面２００で表示するように構成することも可能である。さらにまた、かかるデータの集計にあたっては、対象物の２値画像において、寸法の測定位置を予め教示しておくことにより、上記２値化閾値の設定手順を自動化するように構成することも可能である。そして、以上の機能を実行する際には、グラフィックインタフェースによるメニュー２０５に集約すると使い勝手がよい。
【０１１５】
なお、この実施形態の検査アルゴリズムの構成においては、最小線幅と最小導体間隔を実寸通り検出する基本閾値ｔｈ１が各々異なる場合を想定し、図６に示したように、膨張系と収縮系とで個別に２値化手段１０１，１０２を設けた。但し、これら膨張系，収縮系で基本閾値ｔｈ１が一致するか、またはその差が無視できる条件下では、膨張系と収縮系とでの２値化を共通化してもよい。
【０１１６】
また、パターンの検出に際しては、シェーディング補正や暗レベル補正を行なった上で２値化閾値の設定を行なうことにより、パラメータの標準化が可能となる。即ち、これにより、材質の変化や検出系の特性に依存するような明るさ変化の影響を最小限に抑制する効果がある。
【０１１７】
【発明の効果】
以上説明したように、本発明によれば、所定の画像入力手段により入力したパターンを、実寸通り高精度に２値化検出することが可能となり、パターン上の欠陥をより安定かつ精密に検査することができるし、２値化の際には、検出対象欠陥の限界値サンプルを使用することにより、容易に２値化閾値を設定することができる。これにより、従来の検査では対応が困難であったような微小欠陥を検出することが可能になるという優れた効果が得られる。
【図面の簡単な説明】
【図１】本発明による配線パターン検査方法の一実施形態を示す図である。
【図２】図１における２次微分画像を得るための２次微分オペレータの一具体例を示す図である。
【図３】図１における領域分割のための領域分割オペレータの具体例を示す図である。
【図４】配線パターン上の検査対象欠陥の種類を示す図である。
【図５】欠陥検出のためのパターンの連結関係抽出処理を示す図である。
【図６】本発明による配線パターン検査方法を用いた検査アルゴリズムの全体構成を示す図である。
【図７】パターンの最小幅と領域分割オペレータのサイズとの関係を示す図である。
【図８】パターンエッジ幅と領域分割オペレータのサイズとの関係を示す図である。
【図９】図１に示した各２値化閾値の設定基準を示す図である。
【図１０】設定閾値とピンホール欠陥，飛散欠陥の検出サイズとの関係を示す図である。
【図１１】図６に示す画像表示手段における表示画面の一具体例を示す図である。
【符号の説明】
１ 配線パターン
２ ヘアライン断線
３ ピンホール欠陥
４ ヘアラインショート
５ 飛散欠陥
６ 基材部
１０ 濃淡画像
１１ ２次微分画像
１００ パターン検出／画像入力手段
１０１，１０２ ２値化手段
１３１，１３２ 画像表示手段
ｔｈ１ 基準閾値
ｔｈＨ 飛散検出用の閾値
ｔｈＬ ピンホール検出用の閾値
ｔｈ２ ヘアラインショート検出用の閾値
ｔｈ３ ヘアライン断線検出用の閾値[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a pattern binarization method in industrial image processing, and in particular, detects a wiring pattern with high accuracy in an automatic appearance inspection apparatus such as a printed wiring board, a ceramic substrate, and a green sheet.Wiring pattern inspection methodAbout.
[0002]
[Prior art]
In a conventional inspection apparatus such as a printed circuit board, for example, an appearance inspection apparatus, as a pattern detection method, a simple binarization, that is, a binarization process using a single binarization threshold is a general method. In particular, in the case of the binarization device described in Japanese Patent Application Laid-Open No. 1-142988, in order to stably detect a thin hairline-shaped disconnection defect occurring in a wiring pattern, a binary image with a fixed threshold value and a binary differential of a second derivative are used. And a method of synthesizing the image with the coded image. According to this method, it is possible to detect a wiring pattern to be detected at a fixed threshold value, and detect a hairline-shaped disconnection defect or a short defect that short-circuits between wiring patterns by using a second derivative image to reveal the defect. Configuration.
[0003]
In addition, "JSPE-59-06", 1993 pp. A high-precision printed circuit board pattern inspection apparatus based on comparison with a design pattern in 993-1000 describes in detail a binarization method for extracting minute amplitude defects. This method also detects minute defects by applying an operator based on differentiation.
[0004]
[Problems to be solved by the invention]
In the above-mentioned prior art, a thin hairline-shaped defect portion appears as a steep signal change by performing a second-order differentiation of a detected grayscale image of a wiring pattern, thereby detecting such a defect.
[0005]
However, a pinhole-shaped defect (hereinafter referred to as a pinhole defect) occurring inside the conductor pattern or a semi-short defect in which the conductor interval is equal to or less than a reference value (design specification) although the conductor is not actually short-circuited. In general, a signal change in a grayscale image becomes gentle, so that the effect of the appearance of such a defect by secondary differentiation processing is small, and a fine pinhole defect has a low signal contrast. It cannot be detected with the binarization threshold set in this case.
[0006]
On the other hand, as defects occurring in the wiring pattern, insufficient defects (also referred to as missing defects, such as defects such as pinholes, disconnection and half-disconnection) and excess defects (also referred to as protruding defects). Defects such as scattering, isolated points, semi-shorts and short-circuits correspond to these), but it is difficult to treat these defect modes equally with a single binarization threshold.
[0007]
An object of the present invention is to solve such a problem and to stably extract a detection target from an image without depending on a difference in a graphic property due to a shape or an area of the detection target or a change in a target pattern density. Made it possibleWiring pattern inspection methodIs to provide.
[0008]
[Means for Solving the Problems]
In order to achieve the above object, the present invention providesMeans for inputting a grayscale image of a wiring pattern formed on an object to be inspected, such as a printed board or a ceramics substrate; second differential processing means for converting the grayscale image into a second differential image; A plurality of binarized differential image binarization means for binarizing with a predetermined different threshold, a plurality of gray image binarization means for binarizing the grayscale image with a predetermined threshold different from each other; Whether the pixel processed by the second-order differential image binarization unit and the grayscale image binarization unit is a pixel of the wiring pattern portion or a pixel of the base material portion of the object; Area determining means for determining whether or not the pixel is a pixel at the boundary of the wiring pattern, wherein the output of the secondary differential image binarizing means and the output of the grayscale image binarizing means are output by the area determining means. By selecting and synthesizing according to the output, It is intended to obtain a binarized image.
[0009]
The area determining means counts the number of effective pixels of a predetermined pixel value in an N × M (N and M are integers) area dividing operator for the output image of the grayscale image binarizing means, and according to the counted value. The above determination is made.
[0014]
BEST MODE FOR CARRYING OUT THE INVENTION
As shown in FIG. 4, the wiring pattern formed on a printed circuit board, a ceramic substrate, or a green sheet (hereinafter, represented by a printed circuit board) has a disconnection of the wiring pattern 1 or a width of the wiring pattern 1 as a reference value (design value). Specified defects) (defective defects or concave defects) typified by defects such as semi-disconnections and pinholes, and excess defects (typically defects such as semi-shorts, shorts, and scattering). Protrusion defect or convex defect). A wiring pattern inspection device such as a visual inspection device inspects the presence or absence of such a defect.
[0015]
FIG. 5 is a diagram showing a pattern connection relation comparison method which is an example of such a defect inspection algorithm in the wiring pattern inspection apparatus. Since the details of such an inspection algorithm are described in Japanese Patent No. 1589793, the principle will be briefly described here.
[0016]
In the drawing, the connection relation comparison method binarizes a grayscale image of a detected wiring pattern and expands the binary pattern into a thick line to form a semi-short portion (a portion where the pattern interval is equal to or less than a reference value). An expansion system that converts an image pattern into a short-circuited image pattern, and a contraction that converts this binary pattern into a thin line shape to convert a semi-disconnected portion (a portion whose pattern width is equal to or smaller than a reference value) into a disconnected state. And the connection between pads (circular portion at the end of the pattern) is extracted and designed for the image pattern obtained by the expansion system and the image pattern obtained by the contraction system. Compared with the information, it detects a defect such as a semi-short circuit or a semi-disconnection.
[0017]
This is due to the inspection rule that a conductor is determined to be defective when the conductor interval is equal to or less than a specified interval, or a defect when the conductor width is equal to or less than the specified width. In the case of a shrinkage system, the conductor is cut off with the conductor width less than a specified value being eliminated, and a defect is inspected. Therefore, when the expansion pattern is formed from the original wiring pattern, the expansion amount is the minimum conductor interval, and the contraction amount is the minimum pattern width.
[0018]
Here, the extraction process of the connection relationship means that a number (pad number) is assigned to the center of the pad, the pad number is propagated along the pattern, and when this reaches another pad number, that pad number and the starting point This is a method of storing as if there is a connection with the pad number.
[0019]
According to FIG. 5, in the case of the expansion system, the connection relationship that the pad number (1) is connected to the pad numbers (2), (3), and (4) is extracted. Since (1) and the pad number (2) are connected to the pad number (3) and the pad number (4), respectively, the extracted connection relation is compared with the design information. This indicates that there is a defect such as a semi-short or short between the pad number (1) and the pad numbers (3) and (4).
[0020]
Further, in the case of the contraction system, the connection relationship that only the pad numbers (3) and (4) are connected is extracted in the same manner. According to the design information, the pad numbers (1) and (4) are obtained. Since the connection between 2 and 4 is also performed, it can be seen that there is a defect such as a partial disconnection or a disconnection between the pad numbers 1 and 2.
[0021]
FIG. 6 is a diagram showing the entire configuration of the inspection algorithm of the wiring pattern inspection device.
[0022]
In the figure, a pattern detection unit 100 photoelectrically converts a light image from a printed circuit board to obtain a signal of a grayscale image of the surface of the printed circuit board. This grayscale image is sent to an expansion system and a contraction system.
[0023]
In the expansion system, the grayscale image of the wiring pattern is binarized 101, the obtained binary pattern is expanded 110a as described above, and the expansion pattern is reduced and changed while maintaining the connection relationship. After performing the data compression processing 111a, the data is temporarily stored 112 in the image memory. The same applies to the reduction system. The grayscale image of the wiring pattern is binarized 102, the obtained binary pattern is reduced 120 as described above, and the reduced pattern is imaged while maintaining the connection relationship. After performing the data compression process 111b for reducing and changing the size, the image is temporarily stored 113 in the image memory.
[0024]
Thereafter, the expansion pattern and the reduction pattern are read from the image memory to extract 114 their respective connections, and the extracted results are compared with the design information 115 to determine a defect 116. A connection relationship portion including a defect in the pattern and the contraction pattern is extracted 117, and a defect position is specified 118.
[0025]
In such an inspection of the wiring pattern, in the expansion system, not only the short circuit as shown in FIG. 4 but also a small surplus system defect such as a half short circuit, a narrow hairline short circuit, a fine scattering, and an isolated point are clearly detected as defects. It is necessary to obtain an expansion pattern that makes it possible and eliminates insufficient defects such as hairline disconnection and pinholes. In addition, for a contraction system, disconnection or half disconnection as shown in FIG. Needless to say, shortage defects such as hairline breaks and fine pinholes with narrow widths can be clearly detected as defects, and a shrinkage pattern that eliminates redundant defects such as hairline shorts, fine scattering, and isolated points can be obtained. It is necessary to.
[0026]
Conventionally, as the binarization processes 101 and 102, the detected grayscale image is binarized by a simple binarization method using a single binarization threshold, and the binarization process is performed by an expansion system and a contraction system. However, it has not been possible to set a binarization threshold value capable of detecting all of the surplus and deficiency defects as described above. On the other hand, the present invention enables binarization 101 and 102 so that all of the above-mentioned defects can be detected by processing in the expansion system and the contraction system.
[0027]
Hereinafter, according to the present inventionWiring pattern inspection methodAn embodiment will be described.
In this embodiment, binarization using three threshold values and binarization using secondary differentiation are used together, and five types of threshold values for the binarization can be appropriately set. First, the five types of thresholds will be described with reference to FIG.
[0028]
FIG. 1A shows a specific example of a wiring pattern on a printed circuit board, where 1 is a wiring pattern, 2 is a broken hairline, 3 is a pinhole, 4 is a hairline short, 5 is flying, and 6 is printed. This is the base portion of the substrate.
[0029]
In FIG. 1A, it is assumed that the surface of the base portion 6 of the printed circuit board has high brightness and is bright, and the wiring pattern 1 having low brightness and dark is formed on this surface. In this case, hairline breaks 2 and pinholes 3 are generated in the wiring pattern 1, and a hairline short 4 is formed between the wiring pattern 1 and another wiring pattern. It is assumed that scattering 5 has occurred.
[0030]
As shown in FIG. 1B, the grayscale image signal 10 read along the line AA on the printed board includes a low-level portion for the wiring pattern 1 and a high-level portion for the base portion 6, The level change portion at the boundary between the low level portion and the high level portion becomes the edge portion of the wiring pattern 1.
[0031]
Here, the presence of the hairline disconnection 2 in the dark wiring pattern 1 is brighter than the wiring pattern 1 portion, but is darker and thinner than the base portion 6, and approaches the brightness of the wiring pattern 1 as it becomes thinner. Therefore, in the grayscale image signal 10, the level of the hairline break 2 is slightly higher than the level of the wiring pattern 1 as shown in FIG. Also, the pinhole 3 has the brightness between the wiring pattern 1 and the base portion 6 as in the case of the hairline disconnection 2, but has a larger area than the hairline disconnection 2. As shown in FIG. 1 (b), the level is higher than that of the hairline disconnection 2.
[0032]
Further, since the hairline short 4 and the scattering 5 are surrounded by the base material 6, the hairline short 4 and the scattering 5 have brightness close to that of the base material portion 6. Comes closer to the base member 6. Therefore, in the grayscale image signal 10, as shown in FIG. 1B, the level of the hairline short 4 and the scattering 5 is lower than the level of the base 6, but the level of the hairline disconnection 2 and the pinhole 3. And the level difference between them increases as the difference in brightness (contrast) between the base member 6 and the wiring pattern 1 increases. Also in this case, the scattering 5 has an area with respect to the hairline short 4, and therefore, in the grayscale image signal 10, the level of the scattering 5 is lower than the level of the hairline short 4.
[0033]
As described above, the signal level of the grayscale image signal 10 is significantly different between the shortage defect such as the hairline disconnection 2 and the pinhole 3 and the surplus defect such as the hairline short 4 and the scattering 5. In a binarization method using a single binarization threshold, a binarization pattern including all of these defects cannot be obtained.
[0034]
In this embodiment, as shown in FIG. 1B, for a grayscale image signal 10,
th1: basic threshold
thH: high threshold for scattering detection
thL: low threshold for pinhole detection
Is used.
[0035]
Further, the detection of the hairline disconnection 2 and the hairline short 4 uses a binarization method of the secondary differential signal of the grayscale image signal 10 together. FIG. 1C shows the secondary differential signal 11. Since the signal change of the portion of the hairline break 2 in the grayscale image signal 10 is steeper than that of the portion of the pinhole 3, the portion of the hairline break 2 in the secondary differential signal 11 is changed as shown in FIG. The amplitude is larger than that of the pinhole 3. Similarly, in the second derivative signal 11, the portion of the hairline short 4 has a larger amplitude than the portion of the scattering 5. From this, for this second derivative signal 11,
th2: hairline short detection threshold value for the second derivative image
th3: threshold for detecting hairline breakage in the second derivative image
Set.
[0036]
FIG. 1D shows a binary pattern A obtained by binarizing the grayscale image signal 10 in FIG. 1B with an appropriately set basic threshold value th1.₁Where I is the brightness (shading value) of the grayscale image signal.
[0037]
(Equation 1)

[0038]
It is. FIG. 1H shows a binary pattern A obtained by binarizing the grayscale image signal 10 in FIG. 1B with an appropriately set scattering detection threshold thH._HIndicating that
[0039]
(Equation 2)

[0040]
It is. FIG. 1J shows a binary pattern A obtained by binarizing the grayscale image signal 10 in FIG. 1B with an appropriately set pinhole detection threshold thL._LIndicating that
[0041]
(Equation 3)

[0042]
It is.
[0043]
FIG. 1 (i) shows a binary pattern A obtained by binarizing the second derivative signal 11 in FIG. 1 (c) with an appropriately set hairline short detection threshold th2.₂And the second derivative is I ″_a, I "_b, I "_c, I "_dAs
[0044]
(Equation 4)

[0045]
It is. Here, the second derivative I ″_aIs obtained by applying a 5 × 5 second-order differential operator for each pixel shown in FIG. 2 (a) to the grayscale image signal 10 shown in FIG. 1 (b). I "_b, I "_c, I "_dAre obtained by applying a 5 × 5 second-order differential operator in pixel units shown in FIGS. 2B, 2C and 2D to the grayscale image signal 10, respectively. FIG. 1 (k) shows a binary pattern A obtained by binarizing the second derivative signal 11 in FIG. 1 (c) with a hairline short detection threshold th3 appropriately set.₃Indicating that
[0046]
(Equation 5)

[0047]
It is.
[0048]
Note that the second derivative operator shown in FIG. 2 (a) makes the pattern edge along the horizontal direction obvious, and the second derivative operator shown in FIG. 2 (b) makes the pattern edge along the upward right direction appear. The second derivative operator shown in FIG. 2 (c) manifests a pattern edge along the vertical direction, and the second derivative operator shown in FIG. 2 (d) shows a pattern edge along the downward right direction. Is to be manifested. That is, these secondary differential operators make the edge portion of the input grayscale image obvious, and the steeper the edge portion, the higher the secondary differential value I "_a, I "_b, I "_c, I "_dBecomes larger. Accordingly, as shown in FIG. 1C, the second derivative of the hairline break 2 and the hairline short 4 is larger than the second derivative of the pinhole 3 and the scattering 5.
[0049]
By the way, when the binarization by the above equations 1 to 5 is performed, the binary pattern A based on the reference threshold₁As shown in FIG. 1D, “1” is set at the wiring pattern 1 and “0” is set at the base 6 at the pattern edge. That is, it is a signal whose level is inverted at the pattern edge.
[0050]
Binary pattern A based on scattering detection threshold thH of grayscale image 10_HIs "1" for a defect such as a hairline short 4 or a scatter 5 existing in the base member 6 as shown in FIG. 1 (h), and by appropriately setting the scatter detection threshold thH, the scatter 5 Is included as a signal having a time width corresponding to the actual size of the width of the scatter 5. However, this binary pattern A_HMay include not only the signal component for the scatter 5 but also the signal component for the hairline short 4, and is always “1” in the wiring pattern 1.
[0051]
The binary pattern A based on the pinhole detection threshold thL of the grayscale image 10_L1 (j) becomes “0” for a defect such as a hairline break 2 or a pinhole 3 existing in the wiring pattern 1, and by appropriately setting the pinhole detection threshold thL, as shown in FIG. The signal component for the hole 3 is included as a signal having a time width corresponding to the actual size of the width of the pinhole 3. In the part of the base material part 6, the binary pattern A_LIs always "0".
[0052]
Differential binary pattern A based on hairline short detection threshold th2 of secondary differential pattern 11₂As shown in FIG. 1 (i), by appropriately setting the hairline short detection threshold th2, the signal component for the hairline short 4 has a time width “1” corresponding to the actual size of the width of the defect. Although it is included as a signal, the edge portion of the pinhole 3 and the scattering 5 and the edge portion of the wiring pattern 1 are also included according to the steepness of the edge portion.
[0053]
Differential binary pattern A based on hairline disconnection detection threshold th3 of secondary differential pattern 11₂As shown in FIG. 1 (k), by appropriately setting the threshold value th3 for detecting a hairline break, the signal component for the hairline break 2 has a time width “0” corresponding to the actual size of the width of the defect. Although it is included as a signal, an edge portion such as the pinhole 3 or the wiring pattern 1 is also included depending on the steepness of the edge portion.
[0054]
Now, in the above binary pattern, the binary pattern A_L, A₃Are logically processed to obtain a binary pattern (A) including a signal component of a time width corresponding to the actual size of each defect of the hairline disconnection 2 and the pinhole 3._L・ A₃) Is obtained. Similarly, binary pattern A_H, A₂Are ANDed to form a binary pattern (A) including signal components of a time width corresponding to the actual size of each defect of the hairline short 4 and the scattering 5._H+ A₂) Is obtained. Here, * represents a logical product and + represents a logical sum.
[0055]
By the way, the binary pattern (A_L・ A₃), The hairline disconnection 2 and the pinhole 3 are defects existing in the area of the wiring pattern 1. Therefore, in order to obtain signal components of these defects, the binary pattern (A_L・ A₃) Only needs to be extracted in the area of the wiring pattern 1. Similarly, the binary pattern (A_H+ A₂), The hairline short 4 and the scatter 5 are defects existing in the region of the base member 6. Therefore, in order to obtain signal components of these defects, the binary pattern (A_H+ A₂), It is necessary to extract only the part of the area of the base member 6. The edge of the wiring pattern 1 is detected by binarization using the thresholds th1, thH, thL, th2, and th3, but is not a defect. Further, in the binarization using the thresholds th1, thH, thL, th2, and th3, the edge portion of the wiring pattern 1 is not correctly detected. That is, the basic threshold value th1 is for detecting the edge portion of the wiring pattern 1. By appropriately setting the basic threshold value th1, the edge portion of the wiring pattern 1 is detected at a correct position. The binary pattern A shown in FIG.₁Indicates the edge portion of the wiring pattern 1.
[0056]
Based on the above, the hairline disconnection 2 and the pinhole 3 have the binary pattern (A_L・ A₃), The hairline short 4 and the scattering 5 are binary patterns (A_H+ A₂), The edge of the wiring pattern 1 is a binary pattern A₁From each other.
[0057]
Therefore, in this embodiment, the signal is further divided in the time axis direction into a region of the wiring pattern 1 excluding the edge portion, an edge region including the edge portion of the wiring pattern 1, and a region of the base portion 6. . FIG. 1E shows a binary pattern A representing the area of the wiring pattern 1._NL(Note that the area of the wiring pattern 1 is also the wiring pattern area A._NLFIG. 1F shows a binary pattern A representing an edge region of the wiring pattern 1._NM(Note that the edge area is also the edge area A_NMFIG. 1 (g) shows a binary pattern A representing the area of the base member 6._NH(Note that the region of the base portion 6 is also the base portion region A_NH) Respectively. Area A_NL, A_NM, A_NHIs the number of effective pixels of “1” around the pixel of interest in the region dividing operator described below, and the determination value is N as described later._L, N_HIs represented by the following equation (6).
[0058]
(Equation 6)

[0059]
However,
A_NL+ A_NM+ A_NH= 1
It is.
[0060]
In order to detect the hairline break 2 and the pinhole 3 in the wiring pattern area, the binary pattern (A_L・ A₃), A_NL, That is,
A_NL・ (A_L・ A₃)
In order to detect the hairline 4 and the scattering 5 in the region of the base material portion 6, the binary pattern (A_H+ A₂), A_NH, That is,
A_NH・ (A_H+ A₂)
Perform In order to detect the edge of the wiring pattern 1 in the edge area,
, A_NM・ (A₁・ A₃+ A₁・ A₂)
Is performed. In the above equation, A₁・ A₃Is the level in the edge region and A in the wiring pattern region._NL・ (A_L・ A₃) Is used to make the levels equal to each other.₁・ A₂Is the level in the edge area when the binary pattern A1 is inverted in level from the level shown in FIG._NH・ (A_H+ A₂This is an operation for matching the level of ()).
[0061]
Therefore, the binarized output A obtained by the above binarization processing_RIs
[0062]
(Equation 7)

[0063]
It becomes. This output A_RIs shown in FIG.
[0064]
Next, the above-described area division will be described.
[0065]
Here, such a wiring pattern area A_NL, Edge area A_NM, Base material area A_NHIs divided into binary patterns A based on the basic threshold th1 shown in FIG.₁And the binary pattern A₁The pixel of interest in_NL, A_NM, A_NHIt is determined which one of the above belongs. This area division operator is a binary pattern A₁Is applied to the two-dimensional array of pixels in the region A. Depending on how the target pixel is surrounded by the surrounding effective pixels serving as the determination reference, the target pixel_NL, A_NM, A_NHIt is determined which one of the above belongs.
[0066]
FIG. 3 shows a specific example of the area dividing operator. In this specific example, the effective pixels are pixels B on the circumference around the target pixel P, and it is determined whether or not these effective pixels B are “1”. The number N of effective pixels B is counted, and the target pixel P is set in the area A according to the count value N._NL, A_NM, A_NHIt is determined which one of the above belongs.
[0067]
Here, when the diameter of the circumference where the peripheral pixel B exists is n pixels in pixel units, this area dividing operator is called an operator of size n × n. The operator size is set according to the line width of the wiring pattern 1 and the edge width of the wiring pattern. In the case of an inspection apparatus for inspecting a printed circuit board having a wiring pattern with a different line width or an edge width, An area dividing operator having an optimum size is provided in advance for each printed circuit board, and an area dividing operator to be used is made different depending on the printed circuit board to be inspected.
[0068]
FIG. 3 shows an area division operator having a size of 7 × 7, 9 × 9, and 11 × 11. In the area division operator of size 7 × 7, the number of effective pixels B is 16 pixels, and two judgment values N_L= 14, N_H= 2 is set. In the area division operator having a size of 9 × 9, the number of effective pixels B is 24, and two determination values N_L= 22, N_H= 2 is set and the effective pixel B is 28 pixels in the area division operator of size 11 × 11, and two determination values N_L= 25, N_H= 3 is set.
[0069]
The determination value N of such a region division operator_L, N_HFrom the above and the number N of effective pixels B of “1”, the area A to which the target pixel P belongs_NL, A_NM, A_NHIs determined.
[0070]
By the way, if the above-mentioned defect does not occur, when all the effective pixels B are “1”, it can be determined that the target pixel P is in the area of the wiring pattern, and all the effective pixels B are “0”. At this time, it can be determined that the pixel of interest P is in the region of the base portion. Further, when some of the effective pixels B are "1" and the effective pixels B of the stiffness are "0", the pixel of interest P is Can be determined to belong to the region of the edge portion of. For this reason, for example, in a 7 × 7 area division operator, N_L= 16, N_H= 0 is also conceivable.
[0071]
However, in this case, for example, in a 7 × 7 area division operator, the wiring pattern area A_NLIn the case where the hairline break having a width of 1 pixel or less crosses the circumference where the effective pixels B are arranged, the two effective pixels become “0”, the number of effective pixels B of “1” becomes N = 14, and N < N_LTherefore, the wiring pattern area A_NLIn the area A at the edge of the wiring pattern_NMWill be erroneously determined. Therefore, in FIG._LIs smaller than the total number of the effective pixels B by the width of the hairline disconnection. Similarly, the determination value N_HIn this case, a value corresponding to the width of the hairline short is given._NHTo the area A at the edge of the wiring pattern._NMIs not determined by mistake.
[0072]
Then, as described above, the determination value N_L, N_HCan be set to correctly determine each area of the wiring pattern, its edge portion, and the base material portion._NLHairline disconnection in the area, and the area A of the substrate part_NHHairline short circuit can be detected.
[0073]
Note that the width of the hairline disconnection or the hairline short may be different depending on the size of the region dividing operator, and is determined according to the width of such a defect in the wiring pattern using the region dividing operator. Is what is done.
[0074]
Further, here, the effective pixels B are arranged on the circumference that is isotropically arranged with respect to the target pixel P. However, the present invention is not limited to this. The range of all pixels in the center circle or all pixels in the square centered on the pixel of interest can be appropriately changed according to the graphic properties and specifications of the inspection object.
[0075]
Next, the relationship between the area dividing operator and the line width of the wiring pattern will be described with reference to FIG.
[0076]
FIG. 7A shows an area dividing operator (here, an operator size of 5 × 5) having a size equal to or less than the line width of the wiring pattern 1 (here, this line width is 5 pixels). In the drawing, the position of the area dividing operator with respect to the wiring pattern 1 changes in the order of (i), (ii), and (iii).
[0077]
In the state (i), all the effective pixels B around the target pixel P are in the area of the wiring pattern 1, and all the effective pixels B at this time are “1”. Here, assuming that the ratio of the number N of effective pixels B of “1” to the number of all effective pixels B is α, in the state of (i), α = 12/12. Next, in the state of (ii) in which the region dividing operator has moved rightward by two pixels with respect to the wiring pattern 1, the five effective pixels B on the right side protrude from the wiring pattern 1 and become "0". , And the ratio α = 7/12. Further, when the area dividing operator moves rightward by one pixel with respect to the wiring pattern 1 (iii), the right seven effective pixels B protrude from the wiring pattern 1 and become “0”. , And the ratio α = 5/12.
[0078]
As described above, when the operator size is smaller than the line width of the wiring pattern 1, when the area dividing operator is within the wiring pattern 1, the ratio α = 1 becomes the maximum, and the area dividing operator deviates from the wiring pattern 1. Α becomes smaller as it goes.
[0079]
On the other hand, FIG. 7B shows a case where the operator size is larger than the line width of the wiring pattern 1 (this line width is also set to 5 pixels) (here, the operator size is 7 × 7). It is shown. Also in this case, similarly to the case of FIG. 7A, the operation of the area dividing operator crossing the wiring pattern 1 in the right direction in the drawing is shown, and the position of the area dividing operator with respect to the wiring pattern 1 is (a). , (B), and (c) in this order.
[0080]
In the state (a), the target pixel P is located at the center of the wiring pattern 1 in the width direction. In this case, since the operator size is 7 × 7 with respect to the line width of five pixels, three effective pixels B on both the right and left sides protrude from the area of the wiring pattern 1 and are “0”. Therefore, in this state (a), α = 10/12. Next, in the state of (b) in which the area dividing operator has moved to the right by one pixel with respect to the wiring pattern 1, only the five effective pixels B on the right protrude from the wiring pattern 1 and become "0". At times, the ratio α is 11/12. Further, when the area dividing operator moves rightward by one pixel with respect to the wiring pattern 1 (c), the seven effective pixels B on the right protrude from the wiring pattern 1 and become “0”. , And the ratio α = 9/12.
[0081]
As described above, when the operator size is larger than the line width of the wiring pattern 1, it is more likely that the region dividing operator is slightly offset from the wiring pattern 1 to one side than the case where the target pixel P is at the center in the width direction of the wiring pattern 1. The ratio α becomes large, and thereafter, as the area dividing operator departs from the wiring pattern 1, α becomes smaller.
[0082]
From the above, when the operator size is equal to or smaller than the line width of the wiring pattern 1, when the area dividing operator is within the wiring pattern 1, the ratio α = 1, which is the maximum, and Α decreases monotonically as it deviates, but when the operator size is larger than the line width of the wiring pattern 1, the pixel of interest P is slightly shifted from the center of the wiring pattern 1 in the width direction. Α increases, and the relationship between α and the positional relationship between the area dividing operator and the wiring pattern becomes unnatural.
[0083]
Therefore, the size of the region dividing operator needs to be smaller than the minimum line width of the wiring pattern 1 on the printed circuit board. As a result, the maximum size of the usable region dividing operator is determined according to the minimum line width of the wiring pattern 1 on the printed board to be inspected.
[0084]
FIG. 8 is a diagram illustrating the minimum size of a usable area dividing operator for the wiring pattern 1 on the printed board to be inspected.
[0085]
As shown in FIG. 8, the edge portion of the grayscale image signal 10 of the wiring pattern 1 has a finite period of an inclined waveform in which the brightness (grayscale value) changes continuously as shown in FIG. I have. Here, in this edge portion, the number of pixels of brightness included in the brightness range from the brightness of the pinhole detection threshold thL to the brightness of the scatter detection threshold thH is hereinafter referred to as the number of brightness change pixels.
[0086]
FIG. 8A shows a binary pattern near an edge when operator size> number of brightness change pixels. In this case, the area A of the edge determined by the area division operator_NMIncludes the entire edge portion of the grayscale image signal 10, and therefore, the region A of this edge portion_NM  Binary pattern A obtained at basic threshold th1 only within₁Edge E₀Is obtained.
[0087]
On the other hand, as shown in FIG. 8B, when the operator size <the number of brightness change pixels, the edge area A determined by the area division operator_NMIs smaller than the range of the edge portion of the grayscale image signal 10, and the area A of the wiring pattern 1 is determined by the area dividing operator._NLArea A of the base member 6_NHAlso enter the area of the edge portion. For this reason, of course, the area A_NMThe binary pattern A obtained at the basic threshold th1₁Edge E₀Is obtained, but the area A_NLHowever, the binary pattern A obtained by the pinhole detection threshold thL_LThe edge E for this edge_LBut also the area A_NHHowever, the binary pattern A obtained by the scattering detection threshold thH_HThe edge E for this edge_HRespectively occur.
[0088]
As described above, if the operator size is smaller than the number of brightness change pixels, the area A_NL, A_NM, A_NH, An unnatural situation occurs in which the same edge portion in the wiring pattern 1 is erroneously detected. Therefore, the size of the region dividing operator needs to be equal to or larger than the width of the edge of the wiring pattern 1.
[0089]
Note that the area division operator can take only discrete values of 7 × 7, 9 × 9, and 11 × 11 in size, but the threshold N_H, N_LIs provided, it is possible to realize an area dividing operator having a size between such discrete actual sizes. That is, the threshold N_H, N_L, The same effect as reducing the operator size can be obtained.
[0090]
In this embodiment, as described above, the pinhole defect is revealed at the low threshold value thL, and the scattering defect is manifested at the high threshold value thH, and the hairline disconnection or the hairline short is detected exclusively by the second derivative threshold. As a result, the binarized output A represented by the above equation (7)_RThe threshold values th1, thH, thL, th2, and th3 are set in the case of the defect inspection apparatus shown in FIG. 6 by the binarization means 101 and 102 and the above-described equation (7). This is performed using image display means 131 and 132 for displaying the binarized image to be displayed.
[0091]
The binarizing means 101 and 102 can switch the binarizing function by appropriately setting the threshold. That is, in order to perform binarization using second-order differentiation, the threshold may be set to be th1 = thH = thL. As a result, A 1₁= A_H= A_LTherefore, the binarized output signal A_RIs

Thus, binarization using the conventional second derivative can be realized. However, in the above equation (7), A_NM・ A₁・ A₂Is A_NM= 1 (pattern edge area) and A₁= 0 represents a second-order differential binary image based on a threshold value th2 in an area where A = 0_NH・ A₃・ A₂Is A_NH= 1, but in this area A₃= 1 and A₁= 1. Therefore, in this case, (A_NM・ A₁・ A₂+ A_NH・ A₃・ A₂) Is A₁・ A₂And the above binary output signal A_R  Is obtained.
[0092]
On the other hand, in order to function as a three-threshold binarization without using the second derivative binarization, the threshold of the second derivative image I ″ may be set to th2 = ∞, th3 = −∞. , A₂= 0, and according to Equation 5, A₃= 1, this is substituted into Equation 7 above, and the binary output signal A_RIs
A_R= A_NL・ A_L+ A_NM・ A₁+ A_NH・ A_H
Thus, a three-threshold binary pattern is obtained.
[0093]
Furthermore, when functioning as simple binarization (binarization using only the threshold th1), th1 = thH = thL, th2 = ∞, and th3 = −∞. Thus, A₁= A_H= A_L, A₂= 0, A₃= 1 and A_NL+ A_NM+ A_NH= 1, and when this is substituted into the above equation 7, the binary output signal A_RIs
A_R= A₁
It becomes.
[0094]
As described above, in this embodiment, it is possible to cope with the detection of various types of defects, and it is possible to flexibly switch functions.
[0095]
Next, a method of setting the binarization thresholds th1, th2, th3, thH, and thL for accurately detecting a pattern will be described.
[0096]
For this setting, a limit defect sample of each defect mode (that is, a type of defect such as a broken hairline or a short hairline) is used as a reference. The reference sample may be a real defect or a simulated defect. That is, there are a hairline disconnection or a hairline short having a minimum width to be detected, a half short having a minimum conductor interval, a half disconnection defect having a minimum line width, a pinhole having a minimum detection size, and a scattering defect.
[0097]
FIG. 9 shows a specific example of the criteria for setting each binarization threshold in the expansion system and the contraction system in FIG.
[0098]
When setting each threshold value in the expansion system of FIG. 6, a grayscale image having the above-mentioned sample is supplied from the pattern detection unit 100 to the binarization unit 101, and the binarization pattern A obtained here is supplied._R  Is displayed on the image display means 131, and the threshold value is adjusted to set the optimum threshold value. Hereinafter, the setting method will be described with reference to FIG.
[0099]
In the case of the expansion system, the reference threshold th1 for pattern detection is such that the detection target is a short-circuit with the edge of the pattern (a defect in which the conductor pattern interval is narrower than the minimum interval specified: see FIG. 4). As described above, the binarized pattern obtained at the reference threshold value th1 is set to a threshold value such that the half short-circuit is always short-circuited when expanded by the specified width. This is a reference when determining the threshold value thH for detection. Therefore, it is necessary to set the threshold value th1 so that the minimum conductor interval (not a defect) defined by the binarized pattern obtained by binarizing the grayscale image with the reference threshold value th1 is detected to the actual size. is there. This means that if there is no short-circuit, even if the space between the conductor patterns is a prescribed minimum, no short-circuit will occur when expanded. In order to set the reference threshold th1 in this manner, the above-mentioned number obtained from this sample is obtained by using th1 = thH = thL, th2 = ∞, th3 = −∞, and using a grayscale image of the minimum conductor interval as a sample. 7 binary output signal A_RA_R= A₁While adjusting the reference threshold th1 while displaying the image on the image display unit 131.
[0100]
In the case of the expansion system, the threshold th2 is used to binarize the second derivative image, and a hairline short or an isolated point as a surplus system defect is used as a defect to be inspected. The defect is set so as to be detected at the actual size. For this purpose, a gray image of a sample having a defect of such a size is obtained, and is binarized by the reference threshold th1 as described above. Value pattern A₁Is obtained, and a second derivative image of the grayscale image is obtained, binarized by the threshold value th2, and the threshold value th2 is adjusted so that the defect having the minimum size is detected in the actual size. In this case, the binarized pattern A of the binary image based on the threshold value th2 of the secondary differential image is set so that the defect to be inspected is extracted and displayed on the image display unit 131.₁Is extracted and supplied to the image display means 131. Therefore, by setting th1 = thH = thL and th3 = −∞, A₁= A_H= A_L= 0 (therefore, A₁= 1) and A₃= 1, the binary output signal A_RFrom the above equation 7,
A_R= A_NM・ A₁・ A₂+ A_NH・ A₂= (A_NM・ A₁+ A_NH) ・ A₂
It is assumed that. Area (A_NM・ A₁+ A_NH) Is A, as is apparent from FIG.₁= 1.
[0101]
In the case of the expansion system, the threshold value thH is for binarizing the grayscale image and detecting scattering as a surplus system defect, and the defect having the minimum size detected in the binarized image is detected in the actual size. Is set to In this case, the binarized pattern A of the binary image based on the threshold value thH of the grayscale image is selected so that the inspection target defect is extracted and displayed on the image display unit 131.₁Is extracted and supplied to the image display means 131. Therefore, by setting th1 = thL, th2 = ∞, th3 = −∞, A₁= A_L= A₂= 0 (therefore, A₁= 1) and A₃= 1, the binary output signal A_RFrom the above equation 7,
A_R= A_NH・ A_H
It is assumed that. Area A_NHAs is clear from FIG. 1, A₁= 1.
[0102]
The threshold value th3 is for detecting a hairline disconnection which is an insufficient defect in the conductor pattern, and the threshold value thL is for detecting a pinhole which is an insufficient defect in the conductor pattern. , These thresholds th3, thL are not used, and are set to the same value as the contraction system as an example.
[0103]
Next, when setting each threshold value in the contraction system of FIG. 6, a grayscale image having a sample is supplied from the pattern detection unit 100 to the binarization unit 102, and the binarization pattern A obtained here is obtained._RIs displayed on the image display means 132, and the threshold value is adjusted to set the optimum threshold value. Hereinafter, the setting method will be described with reference to FIG.
[0104]
In the case of the contraction type, the reference threshold th1 for pattern detection is that the detection target is an edge portion of the pattern and a semi-disconnection (a defect in which the width of the conductor pattern is smaller than the minimum width defined: see FIG. 4). As described above, when the binarized pattern obtained at the reference threshold th1 is reduced by the specified width, the threshold value is set such that the half-break is always broken. This is a reference when determining the threshold thL for pinhole detection. Therefore, it is necessary to set the threshold th1 so that the minimum conductor width (not a defect) defined in the binarized pattern obtained by binarizing the grayscale image with the reference threshold th1 is detected to the actual size. is there. This means that if there is no half-break, even if the width of the conductor pattern is a prescribed minimum, no break occurs when the conductor pattern contracts. In order to set the reference threshold th1 in this way, the above-mentioned number obtained from this sample is set by using th1 = thH = thL, th2 = ∞, th3 = −∞, and using a grayscale image of the minimum conductor width as a sample. 7 binary output signal A_RA_R= A₁While adjusting the reference threshold th1 while displaying the image on the image display unit 131.
[0105]
In the case of a contraction system, the threshold value th3 is for binarizing a second derivative image, and a hairline disconnection as an insufficient system defect is set as a defect to be inspected. For this purpose, a gray image of a sample having a defect of such a size is obtained, and is binarized by the reference threshold th1 as described above. A₁Is obtained, a second derivative image of the grayscale image is obtained, and the binarized image is binarized by the threshold value th3 to adjust the threshold value th3 so that the defect having the minimum size is detected in the actual size. In this case, the binarized pattern A of the binary image based on the threshold value th3 of the secondary differential image is set so that the inspection target defect is extracted and displayed on the image display unit 132.₁Is extracted and supplied to the image display means 132. Therefore, by setting th1 = thH = thL and th2 = ∞, A₁= A_H= A_L= 1 (hence A₁= 0) and A₂= 0 and the binary output signal A_RFrom the above equation 7,
A_R= A_NL・ A₃+ A_NM・ A₃= (A_NL+ A_NM) ・ A₃
(However, A_NHA in the area of = 1₁= A_H= 0, so A_NH・ A₃・ A_H= 0)
It is assumed that. Area (A_NL+ A_NM) Is A, as is apparent from FIG.₁= 1.
[0106]
In the case of a contraction system, the threshold thL is used to binarize a grayscale image to detect a pinhole as a defect of insufficient system, and a defect having a minimum size to be detected in the binarized image is detected in an actual size. Is set to In this case, the binarized pattern A of the binary image based on the threshold value thL of the grayscale image is selected so that the inspection target defect is extracted and displayed on the image display unit 132.₁Is extracted and supplied to the image display means 132. Therefore, by setting th1 = thH, th2 = ∞, th3 = −∞, A₁= A_H= 1 (hence A₁= 0) and A₂= 0, A₃= 1, the binary output signal A_RFrom the above equation 7,
A_R= A_NL・ A_L+ A_NM・ A₁
(However, A_NHA in the area of = 1₁= A_H= 0, so A_NH・ A₃・ A_H= 0)
It is assumed that. Area A_NLAs is clear from FIG. 1, A₁= 1 and the divided area A_NM(Ie, at the pattern edge)₁= 1.
[0107]
The threshold value th2 is for detecting a hairline short which is an excess defect in the conductor pattern, and the threshold value thH is for detecting scattering which is an extra defect in the conductor pattern. These thresholds th2 and thH are not used, and are set to the same value as the expansion system as an example.
[0108]
As described above, in the expansion system of FIG. 6, the threshold value th1, the threshold value th2 for detecting a hairline short / isolated point, and the threshold value thH for scattering detection are determined by using the binarizing unit 101 and the image display unit 131. In the contraction system shown in FIG. 6, a reference threshold th1, a threshold th2 for hairline disconnection detection, and a threshold thL for pinhole detection are set using the binarizing means 102 and the image display means 132. Is done.
[0109]
In the above setting, the relationship between the detection dimensions of the target defect may be obtained using the binarization threshold as a parameter. FIG. 10 shows a setting example of the threshold thL for detecting pinholes and the threshold thH for detecting scattering.
[0110]
FIG. 10A shows the detection size of a pinhole defect when the threshold thL is changed for a pinhole defect having a diameter of 25 μm. As is apparent from this detection result, the detection size decreases as the threshold thL is increased. From the relationship shown in the drawing, in order to accurately detect a pinhole defect, the threshold thL must be set to 50 gradations. It turns out that it is good. Here, the reason why the detected defect size changes depending on the threshold value is due to the modulation of the pattern waveform in the grayscale image. As the reference threshold th1, a value obtained from the minimum conductor interval by the above procedure is used.
[0111]
FIG. 10B shows the detection size of the scattering defect when the threshold value thH for the scattering defect having a diameter of 28 μm is changed. As is apparent from this detection result, the detection size increases as the threshold thH is increased. From the relationship shown in the drawing, in order to accurately detect a scattered defect, the threshold thH must be set to 90 to 110 gradations. It can be seen that the range should be within the range, and in particular, 100 gradations are optimal.
[0112]
When the above-described binarization threshold is determined, the display of the respective binarized images and the aggregation of data relating to the detection size of the reference defect when the threshold is used as a parameter are performed by the image display means 131 and 132 in FIG. It is. These image display means 131 and 132 can be constituted by, for example, a personal computer and a monitor.
[0113]
FIG. 11 is a diagram showing a specific example of the display screen 200 on the image display means 131 and 132.
[0114]
In the figure, the functions of the image display means 131 and 132 include the storage function of the binary image 201 when the respective threshold values are changed and the display function on the display screen 200, the five threshold values th1, th2, th3 and thL. , And thH, a function of selectively displaying a binary image on the display screen 200 and a function of comparing and displaying the original image 202 and the binary image on the display screen 200 are also possible. Further, it is also possible to configure so that an arbitrary local region 203 of the binary image is enlarged and displayed at a predetermined magnification on the display screen 200. Further, it is also possible to configure so that the binarization threshold and the data regarding the defect detection size are totaled and the graph 204 as described in FIG. 10 is displayed on the display screen 200. In addition, it is possible to automate the procedure for setting the above-mentioned binarization threshold value by teaching the measurement position of the dimensions in advance in the binary image of the object when summing up such data. is there. When executing the above functions, it is convenient to collect them in the menu 205 by the graphic interface.
[0115]
Note that, in the configuration of the inspection algorithm of this embodiment, it is assumed that the basic threshold th1 for detecting the minimum line width and the minimum conductor interval according to the actual size is different from each other, and as shown in FIG. , Binarizing means 101 and 102 are individually provided. However, under the condition that the basic threshold value th1 in these expansion system and contraction system is the same or the difference is negligible, the binarization of the expansion system and the contraction system may be shared.
[0116]
Further, upon detection of a pattern, by performing shading correction and dark level correction and then setting a binarization threshold, parameters can be standardized. In other words, this has the effect of minimizing the effect of brightness changes that depend on material changes and detection system characteristics.
[0117]
【The invention's effect】
As described above, according to the present invention, a pattern input by a predetermined image input unit can be binarized and detected with high accuracy as the actual size, and a defect on the pattern can be more stably and precisely inspected. In the case of binarization, the binarization threshold can be easily set by using the limit value sample of the defect to be detected. As a result, an excellent effect is obtained in that it is possible to detect a minute defect that has been difficult to deal with in the conventional inspection.
[Brief description of the drawings]
FIG. 1 according to the inventionWiring pattern inspection methodFIG. 3 is a diagram showing an embodiment of the present invention.
FIG. 2 is a diagram showing a specific example of a second derivative operator for obtaining a second derivative image in FIG. 1;
FIG. 3 is a diagram showing a specific example of an area division operator for area division in FIG. 1;
FIG. 4 is a diagram illustrating types of defects to be inspected on a wiring pattern.
FIG. 5 is a diagram showing a pattern connection relation extraction process for detecting a defect;
FIG. 6 according to the inventionWiring pattern inspection methodFIG. 3 is a diagram illustrating an entire configuration of an inspection algorithm using the.
FIG. 7 is a diagram showing a relationship between a minimum width of a pattern and a size of an area dividing operator.
FIG. 8 is a diagram illustrating a relationship between a pattern edge width and a size of an area dividing operator.
FIG. 9 is a diagram showing setting criteria for each binarization threshold shown in FIG. 1;
FIG. 10 is a diagram illustrating a relationship between a set threshold value and a detection size of a pinhole defect and a scattering defect.
11 is a diagram showing a specific example of a display screen on the image display means shown in FIG.
[Explanation of symbols]
1 Wiring pattern
2 Hairline disconnection
3 Pinhole defect
4 Hairline Short
5 scattering defects
6 Base material
10 Shading image
11 Second derivative image
100 pattern detection / image input means
101,102 binarization means
131, 132 image display means
th1 Reference threshold
thH Threshold for scattering detection
thL Threshold for pinhole detection
th2 Threshold for detecting hairline short
th3 Threshold value for hairline disconnection detection

Claims (2)

Means for inputting a grayscale image of a wiring pattern formed on an object to be inspected such as a printed circuit board or a ceramic substrate ,
Secondary differential processing means for converting the grayscale image into a secondary differential image;
A plurality of secondary differential image binarizing means for binarizing the secondary differential image with predetermined thresholds different from each other;
A plurality of grayscale image binarization means for binarizing the grayscale image with predetermined different thresholds;
Pixels processed by the secondary differential image binarization unit and the grayscale image binarization unit in the grayscale image are pixels of the wiring pattern portion or pixels of a base portion of the object. Or an area determining means for determining whether the pixel is a pixel at the boundary of the wiring pattern.
With
By selecting and combining the output of the secondary differential image binarization unit and the output of the grayscale image binarization unit according to the output of the area determination unit, the binarized image of the grayscale image is selected. A wiring pattern inspection method, characterized by obtaining: In claim 2,
The area determining means counts the number of effective pixels of a predetermined pixel value in an N × M (N, M is an integer) area dividing operator with respect to the output image of the grayscale image binarizing means, and A wiring pattern inspection method, wherein the determination is made in response to the request.

Images