JP3884834B2 - Defect inspection method and apparatus - Google Patents

Description

【００２３】
【数２】

【００２４】
よって、カメラから取り込んだ各色のＲＧＢ値を（数１）（数２）に代入してｘｙＹ値を算出し、この値が色彩計で計測したｘｙＹ値と一致するようなａ₁₁〜ａ₃₃を求めれば、カメラ固有の変換パラメータを求めることになる。未知のパラメータは９個なので、最低３色のＲＧＢ値（Ｒ₁，Ｇ₁，Ｂ₁）〜（Ｒ₃，Ｇ₃，Ｂ₃）とそれに対応する色彩計のｘｙＹ値（ｘ₁，ｙ₁，Ｙ₁）〜（ｘ₃，ｙ₃，Ｙ₃）でパラメータは算出できる。
【００２５】
（数２）よりＸＹＺはｘｙＹ値から（数３）で算出できるので、
【００２６】
【数３】

【００２７】
色彩計の３色のｘｙＹ値を（数３）に代入し、ＸＹＺを求め、（数１）に代入する（（数４））。
【００２８】
【数４】

【００２９】
これにより、カメラ固有の変換パラメータａ₁₁〜ａ₃₃を求め（６０４）、カメラのＲＧＢ値から色彩計の値と等しいｘｙＹ値を求めることが可能となる。
【００３０】
キャリブレーションによりあらかじめ算出したカメラ固有の変換パラメータを用いて、カメラから得られたＲＧＢ値をｘｙＹ値に色度変換し、画像中の色度分布を算出した後、２０３では、画像中から正常な色度値、すなわち、欠陥でない部分の色度を基準値として算出する。まず、図７上のグラフのように、画像中の各画素の色度ｘ、ｙを調べ、各ｘ、ｙ値をとる画素数をカウントし、２次元度数分布を作る。そして、画像中で最も画素数の多いｘ、ｙ値を調べる。画像中の大部分が欠陥ではない部分であることから、２次元度数分布のピーク値のｘ、ｙ値を調べ、図７下のグラフに示すようにピーク値、すなわち最も画素数が多いｘ、ｙ値を基準色のｘｙ色度値とする。
【００３１】
２０４では、この基準色に対する画像上の各位置での色相、色差を算出する。基準色の色度を（ｘｃ、ｙｃ）、画像上の位置（ｉ、ｊ）での色度を（ｘｉｊ、ｙｉｊ）とすると、図８に示すように位置（ｉ、ｊ）での色相を、色度図上での基準色に対する向きで算出する。その算出式を（数５）に示す。
【００３２】
【数５】

【００３３】
また、図９に示すように位置（ｉ，ｊ）での色差を、色度図上での基準色からの距離で算出する。その算出式を（数６）に示す。
【００３４】
【数６】

【００３５】
以上のように算出した基準色に対する画像の各位置での色相・色差より、図１０に示すように、色相で欠陥として検出したい色合いを限定し、（図では、色相θが、θ１≦θ≦θ２の範囲）し、色差で基準色との色の違いの程度を限定する（図では、色差ｄが、ｄ１≦ｄ≦ｄ２の範囲）。そしてこの範囲内にある部分を欠陥候補領域として抽出する。
【００３６】
このように色相と色差で範囲を限定して求めた欠陥候補の中には、欠陥として検出する必要のないものもある。例えば、基準色に対し、徐々に色度が変化していくようなものは、人間にはあまり気にならない。むしろ、周囲の色に対し、急激に色度が変化した部分や、輪郭がはっきりとした領域は、かなり目立つ欠陥である。そこで、このように周囲の色に対する色の変化がなだらかなものは正常部と見なし、変化が急激なもののみを欠陥と見なす場合、２０６では、欠陥候補領域について、周囲との色差の変化量を求め、その値がある一定値以上のものだけを欠陥とする。
【００３７】
図１１を用いて説明すると、（Ａ）は２０５により抽出された欠陥候補領域である。（Ｂ）の１１２´は（Ａ）の１１２上の色差である。更に、１１２上の各位置での色差の変化量、すなわち、１１２' を微分したものが、（Ｃ）の１１３である。このように周囲との色差の変化量が小さいものは、微分値も小さくなる。そこで、（Ｃ）に示すように、微分値が、ある一定値１１４より大きいもののみを欠陥領域とする。その結果、（Ｄ）のように色差が大きく、かつ色変化量が大きい、すなわち輪郭が鮮明な領域のみが検出される。
【００３８】
次に、しきい値１１４の決定方法を図１２を用いて説明する。図１２（Ａ）のグラフは、縦軸を色相と色差により抽出された各欠陥候補領域内の色差の最大値、横軸を各欠陥領域の輪郭部分の色差微分値の最大値にとり、真の欠陥領域での値を×で、欠陥でない領域での値を○でプロットしてある。また、１２１Ａ１は各色差微分値の度数分布、１２１Ａ２は、色差値の度数分布である。欠陥部と正常部が明らかに分かれている場合にはしきい値１１４は１２（Ａ）に示すように、１２１Ａ１と１２１Ａ２の度数分布の谷のピーク値を通り、プロットされた点の慣性主軸１２１Ａ３に垂直な直線１２１Ａ４が示す値とする。また、欠陥部と正常部が分離していない場合、すなわち、度数分布の谷のピークがない場合には、しきい値を示す直線は１２１Ｂ４となる。つまり、全ての欠陥候補領域を欠陥として検出し、見落とし、見逃しがないようにする。
【００３９】
更に、欠陥と見なす最低寸法や、面積等が決まっている場合には、２０６にて検出された欠陥領域の形状を２０７で計測する。画像で寸法計測を行う場合、同じものを撮像しても、その倍率や、焦点が合っているか否かで画像として取り込んだ時のサイズが異なる。そこで、このような撮像条件が変わったときのスケールを教示するための寸法ティーチングをあらかじめ、行っておく。まず、実寸が既知のカメラ校正用パターンをカメラで撮像する。パターンは、線、円など検出したい欠陥の形状に近いものにする。コンピュータ１４では、撮像したカメラ校正用パターンの画素サイズと、実寸からスケール（１画素の大きさ）を算出し、保持しておく。そして、検出された欠陥領域の実寸法などを、保持されているスケールを用いて計測し、欠陥と見なす最低寸法より大きい場合や縦横比などの形状が規定値を越えていれば欠陥として検出する。欠陥が検出された場合には、塗装表面の塗り直し等の対策を行う。
【００４０】
検出結果は表示装置１６に表示され、出力装置１７から出力される。その一例を図１３に示す。１３１aは検査対象物の原画像データ、１３１bは検出結果、１３１cは欠陥の検出された位置を通し番号で示したものである。また、１３１dは１３１cで示した番号に対応した欠陥の詳細データである。位置、寸法や面積等の形状、色度、欠陥レベルなど、各評価方法に応じた情報を欠陥毎に出力する。また、これらの検査結果は記憶装置１８に格納される。その例を図１４に示す。検査対象範囲が広い場合、いくつかの画像に分割して撮像、検査を行うが、この時の分割は、１４０１に示すように、撮像範囲が重なるようになっている。１４０１ａ、１４０１ｂ、１４０１ｃは、１４０１の分割画像であるが、それぞれに対して検査が行われる。その結果は、１４０２に示すように検査対象品毎に全画像の情報が１つにまとまって格納される。
【００４１】
さらに、図１３で示した個々の画像毎の詳細データは、１４０３にあるように１４０２の下に格納されるので、ユーザは、まず、製品毎のデータ１４０２を調べ、更に欠陥のあった部分について１４０３の詳細に見たい場合は、その製品のID番号と、画像NOから、必要データのみを引き出すことができる。
【００４２】
本発明は以上に説明したように構成され、塗装表面を自動で検査し、定量的に評価するので、画像及び定量的な数値データを検査データとして得ることができる。これにより得られたデータを、品質保証用のデータとして出荷製品に添付することが可能となる。また、経時変化等による劣化の検査が必要な塗装面に対しては、定期的に検査を行い、図１３に示したデータを時系列で保持し、これらを比較することにより品質管理を行うこともできる。すなわち、品質の変化を画像及びデジタルデータで保持、チェックすることにより、欠陥が致命的なものになる前に、塗り直しなどの対策が可能となる。
【００４３】
次に、第２の実施例として、鋳物の表面欠陥検査の１つである浸透探傷試験の自動評価を例にとって説明する。浸透探傷試験では、割れなどの表面欠陥部分に入り込んだ赤色、もしくは蛍光緑などの浸透液を白色の現像液で浮き出させて、目視によりその色の濃さ等から欠陥を検出する。浸透液が浮き出てきた部分の中には、表面に残っていた浸透液がにじみ出てきたものもあり、これは疑似欠陥と呼ばれ、真の欠陥ではない。このため、自動で検査する場合、検査対象面を、疑似欠陥部を含めた正常部と欠陥部のどちらか２つのモードに識別しなければならない。そこで、浸透液が赤色の場合の検査手順を図１５を用いて説明する。まず、浸透探傷試験面をカメラで撮像し、ＲＧＢデータとして画像入力ボードを介してコンピュータへ取り込む（１５１）。コンピュータでは、得られたＲＧＢデータを色度・輝度に変換する（１５２）。
【００４４】
ここで、従来、人間は試験面を検査する際に、周囲の白色部に対する赤色の濃さや輪郭の強さから欠陥と疑似欠陥を識別している。そこで、本発明による自動検査においても、各画像毎に基準となる白色を決定する（１５３）。決定の仕方は、前述の図７で説明した通り、画像中の色度分布から各色度値をとる画素数をカウントし、最も画素数の多かった色度値を基準白色とする。そして、基準白色に対する画像上の各位置での色相、色差を算出する（１５４）。これらの色相、色差から、疑似欠陥も含めた欠陥候補領域を抽出する（１５５）。
【００４５】
欠陥、疑似欠陥は、浸透液が赤色の場合、色度図上でピンク、赤へかけての領域の値をとる。そこで、基準白色に対する色相領域の指定は、図１６の領域１６１となるようなθ１、θ２を入力する。色差の値は、サンプルデータから、見逃しがないように、疑似欠陥も含めた範囲となるように指定する。このように疑似欠陥も含めた範囲を指定し、欠陥候補領域を抽出した後、色差の微分値で欠陥のみを検出する（１５６）。そして、検出された領域について、面積、縦横比、寸法など形状計測を行い（１５７）、浸透探傷試験の判定基準で規定されている寸法や形状に該当するもののみを真の欠陥部として検出し、表示装置、出力装置に出力する（１５８）。画像データ、検査データなどは、記憶装置に格納される（１５９）。各手順の詳細は、塗装面の欠陥検査と同様である。
【００４６】
このようにして、色調の違いにより２つのモードに識別する本発明により、浸透探傷試験像を正常部分（白色部分及び疑似欠陥部分）と欠陥部分の２つのモードに自動で識別する。これにより、表面欠陥の浸透探傷試験像の評価を自動で行うことができる。
【００４７】
第３の実施例として、２種の液体を撹拌させた時の撹拌の進行度チェックを例に説明すると、撹拌する２種以上の液体の色が異なっている場合、それらが十分に混ざり合っているか否かは液体の色合いで検査する。
【００４８】
図１７は、２色の液体の攪拌を例にとって、１７１ａと１７１ｂを攪拌し、混ざり合った部分を１７１ｃで示している。１７１ａと１７１ｂの混ざり具合のチェックは、図の色が、１７１ｃであるか否かの２つのモードに識別することで行う。その手順を、図１８を用いて説明する。
【００４９】
まず、Ａ色とＢ色の液体を攪拌している像をカメラで撮像し、ＲＧＢデータとして画像入力ボードを介してコンピュータへ取り込む（１８１）。コンピュータでは、得られたＲＧＢデータを色度・輝度に変換する（１８２）。この色度分布の中から、Ａ色とＢ色が混ざり合ってできた色Ｃを、基準色として算出し（１８３）、基準色に対する画像上の各位置での色相、色差を算出する（１８４）。これらの色相、色差から色Ｃでない色領域、すなわち混ざり合っていない領域（未混合色領域）を抽出する（１８５）。そして、基準色、すなわち混合色Ｃと、Ｃでない色との２つのモードに得られた像を識別し、その割合を算出する（１８６）。この割合により、攪拌の進行度をチェックする。これをリアルタイムで繰り返し、十分に攪拌されていれば、攪拌を停止する。
【００５０】
このようにして、色調により２つのモードに識別する本発明により、２色以上の液体の攪拌中の色を、混合済みの色とそれ以外の色の２つのモードに自動で識別する。これにより、攪拌の進行度を自動でチェックすることができる。
【００５１】
第４の実施例として、ＣＲＴや液晶、プラズマディスプレイなどの電子ディスプレイの発光状態の検査を例にとって説明する。電子ディスプレイの発光状態検査では、表示管面全体に特定の一色（特に白）を表示し、その均一性を評価するものである。
【００５２】
すなわち、図１９に示すように、その一部が赤みがかっていたり、黄色くなっているなどの全体の色調とは微妙に異なる色のついた領域を色むらとして検出する。このため、自動で検査する場合、検査対象面であるディスプレイの発光管面を、均一発光部と色むら部のどちらか２つのモードに識別しなければならない。
【００５３】
検査装置の構成の一例を図２０に示すと、構成はほぼ図１と同様であるが、検査対象物である電子ディスプレイ２１１は自発光物なので、照明手段は必要ない。
【００５４】
その検査手順を図２を用いて説明すると、発光管面をカメラで撮像し、ＲＧＢデータとして画像入力ボードを介してコンピュータへ取り込む（２０１）。コンピュータでは、得られたＲＧＢデータを色度・輝度に変換する（２０２）。ここで、人間が発光状態を検査する際には、むらの少ない管面中央部との比較により、周辺部の評価を行っている。また、周りとの色度のコントラストによりむらを検出している。
【００５５】
そこで、本発明による自動検査においても、各対象ディスプレイ毎に基準となる色を決定する（２０３）。決定の仕方は、人間が管面中央との比較により評価を行うのと同様に、管面中央付近の平均色度で求める。そして、基準色に対する管面各位置での色相、色差を算出（２０４）し、これらの色相、色差から、色むら候補領域を抽出する（２０５）。色むらは任意の色なので、抽出する色相の範囲θ１、θ２を様々に変えていき、あらゆる色合いを検出できるようにする。色差の値ｒ１、ｒ２は、合格限度値などを指定する。
【００５６】
ここで抽出された色むら候補領域の中には、人間にとってはむらと感じないものも含まれている。例えば、管面中央から周辺にかけて徐々に色度が変化していくような場合、人間はあまり、色度の変化を感じない。逆に、周りとの色度の変化が急激な時、すなわちコントラストが強い時、むらが強調されて見える。そこで、抽出された各色むら候補領域について、色差の微分値を算出し（２０６）、その値が大きい領域のみを欠陥として検出する。そして、検出された領域について、形状計測を行い（２０７）、判定基準値以上の大きさを持つ領域のみを色むら部として検出し、表示装置、出力装置に出力する（２０８）。画像データ、検査データなどは、記憶装置に格納される（２０９）。
【００５７】
各手順の詳細は、塗装面の欠陥検査と同様である。このようにして、色調の違いにより２つのモードに識別する本発明により、電子ディスプレイの発光管面を正常部分と色むら部分の２つのモードに自動で識別する。これにより、発光状態の色均一性の評価を自動で行うことができる。
【００５８】
以上、本発明の実施例について説明したが、本発明は、上記説明した実施例に限定されるものではなく、例えば物体表面の汚れの検査や、物体表面の色調の度合いの検査など、色と欠陥との間に相関関係があるものであれば、広く応用することができる。
【００５９】
【発明の効果】
本発明は、以上説明したように構成されているので以下に記載されるような効果がある。
【００６０】
即ち、本発明によれば、従来は作業者の目視による官能的な検査に頼っていた色に基づく表面欠陥の検査を、自動化することにより定量的な評価を行えるようになり、安定した欠陥検査が行えるようになった。
【００６１】
即ち、本発明によれば、色の違いから、検査対象物をカラー画像として取り込み、その画像データから色度・輝度を算出し、算出した色度の分布とその形状から欠陥のみを検出することにより、塗装むらや傷などの表面欠陥を検出する表面検査を自動化することができ、欠陥の定量的な評価が可能になった。これにより、表面検査の高効率化と信頼性の向上を図ることができるようになった。
【００６２】
また、あらかじめ、検査対象物の色度を高精度に計測するためのキャリブレーションと、面積、寸法を精度良く計測するためのティーチングを行うことにより、カメラで撮像した検査対象物のカラー画像データを用いて、高精度な色度測定と、形状計測が行えるようになった。
【００６３】
さらに、得られた色度分布から、基準となる色を自動で算出、それに対する色相、色差、及び色差の変化量を求めることにより、微妙な色調の違いを検出することができ、これにより、検査対象物を正常部と異常部とに自動で識別することが可能となった。
【００６４】
さらに、このように自動で識別することにより、表面の浸透探傷試験像の自動評価を可能とする。また、2色以上の液体の攪拌の進行度を自動でチェックすることができる。また、ＣＲＴや液晶などの電子ディスプレイの発光状態の均一性を自動評価できる。
【００６５】
さらに、画像データ及び識別結果の数値データを格納、保持することにより、欠陥検査においては、品質保証用のデータとして出荷製品に添付することが可能となる。また、経時変化等による劣化の検査が必要な対象物に対しては、定期的に検査を行い、データを時系列で保持し、これらを比較することにより品質管理を行うこともできる。これにより、欠陥などが致命的なものになる前に、対策が可能となる。
【００６６】
さらに、攪拌の進行度の自動チェックにより、攪拌装置のモニタリングが可能となる。
【００６７】
さらに、電子ディスプレイの均一性の定量評価による品質向上の実現が可能となる。
【図面の簡単な説明】
【図１】本発明の塗装面の表面検査装置の構成の一例を示す図である。
【図２】本発明の塗装面の表面検査装置の動作手順の例を示すフローチャート図である。
【図３】本発明の塗装面の表面検査装置において、偏光フィルタを用いた例を示す図である。
【図４】ＣＩＥ（国際照明委員会）の規定するｘｙ色度図である。
【図５】本発明のカラーキャリブレーション及び寸法ティーチングに用いるカメラ校正パレットの一例を示す図である。
【図６】本発明のカラーキャリブレーションの動作手順の例を示すフローチャート図である。
【図７】本発明の塗装面の表面検査装置において、基準となる色度を画像中から算出する方法を示す図である。
【図８】本発明の本発明の塗装面の表面検査装置において、色相画像の算出方法を示す図である。
【図９】本発明の本発明の塗装面の表面検査装置において、色差画像の算出方法を示す図である。
【図１０】発明の本発明の塗装面の表面検査装置において、色相と色差から欠陥候補領域を抽出する方法の例を示す図である。
【図１１】発明の本発明の塗装面の表面検査装置において、抽出された欠陥候補領域から、色差の変化量により欠陥領域のみを検出する例を示す図である。
【図１２】発明の本発明の塗装面の表面検査装置において、欠陥領域のみを検出する際の、２値化しきい値の決定方法の例を示す図である。
【図１３】本発明の検査結果の出力例を示す図である。
【図１４】本発明の検査結果の格納例を示す図である。
【図１５】本発明の表面の浸透探傷試験像の検査の処理手順の例を示すフローチャート図である。
【図１６】本発明の表面の浸透探傷試験像の検査において、ＣＩＥ色度図上での計測色度の範囲を示す図である。
【図１７】攪拌中の液体の例を示す図である。
【図１８】本発明の攪拌の進行度チェックの処理手順の例を示すフローチャート図である。
【図１９】電子ディスプレイの発光管面の一例を示す図である。
【図２０】本発明の電子ディスプレイの発光状態均一性評価装置の構成の一例を示す図である。
【符号の説明】
１１・・・検査対象物、１２・・・照明手段、１３・・・カメラ、１４・・・コンピュータ、１５・・・画像入力ボード、１６・・・表示装置、１７・・・出力手段、１８・・・記憶装置、３１・・・偏光フィルタ、５１・・・カメラ校正用パレット、５２・・・色彩計。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a method and apparatus for inspecting a defect of a sample, and more particularly to a defect inspection method and apparatus optimal for detecting a defect by a color difference such as a painted surface or a casting surface.
[0002]
[Prior art]
The surface inspection of the painted surface will be described as an example. Inspection of defects such as uneven coating, faintness, unpainted surface, and scratches on the painted surface is performed visually. Specifically, human beings check whether there is a color different from the paint color such as an uneven portion or a ground color with respect to the color tone of the entire inspection surface.
[0003]
Also, the surface defect inspection of castings will be described as an example. In the surface defect inspection by the penetrant flaw detection test, a white developer is sprayed on the entire inspection surface, and the defect portion is determined based on the color density and shape of the penetrating liquid that has oozed out. Is detected by visual inspection of the inspector.
[0004]
[Problems to be solved by the invention]
The method of inspecting color unevenness and color difference, which is the above inspection method, is inefficient and inevitable due to human error such as oversight and variations in inspection results due to individual differences, and defect Dimension measurement is also ambiguous, which causes a decrease in the reliability of inspection results.
[0005]
On the other hand, when trying to measure colors digitally, it is common to use a color meter, but since the color meter is a spot measurement, it is impossible to measure at a multipoint, and a wide range such as surface inspection. It is not suitable for accurate measurement.
[0006]
In order to solve the problems of the conventional inspection technique, the object of the present invention is to image the inspection surface with a camera capable of color measurement on the surface, measure the color from the obtained image data, and An object of the present invention is to provide an inspection method and apparatus for automatically detecting defects such as coating unevenness and scratches on an inspection surface accurately by detecting different portions and measuring their shapes, and performing quantitative evaluation. Furthermore, the present apparatus is introduced into a production line and the like, and surface inspection is automatically performed to provide a high-quality product and to construct an efficient and highly reliable inspection system. It is another object of the present invention to provide an inspection method for measuring subtle differences in color tone from the entire color image of an inspection object and automatically identifying the inspection object in two modes based on the difference in color.
[0007]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a defect inspection apparatus for detecting surface defects such as coating unevenness and scratches from the difference in color, a means for capturing an inspection object as a color image, and a color from the captured image data. Chromaticity conversion means for calculating degree / luminance, means for detecting a defect candidate part from the calculated chromaticity distribution, means for measuring the shape of the defect candidate part and detecting only the defect from the shape, and detection thereof And a means for outputting and storing the result.
[0008]
In addition, in order to remove the reflected light of the inspection object due to illumination, improve the S / N ratio, and obtain an image with less noise, a polarization filter is inserted into the illumination light source and camera, and the optimal orientation of the polarization filter is imaged. It is further provided with a means for automatically adjusting while taking the correspondence with.
[0009]
Further, in order to detect a difference in color tone in the color image data of the inspection object and discriminate the inspection object into two modes of a normal part and an abnormal part, it is necessary to measure a subtle color difference. Therefore, color calibration means for measuring the chromaticity of the inspection object with high accuracy is provided. In addition, a dimensional teaching means for accurately measuring the area and shape of the region identified from the image of the inspection object is provided.
[0010]
Further, in order to detect a difference in color tone and discriminate an inspection object into two modes, ie, a normal part and an abnormal part, a means for measuring the difference in color tone based on a hue and a color difference with respect to a reference color, and a reference There is further provided means for measuring a difference in color tone based on a change amount of the color difference with respect to the color. In addition, a means for discriminating the inspection object into two modes based on the results measured from these two means is provided. Further, there is provided means for automatically calculating a reference color from the captured image data when a portion having a different color tone is extracted by comparison with the reference color.
[0011]
That is, in the present invention, in order to achieve the above-described object, a defect inspection apparatus for inspecting the surface defect of the sample by detecting the color of the surface of the sample, image capturing means for capturing the sample and capturing it as a color image; A defect candidate portion is extracted from the chromaticity / luminance calculation means for calculating chromaticity / luminance from the color image data captured by the image capturing means, and the chromaticity / luminance distribution calculated by the chromaticity / luminance calculation means. Defect candidate extracting means, defect detecting means for measuring the shape of the defect candidate portion extracted by the defect candidate extracting means and detecting only the defect from the shape, and output means for outputting the detection result detected by the defect detecting means And storage means for storing the detection results detected by the defect detection means.
[0012]
Further, in the present invention, an illuminating unit that illuminates the sample with polarized light, an imaging unit that captures an image of the sample illuminated by the illuminating unit through a polarizing filter to obtain a color image of the sample, and the imaging unit captures the image. A defect inspection apparatus is configured to include a defect detection unit that detects a defect of the sample based on the color image of the sample and an output unit that outputs a result detected by the defect detection unit.
[0013]
Furthermore, in the present invention, a defect inspection method for detecting the surface defect of the sample by detecting the color of the surface of the sample, obtaining a color image by imaging the sample, and calculating the chromaticity / luminance from the obtained color image data. The defect candidate part is extracted from the calculated chromaticity / luminance distribution, the shape of the extracted defect candidate part is measured, only the defect is detected from the shape, and information on the detected defect is displayed on the screen. It was made to display by displaying.
[0014]
Furthermore, in the present invention, the sample is illuminated with polarized light, the illuminated sample is imaged through a polarizing filter to obtain a color image of the sample, and a defect in the sample is detected based on the color image of the sample. This detection result is output for output.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS.
[0016]
As a first embodiment, a case where a surface inspection of a coated surface is detected in which a portion that is different from the surrounding color due to uneven coating, unpainted, scratches, etc. is detected as a defect in a uniformly coated surface. To do.
[0017]
FIG. 1 shows an example of the configuration of an apparatus according to the present invention. According to this, 11 is an inspection object whose surface is painted, 12 is an illumination means for illuminating the inspection object, 13 is an imaging means for capturing RGB signals of the inspection object using a CCD camera or the like, and 14 is an image input board 15. A computer for measuring the color of an inspection object and detecting a defect; 16 a display means such as a display for displaying the processing result; 17 an output means such as a printer for outputting the processing result; and 18 a processing result. Is a storage device for storing and saving.
[0018]
FIG. 2 shows an example of the operation according to the present invention. First, the inspection object 11 illuminated by the illumination 12 is imaged by the camera 13, and is taken into the computer 14 as RGB data via the image input board 15 (201). The computer 14 converts the obtained RGB data into chromaticity / luminance (202). From this chromaticity distribution, a normal color, that is, a color that is not a defective portion is calculated as a reference color (203), and a hue and a color difference at each position on the image with respect to the reference color are calculated (204). From these hues and color differences, an abnormal color area, that is, a defect candidate area is extracted (205). For each region extracted as a defect candidate, a change amount of the color difference from the surrounding area, that is, a differential value of the color difference is obtained (206), and shape measurement such as area, aspect ratio, and dimension is performed (207). Then, only a region having a large amount of change in color difference from the surroundings and having a size / area larger than a specified value is detected as a true defect portion and output to the display device 16 and the output device 17 (208). Image data, inspection data, and the like are stored in the storage device 18 (209).
[0019]
Next, each operation will be described in detail. When a color image of the inspection object 201 is captured, if the object is a metal or the like, regular reflection due to illumination occurs in a thin paint part or a curved part, and the image data of that part is saturated in the obtained image. Value. In order to remove such reflected light and increase the S / N ratio of the image, a polarizing filter 31 is attached to both the CCD camera 13 and the illumination device 12 as shown in FIG. It automatically adjusts while examining the image in the minimum orientation, and captures the image with the best S / N ratio.
[0020]
In color inspection, it is necessary to quantitatively evaluate the color. Therefore, in 202, the RGB data of the captured color image is converted into chromaticity x, y and luminance Y defined by the CIE (International Lighting Commission), and inspection is performed using these. A representation of the chromaticity x, y in two-dimensional orthogonal coordinates is called a chromaticity diagram and is shown in FIG. In the chromaticity diagram, each color is arranged around white as a center, and each color becomes brighter as it goes away from white. Hereinafter, the hue is called hue, the vividness of each color is called saturation, and the distance between two chromaticity values on the chromaticity diagram is called color difference. In this method, in order to convert RGB data into chromaticity x, y, and luminance Y with high accuracy, color calibration is performed in advance using a camera calibration palette 51 as shown in FIG. The flow of the process is shown in FIG.
[0021]
The camera calibration palette 51 is painted with three or more colors (4 in 51). This is imaged by the CCD camera 13 (601), and the RGB value of each color is calculated (602). Further, the chromaticity x and y and the luminance y are measured by the colorimeter 52 (603). Here, the relationship between the RGB value and the xyY value is expressed by (Equation 1) and (Equation 2).
[0022]
[Expression 1]

[0023]
[Expression 2]

[0024]
Therefore, the xyY value is calculated by substituting the RGB value of each color captured from the camera into (Equation 1) and (Equation 2), and this value matches the xyY value measured by the colorimeter. ₁₁ ~ A ₃₃ If this is obtained, a conversion parameter unique to the camera is obtained. Since there are nine unknown parameters, RGB values (R ₁ , G ₁ , B ₁ ) ~ (R _Three , G _Three , B _Three ) And the corresponding xyY value (x ₁ , Y ₁ , Y ₁ ) To (x _Three , Y _Three , Y _Three ) To calculate the parameters.
[0025]
From (Equation 2), XYZ can be calculated by (Equation 3) from the xyY value.
[0026]
[Equation 3]

[0027]
The xyY values of the three colors of the colorimeter are substituted into (Equation 3) to obtain XYZ, and are substituted into (Equation 1) ((Equation 4)).
[0028]
[Expression 4]

[0029]
As a result, the camera-specific conversion parameter a ₁₁ ~ A ₃₃ (604) and an xyY value equal to the colorimeter value can be obtained from the RGB values of the camera.
[0030]
After converting the RGB value obtained from the camera into an xyY value using the conversion parameter specific to the camera calculated in advance by calibration and calculating the chromaticity distribution in the image, in 203, the normal value from the image is obtained. The chromaticity value, that is, the chromaticity of the portion that is not defective is calculated as a reference value. First, as shown in the graph in FIG. 7, the chromaticity x, y of each pixel in the image is examined, the number of pixels taking each x, y value is counted, and a two-dimensional frequency distribution is created. Then, the x and y values having the largest number of pixels in the image are examined. Since most of the image is a non-defective part, the x and y values of the peak value of the two-dimensional frequency distribution are examined, and as shown in the lower graph of FIG. Let the y value be the xy chromaticity value of the reference color.
[0031]
In 204, the hue and color difference at each position on the image with respect to the reference color are calculated. If the chromaticity of the reference color is (xc, yc) and the chromaticity at the position (i, j) on the image is (xij, yij), the hue at the position (i, j) is as shown in FIG. The calculation is performed in the direction with respect to the reference color on the chromaticity diagram. The calculation formula is shown in (Formula 5).
[0032]
[Equation 5]

[0033]
Further, as shown in FIG. 9, the color difference at the position (i, j) is calculated by the distance from the reference color on the chromaticity diagram. The calculation formula is shown in (Formula 6).
[0034]
[Formula 6]

[0035]
From the hue and color difference at each position of the image with respect to the reference color calculated as described above, the hue to be detected as a defect in the hue is limited as shown in FIG. 10 (in the figure, the hue θ is θ1 ≦ θ ≦ and the degree of color difference from the reference color is limited by the color difference (in the figure, the color difference d is in the range of d1 ≦ d ≦ d2). Then, a portion within this range is extracted as a defect candidate region.
[0036]
Thus, some defect candidates obtained by limiting the range by hue and color difference do not need to be detected as defects. For example, a person whose chromaticity gradually changes with respect to the reference color is not much of a concern for humans. Rather, a portion in which the chromaticity has suddenly changed relative to the surrounding color or a region with a clear outline is a considerably conspicuous defect. Thus, when a gentle color change with respect to the surrounding colors is regarded as a normal part and only a sudden change is regarded as a defect, in 206, the amount of change in color difference from the surrounding is determined for the defect candidate area. Only those having a certain value or more are determined as defects.
[0037]
Referring to FIG. 11, (A) is a defect candidate area extracted in 205. 112 'in (B) is a color difference on 112 in (A). Further, 113 of (C) is a color difference change amount at each position on 112, that is, a derivative of 112 ′. In this way, when the amount of change in color difference from the surroundings is small, the differential value is also small. Therefore, as shown in (C), only those having a differential value larger than a certain fixed value 114 are set as defect areas. As a result, only a region having a large color difference and a large color change amount, that is, a sharp outline, as shown in FIG.
[0038]
Next, a method for determining the threshold value 114 will be described with reference to FIG. In the graph of FIG. 12A, the vertical axis represents the maximum value of the color difference in each defect candidate area extracted by hue and color difference, and the horizontal axis represents the maximum value of the color difference differential value of the contour portion of each defect area. The value in the defect area is plotted with x, and the value in the non-defect area is plotted with ◯. 121A1 is a frequency distribution of each color difference differential value, and 121A2 is a frequency distribution of color difference values. When the defective part and the normal part are clearly separated, the threshold value 114 passes through the peak values of the valleys of the frequency distributions of 121A1 and 121A2, as indicated by 12 (A), and the inertial principal axis 121A3 of the plotted point The value is indicated by a straight line 121A4 perpendicular to. Further, when the defective portion and the normal portion are not separated, that is, when there is no peak of the valley of the frequency distribution, the straight line indicating the threshold value is 121B4. That is, all defect candidate areas are detected as defects so as not to be overlooked and overlooked.
[0039]
Further, when the minimum dimension, area, etc., to be regarded as a defect are determined, the shape of the defect area detected at 206 is measured at 207. When measuring the dimensions of an image, even if the same image is captured, the size when captured as an image differs depending on the magnification and whether the image is in focus. Therefore, dimensional teaching for teaching the scale when such imaging conditions change is performed in advance. First, a camera calibration pattern whose actual size is known is imaged by the camera. The pattern should be close to the shape of the defect to be detected, such as a line or circle. The computer 14 calculates a scale (size of one pixel) from the pixel size and the actual size of the captured camera calibration pattern, and holds them. Then, the actual size of the detected defect area is measured using a held scale, and if it is larger than the minimum size considered as a defect or the shape such as the aspect ratio exceeds the specified value, it is detected as a defect. . If a defect is detected, take measures such as repainting the painted surface.
[0040]
The detection result is displayed on the display device 16 and output from the output device 17. An example is shown in FIG. 131a is the original image data of the inspection object, 131b is the detection result, and 131c is the serial number indicating the position where the defect is detected. 131d is detailed data of defects corresponding to the numbers indicated by 131c. Information corresponding to each evaluation method, such as position, shape such as dimension and area, chromaticity, and defect level, is output for each defect. These inspection results are stored in the storage device 18. An example is shown in FIG. When the inspection target range is wide, imaging and inspection are performed by dividing the image into several images. In this division, the imaging ranges overlap as shown by 1401. Reference numerals 1401a, 1401b, and 1401c are 1401 divided images, and each of them is inspected. As a result, as shown in 1402, information on all the images is stored together for each inspection object.
[0041]
Further, since the detailed data for each image shown in FIG. 13 is stored under 1402 as in 1403, the user first checks the data 1402 for each product, and further on the defective part. When looking at the details of 1403, only necessary data can be extracted from the ID number of the product and the image NO.
[0042]
Since the present invention is configured as described above and automatically inspects and quantitatively evaluates the coating surface, an image and quantitative numerical data can be obtained as inspection data. The data thus obtained can be attached to the shipped product as quality assurance data. In addition, for coated surfaces that need to be inspected for deterioration due to changes over time, etc., periodic inspection is performed, and the data shown in FIG. 13 is maintained in time series, and quality control is performed by comparing them. You can also. That is, by retaining and checking the change in quality with images and digital data, it is possible to take measures such as repainting before the defect becomes fatal.
[0043]
Next, as a second embodiment, an automatic evaluation of a penetration flaw detection test which is one of surface defect inspections for castings will be described as an example. In the penetrant flaw detection test, a penetrant such as red or fluorescent green that has entered a surface defect portion such as a crack is raised with a white developer, and the defect is detected from the color density by visual inspection. Some of the parts where the penetrant liquid has come out have oozed out the penetrant liquid remaining on the surface, which is called a pseudo defect and is not a true defect. For this reason, when inspecting automatically, the surface to be inspected must be identified into two modes, either a normal part including a pseudo defect part or a defect part. Therefore, the inspection procedure when the penetrant is red will be described with reference to FIG. First, the penetrant flaw detection test surface is imaged with a camera, and is captured as RGB data into a computer via an image input board (151). The computer converts the obtained RGB data into chromaticity / luminance (152).
[0044]
Here, conventionally, when a human inspects a test surface, he / she distinguishes a defect from a pseudo defect based on the intensity of red and darkness with respect to the surrounding white portion. Therefore, also in the automatic inspection according to the present invention, the reference white color is determined for each image (153). As described above with reference to FIG. 7, the number of pixels taking each chromaticity value is counted from the chromaticity distribution in the image, and the chromaticity value having the largest number of pixels is set as the reference white. Then, the hue and color difference at each position on the image with respect to the reference white color are calculated (154). A defect candidate region including a pseudo defect is extracted from these hues and color differences (155).
[0045]
When the penetrant is red, the defect and the pseudo-defect take values in the region from pink to red on the chromaticity diagram. Therefore, to specify the hue region for the reference white, θ1 and θ2 that are the region 161 in FIG. 16 are input. The color difference value is specified from the sample data so as to be in a range including pseudo defects so as not to be overlooked. Thus, after specifying the range including the pseudo defect and extracting the defect candidate area, only the defect is detected by the differential value of the color difference (156). Then, shape measurement such as area, aspect ratio, and dimensions is performed on the detected area (157), and only those that correspond to the dimension and shape defined by the determination standard of the penetrant flaw detection test are detected as true defects. The data is output to the display device and the output device (158). Image data, inspection data, and the like are stored in the storage device (159). The details of each procedure are the same as the defect inspection of the painted surface.
[0046]
In this way, according to the present invention for discriminating between two modes according to the difference in color tone, the penetration flaw detection test image is automatically discriminated into two modes of a normal portion (white portion and pseudo-defect portion) and a defective portion. Thereby, the evaluation of the penetration inspection image of the surface defect can be automatically performed.
[0047]
As a third example, an example of checking the progress of stirring when two kinds of liquids are stirred will be described. If the colors of two or more kinds of liquids to be stirred are different, they are sufficiently mixed. It is inspected by the color of the liquid.
[0048]
FIG. 17 shows a mixed portion 171c where 171a and 171b are stirred and a mixture of two colors of liquid is taken as an example. The check of the degree of mixing of 171a and 171b is performed by discriminating into two modes of whether the color of the figure is 171c or not. The procedure will be described with reference to FIG.
[0049]
First, an image in which liquids of A color and B color are being agitated is picked up by a camera and captured as RGB data into a computer via an image input board (181). The computer converts the obtained RGB data into chromaticity / luminance (182). From this chromaticity distribution, a color C formed by mixing the A and B colors is calculated as a reference color (183), and the hue and color difference at each position on the image with respect to the reference color are calculated (184). ). From these hues and color differences, a color area that is not color C, that is, a non-mixed area (unmixed color area) is extracted (185). Then, the image obtained in the two modes of the reference color, that is, the mixed color C and the color other than C is identified, and the ratio is calculated (186). The progress of stirring is checked by this ratio. This is repeated in real time, and if it is sufficiently stirred, stirring is stopped.
[0050]
In this way, according to the present invention for discriminating between two modes according to the color tone, the stirring color of two or more liquids is automatically discriminated into two modes of mixed color and other colors. Thereby, the progress of stirring can be automatically checked.
[0051]
As a fourth embodiment, a description will be given of an inspection of the light emission state of an electronic display such as a CRT, a liquid crystal, or a plasma display. In the light emission state inspection of an electronic display, a specific color (particularly white) is displayed on the entire display tube surface, and its uniformity is evaluated.
[0052]
That is, as shown in FIG. 19, a region having a color slightly different from the overall color tone, such as a part thereof being reddish or yellowish, is detected as color unevenness. For this reason, when inspecting automatically, the light emitting tube surface of the display, which is the inspection target surface, must be identified as one of the two modes of the uniform light emitting portion and the uneven color portion.
[0053]
FIG. 20 shows an example of the configuration of the inspection apparatus. The configuration is almost the same as that in FIG. 1, but the electronic display 211 that is the inspection target is a self-luminous object, so that no illumination means is necessary.
[0054]
The inspection procedure will be described with reference to FIG. 2. The arc tube surface is imaged by a camera, and is captured as RGB data into a computer via an image input board (201). The computer converts the obtained RGB data into chromaticity / luminance (202). Here, when a human inspects the light emission state, the peripheral portion is evaluated by comparison with the central portion of the tube surface with little unevenness. Further, unevenness is detected based on the contrast of chromaticity with the surroundings.
[0055]
Therefore, in the automatic inspection according to the present invention, a reference color is determined for each target display (203). The method of determination is obtained by the average chromaticity near the center of the tube surface, in the same manner as when a human performs evaluation by comparison with the center of the tube surface. Then, the hue and color difference at each position on the tube surface with respect to the reference color are calculated (204), and an uneven color candidate region is extracted from these hue and color difference (205). Since the color unevenness is an arbitrary color, the hue ranges [theta] 1 and [theta] 2 to be extracted are variously changed so that all hues can be detected. The color difference values r1 and r2 specify a pass limit value or the like.
[0056]
Some of the color unevenness candidate areas extracted here include those that humans do not feel uneven. For example, when the chromaticity gradually changes from the center of the tube surface to the periphery, humans do not feel much change in chromaticity. On the other hand, when the change in chromaticity with the surroundings is abrupt, that is, when the contrast is strong, unevenness appears to be emphasized. Therefore, a differential value of the color difference is calculated for each extracted color unevenness candidate region (206), and only a region having a large value is detected as a defect. Then, the shape of the detected region is measured (207), and only a region having a size equal to or larger than the determination reference value is detected as a color unevenness portion and is output to the display device and the output device (208). Image data, inspection data, and the like are stored in the storage device (209).
[0057]
The details of each procedure are the same as the defect inspection of the painted surface. In this way, according to the present invention for discriminating between two modes according to the difference in color tone, the light emitting tube surface of the electronic display is automatically discriminated into two modes of a normal portion and an uneven color portion. Thereby, evaluation of the color uniformity of the light emission state can be performed automatically.
[0058]
As described above, the embodiments of the present invention have been described. However, the present invention is not limited to the above-described embodiments. For example, the color of the object surface is inspected for dirt or the surface tone of the object. If there is a correlation with the defect, it can be widely applied.
[0059]
【The invention's effect】
Since the present invention is configured as described above, the following effects can be obtained.
[0060]
That is, according to the present invention, it has become possible to perform quantitative evaluation by automating the inspection of surface defects based on colors, which has conventionally relied on the sensual inspection by the operator's visual inspection, and stable defect inspection. Can be done.
[0061]
That is, according to the present invention, an inspection object is captured as a color image from the difference in color, chromaticity / luminance is calculated from the image data, and only a defect is detected from the calculated chromaticity distribution and its shape. This makes it possible to automate surface inspections that detect surface defects such as coating unevenness and scratches, and enables quantitative evaluation of defects. As a result, the efficiency of the surface inspection and the improvement of the reliability can be achieved.
[0062]
In addition, by performing calibration for measuring the chromaticity of the inspection object with high accuracy and teaching for measuring the area and dimensions with high accuracy, the color image data of the inspection object captured by the camera is obtained. It is now possible to perform highly accurate chromaticity measurement and shape measurement.
[0063]
Furthermore, from the obtained chromaticity distribution, it is possible to detect a subtle difference in color tone by automatically calculating a reference color, obtaining a hue, a color difference, and a change amount of the color difference with respect to the reference color, The inspection object can be automatically identified as a normal part and an abnormal part.
[0064]
Furthermore, by automatically identifying in this way, it is possible to automatically evaluate the surface penetrant inspection image. In addition, it is possible to automatically check the progress of stirring of liquids of two or more colors. In addition, the uniformity of the light emission state of an electronic display such as a CRT or a liquid crystal can be automatically evaluated.
[0065]
Furthermore, by storing and holding the image data and the numerical data of the identification result, it becomes possible to attach to the shipped product as quality assurance data in the defect inspection. In addition, for an object that needs to be inspected for deterioration due to changes over time, quality control can be performed by periodically inspecting, holding data in time series, and comparing them. As a result, measures can be taken before defects and the like become fatal.
[0066]
Furthermore, the stirrer can be monitored by an automatic check of the progress of the stirrer.
[0067]
Furthermore, quality improvement can be realized by quantitative evaluation of the uniformity of the electronic display.
[Brief description of the drawings]
FIG. 1 is a diagram showing an example of the configuration of a painted surface inspection apparatus according to the present invention.
FIG. 2 is a flowchart showing an example of an operation procedure of the painted surface inspection apparatus according to the present invention.
FIG. 3 is a view showing an example in which a polarizing filter is used in the painted surface inspection apparatus of the present invention.
FIG. 4 is an xy chromaticity diagram defined by CIE (International Lighting Commission).
FIG. 5 is a diagram illustrating an example of a camera calibration palette used for color calibration and dimensional teaching according to the present invention.
FIG. 6 is a flowchart illustrating an example of an operation procedure of color calibration according to the present invention.
FIG. 7 is a diagram illustrating a method for calculating a reference chromaticity from an image in the painted surface inspection apparatus according to the present invention.
FIG. 8 is a diagram showing a hue image calculation method in the painted surface inspection apparatus according to the present invention.
FIG. 9 is a diagram illustrating a method for calculating a color difference image in the painted surface inspection apparatus according to the present invention.
FIG. 10 is a diagram showing an example of a method for extracting a defect candidate region from a hue and a color difference in the painted surface inspection apparatus according to the present invention.
FIG. 11 is a diagram illustrating an example in which only a defect area is detected from an extracted defect candidate area based on a color difference change amount in the painted surface inspection apparatus according to the present invention;
FIG. 12 is a diagram showing an example of a binarization threshold value determining method when only a defective area is detected in the painted surface inspection apparatus according to the present invention.
FIG. 13 is a diagram showing an output example of the inspection result of the present invention.
FIG. 14 is a diagram showing an example of storing inspection results according to the present invention.
FIG. 15 is a flowchart showing an example of a processing procedure for inspecting a surface penetrant test image of the present invention.
FIG. 16 is a diagram showing a range of measured chromaticity on the CIE chromaticity diagram in the inspection of the surface penetrant flaw test image of the present invention.
FIG. 17 is a diagram showing an example of a liquid being stirred.
FIG. 18 is a flowchart showing an example of a processing procedure for stirring progress check according to the present invention.
FIG. 19 is a diagram showing an example of a light emitting tube surface of an electronic display.
FIG. 20 is a diagram showing an example of a configuration of a light emission state uniformity evaluation apparatus for an electronic display according to the present invention.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 11 ... Inspection object, 12 ... Illuminating means, 13 ... Camera, 14 ... Computer, 15 ... Image input board, 16 ... Display apparatus, 17 ... Output means, 18 ... Storage device, 31 ... Polarizing filter, 51 ... Camera calibration palette, 52 ... Colorimeter.

Claims (9)

A defect inspection apparatus for inspecting a surface defect of the sample by detecting a color of the surface of the sample, the image capturing unit capturing the sample and capturing it as a color image, and the color image captured by the image capturing unit. Chromaticity / luminance calculation means for calculating chromaticity / luminance from the data using conversion parameters specific to the image capturing means, and determining the reference color from the chromaticity distribution calculated by the chromaticity / luminance calculation means Defect candidate area extraction for obtaining a hue / color difference at each position of the color image with respect to the reference color and extracting a defect candidate area using information on the hue / color difference at each position of the reference color and the obtained color image its size by measuring means and, defects the detected detected defects from the color difference of the amount of change with the surrounding defect candidate regions extracted by said defect candidate area extracting means and said defect candidate area the size and shape A defect detection means for detecting a true defect from the shape, an output means for outputting a detection result detected by the defect detection means, and a storage means for storing the detection result detected by the defect detection means Defect inspection equipment.   The inspection apparatus further includes an illuminating unit, the image capturing unit includes a camera unit that captures a color image by imaging the sample, and the illuminating unit and the camera unit each include a polarizing filter. The defect inspection apparatus according to claim 1.   The chromaticity / luminance calculation means measures the conversion parameters specific to the image capturing means, the RGB data of the color image captured by capturing the camera calibration palette with the image capturing means, and the camera calibration palette with a colorimeter. The defect inspection apparatus according to claim 1, wherein the defect inspection apparatus is obtained by using data of chromaticity x, y and luminance Y obtained in this way. Illuminating means for illuminating the sample with polarized light, imaging means for obtaining an image of the sample by imaging the sample illuminated by the illuminating means through a polarizing filter, and color of the sample imaged by the imaging means Conversion means for converting RGB data of an image into chromaticity x, y and luminance Y using conversion parameters specific to the imaging means, and reference colors and reference colors using chromaticity x, y converted by the conversion means The hue and the color difference at each position on the color image with respect to the chromaticity x, y of the sample are obtained, and the defect candidate area is determined using the obtained reference color and the information on the hue / color difference at each position of the color image of the sample. A process of detecting a defect from the amount of change in color difference between the extracted defect candidate region and the periphery of the defect candidate region, measuring the size and shape of the detected defect, and detecting a true defect from the size and shape Means and processing by the processing means Defect inspection apparatus characterized by comprising a display means for displaying on the screen together with the original image data of the sample results.   The conversion means includes a conversion parameter specific to the imaging means, RGB data of a color image obtained by imaging a camera calibration palette with the imaging means, and chromaticity obtained by measuring the camera calibration palette with a colorimeter. The defect inspection apparatus according to claim 4, wherein the defect inspection apparatus is obtained using data of x, y and luminance Y. A defect inspection method for inspecting a surface defect of a sample by detecting a color of the surface of the sample, wherein the sample is imaged with a color camera to obtain a color image, and the color camera is obtained from the obtained color image data Chromaticity / luminance is calculated using specific conversion parameters, a reference color is determined from the calculated distribution of chromaticity , and hue / color difference at each position of the color image with respect to the determined reference color is obtained. A defect candidate area is extracted using information on hue and hue / color difference at each position of the obtained color image, and a defect is detected from the amount of change in color difference between the extracted defect candidate area and the periphery of the defect candidate area. A defect inspection method characterized by measuring the size and shape of the detected defect, detecting a true defect from the size and shape, and displaying information on the detected true defect on a screen.   The conversion parameters unique to the color camera are RGB data of a color image captured by capturing a camera calibration palette with the color camera, and chromaticity and luminance data obtained by measuring the camera calibration palette with a colorimeter; The defect inspection method according to claim 6, wherein the defect inspection method is calculated using The sample is illuminated with polarized light, the illuminated sample is imaged with a color camera through a polarization filter to obtain a color image of the sample, and RGB data of the color image of the sample is converted into conversion parameters unique to the color camera. chromaticity x was used to convert the y and luminance Y, the converted chromaticity x, the reference color by using the y and the reference color chromaticity x, hue at each position on the color image for y, A color difference is obtained , a defect candidate area is extracted using the obtained reference color and hue / color difference information at each position of the color image of the sample, and the extracted defect candidate area and the periphery of the defect candidate area A process of detecting a defect from the amount of change in color difference, detecting a true defect from the size and shape of the detected defect, and displaying the processed result on the screen together with the original image data of the sample Defect inspection method. The conversion parameters unique to the color camera, RGB data of a color image obtained by imaging a camera calibration palette with the color camera, and chromaticity x, y and luminance obtained by measuring the camera calibration palette with a colorimeter The defect inspection method according to claim 8 , wherein the defect inspection method is obtained using Y data.

Images