JP2004132712A

JP2004132712A - Method and apparatus for inspecting surface flaw in strip-like object

Info

Publication number: JP2004132712A
Application number: JP2002294555A
Authority: JP
Inventors: Toshio Akagi; 赤木　俊夫; Kazukimi Yamada; 山田　和公
Original assignee: Nippon Steel Corp
Current assignee: Nippon Steel Corp
Priority date: 2002-10-08
Filing date: 2002-10-08
Publication date: 2004-04-30
Anticipated expiration: 2022-10-08
Also published as: JP4018962B2

Abstract

<P>PROBLEM TO BE SOLVED: To provide a surface crack inspection method and a surface crack inspecting apparatus of a strip-like body for precisely determining pseudo flaws and harmful ones. <P>SOLUTION: In the surface crack inspection method of the strip-like body for imaging the surface of the traveling strip-like body, performing the image processing of the captured image, and inspecting the surface flaws, a primary decision of flaw type and harmful/harmless is performed to each abnormal section extracted in frame units from the captured image. Information of an abnormal section being generated continuously in the traveling direction of the strip-like body is buffered at a specific buffering maximum processing length or less for grouping. The secondary decision of the harmful/harmless is made, based on information on an abnormal section that is determined to be harmful by the primary decision for each group of the abnormal section and an abnormal section that is determined to be harmless. <P>COPYRIGHT: (C)2004,JPO

Description

【０００１】
【発明の属する技術分野】
この発明は、金属、プラスチックその他材料からなる帯状体を移動しながら表面疵を光学的に検査する帯状体の表面疵検査装置に関する。
【０００２】
【従来の技術】
鋼板などの帯状体について、表面の疵、汚れ、色調などの表面疵を光学的に検査する表面疵検査が広く行なわれている。この表面疵検査では、帯状体をこれの長手方向に送りながら帯状体面をビデオカメラで撮像する。撮像画像に基づいて異常部を検出し、その位置、形状、輝度などの特徴量から疵の種類および有害度の判定処理を行う。
【０００３】
このとき、検出された異常部は必ずしも有害ではなく、無害なものであることも多い。例えば、帯状鋼板の製造ラインでは、製造条件の違いによって油、水滴、軽い汚れなどが鋼板表面に残存する場合が多くある。これら油などは、最終製品の品質に影響がない限り無害扱いされる。このため、表面疵検査では無害な油、水滴、軽い汚れなど（以下、疑似疵という）と最終製品の品質に影響を与えるへゲ疵などの有害疵とを正しく識別することが求められる。しかし、疑似疵であっても製造条件に応じて様々な形態を持つため、撮像画像では形状や輝度などが有害疵のものと非常に似通ったものとなる。したがって、各フレーム単位や異常部単位で疵の判定を行う場合、疑似疵が有害と誤判定されること（以下、過検出という）がしばしば発生し、表面疵検査の信頼性を低下させる要因となっていた。
【０００４】
形状や輝度などが似通った疵どうしの識別精度を上げる技術として、限定された領域での疵情報と、広い領域（以下、マクロ領域という）から得られる疵分布形態情報とを総合して識別する方法（例えば、特許文献１参照）が考えられる。しかし、へゲ疵、スリバー疵などのような鋼板表面の典型的な有害疵に比べ、過検出要因となる疑似疵は、操業条件に応じて形状、輝度、発生頻度などが多様に変化する。このために、従来のように判定基準を一律に定めた判定方法では、精度の高い判定を行うことは難しかった。
【０００５】
また、疑似疵と有害疵が同じマクロ領域に発生することも多いが、それぞれの個数や個数比は様々であるため、例えば上記特許文献１の実施例のように、マクロ領域内での個数・個数比情報から判定をやり直す方法では判定精度を上げることが難しかった。
【０００６】
【特許文献１】
特開平４−１１０７５８号公報（第１ページ、下左欄、特許請求の範囲請求、および第４ページ、下右欄〜第５ページ、下右欄）
【０００７】
【発明が解決しようとする課題】
この発明の課題は、疑似疵と有害疵とを高精度で判定することができる帯状体の表面疵検査方法およびその装置を提供することである。
【０００８】
【課題を解決するための手段】
この発明の帯状体の表面疵検査方法は、移動する帯状体の表面を撮像し、撮像画像を画像処理して表面疵を検査する帯状体の表面疵検査方法において、撮像画像からフレーム単位で抽出した各異常部について疵種および有害／無害の１次判定を行ない、帯状体の移動方向に連続して発生する異常部の情報を所定のバッファリング最大処理長さ以下でバッファリングしてグループ化し、前記異常部のグループごとに前記１次判定で有害と判定された異常部および無害と判定された異常部の情報に基づいて有害／無害を２次判定する。
【０００９】
上記帯状体の表面疵検査方法において、２次判定が、１次判定で有害と判定された異常部の個数、および無害と判定された異常部の個数に基づいて有害／無害を判定し、ついで前記１次判定で有害と判定された異常部および無害と判定された異常部の特徴量により有害／無害を判定するようにしてもよい。この帯状体の表面疵検査方法において、１次判定で無害と判定された異常部の特徴量分布の代表値と１次判定で有害と判定された各異常部との間のノルム値により、有害と判定された異常部について有害／無害を判定するようにしてもよい。
【００１０】
この発明の他の帯状体の表面疵検査方法は、移動する帯状体の表面を撮像し、撮像した画像信号に基づいて表面疵を検査する帯状体の表面疵検査方法において、
（ａ）　撮像画像からフレーム単位で異常部を抽出すること
（ｂ）　抽出した各異常部について異常部の特徴量を求めること
（ｃ）　各異常部について、特徴量に基づき異常部の疵種および有害／無害の１次判定すること
（ｄ）　帯状体の移動方向に連続して発生する異常部の情報を、所定の帯状体最大処理長さ以下でバッファリングしてグループ化すること
（ｅ）　前記異常部のグループごとに、有害と１次判定された異常部の個数の、異常部すべての個数に対する個数比Ｐを求めること
（ｆ）　前記異常部のグループごとに、個数比Ｐが閾値Ｐｏ超えるかどうかを判断すること
（ｇ１）　個数比Ｐが閾値Ｐｏを超えるグループの場合、有害／無害の１次判定結果をそのまま維持すると２次判定すること
（ｇ２）　個数比Ｐが閾値Ｐｏ以下であるグループの場合、無害と１次判定された異常部の特徴量分布の代表値と有害と１次判定された各異常部の特徴量との間のノルム値Ｎを求めること
（ｈ）　ノルム値Ｎが閾値Ｎｏを超えるかどうかを判断すること
（ｉ１）　ノルム値Ｎが閾値Ｎｏを超える場合、１次判定で有害とされた異常部をそのまま有害と２次判定すること
（ｉ２）　ノルム値Ｎが閾値Ｎｏ以下である場合、１次判定で有害とされた異常部を無害と２次判定すること
からなっている。
【００１１】
この発明の帯状体の表面疵検査装置は、移動する帯状体の表面を撮像する撮像装置と、撮像画像を処理して疵種および有害／無害を判定する画像処理装置とを備えた帯状体の表面疵検査装置において、撮像画像からフレーム単位で抽出した各異常部について疵種および有害／無害を判定する１次判定手段と、帯状体の移動方向に連続して発生する異常部の情報を所定のバッファリング最大処理長さ以下でバッファリングしてグループ化するバッファリング手段と、前記異常部のグループごとに前記１次判定で有害と判定された異常部および無害と判定された異常部の疵情報に基づいて有害／無害を判定する２次判定手段とからなっている。
【００１２】
上記帯状体の表面疵検査装置において、２次判定手段を、１次判定で有害と判定された異常部の個数および無害と判定された異常部の個数に基づいて有害／無害を判定する第１判定手段と、前記１次判定で有害と判定された異常部および無害と判定された異常部の特徴量により有害／無害を判定する第２判定手段とで構成してもよい。この表面疵検査装置で、第２判定手段が、１次判定で無害と判定された異常部の特徴量分布の代表値と１次判定で無害と判定された各異常部との間のノルム値により、有害と判定された異常部について有害／無害を判定するようにしてもよい。
【００１３】
この発明の他の表面疵検査装置は、移動する帯状体の表面を撮像する撮像装置と、撮像画像を処理して疵種および有害／無害を判定する画像処理装置とを備えた帯状体の表面疵検査装置において、
撮像画像からフレーム単位で異常部を抽出する異常部抽出手段と、
抽出した各異常部について異常部の特徴量を求める特徴量計測手段と、
各異常部について、特徴量に基づき異常部の疵種および有害／無害を判定する１次判定手段と、
帯状体の移動方向に連続して発生する異常部の情報を所定のバッファリング最大処理長さ以下でバッファリングしてグループ化するバッファリング手段と、
前記異常部のグループごとに、有害と１次判定された異常部の個数の、異常部すべての個数に対する個数比Ｐを求める個数比計測手段と
前記異常部のグループごとに、個数比Ｐが閾値Ｐｏ超えるかどうかを判断する個数比判断手段と、
個数比Ｐが閾値Ｐｏを超えるグループの場合、有害／無害の１次判定結果をそのまま維持すると判定する２次判定手段と、
個数比Ｐが閾値Ｐｏ以下であるグループの場合、無害と１次判定された異常部の特徴量分布の代表値と有害と１次判定された各異常部の特徴量との間のノルム値Ｎを求めることノルム値計測手段と、
ノルム値Ｎが閾値Ｎｏを超えるかどうかを判断するノルム値判断手段と、
ノルム値Ｎが閾値Ｎｏを超える場合、１次判定で有害とされた異常部を有害と２次判定する第１判定手段と、
ノルム値Ｎが閾値Ｎｏ以下である場合、１次判定で有害とされた異常部を無害と２次判定する第２判定手段と
からなっている。
【００１４】
【発明の実施の形態】
帯状体、例えば鋼板の製造プロセスでは、疑似疵は様々な形状、輝度などを持つが、互いに近い領域内の疑似疵は、油、水滴、軽い汚れなどの発生要因が共通で操業条件も近いため、撮像画像でも似たような形状、輝度を持つことが多い。また、疑似疵は、ロールについた油や水が鋼板上に転写されて生じる場合などを典型的な例として、一般に帯状体の移動方向に連続して発生することが多い。疑似疵が連続的に発生している場合、その特徴量の類似度は高いものとなる。
【００１５】
そこで、この発明では、上述のように帯状体の移動方向へ連続性を持つ異常部の情報をバッファリングして、グループとして扱う。そして、個々の異常部単位で１次判定したのちに、グループ内に存在する各異常部の情報から、再度、判定をやり直す２次判定を行なう。２次判定には、１次判定で有害な疵、あるいは無害と判定された異常部の個数と共に、有害な疵と判定された異常部の位置、形状、輝度などの情報と、無害と１次判定された異常部の位置、形状、輝度などの情報とが利用される。２次判定を行なうことにより、従来の画像フレーム単位、異常部単位では有害疵と判定されるような特徴量を有する疑似疵でも、グループ内の無害と１次判定された他の異常部と類似度が高ければ、同種の無害であるとみなすことができ、過検出の頻度を抑制できる。逆に、類似度が低ければ、１次判定の結果通り有害疵とみなすことによって、真の有害疵を無害と誤って判定し直し、結果として有害疵を見逃してしまうことを防ぐことができる。
【００１６】
図１〜図３を参照して、この発明の１実施の形態について説明する。図１は表面疵検査装置の概略図であり、図２は上記装置の主要部のブロック図である。図３は画像処理のフローチャートである。以下、帯状体が帯状鋼板である場合について説明する。
【００１７】
通板方向に移動する鋼板１の表面をビデオカメラ１０で撮像する。ビデオカメラ１０からの画像信号は、画像処理装置２０のコンピュータ２２にデータバス３４およびデータ入出力制御装置２６を介して主記憶装置２８に転送される。主記憶装置２８は画像処理プログラムを実行し、送られてきた撮像画像について、１次判定、バッファリング、２次判定などの処理を行なう。画像処理プログラムは、ハードデスクなどの補助記憶装置３０に保存されている。画像処理に必要な閾値などの設定項目は、キーボード４０から入力され、補助記憶装置３０に保存されている。画像処理された撮像画像のデータは補助記憶装置３０に保存され、またグラフィックスボード３２を通して表示装置４２に出力される。表示装置４２は、鋼板面の画像とともに、疵種、有害／無害などの判定結果を表示する。
【００１８】
上記画像処理について、図３のフローチャートに従って説明する。
ステップＳ１１
撮像画像は濃度補正、平滑化などの前処理を行なったのち、撮像画像からフレーム単位で異常部を抽出する。撮像画像についてエッジ抽出、画像強調などの処理を行なって異常部を抽出する。
【００１９】
ステップＳ１２
抽出した各異常部について、異常部の位置、形状、輝度などを撮像画像上で計測して特徴量を求める。形状の特徴量は、異常部に外接する長方形の幅、長さ、長さと幅の比、異常部の面積、異常部の周囲長などがあり、疵種、異常部の形状などによって適宜選択される。なお、異常部の鋼板長手方向位置は、図１に示す搬送ローラ１５に設けられたパルスジェネレータ１６からのパルス信号から求め、幅方向位置は撮像画像上で求める。
【００２０】
ステップＳ１３
各異常部について、特徴量を判定基準と比較して異常部の疵種および有害／無害を１次判定する。判定基準は、疵種および有害／無害について実操業で得られたデータを基にして作成され、コンピュータ２２の補助記憶装置３０に表形式で格納されている。
【００２１】
ステップＳ１４
１次判定した各異常部について、通板方向に連続する異常部の情報をバッファリングしてグループ化する。バッファリングする異常部の情報として、前記異常部の特徴量（位置、形状、輝度など）の外に１次判定結果、画像などがある。異常部の特徴量および１次判定結果は必須であるが、その他は検査条件によって適宜選択される。
【００２２】
バッファリングは、板幅方向で異なる位置の異常部はそれぞれ別々にバッファリングして別のグループとして処理する。バッファリングの長さは、あらかじめ設定したバッファリング最大処理長さＢ以下とする。バッファリング最大処理長さＢは、撮像点Ｓから表示装置４２までの距離Ａと、バッファリング終了から画像処理により２次判定結果が得られるまでの時間に帯状体が移動する距離との差である。バッファリング最大処理長さＢを超えると、バッファリングしたグループが表示装置４２に到達したときに２次判定結果を検査員Ｍに示すことができない。バッファリングした連続する異常部の長さがバッファリング最大処理長さＢより小さい場合、その連続する異常部の長さでバッファリングする。撮像部の撮像点Ｓからの移動距離は、上記パルスジェネレータ１６からのパルス信号により計測する。パルス信号はデータバス３４およびデータ入出力制御装置２６を介して主記憶装置２８に転送され、ＣＰＵ２４で移動距離が演算される。
【００２３】
異常部が、通板方向に平行な直線からずれることがある。このような場合、板幅方向に基準点からのずれの許容範囲を設け、許容範囲内の異常部はそのグループに含めるようにするとよい。図４は、バッファリングされる各グループの先頭の異常部を基準点として、許容範囲αを設定し、後続の異常部をグルーピングしていく方法を示している。図５は、鋼板エッジ部に生じた異常部のグルーピングの例を模式的に示している。図中、小さな長方形は、鋼板上で検出された異常部の位置を示している。グループ１は、連続する異常部がバッファリング最大処理長さＢに達する前に途切れたために、バッファリングを中断して次の２次判定に移った場合を示す。一方、グループ２、グループ３は異常部の長さがバッファリング最大処理長さＢより長いため、バッファリング最大処理長さＢに達した時点で、バッファリングを中断して、次の２次判定に移った場合を示す。
【００２４】
ステップＳ１５
バッファリングしたグループごとに１次判定で有害と判定された異常部および無害と判定された異常部の特徴量に基づいて有害／無害を２次判定する。例えば、１次判定で有害と判定された異常部の個数、および無害と判定された異常部の個数に基づいて有害／無害を２次判定し、ついで１次判定で有害と判定された異常部および無害と判定された異常部の特徴量により有害／無害を２次判定するようにしてもよい。
【００２５】
つぎに、画像処理の他の形態について、図６に示すフローチャートに従って説明する。
ステップＳ２１〜２４
図３のステップＳ１１〜１４と同じであるので、その説明は省略する。
【００２６】
ステップＳ２５
バッファリングしたグループごとに、有害と１次判定された異常部の個数ｎの、有害・無害を含めた異常部すべての個数ｎｏに対する個数比Ｐ（＝ｎ／ｎｏ）を求める。
【００２７】
ステップＳ２６
グループごとに、個数比Ｐが閾値Ｐｏ超えるかどうかを判断する。
【００２８】
ステップＳ２７Ｙ
個数比Ｐが閾値Ｐｏを超えるグループの場合、有害／無害の１次判定結果をそのまま維持すると２次判定する。
【００２９】
ステップＳ２７Ｎ
個数比Ｐが閾値Ｐｏ以下であるグループの場合、無害と１次判定された異常部の特徴量分布の代表値と有害と１次判定された各異常部の特徴量との間のノルム値Ｎを求める。ノルム値は、特徴量空間で上記代表値と有害と１次判定された異常部の特徴量との間の距離を表わす。特徴量分布の代表値として平均値、中央値などが用いられる。ノルム値として、ユークリッド距離、標準ユークリッド距離またはマハラノビス距離が用いられる。これら代表値およびノルム値については、操業実績により判定精度の高いものを採用する。
【００３０】
ステップＳ２８
ノルム値Ｎが閾値Ｎｏを超えるか、どうかを判断する。
【００３１】
ステップＳ２９Ｎ
ノルム値Ｎが閾値Ｎｏを超える場合、１次判定で有害とされた異常部を無害に修正すると２次判定する。
【００３２】
ステップＳ２９Ｙ
ノルム値Ｎが閾値Ｎｏ以下である場合、１次判定で有害とされた異常部を無害と２次判定する。
【００３３】
この発明は、上記実施の形態に限られるものではない。帯状体は鋼板に限らずアルミニウム板、プラスチック板、紙などであってもよい。鉄鋼のスラブ、厚板なども、異常部近傍の領域（長手方向、幅方向に限らず、ある幅を持つ領域）の情報を用いることで、この発明を利用することができる。２次判定に用いる特徴量は輝度に限らず位置、形状などであってもよく、また特徴量の数は３以上であってもよい。
【００３４】
【実施例】
帯状鋼板について、製造ライン上で本発明法および従来法によりそれぞれ表面疵検査を行なった。鋼板の幅は１２００ｍｍであり、試験長さは６００ｍであった。鋼板の平均移動速度は、２００ｍ／ｍｉｎであった。表面疵検出装置の設置箇所で鋼板上に水滴が残っており、撮像画像では水滴の形状・輝度が有害のヘゲ疵と似ていた。
【００３５】
ヘゲ疵および水滴と１次判定された２種類の異常部について、個数比Ｐによる２次判定を行なった。更に、個数比Ｐによる２次判定で無害と判定された異常部について、輝度特徴量Ａ、輝度特徴量Ｂの２つの特徴量により２次判定を行なった。
【００３６】
図７は、ヘゲ疵および水滴と１次判定された２種類の異常部を含むグループの２つの特徴量Ａ、Ｂを示している。ここで、個数比Ｐの閾値は０．３であり、ノルム値Ｎの閾値も０．３である。図７（ａ）ではヘゲ疵の個数比Ｐが３／５であり、閾値を超えるため、１次判定で有害と判定されたヘゲ疵はそのまま有害と２次判定した。図７（ｂ）では、ヘゲ疵の個数比Ｐは３／２３と閾値以下である。また、ヘゲ疵のノルム値Ｎは、それぞれ０．１０、０．１４、０．２０であり、閾値０．３以下であるので、水滴との類似度が高い。したがって、１次判定で有害とされたが、ヘゲ疵は水切り不良による水滴と判断され、無害に修正された。図７（ｃ）では、ヘゲ疵の個数比Ｐは１／２６と閾値以下である。しかし、ヘゲ疵のノルム値Ｎは０．４０であり、閾値０．３を超えるので、水滴との類似度が低い。したがって、１次判定で有害とされたヘゲ疵はそのまま有害と２次判定された。
【００３７】
この発明による帯状鋼板１ｋｍ当り過検出の発生回数は０．１回であったが、従来法では１．５回であった。過検出頻度が１／１５と大幅に低減し、表面疵検査の信頼性が向上した。
【００３８】
【発明の効果】
この発明では、有害疵に似た無害の汚れ、油、水滴などが連続して発生しても、これらを有害疵とする誤判定が大幅に減少し、過検出の発生を抑制することができる。この結果、鋼板などの帯状体の生産性向上およびコスト低減を図ることができる。
【図面の簡単な説明】
【図１】この発明の表面疵検査を実施する表面疵検査装置の概略図でる。
【図２】表面疵検査装置の主要部を構成する画像処理装置のブロック図である。
【図３】この発明における画像処理の１形態を示すフローチャートである。
【図４】異常部のグルーピングにおいて、異常部が板幅方向にずれる場合の許容範囲の設定を説明する模式図である。
【図５】鋼板エッジ部に生じた異常部のグルーピング例を示す模式図である。
【図６】この発明における画像処理の他の形態を示すフローチャートである。
【図７】個数比およびノルム値により２次判定を行なう例を示す図であり、ヘゲ疵および水滴の特徴量分布図である。
【符号の説明】
１　鋼板（帯状体）　　　　　　　　１０　ビデオカメラ
１６　パルスジェネレータ　　　　　　　２０　画像処理装置
２２　コンピュータ
Ｓ　撮像点　　　　　　　　　　　　　Ａ　撮像点から検査員までの距離
Ｂ　バッファリング最大処理長さ[0001]
TECHNICAL FIELD OF THE INVENTION
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for inspecting surface defects of a band, which optically inspects surface defects while moving the band formed of metal, plastic, or other material.
[0002]
[Prior art]
2. Description of the Related Art A surface flaw inspection for optically inspecting a surface flaw such as a surface flaw, dirt, and color tone of a belt-like body such as a steel sheet is widely performed. In the surface flaw inspection, the belt-like body is imaged with a video camera while the belt-like body is fed in the longitudinal direction. An abnormal portion is detected based on the captured image, and a process of determining the type and harmfulness of a flaw is performed based on feature amounts such as a position, a shape, and luminance.
[0003]
At this time, the detected abnormal part is not necessarily harmful, and is often harmless. For example, in a production line of strip-shaped steel sheets, oil, water droplets, light stains, and the like often remain on the steel sheet surface due to differences in manufacturing conditions. These oils are treated harmless unless they affect the quality of the final product. For this reason, in the surface flaw inspection, it is required to correctly distinguish harmless oil, water drops, light dirt, etc. (hereinafter referred to as pseudo flaws) from harmful flaws, such as hedging flaws, which affect the quality of the final product. However, even a pseudo flaw has various forms according to manufacturing conditions, and thus the captured image is very similar in shape and brightness to that of a harmful flaw. Therefore, when flaws are determined on a frame basis or on an abnormal part basis, false flaws are often erroneously determined to be harmful (hereinafter referred to as over-detection), which is a factor that reduces the reliability of surface flaw inspection. Had become.
[0004]
As a technique for improving the identification accuracy of flaws having similar shapes and luminances, flaw information in a limited area and flaw distribution form information obtained from a wide area (hereinafter referred to as a macro area) are comprehensively identified. A method (for example, see Patent Document 1) is conceivable. However, compared to typical harmful flaws on the steel sheet surface, such as hedging flaws, sliver flaws, etc., pseudo flaws that cause overdetection have various changes in shape, brightness, occurrence frequency, etc. according to operating conditions. For this reason, it has been difficult to make a highly accurate determination by the conventional determination method in which the determination criteria are uniformly set.
[0005]
In addition, pseudo flaws and harmful flaws often occur in the same macro area, but since the numbers and number ratios of the flaws are various, for example, as in the example of Patent Document 1, the number of It was difficult to increase the determination accuracy by the method of performing the determination again from the number ratio information.
[0006]
[Patent Document 1]
JP-A-4-110758 (first page, lower left column, claims, and fourth page, lower right column to page 5, lower right column)
[0007]
[Problems to be solved by the invention]
An object of the present invention is to provide a method and an apparatus for inspecting a surface flaw of a belt-like body capable of determining a pseudo flaw and a harmful flaw with high accuracy.
[0008]
[Means for Solving the Problems]
In the method for inspecting a surface flaw of a belt-like body according to the present invention, in the method for inspecting a surface flaw of a belt-like body in which a surface of a moving belt-like body is imaged, and a picked-up image is image-processed to inspect the surface flaws, the method is extracted in frame units from the picked-up image. The primary determination of flaw type and harmful / harmless is performed for each abnormal part, and information on abnormal parts continuously occurring in the moving direction of the band is buffered and grouped below a predetermined buffering maximum processing length. A harmful / harmless judgment is made for each of the abnormal part groups based on the information of the abnormal part determined to be harmful in the primary determination and the abnormal part determined to be harmless.
[0009]
In the above-mentioned method for inspecting the surface flaw of a belt-like body, the secondary determination determines harmful / harmless based on the number of abnormal portions determined to be harmful in the primary determination and the number of abnormal portions determined to be harmless in the primary determination. Harmful / harmless may be determined based on the characteristic amount of the abnormal portion determined to be harmful in the primary determination and the abnormal portion determined to be harmless. In this method for inspecting the surface flaw of a belt-like body, the harmfulness is determined by the norm value between the representative value of the characteristic amount distribution of the abnormal portion determined to be harmless in the primary determination and each abnormal portion determined to be harmful in the primary determination. Harmful / harmless may be determined for the abnormal part determined as above.
[0010]
Another method for inspecting a surface flaw of a belt-like body of the present invention is a method of inspecting a surface flaw of a belt-like body for imaging a surface of a moving belt-like body and inspecting a surface flaw based on an image signal obtained by imaging.
(A) extracting an abnormal part from the captured image in frame units; (b) obtaining a characteristic amount of the abnormal part with respect to each extracted abnormal part; and (c) for each abnormal part, a defect type of the abnormal part based on the characteristic amount and Primary determination of harmful / harmless (d) Information of abnormal parts continuously occurring in the moving direction of the band is buffered and grouped below a predetermined band maximum processing length (e). For each group of abnormal parts, determine the number ratio P of the number of abnormal parts that are primarily determined to be harmful to the number of all abnormal parts. (F) For each group of abnormal parts, the number ratio P is a threshold Po. (G1) In the case of a group in which the number ratio P exceeds the threshold Po, a secondary determination is made to maintain the harmful / harmless primary determination result as it is (g2) When the number ratio P is equal to or less than the threshold Po A guru (H) norm value N between the representative value of the feature value distribution of the abnormal portion that is primarily determined to be harmless and the feature value of each abnormal portion that is primarily determined to be harmful. (I1) When the norm value N exceeds the threshold value No, the abnormal part determined to be harmful in the primary determination is secondarily determined to be harmful as it is in the primary determination (i2). When the difference is equal to or smaller than the threshold No, the abnormal part determined to be harmful in the primary determination is secondarily determined to be harmless.
[0011]
A surface defect inspection apparatus for a belt-shaped body according to the present invention includes: an imaging device that images a surface of a moving band-shaped body; and an image processing device that processes a captured image to determine a flaw type and harmful / harmless. In the surface flaw inspection device, primary determination means for determining the type of flaw and harmful / harmless for each abnormal part extracted in frame units from the captured image, and information on abnormal parts continuously occurring in the moving direction of the band-shaped object are determined. Buffering means for buffering and grouping by the buffering maximum processing length less than or equal to the maximum length of processing, and a defect of an abnormal part determined to be harmful in the primary determination and an abnormal part determined to be harmless for each group of the abnormal parts. It is composed of secondary judgment means for judging harmful / harmless based on the information.
[0012]
In the above-mentioned apparatus for inspecting a surface flaw of a belt-like body, the secondary judging means judges the harmfulness / harmlessness based on the number of abnormal parts judged to be harmful in the primary judgment and the number of abnormal parts judged to be harmless. The determination unit may include a second determination unit that determines harmfulness or harmlessness based on a characteristic amount of the abnormal part determined to be harmful in the primary determination and the abnormal part determined to be harmless. In this surface flaw inspection apparatus, the second determining means determines a norm value between a representative value of the characteristic amount distribution of the abnormal portion determined to be harmless in the primary determination and each abnormal portion determined to be harmless in the primary determination. Thus, the harmful / harmless determination may be made for the abnormal part determined to be harmful.
[0013]
According to another aspect of the present invention, there is provided a surface flaw inspection apparatus including: an imaging apparatus configured to capture an image of a surface of a moving band; and an image processing apparatus configured to process a captured image to determine a flaw type and harmfulness / harmlessness. In the flaw inspection device,
Abnormal portion extracting means for extracting an abnormal portion from the captured image in frame units;
A feature amount measuring means for obtaining a feature amount of the abnormal portion for each extracted abnormal portion;
Primary determining means for determining, for each abnormal part, a flaw type and harmful / harmless of the abnormal part based on the characteristic amount;
Buffering means for buffering and grouping the information of the abnormal part which continuously occurs in the moving direction of the belt-shaped body at a predetermined buffering maximum processing length or less,
For each group of abnormal parts, a number ratio measuring means for calculating a number ratio P of the number of abnormal parts determined as primary to be harmful to the number of all abnormal parts, and for each group of abnormal parts, a number ratio P is a threshold. A number ratio determining means for determining whether Po is exceeded;
A secondary determination unit that determines that the harmful / harmless primary determination result is to be maintained as it is in a group where the number ratio P exceeds the threshold Po;
In the case of the group in which the number ratio P is equal to or less than the threshold value Po, the norm value N between the representative value of the characteristic amount distribution of the abnormal part which is primarily determined to be harmless and the characteristic amount of each abnormal part which is primarily determined to be harmful A norm value measuring means;
Norm value determining means for determining whether or not the norm value N exceeds a threshold No;
When the norm value N exceeds the threshold No, first determining means for secondary determining that the abnormal part determined to be harmful in the primary determination is harmful,
When the norm value N is equal to or smaller than the threshold No, the second determination means is configured to secondarily determine that the abnormal portion determined to be harmful in the primary determination is harmless.
[0014]
BEST MODE FOR CARRYING OUT THE INVENTION
In the manufacturing process of strips, for example, steel plates, pseudo flaws have various shapes, brightness, etc., but pseudo flaws in areas close to each other are common due to common factors such as oil, water droplets, light dirt, and similar operating conditions. Also, the captured image often has a similar shape and brightness. In addition, the pseudo flaw is generally generated continuously in the moving direction of the strip, typically as a typical example in which oil or water on a roll is transferred onto a steel plate. When the pseudo flaw is continuously generated, the similarity of the feature amount is high.
[0015]
Therefore, in the present invention, as described above, information on abnormal portions having continuity in the moving direction of the band is buffered and handled as a group. Then, after the primary determination is performed for each abnormal part, a secondary determination is performed again from the information of each abnormal part existing in the group. The secondary determination includes the number of abnormal parts determined to be harmful or harmless in the primary determination, information on the position, shape, brightness, etc. of the abnormal part determined to be harmful, Information such as the position, shape, and brightness of the determined abnormal part is used. By performing the secondary determination, even a pseudo flaw having a characteristic amount that is determined to be a harmful flaw in the conventional image frame unit or abnormal part unit is similar to other abnormal parts that are primarily determined to be harmless in the group. If the degree is high, it can be regarded as harmless of the same kind, and the frequency of overdetection can be suppressed. Conversely, if the degree of similarity is low, the harmful flaw can be erroneously determined to be harmless again by regarding the harmful flaw as harmless as a result of the primary determination, thereby preventing the harmful flaw from being overlooked as a result.
[0016]
An embodiment of the present invention will be described with reference to FIGS. FIG. 1 is a schematic diagram of a surface flaw inspection device, and FIG. 2 is a block diagram of a main part of the device. FIG. 3 is a flowchart of the image processing. Hereinafter, the case where the strip is a strip-shaped steel plate will be described.
[0017]
The video camera 10 captures an image of the surface of the steel plate 1 moving in the passing direction. An image signal from the video camera 10 is transferred to the computer 22 of the image processing device 20 to the main storage device 28 via the data bus 34 and the data input / output control device 26. The main storage device 28 executes an image processing program and performs processing such as primary determination, buffering, and secondary determination on the transmitted captured image. The image processing program is stored in an auxiliary storage device 30 such as a hard desk. Setting items such as threshold values required for image processing are input from the keyboard 40 and stored in the auxiliary storage device 30. The data of the captured image subjected to the image processing is stored in the auxiliary storage device 30 and is output to the display device 42 through the graphics board 32. The display device 42 displays the image of the steel plate surface and the determination result of the flaw type, harmful / harmless, and the like.
[0018]
The image processing will be described with reference to the flowchart of FIG.
Step S11
After performing pre-processing such as density correction and smoothing on the captured image, abnormal portions are extracted from the captured image in frame units. An abnormal part is extracted by performing processes such as edge extraction and image enhancement on the captured image.
[0019]
Step S12
For each of the extracted abnormal portions, the position, shape, luminance, and the like of the abnormal portion are measured on the captured image to obtain a feature amount. The feature amount of the shape includes a width, a length, a ratio of the length to the width of the rectangle circumscribing the abnormal part, an area of the abnormal part, a peripheral length of the abnormal part, and the like, and is appropriately selected depending on a flaw type, a shape of the abnormal part, and the like. You. The position of the abnormal portion in the longitudinal direction of the steel sheet is obtained from a pulse signal from a pulse generator 16 provided on the transport roller 15 shown in FIG. 1, and the position in the width direction is obtained on a captured image.
[0020]
Step S13
For each abnormal part, the feature amount is compared with a determination criterion, and the type of flaw and harmful / harmless of the abnormal part are primarily determined. The criterion is created based on data obtained in actual operation for the flaw type and harmful / harmless, and is stored in the auxiliary storage device 30 of the computer 22 in a table format.
[0021]
Step S14
For each of the first-order determined abnormal portions, information on the abnormal portions that are continuous in the sheet passing direction is buffered and grouped. The information of the abnormal part to be buffered includes a primary determination result, an image, and the like, in addition to the characteristic amount (position, shape, luminance, and the like) of the abnormal part. The characteristic amount of the abnormal part and the primary determination result are essential, but others are appropriately selected according to the inspection conditions.
[0022]
In the buffering, abnormal portions at different positions in the plate width direction are separately buffered and processed as separate groups. The buffering length is set to be equal to or less than a preset buffering maximum processing length B. The maximum buffering processing length B is the difference between the distance A from the imaging point S to the display device 42 and the distance that the band moves during the time from the end of buffering to the time when the secondary determination result is obtained by the image processing. is there. If the buffering maximum processing length B is exceeded, the secondary determination result cannot be shown to the inspector M when the buffered group reaches the display device 42. If the length of the buffered continuous abnormal part is smaller than the buffering maximum processing length B, buffering is performed with the length of the continuous abnormal part. The moving distance of the imaging unit from the imaging point S is measured by a pulse signal from the pulse generator 16. The pulse signal is transferred to the main storage device 28 via the data bus 34 and the data input / output control device 26, and the movement distance is calculated by the CPU 24.
[0023]
The abnormal portion may deviate from a straight line parallel to the passing direction. In such a case, an allowable range of deviation from the reference point may be provided in the plate width direction, and abnormal portions within the allowable range may be included in the group. FIG. 4 shows a method of setting an allowable range α using the abnormal part at the head of each buffered group as a reference point and grouping the subsequent abnormal parts. FIG. 5 schematically shows an example of grouping of abnormal portions generated at the steel plate edge portion. In the drawing, a small rectangle indicates the position of the abnormal portion detected on the steel plate. Group 1 shows a case where the buffering is interrupted and the process proceeds to the next secondary determination because the continuous abnormal portion is interrupted before reaching the buffering maximum processing length B. On the other hand, since the length of the abnormal portion is longer than the maximum buffering processing length B in groups 2 and 3, buffering is interrupted when the buffering maximum processing length B is reached, and the next secondary determination is made. Shows the case where it has moved to.
[0024]
Step S15
For each buffered group, harmful / harmless is secondarily determined based on the characteristic amount of the abnormal part determined to be harmful in the primary determination and the abnormal part determined to be harmless. For example, a secondary determination of harmful / harmless is made based on the number of abnormal parts determined to be harmful in the primary determination and the number of abnormal parts determined to be harmless, and then the abnormal part determined to be harmful in the primary determination Harmful / harmless may be secondarily determined based on the feature amount of the abnormal portion determined to be harmless.
[0025]
Next, another embodiment of the image processing will be described with reference to the flowchart shown in FIG.
Steps S21 to S24
Steps S11 to S14 in FIG. 3 are the same, and a description thereof will be omitted.
[0026]
Step S25
For each buffered group, the number ratio P (= n / no) of the number n of the abnormal parts that are primarily determined to be harmful to the number no of all abnormal parts including harmful and harmless is calculated.
[0027]
Step S26
It is determined whether the number ratio P exceeds the threshold Po for each group.
[0028]
Step S27Y
In the case of a group in which the number ratio P exceeds the threshold Po, a secondary determination is made to maintain the harmful / harmless primary determination result as it is.
[0029]
Step S27N
In the case of the group in which the number ratio P is equal to or less than the threshold value Po, the norm value N between the representative value of the characteristic amount distribution of the abnormal part which is primarily determined to be harmless and the characteristic amount of each abnormal part which is primarily determined to be harmful Ask for. The norm value indicates a distance between the representative value in the feature amount space and the feature amount of the abnormal part that is primarily determined to be harmful. An average value, a median value, or the like is used as a representative value of the feature amount distribution. As the norm value, a Euclidean distance, a standard Euclidean distance or a Mahalanobis distance is used. As these representative values and norm values, those having high determination accuracy depending on the operation results are adopted.
[0030]
Step S28
It is determined whether or not the norm value N exceeds the threshold No.
[0031]
Step S29N
When the norm value N exceeds the threshold No, a secondary determination is made to harmlessly correct the abnormal part determined to be harmful in the primary determination.
[0032]
Step S29Y
When the norm value N is equal to or smaller than the threshold No, the abnormal part determined to be harmful in the primary determination is secondarily determined to be harmless.
[0033]
The present invention is not limited to the above embodiment. The strip is not limited to a steel plate, but may be an aluminum plate, a plastic plate, paper, or the like. The present invention can also be used for iron and steel slabs, thick plates, and the like by using information on regions near abnormal portions (regions having a certain width, not limited to the longitudinal direction and the width direction). The feature amount used for the secondary determination is not limited to the luminance, but may be a position, a shape, or the like. The number of feature amounts may be three or more.
[0034]
【Example】
With respect to the strip-shaped steel sheet, surface flaw inspection was performed on the production line by the method of the present invention and the conventional method. The width of the steel plate was 1200 mm and the test length was 600 m. The average moving speed of the steel sheet was 200 m / min. Water droplets remained on the steel plate at the place where the surface flaw detection device was installed, and the shape and brightness of the water droplets were similar to harmful barbed flaws in the captured image.
[0035]
A secondary determination was performed based on the number ratio P for the two types of abnormal portions that were primary determined as barbed flaws and water droplets. Further, for the abnormal part determined to be harmless by the secondary determination based on the number ratio P, a secondary determination was performed using two characteristic amounts, that is, the luminance characteristic amount A and the luminance characteristic amount B.
[0036]
FIG. 7 shows two feature values A and B of a group including two types of abnormal portions that are primarily determined as a barbed flaw and a water droplet. Here, the threshold value of the number ratio P is 0.3, and the threshold value of the norm value N is also 0.3. In FIG. 7A, since the number ratio P of the barbed flaws is 3/5, which exceeds the threshold value, the barbed flaws determined to be harmful in the primary determination are secondarily determined to be harmful as they are. In FIG. 7B, the number ratio P of the barbed flaws is 3/23, which is below the threshold value. Further, the norm values N of the barb flaws are 0.10, 0.14, and 0.20, respectively, and are equal to or smaller than the threshold value 0.3, so that the similarity with the water droplet is high. Therefore, although it was determined to be harmful in the primary determination, the barbed flaw was determined to be a water droplet due to poor drainage, and was corrected harmlessly. In FIG. 7C, the number ratio P of the barbed flaw is 1/26, which is equal to or less than the threshold value. However, since the norm value N of the barbed flaw is 0.40, which exceeds the threshold value 0.3, the similarity with the water droplet is low. Therefore, the barbed flaws determined to be harmful in the primary determination were directly determined to be harmful in the secondary determination.
[0037]
The number of occurrences of overdetection per km of the strip-shaped steel sheet according to the present invention was 0.1 times, but was 1.5 times in the conventional method. The overdetection frequency was greatly reduced to 1/15, and the reliability of surface flaw inspection was improved.
[0038]
【The invention's effect】
According to the present invention, even if harmless stains, oils, water droplets, etc. similar to harmful flaws are continuously generated, erroneous determination of these as harmful flaws is greatly reduced, and occurrence of overdetection can be suppressed. . As a result, it is possible to improve the productivity and reduce the cost of the strip such as a steel plate.
[Brief description of the drawings]
FIG. 1 is a schematic view of a surface flaw inspection apparatus for performing a surface flaw inspection according to the present invention.
FIG. 2 is a block diagram of an image processing apparatus constituting a main part of the surface flaw inspection apparatus.
FIG. 3 is a flowchart illustrating one embodiment of image processing according to the present invention.
FIG. 4 is a schematic diagram illustrating setting of an allowable range in a case where an abnormal portion is shifted in a plate width direction in grouping of the abnormal portions.
FIG. 5 is a schematic diagram showing an example of grouping of abnormal portions generated at an edge portion of a steel plate.
FIG. 6 is a flowchart illustrating another embodiment of the image processing according to the present invention.
FIG. 7 is a diagram illustrating an example in which a secondary determination is performed based on a number ratio and a norm value, and is a characteristic amount distribution diagram of a flaw and a water droplet.
[Explanation of symbols]
Reference Signs List 1 steel plate (band) 10 video camera 16 pulse generator 20 image processing device 22 computer S imaging point A distance from imaging point to inspector B buffering maximum processing length

Claims

In a method of inspecting a surface flaw of a band which images a surface of a moving band and inspects a surface flaw by performing image processing on the picked-up image, a defect type and a harmful / harmless harmless / harmless portion are extracted for each abnormal portion extracted in frame units from the picked-up image. A primary determination is performed, and information on abnormal portions that occur continuously in the moving direction of the band is buffered and grouped below a predetermined buffering maximum processing length, and the primary determination is performed for each group of the abnormal portions. A method of inspecting a surface flaw of a band-shaped body, wherein harmfulness / harmlessness is secondarily determined based on information of an abnormal part determined to be harmful and an abnormal part determined to be harmless.

The secondary determination determines harmful / harmless based on the number of abnormal portions determined to be harmful in the primary determination and the number of abnormal portions determined to be harmless, and then determined as harmful in the primary determination. The method for inspecting a surface flaw of a belt-like body according to claim 1, wherein harmfulness / harmlessness is determined based on a characteristic amount of the abnormal part and the abnormal part determined to be harmless.

Harmful / harmless for an abnormal part determined to be harmful, based on the norm value between the representative value of the feature value distribution of the abnormal part determined to be harmless in the primary determination and each abnormal part determined to be harmful in the primary determination The surface flaw inspection method according to claim 2, wherein:

In the method for inspecting the surface flaws of a band, which images the surface of the moving band, and inspects the surface flaws based on the captured image signal,
(A) extracting an abnormal part from the captured image in frame units; (b) obtaining a characteristic amount of the abnormal part with respect to each extracted abnormal part; and (c) for each abnormal part, a defect type of the abnormal part based on the characteristic amount and Primary determination of harmful / harmless (d) Information of abnormal parts continuously occurring in the moving direction of the band is buffered and grouped below a predetermined band maximum processing length (e). For each group of abnormal parts, determine the number ratio P of the number of abnormal parts that are primarily determined to be harmful to the number of all abnormal parts. (F) For each group of abnormal parts, the number ratio P is a threshold Po. (G1) In the case of a group in which the number ratio P exceeds the threshold Po, a secondary determination is made to maintain the harmful / harmless primary determination result as it is (g2) When the number ratio P is equal to or less than the threshold Po A guru (H) norm value N between the representative value of the feature value distribution of the abnormal portion that is primarily determined to be harmless and the feature value of each abnormal portion that is primarily determined to be harmful. (I1) When the norm value N exceeds the threshold value No, the abnormal part determined to be harmful in the primary determination is secondarily determined to be harmful as it is in the primary determination (i2). A method for inspecting a surface flaw of a band-shaped body, comprising: when a threshold value is equal to or less than a threshold value, performing a secondary determination that an abnormal portion determined to be harmful in the primary determination is harmless.

In a belt-like surface flaw inspection device including an imaging device that captures the surface of a moving belt-like material and an image processing device that processes the captured image to determine the type of flaw and harmfulness / harmlessness, a frame unit is obtained from the captured image. Primary determining means for determining the type of flaw and harmfulness / harmlessness for each extracted abnormal portion, and buffering information on abnormal portions which continuously occur in the moving direction of the band with a predetermined buffering maximum processing length or less. Buffering means for grouping the abnormal parts, and secondary judgment for judging harmful / harmless based on the information of the abnormal part judged to be harmful in the primary judgment and the abnormal part judged to be harmless for each group of the abnormal parts. Means for inspecting the surface flaw of a belt-like body.

A first determination unit configured to determine harmful / harmless based on the number of abnormal portions determined to be harmful in the primary determination and the number of abnormal portions determined to be harmless in the primary determination; 7. The apparatus for inspecting a surface defect of a belt-like body according to claim 6, further comprising a second determination unit that determines harmfulness / harmlessness based on characteristic amounts of the abnormal part determined to be harmful and the abnormal part determined to be harmless.

The second determination means is determined to be harmful by the norm value between the representative value of the feature distribution of the abnormal part determined to be harmless by the primary determination and each abnormal part determined to be harmless by the primary determination. The surface flaw inspection device according to claim 6, wherein the abnormal portion is determined to be harmful or harmless.

An imaging apparatus for imaging the surface of a moving band, and an image processing apparatus for processing a captured image to determine a flaw type and harmful / harmless, and a surface flaw inspection apparatus for a band,
Abnormal portion extracting means for extracting an abnormal portion from the captured image in frame units;
A feature amount measuring means for obtaining a feature amount of the abnormal portion for each extracted abnormal portion;
Primary determining means for determining, for each abnormal part, a flaw type and harmful / harmless of the abnormal part based on the characteristic amount;
Buffering means for buffering and grouping the information of the abnormal part which continuously occurs in the moving direction of the belt-shaped body at a predetermined buffering maximum processing length or less,
A number ratio measuring means for obtaining a ratio P of the number of abnormal parts which are determined to be harmful to the number of all abnormal parts for each group of abnormal parts, and a number ratio P for each abnormal group. Means for determining whether the number exceeds
If the number ratio P exceeds the threshold value Po, the secondary determination unit determines to maintain the harmful / harmless primary determination result as it is,
In the case of the group in which the number ratio P is equal to or less than the threshold value Po, the norm value N between the representative value of the characteristic amount distribution of the abnormal part which is primarily determined to be harmless and the characteristic amount of each abnormal part which is primarily determined to be harmful A norm value measuring means for determining
Norm value determining means for determining whether or not the norm value N exceeds a threshold No;
When the norm value N exceeds the threshold No, first determining means for secondary determining that the abnormal part determined to be harmful in the primary determination is harmful,
When the norm value N is equal to or smaller than a threshold No, a second determining means for secondary determining that the abnormal portion determined to be harmful in the primary determination is harmless is secondly determined.