JP4422923B2 - Imprint collation apparatus, imprint collation method, and program for causing computer to execute the method - Google Patents

Description

【００５９】
つぎに、平滑化部９０２は登録パターンおよび回転補正部９０１による回転補正後の評価パターンそれぞれに平滑化フィルタをかける機能部である。平滑化フィルタをかけることで、画素単位の微小な印影のずれを吸収させることができる。
【００６０】
つぎに、距離値算出部９０３は平滑化部９０２で平滑化後の登録パターンと評価パターンとの距離値Ｄを算出する機能部である。この距離値の算出にはどのような計算式をもちいるのであってもよいが、実施の形態１では下記計算式によることとする。
【数２】

【００６１】
なお、画素値の差の２乗でなくその絶対値の総和を取るようにしてもよい。また、登録パターン・評価パターンのいずれにおいても画素値は１（画素ありすなわち黒）か０（画素なしすなわち白）のいずれかであり、同一位置の画素がともに黒であるかともに白である場合の画素値の差は常に０、一方が黒で他方が白である場合の画素値の差は常に１であることから、この差の２乗や絶対値の総和を取るのでなく、端的に両パターン間で値が異なる画素の総数を取るようにしてもよい。画素値の差の２乗の総和に代えて、画素値の差の絶対値の総和や、あるいは画素値の異なる画素の総数により距離値を把握することで、計算量を低減させ処理速度を向上させることができる。
【００６２】
つぎに、一致判定部９０４は距離値算出部９０３で算出された上記距離値とあらかじめ保持している閾値との大小比較から、登録パターンと評価パターンとの一致の有無を判定する機能部である。具体的には、距離値が閾値を下回ったときに両者は一致したと判定し、逆に距離値が閾値以上であるときは両者は一致しないと判定する。
【００６３】
なお、一致判定部９０４により評価パターンが登録パターンと一致しないと判定された場合、回転補正部９０１は評価パターン内の回転補正後の印影をさらに１８０度回転させる。したがって、元の評価パターンからの印影の回転角度はＫ＋１８０度となる。
【００６４】
このように半回転させると、人間にはそのようには認知できないが数学的には類似するパターンが現れるため、再度照合をおこなうことで照合率を確実に向上させることができる。上述のように第１照合部２０３では圧縮後の画像同士を照合するため、処理量が低減する代わりに照合率も低下してしまう。そこで、このように半回転後の再照合を含ませることで、圧縮に起因する照合率の低下分を補おうとするものである。なお、圧縮なしで照合をおこなうよりは、圧縮して照合の上さらに再照合をおこなったほうが、全体としての処理量は格段に少なくて済む。
【００６５】
そして、再照合によっても評価パターンが登録パターンと一致しないと判定された場合に限り、第１照合部２０３は前処理部２０２から入力した評価パターンをつぎの照合部、すなわち第２照合部２０４に出力する。
【００６６】
図２の残りの機能部の説明に戻り、つぎに図１０は実施の形態１による第２照合部２０４の機能構成をより詳細に示す説明図である。これは図９に示す第１照合部２０３の機能構成に、再評価パターン作成部１０００を追加したものである。
【００６７】
再評価パターン作成部１０００は、入力した評価パターンからその再評価パターンを作成する機能部である。この再評価パターンとは、評価パターン内の印影の輪郭を強調して、より明瞭な印影としたものである。たとえば、図８に示す評価パターンには印影のかすれが見られるので、エッジ強調フィルタをかけてこのかすれを補正したものを再評価パターンとする。
【００６８】
このフィルタとしては、実施の形態１では処理量の少なさのわりに良好なエッジ抽出がおこなえることで知られるｓｏｂｅｌフィルタを利用するが、画像の輪郭を強調できるものであれば、他のどのようなフィルタを利用するのであってもよい。フィルタ後、エッジ強調前の画像とエッジ強調後の画像とを重ね合わせたものを再評価パターンとする。
【００６９】
そして、圧縮部１００１、回転補正部１００２、平滑化部１００３、距離値算出部１００４および一致判定部１００５は、評価パターンの代わりにこの再評価パターンに対して、対応する図９の各部と同一の処理をおこなう。そして、再評価パターンが登録パターンと一致しない場合に、第２照合部２０４は再評価パターン作成部１０００で作成した再評価パターンをつぎの照合部、すなわち第３照合部２０５に出力する。
【００７０】
図２の残りの機能部の説明に戻り、つぎに図１１は実施の形態１による第３照合部２０５の機能構成をより詳細に示す説明図である。これは図１０に示す第２照合部２０４の機能構成から、圧縮部１００１を削除した構成である。
【００７１】
ただし、第３照合部２０５の回転補正部１１０１および距離値算出部１１０３は、第１照合部２０３の回転補正部９０１や距離値算出部９０３、あるいは第２照合部２０４の回転補正部１００２や距離値算出部１００４のように、評価パターンあるいは再評価パターンが登録パターンと一致しない場合に当該パターン内の印影を半回転して再照合する機能を有していない。これは第３照合部２０５では、第１照合部２０３や第２照合部２０４におけるような圧縮をおこなわないので、圧縮の分の照合率の低下を再照合により埋め合わせる必要がないためである。
【００７２】
なお、この第３照合部２０５でも登録パターンとの照合は評価パターンでなく再評価パターンについておこなうが、第２照合部２０４で作成された再評価パターンを流用できるため、実際には再評価パターン作成部１１００による再評価パターンの再作成はおこなっていない。上記以外の各部の機能は、対応する図９や図１０の各部の機能と同一である。そして、再評価パターンが登録パターンと一致しない場合に、第３照合部２０５は第２照合部２０４から入力した再評価パターンをつぎの照合部、すなわち第４照合部２０６に出力する。
【００７３】
図２の残りの機能部の説明に戻り、つぎに図１２は実施の形態１による第４照合部２０６の機能構成をより詳細に示す説明図である。これは図１１に示す第３照合部２０５の機能構成に、位置補正部１２０２を追加した構成である。
【００７４】
位置補正部１２０２は、登録パターン内の印影に対する再評価パターン内の印影の、相対的な位置ずれを補正する機能部であり、印影領域特定部１２０２ａ、辺位置補正部１２０２ｂおよび相関値算出部１２０２ｃを含む構成である。
【００７５】
位置補正部１２０２は、まずその印影領域特定部１２０２ａにより、登録パターンおよび回転補正部１２０１による回転補正後の再評価パターンについてそれぞれの印影が存在する領域、すなわち印影領域を特定する。この印影領域の特定は、図３に示す前処理部２０２の印影領域特定部３０１ａと同様、画素値のヒストグラムにもとづいておこなう。
【００７６】
つぎに、位置補正部１２０２はその辺位置補正部１２０２ｂにより、再評価パターン内の印影領域のいずれか一辺が、登録パターン内の印影領域の対応する一辺と重なり合うような位置に、再評価パターン内の印影を移動する。この対応する辺とは、上辺に対しては上辺、下辺に対しては下辺、左辺に対しては左辺、右辺に対しては右辺である。
【００７７】
図１３は、実施の形態１による辺位置補正部１２０２ｂがおこなう印影の移動の様子を示す説明図である。登録パターン内の印影領域の上辺と、再評価パターン内の印影領域の上辺とが一致するよう、再評価パターン内の印影を移動したものが同図（ａ）である。同様に、下辺同士を一致させたものが同図（ｂ）、左辺同士を一致させたものが同図（ｃ）、右辺同士を一致させたものが同図（ｄ）である。このように、辺位置補正部１２０２ｂでは印影領域の４辺それぞれの重ね合わせをおこなって、４つの補正結果を作成する。
【００７８】
つぎに、位置補正部１２０２はその相関値算出部１２０２ｃにより、それぞれの補正後の再評価パターンと登録パターンとの相関値を算出し、このうち相関値が最大となるものを選択する。そして、平滑化部１２０３での平滑化、距離値算出部１２０４での距離値算出および一致判定部１２０５での一致判定は、それぞれこの選択した再評価パターンについておこなう。
【００７９】
なお、相関値算出部１２０２ｃではどのようにして上記相関値を求めるのであってもよいが、実施の形態１では同一位置の画素のうち登録パターンと再評価パターンとで値が異なる画素の総数、すなわち登録パターンでは白であるのに再評価パターンでは黒になっている画素や、あるいは逆に登録パターンでは黒であるのに再評価パターンでは白になっている画素の総数を両パターンの相関値としている。
【００８０】
これもパターン間の相関のなさを示すという点では、一種の相関値ではあるものの、一般的な相関値と異なり値が小さいほど両者の相関は高くなる。そのため、位置補正部１２０２は実際には、相関値算出部１２０２ｃにより算出された相関値が最大となるものでなく、最小となる再評価パターンを一つ選択している。
【００８１】
なお、この第４照合部２０６でも登録パターンとの照合は評価パターンでなく再評価パターンについておこなうが、第２照合部２０４で作成された再評価パターンを流用できるため、実際には再評価パターン作成部１２００による再評価パターンの再作成はおこなっていない。上記以外の各部の機能は、対応する図９、図１０あるいは図１１の各部の機能と同一である。
【００８２】
なお、未知パターン入力部２０１、前処理部２０２、第１照合部２０３、第２照合部２０４、第３照合部２０５および第４照合部２０６は、それぞれＨＤ１０５やＦＤ１０７、あるいはＣＤ−ＲＯＭ１１５などの各種記録媒体からＲＡＭ１０３に読み出されたプログラムの命令にしたがって、ＣＰＵ１０１などが命令処理を実行することにより、各部の機能を実現するものである。
【００８３】
（実施の形態２による印影照合装置）
さて、上述した実施の形態１による印影照合装置では、第４照合部２０６で登録パターン内の印影と再評価パターン内の印影との角度ずれを補正した後、さらに印影領域の上辺・下辺・左辺・右辺の重ね合わせをおこなうことで印影の位置ずれを補正したが、このときさらに詳細に、４辺だけでなく左上隅・左下隅・右上隅・右下隅の４頂点の重ね合わせもおこなうようにしてもよい。また、角度ずれの補正の後だけでなく、当該補正の前にも位置ずれの補正をおこなうようにしてもよい。
【００８４】
実施の形態２による印影照合装置のハードウエア構成は、図１に示した実施の形態１による印影照合装置のそれと同様であるので説明を省略する。図１４は、実施の形態２による印影照合装置の構成を機能的に示す説明図である。これは図２に示す、実施の形態１による印影照合装置の機能構成において、第３照合部２０５および第４照合部２０６をそれぞれ第５照合部１４０５および第６照合部１４０６に置き換えたものである。この第５照合部１４０５および第６照合部１４０６は、それぞれ下記に示す方法で登録パターンと再評価パターンとの照合をおこなう機能部である。
【００８５】
（方法５）
第５照合部１４０５における照合方法である。上述の再評価パターン内の印影をｍ度刻みに回転して、各角度における登録パターンとの相関値を求め、この相関値が最大となる角度において、さらに再評価パターン内の印影を移動することで、印影領域の各辺および各頂点の位置合わせをおこなう。その後両パターン間の距離値を求めて、所定の閾値との比較により両者の一致の有無を判定する。方法４との差異は、辺同士だけでなく頂点同士の位置合わせもおこなうこと、位置合わせをおこなわない場合もあることの二点である。
【００８６】
（方法６）
第６照合部１４０６における照合方法である。まず、上述の再評価パターン内の印影を移動することで、登録パターン内の印影領域との各辺および各頂点の位置合わせをおこなう。つぎに、再評価パターン内の印影をｍ度刻みに回転して、各角度における登録パターンとの相関値を求め、この相関値が最大となる角度において、再度上記と同様の位置合わせをおこなう。その後両パターン間の距離値を求めて、所定の閾値との比較により両者の一致の有無を判定する。方法５との差異は、方法５と同様の位置合わせを印影の回転の前後に重複しておこなう点のみである。
【００８７】
図１５は、実施の形態２による第５照合部１４０５の機能構成をより詳細に示す説明図である。これは図１２に示す、実施の形態１による第４照合部２０６の位置補正部１２０２に、角位置補正部１５０２ｃを追加した構成である。図１６は、実施の形態２による角位置補正部１５０２ｃがおこなう印影の移動の様子を示す説明図である。同図に示すのは、（ａ）、（ｂ）のいずれも印影領域の左上隅の頂点同士を一致させる例である。角位置補正部１５０２ｃは、同様に左下隅・右上隅および右下隅の各頂点の重ね合わせもおこなう。
【００８８】
そして、相関値算出部１５０２ｄは、辺位置補正部１５０２ｂまたは角位置補正部１５０２ｃによる補正後の８つの再評価パターンそれぞれについて、登録パターンとの相関値を算出する。また、これとあわせて補正前の再評価パターンと登録パターンとの相関値も算出する。なお、上述のように登録パターン・再評価パターンのいずれにおいても印影の中心は画像の中心に来るようあらかじめ調整されているので、補正前の再評価パターンは、登録パターンと印影領域の中心同士を重ね合わせたものであると見ることもできる。
【００８９】
この後、上記９つの再評価パターンのうち登録パターンとの相関値が最大となるものを選択し、この選択した再評価パターンについて平滑化部１５０３での平滑化、距離値算出部１５０４での距離値算出および一致判定部１５０５での一致判定をそれぞれおこなう。そして、この再評価パターンが登録パターンと一致しない場合に、第５照合部１４０５は第２照合部１４０４から入力した再評価パターンをつぎの照合部、すなわち第６照合部１４０６に出力する。
【００９０】
つぎに、図１７は実施の形態２による第６照合部１４０６の機能構成をより詳細に示す説明図である。これは図１５に示す第５照合部１４０５の機能構成と同一であって、ただその位置補正部１７０２は回転補正部１７０１による回転補正の後だけでなく、その前にも同様の位置補正をおこなう機能を有している。
【００９１】
すなわち、第５照合部１４０５から入力した、その印影領域の中心が登録パターンのそれと一致している再評価パターンと、その印影領域のいずれかの辺または頂点が登録パターンのそれと一致するように補正後の８つの再評価パターンとについて、それぞれ登録パターンとの相関を調べ、最も相関の高いものを一つ選択する。
【００９２】
そして、この選択した再評価パターンにおいて、印影を回転させて登録パターンとの相関が最も高くなる角度を求め、この角度に印影を回転しただけの再評価パターンと、その上さらに印影領域の各辺または各頂点が一致するように補正した８つの再評価パターンについて、登録パターンとの相関を調べる。そして、最も相関の高いものについて登録パターンとの距離値の算出、および一致の有無の判定をおこなう。
【００９３】
このように回転補正の前にも位置補正をおこなうと、印鑑が真円である場合、同一印鑑から採取された印影であれば外周円同士が一致する位置に補正される可能性が高い。回転補正の前に、印影の形状のみに注目したおおまかなマッチングをおこなっておくようにする、といってもよい。あらかじめ外周円の位置を合わせておくことで、その後に算出される回転角度や距離値がより正確なものとなり、したがって本来なら登録パターン内の印影と一致するはずのものが、印影位置のずれにより誤って不一致と判定される確率を下げることができる。
【００９４】
なお、未知パターン入力部１４０１、前処理部１４０２、第１照合部１４０３、第２照合部１４０４、第５照合部１４０５および第６照合部１４０６は、それぞれＨＤ１０５やＦＤ１０７、あるいはＣＤ−ＲＯＭ１１５などの各種記録媒体からＲＡＭ１０３に読み出されたプログラムの命令にしたがって、ＣＰＵ１０１などが命令処理を実行することにより、各部の機能を実現するものである。
【００９５】
図１８は、実施の形態１および２で使用した方法１〜６の手順を対比して示す説明図である。方法１はステップＳ１８０２の、圧縮部９００による登録パターン・評価パターンの圧縮から始まり、ステップＳ１８０３は省略して続くステップＳ１８０４で、印影回転部９０１ａ・相関値算出部９０１ｂによる３度刻みの印影の回転と各角度での相関値の算出とをおこなう。
【００９６】
続くステップＳ１８０５で、回転角度算出部９０１ｃにより相関値が最大となるＫを回転角度と特定（算出）し、ステップＳ１８０６で回転補正部９０１により、この角度Ｋだけの回転補正を評価パターンについておこなう。ステップＳ１８０７およびステップＳ１８０８は省略し、つぎにステップＳ１８０９で、平滑化部９０２による両パターンの平滑化、続くステップＳ１８１０で、距離値算出部９０３による距離値Ｄの算出、さらにステップＳ１８１１で、一致判定部９０４による上記距離値と閾値との大小比較、すなわち両パターンの一致の有無の判定をおこなう。
【００９７】
そして、ステップＳ１８１１で不一致と判定された場合に限り、続くステップＳ１８１２で回転補正部９０１により、評価パターン内の印影をさらに半回転させた上、ステップＳ１８１３で距離値算出部９０３による距離値Ｄの再算出、ステップＳ１８１４で一致判定部９０４による一致の有無の再判定をおこなう。
【００９８】
方法２はステップＳ１８０１の、再評価パターン作成部１０００による再評価パターンの作成、すなわち評価パターンに対するエッジ強調から始まり、その後ステップＳ１８０２以降の処理は上記方法１と同一である。
【００９９】
方法３も同様にステップＳ１８０１から始まり、ステップＳ１８０２およびステップＳ１８０３は省略して、未圧縮の登録パターン・再評価パターンについて、ステップＳ１８０４以降の処理を上記方法１および２と同様におこなう。ただし、ステップＳ１８１１で不一致と判定された場合の、ステップＳ１８１２〜ステップＳ１８１４の処理は省略する。
【０１００】
方法４および５は、ステップＳ１８０６とステップＳ１８０９との間でステップＳ１８０７またはステップＳ１８０８、すなわち位置補正部１２０２または１５０２による位置補正をおこなうほかは、上記方法３と同一の処理である。ただし、この位置ずれ補正は方法４においては印影領域の各辺について、方法５においては各辺および各頂点についておこなう。
【０１０１】
また、方法６はステップＳ１８０１とステップＳ１８０４との間でステップＳ１８０３、すなわち位置補正部１７０２による位置補正をおこなうほかは、上記方法５と同一の処理である。なお、このステップＳ１８０３の処理はステップＳ１８０８の処理と同一であり、要するにステップＳ１８０６の回転補正の前後で同一の位置補正を繰り返していることになる。
【０１０２】
つぎに、図１９はステップＳ１８０７における位置補正の手順を詳細に示す説明図である。ステップＳ１９０１で、第４照合部２０６はその印影領域特定部１２０２ａにより、登録パターンおよび回転補正部１２０１による回転補正後の再評価パターンそれぞれについて、その印影領域を特定する。
【０１０３】
そして、ステップＳ１９０２で辺位置補正部１２０２ｂにより、まず両パターンの印影領域の上辺同士が一致するように、再評価パターン内の印影の移動をおこなう。つぎに、ステップＳ１９０３〜ステップＳ１９０５で同様に、印影領域の下辺同士、左辺同士、および右辺同士が一致するような印影移動をそれぞれおこなう。なお、このステップＳ１９０２〜ステップＳ１９０５の補正は、第３照合部２０５から入力した再評価パターンについてそれぞれ独立におこなうので、その順序は図示の順序とは異なっていてもよい。
【０１０４】
そして、ステップＳ１９０６で相関値算出部１２０２ｃにより、ステップＳ１９０２〜ステップＳ１９０５によるそれぞれの移動後の再評価パターンと、登録パターンとの相関値を算出する。そして、ステップＳ１９０７で位置補正部１２０２は、上記４つの再評価パターンのうち登録パターンとの相関が最も高いものを一つ選択する。
【０１０５】
つぎに、図２０はステップＳ１８０３およびステップＳ１８０８における位置補正の手順を詳細に示す説明図である。ステップＳ２００１で、第５照合部１４０５の印影領域特定部１５０２ａまたは第６照合部１４０６の印影領域特定部１７０２ａにより、登録パターンおよび再評価パターンのそれぞれについて印影領域を特定する。
【０１０６】
そして、ステップＳ２００２〜ステップＳ２００５で辺位置補正部１５０２ｂまたは１７０２ｂにより、図１９のステップＳ１９０２〜ステップＳ１９０５と同様に、印影領域の対応する辺同士が一致するような位置への印影の移動をそれぞれおこなう。つぎに、ステップＳ２００６〜ステップＳ２００９で角位置補正部１５０２ｃまたは１７０２ｃにより、印影領域の対応する頂点同士が一致するような位置への印影の移動をそれぞれおこなう。なお、このステップＳ２００２〜ステップＳ２００９の順序は入れ違っていてもよい。
【０１０７】
そして、ステップＳ２０１０で相関値算出部１５０２ｄまたは１７０２ｄにより、ステップＳ２００２〜ステップＳ２００９によるそれぞれの移動後の再評価パターン、および移動前の再評価パターンについて、登録パターンとの相関値をそれぞれ算出する。そして、ステップＳ２０１１で位置補正部１５０２または１７０２は、上記９つの再評価パターンのうち登録パターンとの相関が最も高いものを一つ選択する。
【０１０８】
（実施の形態１および２による印影照合装置の性能評価）
さて、実施の形態１および２で説明した印影照合装置の実際の照合能力であるが、本出願人は下記に示す照合実験によりその性能評価をおこなっているので、この発明による印影照合装置の有用性を示すため、以下上記実験の手順と結果とを詳細に説明する。
【０１０９】
実験に使用する印影の原画像を図２１に示す。この実験では一つの印鑑から、１色につき図示するような３６個の印影を採取する。このとき、いったん印鑑にインクなどを付けたら、連続して３つの印影を作成するようにする。たとえば、図２１において２段目最左端の印影は１段目最左端の印影の押捺後そのまま朱肉を付けずに押捺したものであり、３段目最左端の印影は２段目最左端の印影の押捺後、さらにそのまま朱肉を付けずに押捺したものである。
【０１１０】
このように、３連続の押印を１２回繰り返すことで３６個の印影を得るが、このうち半分の６回では、意図的に印影に回転を持たせるようにする。図２１において１〜３段目の印影は普通にまっすぐ押捺したもの、４〜６段目の印影は意図的に印鑑を回転させて押捺したものである。
【０１１１】
実験では４９種類の印鑑を使用するので、採取する印影の総数は１色あたり１７６４（＝３６×４９）個である。採取した印影は１個ずつ、「ｎ＋印鑑番号＋インクの色＋ｙｎｏ＋ｘｎｏ．ｂｍｐ」という名称のファイルに保存する。ここで「印鑑番号」とは、４９種類の印鑑を識別するために各印鑑に付与する「０００」から「０４８」までのいずれかの数値である。
【０１１２】
「インクの色」とは朱肉を表す「ｏｒｇ」、青のスタンプインクを表す「ｂｌｅ」、黒のスタンプインクを表す「ｂｌｋ」および赤のスタンプインクを表す「ｒｅｄ」のいずれかの文字列であるが、当面は朱肉のみを考えるものとする。また、「ｙｎｏ」「ｘｎｏ」は図２１に示すように、両者の組み合わせにより同一印鑑について採取された３６個の印影を識別することのできる「００」から「０５」までの数値である。
【０１１３】
したがって、たとえば図２１に示す１段目最左端の「長谷川」はファイル「ｎ０００ｏｒｇ００００．ｂｍｐ」として、また６段目最左端の「長谷川」はファイル「ｎ０００ｏｒｇ０５００．ｂｍｐ」として、それぞれ保存されることになる。また、同図の「長谷川」とは別の印鑑、たとえば「青山」から採取されたある印影は「ｎ０２３ｏｒｇ００００．ｂｍｐ」、「池谷」から採取されたある印影は「ｎ０３８ｏｒｇ００００．ｂｍｐ」などに保存されることになる。なお、画像サイズはいずれも２５６×２５６ピクセルであり、各画素の濃淡値は２５６階調で表現される。
【０１１４】
さらに、これらの原画像からその２値化画像、評価パターンおよび再評価パターンを、それぞれ実施の形態１および２と同様にして作成する。これらの２値化画像を図２２に、評価パターンを図２３に、再評価パターンを図２４にそれぞれ示す。図２２に示す２値化画像では左端に寄ってしまっていた６段目最左端の印影が、その欠けやかすれにもかかわらず、図２３に示す評価パターンでは画像全体の中央に正確に移動されていることが分かる。また、図２３に示す評価パターンではかすれていた３段目や６段目の印影が、図２４に示す再評価パターンではよりくっきりと明瞭になっていることが分かる。
【０１１５】
なお、同一の原画像から作成された２値化画像、評価パターンおよび再評価パターンは、それぞれ原画像と同一名称のファイルとして保存され、位置するディレクトリの差異によって区別される。また、図２３に示すヒストグラム２３００や印影領域２３０１、画像の縦横中心線２３０２などは、図２２との差異が分かりやすいように表示しているものであって、評価パターンのファイル内に実際にこれらのデータが格納されているわけではない。
【０１１６】
実験では図２３に示す評価パターンのうち、ファイル「ｎ＊＊＊ｏｒｇ００００．ｂｍｐ」（＊＊＊は「０００」から「０４８」までのいずれかの印鑑番号）に格納されている４９個の評価パターンをそれぞれ登録パターンとする。そして、これら４９個の登録パターンと残り１７１５個の評価パターンとのすべての組み合わせについて、上述の方法１により順次距離値を算出する。また、この４９個の登録パターンと、上記１７１５個の評価パターンから作成された１７１５個の再評価パターンとのすべての組み合わせについて、上述の方法２〜６による距離値算出を順次おこなう。
【０１１７】
そして、これらの距離値から逆に、少なくとも別個の印鑑から採取された登録パターンと入力パターンとは必ず不一致と判定されるような閾値を、４９個の登録パターンごと、かつ上記６つの方法ごとに求める。
【０１１８】
図２３に示す印鑑番号「０００」の印鑑「長谷川」について、上記閾値を求める場合を例として説明する。まず、図２３の１段目最左端に示す「長谷川」すなわちこの印鑑の登録パターンと、残り３５個の「長谷川」すなわち登録パターンと同一の印鑑から採取された評価パターンとの距離値を方法１によって求める。左上隅の「長谷川」から見て残り３５個の「長谷川」は真印の関係にあるので、この登録パターンと各評価パターンとの間で算出される３５個の距離値を、まとめてこの登録パターンの真印距離と呼ぶ。そして、算出した真印距離を「ｎ＋印鑑番号＋インクの色＋登録パターンのｙｎｏ＋登録パターンのｘｎｏ．ｔｒｕｅ」という名称の距離値ファイル、この例では「ｎ０００ｏｒｇ００００．ｔｒｕｅ」として保存する。
【０１１９】
つぎに、図２３の１段目最左端に示す「長谷川」すなわちこの印鑑の登録パターンと、同図では背後に隠れて見えない１７２８（＝３６×４８）個の「青山」「池谷」など、すなわち登録パターンとは別個の印鑑から採取された評価パターンとの距離値を方法１によって求める。「長谷川」から見て「青山」「池谷」などは偽印の関係にあるので、この登録パターンと各評価パターンとの間で算出される１７２８個の距離値を、まとめてこの登録パターンの偽印距離と呼ぶ。そして、算出した偽印距離を「ｎ＋印鑑番号＋インクの色＋登録パターンのｙｎｏ＋登録パターンのｘｎｏ．ｆａｋｅ」という名称の距離値ファイル、この例では「ｎ０００ｏｒｇ００００．ｆａｋｅ」として保存する。
【０１２０】
さらに、方法２〜６による距離値の算出も同様におこない、各方法により算出された「長谷川」の真印距離は上記「ｎ０００ｏｒｇ００００．ｔｒｕｅ」に、偽印距離は上記「ｎ０００ｏｒｇ００００．ｆａｋｅ」に、それぞれ追記してゆく。ただし、方法２〜６による距離値算出については、対象となるのは評価パターンではなく再評価パターンである。また、残り４８個の登録パターンについても同様に、方法１〜６により求めた真印距離のファイルと偽印距離のファイルとをそれぞれ作成してゆく。
【０１２１】
求める閾値、すなわち少なくとも別個の印鑑から採取された登録パターンと入力パターンとは必ず不一致と判定されるような閾値とは、「長谷川」かつ方法１の場合、具体的には「ｎ０００ｏｒｇ００００．ｆａｋｅ」に記述された偽印距離のうち、方法１により算出された１７２８個の偽印距離の最小値である。すなわち、登録パターン「長谷川」が採取された印鑑とは別の印鑑から採取された評価パターンの方法１による距離値は、必ずこの最小値以上となることから、逆に方法１による距離値がこの最小値以上である評価パターンは登録パターンとは一致しないと判定するようにしておけば、登録パターン「長谷川」と評価パターン「青山」「池谷」などは常に不一致と判定されることになる。あるいは、偽印距離の最小値を閾値とした判定では、登録パターンに対して偽印の関係にある評価パターンが偽印として棄却される確率は１００％となるといってもよい。
【０１２２】
ただし、登録パターン「長谷川」と同一の印鑑から採取された３５個の評価パターン「長谷川」の中には、別の印鑑から採取された「青山」「池谷」などの評価パターンよりも距離値の大きくなるものがあり、このような「長谷川」は真印であるにもかかわらず上記閾値では偽印と判定されてしまうので、偽印が常に偽印とされるからと言って、必ずしも真印が常に真印とされるわけではない。言い換えれば、偽印棄却率１００％であることは必ずしも真印受理率１００％であることを意味しない。
【０１２３】
このような判定誤りは、要するに真印距離と偽印距離とが特定の値を境に完全に分離されるような、理想的な距離値の算出ができていないことを示しており、実施の形態１および２で説明した四段形の印影照合は、種々の方式による距離値算出を段階的に組み合わせることで、距離値による真印と偽印の切り分けの精度を向上させる試みと言うこともできる。また、この実験の目的も、偽印棄却率１００％の条件下での真印受理率が実施の形態１および２による印影照合装置では何％となるか（理想的には１００％）を測定する点にある。
【０１２４】
実験では実施の形態１による印影照合装置の登録パターン記憶部２００に、上記４９個の登録パターンを、またその未知パターン入力部２０１に、登録パターンの原画像を除く１７１５（＝３６×４９−４９）個の印影の原画像を、その第１照合部２０３〜第４照合部２０６に、それぞれ方法１〜４により算出された各登録パターンの閾値を与える。そして、これら４９個の登録パターンと１７１５個の入力パターンとの照合を順次おこなう。
【０１２５】
図２５は、上記照合実験の結果を示すグラフである。同グラフは、実施の形態１による印影照合装置に偽印棄却率１００％となるような上記閾値を与えたときの、同装置の照合率を示している。また、印影のかすれの有無の影響を調べるため、照合率は朱肉を付けて１回目に押捺した印影５３９（＝１１×４９）個、それに続けて２回目に押捺した印影５８８（＝１２×４９）個、さらにそれに続けて３回目に押捺した印影５８８（＝１２×４９）個ごとに算出している。なお、グラフ中の数字はどの登録パターンとも一致しない、言い換えればどの登録パターンに対しても真印ではないとして棄却された印影の個数である。
【０１２６】
たとえば、方法１すなわち第１照合部２０３による照合で棄却された印影は２７（＝２＋８＋１７）個であり、この後方法２すなわち第２照合部２０４による照合で棄却された印影は７（＝０＋２＋５）個であって、これは方法１で棄却された２７個の印影のうち２０個が方法２によって新たに受理された、いわば救済されたことを示している。そして、第１照合部２０３〜第４照合部２０６を経て最終的に棄却される印影は０個となっており、これはすなわち方法１〜４を組み合わせることにより、偽印棄却率１００％の条件下での真印受理率を１００％とすることができたことを示している。
【０１２７】
しかも図２１に示すように、本実験では入力パターン中に登録パターンとの角度ずれや位置ずれの大きいものを意図的に多く含ませているが、上述した印影の中心位置の補正、角度ずれや位置ずれの補正により、これらの入力パターンについても第１照合部２０３〜第４照合部２０６のいずれかによっていずれかの登録パターンとの同一性が確認されている。すなわち、実施の形態１による印影照合装置は、あるべき位置に押捺されていなかったりあるべき角度で押捺されていなかったりといった問題のある印影でも、登録されているいずれかの印鑑のものである限り、それがいずれかの登録印鑑のものであると必ず特定できたということである。
【０１２８】
また、図２１に示すように、本実験では欠けやかすれなど明瞭性に問題のある入力パターンを意図的に多く使用しているが、照合を評価パターンでなくエッジ強調後の再評価パターンによりおこなうことで、たとえば朱肉を付けて３回目に押捺した印影の照合率を９７．１％（＝（５８８−１７）／５８８）から９９．１％（＝（５８８−５）／５８８）まで上昇させることができている。
【０１２９】
ちなみに、上記はまず第１照合部２０３に１７１５個の入力パターンを与え、ここで棄却されたもののみを第２照合部２０４に与え、さらにここでも棄却されたものを第３照合部２０５に与え・・・のように、段階的・直列的に照合をおこなった場合の結果であるが、上記各部に独立に１７１５個の入力パターンを与え、個別的・並列的に照合をおこなった場合の実験結果を図２６に示す。未圧縮の再評価パターンをもちいて角度ずれと位置ずれの双方の補正をおこなう方法４は、それ単独でも照合率が１００％となっている。
【０１３０】
また、図２５および図２６は登録パターン・入力パターンのいずれも朱肉により押捺されていた場合の照合結果であったが、つぎに朱肉により押捺された登録パターンと、朱肉以外により押捺された入力パターンとの照合実験をおこなう。実施の形態１による印影照合装置の未知パターン入力部２０１に、上記で与えた朱肉による１７１５個の印影に加え、さらに青のスタンプインクにより押捺した１７６４（＝３６×４９）個の印影、黒のスタンプインクにより押捺した１７６４個の印影、および赤のスタンプインクにより押捺した１７６４個の印影の、合計７００７個の原画像を与えて同様に照合をおこなう。
【０１３１】
図２７は、上記照合実験の結果を示すグラフである。方法１〜４を経て最終的に棄却される印影の個数は、朱肉では０個であるが、青インクでは３個、黒インクでは５個、赤インクでは４個である。すなわち、合計１２個の印影がどの印鑑のものでもないとして誤って棄却されたことになる。
【０１３２】
これに対し、同一の原画像を実施の形態２による印影照合装置の未知パターン入力部１４０１に与えて、同様に照合をおこなったときの実験結果を図２８に示す。方法１、２、５および６を経て最終的に棄却される印影は２個に減少している。これはすなわち、方法３および４をより精度の高い方法５および６に置き換えることで、照合率を向上させることができたことを示している。なお、この最終的に棄却された２個は、いずれもかなりにじみのひどい印影であった。
【０１３３】
【発明の効果】
以上説明したように請求項１に記載の発明によれば、印影画像を内部に含む画像について前記印影画像が存在する印影領域を特定する印影領域特定手段と、前記印影領域特定手段によって特定された印影領域の中心位置を算出する印影中心算出手段と、前記画像の中心位置と前記印影中心算出手段によって算出された印影領域の中心位置とが一致するような位置に前記印影画像の位置を補正する中心位置補正手段と、を備えたので、画像内の印影の位置は当該画像の中央に来るようにあらかじめ調整され、これによって、押捺された位置に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３４】
また、請求項２に記載の発明によれば、前記請求項１に記載の発明において、前記印影領域特定手段が、前記画像内の任意の位置を中心として前記印影画像をそれぞれ異なる角度だけ回転させた状態でのそれぞれの印影領域のうち面積が最大となるものを前記画像の印影領域として特定するので、画像内の印影はたとえ欠けやかすれなどがあってもその位置が当該画像の中央に来るようにあらかじめ調整され、これによって、押捺された位置や明瞭性に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３５】
また、請求項３に記載の発明によれば、印影画像を内部に含む第１の画像内の前記印影画像をその中心位置を中心として回転させる回転補正手段と、前記回転補正手段によってその印影画像が回転された前記第１の画像、および印影画像を内部に含む第２の画像のそれぞれについて、前記印影画像が存在する印影領域を特定する印影領域特定手段と、前記印影領域特定手段によって特定された前記第１の画像の印影領域の一辺と前記第２の画像の印影領域の対応する一辺とが一致するような位置に前記第１の画像内の印影画像の位置を補正する辺位置補正手段と、前記辺位置補正手段によってその印影画像の位置が補正された前記第１の画像と、前記第２の画像との距離値を算出する距離値算出手段と、前記距離値算出手段によって算出された距離値にもとづいて前記第１の画像内の印影画像と前記第２の画像内の印影画像とが一致するか否かを判定する一致判定手段と、を備えたので、各画像内の印影はその相対的な角度ずれを補正された後、さらにその相対的な位置ずれを４ヶ所で補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３６】
また、請求項４に記載の発明によれば、前記請求項３に記載の発明において、さらに、前記印影領域特定手段によって特定された前記第１の画像の印影領域の一頂点と前記第２の画像の印影領域の対応する一頂点とが一致するような位置に前記第１の画像内の印影画像の位置を補正する角位置補正手段を備え、前記距離値算出手段は、前記辺位置補正手段または前記角位置補正手段によってその印影画像の位置が補正された前記第１の画像と、前記第２の画像との距離値を算出するので、各画像内の印影はその相対的な角度ずれを補正された後、さらにその相対的な位置ずれを８ヶ所で補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３７】
また、請求項５に記載の発明によれば、印影画像を内部に含む第１の画像および第２の画像のそれぞれについて前記印影画像が存在する印影領域を特定する印影領域特定手段と、前記印影領域特定手段によって特定された前記第１の画像の印影領域の一辺と前記第２の画像の印影領域の対応する一辺とが一致するような位置に前記第１の画像内の印影画像の位置を補正する辺位置補正手段と、前記辺位置補正手段によってその位置が補正された前記第１の画像内の印影画像をその中心位置を中心として回転させる回転補正手段と、前記回転補正手段によってその印影画像が回転された前記第１の画像と、前記第２の画像との距離値を算出する距離値算出手段と、前記距離値算出手段によって算出された距離値にもとづいて前記第１の画像内の印影画像と前記第２の画像内の印影画像とが一致するか否かを判定する一致判定手段と、を備えたので、各画像内の印影はその相対的な位置ずれを４ヶ所で補正された後、さらにその相対的な角度ずれを補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３８】
また、請求項６に記載の発明によれば、前記請求項５に記載の発明において、さらに、前記印影領域特定手段によって特定された前記第１の画像の印影領域の一頂点と前記第２の画像の印影領域の対応する一頂点とが一致するような位置に前記第１の画像内の印影画像の位置を補正する角位置補正手段を備え、前記回転補正手段は、前記辺位置補正手段または前記角位置補正手段によってその位置が補正された前記第１の画像内の印影画像を回転させるので、各画像内の印影はその相対的な位置ずれを８ヶ所で補正された後、さらにその相対的な角度ずれを補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１３９】
また、請求項７に記載の発明によれば、前記請求項５または請求項６に記載の発明において、さらに、前記回転補正手段によってその印影画像が回転された前記第１の画像、および前記第２の画像のそれぞれについて、前記印影画像が存在する印影領域を特定する第２の印影領域特定手段と、前記第２の印影領域特定手段によって特定された前記第１の画像の印影領域の一辺と前記第２の画像の印影領域の対応する一辺とが一致するような位置に前記第１の画像内の印影画像の位置を補正する第２の辺位置補正手段と、を備え、前記距離値算出手段は、前記第２の辺位置補正手段によってその印影画像の位置が補正された前記第１の画像と、前記第２の画像との距離値を算出するので、各画像内の印影はその相対的な位置ずれを４ヶ所で補正された後、さらにその相対的な角度ずれを補正された上、再度その相対的な位置ずれを４ヶ所で補正された後に照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１４０】
また、請求項８に記載の発明によれば、前記請求項７に記載の発明において、さらに、前記第２の印影領域特定手段によって特定された前記第１の画像の印影領域の一頂点と前記第２の画像の印影領域の対応する一頂点とが一致するような位置に前記第１の画像内の印影画像の位置を補正する第２の角位置補正手段を備え、前記距離値算出手段は、前記第２の辺位置補正手段または前記第２の角位置補正手段によってその印影画像の位置が補正された前記第１の画像と、前記第２の画像との距離値を算出するので、各画像内の印影はその相対的な位置ずれを４ヶ所または８ヶ所で補正された後、さらにその相対的な角度ずれを補正された上、再度その相対的な位置ずれを８ヶ所で補正された後に照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合装置が得られるという効果を奏する。
【０１４１】
また、請求項９に記載の発明によれば、印影画像を内部に含む画像について前記印影画像が存在する印影領域を特定する印影領域特定工程と、前記印影領域特定工程で特定された印影領域の中心位置を算出する印影中心算出工程と、前記画像の中心位置と前記印影中心算出工程で算出された印影領域の中心位置とが一致するような位置に前記印影画像の位置を補正する中心位置補正工程と、を含んだので、画像内の印影の位置は当該画像の中央に来るようにあらかじめ調整され、これによって、押捺された位置に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合方法が得られるという効果を奏する。
【０１４２】
また、請求項１０に記載の発明によれば、印影画像を内部に含む第１の画像内の前記印影画像をその中心位置を中心として回転させる回転補正工程と、前記回転補正工程でその印影画像が回転された前記第１の画像、および印影画像を内部に含む第２の画像のそれぞれについて、前記印影画像が存在する印影領域を特定する印影領域特定工程と、前記印影領域特定工程で特定された前記第１の画像の印影領域の一辺と前記第２の画像の印影領域の対応する一辺とが一致するような位置に前記第１の画像内の印影画像の位置を補正する辺位置補正工程と、前記辺位置補正工程でその印影画像の位置が補正された前記第１の画像と、前記第２の画像との距離値を算出する距離値算出工程と、前記距離値算出工程で算出された距離値にもとづいて前記第１の画像内の印影画像と前記第２の画像内の印影画像とが一致するか否かを判定する一致判定工程と、を含んだので、各画像内の印影はその相対的な角度ずれを補正された後、さらにその相対的な位置ずれを４ヶ所で補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合方法が得られるという効果を奏する。
【０１４３】
また、請求項１１に記載の発明によれば、印影画像を内部に含む第１の画像および第２の画像のそれぞれについて前記印影画像が存在する印影領域を特定する印影領域特定工程と、前記印影領域特定工程で特定された前記第１の画像の印影領域の一辺と前記第２の画像の印影領域の対応する一辺とが一致するような位置に前記第１の画像内の印影画像の位置を補正する辺位置補正工程と、前記辺位置補正工程でその位置が補正された前記第１の画像内の印影画像をその中心位置を中心として回転させる回転補正工程と、前記回転補正工程でその印影画像が回転された前記第１の画像と、前記第２の画像との距離値を算出する距離値算出工程と、前記距離値算出工程で算出された距離値にもとづいて前記第１の画像内の印影画像と前記第２の画像内の印影画像とが一致するか否かを判定する一致判定工程と、を含んだので、各画像内の印影はその相対的な位置ずれを４ヶ所で補正された後、さらにその相対的な角度ずれを補正された上で照合され、これによって、押捺された位置や角度に問題のある印影であっても照合元の印影との照合を高精度におこなうことが可能な印影照合方法が得られるという効果を奏する。
【０１４４】
また、請求項１２に記載の発明によれば、前記請求項９〜請求項１１のいずれか一つに記載された方法をコンピュータに実行させることが可能なプログラムが得られるという効果を奏する。
【図面の簡単な説明】
【図１】実施の形態１による印影照合装置のハードウエア構成を示す説明図である。
【図２】実施の形態１による印影照合装置の構成を機能的に示す説明図である。
【図３】実施の形態１による前処理部の機能構成をより詳細に示す説明図である。
【図４】印影が画像の中央に位置していない入力パターンの一例を示す説明図である。
【図５】印影に欠けやかすれがある場合に印影領域が正確に求まらない様子を示す説明図である。
【図６】印影に欠けやかすれがある場合にも印影領域を正確に求めるために、印影を回転させる様子を示す説明図である。
【図７】印影に欠けやかすれがある場合にも印影領域を正確に求めるために、印影を回転させる様子を示す他の説明図である。
【図８】図４に示す入力パターンから作成される評価パターンの一例を示す説明図である。
【図９】実施の形態１による第１照合部の機能構成をより詳細に示す説明図である。
【図１０】実施の形態１による第２照合部の機能構成をより詳細に示す説明図である。
【図１１】実施の形態１による第３照合部の機能構成をより詳細に示す説明図である。
【図１２】実施の形態１による第４照合部の機能構成をより詳細に示す説明図である。
【図１３】実施の形態１による辺位置補正部がおこなう印影の移動の様子を示す説明図である。
【図１４】実施の形態２による印影照合装置の構成を機能的に示す説明図である。
【図１５】実施の形態２による第５照合部の機能構成をより詳細に示す説明図である。
【図１６】実施の形態２による角位置補正部がおこなう印影の移動の様子を示す説明図である。
【図１７】実施の形態２による第６照合部の機能構成をより詳細に示す説明図である。
【図１８】実施の形態１および２で使用した方法１〜６の手順を対比して示す説明図である。
【図１９】図１８に示す位置補正の手順を詳細に示す説明図である。
【図２０】図１８に示す別の位置補正の手順をより詳細に示す説明図である。
【図２１】実施の形態１および２による印影照合装置の性能評価のために使用する原画像の例を示す説明図である。
【図２２】図２１に示す原画像から作成された２値化画像の例を示す説明図である。
【図２３】図２２に示す２値化画像から作成された評価パターンの例を示す説明図である。
【図２４】図２３に示す評価パターンから作成された再評価パターンの例を示す説明図である。
【図２５】実施の形態１による印影照合装置に、朱肉による入力パターン１７１５個と偽印棄却率１００％となる閾値とを与えたときの、同装置の照合率を示すグラフである。
【図２６】実施の形態１による印影照合装置に、朱肉による入力パターン１７１５個と偽印棄却率１００％となる閾値とを与えたときの、同装置の照合率を示すグラフである。
【図２７】実施の形態１による印影照合装置に、朱肉および青・黒・赤のスタンプインクによる入力パターン７００７個と偽印棄却率１００％となる閾値とを与えたときの、同装置の照合率を示すグラフである。
【図２８】実施の形態２による印影照合装置に、朱肉および青・黒・赤のスタンプインクによる入力パターン７００７個と偽印棄却率１００％となる閾値とを与えたときの、同装置の照合率を示すグラフである。
【符号の説明】
１００ バスまたはケーブル
１０１ ＣＰＵ
１０２ ＲＯＭ
１０３ ＲＡＭ
１０４ ＨＤＤ
１０５ ＨＤ
１０６ ＦＤＤ
１０７ ＦＤ
１０８ ディスプレイ
１０９ Ｉ／Ｆ
１１０ 通信ケーブル
１１１ キーボード
１１２ マウス
１１３ スキャナ
１１４ プリンタ
１１５ ＣＤ−ＲＯＭ
１１６ ＣＤ−ＲＯＭドライブ
２００，１４００ 登録パターン記憶部
２０１，１４０１ 未知パターン入力部
２０２，１４０２ 前処理部
２０３，１４０３ 第１照合部
２０４，１４０４ 第２照合部
２０５ 第３照合部
２０６ 第４照合部
３００ ２値化部
３０１ 評価パターン作成部
９００，１００１ 圧縮部
９０１，１００２，１１０１，１２０１，１５０１，１７０１ 回転補正部
９０２，１００３，１１０２，１２０３，１５０３，１７０３ 平滑化部
９０３，１００４，１１０３，１２０４，１５０４，１７０４ 距離値算出部
９０４，１００５，１１０４，１２０５，１５０５，１７０５ 一致判定部
１０００，１１００，１２００，１５００，１７００ 再評価パターン作成部
１２０２，１５０２，１７０２ 位置補正部
１４０５ 第５照合部
１４０６ 第６照合部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an imprint collation apparatus that collates an imprint in an image with an imprint in another image, an imprint collation method, and a program that causes a computer to execute the method.
[0002]
[Prior art]
Researches such as fingerprint verification and writer verification have been conducted as technologies for automating and mechanizing personal identification. However, in Japan and the Eastern countries, identification of individuals through verification of seals or seals is still important. It has significance. At present, the verification of seals is performed by human eyes, and there is a strong demand for improvements in accuracy and speed through automation and mechanization, and reduction in the work burden associated with an increase in the number of registered seals.
[0003]
[Problems to be solved by the invention]
However, whether or not to match a certain imprint with a certain imprint is ultimately determined by the degree of coincidence between images in pixel units. If there is a problem such as not being imprinted (so-called “distortion”), missing or faint, and unclear, the correlation value of the pixel with the imprint of the verification source will decrease, resulting in a decrease in the verification rate. There was a problem of end.
[0004]
In order to solve the above-described problems caused by the prior art, the present invention enables high-precision collation with the imprint of the collation source even if the imprint has a problem with the stamped position, angle, or clarity. An object of the present invention is to provide an imprint verification device, an imprint verification method, and a program for causing a computer to execute the method.
[0005]
[Means for Solving the Problems]
In order to solve the above-described problems and achieve the object, an imprint checking apparatus according to the present invention includes an imprint region specifying unit that specifies an imprint region in which the imprint image exists for an image including the imprint image therein, and the imprint region. An imprint center calculating means for calculating the center position of the imprint area specified by the specifying means; and the imprint at a position such that the center position of the image coincides with the center position of the imprint area calculated by the imprint center calculating means. Center position correcting means for correcting the position of the image.
[0006]
According to this invention, the position of the imprint in the image is adjusted in advance so as to come to the center of the image.
[0007]
Further, in the imprint verification device according to the present invention, the imprint area specifying means has an area in each imprint area in a state where the imprint image is rotated by a different angle around an arbitrary position in the image. The largest one is specified as the imprint region of the image.
[0008]
According to the present invention, the imprint in the image is adjusted in advance so that the position is at the center of the image even if there is a chipping or fading.
[0009]
In addition, the imprint checking apparatus according to the present invention includes a rotation correction unit that rotates the imprint image in the first image including the imprint image therein around the center position, and the imprint image is rotated by the rotation correction unit. For each of the first image and the second image containing the imprint image therein, an imprint area specifying means for specifying an imprint area in which the imprint image exists, and the imprint area specifying means specifying the imprint area specifying means Edge position correcting means for correcting the position of the imprint image in the first image at a position where one side of the imprint area of the first image and a corresponding side of the imprint area of the second image coincide with each other; Distance value calculating means for calculating a distance value between the first image whose position of the imprinted image is corrected by the side position correcting means and the second image, and calculated by the distance value calculating means. It was based on the distance value, characterized in that and a determining coincidence determining means for determining whether or not the seal impression and the image matches in imprint image and the second image within the first image.
[0010]
According to the present invention, after the relative angular deviation is corrected, the imprints in each image are collated after the relative positional deviation is corrected at four locations.
[0011]
Further, in the imprint checking device according to the present invention, one apex of the imprint area of the first image specified by the imprint area specifying means and one corresponding apex of the imprint area of the second image coincide with each other. Angular position correction means for correcting the position of the imprint image in the first image at such a position, and the distance value calculation means uses the side position correction means or the angular position correction means to determine the position of the imprint image. A distance value between the first image in which the correction is corrected and the second image is calculated.
[0012]
According to the present invention, after the relative angular deviation is corrected, the imprints in each image are collated after the relative positional deviation is corrected at eight locations.
[0013]
In addition, the imprint collation apparatus according to the present invention includes an imprint area specifying unit that specifies an imprint area in which the imprint image exists for each of the first image and the second image including the imprint image therein, and the imprint area specifying The position of the imprint image in the first image is corrected to a position where one side of the imprint area of the first image specified by the means coincides with one side corresponding to the imprint area of the second image. Side position correction means, rotation correction means for rotating the imprint image in the first image whose position has been corrected by the side position correction means, with the center position as a center, and the imprint image by the rotation correction means. Distance value calculating means for calculating a distance value between the rotated first image and the second image, and the first image in the first image based on the distance value calculated by the distance value calculating means. A coincidence determining means for determining whether or not the imprint image in imprint image and the second image are matched, and further comprising a.
[0014]
According to the present invention, the imprints in each image are collated after correcting their relative positional deviations at four locations and further correcting their relative angular deviations.
[0015]
Further, in the imprint checking device according to the present invention, one apex of the imprint area of the first image specified by the imprint area specifying means and one corresponding apex of the imprint area of the second image coincide with each other. An angular position correction unit that corrects the position of the imprint image in the first image at such a position, and the rotation correction unit is corrected by the side position correction unit or the angular position correction unit. The imprint image in the first image is rotated.
[0016]
According to the present invention, the imprints in each image are collated after correcting their relative positional deviations at eight locations and further correcting their relative angular deviations.
[0017]
In addition, the imprint collation apparatus according to the present invention further includes an imprint area where the imprint image exists for each of the first image and the second image obtained by rotating the imprint image by the rotation correction unit. The second imprint area specifying means to be specified matches one side of the imprint area of the first image specified by the second imprint area specifying means with the corresponding one side of the imprint area of the second image. And a second side position correction unit that corrects the position of the imprint image in the first image at such a position, and the distance value calculation unit uses the second side position correction unit to correct the imprint image. A distance value between the first image whose position is corrected and the second image is calculated.
[0018]
According to the present invention, after the relative misregistration of the imprint in each image is corrected at four locations, the relative angular misalignment is further corrected, and the relative misregistration is again corrected at four locations. It is verified after being corrected by.
[0019]
The imprint collation apparatus according to the present invention further includes a vertex corresponding to one vertex of the imprint area of the first image specified by the second imprint area specifying means and the imprint area of the second image. And a second angular position correction unit that corrects the position of the imprint image in the first image at a position such that the distance value calculation unit matches the second side position correction unit or the second side position correction unit. A distance value between the first image whose position of the imprint image is corrected by the second angular position correcting unit and the second image is calculated.
[0020]
According to the present invention, after the relative misregistration of each seal in each image is corrected at four or eight locations, the relative angular misalignment is further corrected, and the relative misregistration is again performed. Are corrected after being corrected at 8 locations.
[0021]
In addition, the imprint collation method according to the present invention includes an imprint region specifying step for specifying an imprint region in which the imprint image exists for an image including the imprint image therein, and a center position of the imprint region specified in the imprint region specifying step. And a center position correction step of correcting the position of the imprint image to a position such that the center position of the image coincides with the center position of the imprint region calculated in the imprint center calculation step. , Including.
[0022]
According to this invention, the position of the imprint in the image is adjusted in advance so as to come to the center of the image.
[0023]
The imprint checking method according to the present invention includes a rotation correction step of rotating the imprint image in the first image including the imprint image therein around the center position, and the imprint image is rotated in the rotation correction step. For each of the first image and the second image that includes the imprint image therein, the imprint region specifying step that specifies the imprint region where the imprint image exists, and the imprint region specifying step that are specified in the imprint region specifying step A side position correcting step of correcting the position of the imprint image in the first image at a position such that one side of the imprint region of the first image and a corresponding side of the imprint region of the second image coincide with each other; A distance value calculating step for calculating a distance value between the first image whose position of the imprint image has been corrected in the side position correcting step and the second image, and a distance calculated in the distance value calculating step Before based on value A match determining step determines whether the imprint image of the first imprint image and said second image in the image coincides, characterized in that it contains.
[0024]
According to the present invention, after the relative angular deviation is corrected, the imprints in each image are collated after the relative positional deviation is corrected at four locations.
[0025]
In addition, an imprint checking method according to the present invention includes an imprint area specifying step for specifying an imprint area in which the imprint image exists for each of the first image and the second image including the imprint image, and the imprint area specifying The position of the imprinted image in the first image is corrected to a position where one side of the imprinted area of the first image specified in the step and a corresponding side of the imprinted area of the second image coincide with each other. A side position correction step, a rotation correction step for rotating the imprint image in the first image whose position has been corrected in the side position correction step, around the center position, and the imprint image in the rotation correction step. A distance value calculating step for calculating a distance value between the rotated first image and the second image, and an imprint in the first image based on the distance value calculated in the distance value calculating step. Image and second image And determining matching determination step whether imprint and image matches the inner, characterized by including.
[0026]
According to the present invention, the imprints in each image are collated after correcting their relative positional deviations at four locations and further correcting their relative angular deviations.
[0027]
Moreover, according to the program concerning this invention, it becomes possible to make a computer perform the method mentioned above.
[0028]
DETAILED DESCRIPTION OF THE INVENTION
Exemplary embodiments of an imprint collation apparatus, an imprint collation method, and a program for causing a computer to execute the method according to the present invention will be described below in detail with reference to the accompanying drawings.
[0029]
(Imprint checking device according to Embodiment 1)
First, a hardware configuration of the seal collation apparatus according to the embodiment of the present invention will be described. FIG. 1 is an explanatory diagram illustrating a hardware configuration of the seal collation apparatus according to the first embodiment. In the figure, 101 indicates a CPU that controls the entire apparatus, 102 indicates a ROM that stores basic input / output programs, and 103 indicates a RAM that is used as a work area of the CPU 101.
[0030]
Reference numeral 104 denotes an HDD (hard disk drive) that controls reading / writing of data with respect to the HD (hard disk) 105 according to the control of the CPU 101, and 105 denotes an HD that stores data written according to the control of the HDD 104. Yes. Reference numeral 106 denotes an FDD (floppy disk drive) that controls reading / writing of data with respect to the FD (floppy disk) 107 under the control of the CPU 101, and 107 denotes a removable FD that stores data written according to the control of the FDD 106. Respectively.
[0031]
Reference numeral 108 denotes a cursor, menu, window, or display that displays various data such as characters and images. Reference numeral 109 denotes a network board that is connected to a network NET via a communication cable 110 and functions as an interface between the network and the CPU 101. Respectively. Reference numeral 111 denotes a keyboard having a plurality of keys for inputting characters, numerical values, and various instructions, and 112 denotes a mouse for selecting and executing various instructions, selecting a processing target, moving a cursor, and the like. ing.
[0032]
Reference numeral 113 denotes a scanner that optically reads characters and images, 114 denotes a printer that prints characters and images according to the control of the CPU 101, 115 denotes a CD-ROM that is a removable recording medium, and 116 denotes a CD-ROM 115. Reference numeral 100 denotes a CD-ROM drive for controlling the reading of data with respect to the above.
[0033]
Next, a functional configuration of the seal collation apparatus according to the embodiment of the present invention will be described. FIG. 2 is an explanatory diagram functionally showing the configuration of the seal collation apparatus according to the first embodiment. The imprint matching apparatus according to the first embodiment includes a registered pattern storage unit 200, an unknown pattern input unit 201, a preprocessing unit 202, a first matching unit 203, a second matching unit 204, a third matching unit 205, and a fourth matching unit 206. It is the structure containing.
[0034]
The registered pattern storage unit 200 is a functional unit that retains as many stamp images as the number of registered seals that are collected in advance for each seal. It is assumed that these images in the registered pattern storage unit 200, that is, registered patterns, have been subjected to preprocessing such as binarization and center position correction, which will be described later.
[0035]
The unknown pattern input unit 201 is a functional unit that samples and quantizes an imprint imprinted on a paper surface and captures it as 256 × 256 pixels and 256 gradation image data. The first collation unit 203 to the fourth collation unit 206 to be described later sequentially collate the input patterns input from the unknown pattern input unit 201 through the preprocessing unit 202 to be described later with the respective registered patterns in the registration pattern storage unit 200. I will do it.
[0036]
The preprocessing unit 202 is a functional unit that performs preprocessing, which will be described below, in order to improve the accuracy of collation in the first collation unit 203 to the fourth collation unit 206 in the subsequent stage for the input pattern from the unknown pattern input unit 201. is there. FIG. 3 is an explanatory diagram showing the functional configuration of the preprocessing unit 202 according to the first embodiment in more detail. The preprocessing unit 202 according to the first embodiment includes a binarization unit 300 and an evaluation pattern creation unit 301.
[0037]
The binarization unit 300 is a functional unit that binarizes the multi-valued image input from the unknown pattern input unit 201, and includes a histogram calculation unit 300a and a binarization threshold value determination unit 300b. The binarization unit 300 first calculates a histogram of pixel values of the input pattern by the histogram calculation unit 300a. Next, the binarization threshold value determination unit 300b determines a binarization threshold value from a maximum value and a minimum value of the histogram by a method called “Otsu's differential binarization method”. Then, before and after this threshold value, each pixel value in the input pattern is binarized.
[0038]
The evaluation pattern creation unit 301 is a functional unit that creates an evaluation pattern that is used for collation in a first collation unit 203 to a fourth collation unit 206, which will be described later, from an input pattern after binarization by the binarization unit 300. Yes, it includes a seal imprint area specifying unit 301a and a center position correcting unit 301b.
[0039]
This evaluation pattern is obtained by moving an imprint image (hereinafter simply referred to as “imprint”) in the input pattern after binarization so that the center position thereof coincides with the center position of the entire image. For example, the imprint shown in FIG. 4 is slightly offset from the whole image, but the position of the imprint may deviate from the center of the image in this way. The evaluation pattern is used as an evaluation pattern, and this evaluation pattern is used for collation described later.
[0040]
The evaluation pattern creation unit 301 first calculates a histogram of pixel values in the x-axis direction and a histogram of pixel values in the y-axis direction of the input pattern after binarization by the imprint region specifying unit 301a. Next, the histogram value range is normalized from 0 to 1, and the image edge portion below a certain threshold value (for example, 0.06) is cut off, and the remaining area is cut out as an imprint area, that is, an imprint area. . In this calculation method, the imprint area is a rectangle circumscribing the imprint.
[0041]
However, when the imprint area is simply cut out based on the histogram in this way, there is a problem that the imprint area is likely to be cut out erroneously with respect to an imprint having a chipped or blurred image. FIG. 5 is an explanatory diagram showing a state in which an imprint area is not accurately obtained when the imprint is missing or blurred.
[0042]
In the normalized histogram 500 in the x-axis direction, the left end and the right end where the accumulated pixel value is less than the threshold value are discarded, and the range indicated by 501 is extracted as the existence range of the imprint on the x-axis. Similarly, in the histogram 502 in the y-axis direction, the range indicated by 503 is cut out as the existence range of the imprint on the y-axis. Has been truncated. For this reason, the actual imprint is not contained in the imprint area 504 specified as described above, and a portion with a chipped or blurred portion protrudes.
[0043]
Therefore, the imprint is rotated to cut out the imprint area at a plurality of rotation angles, and an imprint area where the area is the maximum, that is, an imprint area that is not reduced due to missing or fading of the imprint is obtained. 6 and 7 are explanatory diagrams showing a state in which the seal is rotated in order to accurately obtain the seal region even when the seal is missing or faint.
[0044]
First, FIG. 6 shows a state in which the imprint shown in FIG. 5 is rotated by 30 degrees. In this case as well, the imprint area 600 specified from the histogram is smaller than the original imprint area as in FIG. Yes. On the other hand, FIG. 7 shows a state in which the imprint shown in FIG. 5 is rotated by 60 degrees, and the imprint area 700 specified from the histogram coincides with the original imprint area and includes the entire imprint in the area. It is like that. The imprint area specifying unit 301a specifies the imprint areas 504, 600, and 700 for each rotation angle of 0 degrees, 30 degrees, and 60 degrees, and among these, 700 having the largest area is set as the imprint area. Thereby, it is possible to avoid an imprint error in the imprint region caused by missing or faint imprints.
[0045]
Note that the rotation center of the imprint may be selected as long as the rotated imprint does not protrude outside the image, but here, the center of the entire image including the imprint, that is, the point 505 shown in FIG. Is the center of rotation.
[0046]
When the seal imprint area is obtained as described above, the evaluation pattern creating unit 301 then calculates the center position (hereinafter simply referred to as “center”) of the area by the center position correcting unit 301b. The center of the imprint area is a point where the distance from the left side and the right side of the area, the distance from the upper side, and the distance from the lower side are equal to each other. In the imprint area 700 shown in FIG. Is a point (center) 701. Note that the center of the imprint area 700 can be simply referred to as the center of the imprint.
[0047]
Then, the center position correction unit 301b determines that the center 701 of the imprinted region is within the imprinted region 700 so that the center of the entire image, that is, the intersection (center) 704 between the vertical centerline 702 and the horizontal centerline 703 of the image coincides. The image, that is, the seal impression is moved. In the case of FIG. 7, the entire imprint is moved diagonally left upward so that the center 701 of the imprint area coincides with the center 704 of the image. After the movement, the center of the imprint area 701, that is, the center 704 of the image is the center, the angle of the imprint is rotated when the imprint area is specified, in the example of FIG. Keep it. The input pattern after the above processing is the evaluation pattern here.
[0048]
FIG. 8 is an explanatory diagram showing an example of an evaluation pattern created from the input pattern shown in FIG. In FIG. 4 before correcting the center position, the imprint was slightly below the entire image, but in FIG. 8 after correction, the imprint is just in the center of the image.
[0049]
Returning to FIG. 2, the first collation unit 203, the second collation unit 204, the third collation unit 205, and the fourth collation unit 206 are respectively registered patterns and evaluation patterns, or re-evaluation patterns to be described later, by the methods described below. It is a functional part which performs collation with i.e., determination of whether or not there is a match between both.
[0050]
(Method 1)
This is a verification method in the first verification unit 203. The registered pattern and the above-described evaluation pattern are compressed. Then, the imprint in the evaluation pattern is rotated in units of m degrees to obtain a correlation value with the registered pattern at each angle, and a distance value with respect to the registered pattern at an angle at which this correlation value is maximum is obtained. Whether or not there is a match with the registered pattern is determined by comparing the value with a predetermined threshold value. Also, only when it is determined that there is a mismatch, the seal in the evaluation pattern is further rotated 180 degrees from the angle at which the correlation value is maximum, and the calculation of the distance value at this angle and the determination of the presence or absence of the match are performed again.
[0051]
(Method 2)
This is a collation method in the second collation unit 204. From the above-described evaluation pattern, a re-evaluation pattern in which the blur is further corrected is created. Next, the registered pattern and the reevaluation pattern are compressed. Then, the imprint in the re-evaluation pattern is rotated in increments of m degrees to obtain a correlation value with the registered pattern at each angle, and a distance value with the registered pattern at an angle at which this correlation value is maximum is obtained. Whether or not there is a match with the registered pattern is determined by comparing the distance value with a predetermined threshold value. Also, only when it is determined that there is a mismatch, the imprint in the re-evaluation pattern is further rotated 180 degrees from the angle at which the correlation value is maximized, and the calculation of the distance value at this angle and the determination of whether or not there is a match are performed again. . The only difference from Method 1 is that a re-evaluation pattern is used instead of an evaluation pattern for matching with a registered pattern.
[0052]
(Method 3)
This is a verification method in the third verification unit 205. Rotate the imprint in the re-evaluation pattern in units of m degrees to obtain a correlation value with the registered pattern at each angle, obtain a distance value with the registered pattern at an angle at which this correlation value is maximum, Whether or not there is a match with the registered pattern is determined by comparing the distance value with a predetermined threshold value. The difference from the method 2 is that the registration pattern and the re-evaluation pattern are not compressed, and instead, the re-determination after a half rotation is not performed even if it is determined to be inconsistent.
[0053]
(Method 4)
This is a collation method in the fourth collation unit 206. Rotate the imprint in the above-mentioned reevaluation pattern in units of m degrees to obtain a correlation value with the registered pattern at each angle, and further move the imprint in the reevaluation pattern at an angle at which this correlation value is maximum. Then, alignment of each side of the seal impression area is performed. Thereafter, a distance value between the two patterns is obtained, and whether or not they match is determined by comparison with a predetermined threshold value. The only difference from Method 3 is that the distance value is calculated after correcting the relative displacement of the imprint position at the angle where the correlation value is maximum.
[0054]
Hereinafter, functions of the first verification unit 203 to the fourth verification unit 206 will be sequentially described. First, FIG. 9 is an explanatory diagram showing the functional configuration of the first verification unit 203 according to the first embodiment in more detail. The first collation unit 203 according to the first embodiment includes a compression unit 900, a rotation correction unit 901, a smoothing unit 902, a distance value calculation unit 903, and a coincidence determination unit 904.
[0055]
The compression unit 900 is a functional unit that compresses a registered pattern and an evaluation pattern to be collated from each other, that is, from 256 × 256 pixels to 64 × 64 pixels. Specifically, the image before compression is divided into 4096 (= 64 × 64) blocks composed of 4 pixels × 4 pixels, and the average of the pixel values of 16 pixels in each block is taken. Each pixel value of the image after compression is used.
[0056]
The rotation correction unit 901 is a functional unit that corrects a relative angular deviation of the imprint in the evaluation pattern with respect to the imprint in the registered pattern, and includes an imprint rotation unit 901a, a correlation value calculation unit 901b, and a rotation angle calculation unit 901c. It is a configuration.
[0057]
The rotation correcting unit 901 first rotates the imprint in the evaluation pattern by a predetermined angle unit, for example, in increments of 3 degrees, by the imprint rotating unit 901a. The center of rotation is the center of the imprint that matches the center of the image by the above-described preprocessing. Next, the rotation correction unit 901 uses the correlation value calculation unit 901b to calculate the correlation value C (k) between the evaluation pattern and the registered pattern at each angle k of 0 degrees, 3 degrees, 6 degrees,. calculate. Then, the rotation angle calculation unit 901c specifies the angle K when C (k) is maximized, that is, the relative rotation angle K of the imprint in the evaluation pattern with respect to the imprint in the registered pattern. Finally, the rotation correction unit 901 rotates the seal in the original evaluation pattern by the angle K calculated by the rotation angle calculation unit 901c with the center as the center.
[0058]
The correlation value calculation unit 901b may calculate any correlation value as long as the degree of correlation between the two patterns can be calculated. In the first embodiment, the following calculation is used. According to the formula.
[Expression 1]

[0059]
Next, the smoothing unit 902 is a functional unit that applies a smoothing filter to each of the registered pattern and the evaluation pattern after the rotation correction by the rotation correction unit 901. By applying a smoothing filter, it is possible to absorb a minute shift in imprint on a pixel basis.
[0060]
Next, the distance value calculation unit 903 is a functional unit that calculates the distance value D between the registered pattern and the evaluation pattern that have been smoothed by the smoothing unit 902. Any calculation formula may be used to calculate the distance value, but in the first embodiment, the following calculation formula is used.
[Expression 2]

[0061]
In addition, you may make it take the sum total of the absolute value instead of the square of the difference of a pixel value. Also, in both the registered pattern and the evaluation pattern, the pixel value is either 1 (with pixels, ie, black) or 0 (without pixels, ie, white), and the pixels at the same position are both black or white. The pixel value difference is always 0, and when one is black and the other is white, the pixel value difference is always 1. Therefore, instead of taking the square of this difference or the sum of absolute values, You may make it take the total of the pixel from which a value differs between patterns. Instead of the sum of the squares of the pixel value differences, grasp the distance value by the sum of the absolute values of the pixel value differences or the total number of pixels having different pixel values, thereby reducing the amount of calculation and improving the processing speed. Can be made.
[0062]
Next, the coincidence determination unit 904 is a functional unit that determines whether or not there is a match between the registered pattern and the evaluation pattern based on a magnitude comparison between the distance value calculated by the distance value calculation unit 903 and a threshold value stored in advance. . Specifically, when the distance value is below the threshold value, it is determined that the two values match. Conversely, when the distance value is equal to or greater than the threshold value, the two values are determined not to match.
[0063]
If the match determination unit 904 determines that the evaluation pattern does not match the registered pattern, the rotation correction unit 901 further rotates the imprint after rotation correction in the evaluation pattern by 180 degrees. Therefore, the rotation angle of the impression from the original evaluation pattern is K + 180 degrees.
[0064]
When half-rotated in this way, humans cannot recognize it as such, but mathematically similar patterns appear. Therefore, the collation rate can be reliably improved by performing collation again. As described above, since the first collation unit 203 collates the compressed images, the collation rate is lowered instead of reducing the processing amount. Thus, by including re-collation after half rotation in this way, an attempt is made to compensate for the decrease in the collation rate due to compression. Note that the processing amount as a whole can be significantly reduced by performing re-collation after compression and collation, rather than performing collation without compression.
[0065]
Only when it is determined that the evaluation pattern does not match the registered pattern even after re-collation, the first collation unit 203 sends the evaluation pattern input from the preprocessing unit 202 to the next collation unit, that is, the second collation unit 204. Output.
[0066]
Returning to the description of the remaining functional units in FIG. 2, FIG. 10 is an explanatory diagram showing the functional configuration of the second verification unit 204 according to the first embodiment in more detail. This is obtained by adding a re-evaluation pattern creation unit 1000 to the functional configuration of the first verification unit 203 shown in FIG.
[0067]
The reevaluation pattern creation unit 1000 is a functional unit that creates the reevaluation pattern from the input evaluation pattern. The re-evaluation pattern is a pattern that is more clear by emphasizing the outline of the imprint in the evaluation pattern. For example, since a faint impression is seen in the evaluation pattern shown in FIG. 8, an edge emphasis filter is applied to correct the fading as a re-evaluation pattern.
[0068]
As the filter, a sobel filter known to be able to perform good edge extraction in spite of a small amount of processing is used in the first embodiment. However, any other filter can be used as long as it can enhance the contour of an image. A filter may be used. After filtering, an image obtained by superimposing an image before edge enhancement and an image after edge enhancement is used as a reevaluation pattern.
[0069]
Then, the compression unit 1001, the rotation correction unit 1002, the smoothing unit 1003, the distance value calculation unit 1004, and the coincidence determination unit 1005 are the same as the corresponding units in FIG. 9 for this re-evaluation pattern instead of the evaluation pattern. Perform processing. When the reevaluation pattern does not match the registered pattern, the second collation unit 204 outputs the reevaluation pattern created by the reevaluation pattern creation unit 1000 to the next collation unit, that is, the third collation unit 205.
[0070]
Returning to the description of the remaining functional units in FIG. 2, FIG. 11 is an explanatory diagram showing the functional configuration of the third verification unit 205 according to the first embodiment in more detail. This is a configuration in which the compression unit 1001 is deleted from the functional configuration of the second verification unit 204 shown in FIG.
[0071]
However, the rotation correction unit 1101 and the distance value calculation unit 1103 of the third verification unit 205 are the rotation correction unit 901 and the distance value calculation unit 903 of the first verification unit 203, or the rotation correction unit 1002 and the distance of the second verification unit 204. Unlike the value calculation unit 1004, when the evaluation pattern or the re-evaluation pattern does not match the registered pattern, the value calculation unit 1004 does not have a function of re-collating the imprint in the pattern by half rotation. This is because the third collation unit 205 does not perform compression as in the first collation unit 203 and the second collation unit 204, and therefore it is not necessary to compensate for the decrease in the collation rate corresponding to the compression.
[0072]
In this third matching unit 205, the matching with the registered pattern is performed not on the evaluation pattern but on the reevaluation pattern. However, since the reevaluation pattern created by the second matching unit 204 can be used, the reevaluation pattern is actually created. The re-evaluation pattern is not recreated by the unit 1100. The function of each part other than the above is the same as the function of each part in the corresponding FIG. 9 or FIG. When the reevaluation pattern does not match the registered pattern, the third matching unit 205 outputs the reevaluation pattern input from the second matching unit 204 to the next matching unit, that is, the fourth matching unit 206.
[0073]
Returning to the description of the remaining functional units in FIG. 2, FIG. 12 is an explanatory diagram showing the functional configuration of the fourth collating unit 206 according to the first embodiment in more detail. This is a configuration in which a position correction unit 1202 is added to the functional configuration of the third matching unit 205 shown in FIG.
[0074]
The position correction unit 1202 is a functional unit that corrects a relative positional shift of the imprint in the re-evaluation pattern with respect to the imprint in the registered pattern. The imprint region specifying unit 1202a, the side position correction unit 1202b, and the correlation value calculation unit 1202c. It is the structure containing.
[0075]
First, the position correction unit 1202 uses the imprint region specifying unit 1202a to specify a region in which each imprint exists, that is, an imprint region for the registered pattern and the re-evaluation pattern after the rotation correction by the rotation correction unit 1201. This imprint area is specified based on a histogram of pixel values, similar to the imprint area specifying section 301a of the preprocessing section 202 shown in FIG.
[0076]
Next, the position correction unit 1202 uses the side position correction unit 1202b to place the imprint pattern in the re-evaluation pattern at a position where any one of the imprint areas in the re-evaluation pattern overlaps with the corresponding one of the imprint areas in the registration pattern. Move the imprint of. The corresponding sides are the upper side for the upper side, the lower side for the lower side, the left side for the left side, and the right side for the right side.
[0077]
FIG. 13 is an explanatory diagram illustrating a state of the imprint movement performed by the side position correcting unit 1202b according to the first embodiment. FIG. 5A shows the imprint in the re-evaluation pattern moved so that the upper side of the imprint area in the registered pattern matches the upper side of the imprint area in the re-evaluation pattern. Similarly, FIG. 4B shows the case where the lower sides are matched, FIG. 5C shows the case where the left sides match, and FIG. 6D shows the case where the right sides match. In this way, the side position correction unit 1202b superimposes each of the four sides of the imprint region to create four correction results.
[0078]
Next, the position correction unit 1202 calculates the correlation value between the corrected re-evaluation pattern and the registered pattern by the correlation value calculation unit 1202c, and selects the one having the maximum correlation value. Then, the smoothing in the smoothing unit 1203, the distance value calculation in the distance value calculation unit 1204, and the coincidence determination in the coincidence determination unit 1205 are performed for the selected re-evaluation pattern.
[0079]
The correlation value calculation unit 1202c may obtain the correlation value in any way, but in the first embodiment, among the pixels at the same position, the total number of pixels having different values between the registered pattern and the reevaluation pattern, That is, the total number of pixels that are white in the re-evaluation pattern but black in the re-evaluation pattern, or vice versa, is the correlation value of both patterns. It is said.
[0080]
This is also a kind of correlation value in that it indicates no correlation between patterns, but unlike a general correlation value, the smaller the value, the higher the correlation between the two. Therefore, the position correction unit 1202 actually selects one re-evaluation pattern that minimizes the correlation value calculated by the correlation value calculation unit 1202c instead of maximizing the correlation value.
[0081]
The fourth matching unit 206 also matches the registered pattern with the re-evaluation pattern instead of the evaluation pattern. However, since the re-evaluation pattern created by the second matching unit 204 can be used, the re-evaluation pattern is actually created. The re-evaluation pattern is not recreated by the unit 1200. The function of each part other than the above is the same as the function of each part in FIG. 9, FIG. 10, or FIG.
[0082]
The unknown pattern input unit 201, the preprocessing unit 202, the first collation unit 203, the second collation unit 204, the third collation unit 205, and the fourth collation unit 206 are each an HD 105, an FD 107, or a CD-ROM 115. The functions of the respective units are realized by the CPU 101 and the like executing instruction processing in accordance with the instructions of the program read from the recording medium to the RAM 103.
[0083]
(Imprint checking device according to Embodiment 2)
In the seal collation apparatus according to Embodiment 1 described above, after the fourth collation unit 206 corrects the angular deviation between the seal in the registered pattern and the seal in the reevaluation pattern, the top, bottom, and left sides of the seal area are further corrected.・ The position of the seal was corrected by superimposing the right side. At this time, in addition to the four sides, the four vertices of the upper left corner, lower left corner, upper right corner, and lower right corner were also superimposed. May be. Further, the positional deviation may be corrected not only after the angular deviation correction but also before the correction.
[0084]
The hardware configuration of the seal collation apparatus according to the second embodiment is the same as that of the seal collation apparatus according to the first embodiment shown in FIG. FIG. 14 is an explanatory diagram functionally showing the configuration of the seal collation apparatus according to the second embodiment. This is obtained by replacing the third collation unit 205 and the fourth collation unit 206 with a fifth collation unit 1405 and a sixth collation unit 1406, respectively, in the functional configuration of the stamp collation apparatus according to the first embodiment shown in FIG. . The fifth collating unit 1405 and the sixth collating unit 1406 are functional units that perform collation between the registered pattern and the reevaluation pattern by the following methods.
[0085]
(Method 5)
This is a collation method in the fifth collation unit 1405. Rotate the imprint in the above-mentioned reevaluation pattern in units of m degrees to obtain a correlation value with the registered pattern at each angle, and further move the imprint in the reevaluation pattern at an angle at which this correlation value is maximum. Then, each side and each vertex of the seal imprint area are aligned. Thereafter, a distance value between the two patterns is obtained, and whether or not they match is determined by comparison with a predetermined threshold value. The difference from the method 4 is that not only the edges but also the vertices are aligned, and the alignment may not be performed.
[0086]
(Method 6)
This is a collation method in the sixth collation unit 1406. First, each side and each vertex are aligned with the imprint region in the registered pattern by moving the imprint in the re-evaluation pattern. Next, the imprint in the re-evaluation pattern is rotated in units of m degrees to obtain a correlation value with the registered pattern at each angle, and the same alignment as described above is performed again at the angle at which the correlation value is maximized. Thereafter, a distance value between the two patterns is obtained, and whether or not they match is determined by comparison with a predetermined threshold value. The only difference from the method 5 is that the same alignment as the method 5 is performed before and after the rotation of the seal impression.
[0087]
FIG. 15 is an explanatory diagram showing the functional configuration of the fifth collation unit 1405 according to Embodiment 2 in more detail. This is a configuration in which an angular position correction unit 1502c is added to the position correction unit 1202 of the fourth matching unit 206 according to Embodiment 1 shown in FIG. FIG. 16 is an explanatory diagram illustrating a state of imprint movement performed by the angular position correction unit 1502c according to the second embodiment. In the figure, both (a) and (b) are examples in which the vertices at the upper left corner of the imprint region coincide with each other. Similarly, the corner position correction unit 1502c also superimposes the vertices of the lower left corner, the upper right corner, and the lower right corner.
[0088]
Then, the correlation value calculation unit 1502d calculates a correlation value with the registered pattern for each of the eight re-evaluation patterns after correction by the side position correction unit 1502b or the corner position correction unit 1502c. In addition, a correlation value between the re-evaluation pattern before correction and the registered pattern is also calculated. Note that, as described above, in both the registered pattern and the re-evaluation pattern, the center of the imprint is adjusted in advance so as to come to the center of the image. It can also be seen as a superposition.
[0089]
Thereafter, the one having the maximum correlation value with the registered pattern is selected from the nine re-evaluation patterns, the smoothing by the smoothing unit 1503 is performed on the selected re-evaluation pattern, and the distance by the distance value calculating unit 1504 The value calculation and coincidence determination by the coincidence determination unit 1505 are performed. If this reevaluation pattern does not match the registered pattern, the fifth collation unit 1405 outputs the reevaluation pattern input from the second collation unit 1404 to the next collation unit, that is, the sixth collation unit 1406.
[0090]
Next, FIG. 17 is an explanatory diagram showing the functional configuration of the sixth collating unit 1406 according to the second embodiment in more detail. This is the same as the functional configuration of the fifth collation unit 1405 shown in FIG. 15, and the position correction unit 1702 performs the same position correction not only after the rotation correction by the rotation correction unit 1701 but also before that. It has a function.
[0091]
That is, correction is performed so that the re-evaluation pattern in which the center of the imprint area matches that of the registered pattern, and any side or vertex of the imprint area matches that of the registered pattern, which is input from the fifth matching unit 1405. For the subsequent eight re-evaluation patterns, the correlation with the registered pattern is examined, and the one with the highest correlation is selected.
[0092]
Then, in this selected re-evaluation pattern, the impression is rotated to obtain the angle where the correlation with the registered pattern is the highest, and the re-evaluation pattern in which the imprint is just rotated to this angle, and each side of the imprint region Alternatively, the correlation between the eight re-evaluation patterns corrected so that the respective vertices coincide with the registered pattern is examined. Then, the distance value with the registered pattern is calculated for the one having the highest correlation and the presence / absence of matching is determined.
[0093]
As described above, if the position correction is performed before the rotation correction, if the seal stamp is a perfect circle, there is a high possibility that if the seal stamp is taken from the same seal stamp, the outer peripheral circles are matched. Before the rotation correction, it may be said that rough matching is performed with attention paid only to the shape of the imprint. By aligning the position of the outer circumference circle in advance, the rotation angle and distance value calculated after that become more accurate, and therefore what should be consistent with the imprint in the registered pattern is due to the deviation of the imprint position. It is possible to reduce the probability of erroneously determining a mismatch.
[0094]
The unknown pattern input unit 1401, the preprocessing unit 1402, the first collation unit 1403, the second collation unit 1404, the fifth collation unit 1405, and the sixth collation unit 1406 are each of various types such as HD 105, FD 107, or CD-ROM 115. The functions of the respective units are realized by the CPU 101 and the like executing instruction processing in accordance with the instructions of the program read from the recording medium to the RAM 103.
[0095]
FIG. 18 is an explanatory diagram showing the procedure of methods 1 to 6 used in the first and second embodiments in comparison. Method 1 starts from compression of the registered pattern / evaluation pattern by the compression unit 900 in step S1802, and step S1804 omits step S1803. In step S1804, rotation of the imprint in increments of 3 by the imprint rotation unit 901a and the correlation value calculation unit 901b. And the correlation value at each angle is calculated.
[0096]
In the following step S1805, the rotation angle calculation unit 901c specifies (calculates) K having the maximum correlation value as the rotation angle, and in step S1806, the rotation correction unit 901 performs rotation correction only for this angle K on the evaluation pattern. Steps S1807 and S1808 are omitted. Next, in step S1809, both patterns are smoothed by the smoothing unit 902. In the subsequent step S1810, the distance value D is calculated by the distance value calculating unit 903. Further, in step S1811, the coincidence determination is performed. The unit 904 compares the distance value with the threshold value, that is, determines whether or not there is a match between both patterns.
[0097]
Only when it is determined that there is a mismatch in step S1811, in step S1812, the rotation correction unit 901 further rotates the imprint in the evaluation pattern by a half rotation, and in step S1813, the distance value calculation unit 903 calculates the distance value D. In recalculation, in step S1814, the coincidence determination unit 904 re-determines whether there is a match.
[0098]
The method 2 starts from the creation of the re-evaluation pattern by the re-evaluation pattern creation unit 1000 in step S1801, that is, the edge enhancement for the evaluation pattern. Thereafter, the processing after step S1802 is the same as the method 1 described above.
[0099]
Similarly, the method 3 starts from step S1801, omits steps S1802 and S1803, and performs the processing from step S1804 on the uncompressed registered pattern / re-evaluation pattern in the same manner as the above methods 1 and 2. However, the processing in steps S1812 to S1814 when it is determined that there is a mismatch in step S1811 is omitted.
[0100]
The methods 4 and 5 are the same as the method 3 except that step S1807 or step S1808, that is, position correction by the position correction unit 1202 or 1502 is performed between step S1806 and step S1809. However, this misregistration correction is performed for each side of the seal impression region in the method 4 and for each side and each vertex in the method 5.
[0101]
The method 6 is the same as the method 5 except that step S1803, that is, position correction by the position correction unit 1702 is performed between step S1801 and step S1804. Note that the processing in step S1803 is the same as the processing in step S1808, and in short, the same position correction is repeated before and after the rotation correction in step S1806.
[0102]
Next, FIG. 19 is an explanatory diagram showing in detail the procedure of position correction in step S1807. In step S1901, the fourth collation unit 206 uses the imprint region specifying unit 1202a to specify the imprint region for each of the registered pattern and the reevaluation pattern after the rotation correction by the rotation correction unit 1201.
[0103]
In step S1902, the side position correction unit 1202b first moves the imprint in the re-evaluation pattern so that the upper sides of the imprint areas of both patterns match. Next, in steps S1903 to S1905, the imprint movement is performed in such a manner that the lower sides, the left sides, and the right sides of the imprint region coincide with each other. Note that the corrections in steps S1902 to S1905 are performed independently for the re-evaluation patterns input from the third verification unit 205, and therefore, the order may be different from the order shown.
[0104]
In step S1906, the correlation value calculation unit 1202c calculates a correlation value between the re-evaluation pattern after movement in steps S1902 to S1905 and the registered pattern. In step S1907, the position correction unit 1202 selects one of the four reevaluation patterns that has the highest correlation with the registered pattern.
[0105]
Next, FIG. 20 is an explanatory diagram showing in detail the procedure of position correction in steps S1803 and S1808. In step S2001, the imprint region specifying unit 1502a of the fifth collating unit 1405 or the imprint region specifying unit 1702a of the sixth collating unit 1406 specifies imprint regions for each of the registered pattern and the reevaluation pattern.
[0106]
Then, in step S2002 to step S2005, the side position correction unit 1502b or 1702b moves the imprint to a position where the corresponding sides of the imprint region coincide with each other, as in step S1902 to step S1905 of FIG. . Next, in step S2006 to step S2009, the angular position correction unit 1502c or 1702c moves the imprint to a position where corresponding vertices in the imprint area coincide with each other. Note that the order of steps S2002 to S2009 may be reversed.
[0107]
In step S2010, the correlation value calculation unit 1502d or 1702d calculates a correlation value with the registered pattern for each of the re-evaluation pattern after movement and the re-evaluation pattern before movement in steps S2002 to S2009. In step S2011, the position correction unit 1502 or 1702 selects one of the nine re-evaluation patterns that has the highest correlation with the registered pattern.
[0108]
(Performance Evaluation of Imprint Collation Device According to Embodiments 1 and 2)
Now, it is the actual collation capability of the seal collation apparatus described in the first and second embodiments. Since the applicant has evaluated the performance by the collation experiment shown below, the imprint collation apparatus according to the present invention is useful. In order to show the characteristics, the procedure and results of the experiment will be described in detail below.
[0109]
An original image of an imprint used for the experiment is shown in FIG. In this experiment, 36 seals as shown in the figure are collected from one seal. At this time, once ink or the like is applied to the seal, three seals are created in succession. For example, in FIG. 21, the imprint on the leftmost end of the second row is an imprinted stamp without applying vermilion after the imprint of the leftmost end of the first step, and the imprint on the leftmost end of the third step is the imprint on the leftmost end of the second step. After the stamping, it was stamped without adding vermilion.
[0110]
In this way, 36 imprints are obtained by repeating 3 consecutive imprints 12 times, but in 6 of these half, the imprint is intentionally rotated. In FIG. 21, the 1st to 3rd level imprints are normally printed straight, and the 4th to 6th level imprints are intentionally rotated and stamped.
[0111]
Since 49 kinds of seals are used in the experiment, the total number of stamps collected is 1764 (= 36 × 49) per color. The collected stamps are saved one by one in a file named “n + seal number + ink color + yno + xno.bmp”. Here, the “seal number” is any numerical value from “000” to “048” given to each seal to identify 49 types of seals.
[0112]
“Ink color” is a character string of “org” representing vermilion, “ble” representing blue stamp ink, “blk” representing black stamp ink, and “red” representing red stamp ink. However, for now, we will consider only vermilion. Further, as shown in FIG. 21, “yno” and “xno” are numerical values from “00” to “05” that can identify 36 imprints collected for the same seal by the combination of both.
[0113]
Therefore, for example, “Hasegawa” at the leftmost end of the first row shown in FIG. 21 is saved as a file “n000org0000.bmp”, and “Hasegawa” at the leftmost end of the sixth row is saved as a file “n000org0500.bmp”. Become. In addition, a seal imprinted from “Hasegawa” in the figure, for example, a certain seal imprinted from “Aoyama” is stored in “n023org0000.bmp”, a seal imprinted from “Iketani” is stored in “n038org0000.bmp”, etc. Will be. Note that the image size is 256 × 256 pixels, and the gray value of each pixel is expressed in 256 gradations.
[0114]
Further, a binarized image, an evaluation pattern, and a reevaluation pattern are created from these original images in the same manner as in the first and second embodiments. These binarized images are shown in FIG. 22, the evaluation pattern is shown in FIG. 23, and the reevaluation pattern is shown in FIG. In the binarized image shown in FIG. 22, the imprint on the leftmost edge of the sixth stage, which has been close to the left edge, is accurately moved to the center of the entire image in the evaluation pattern shown in FIG. I understand that In addition, it can be seen that the third and sixth level imprints that were blurred in the evaluation pattern shown in FIG. 23 are clearer and clearer in the reevaluation pattern shown in FIG.
[0115]
Note that the binarized image, the evaluation pattern, and the reevaluation pattern created from the same original image are each stored as a file having the same name as the original image, and are distinguished by the difference in the directory where they are located. Further, the histogram 2300, the seal impression area 2301, the vertical and horizontal center lines 2302 shown in FIG. 23 are displayed so that the difference from FIG. 22 is easily understood, and these are actually included in the evaluation pattern file. Is not stored.
[0116]
In the experiment, 49 evaluations stored in the file “n *** org0000.bmp” (*** is any seal number from “000” to “048”) among the evaluation patterns shown in FIG. Each pattern is a registered pattern. Then, the distance value is sequentially calculated by the above-described method 1 for all the combinations of these 49 registered patterns and the remaining 1715 evaluation patterns. Further, the distance values are calculated sequentially by the above methods 2 to 6 for all combinations of the 49 registered patterns and the 1715 reevaluation patterns created from the 1715 evaluation patterns.
[0117]
Then, conversely from these distance values, a threshold value that always determines that the registered pattern collected from at least a separate seal does not match the input pattern is set for each of the 49 registered patterns and for each of the above six methods. Ask.
[0118]
The case where the above threshold value is obtained for the seal stamp “Hasegawa” with the seal number “000” shown in FIG. 23 will be described as an example. First, a distance value between “Hasegawa” shown at the leftmost end of the first row in FIG. 23, that is, the registered pattern of this seal, and the remaining 35 “Hasegawa”, that is, evaluation patterns collected from the same seal as the registered pattern, is used as method 1 Ask for. Since the remaining 35 “Hasegawa” as viewed from the “Hasegawa” in the upper left corner are in a true relationship, the 35 distance values calculated between this registration pattern and each evaluation pattern are collectively registered. This is called the true mark distance of the pattern. Then, the calculated true mark distance is saved as a distance value file named “n + seal number + ink color + yno of registered pattern + xno.true of registered pattern”, in this example, “n000org0000.true”.
[0119]
Next, “Hasegawa” shown at the leftmost end of the first row in FIG. 23, that is, the registration pattern of this seal, 1728 (= 36 × 48) “Aoyama”, “Ikeda”, etc. hidden behind in the figure, That is, the distance value between the registered pattern and the evaluation pattern collected from a seal stamp is obtained by the method 1. Since “Aoyama”, “Ikeda”, and the like are viewed as “Hasegawa” from the perspective of “Hasegawa”, 1728 distance values calculated between this registered pattern and each evaluation pattern are collectively displayed as false in this registered pattern. This is called the mark distance. The calculated false stamp distance is saved as a distance value file named “n + seal number + ink color + yno of registered pattern + xno.fake of registered pattern”, in this example, “n000org0000.fake”.
[0120]
Further, the distance values are calculated in the same manner as in methods 2 to 6, and the true mark distance of “Hasegawa” calculated by each method is the above “n000org0000.true”, the false mark distance is the above “n000org0000.fake”, Each will be added. However, regarding the distance value calculation by the methods 2 to 6, the target is not the evaluation pattern but the reevaluation pattern. Similarly, for the remaining 48 registered patterns, a file of true mark distance and a file of false mark distance obtained by methods 1 to 6 are respectively created.
[0121]
The threshold value to be obtained, that is, the threshold value for which it is determined that the registered pattern collected from at least a separate seal does not match the input pattern is “Hasegawa” and Method 1, specifically “n000org0000.fake”. Among the described false marking distances, it is the minimum value of 1728 false marking distances calculated by the method 1. That is, the distance value according to the method 1 of the evaluation pattern collected from a seal different from the seal where the registered pattern “Hasegawa” was collected is always equal to or greater than the minimum value. If it is determined that the evaluation pattern equal to or greater than the minimum value does not match the registered pattern, the registered pattern “Hasegawa” and the evaluation patterns “Aoyama”, “Ikeya”, etc. are always determined to be inconsistent. Alternatively, in the determination using the minimum value of the false marking distance as a threshold value, it may be said that the probability that an evaluation pattern having a false marking relationship with the registered pattern is rejected as a false marking is 100%.
[0122]
However, among the 35 evaluation patterns “Hasegawa” collected from the same seal as the registered pattern “Hasegawa”, the distance value is higher than evaluation patterns such as “Aoyama” and “Ikeda” collected from another seal. Some of these “Hasegawa” are true marks, but the above threshold values are determined to be false marks. Is not always true. In other words, a false mark rejection rate of 100% does not necessarily mean a true mark acceptance rate of 100%.
[0123]
In short, such a determination error indicates that the ideal distance value cannot be calculated so that the true mark distance and the false mark distance are completely separated from each other at a specific value. The four-stage imprint verification described in the first and second embodiments may be said to be an attempt to improve the accuracy of separating true marks and false marks by distance values by stepwise combining distance value calculation by various methods. it can. The purpose of this experiment is also to measure what percentage (ideally 100%) the true seal acceptance rate under the conditions of 100% false seal rejection rate in the seal imprinting apparatus according to the first and second embodiments. There is in point to do.
[0124]
In the experiment, the above-mentioned 49 registered patterns are stored in the registered pattern storage unit 200 of the imprint matching apparatus according to the first embodiment, and the original image of the registered pattern is deleted 1715 (= 36 × 49−49) in the unknown pattern input unit 201. ) The threshold value of each registered pattern calculated by the methods 1 to 4 is given to the first collation unit 203 to the fourth collation unit 206 for the original images of the imprints. Then, these 49 registered patterns and 1715 input patterns are collated sequentially.
[0125]
FIG. 25 is a graph showing the results of the verification experiment. The graph shows the verification rate of the device when the above threshold value giving a false seal rejection rate of 100% is given to the seal impression verification device according to the first embodiment. Further, in order to investigate the influence of the presence or absence of imprinting, the collation rate is 539 imprints (= 11 × 49) imprinted for the first time with vermilion, followed by imprints 588 imprinted for the second time (= 12 × 49). ), And for every 588 imprints (= 12 × 49) imprinted for the third time. It should be noted that the numbers in the graph are the number of imprints that do not match any registered pattern, in other words, are rejected as not being true for any registered pattern.
[0126]
For example, there are 27 (= 2 + 8 + 17) impressions rejected by the collation by the first collation unit 203, and 7 imprints rejected by the collation by the second collation unit 204 (= 0 + 2 + 5). This indicates that 20 of the 27 impressions rejected in Method 1 were newly accepted by Method 2 and were relieved. Then, the number of impressions finally rejected through the first verification unit 203 to the fourth verification unit 206 is zero, that is, by combining the methods 1 to 4, the condition of a false seal rejection rate of 100% It shows that the true mark acceptance rate below could be 100%.
[0127]
In addition, as shown in FIG. 21, in this experiment, the input pattern intentionally includes a large number of angular deviations and positional deviations from the registered pattern. By correcting the positional deviation, the identity of any of these input patterns with any of the registered patterns is confirmed by any one of the first matching unit 203 to the fourth matching unit 206. In other words, the imprint checking apparatus according to the first embodiment can be used for any registered stamps even if there is a problem such as imprinting that is not stamped at a certain position or that is not stamped at a certain angle. This means that it was always possible to identify that it belongs to one of the registered seals.
[0128]
Further, as shown in FIG. 21, in this experiment, many input patterns having a problem with clarity such as chipping or fainting are intentionally used, but collation is performed not by the evaluation pattern but by the re-evaluation pattern after edge enhancement. Thus, for example, the verification rate of the imprint imprinted for the third time with vermilion is increased from 97.1% (= (588-17) / 588) to 99.1% (= (588-5) / 588). Is able to.
[0129]
Incidentally, the above first gives 1715 input patterns to the first collation unit 203, only those rejected here are given to the second collation unit 204, and also those rejected here are given to the third collation unit 205. This is the result when collating step by step and serially as shown in the above, but 1715 input patterns were given to each of the above parts independently, and the experiment was carried out individually and in parallel. The results are shown in FIG. Method 4 that corrects both the angle shift and the position shift using the uncompressed re-evaluation pattern has a matching rate of 100% even by itself.
[0130]
FIG. 25 and FIG. 26 show the collation results when both the registration pattern and the input pattern are imprinted with vermilion. Next, the registration pattern imprinted with vermilion and the input pattern imprinted with other than vermilion A collation experiment is performed. In addition to the 1715 imprints by vermilion given above, 1764 (= 36 × 49) imprints printed with blue stamp ink are added to the unknown pattern input unit 201 of the seal collation apparatus according to the first embodiment. A total of 7007 original images of 1764 imprints imprinted with stamp ink and 1764 imprint imprinted with red stamp ink are given and collated in the same manner.
[0131]
FIG. 27 is a graph showing the results of the verification experiment. The number of impressions finally rejected through the methods 1 to 4 is 0 for vermilion, but 3 for blue ink, 5 for black ink, and 4 for red ink. That is, a total of 12 seals are mistakenly rejected as not belonging to any seal.
[0132]
On the other hand, FIG. 28 shows an experimental result when the same original image is given to the unknown pattern input unit 1401 of the seal impression collation apparatus according to Embodiment 2 and collation is performed in the same manner. Imprints that are ultimately rejected via methods 1, 2, 5 and 6 have been reduced to two. This indicates that the verification rate can be improved by replacing the methods 3 and 4 with the more accurate methods 5 and 6. Note that the two finally rejected were both immensely smudged.
[0133]
【The invention's effect】
As described above, according to the first aspect of the present invention, the imprint area specifying means for specifying the imprint area in which the imprint image exists is specified for the image including the imprint image inside, and the imprint area specifying means specifies the imprint area specifying means. Imprint center calculating means for calculating the center position of the imprint area, and correcting the position of the imprint image to a position where the center position of the image and the center position of the imprint area calculated by the imprint center calculating means match. Center position correction means, so that the position of the imprint in the image is adjusted in advance so as to be in the center of the image, so that even if there is a problem with the imprinted position, the imprint of the collation source There is an effect that it is possible to obtain an imprint verification device capable of performing high-accuracy verification.
[0134]
According to the invention described in claim 2, in the invention described in claim 1, the imprint region specifying means rotates the imprint image by different angles around an arbitrary position in the image. Since the imprinted area of the image has the largest area as the imprinted area of the image, the imprinted image in the image is located at the center of the image even if it is missing or faint As a result, it is possible to obtain an imprint collation apparatus that can perform collation with the imprint of the collation source with high accuracy even if the imprint is imprinted or has a problem with clarity. Play.
[0135]
According to the third aspect of the present invention, rotation correction means for rotating the imprint image in the first image including the imprint image therein around its center position, and the imprint image by the rotation correction means. For each of the first image in which the image is rotated and the second image including the imprint image therein, the imprint area specifying means for specifying the imprint area where the imprint image exists, and the imprint area specifying means. Side position correction means for correcting the position of the imprinted image in the first image at a position where one side of the imprinted region of the first image and a corresponding side of the imprinted region of the second image coincide with each other. And a distance value calculating means for calculating a distance value between the first image whose position is corrected by the side position correcting means and the second image, and a distance value calculating means. Since there is provided a coincidence determining means for determining whether or not the imprint image in the first image and the imprint image in the second image match based on the distance value, the imprint in each image is After the relative angular deviation is corrected, the relative positional deviation is further corrected at four locations and collation is performed, so that even imprints with a problem with the stamped position and angle are collated. An effect is obtained in that an imprint verification apparatus capable of performing high-precision collation with the original imprint is obtained.
[0136]
According to a fourth aspect of the present invention, in the third aspect of the present invention, the vertex of the imprint area of the first image specified by the imprint area specifying means and the second An angular position correction unit that corrects the position of the imprint image in the first image at a position where a corresponding vertex of the imprint region of the image coincides, and the distance value calculation unit includes the side position correction unit Alternatively, since the distance value between the first image whose position of the imprinted image is corrected by the angular position correcting means and the second image is calculated, the imprint in each image has its relative angular deviation. After the correction, the relative misregistration is further corrected at eight locations, and collation is performed. As a result, even if there is a problem with the imprinted position and angle, collation with the imprint of the collation source is performed. Imprinting that can be performed with high accuracy Apparatus an effect that can be obtained.
[0137]
According to the invention described in claim 5, the imprint area specifying means for specifying the imprint area where the imprint image exists for each of the first image and the second image including the imprint image therein, and the imprint The position of the imprinted image in the first image is set to a position where one side of the imprinted area of the first image specified by the area specifying unit coincides with one side corresponding to the imprinted area of the second image. Side position correcting means to be corrected, rotation correcting means for rotating the imprint image in the first image whose position is corrected by the side position correcting means, with the center position as a center, and imprinting by the rotation correcting means Distance value calculating means for calculating a distance value between the first image obtained by rotating the image and the second image, and the first image in the first image based on the distance value calculated by the distance value calculating means. of And a coincidence determining means for determining whether or not the shadow image and the imprint image in the second image coincide with each other, so that the imprint in each image has its relative positional deviation corrected at four locations. After that, the relative angular deviation is corrected and collation is performed, so that even if the imprinted position and angle have a problem, collation with the imprint of the collation source is performed with high accuracy. There is an effect that an imprint checking device capable of performing the above is obtained.
[0138]
According to the invention described in claim 6, in the invention described in claim 5, the vertex of the imprint area of the first image specified by the imprint area specifying means and the second An angular position correction unit that corrects the position of the imprint image in the first image at a position where a corresponding vertex of the imprint region of the image coincides, and the rotation correction unit includes the side position correction unit or Since the imprint image in the first image whose position has been corrected by the angular position correcting means is rotated, the imprint in each image is corrected for its relative positional shift at eight locations, and then further relative to the imprint image. Imprint verification device that can collate after correcting the angular misalignment, and can collate with the imprint of the collation source with high accuracy even if there is a problem with the imprinted position and angle The effect that can be obtained Unlikely to.
[0139]
According to the invention described in claim 7, in the invention described in claim 5 or 6, the first image obtained by rotating the imprint image by the rotation correcting unit, and the first image For each of the two images, a second imprint area specifying means for specifying an imprint area in which the imprint image exists, and one side of the imprint area of the first image specified by the second imprint area specifying means; Second distance position correction means for correcting the position of the imprint image in the first image at a position where the corresponding one side of the imprint area of the second image matches, and calculating the distance value The means calculates a distance value between the first image whose position of the imprinted image is corrected by the second side position correcting means and the second image, so that the imprint in each image is a relative value thereof. Correct misalignment at four locations After that, the relative angular misalignment is further corrected, and the relative misalignment is corrected again at four locations, and then collation is performed. As a result, an imprint having a problem with the stamped position and angle is obtained. Even if it exists, there exists an effect that the imprint verification apparatus which can perform collation with the imprint of the collation origin with high precision is obtained.
[0140]
According to the invention described in claim 8, in the invention described in claim 7, further, the vertex of the imprint area of the first image specified by the second imprint area specifying means and the A second angular position correcting unit that corrects the position of the imprint image in the first image at a position such that a corresponding vertex of the imprint region of the second image coincides; and the distance value calculating unit includes: Since the distance value between the first image and the second image whose position of the imprint image has been corrected by the second side position correcting unit or the second angular position correcting unit is calculated, Imprints in the image were corrected for their relative misalignment at 4 or 8 locations, then for their relative angular misalignment, and again for their relative misalignment at 8 locations. Will be verified later, so that the stamped position and corner An effect that matching seal imprint verification device capable of performing a high precision and a there be matching original imprint imprint problematic in is obtained.
[0141]
According to the invention of claim 9, an imprint area specifying step for specifying an imprint area in which the imprint image exists for an image including an imprint image therein, and an imprint area specified in the imprint area specifying step An imprint center calculating step for calculating a center position, and a center position correction for correcting the position of the imprint image to a position where the center position of the image coincides with the center position of the imprint region calculated in the imprint center calculating step. The position of the imprint in the image is adjusted in advance so that it comes to the center of the image, so that even if there is a problem with the imprinted position, it can be compared with the imprint of the verification source. It is possible to obtain an imprint checking method capable of performing the above with high accuracy.
[0142]
According to the invention of claim 10, a rotation correction step of rotating the imprint image in the first image including the imprint image therein around its center position, and the imprint image in the rotation correction step. For each of the first image and the second image including the imprint image therein, the imprint region specifying step for specifying the imprint region where the imprint image exists, and the imprint region specifying step. An edge position correcting step for correcting the position of the imprinted image in the first image at a position where one side of the imprinted area of the first image coincides with a corresponding side of the imprinted area of the second image. And a distance value calculating step for calculating a distance value between the first image whose position of the imprint image has been corrected in the side position correcting step and the second image, and a distance value calculating step. Based on distance value before And a matching determination step for determining whether or not the imprint image in the first image matches the imprint image in the second image. Therefore, the imprint in each image has a relative angular shift. After the correction, the relative misregistration is further corrected at four locations, and collation is performed. As a result, even if there is a problem with the imprinted position and angle, collation with the imprint of the collation source is performed. It is possible to obtain an imprint checking method capable of performing the above with high accuracy.
[0143]
According to the invention of claim 11, an imprint area specifying step for specifying an imprint area where the imprint image exists for each of the first image and the second image including the imprint image therein, and the imprint The position of the imprinted image in the first image is set to a position where one side of the imprinted area of the first image specified in the area specifying step matches one side of the imprinted area of the second image. A side position correction step to be corrected, a rotation correction step for rotating the imprint image in the first image whose position has been corrected in the side position correction step about the center position, and the imprint in the rotation correction step A distance value calculating step for calculating a distance value between the first image obtained by rotating the image and the second image, and the first image in the first image based on the distance value calculated in the distance value calculating step. Imprint image and the second image And a matching determination step for determining whether or not the imprint image in the image coincides with each other. Therefore, the imprint in each image has its relative positional deviation corrected at four positions, and then the relative image shift is further performed. After checking the angle deviation, it is verified and this provides an imprint verification method that enables highly accurate verification with the imprint of the verification source even if there is a problem with the imprinted position or angle. There is an effect that is.
[0144]
According to the invention described in claim 12, there is an effect that a program capable of causing a computer to execute the method described in any one of claims 9 to 11 is obtained.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram showing a hardware configuration of an imprint checking apparatus according to a first embodiment.
FIG. 2 is an explanatory diagram functionally showing the configuration of the seal collation apparatus according to the first embodiment.
FIG. 3 is an explanatory diagram showing in more detail the functional configuration of a preprocessing unit according to the first embodiment;
FIG. 4 is an explanatory diagram illustrating an example of an input pattern in which an imprint is not positioned at the center of an image.
FIG. 5 is an explanatory diagram showing a state in which an imprint area is not accurately obtained when there is a lack or fading in the imprint.
FIG. 6 is an explanatory diagram showing a state in which an imprint is rotated in order to accurately obtain an imprint area even when the imprint is missing or faint.
FIG. 7 is another explanatory diagram showing a state in which an imprint is rotated in order to accurately obtain an imprint area even when the imprint is missing or faint.
8 is an explanatory diagram showing an example of an evaluation pattern created from the input pattern shown in FIG.
FIG. 9 is an explanatory diagram showing in more detail the functional configuration of the first verification unit according to the first embodiment.
FIG. 10 is an explanatory diagram showing in more detail the functional configuration of a second verification unit according to the first embodiment.
FIG. 11 is an explanatory diagram showing the functional configuration of a third matching unit according to Embodiment 1 in more detail.
12 is an explanatory diagram showing in more detail the functional configuration of a fourth verification unit according to Embodiment 1. FIG.
FIG. 13 is an explanatory diagram showing a state of imprint movement performed by the side position correcting unit according to the first embodiment.
FIG. 14 is an explanatory diagram functionally showing the configuration of the seal collation apparatus according to the second embodiment.
FIG. 15 is an explanatory diagram showing in more detail the functional configuration of a fifth collating unit according to the second embodiment;
FIG. 16 is an explanatory diagram illustrating a state of imprint movement performed by the angular position correction unit according to the second embodiment;
FIG. 17 is an explanatory diagram showing in more detail the functional configuration of a sixth collation unit according to the second embodiment;
FIG. 18 is an explanatory diagram showing a comparison of procedures of methods 1 to 6 used in the first and second embodiments.
FIG. 19 is an explanatory diagram showing in detail the procedure of position correction shown in FIG. 18;
FIG. 20 is an explanatory diagram showing another position correction procedure shown in FIG. 18 in more detail.
FIG. 21 is an explanatory diagram showing an example of an original image used for performance evaluation of the seal collation apparatus according to the first and second embodiments.
22 is an explanatory diagram showing an example of a binarized image created from the original image shown in FIG.
23 is an explanatory diagram showing an example of an evaluation pattern created from the binarized image shown in FIG.
24 is an explanatory diagram showing an example of a re-evaluation pattern created from the evaluation pattern shown in FIG.
FIG. 25 is a graph showing a matching rate of the device when 1715 input patterns by vermilion and a threshold value with a false imitation rejection rate of 100% are given to the seal impression matching device according to the first embodiment.
FIG. 26 is a graph showing the collation rate of the device when 1715 input patterns by vermilion and a threshold value with a false imitation rejection rate of 100% are given to the seal impression collation device according to the first embodiment.
FIG. 27 shows the collation of the imprint checking apparatus according to the first embodiment when 7007 input patterns using red, blue, black, and red stamp inks and a threshold value with a false imprint rejection rate of 100% are given. It is a graph which shows a rate.
FIG. 28 shows the collation of the imprint verification apparatus according to the second embodiment when 7007 input patterns using vermillion, blue, black, and red stamp ink and a threshold value with a false seal rejection rate of 100% are given. It is a graph which shows a rate.
[Explanation of symbols]
100 bus or cable
101 CPU
102 ROM
103 RAM
104 HDD
105 HD
106 FDD
107 FD
108 display
109 I / F
110 Communication cable
111 keyboard
112 mouse
113 scanner
114 printer
115 CD-ROM
116 CD-ROM drive
200, 1400 Registered pattern storage unit
201,1401 Unknown pattern input part
202,1402 Pre-processing unit
203,1403 first verification unit
204, 1404 Second verification unit
205 Third verification unit
206 Fourth verification unit
300 Binarization part
301 Evaluation pattern creation unit
900,1001 compression unit
901, 1002, 1101, 1201, 1501, 1701 Rotation correction unit
902, 1003, 1102, 1203, 1503, 1703 Smoothing unit
903, 1004, 1103, 1204, 1504, 1704 Distance value calculation unit
904, 1005, 1104, 1205, 1505, 1705 Match determination unit
1000, 1100, 1200, 1500, 1700 Re-evaluation pattern creation unit
1202, 1502, 1702 Position correction unit
1405 Fifth verification unit
1406 Sixth verification unit

Claims (3)

An imprint area specifying means for specifying an imprint area in which the imprint image exists for an image including the imprint image therein;
An imprint center calculating means for calculating a center position of the imprint area specified by the imprint area specifying means;
Center position correction means for correcting the position of the imprint image at a position such that the center position of the image and the center position of the imprint area calculated by the imprint center calculation means match;
Bei to give a,
The imprinted area specifying means has, as an imprinted area of the image, an area having the largest area among the imprinted areas in a state where the imprinted image is rotated by different angles around an arbitrary position in the image. A seal impression device characterized by specifying . An imprint area specifying step for specifying an imprint area in which the imprint image exists for an image including the imprint image therein;
An imprint center calculating step of calculating a center position of the imprint region specified in the imprint region specifying step;
A center position correcting step of correcting the position of the imprint image at a position such that the center position of the image and the center position of the imprint region calculated in the imprint center calculating step match;
Only including,
In the imprint area specifying step, an imprint area of the image having the largest area among the imprint areas in a state where the imprint image is rotated by different angles around an arbitrary position in the image is used as the imprint area of the image. An imprint matching method characterized by specifying . Computer
  An imprint area specifying means for specifying an imprint area in which the imprint image exists for an image including the imprint image therein;
  An imprint center calculating means for calculating a center position of the imprint area specified by the imprint area specifying means;
  Center position correcting means for correcting the position of the imprint image to a position such that the center position of the image coincides with the center position of the imprint area calculated by the imprint center calculating means;
  An imprint checking program that realizes the function as
  The imprinted area specifying means has, as an imprinted area of the image, an area having the largest area among the imprinted areas in a state where the imprinted image is rotated by different angles around an arbitrary position in the image. Imprint collation program characterized by specifying.

Images