JP4165203B2 - Image collation device, image collation method, and image collation program - Google Patents

Description

【００４１】
ＣＰＵ１１は、各指紋画像から抽出した平均値の最大値Ｍａｘ（Ａ（ｉ））と最小値Ｍｉｎ（Ａ（ｉ））と、その平均値Ａｖｅ（Ａ（ｉ））との割合Ｒａを計算する（ステップＣ３）。ここで、最大値と最小値との差とその平均値との割合Ｒａが予め設定された閾値Ｒａｔｈ以下であるかを判別する（ステップＣ４）。
【００４２】
ここで、割合Ｒａが予め設定された閾値Ｒａｔｈ以下でなかった場合には、照合アルゴリズムとしてパターンマッチング方式を選択し（ステップＣ６）、割合Ｒａが予め設定された閾値Ｒａｔｈ以下であった場合には、照合アルゴリズムとして特徴点方式を選択する（ステップＣ５）。
【００４３】
以下、画像照合装置１０は、第１実施形態と同様にして、選択された照合アルゴリズムにより、照合対象とする指紋画像に対する照合を実行する。
【００４４】
このようにして、ｎ回採取された指紋画像から得られた画素値の平均値に基づいて、照合アルゴリズムを選択することができるので、指紋の状態が安定しないことにより平均値がばらついて特徴点方式を用いるのが好ましくない場合には、パターンマッチング方式を利用することができる。画素値の平均値を算出する処理は、特徴点抽出の処理よりも通常では処理負担が軽いので、このスイッチングアルゴリズムを用いることで、全体の処理時間に対してデメリットとならない範囲で照合アルゴリズムの切り替えを実行し、総合的に性能の良い画像照合装置を提供することができる。
【００４５】
（第４実施形態）
次に、本発明の第４実施形態について説明する。第４実施形態では、第１実施形態におけるスイッチングアルゴリズム（図２、ステップＳ２）が異なっているだけであるので、スイッチングアルゴリズムについて説明する。
【００４６】
第３実施形態では、指紋画像の画素の濃度値（指紋画像の濃度値の平均値）に基づいて照合アルゴリズムを選択しているが、平均値が同じであっても標準偏差の差が大きい場合には性質の異なる指紋画像としてみることができる。第４実施形態のスイッチアルゴリズムでは、平均値だけでなく標準偏差σに基づいて照合アルゴリズムを選択する。
【００４７】
図７には、第４実施形態におけるスイッチングアルゴリズムのフローチャートを示している。
【００４８】
ＣＰＵ１１は、生体画像読み取り装置１６によって指紋画像をｎ回採取し、各指紋画像のそれぞれについて画素値の平均値と標準偏差を算出する。平均値は、第３実施形態において示す式（１）に従って算出し、標準偏差については以下に示す式（１）に従って算出する。ここで、各指紋画像から算出した画素値の標準偏差をσｉ（ｉ＝１〜ｎ）とする（ステップＤ１，Ｄ２）。
【００４９】
【数２】

【００５０】
ＣＰＵ１１は、各指紋画像から抽出した平均値の最大値Ｍａｘ（Ａ（ｉ））と最小値Ｍｉｎ（Ａ（ｉ））と、その平均値Ａｖｅ（Ａ（ｉ））との割合Ｒａを計算する（ステップＤ３）。ここで、最大値と最小値との差とその平均値との割合Ｒａが予め設定された閾値Ｒａｔｈ以下であるかを判別する（ステップＤ４）。
【００５１】
ここで、割合Ｒａが予め設定された閾値Ｒａｔｈ以下でなかった場合には、照合アルゴリズムとしてパターンマッチング方式を選択する（ステップＤ８）。
【００５２】
一方、割合Ｒａが予め設定された閾値Ｒａｔｈ以下であった場合には、ＣＰＵ１１は、各指紋画像から抽出した画素値の標準偏差の最大値Ｍａｘ（σ（ｉ））と最小値Ｍｉｎ（σ（ｉ））と、その平均値Ａｖｅ（σ（ｉ））との割合Ｒｓを計算する（ステップＤ５）。ここで、最大値と最小値との差とその平均値との割合Ｒｓが予め設定された閾値Ｒｓｔｈ以下であるかを判別する（ステップＤ６）。
【００５３】
ここで、割合Ｒｓが予め設定された閾値Ｒｓｔｈ以下でなかった場合には、照合アルゴリズムとしてパターンマッチング方式を選択する（ステップＤ８）。
【００５４】
一方、割合Ｒｓが予め設定された閾値Ｒｓｔｈ以下であった場合には、ＣＰＵ１１は、照合アルゴリズムとして特徴点方式を選択する（ステップＤ７）。
【００５５】
以下、画像照合装置１０は、第１実施形態と同様にして、選択された照合アルゴリズムにより、照合対象とする指紋画像に対する照合を実行する。
【００５６】
このようにして、ｎ回採取された指紋画像から得られた画素値の平均値と標準偏差に基づいて、照合アルゴリズムを選択することができるので、平均値がばらついていないとしても標準偏差にばらつきがあって特徴点方式を用いるのが好ましくない場合には、パターンマッチング方式を利用することができる。画素値の平均値だけでなく標準偏差に基づいて照合アルゴリズムの切り替えを実行することで、より正確な切り替えが可能となり、性能の良い画像照合装置を提供することができる。
【００５７】
なお、前述した第４実施形態では、標準偏差σに基づいて照合アルゴリズムを選択しているが、分散値（標準偏差σの二乗値）に基づくものとしても良い。
【００５８】
また、前述した第２〜第４実施形態では、スイッチングアルゴリズムにおいて、指紋画像から抽出される特徴点数のばらつきを、複数種類の特徴点、例えば端点と分岐などを合わせて判断しているが、特徴点の種類毎の数を独立して判断するようにしても良い。こり場合、特徴点の種類によって優先度をつける、重み付けをするといったことをしても良い。
【００５９】
（第５実施形態）
次に、本発明の第５実施形態について説明する。第２〜第４実施形態では、指紋画像の照合時に複数回（ｎ回）の指紋画像の採取を行っているが、第５実施形態では、時間を経て行われる複数回の照合時に取得された指紋画像をもとに、スイッチングアルゴリズムにより照合アルゴリズムを切り替えるようにする。
【００６０】
図８は、第５実施形態における画像照合装置１０の処理の流れを説明するための図である。
【００６１】
画像照合装置１０は、画像照合処理を行なう前に、照合時に用いる登録画像を生成するために、生体画像読み取り装置１６から画像の読み取りを行なう（図８（１））。ＣＰＵ１１は、生体画像読み取り装置１６によって読み取られた画像をもとに、照合対象とする生体の画像を照合するための登録画像を生成する。ここでは、画像照合装置１０に用意された複数の照合アルゴリズム、ここでは特徴点方式とパターンマッチング方式のそれぞれによる処理を実行して、入力された指紋画像を各方式に対応する登録画像に変換する（図８（２）（３）、図８（４）（５））。以下、生体画像読み取り装置１６から入力された指紋画像に対して、それぞれの方式に応じた登録画像に基づく照合を行なう。
【００６２】
画像照合装置１０は、ある時点の照合時（ｄ１）に生体画像読み取り装置１６から入力された指紋画像（照合画像（ｄ１）（図８（６）））の照合に用いる照合アルゴリズムを、スイッチングアルゴリズムにより選択する（図８（７））。例えば、これまでに入力された指紋画像が１つである場合には、前述した第１実施形態と同様にしてスイッチングアルゴリズムにより照合アルゴリズムを選択する。ここで、スイッチングアルゴリズムにより特徴点方式が選択されると、ＣＰＵ１１は、入力された照合画像（ｄ１）について、予め登録されている特徴点方式による登録画像（ｄ１）（図８（８））をもとにして、特徴点方式での照合処理を実行する（図８（９））。
【００６３】
画像照合装置１０は、次のある時点の照合時（ｄ２）に生体画像読み取り装置１６から入力された指紋画像（照合画像（ｄ２）（図８（１０）））の照合に用いる照合アルゴリズムを、スイッチングアルゴリズムにより選択する（図８（１１））。例えば、現在（照合時（ｄ１））と過去（照合時（ｄ１））を含めて複数の指紋画像が採取されている場合には、前述した第２〜第４実施形態の何れかと同様にして、複数の指紋画像をもとにしたスイッチングアルゴリズムにより照合アルゴリズムを選択する（図８（１１））。ここで、スイッチングアルゴリズムによりパターンマッチング方式が選択されると、ＣＰＵ１１は、入力された照合画像（ｄ２）について、予め登録されているパターンマッチング方式による登録画像（ｄ２）（図８（１２））をもとにして、パターンマッチング方式での照合処理を実行する（図８（１３））。
【００６４】
すなわち、第５実施形態における画像照合装置１０は、生体画像読み取り装置１６によって読み取られた照合対象とする画像（照合画像）を含む、過去に読み取られた照合画像、例えば照合時（ｄ１，ｄ２，…）において生体画像読み取り装置１６から読み取られた指紋画像をもとに、それぞれの時点における照合対象とする指紋画像の照合に用いる照合アルゴリズムを選択することができる。
【００６５】
このようにして、登録時に各照合アルゴリズムに対応する登録画像を保存しておくことで、照合時に実行されるスイッチングアルゴリズムによって、照合に用いられる照合アルゴリズムが何れに選択されても、その照合アルゴリズムに対応する登録画像をもとに精度の良い安定した照合を行なうことができる。また、各時点の照合時に採取される時間を経た複数の指紋画像（照合画像）をもとに、スイッチングアルゴリズムによって照合アルゴリズムの選択を行なうので、時間経過に伴う指紋画像の変動に対して、より柔軟に対応することができるようになる。
【００６６】
なお、前述した各実施形態（第１〜第５実施形態）の説明では、照合アルゴリズムとして特徴点方式とパターンマッチング方式の方式の２種類の方式を用いているが、その他の３種類以上の照合アルゴリズムを用いることも可能である。また、指紋画像を照合対象ととして説明しているが、その他の画像、例えば虹彩、網膜、静脈パターン、掌形、耳介等のバイオメトリクス画像に対しても適用可能であり、それぞれのバイオマトリクス画像に応じた照合アルゴリズムを複数用いることができる。また、複数のバイオメトリクス画像を対象とし、それぞれのバイオメトリクス画像に最適な複数の照合アルゴリズムを利用して照合する画像照合装置を構成することも可能である。これにより、総合的に性能の良い画像照合装置を提供することができる。また、画像パターン同士を比較するパターンマッチング方式の照合アルゴリズムとして、特に一方の画像に複数個のテンプレート配置し、他方の画像に対して、その相関係数が最大となる領域を検出し、両者の位置関係の相違により２つの画像の同一性を評価する方式を用いることが可能である。
【００６７】
また、本発明は、前述した各実施形態に限定されるものではなく、実施段階ではその要旨を逸脱しない範囲で種々に変形することが可能である。また、前述した実施形態で実行される機能は可能な限り適宜組み合わせて実施しても良い。前述した実施形態には種々の段階の発明が含まれており、開示される複数の構成要件における適宜の組み合わせにより種々の発明が抽出され得る。例えば、実施形態に示される全構成要件から幾つかの構成要件が削除されても、効果が得られるので有れば、この構成要件が削除された構成が発明として抽出され得る。
【００６８】
また、前述した各実施形態において記載した処理は、コンピュータに実行させることのできる画像プログラムとして、例えば磁気ディスク（フレキシブルディスク、ハードディスク等）、光ディスク（ＣＤ−ＲＯＭ、ＤＶＤ等）、半導体メモリなどの記録媒体に書き込んで各種装置に提供することができる。また、通信媒体により伝送して各種装置に提供することも可能である。画像照合装置を実現するコンピュータは、記録媒体に記録された画像照合プログラムを読み込み、または通信媒体を介して画像照合プログラムを受信し、このプログラムによって動作が制御されることにより、上述した処理を実行する。
【００６９】
【発明の効果】
以上のように、本発明によれば、照合対象とする生体の画像をもとに、複数用意されている照合アルゴリズムから読み取られた画像の照合に用いる照合アルゴリズムを選択し、この選択した照合アルゴリズムに従って画像の照合を行なうので、照合対象とする画像に応じた適切な照合アルゴリズムに従って照合を実行することが可能となるものである。
【図面の簡単な説明】
【図１】本発明の実施形態における画像照合装置１０の構成を示すブロック図。
【図２】本発明の実施形態における画像照合装置１０の画像照合処理について説明するための示すフローチャート。
【図３】第１実施形態における画像照合装置１０の画像照合処理について説明するための概念図。
【図４】第１実施形態におけるスイッチングアルゴリズムのフローチャート。
【図５】第２実施形態におけるスイッチングアルゴリズムのフローチャート。
【図６】第３実施形態におけるスイッチングアルゴリズムのフローチャート。
【図７】第４実施形態におけるスイッチングアルゴリズムのフローチャート。
【図８】第５実施形態における画像照合装置１０の処理の流れを説明するための図。
【符号の説明】
１０…画像照合装置１０、１１…ＣＰＵ１１、１２…記憶装置１２、１３…ＲＡＭ１３、１４…入力装置１４、１５…出力装置１５、１６…生体画像読み取り装置１６。[0001]
BACKGROUND OF THE INVENTION
The present invention provides an image of a living body (biometric image) for personal authentication, Especially fingerprints The present invention relates to an image collation apparatus, an image collation method, and an image collation program for collating images.
[0002]
[Prior art]
In recent years, in order to authenticate an individual, an image collation apparatus that collates a living body image (biometric image), for example, a fingerprint, a palm print, an iris, a vein pattern, and an auricle image has been used.
[0003]
Conventionally, a pattern data matching device for matching fingerprint or palm print pattern data is disclosed in Japanese Patent Laid-Open No. 2001-67474. In this prior art, a plurality of collation programs are received from a host device, and collation program designation information is received from the host device in addition to each feature point information as a large number of file pattern data information to be collated. The control unit of the pattern data collation apparatus reads the collation program designation information attached to the file pattern data information designated by the host apparatus, and executes comparison collation based on the collation program indicated by this information.
[0004]
[Patent Document 1]
JP 2001-67474 A
[0005]
[Problems to be solved by the invention]
As described above, in the conventional pattern data collation apparatus, a plurality of collation programs are prepared to attempt collation accuracy and collation performance. However, collation is executed using a collation program designated by the host apparatus side. There are only. That is, it is unclear whether the collation program specified by the host device is suitable for the file pattern data information to be collated, and even if a plurality of collation programs are prepared, the collation accuracy and collation performance cannot always be improved. .
[0006]
The present invention has been made in view of the above problems, and provides an image matching apparatus, an image matching method, and an image matching program capable of performing matching according to an appropriate matching algorithm according to an image to be verified. The purpose is to provide.
[0007]
[Means for Solving the Problems]
The image collation device of the present invention is a collation target. fingerprint Images of More than once Biological image reading means for reading; Pixel density calculating means for calculating an average value of pixel density for each of the plurality of images read by the biological image reading means, and calculating variation in the average value of pixel density for each image calculated by the pixel density calculating means. If the variation of the average value of the pixel density for each image calculated by the variation calculation unit and the variation calculation unit is equal to or less than a threshold value, the verification is performed based on the feature points extracted from the image. Select a feature point method, and if it is not less than the threshold value, a pattern matching method that performs matching based on the pattern of the image A collation algorithm selecting unit to be selected, and a collating unit for collating the image read by the biological image reading unit in accordance with the collation algorithm selected by the collation algorithm selecting unit.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
[0010]
(First embodiment)
FIG. 1 is a block diagram illustrating a configuration of an image matching device 10 according to the first embodiment. The image collation apparatus 10 is realized by, for example, a computer that reads a program recorded on various recording media and whose operation is controlled by the read program.
[0011]
As shown in FIG. 1, in the image collating apparatus 10, a CPU 11 is connected to a storage device 12, a RAM 13, an input device 14, an output device 15, and a biological image reading device 16.
[0012]
The CPU 11 controls the entire image collating apparatus 10 and executes various processes using the RAM 13 as a work memory by executing a program recorded in a storage device 12 such as a hard disk device or a flash memory. The CPU 11 executes collation with respect to a biological image (hereinafter referred to as a biometric image) input from the biological image reading device 16 by executing an image collation program recorded in the storage device 12.
[0013]
The storage device 12 stores an image matching program for image matching processing, in addition to a control program that controls the overall operation of the image matching device 10. The image matching program includes a plurality of programs based on a matching algorithm used for matching against biometric images. The matching algorithm includes an algorithm based on a feature point method (maneuver method) that performs matching based on feature points extracted from a biometric image, and a pattern matching method that performs matching based on a pattern of the biometric image. Note that programs corresponding to other collation algorithms may be included.
[0014]
The RAM 13 stores programs and data accessed by the CPU 11.
[0015]
The input device 14 includes buttons and a pointing device, and is used for inputting data and various instructions.
The output device 15 includes a display for displaying various information, a speaker for outputting sound, and the like. The display displays a screen accompanying execution of various programs by the CPU 11.
[0016]
The biometric image reading device 16 is a device that reads a biometric image to be collated by image collation processing. The biological image reading device 16 is configured with a structure corresponding to the type of biometric image to be processed. For example, when a biometric image to be collated is a fingerprint image, the biometric image reading device 16 has a reading surface for contacting a fingertip with a size corresponding to the size of a general human finger. Provided. The biological image reading device 16 is provided with a CCD (Charge Coupled Device) sensor that reads a fingerprint of a finger in contact with the reading surface as an optical image. The fingerprint image data read by the CCD sensor is recorded in the RAM 13 and used for image collation processing. The biometric image reading device 16 is provided with a fingertip positioning mechanism or the like that stabilizes the fingertip that is in contact with the reading surface so that a high-quality image can be read. Note that, for example, when a palm print, iris, retina, vein pattern, or auricle image is used as the biometric image, the biometric image reading device 16 is provided with a mechanism for reading each image. For example, if a vein pattern image is detected from the palm of the hand, it can be taken using the fact that only the vein portion in the subcutaneous tissue appears black by irradiating the palm with near infrared rays. Further, the image of the pinna only needs to be photographed in appearance without requiring a positive identification action from a discriminator that requires authentication. In the auricle, a plurality of components classified into a raised portion, a depressed portion, and a flat portion can be considered from the shape of the pearl bone and the depression formed by them, and an individual can be identified from these differences.
[0017]
Next, image collation processing of the image collation apparatus 10 in the first embodiment will be described with reference to the flowchart shown in FIG. 2 and the conceptual diagram shown in FIG. The image matching process is realized by the CPU 11 executing the image matching program stored in the storage device 12. In the following description, a case where collation is performed on a fingerprint image as a biometric image will be described as an example.
[0018]
First, the CPU 11 causes the biological image reading device 16 to read a biological image to be processed (step S1). The biometric image reading device 16 optically scans the fingertip portion in contact with the fingerprint contact surface, reads the fingerprint image, and stores the fingerprint image data in the RAM 13 (FIG. 3 (1)). The CPU 11 performs correction processing on the fingerprint image data stored in the RAM 13. In the correction process, image processing such as enlargement, reduction, and rotation is performed on the digitized fingerprint image data, for example, by removing distortion and noise.
[0019]
Next, the CPU 11 executes a switching algorithm for selecting a collation algorithm to be used for collation of the image based on the fingerprint image data subjected to the correction process (step S2) (FIG. 3 (2)).
[0020]
FIG. 4 shows a flowchart of the switching algorithm in the first embodiment.
First, the CPU 11 detects a feature point of the fingerprint from the fingerprint image data (step A1). As the feature points of the fingerprint, the end points and branch point structures of the ridge (the peak portion of the fingerprint) are extracted. The CPU 11 counts the number of feature points extracted from the fingerprint image (step A2).
[0021]
Here, the CPU 11 selects a matching algorithm according to the number of feature points from which fingerprint images have been extracted. First, the CPU 11 determines whether the number of feature points is greater than or equal to a preset lower limit value. If the number of feature points is not greater than or equal to the lower limit value (step A3, No), the pattern matching method is used as a matching algorithm. Is selected (step A6).
[0022]
On the other hand, if the number of feature points is equal to or greater than the lower limit value (step A3, Yes), it is further determined whether the number of feature points is equal to or less than a preset upper limit value (step A4). Here, if the number of feature points is not less than or equal to the upper limit value (step A4, No), a pattern matching method is selected as a matching algorithm (step A6). If the number of feature points is equal to or less than the upper limit value (step A4, Yes), the feature point method is used as a matching algorithm (step A5).
[0023]
The CPU 11 determines the collation algorithm used for collation of the input fingerprint image as the collation algorithm selected by the switching algorithm, collates the fingerprint image according to this collation algorithm, and outputs the collation result (steps S3 and S4). (FIG. 3 (3) (4)).
[0024]
In the first embodiment, when the number of feature points is a number between a preset lower limit value and upper limit value, the feature point method is selected as the matching algorithm, and in the case of the number of other feature points. Selects the pattern matching method. In the feature point method, if the number of extracted feature points is too small (less than the lower limit value), the matching cannot be performed with high accuracy, and if the number of feature points is too large (exceeding the upper limit value), the feature that changes Since the number of point elements increases and accuracy decreases, it is not adopted.
[0025]
The feature point method takes a long time to extract the feature points, but after that, it becomes a search operation for the registered feature point information based on the extracted feature points. High-speed verification is possible. However, since it is premised on that necessary and sufficient feature points can be extracted, it is not preferable to apply to a fingerprint image with few feature points or too many feature points. On the other hand, in the pattern matching method, various types of image processing are performed on the fingerprint pattern itself, and comparison is performed by comparing the image pattern with a pre-registered image pattern to obtain a similarity. The pattern matching method has a disadvantage that it has a large amount of data. For example, when it is targeted for fingerprint images, it can cope with various finger conditions (effects on fingerprint images such as dry skin, sweat, and scratches). There is an excellent point. In addition, since there is no need to extract feature points, there is an advantage that the uncorresponding rate is small. However, since it is necessary to repeatedly compare the images, there is a disadvantage that it takes longer time to collate than the feature point method.
[0026]
In the first embodiment, most fingerprint images from which feature points can be extracted are supported by a matching algorithm based on the feature point method, and pattern matching is applied to a small number of fingerprint images that are difficult to extract features. Collation is performed by a collation algorithm based on a method, and a fingerprint image for which an appropriate feature point cannot be extracted can be reliably collated without being discarded as unsupported.
[0027]
As described above, the first embodiment uses the advantage that the feature point method is high speed, and also uses the pattern matching method that compensates for the disadvantage that the feature point method is unsupported at a high speed. It is possible to provide an image collation apparatus that can perform simple processing.
[0028]
(Second Embodiment)
Next, a second embodiment of the present invention will be described. In the second embodiment, since only the switching algorithm (FIG. 2, step S2) in the first embodiment is different, the switching algorithm will be described.
[0029]
In the feature point method, the fingerprint is determined when the feature point information matches at a certain rate, rather than using all the feature points when the feature points are extracted. In this case, if the variation in the number of feature points exceeds the number of feature points for matching, there is a possibility that accuracy will be deteriorated in matching by the feature point method. In the switching algorithm of the second embodiment, the matching algorithm is selected based on the variation in the number of feature points.
[0030]
FIG. 5 shows a flowchart of the switching algorithm in the second embodiment.
[0031]
The CPU 11 collects fingerprint images n times by the biological image reading device 16 and extracts feature points from each of the fingerprint images. Here, the number of feature points extracted from each fingerprint image is Mi (i = 1 to n) (steps B1 and B2).
[0032]
The CPU 11 calculates a ratio R between the maximum value Max (Mi) and the minimum value Min (Mi) of the number of feature points extracted from each fingerprint image and the average value Ave (Mi) (step B3). Here, it is determined whether the ratio R between the difference between the maximum value and the minimum value and the average value is equal to or less than a preset threshold value Rt (step B4).
[0033]
Here, when the ratio R is not equal to or less than the preset threshold value Rth, a pattern matching method is selected as a matching algorithm (step B6), and when the ratio R is equal to or less than the preset threshold value Rth. Then, the feature point method is selected as the matching algorithm (step B5).
[0034]
Thereafter, the image collating apparatus 10 performs collation on the fingerprint image to be collated with the selected collation algorithm, as in the first embodiment.
[0035]
In this way, since the matching algorithm can be selected based on the variation in the number of feature points obtained from the fingerprint image collected n times, it is not preferable to use the feature point method because the number of feature points varies. In some cases, a pattern matching method can be used.
[0036]
(Third embodiment)
Next, a third embodiment of the present invention will be described. In the third embodiment, since only the switching algorithm (FIG. 2, step S2) in the first embodiment is different, the switching algorithm will be described.
[0037]
When fingerprint images are collected multiple times, the density value of the fingerprint image pixels (average density value of the fingerprint image) due to the influence of the condition of the fingerprint, dirt on the fingertips, presence or absence of sweat, etc. May change each time it is collected. When the average value of the density values of the fingerprint image varies every time it is collected, there is a possibility that a stable image cannot be obtained at the feature point extraction stage when the feature point method is used. In the switch algorithm of the third embodiment, the matching algorithm is selected based on the average value of the density values of the fingerprint image.
[0038]
FIG. 6 shows a flowchart of the switching algorithm in the third embodiment.
[0039]
The CPU 11 collects fingerprint images n times by the biological image reading device 16 and calculates an average value of pixel values for each fingerprint image according to the following equation (1). Here, an average value of pixel values calculated from each fingerprint image is Ai (i = 1 to n) (steps C1 and C2).
[0040]
[Expression 1]

[0041]
The CPU 11 calculates a ratio Ra between the maximum value Max (A (i)) and the minimum value Min (A (i)) of the average value extracted from each fingerprint image and the average value Ave (A (i)). (Step C3). Here, it is determined whether the ratio Ra between the difference between the maximum value and the minimum value and the average value is equal to or less than a preset threshold value Rath (step C4).
[0042]
Here, when the ratio Ra is not less than or equal to the preset threshold value Rath, a pattern matching method is selected as a matching algorithm (step C6), and when the ratio Ra is less than or equal to the preset threshold value Rath Then, a feature point method is selected as a matching algorithm (step C5).
[0043]
Thereafter, the image collating apparatus 10 performs collation on the fingerprint image to be collated with the selected collation algorithm, as in the first embodiment.
[0044]
In this way, since the collation algorithm can be selected based on the average value of the pixel values obtained from the fingerprint image taken n times, the average value varies due to the unstable state of the fingerprint. If it is not preferable to use a method, a pattern matching method can be used. The processing to calculate the average value of the pixel values is usually lighter than the feature point extraction processing, so by using this switching algorithm, the matching algorithm can be switched within a range where there is no disadvantage to the overall processing time. Can be executed to provide a comprehensively high-performance image matching apparatus.
[0045]
(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described. In the fourth embodiment, since only the switching algorithm (FIG. 2, step S2) in the first embodiment is different, the switching algorithm will be described.
[0046]
In the third embodiment, the matching algorithm is selected based on the density value of the pixels of the fingerprint image (the average value of the density values of the fingerprint image). However, even if the average value is the same, the difference in standard deviation is large. Can be seen as fingerprint images with different properties. In the switch algorithm of the fourth embodiment, the matching algorithm is selected based on the standard deviation σ as well as the average value.
[0047]
FIG. 7 shows a flowchart of the switching algorithm in the fourth embodiment.
[0048]
The CPU 11 collects fingerprint images n times by the biological image reading device 16 and calculates an average value and a standard deviation of pixel values for each fingerprint image. The average value is calculated according to Equation (1) shown in the third embodiment, and the standard deviation is calculated according to Equation (1) shown below. Here, the standard deviation of the pixel value calculated from each fingerprint image is set to σi (i = 1 to n) (steps D1 and D2).
[0049]
[Expression 2]

[0050]
The CPU 11 calculates the ratio Ra between the maximum value Max (A (i)) and the minimum value Min (A (i)) of the average value extracted from each fingerprint image and the average value Ave (A (i)). (Step D3). Here, it is determined whether the ratio Ra between the difference between the maximum value and the minimum value and the average value is equal to or less than a preset threshold value Rath (step D4).
[0051]
Here, when the ratio Ra is not less than or equal to the preset threshold value Rath, a pattern matching method is selected as a matching algorithm (step D8).
[0052]
On the other hand, when the ratio Ra is equal to or less than the preset threshold value Rath, the CPU 11 determines the maximum value Max (σ (i)) and the minimum value Min (σ ( A ratio Rs between i)) and the average value Ave (σ (i)) is calculated (step D5). Here, it is determined whether the ratio Rs between the difference between the maximum value and the minimum value and the average value is equal to or less than a preset threshold value Rsth (step D6).
[0053]
If the ratio Rs is not less than or equal to the preset threshold value Rsth, a pattern matching method is selected as a matching algorithm (step D8).
[0054]
On the other hand, when the ratio Rs is equal to or less than the preset threshold value Rsth, the CPU 11 selects a feature point method as a matching algorithm (step D7).
[0055]
Thereafter, the image collating apparatus 10 performs collation on the fingerprint image to be collated with the selected collation algorithm, as in the first embodiment.
[0056]
In this way, the collation algorithm can be selected based on the average value and standard deviation of the pixel values obtained from the fingerprint image taken n times, so that even if the average value does not vary, the standard deviation varies. If it is not desirable to use the feature point method, the pattern matching method can be used. By executing the switching of the matching algorithm based on the standard deviation as well as the average value of the pixel values, more accurate switching can be performed, and an image matching apparatus with good performance can be provided.
[0057]
In the fourth embodiment described above, the matching algorithm is selected based on the standard deviation σ, but may be based on a variance value (square value of the standard deviation σ).
[0058]
In the above-described second to fourth embodiments, the switching algorithm determines the variation in the number of feature points extracted from the fingerprint image by combining a plurality of types of feature points, for example, end points and branches. You may make it judge independently the number for every kind of point. In this case, priority may be given or weighted depending on the type of feature points.
[0059]
(Fifth embodiment)
Next, a fifth embodiment of the present invention will be described. In the second to fourth embodiments, fingerprint images are collected a plurality of times (n times) at the time of fingerprint image collation. In the fifth embodiment, the fingerprint images are obtained at a plurality of times of collation performed over time. Based on the fingerprint image, the matching algorithm is switched by the switching algorithm.
[0060]
FIG. 8 is a diagram for explaining a processing flow of the image collating apparatus 10 according to the fifth embodiment.
[0061]
Prior to performing the image matching process, the image matching device 10 reads an image from the biological image reading device 16 in order to generate a registered image used at the time of matching (FIG. 8 (1)). Based on the image read by the biometric image reading device 16, the CPU 11 generates a registered image for collating a biometric image to be collated. Here, processing by a plurality of matching algorithms prepared in the image matching device 10, here, the feature point method and the pattern matching method is executed, and the inputted fingerprint image is converted into a registered image corresponding to each method. (FIGS. 8 (2) (3) and 8 (4) (5)). Hereinafter, the fingerprint image input from the biometric image reading device 16 is collated based on the registered image corresponding to each method.
[0062]
The image matching device 10 uses a switching algorithm as a matching algorithm used for matching a fingerprint image (matched image (d1) (FIG. 8 (6))) input from the biological image reading device 16 at the time of matching (d1). To select (FIG. 8 (7)). For example, when only one fingerprint image has been input so far, the collation algorithm is selected by the switching algorithm in the same manner as in the first embodiment described above. Here, when the feature point method is selected by the switching algorithm, the CPU 11 selects a registered image (d1) (FIG. 8 (8)) registered in advance using the feature point method registered for the input collation image (d1). Based on this, collation processing by the feature point method is executed (FIG. 8 (9)).
[0063]
The image collation apparatus 10 uses a collation algorithm used for collation of the fingerprint image (collation image (d2) (FIG. 8 (10))) input from the biological image reading device 16 at the next collation (d2) at a certain time. The selection is made according to the switching algorithm (FIG. 8 (11)). For example, when a plurality of fingerprint images are collected including the present (during collation (d1)) and the past (during collation (d1)), the same as in any of the second to fourth embodiments described above. A collation algorithm is selected by a switching algorithm based on a plurality of fingerprint images (FIG. 8 (11)). Here, when the pattern matching method is selected by the switching algorithm, the CPU 11 uses the registered image (d2) (FIG. 8 (12)) registered in advance for the input collation image (d2). Based on the above, collation processing by the pattern matching method is executed (FIG. 8 (13)).
[0064]
That is, the image collation apparatus 10 according to the fifth embodiment includes a collation image read in the past, including an image to be collated (collation image) read by the biological image reading apparatus 16, for example, at the time of collation (d1, d2, and d2). ...) based on the fingerprint image read from the biometric image reading device 16, a collation algorithm used for collation of the fingerprint image to be collated at each time point can be selected.
[0065]
In this way, by storing a registered image corresponding to each collation algorithm at the time of registration, regardless of which collation algorithm used for collation is selected by the switching algorithm executed at the time of collation, Accurate and stable collation can be performed based on the corresponding registered image. In addition, since a matching algorithm is selected by a switching algorithm based on a plurality of fingerprint images (collation images) that have been collected at the time of collation at each time point, the fluctuation of the fingerprint image with the passage of time is more It becomes possible to respond flexibly.
[0066]
In the description of each of the above-described embodiments (first to fifth embodiments), two types of methods, that is, a feature point method and a pattern matching method are used as a matching algorithm, but other three or more types of matching are used. It is also possible to use an algorithm. In addition, the fingerprint image is described as the target of collation, but it can also be applied to other images such as iris, retina, vein pattern, palm shape, auricle and other biometric images. A plurality of matching algorithms corresponding to images can be used. It is also possible to configure an image matching apparatus that targets a plurality of biometric images and performs matching using a plurality of matching algorithms that are optimal for the respective biometric images. Thereby, it is possible to provide an image collating apparatus having a comprehensively high performance. In addition, as a matching algorithm of a pattern matching method for comparing image patterns, a plurality of templates are arranged in one image, and an area where the correlation coefficient is maximum is detected for the other image. It is possible to use a method for evaluating the identity of two images based on the difference in positional relationship.
[0067]
Further, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention at the stage of implementation. Further, the functions executed in the above-described embodiments may be combined as appropriate as possible. The embodiments described above include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if several constituent requirements are deleted from all the constituent requirements shown in the embodiment, since an effect can be obtained, a configuration from which this constituent requirement is deleted can be extracted as an invention.
[0068]
The processing described in each of the above-described embodiments is an image program that can be executed by a computer, such as a recording on a magnetic disk (flexible disk, hard disk, etc.), an optical disk (CD-ROM, DVD, etc.), a semiconductor memory, etc. It can be written on a medium and provided to various devices. It is also possible to transmit to a variety of devices by transmitting via a communication medium. The computer that implements the image collation apparatus reads the image collation program recorded on the recording medium or receives the image collation program via the communication medium, and performs the above-described processing by controlling the operation by this program. To do.
[0069]
【The invention's effect】
As described above, according to the present invention, a collation algorithm used for collation of images read from a plurality of collation algorithms prepared based on a biological image to be collated is selected, and the selected collation algorithm is selected. Therefore, it is possible to perform collation according to an appropriate collation algorithm corresponding to the image to be collated.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an image collating apparatus 10 according to an embodiment of the present invention.
FIG. 2 is a flowchart for explaining image collation processing of the image collation apparatus 10 according to the embodiment of the present invention.
FIG. 3 is a conceptual diagram for explaining image matching processing of the image matching apparatus 10 according to the first embodiment.
FIG. 4 is a flowchart of a switching algorithm in the first embodiment.
FIG. 5 is a flowchart of a switching algorithm in the second embodiment.
FIG. 6 is a flowchart of a switching algorithm in the third embodiment.
FIG. 7 is a flowchart of a switching algorithm in the fourth embodiment.
FIG. 8 is a view for explaining a processing flow of an image collating apparatus 10 according to a fifth embodiment.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 10 ... Image collation apparatus 10,11 ... CPU11,12 ... Storage device 12,13 ... RAM13,14 ... Input device 14,15 ... Output device 15,16 ... Biological image reading device 16.

Claims (4)

Biometric image reading means for reading a fingerprint image to be verified multiple times ;
Pixel density calculating means for calculating an average value of pixel density for each of the plurality of images read by the biological image reading means;
Variation calculating means for calculating a variation in the average value of the pixel density for each image calculated by the pixel density calculating means;
If the variation in the average value of the pixel density for each image calculated by the variation calculating means is less than or equal to the threshold value, a feature point method for matching based on the feature points extracted from the image is selected as a matching algorithm. If not less than the threshold , the biometric image reading unit reads the biometric image reading unit according to the matching algorithm selected by the matching algorithm selection unit that selects a pattern matching method for matching based on the pattern of the image. An image collating apparatus comprising collating means for collating images. The collation algorithm selection means includes
  Pixel standard deviation calculating means for calculating a standard deviation value for each of the plurality of images read by the biological image reading means;
  Standard deviation variation calculating means for calculating the variation of the standard deviation value for each image calculated by the pixel standard deviation calculating means,
  The matching algorithm selection unit selects the pattern matching method when the variation of the average value of the pixel density for each image calculated by the variation calculation unit is equal to or less than the threshold value, but the variation of the standard deviation value is not equal to or less than the threshold value. The image collating apparatus according to claim 1, which is configured as described above.   A biometric image reading step of reading a fingerprint image to be verified multiple times;
  A pixel density calculating step for calculating an average value of pixel density for each of the plurality of images read by the biological image reading step;
  A variation calculating step for calculating a variation in the average value of the pixel density for each image calculated by the pixel density calculating step;
  If the variation in the average value of the pixel density for each image calculated by the variation calculating step is less than or equal to the threshold value, a feature point method for matching based on the feature points extracted from the image is selected as a matching algorithm. A matching algorithm selection step for selecting a pattern matching method for matching based on the pattern of the image if not below the threshold;
  A collation step of collating the image read by the biometric image reading step according to the collation algorithm selected by the collation algorithm selection step;
An image collating method comprising:   The computer used for the image verification device
  Biological image reading means for reading a fingerprint image to be collated multiple times,
  Pixel density calculation means for calculating an average value of pixel density for each of the plurality of images read by the biological image reading means;
  Variation calculating means for calculating variation in average value of pixel density for each image calculated by the pixel density calculating means;
  If the variation in the average value of the pixel density for each image calculated by the variation calculating means is less than or equal to the threshold value, a feature point method for matching based on the feature points extracted from the image is selected as a matching algorithm. , A matching algorithm selection means for selecting a pattern matching method for matching based on the pattern of the image if not below a threshold value,
  An image collation program for functioning as a collation unit that collates an image read by the biological image reading unit according to the collation algorithm selected by the collation algorithm selection unit.

Images