JP5377580B2 - Authentication device for back of hand and authentication method for back of hand - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a back of hand authentication system and a back of hand authentication method which verify the identity of a user by imaging the user's back of hand vein pattern and matching a branch point, a branch feature, and an optimum branch point weighted value of the vein pattern providing most stable optimum branch characteristics for an individual. <P>SOLUTION: The back of hand authentication system includes: a back of hand authentication terminal device having a finger imaging camera for taking a picture of user's fingers, a back of hand imaging camera for taking an original image of a user's back of hand, and a video decoder connected to the finger imaging camera and the back of hand imaging camera, which converts the original images of the finger and the back of hand into a digital finger image and a digital back of hand image respectively and outputs them through a communication control unit; and a back of hand authentication server having a processor for authentication processing, an authentication information database for storing a reference image and a vein pattern optimum distribution characteristic information used for authentication processing for each personal identification code, a temporary storage memory for storing information temporarily at the time of authentication processing, and an external device interface for outputting an authentication result to an external device. <P>COPYRIGHT: (C)2012,JPO&amp;INPIT

Description

The present invention relates to an authentication terminal for the back of the hand and an authentication method for the back of the hand.

  Generally, card keys such as magnetic cards, RF cards, and smart cards are used in security facilities and authentication systems that require personal identification for large-scale users or small-scale user access control and security. However, the personal identification method using the card key cannot essentially prevent the loss, theft, misuse, etc. of the card, and in the case of a large-scale user, the number of cards provided to the individual increases. There is a disadvantage that the burden on the company becomes large. A biometric recognition system that automatically recognizes and authenticates the user's physical characteristics to compensate for these disadvantages and can be used for user identification, identity authentication, and security access control. Has been developed and used.

  Biological recognition systems are broadly divided into signature recognition that recognizes human behavioral features, fingerprint recognition that recognizes physical features, face recognition, hand shape recognition, and vein pattern on the back of the hand. Etc. are representative.

  Among the biometric recognition systems, the most commonly known fingerprint recognition system has an advantage as a sufficient personal identification system in that the fingerprint is unique for each user, and is located on the epidermis of the human body. By recognizing a fingerprint pattern to be recognized, personal authentication is performed through a good fingerprint image. However, since the fingerprint recognition system is a method for extracting the features of the fingerprint pattern located on the epidermis, it is difficult to expect a good fingerprint image in the case of workers and military personnel at the work site, so the performance of the system deteriorates. In addition to acting as a major factor, sweat, scratches, and external contamination from the hand may also act as factors that degrade the performance of the system.

  Next, among biometric recognition systems, the face recognition system cannot show stable recognition performance due to external changes in external lighting, makeup, glasses, accessories, etc. with the current technology, and the shape of the hand The recognition system must fix the position of the user's hand, so the convenience of the user is inferior, and the deformation of biological features can easily occur due to rheumatism, arthritis, etc. There are things to do.

  For example, in order to improve the shortcomings of the existing biometric recognition system, the personal identification system using the vein pattern on the back of the hand that recognizes the vein pattern located under the skin disclosed in Patent Document 1 is a user in terms of the design of the device. The method of securing the position of the hand of the user is used, and in terms of the algorithm, a method of extracting the veins of the user by applying high bandwidth processing and binarization processing is used. In other words, the personal identification system using the vein pattern on the back of the hand is equipped with a stick that the user can hold with his hand, and when the user holds the stick with his hand, an image of the back of the user's hand is input. Means for extracting the vein pattern of the back of the user's hand, and for correcting the rotation and movement of the user's hand, the branch point of the upper left side and the lower right side of the branch pattern of the back of the hand extracted by the extraction means Comparing means to use and correct are used.

  FIG. 30 illustrates a case where a user holding a stick in the personal identification system using a vein pattern on the back of the hand of the prior art applies force to the hand. FIG. 31 illustrates the user rotating the hand up and down. The case is illustrated.

  As shown in FIGS. 30 and 31, in the conventional means for attaching the personal identification system using the vein pattern of the back of the hand, the user can rotate up and down with respect to the bar depending on the position where the user grips the bar. Occurrence may cause a change in distance from the CCD camera that captures an image and a distortion of the vein pattern to be extracted (geometry distribution).

  Further, the extraction means extracts the pattern by the high bandwidth process and the binarization process without considering the directionality of the vein pattern to be extracted from the back of the user's hand. Due to such extraction means, the conventional personal identification system using the back vein pattern of the hand partly loses the feature vector (pattern branch point) of the vein pattern due to the pattern directionality.

  In the comparison means, in the method of solving the system false rejection (FalseRejection) caused by the rotation and movement of the user's hand, the upper left branch point and the lower right branch point of the extracted hand back vein pattern are used. If the branch point extracted as the reference point is partially lost, it may act as an important factor in system performance degradation by misrecognizing the reference branch point.

  As described above, among the biometric recognition systems, fingerprint recognition, face recognition, hand shape recognition system, etc. have a problem that the performance of the system is reduced due to external contamination such as damaged fingerprint images, sweat, scratches, etc. Since the system using the vein pattern on the back of the hand uses a feature located under the skin, such a problem can be solved. However, the personal identification system using the caudal vein pattern of the prior art can cause system performance degradation such as false recognition or false rejection due to vein pattern distortion or partial loss of vein pattern pattern branching points. There is a problem.

  FIG. 32 is a method for identifying a hand back vein pattern (HVPEA: Hand Vein Pattern Extraction Algorithm) in order by a conventional method for personal identification using the back vein pattern of the hand.

  As shown in FIG. 32, the conventional technique of extracting the back of the hand of the hand obtains a region of interest (ROI) image in which vein patterns are concentratedly distributed from the original image of the back of the user's hand acquired through the camera. Using a vein pattern emphasizing filter (HVPEF: Hand Vein Pattern Emphasizing Filter) that emphasizes only the vein pattern from the ROI image, the boundary portion between the vein pattern distributed on the back of the hand and the background is enhanced (S1, S2) .

  The vein pattern enhancement filter is in the form of a band-pass filter that emphasizes the boundary portion between the vein pattern distributed on the back of the hand and the background in the ROI image, but the low frequency corresponding to the background portion. (Low-frequency) is removed. The characteristic of such a vein pattern enhancement filter is that the gain of the pass-band due to a wide transition section between the high frequency band and the low frequency band cannot be maintained, and the thickness of the vein pattern to be extracted from the ROI image is determined. Because it is not taken into account, the vein pattern connectivity that is an important feature of the vein pattern is partially lost.

  In the binarization process, an image obtained as a result of processing by the vein pattern enhancement filter is used, and only a vein pattern on the back of the hand is extracted by converting a grayscale image into a binary image. (S3). In the resultant image obtained through the binarization process, the vein pattern to be extracted is divided into black (0) and the remaining background portion is divided into white (255).

  The normalization process is applied to the binarized image separated into the vein pattern and the background to be extracted through the binarization process (S4). That is, in the normalization process, the user's database is uniformly used due to the limited memory usage of the system, and whether or not the authorized user is permitted for the image normalized in the recognition process is determined. Needed.

  Next, the noise removal process between the binarization process and the normalization process is performed through a noise removal filter (NRF: Noise Removal Filter), and then through a binarization process and a normalization process. Then, the unwanted projection caused by the subcutaneous fat layer or the like is removed (S5). As a general method for removing the projection by the noise removal filter, a moving average filter that is linear filtering and a median filter that is nonlinear filtering can be used. The prior art hand vein pattern extraction algorithm uses a noise removal process using a median filter.

  FIG. 33 illustrates the results of the vein pattern enhancement filtering process for the ROI image when the back of hand pattern extraction algorithm of FIG. 32 is executed.

  As shown in FIG. 33, the vein pattern enhancement filter emphasizes the vein pattern from the ROI image (a), and after the binarization and normalization process (b), the vein pattern other than the vein pattern enhanced from the ROI image In the background part of (2), noise is removed (c).

  The conventional arterial vein pattern extraction algorithm of the hand extracts only the vein pattern finally distributed on the back of the hand through the vein pattern enhancement filtering, binarization and normalization processing, and the process of removing the clutter. In this conventional technique, the back-vein pattern extraction algorithm extracts the vein pattern without taking into account the contraction and expansion phenomenon of the vein pattern caused by the external temperature change in the acquired ROI image. There is a problem that vein pattern connectivity is lost and stable algorithm performance cannot be provided.

  Furthermore, as a factor that degrades the performance of the conventional arterial vein pattern extraction algorithm, the vein pattern and background boundary information due to the wide transition zone and the information in the vein pattern are partially lost, and the importance of the vein pattern to be extracted is important. There is a problem in that the vein pattern connectivity, which is a unique feature, is partially lost.

  Such a problem is because a vein pattern is extracted with a single filter characteristics without considering the thickness of the vein pattern to be extracted from the extraction algorithm of the hand caudal vein pattern. The vein pattern connectivity, which is an important feature of the pattern, is partially lost, thereby leading to a significant performance degradation phenomenon of the recognition performance of the system.

JP-A-10-295664

The present invention, unlike the existing biometric system that utilizes a feature located in the epidermis, extracts a vein pattern located in the back of the hand subcutaneously, by utilizing the user's confirmation himself, damaged fingerprint image, sweat, Minimize system performance degradation that may occur due to changes in the external environment such as scratches and accessories, provide more stable system performance, and guarantee the user's convenience to the maximum, and the back of the hand authentication terminal and back of the hand An authentication method is provided.

The authentication terminal on the back of the hand of the present invention is
An authentication terminal for the back of a box-shaped hand having a housing in which an operation part is formed obliquely on the front side and a cavity is formed obliquely inside,
The housing is
An opening is formed obliquely in the lower part, and an infrared illuminator on the inner surface of the upper part and an imaging means that reacts to visible light and infrared bands, and a substance that absorbs infrared rays is applied to the inner surface,
further,
An infrared filter provided below the imaging means and the infrared illuminator;
With
In addition, a stable handle is provided at the lower part of the opening that is on the wall side,
The stable handle is
A horizontal bar,
Two shafts are provided on the horizontal bar with a width allowing the middle finger to be inserted, the two shafts are curved to form a housing under the opening, and a central portion of the two shafts serves as the imaging means. The gist is that it is provided so that it can be photographed.

The method for authenticating the back of the hand of the present invention is to insert a middle finger between the two shafts into a stable handle with two shafts at the center of a horizontal shaft rod, and to connect the middle finger, index finger, ring finger and back of the hand. A finger digital image obtained by photographing a connection area, and a back digital authentication method for authenticating a vein of the back of the hand by reading a digital image of the upper part obtained by photographing the back of the hand ,
Computer
Optimal distribution characteristic data (VPBC) including reference polygons connecting the branch points of the veins of the back of the hand obtained by the certifier inserting a middle finger into the two axes of the stable handle and photographing in advance, and the certifier A first reference boundary line, which is an edge portion of a first curved region where the middle finger and the ring finger are in contact with each other, and the middle finger and the index finger are in contact with each other, which is obtained by inserting the middle finger into the two axes of the stable handle and shooting in advance. Storing a second reference boundary line that is an edge portion of the second curved region in the authentication information database as finger boundary reference information ;
Storing the finger digital image and the back digital image from the back authentication terminal of the hand in a temporary storage memory;
From the upper digital image stored in the temporary storage memory, a vein pattern portion necessary for authentication is extracted as ROI image data, and a thick vein pattern (Spi) and a thin vein pattern (ZPi) are extracted and synthesized. And ROI image extraction step of generating a vein pattern (Vpi) from which the vein composite pattern (Gpi) is binarized and noise is removed,
Furthermore, a position correction step and a vein pattern matching step are performed,
The position correction step includes
Extracting the first curved region and the second curved region as finger boundary images in the finger digital image ;
A skeleton process is performed on the first curved region of the finger boundary image, and the first boundary line that is the edge portion between the middle finger side and the ring finger side and the edge portion that is the middle finger and the index finger are Obtaining a boundary line of 2 ;
A first straight line connecting an intersection point included in the first boundary line and an intersection point included in the second boundary line; a reference intersection point included in the first reference boundary line; and a second line Obtaining a first relative angle, which is a relative angle with a second straight line connecting the reference intersection point included in the reference boundary line, and setting this as a first correction angle;
The relative angle between the opening degree of the first boundary line included in the finger boundary image and the reference opening degree of the first reference boundary line included in the finger boundary reference information is set as a second relative angle. And calculating the relative angle between the opening degree of the second boundary line included in the finger boundary image and the reference opening degree of the second reference boundary line included in the finger boundary reference information. Obtaining an angle as a third relative angle, and setting an average of the second relative angle and the third relative angle as a second correction angle;
Correcting the vein pattern (Vpi) at the first correction angle, and further generating a vein pattern corrected image (mVpi) corrected at the second correction angle;
And
The vein pattern matching step includes:
Find each branch point of the vein pattern corrected image (mVpi), generate a polygon connecting these branch points, and compare the polygon with the reference polygon of the optimum distribution characteristic data (VPBC). Means for determining presence or absence;
Displaying the determination result on a display of the back authentication terminal of the hand;
The gist is to do.

As described above, according to the present invention, the system sensitivity determined by slight changes in the external environment and the input image, and the false rejection rate (FRR: False Rejection due to the rotation and movement of the user).
n Rate), a false recognition rate (FAR) is reduced, and vein pattern recognition performance and system stability are dramatically improved.

It is a block diagram of the back authentication system of the hand by the Example of this invention. It is an external view of the back authentication terminal device of the hand by the Example of this invention. It is side sectional drawing of the back authentication terminal device of the hand by the Example of this invention. It is a flowchart of the authentication process of the back authentication system by the Example of this invention. FIG. 5 is a continuation flowchart of the authentication process of the back authentication system according to the embodiment of the present invention. FIG. 6 is a continuation flowchart of the authentication process of the back authentication system according to the embodiment of the present invention. It is an original image applied to the Example of this invention. It is the histogram to which the P-tile method used for ROI automatic detection applied to the Example of this invention was applied. It is an image to which the threshold value of the histogram is applied. It is the ROI image separated from the original image. (A) An image of an emphasized vein pattern, (b) a binary image, and (c) an image from which annoying projection has been removed by the filter bank vein pattern extraction algorithm according to an embodiment of the present invention. (A) A vein pattern image extracted by a conventional enhancement filter, and (b) a vein pattern image extracted by a normal mode filter of a filter-bank vein pattern extraction algorithm applied to the embodiment of the present invention. (A) It is a processing result image of the extraction algorithm of a prior art, (b) It is a processing result image of the filter bank vein pattern extraction algorithm applied to the Example of this invention. It is a conceptual diagram of the finger image applied to the embodiment of the present invention and the original image of the back of the hand. It is a procedure figure of a finger picture and an original picture of the back of a hand applied to an example of the present invention. It is a flowchart of the positional information calculation applied to the Example of this invention. It is an application figure of the smoothing filter applied to the Example of this invention. (A) Digital original image, (b) Solve mask processed image, (c) Kirsh mask processed image, and (d) Edge mask processed image of the image processing using the mask applied to the Example of this invention. It is a process figure for extracting the angle between fingers with the position correction algorithm applied to the example of the present invention. It is a prediction process figure of the primary rotation amendment angle applied to the example of the present invention. It is a prediction process figure of the secondary rotation correction angle applied to the example of the present invention. (A) and (b) are bifurcation feature diagrams of the vein pattern of the same person inputted with time obtained by the conventional vein pattern extraction algorithm. (A) and (b) are bifurcation feature diagrams of a vein pattern of the same person inputted with time obtained by a position correction algorithm of the prior art. It is a flowchart of the recognition algorithm of the back vein pattern of the hand applied to the Example of this invention. FIG. 25 is a flowchart of a detailed process of the training stage in FIG. 24. It is a process figure which applies the optimal bifurcation characteristic obtained by the recognition algorithm of the vein pattern of the back of the hand applied to the example of the present invention, and performs position correction. FIG. 6 is a process diagram for performing rotation and movement correction using the most stable VPBP in a recognition algorithm for a back vein pattern of a hand applied to an embodiment of the present invention. FIG. 5 is a polygon feature vector diagram obtained by connecting basic VPBPs extracted in an embodiment of the present invention. FIG. 10 is a diagram illustrating a branch characteristic in VPBP using the optimum BPWF obtained in the embodiment of the present invention. It is a figure in the case of the user who grips the stick | stick in the personal identification system using the vein pattern of the back of the hand of a prior art, putting power into a hand. It is a figure in case a user rotates a hand up and down. 5 is a flowchart of a conventional algorithm for extracting a vein pattern on the back of a hand in a personal identification method using the vein pattern on the back of the hand. When the vein pattern extraction algorithm on the back of the hand in FIG. 32 is executed, (a) a vein pattern enhancement filtering image of the ROI image, (b) a binarized and normalized image, and (c) a non-projection image. It is a data structure figure in the authentication information database by the Example of this invention.

(Back authentication system configuration)
As shown in FIG. 1, the back authentication system 100 using the vein pattern of the present invention includes a back authentication terminal device 20 and a back authentication server 30 connected to the network via the communication control unit 27 of the back authentication terminal device 20. Composed. An external device 80 such as an electric lock is connected and controlled via the external device interface 60 provided in the back authentication server 30.

  The back authentication terminal device 20 includes a personal identification code input unit 22 connected to the communication control unit 27 for a user to input a personal identification code, and a communication for displaying a determination result of the input personal identification code. Display means 21 connected to the control unit 27, a micro switch 23 for starting photographing, a finger photographing camera 25 connected to the micro switch 23 for photographing the user's finger image Fi, and connected to the micro switch 23 The back of the hand is photographed with the back image camera 24 for photographing the original image Hi of the user's back, and is connected to the finger photographing camera 25 and the back of the hand photographing camera 24. The photographed finger image Fi and the original image Hi of the back of the hand A video decoder 26 is provided that converts the back of the hand into a digital original image Hdi and outputs it via the communication control unit 27.

  The back of the hand authentication server 30 is connected to the processor 40 that performs the authentication process, the authentication information database 50 that stores the reference image used for authentication and the vein pattern optimum distribution characteristic information for each personal identification code, and the processor 40. A temporary storage memory 70 that is connected and temporarily stores information during authentication processing and an external device interface 60 that is connected to the processor 40 and outputs an authentication result to the external device 80 are provided.

  In more detail, the processor 40 is connected to the authentication information database 50 and has a database access unit 42 for accessing information in the authentication information database 50 and a reference corresponding to the personal identification code input from the personal identification code input means 22. The presence / absence of the image and vein pattern optimum distribution characteristic information is determined, and the personal identification code determination unit 41 that outputs the determination result to the display means 21; the digital finger image Fdi input from the video decoder 26 and the digital original image Hdi of the back of the hand The captured image storage unit 43 stored in the temporary storage memory 70, the ROI image extraction unit 44 that extracts the ROI image Hri from the digital original image Hdi, the position information to be corrected from the digital finger image Hdi is calculated, and the extracted ROI image Position correction for correcting the position of Hri and creating a vein pattern corrected image mVPi And parts 45, including the vein pattern matching section 46 for authenticating by comparing the vein pattern corrected image mVPi the reference image stored in the authentication information database 50 and the vein pattern optimal distribution characteristic information.

  As an embodiment of the back of the hand authentication terminal device 20, an external view and a side sectional view of the back of the hand authentication terminal device 20 are shown in FIGS. As the personal identification code input means 22, the keypad 3 for inputting a personal identification code or a number necessary for the progress of the system, and as the display means 21, the personal identification code determination result and the progress status of the system are written. A liquid crystal screen (LCD) 2 to be displayed is installed on the upper surface. Furthermore, you may install the light emitting diode (LED) 1 for displaying the operating condition of a system with a lamp | ramp.

  In the lower part, a button switch 4, a cover 5 for blocking external illumination, and a stable handle 6 as a support base for supporting a finger when inputting an image relating to the back of the user's hand are installed as a micro switch 23. The stable handle 6 is fixed by two shafts, and a finger insertion portion 11 for inserting a middle finger is formed between the two shafts. This makes it possible to shoot with the back of the hand stabilized.

  Further, a CCD or CMOS camera 7 that reacts to the visible light and infrared bands as the finger photographing camera 25 and the back of the hand photographing camera 24 is located at a certain distance from the cover 5 in order to take an image of the back of the user's hand. The infrared illumination 8 is installed as an illumination for emphasizing the vein pattern on both sides of the camera 7, and the infrared filter 9 is installed at the lower stage of the infrared illumination 8 and the camera 7.

  In particular, a housing 10 is provided between the cover 5 and the camera 7 to secure a focal length for capturing a vein pattern ROI image Hri. The inner surface of the housing 10 is coated with a substance that absorbs infrared rays. Therefore, the influence of external light can be minimized.

  As the personal identification code input means 22, a keypad 3 for inputting the above personal identification code, a card reader for reading information from a key card in which the personal identification code is stored, or the like is used.

  As the display means 21, in addition to the LCD 2 for displaying the result as characters, an LED or the like for displaying the result by the color or blinking of the lamp is used.

  As the micro switch 23, in addition to the push button type button switch 4, a human sensor switch for detecting that the hand is held up is used.

  FIG. 34 shows an example of the structure of data stored in the authentication information database 50. The data is registered for each personal identification code and can be searched using the personal identification code as a search key. The data includes a reference image of a vein pattern that has been captured in advance, and vein pattern branching characteristics (VPBC: Vein Pattern Branch Characteristic) and branch weights (BPWF: Branch Point Weighting Factor) as vein pattern optimum distribution characteristic data. ) Is included. In order to improve the efficient use of memory and the retrieval efficiency of data, database normalization is appropriately performed when storing data. Normalization is normally performed in a system using a database, and is not specified in the back authentication method described below.

  In order to register data in the authentication information database 50, it is natural that an input device such as a keyboard is connected to the back authentication server 30 and is omitted in the drawing.

(Back of hand authentication method)
Next, the back of hand authentication method according to the embodiment of the present invention will be described in detail with reference to the flowcharts shown in FIGS.

  First, the user inputs a personal identification code by the personal identification code input means 22 (S100). The input personal identification code is transmitted to the personal identification code determination unit 41 via the communication control unit 27. The personal identification code determination unit 41 that has received the personal identification code searches the information stored in the authentication information database 50 using the personal identification code as a search key via the database access unit 42 and determines whether or not it has been registered. (S101). The determination result is output to the display means 21, and if not registered, an error is displayed (S107). If registration is confirmed, the process proceeds to image shooting.

  The user who has confirmed registration inserts the middle finger into the finger insertion portion 11 of the stable handle 6 (S102). The finger photographing camera 25 photographs the state of the finger placed on the stable handle 6 as a finger image (Fi) (S103).

  Further, the user fixes the state without moving the hand on the stable handle 6 for a while (S104). In this state, the back of the hand photographing camera 24 photographs the back of the hand to obtain an original image (Hi) (S105).

  The photographed finger image (Fi) and the original back image (Hi) are sent to the video decoder 26 and digitized. The digitized images are stored in the temporary storage memory 70 as digital finger images (Fdi) and digital original images (Hdi) by the photographed image storage unit 43 of the processor 40 via the communication control unit 27 (S106). .

  Subsequently, a region of interest (ROI) in which a vein pattern necessary for authentication is captured is identified from the digital original image (Hdi) of the back of the hand, and only the ROI image (Hri) portion is extracted (S200). .

  In order to extract a ROI image (Hri) in which vein patterns are concentratedly distributed from a digital original image (Hdi) on the back of the user's hand, an accurate ROI position is required. By the way, due to an error in the lens and camera module mounting position, the ROI position appears slightly different for each camera screen. Therefore, an accurate ROI position is determined by an ROI automatic detection algorithm described below.

  First, in the ROI image (Hri) extraction process, all the pixels constituting the digital back image (Hdi) of the back of the hand take the lightness value (0 to 255) as the lightness level on the horizontal axis and the lightness value on the vertical axis. A histogram taking the number of pixels is created (S201).

  Next, using the P-tile method, a threshold value (T) of a gray value serving as a boundary between the ROI region and the background is calculated from the histogram (S202). Due to the structure of the back of the hand authentication terminal device 20, the ROI region in which the necessary vein pattern is photographed may be shifted in every photographing, but the area occupied in the whole photographed digital original image (Hdi) The ratio is constant. Therefore, when the ratio of the area occupied by the object in the image is known, a P-tile method is used in which the P% point of the area ratio is the threshold value (T) from the histogram. In addition to the P-tile method, it is also possible to use a mode method that detects valleys in the histogram.

  A ROI region is determined by extracting pixels having a grayscale value equal to or greater than the threshold value (T) (S203). Further, in order to emphasize the edge of the ROI region, the Sobel filter processing is performed to extract the ROI image (Hri) (S204).

  7 is a digital original image (Hdi), FIG. 8 is a histogram to which the P-tile method is applied, FIG. 9 is an ROI region determined by a threshold value (T) obtained from the histogram of FIG. 8, and FIG. 9 shows ROI images (Hri) separated from the digital original image (Hdi) by 9 respectively.

  Next, in order to read a vein pattern from the extracted ROI image (Hri), a filter-bank vein pattern extraction algorithm (FBHVPEA: Filter-bank Hand-pattern Extraction Algorithm) process is performed (S300). This process minimizes partial loss of vein pattern information from the contraction and expansion of vein patterns caused by changes in the external environment, so that pattern connectivity, an important feature of vein patterns, can be maintained to the maximum. .

  First, in the filter-bank vein pattern extraction algorithm, a thick vein pattern (SPi) is effectively extracted from the vein patterns distributed on the back of the hand using a normal mode filter (S301). Further, thin veins and a narrow vein pattern (ZPi) in a contracted state are effectively extracted by an efficiency mode filter (S302). The thick vein pattern (SPi) and thin vein pattern (ZPi) obtained in the previous step are synthesized to create a vein pattern synthetic image (GPi) of the back of the hand (S303). The normal mode filter maintains the low-band information and pattern boundary information in the vein pattern, and the efficiency mode filter has a relatively low low-band information and a low frequency at the pattern boundary compared to a thick vein. Is effectively emphasized.

  Subsequently, the binarization processing of the vein pattern composite image (GPi) is performed and converted into a binary image (BPi) (S304).

  At this time, in the binarization process used to obtain a binary image (BPi), two methods are considered as a method for determining a threshold value for dividing the pixel into either black or white. can do. One is a method of using the average value of the local mask as a threshold value, and the other is a method of determining the threshold value by a value of a 7-bit plane. According to the result of trying to apply the two methods, more accurate vein pattern information can be obtained when the value of the 7-bit plane is used as a threshold value. It is desirable that the size of the spatial mask used for the binarization process is designed with an 11 × 11 kernel.

  Finally, the binary image (BPi) obtained by the binarization process includes miscellaneous projections caused by the back of the user's hand, bending, the subcutaneous fat layer, and the like. In order to remove this, a median filter of 3 × 3 kernels is applied as noise removal filter processing to create a vein pattern image (VPi) on the back of the hand (S305).

  As described above, the filter-bank vein pattern extraction algorithm is a portion generated by extracting a vein pattern with the characteristics of a single filter without considering the thickness of the vein pattern to be extracted from the existing vein pattern extraction algorithm. The loss of typical vein-pattern connectivity information is improved, and more stable algorithm performance is provided against the contraction and expansion phenomenon of vein patterns caused by changes in the external environment.

  FIG. 11 shows an image of the processing result of the vein pattern enhanced by the filter-bank vein pattern extraction algorithm. FIG. 11A shows a combination of vein patterns obtained by the normal mode filter and the efficiency mode filter. The vein pattern composite image (GPi), (b) is a binary image (BPi) converted into a black and white image by binarization processing, and (c) is a vein pattern image (VPi) from which noise has been removed.

  FIG. 12 shows a comparison between (a) a vein pattern obtained by the conventional enhancement filter processing and (b) a vein pattern obtained by the normal mode filter processing of the vein pattern extraction algorithm of the filter bank. is there. When comparing vein patterns extracted for a partial region of the ROI region, in the vein pattern processing result for a thick vein extracted by the existing vein pattern enhancement filter, information in the vein pattern is lost, Although the pattern connectivity of the vein pattern is partially lost, the processing result by the normal mode filter of the present invention effectively corrects the low-band information and maintains the pattern connectivity of the vein pattern. I understand that.

  However, in the normal mode filter, there is a partial loss of pattern connectivity in the extraction of vein patterns for veins that have shrunk or thin and thin as the external temperature decreases. To correct this, the filter-bank vein pattern extraction algorithm uses an efficiency mode filter. The efficiency mode filter corrects a partial loss of pattern connectivity for a thin and thin vein pattern by reducing a low-band gain corresponding to the background and emphasizing a fine pattern change.

  FIG. 13 shows a comparison between (a) a vein pattern obtained by a conventional extraction algorithm and (b) a vein pattern obtained by a filter-bank vein pattern extraction algorithm. 13A shows a partial loss of pattern connectivity of the vein pattern to be extracted. In FIG. 13B, the partially lost pattern connectivity is also reliably extracted.

  Next, when the user places his / her finger on the stable handle 6, the digital finger image (Fdi) obtained from the finger image (Fi) photographed by the finger photographing camera 25 is analyzed, and the process proceeds to a process of correcting the position ( S400).

  First, a rotation correction value (R) and a movement correction value (S) are calculated from the digital finger image (Fdi) by a position correction algorithm described in detail below (S401). A vein pattern correction image (mVPi) in which the vein pattern of the vein pattern image (VPi) is corrected to a position that can be compared with a reference image of a vein pattern registered in advance by the rotation correction value (R) and the movement correction value (S). ) Is created (S402).

  FIG. 14 is a conceptual diagram of an image photographed for position correction applied to the embodiment of the present invention, and FIG. 15 is an image diagram showing a state until an image is photographed.

  As shown in FIG. 15, when the hand is placed so that the middle finger is inserted into the middle finger insertion portion 11 of the stable handle 6, the index finger, middle finger, and ring finger are photographed. From the photographed finger image (Fi), the direction in which the hand is located is detected using the right and left of the stable handle 6, the intersection of the index finger and the middle finger, and the intersection of the middle finger and the ring finger.

  As shown in FIG. 16, in the position correction algorithm, first, a boundary line between the stable handle 6 and the finger is extracted (S51).

  In the boundary line extraction process, the coordinates of the six pixels necessary for the prediction of the rotation angle are obtained, but it is a low-band filter (Low pass Filter) configured by a determinant of 3 × 3 size. Apply a smoothing filter (Averaging Filter) to the original image. The smoothing filter is the most typical filter among the low-band filters. The low-band filter makes the image appear totally unclear, but when the image is moved to the frequency band, it serves to eliminate a sudden change in value and a particularly conspicuous portion from surrounding pixels which are image portions in the high-frequency band. An advantage of using the smoothing filter is that it is possible to reduce miscellaneous projections such as spot noise and salt and pepper noise during image processing.

  FIG. 17 shows an application example of the smoothing filter applied to the embodiment of the present invention. In the smoothing filter, all matrix values are 1. The smoothing filter of this 3 × 3 matrix is moved from left to right in the original image, and again from top to bottom, one pixel at a time with reference to the center pixel, and the following calculation is repeated. After the pixel values of the overlapping images are multiplied by the matrix values at the same position of the smoothing filter, the total is obtained and averaged to obtain an average value, and the average value is set against the central pixel value. In FIG. 17, the average value enters the position where the pixel value of the original image is 3.

The smoothing filter application method described above is performed in the same manner as the image processing method using a mask applied in the embodiment of the present invention.

  In this way, the optimum edge extraction mask obtained in the experiment is applied to the image obtained by applying the low-band filter, and the boundary line between the stable handle 6 and the finger is extracted. There are various masks for extracting edges, such as a Sobel mask, a Kirsh mask, and an edge mask, but edge masks are generally used in general. However, when extracting an edge, the edge of the edge is not constant, such as other objects appearing in addition to the edge to be obtained. Can not ask for.

  FIG. 18 shows the result of image processing using a mask applied to the embodiment of the present invention. When the boundary lines extracted by applying various masks to the digital finger image (Fdi) of FIG. 18A are compared with each other, the Sobel mask processed image of FIG. 18B and the image of FIG. On the boundary line shown in the Kirsh mask processed image, unnecessary contour lines surrounded by the four sides are also extracted, and traces of miscellaneous projections are seen at several places in the image. In the edge mask processed image of FIG. 18 (d), only the boundary line between the stable handdle 6 and the finger is accurately extracted, and no miscellaneous image is seen.

  In the position correction algorithm, by extracting only the boundary line of the exact position, the next image processing process becomes easier, and the overall performance of the algorithm is improved by reducing the screening and noise removal process. Can do.

  Subsequently, an outer angle line is extracted (S52), an angle between fingers is extracted (S53), and a rotation center point that is the center when rotating the image is obtained for rotation correction (S54), and then the boundary The coordinates of the pixels and the like obtained from the extraction of the lines and outer corner lines and the coordinates of the pixels selected by the rotation center point are stored as integer position information (S55).

  The total number of pixels to be obtained by the position correction algorithm is six, and all the pixels are located on the boundary line, and therefore the rotation correction angle is predicted using the coordinates of the pixels. Also, one of the six pixels is designated and used as the rotation center point of the vein pattern on the back of the hand during rotation correction.

The boundary line between the stable handle 6 and the finger is formed with a thick line, and in order to obtain the coordinates of the pixel, a thinning process and a trace of the outer corner line are necessary. The thinning process normally uses an image processing technique called skeleton, which means that only the center point of characters and objects is left and a single line is made.

  In the position correction algorithm, the upper curved portion (Bay) of the extracted boundary line coincides with the outermost angle of the finger, so that the outer angle of the curved portion on the upper side of the boundary line is relatively smaller than the thinning process. Apply the trace of the outer corner line of the border line that was traced. Further, regarding the coordinates of the pixel, the upper left side of the image with respect to the extracted boundary line is regarded as the value of the coordinate system, and the coordinate value is applied.

FIG. 19 is a position correction algorithm applied to the embodiment of the present invention, and illustrates a process for extracting a corner between fingers. The boundary line by the edge mask process shown in FIG. The extracted image is inverted.

  As shown in FIG. 19, the lowermost pixel (C) and the uppermost pixel (A) and pixel (B) on the left and right sides are obtained from the extracted outer corner line. The number of pixels to be extracted from one boundary line is 3, and two boundary lines of the intersection of the index finger and the middle finger and the intersection of the middle finger and the ring finger are photographed in one digital finger image (Fdi). Will be 6 in total. The position correction algorithm uses the angle between the fingers to predict the rotation angle of the rotation correction, and the angle between the fingers is obtained using three pixels obtained for each boundary line.

In FIG. 19, three points (A, B, C) are three pixels obtained by tracing the outer corner lines of the boundary line. Of these pixels, when a virtual vertical line is drawn around the lowest pixel (C), the line segment connecting the highest pixel (A), the pixel (B) and the lowest pixel (C). And form angles of θ _l and θ _r , respectively. The size of the angle is obtained by substituting into the following formula 1. In formula 1, a _k is the distance in the x-axis direction between two pixels, and b _k forms an angle in the y-axis direction. This is the pixel distance difference. The two angles obtained by Equation 1 are combined to finally obtain the angles θ ₁ and θ ₂ between the fingers.

  Next, one of the two lowest pixels obtained from the boundary line is selected, and the rotation center point is obtained. For the pixel to be the rotation center point, it is desirable to select the lowest pixel extracted from the boundary line between the fingers slightly away from the stable handle 6 in order to reduce the error of the rotation center point to the maximum. This is because the shape of the hand is easily deformed when coming into contact with an object due to the characteristics of the hand, and in particular, the curved portion between the fingers hitting the stable handle 6 has a shape of a naturally formed curved portion. Because it is different.

  The position correction algorithm predicts the rotation correction angle using the pixel coordinates obtained from the finger area after storing the position information as integer data. At that time, rotation correction angles are divided into primary and secondary, and all are obtained twice, and rotation correction is performed using the rotation correction angles thus obtained.

  FIG. 20 shows the process of predicting the angle of the primary rotation correction applied to the embodiment of the present invention. The primary rotation correction is a process for obtaining an angle that can correct the entire rotation of the back of the hand. The rotation correction angle is obtained by calculating the coordinates of the four lowest pixels obtained from the registered finger image and the comparison finger image. Use to find the angle.

  As shown in FIG. 20, pixel (A) and pixel (B) are the lowest-order pixels obtained from registered finger images, and pixel (A ′) and pixel (B ′) are taken for comparison. This is the lowest pixel obtained from the finger image.

  Since the coordinate values of the pixels (A, B, A ′, B ′) are stored in the form of integer data, the data can be called up and obtained simply. Of the four pixels, two pixels obtained from the same finger are connected to each other, and arbitrary line segments a and b are obtained.

  Here, the two line segments obtained from the finger image taken for comparison with the registered finger image must form an intersection with each other so that an angle can be formed between the two line segments. At that time, the intersection is determined by selecting a pixel located on the same side as the rotation center point of the line segment of the finger image taken for comparison with reference to the rotation center point determined in the registered finger image. To move to. In FIG. 20, the pixel (B) is the rotation center point in the registered image, and the pixel located on the same side as the pixel (B) among the line segments of the finger image photographed for comparison is the pixel ( B ′).

Pixels other than the pixel used as the intersection of the finger images taken for comparison are moved by applying the same distance by the distance moved in order to create an intersection. In this manner, the angle at which the remaining line segments are relatively inclined with respect to one line segment as in θ ₃ is obtained.

Here, θ ₃ indicates an angle at which the image of the back of the hand taken for comparison is relatively inclined as compared with the registered image of the back of the hand (reference image). That, theta ₃ is an angle of rotation correction, minutes of the inclined angle, the image is rotated back of the hand to be captured for comparison, to calibrate the relative tilt.

  In the secondary rotation correction, while the curved part between the fingers is in strong contact with the stable handle 6, there is a change such as a change in the angle between the fingers due to the rotation of the back of the hand itself or the fine movement of the back of the hand. Even if it is not, it will be distorted as a whole, so that the angle between the fingers is corrected in order to obtain the rotation correction angle.

FIG. 21 illustrates a process of predicting the angle of the secondary rotation correction applied to the embodiment of the present invention, and shows an angle between fingers by pixels obtained from the extracted boundary line. . θ ₁ is an angle between fingers obtained from a registered image, and θ ₁ ′ is an angle between fingers of a finger image taken for comparison. The two angles (θ ₁ , θ ₁ ′) are images different from each other, but are angles obtained between fingers at the same position. Therefore, for comparison with the angles between fingers obtained from registered images. The angle difference between the fingers of the captured finger image is obtained, and the angle for secondary rotation correction is obtained.

  In FIG. 21, pixel (L) and pixel (L ′) are the top leftmost pixels extracted from the boundary line, and pixel (R) and pixel (R ′) are the top rightmost pixels on the boundary line, Pixel (B) and pixel (B ′) become the lowest pixel of the boundary line. A dotted line that divides the angle between the fingers is a center line of the angle between the fingers.

  With reference to a virtual straight line passing through the pixel (B) and the pixel (B ′), an angle inclined with respect to the center line and the intersection of the pixel (B) and the pixel (B ′) is obtained. When this center line is in the left region with respect to the virtual straight line, it has a positive value, and conversely in the right region, it shows a negative value.

The angle between the center lines of θ ₁ and θ ₁ ′ obtained in this way is inclined, and the image is taken for comparison with reference to the center line of θ ₁ , which is the angle between the fingers obtained from the registered image. The relative inclination angle (θ ₁ −θ ₁ ′) is obtained by subtracting the angle of inclination of the center line of the angle θ ₁ ′ between the obtained fingers from the obtained image.

  By the above method, the relative inclination angle of the angle of the image to be photographed for comparison with the angle of the registered image between the remaining fingers is obtained. The average of the two relative inclination angles obtained in this way is calculated and finally set as the secondary rotation correction angle.

  As the last step, authentication is performed by comparing with authentication information registered in the authentication information database 50 by a probabilistic matching algorithm (PMA: Probability Matching Algorithm) (S500).

  First, the vein pattern optimum distribution characteristic data is read from the authentication information database 50 (S501). The vein pattern optimum distribution characteristic data includes a vein pattern branch characteristic (VPBC) and a branch weight weighting factor (BPWF).

  Subsequently, a polygon having a vertex at each branch point of the vein pattern is created from the vein pattern correction image (mVPi) (S502). Then, the polygon created from the registered vein pattern optimum distribution characteristic is compared with the vertex, and the consistency is determined (S502).

  As a result of matching, it is determined whether or not the comparison value obtained satisfies a reference value that can be authenticated if it matches (S504). If not, authentication is not possible (S505). The license is granted (S506).

(Back of hand vein pattern recognition algorithm)
Here, the vein pattern recognition algorithm on the back of the hand used in the present invention will be described in detail.

  The branch characteristics of the vein pattern on the back of the hand include a branch point, a branch angle, and a pattern length at the branch point. An individual is identified using the branching feature of the vein pattern thus extracted. The most basic branching feature of vein pattern branching characteristics is the branching point of vein patterns, and the existing vein pattern extraction algorithms partially lose the pattern connectivity of vein patterns. In operation, it reacts extremely sensitively to changes in the external environment and changes in the input image. In other words, the existing vein pattern extraction algorithm recognizes all the branch characteristics of the back of the hand vein pattern by applying a collective weighting factor (Uniform Weighting Factor), so the stability as a substantial biological recognition system is greatly increased. There is a disadvantage that it is reduced.

  For example, the partial loss of the branch point of the vein pattern to be extracted due to the change of the input image, the change of the external environment, etc. that may occur when the user repeatedly uses the system mainly decreases the performance of the algorithm. It becomes a factor. This is because the branch feature of the vein pattern of the back of the hand to be extracted is not constant and is fixed, so that the loss of the branch point of the vein pattern acts as a major factor that affects the recognition performance.

  FIG. 22 illustrates branching characteristics of the same person's vein pattern obtained by the prior art vein pattern extraction algorithm. (A) and (b) show the image processing of the same person input with a time difference. The results are illustrated. In (a), the number of branch points appears, but in (b), 14 branch points appear. When such a branch point is applied as a branch feature, the person is erroneously recognized as a different person even though the person is the same person at the recognition stage.

  FIG. 23 illustrates the image processing result of the position correction algorithm of the prior art, and (a) and (b) illustrate the image processing result of the same person input with a time difference. The method of correcting the position change due to the rotation or movement of the user corrects the position using the positions of two branch points (the upper left branch point and the lower right lower branch point). Due to the partial loss of the branch point, which is a unique branch feature, the wrong result is shown.

  The back-of-hand vein pattern recognition algorithm according to an embodiment of the present invention applies probabilistic reliability to the bifurcation characteristics of vein patterns distributed by individuals, and provides the most stable VPBP (Vein-Pattern Branch-Point) for each individual. Extraction improves the stability of the system's recognition performance due to rotation and movement caused by slight changes in the position of the user's hand and the sensitivity of the system caused by changes in the external environment, slight changes in the input image, etc. it can.

  FIG. 24 shows the recognition algorithm of the back vein pattern of the hand applied to the embodiment of the present invention, and FIG. 25 shows the detailed process of the training stage of FIG. 24 in order.

  As illustrated in FIG. 24, the vein pattern recognition algorithm on the back of the hand applies a probabilistic confidence level for the bifurcation characteristics to the individual bifurcation characteristics and a probabilistic confidence level obtained in the training stage. It consists of a main stage that automatically converts individual registration information into branching characteristics and applies a feature factor comparison algorithm (FVMP).

  In the training stage, the vein pattern image extracted by the vein pattern extraction algorithm is used, and whether or not the user is available is determined through a direct comparison algorithm (DMA). The DMA determines the direct similarity (Correlation) between the reference image registered in the authentication information database 50 and the vein pattern image based on the branch characteristic, and determines whether or not individual identification is possible. When the person is determined to be the person, the probabilistic reliability of the bifurcation point is determined (PMA), and the user repeatedly uses the system. A branch characteristic (VPBC) of the pattern is detected, and a probabilistic branch weight (BPWF) is determined for each branch characteristic.

  In the main stage, the individual registration information is automatically converted with the branch characteristic to which the probabilistic reliability obtained in the training stage is applied, and FVMA is applied to identify whether or not the user is authorized. In the recognition process, the FVMA determines whether the user is authorized through the polygonal characteristics and the probabilistic bifurcation characteristics.

  Using individual bifurcation characteristics to which probabilistic reliability is applied, stable recognition performance can be shown, and correction for user rotation and movement is corrected based on the most stable VPBP, and system performance In addition, it has the advantage that the convenience of the user can be improved.

  Therefore, when the vein pattern distribution system on the back of each person's hand is used repeatedly, by extracting the optimum branch characteristics for each individual from the changes in the input image due to the individual's external environment and repeated use of the system while applying the branch characteristics. , Show more stable system performance.

  As shown in FIG. 25, the user registration information is stored in the initial training stage using the system by storing the original vein pattern extracted from the user and recognized by the DMA. Is repeatedly converted to a stochastic matching algorithm (Probability Matching Algorithm), and at the same time, the optimal vein pattern bifurcation characteristics for each user are extracted and used to extract the optimal BPWF (S61). ~ S64). As a result, only the distribution characteristics of the vein pattern extracted from the user are stored in the user registration information (S65).

  FIG. 26 shows a process of correcting the position by applying the optimum branch characteristic extracted by the recognition algorithm for the back of the hand vein pattern applied to the embodiment of the present invention.

  In order to extract a branch point, a thinning algorithm is applied to the region including the branch point thus extracted.

  A slight change in the position of the branch point to be extracted from the process of applying the above thinning algorithm has an allowable range that can occur due to a slight change in the thickness of the vein pattern in the input original image of the same person. As shown, the branch point extracted within the allowable range determines the position of the branch point at the center position of the allowable range.

  FIG. 26A shows the optimal branch characteristics stored in the database, and branch points that are not connected by a line exhibit a low probability of reliability. In other words, a low weighting factor (Low Weighting Factor) is used at a branch point that exhibits a branch characteristic that has a high probability of losing the characteristic of the branch point due to a slight change in the input image that occurs due to changes in the external environment or the repeated use of the system. It comes to apply. On the other hand, a branch point connected by a line shows a high probability of reliability. That is, it means a stable VPBP that shows high reliability with respect to a change in the external environment or a slight change in the input image, and recognizes it by applying a high weighting factor (High Weighting Factor).

  FIG. 26 (b) shows the result obtained by rotation and movement of the user's hand, and FIG. 26 (c) shows the branch feature of the database of FIG. 26 (a) and the rotation with respect to FIG. 26 (b). The correction and movement correction methods are shown. In the vein pattern recognition algorithm on the back of the hand, the most stable VPBP is connected to form a polygon, and the polygons are positioned so as to be the best match, thereby correcting for rotation and movement. As a result, the probabilistic reliability extracted is applied to each bifurcation characteristic for the individual bifurcation characteristics obtained by the recognition algorithm of the back of the hand, and the most stable VPBP and the optimum BPWF are applied. By applying it to the recognition stage, stable recognition performance can be obtained.

  FIG. 27 shows a process of correcting rotation and movement using the most stable VPBP in the recognition algorithm of the back of the hand applied to the embodiment of the present invention. (A) and (b) show the extracted VPBP, and (c) shows the result when the most stable VPBP selected in (a) and (b) is overlaid.

  At this time, when FIG. 29A is regarded as the branching characteristic of the database to be stored, as shown in FIG. 29D, the polygon in which the most stable VPBP of FIG. By optimally applying the bifurcation characteristics, the rotation and movement can be effectively corrected as shown in FIG.

  The back-of-hand vein pattern recognition algorithm finally determines whether or not the user can be identified through FVMA at the recognition stage in which personal identification is performed by a probabilistic matching algorithm (PMA).

  FIG. 28 shows a polygon feature vector (Feature Vector) obtained by connecting the basic VPBPs extracted in the embodiment of the present invention. FIG. 29 uses the optimum BPWF obtained in the embodiment of the present invention. The branching characteristics of the VPBP are shown.

  As shown in FIG. 28, in the recognition stage, first, a polygon created by connecting the most stable VPBPs in the database and a VPBP obtained from the currently input image are connected. When matching the vertices of the polygon, it is recognized by judging the matching with respect to each vertex, the matching with the angle at the vertex of the polygon, the length between the vertices, and the matching degree with respect to the area of the polygon.

  Next, as shown in FIG. 29, in the recognition stage, the optimal BPWF for each branch characteristic is applied through probabilistic reliability, and the branch characteristic (branch point, branch angle at the branch point, branch point) is applied. The number of branches at the branching point and the length at the branching point) are judged and recognized.

  The database for vein recognition stores the branch characteristics and reference images related to the vein pattern of the user while continuously updating the data related to the branch characteristics of the vein pattern for each unique registration number assigned to each user. Yes.

  The above drawings and detailed description of the invention are merely exemplary of the invention, which is merely used to illustrate the invention and is described in the meaning limitation and in the claims. It is not intended to be used with the intention of limiting the scope of the invention. Accordingly, those having ordinary skill in the art can understand that various modifications and other equivalent embodiments can be made therefrom. Accordingly, the true technical protection scope of the present invention should be determined from the technical concept of the appended claims.

DESCRIPTION OF SYMBOLS 1 Light emitting diode 2 Liquid crystal screen 3 Keypad 4 Button switch 5 Cover 6 Stable handle 7 Camera 8 Infrared illumination 9 Infrared filter 10 Housing 11 Finger insertion part 20 Back authentication terminal device 21 Display means 22 Personal identification code input means 23 Micro switch 24 Back of hand Shooting camera 25 Finger shooting camera 26 Video decoder 27 Communication control unit 30 Back of hand authentication server 40 Processor 41 Personal identification code determination unit 42 Database access unit 43 Captured image storage unit 44 ROI image extraction unit 45 Position correction unit 46 Vein pattern matching unit 50 Authentication Information database 60 External device interface 70 Temporary storage memory 80 External device 100 Back of hand authentication system

Claims (12)

An authentication terminal for the back of a box-shaped hand having a housing in which an operation part is formed obliquely on the front side and a cavity is formed obliquely inside,
The housing is
An opening is formed obliquely in the lower part, and an infrared illuminator on the inner surface of the upper part and an imaging means that reacts to visible light and infrared bands, and a substance that absorbs infrared rays is applied to the inner surface,
further,
An infrared filter provided below the imaging means and the infrared illuminator;
With
Furthermore, a stable handle is provided at the lower part on the side that becomes the wall side of the opening,
The stable handle is
A horizontal bar,
Two shafts are provided on the horizontal bar with a width allowing the middle finger to be inserted, the two shafts are curved to form a housing under the opening, and a central portion of the two shafts serves as the imaging means. An authentication terminal on the back of the hand, which is provided so that it can be photographed. The authentication terminal for the back of the hand according to claim 1, wherein the right and left corners of the upper side of the opening are provided with a cover up to the vicinity of the center of the upper side. 3. The authentication terminal for the back of the hand according to claim 1, wherein a micro switch that is pushed by the back of the hand is provided between the left cover and the right cover . The imaging means includes
  4. The back authentication terminal according to claim 1, wherein imaging is started by an operation of the microswitch. Finger obtained by photographing the connection area between the middle finger, index finger, ring finger and back of the hand by inserting the middle finger between the two shafts into a stable handle with two shafts at the center of the horizontal shaft A digital back image authentication method for authenticating a vein of the back of the hand by reading a digital image of the back and a back digital image obtained by photographing the back of the hand ,
Computer
Optimal distribution characteristic data (VPBC) including reference polygons connecting the branch points of the veins of the back of the hand obtained by the certifier inserting a middle finger into the two axes of the stable handle and photographing in advance, and the certifier A first reference boundary line, which is an edge portion of a first curved region where the middle finger and the ring finger are in contact with each other, and the middle finger and the index finger are in contact with each other, which is obtained by inserting the middle finger into the two axes of the stable handle and shooting in advance. Storing a second reference boundary line that is an edge portion of the second curved region in the authentication information database as finger boundary reference information ;
Storing the finger digital image and the back digital image from the back authentication terminal of the hand in a temporary storage memory;
From the upper digital image stored in the temporary storage memory, a vein pattern portion necessary for authentication is extracted as ROI image data, and a thick vein pattern (Spi) and a thin vein pattern (ZPi) are extracted and synthesized. And ROI image extraction step of generating a vein pattern (Vpi) from which the vein composite pattern (Gpi) is binarized and noise is removed,
Furthermore, a position correction step and a vein pattern matching step are performed,
The position correction step includes
Extracting the first curved region and the second curved region as finger boundary images in the finger digital image ;
A skeleton process is performed on the first curved region of the finger boundary image, and the first boundary line that is the edge portion between the middle finger side and the ring finger side and the edge portion that is the middle finger and the index finger are Obtaining a boundary line of 2 ;
A first straight line connecting an intersection point included in the first boundary line and an intersection point included in the second boundary line; a reference intersection point included in the first reference boundary line; and a second line Obtaining a first relative angle, which is a relative angle with a second straight line connecting the reference intersection point included in the reference boundary line, and setting this as a first correction angle;
The relative angle between the opening degree of the first boundary line included in the finger boundary image and the reference opening degree of the first reference boundary line included in the finger boundary reference information is set as a second relative angle. And calculating the relative angle between the opening degree of the second boundary line included in the finger boundary image and the reference opening degree of the second reference boundary line included in the finger boundary reference information. Obtaining an angle as a third relative angle, and setting an average of the second relative angle and the third relative angle as a second correction angle;
Correcting the vein pattern (Vpi) at the first correction angle, and further generating a vein pattern corrected image (mVpi) corrected at the second correction angle;
And
The vein pattern matching step includes:
Find each branch point of the vein pattern corrected image (mVpi), generate a polygon connecting these branch points, and compare the polygon with the reference polygon of the optimum distribution characteristic data (VPBC). Means for determining presence or absence;
Displaying the determination result on a display of the back authentication terminal of the hand;
A method of authenticating the back of the hand, characterized by The optimum distribution characteristic data (VPBC) includes the branch angle of the upper vein pattern from the branch point, the number of branches, and the pixel coordinates of the branch point,
The vein pattern matching step includes:
6. The back of hand authentication method according to claim 5, wherein the pixel coordinates, the number of branch points, and the branch angle of the branch point of the vein pattern of the vein pattern correction image (mVpi) are obtained, and whether or not they match is determined . 7. The back of hand authentication method according to claim 5, wherein the vein pattern of the optimum distribution characteristic data (VPBC) includes a thick vein pattern and a thin vein pattern . 8. The back authentication method according to claim 5, wherein the optimum distribution characteristic data (VPBC) is stored in association with an authentication code . 9. The ROI image is extracted using a P-tile method for calculating a threshold value from a ratio of an area occupied by the ROI in the original image of the back of the hand. Authentication method for back of hand. The first boundary line of the finger boundary image, the opening of the second boundary line,
a _k is the distance in the x axis direction between the two pixels, b _k is the combined two theta _k pointing to the following equation which is the distance difference between pixels forming an angle in the y-axis direction 10. The method for authenticating the back of the hand according to any one of claims 5 to 9 .
The optimum distribution characteristic data (VPBC) of the authentication information database is:
Including the position of each vertex of the reference polygon, the branch angle of the straight line from each vertex, the length between the vertices, the area of the polygon,
The vein pattern matching step includes:
The back of the hand according to any one of claims 5 to 10, wherein a position of each vertex, a branch angle of a straight line from each vertex, a length between the vertices, and a matching degree with respect to a polygonal area are determined and recognized. Authentication method . The vein pattern matching step includes
A branch point weighting value is applied to the branch point of the obtained vein pattern (Vpi), and the position of each vertex, the branch angle of a straight line from each vertex, the length between the vertices, and the degree of matching with respect to the number of branches are recognized and recognized. The method for authenticating a back of a hand according to any one of claims 5 to 11, wherein: