Method for contactlessly detecting fingerprints

Abstract

The invention relates to a method for detecting fingerprints using a 2D image-capturing unit (1) and a 3D image-capturing unit (3), wherein a three-dimensional capture of a finger (Fi) is created using the 3D image-capturing unit (3), wherein the image of the finger (Fi) is determined in the three-dimensional capture, wherein the object distance (di) between the 3D image-capturing unit (3) and the finger (Fi) is determined on the basis of the 3D coordinates of the determined image, wherein the 2D image-capturing unit (1) is adjusted to a focal plane corresponding to the determined object distance (di) and the position and / or location in the 3-dimensional capture, wherein a two-dimensional capture (Bj) of the finger (Fi) is created in the previously adjusted focal plane, wherein the position and / or location of the determined image of the finger (Fi) is mapped onto the two-dimensional capture (Bj), and wherein a finger image (FBi,j) and / or fingerprint (FPi,j) is determined from the region corresponding to the mapped position and / or location.

Description

[0001] METHOD FOR CONTACTLESS FINGERPRINT COLLECTION

[0002] The invention relates to a method for capturing fingerprints with a fingerprint recognition arrangement comprising a 2D image acquisition unit and a 3D image acquisition unit, in particular a mobile portable fingerprint recognition arrangement according to claim 1, a data carrier according to claim 15 and a fingerprint recognition arrangement according to claim 16.

[0003] Fingerprint recognition systems are biometric technologies widely used in a variety of application scenarios. These well-known fingerprint recognition systems typically employ contact-based capture using, for example, optical or capacitive sensors. This means that the user wishing to provide a fingerprint must place their finger on the sensor or its capture area, which some users find uncomfortable or unhygienic. Problems arise when capturing fingerprints with such sensors, as distortions in the captured prints can occur due to elastic deformation resulting from friction between the skin on the fingers and the sensor. Furthermore, fingerprint images created with such sensors can exhibit areas of low contrast due to the properties of the skin on the fingers.Moisture and dirt on the sensor plate are the cause. Potential security problems arise from latent fingerprints remaining on the sensors.

[0004] Another disadvantage of such contact-based detection methods is that they need to be cleaned regularly to prevent latent fingerprints from remaining or to ensure adequate hygiene for the user.

[0005] From EP 3663981 A1 and EP 3663976 A1, methods for the contactless detection of fingerprints are known in which a single 2D image acquisition unit is used and the fingerprints are always captured in sharp focus, since the focus plane is selected according to the object distance or the user moves their hand through the detection area and thus also through the focus plane of the image acquisition unit with a changing distance to the image acquisition unit.

[0006] A disadvantage of these contactless fingerprint capture methods is the comparatively long capture and processing time required to obtain sharp fingerprints, which negatively impacts user acceptance. Instead of precisely determining the object distance for individual fingers, the method measures the object distance for all of the user's fingers simultaneously and uses this combined measurement to select the focal plane. With this approach, it's possible that a finger might not be positioned precisely at the determined object distance, resulting in an unsharp image of that finger. Therefore, it may be necessary to take multiple shots to capture all fingers in focus. Furthermore, this combined object distance measurement for all fingers of a hand also necessitates that the distance measurement...The determined object distance makes it impossible to distinguish between a human hand and a dummy, for example a photograph of a hand.

[0007] The object of the invention is therefore to provide a method and a device for contactless fingerprint detection that overcome the aforementioned disadvantages and in particular have high recognition accuracy while ensuring a short recording and processing time.

[0008] The invention solves this problem with a computer-implemented method for capturing fingerprints with a fingerprint recognition arrangement comprising a 2D image acquisition unit and a 3D image acquisition unit, in particular a mobile portable arrangement with the features of claim 1.

[0009] According to the invention, it is provided that

[0010] - the 3D image acquisition unit is designed to create three-dimensional images in the form of point clouds containing the 3D coordinates of the individual points, and

[0011] - the 2D image capture unit can be focused on a variety of focus planes,

[0012] - the detection range of the 2D image acquisition unit and the detection range of the 3D image acquisition unit overlap, in particular are essentially identical,

[0013] - with the 3D image capture unit, at least one three-dimensional image of at least one finger located within the detection range of the 3D image capture unit is created,

[0014] - the image of the finger is determined in the three-dimensional image,

[0015] - based on the 3D coordinates of the individual points of the determined image of the finger in the three-dimensional recording, the object distance between the 3D image acquisition unit and the finger is determined, - the 2D image acquisition unit is set to a focus plane corresponding to the determined object distance and the position and / or orientation in the three-dimensional recording with a focus plane distance value,

[0016] - at least a two-dimensional image of the finger located within the detection range of the 2D image capture unit, created with the 2D image capture unit in the previously set focus plane,

[0017] - the position and / or location of the determined image of the finger in the three-dimensional image is mapped onto the two-dimensional image, and

[0018] - a finger image and / or fingerprint is determined from the area of ​​the two-dimensional image corresponding to the depicted position and / or location.

[0019] In the following, the term "plane of focus" refers to the area relative to an image capture unit in which objects are sharply imaged onto the sensor for a given lens setting. In practice, these are not true planes, but rather approximately flat areas defined by the shape of the lens used. Each plane of focus is associated with a plane-of-focus distance value, which specifies the distance between the lens of the image capture unit and the plane of focus. The depth of field of an optical system is the length of the interval in which objects are sufficiently sharply imaged.

[0020] Thus, the focal plane whose depth of field surrounds the determined object distance as closely as possible is understood to be the focal plane that renders the area around the determined object distance as sharply as possible.

[0021] In the following, a “recording” refers to the unprocessed “camera image” provided by the 3D image acquisition unit or the 2D image acquisition unit.

[0022] The term "image" of the finger in the three-dimensional or two-dimensional image refers to the area of ​​the respective image that shows the recorded finger, i.e., contains the information or image data that originates from the recorded finger.

[0023] In the following, "position" in the three-dimensional image refers to the location within the three-dimensional image, i.e., in a three-dimensional data space, while "orientation" refers to the orientation, including angles, etc., within the three-dimensional image, i.e., the three-dimensional data space. This approach advantageously allows the position of each finger to be determined quickly and precisely using the three-dimensional image created by the 3D imaging unit, thus reducing the fingerprint capture time. This can also lead to better user acceptance of the fingerprint recognition system and the fingerprint capture process.

[0024] This is made possible in particular by the fact that, instead of a single object distance value for the entire hand or all detected fingers, an object distance can be determined simultaneously for each individual finger, even if the distances of the individual fingers to the fingerprint recognition array differ. This allows the focusing of the optics to be carried out very precisely for each finger, enabling an immediately sharp image.

[0025] Furthermore, the user's fingerprints are always captured with exceptional sharpness by selecting the focus plane that is adapted to the object distance and the position or location of the fingers in the three-dimensional recording or the positioning in relation to the 3D image capture unit, so that they can easily be used for comparison with, for example, existing databases with contact-based fingerprints.

[0026] By mapping the position and / or location of the identified image of the finger from the three-dimensional image onto the two-dimensional image, particularly rapid processing is advantageously ensured, since there is no need to search for the finger or fingertips in the two-dimensional image. Instead, the mapping reveals which pixels or areas in the two-dimensional image correspond to which areas in the three-dimensional image, allowing the section of the two-dimensional image containing the finger image to be selected directly.

[0027] Furthermore, the three-dimensional recording from the 3D image acquisition unit makes it particularly easy and quick to check whether only a photograph is being held over the fingerprint recognition device instead of fingers, thus attempting to fake fingerprints.

[0028] A simple verification of the fingerprints determined according to the invention in common databases or with programs known from the prior art can be ensured if at least a part of the finger image and / or fingerprint is scaled based on the focal plane distance value and the position and / or location in the three-dimensional image, in particular the positioning of the at least one finger in the detection area of ​​the 2D image acquisition unit and / or the 3D image acquisition unit, wherein the scaling is carried out to a predetermined resolution, in particular to a predetermined dot density, preferably specified in DPI, which is directly proportional to the focal plane distance value.

[0029] In connection with the invention, DPI (dots per inch) is used synonymously with PPI (pixels per inch) for resolution or dot density.

[0030] By selecting the scaling according to the distance of the focal plane and the position and / or location in the three-dimensional image, it can be ensured that the images of the user's fingers are ideally scaled according to their position in the detection area of ​​the fingerprint recognition arrangement.

[0031] A simple verification of the fingerprints determined according to the invention in common databases or with programs known from the prior art can be ensured if the scaling to a predetermined resolution, in particular to greater than or equal to 500 DPI, is carried out individually for each finger image and / or each fingerprint.

[0032] To ensure a particularly easy comparison of the captured fingerprint images, which represent the user's fingerprints, with existing fingerprint databases that include fingerprints created using contact-based methods, it may be possible to convert the captured fingerprint images into a fingerprint format, whereby the fingerprints thus converted represent the images of the papillary ridges of the fingers, in particular in the form of a digital grayscale image.

[0033] In the following, a "fingerprint" is understood to be the equivalent of a contact-based fingerprint created from a two-dimensional image in the form of a digital image, in particular a digital grayscale image, which contains only the recorded images of the papillary ridges.

[0034] To reliably ensure that the fingerprint images and / or fingerprints are indeed optimally sharp, it may be necessary to check the quality of the determined fingerprint image and / or fingerprint against a predefined quality measure, in particular image sharpness, and

[0035] - that at least one more two-dimensional image of the finger is created with the 2D image capture unit in the previously set focus plane if the quality of the finger image and / or fingerprint is recognized as not optimal based on the specified quality measure.

[0036] To further increase recognition accuracy and reduce processing time, it may be provided that during the creation of at least one three-dimensional image with the 3D image acquisition unit and / or during the creation of at least one two-dimensional image with the 2D image acquisition unit, at least one finger located in the respective detection area is illuminated.

[0037] In order to determine whether the fingers in the detection area are actually human fingers and whether fingerprints are merely simulated, for example with a photograph, it may be provided that during the creation of at least one three-dimensional image with the 3D image acquisition unit and / or during the creation of at least one two-dimensional image with the 2D image acquisition unit, the at least one finger located in the respective detection area is illuminated with infrared light, in particular with a wavelength of 750-960 nm.

[0038] Such infrared illumination advantageously allows for the detection of veins under the skin or the measurement of blood flow, which can then be used as an additional feature in assessing whether fingerprints are being faked. This method utilizes the fact that infrared light penetrates deeper into the skin, thus making subcutaneous structures like veins visible.

[0039] Additionally or alternatively, it is of course also possible to use lighting during recording that generates light of other or additional wavelengths.

[0040] In order to achieve improved minute recognition, it may be provided that during the creation of the at least one three-dimensional image with the 3D image acquisition unit and / or during the creation of the at least one two-dimensional image with the 2D image acquisition unit, the at least one finger located in the respective detection area is illuminated from changing directions of illumination, wherein it is particularly provided that the height of the finger grooves is determined based on the finger images and / or fingerprints determined under changing directions of illumination.

[0041] By illuminating the finger from different directions, for example from the left, right, below and from "above" from the direction of the fingertips, etc., a shadow is cast on the finger ridges, which varies depending on the direction of illumination, so that the height of the finger ridges can be determined, which contributes to improved minutiae recognition.

[0042] To increase the reliability of fingerprint identification when compared with, for example, fingerprints in a database, it may be possible to...

[0043] - that a large number of three-dimensional images are created with the 3D image acquisition unit and a large number of two-dimensional images are created with the 2D image acquisition unit,

[0044] - that during the creation of the three-dimensional images and / or during the creation of the two-dimensional images, at least one finger located in the respective capture area is rotated and / or angled and

[0045] - that those areas of the two-dimensional images which correspond to the depicted position and / or location of the determined image of the finger in the three-dimensional image and which are sharply captured, are combined to form a complete finger image and / or complete fingerprint.

[0046] This approach advantageously allows for the capture of the sides of the fingers and their inclusion in the fingerprint images. Depending on the angle of the fingertips, various partial fingerprints are captured, and these individual images are used to create a composite fingerprint image that includes the fingerprint from different perspectives. This is particularly beneficial because the sides are crucial for the completeness of fingerprint features. Databases often contain rolled fingerprints that include the sides of the fingers, and reliable comparison and identification are only possible when the sides of the captured fingerprint images are also available for comparison.

[0047] Since optical components of any fingerprint recognition system are usually calibrated at least once in order to determine the corresponding focus plane based on the measured object distance and the position and / or orientation in the three-dimensional image, it can be advantageous to use a focus plane-object distance matrix for selecting the focus plane distance value.

[0048] - wherein the focus plane-object distance matrix is ​​created by calibrating the fingerprint recognition arrangement with a calibration pattern of known dimensions,

[0049] - where for a multitude of known lens configurations of the 2D image acquisition unit and the respective associated focal planes with their respective focal plane distance values, each

[0050] - the calibration pattern is moved through the detection range of the 2D image acquisition unit and / or the 3D image acquisition unit at different distances to the 3D image acquisition unit and the 2D image acquisition unit, whereby two-dimensional and three-dimensional images of the calibration pattern are created in each case,

[0051] - the two-dimensional images are each divided into a number of segments, in particular square ones, whereby the entries of the focus plane-object distance matrix correspond to the segments,

[0052] - those segments of the individual two-dimensional images are identified whose quality is recognized as optimal based on at least one predefined quality measure, in particular image sharpness,

[0053] - for the segments of the individual two-dimensional images identified as optimal, their position and / or orientation in the two-dimensional image is mapped onto the three-dimensional image, and the object distance between the 3D image acquisition unit and the calibration pattern corresponding to the respective mapped position and / or orientation in the three-dimensional image is determined, and

[0054] - The object distance corresponding to each segment of the two-dimensional images identified as optimal is stored in the focus plane-object distance matrix as the focal plane distance value of the respective focal plane, together with the position and / or location in the three-dimensional image and the associated lens configuration, so that the focus plane-object distance matrix contains for each segment of the two-dimensional image the object distance as the focal plane distance value, the position and / or location in the three-dimensional image and the associated lens configuration.

[0055] Accordingly, the focus plane-object distance table contains the corresponding focus plane distance value and the corresponding configuration for adjusting the lens elements for each object distance and each position and / or orientation in the three-dimensional image, so that a sharp image of a finger can be produced.

[0056] If the technical specification of the fingerprint recognition arrangement is not known, e.g., when components for the consumer market are used, it may be provided that, when using a fingerprint recognition arrangement whose technical specification, in particular the camera parameters of the 2D image acquisition unit, is not known in advance, a scaling matrix is ​​used for scaling the finger images and / or fingerprints.

[0057] - wherein the fingerprint recognition arrangement is calibrated with a calibration pattern of known dimensions to create the scaling matrix,

[0058] - where for a variety of lens configurations of the 2D image acquisition unit and the respective associated focus planes with their respective focus plane distance values, each

[0059] - the calibration pattern is moved through the detection range of the 2D image acquisition unit and / or the 3D image acquisition unit at different distances to the 2D image acquisition unit and the 3D image acquisition unit, whereby two-dimensional and three-dimensional images of the calibration pattern are created in each case,

[0060] - the two-dimensional images are each divided into a number of, in particular square, segments, whereby the entries of the scaling matrix correspond to the segments,

[0061] - those segments of the individual two-dimensional images are identified whose quality is recognized as optimal based on at least one predefined quality measure, in particular image sharpness,

[0062] - for the segments of the individual two-dimensional images identified as optimal

[0063] - each of whose position and / or location in the two-dimensional image is mapped onto the three-dimensional image,

[0064] - the object distance between the 3D image acquisition unit and the calibration pattern corresponding to the respective depicted position and / or location in the three-dimensional image is determined, and

[0065] - a calculation of the image resolution, especially in DPI, is performed based on the known dimensions of the calibration pattern, and

[0066] - the image resolution determined in this way, especially in DPI, together with the object distance belonging to the segment recognized as optimal, is stored in the scaling matrix as the focus plane distance value of the respective focus plane and the position and / or location in the three-dimensional image.

[0067] The scaling matrix contains the corresponding image resolutions in DPI for each focus plane distance value and each position and / or orientation in the three-dimensional image. This allows the appropriate scaling to be selected for the captured image of a finger, based on, for example, the focus plane distance value and the position and / or orientation in the three-dimensional image. This enables particularly quick and easy scaling of finger images or fingerprints to a specific DPI value (e.g., 500 DPI).

[0068] To ensure particularly precise scaling of the finger images or fingerprints, or a particularly precise selection of the focus plane distance value, it may be provided that, for the selection of the focus plane distance value based on the focus plane-object distance matrix and / or for the selection of the scaling based on the scaling matrix, an average value of the relevant entries of the focus plane-object distance matrix and / or the scaling matrix is ​​determined if the image of at least one finger in the three-dimensional image extends over several segments.

[0069] If the image of the captured finger in the three-dimensional image extends over an area corresponding to several segments in the two-dimensional image, which have different focus plane distance values, this ensures that an optimal focus plane distance value is selected.

[0070] In order to capture as many fingers as possible, or ideally all fingers of a hand, that are within the detection range of the fingerprint recognition system, and to create corresponding fingerprint images or fingerprints, it may be possible to provide the following:

[0071] - that at least one three-dimensional image of several, in particular four, fingers located within the detection range of the 3D image capture unit is created,

[0072] - that the image of each finger is determined in the three-dimensional recording,

[0073] - that for each individual finger, the distance between the 3D image acquisition unit and the respective finger is determined based on the 3D coordinates of the individual points of the determined image of the respective finger in the three-dimensional recording,

[0074] - that the 2D image capture unit is set to a focus plane corresponding to the determined distance of each finger, and that at least one two-dimensional image of the respective finger is created with the 2D image capture unit in the previously set focus plane,

[0075] - that the position and / or location of the determined images of the individual fingers in the three-dimensional image are each mapped onto the two-dimensional image that was created in the focus plane previously set for the respective finger, and

[0076] - that a finger image and / or fingerprint of the respective finger is determined from the area of ​​the respective two-dimensional image corresponding to the depicted position and / or location.

[0077] In order to reliably detect that a fingerprint is being faked, it may be provided that, based on the 3D coordinates of the individual points of the determined image of one or more fingers in the three-dimensional recording, an analysis of the distance distribution between one or more fingers and the 3D image recording unit is carried out, wherein it is particularly provided that a statistical analysis of the distance distribution is carried out and / or that the distances along a profile running perpendicular to the fingertips are analyzed.

[0078] In this way, it is possible to analyze particularly quickly and reliably whether, for example, a sheet of paper with a photo or a real hand with indentations between the fingers and curves on the fingers is held in the detection area of ​​the fingerprint recognition device.

[0079] According to an advantageous embodiment of a method according to the invention, with which presentation attacks can be excluded and human fingers or hands detected particularly easily and reliably, it can be provided that

[0080] - that during the creation of three-dimensional images with the 3D image capture unit, predefined finger movements are performed with at least one finger that is located in the respective capture area,

[0081] - that the three-dimensional recordings created during the execution of the finger movements are compared with stored three-dimensional recordings of fingers that were created while the same movements were performed, in particular by applying neural networks, and

[0082] - that the presence of a real human hand is inferred based on the comparison. According to a structurally simple embodiment of the invention, the 3D image acquisition unit can be a time-of-flight sensor, and depth images can be created as three-dimensional images by the 3D image acquisition unit. In particular, it is provided that depth images are created in the form of a 2D matrix, the individual entries of which include at least one object distance value for the individual object points in the recording area of ​​the 3D image acquisition unit.

[0083] The invention further relates to a computer-readable storage medium comprising instructions which, when executed by a computer, in particular a processing unit of a fingerprint recognition arrangement, cause it to execute a method according to the invention.

[0084] A further object of the invention is to provide a fingerprint recognition arrangement for capturing fingerprints, in particular designed as a mobile portable terminal device, preferably designed for carrying out a method according to the invention and / or containing a data carrier according to the invention, with which a method according to the invention can be carried out.

[0085] The invention solves this problem with a fingerprint recognition arrangement according to claim 16, comprising a 2D image acquisition unit and a 3D image acquisition unit.

[0086] - wherein the 3D image acquisition unit is designed to create three-dimensional images in the form of point clouds containing the 3D coordinates of the individual points,

[0087] - wherein the 2D image acquisition unit can be focused on a variety of focus planes, and

[0088] - wherein the detection range of the 2D image acquisition unit and the detection range of the 3D image acquisition unit overlap, in particular being essentially identical.

[0089] According to the invention, it is provided that

[0090] - a processing unit is provided which is designed to

[0091] - to cause the 3D image acquisition unit to create at least one three-dimensional image of at least one finger located within the detection range of the 3D image acquisition unit,

[0092] - to determine the image of the finger in the three-dimensional image,

[0093] - to determine the object distance between the 3D image acquisition unit and the finger based on the 3D coordinates of the individual points of the determined image of the finger in the three-dimensional image, - to set the 2D image acquisition unit to at least one focal plane with a focal plane distance value corresponding to the determined object distance and the position and / or location in the three-dimensional image, and, with this setting, to cause at least one two-dimensional image of at least one finger that is located in the detection range of the 2D image acquisition unit to be created,

[0094] - to map the position and / or location of the determined image of the finger in the three-dimensional image onto the two-dimensional image, and

[0095] - to determine a finger image and / or fingerprint from the area of ​​the two-dimensional image corresponding to the depicted position and / or location.

[0096] A fingerprint recognition arrangement designed in this way achieves all the advantages that were described at the beginning of the inventive method and is also extremely compact.

[0097] Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawings.

[0098] The invention is below schematically illustrated in the drawings using particularly advantageous, but not limiting, embodiments and is described by way of example with reference to the drawings.

[0099] The following will show:

[0100] Fig. 1 shows a fingerprint recognition arrangement for carrying out a method according to the invention,

[0101] Fig. 3 shows a schematic diagram of the process of an embodiment of a method according to the invention.

[0102] Fig. 2a is a schematic example of a two-dimensional recording,

[0103] Fig. 2b a schematic example of a finger picture,

[0104] Fig. 2c shows a schematic example of a fingerprint.

[0105] Fig. 1 shows a fingerprint recognition arrangement 100 according to the invention for carrying out a fingerprint detection method according to the invention. In the illustrated embodiment, the fingerprint recognition arrangement 100 comprises a 2D image acquisition unit 1, a processing unit 2, and a 3D image acquisition unit 3. Various data 21 can be stored in the processing unit 2. The 2D image acquisition unit 1 is, for example, a camera whose lens 11 is electronically focusable. One implementation of such a lens is called a liquid lens and is based on the principle of electrowetting, according to which the interface between two liquids changes when a voltage is applied. The optics of the 2D image acquisition unit 1 can be electronically adjusted to a focus plane with submillimeter accuracy.The 2D image acquisition unit 1 can create two-dimensional images B, in the form of grayscale images or color images containing pixel-by-pixel brightness or color information.

[0106] The 3D image acquisition unit 3 of the fingerprint recognition system 100 can, for example, be a time-of-flight (TAF) camera. Such a TAF camera determines the distance to the surface of an object point by point using the time-of-flight method. For this purpose, a light pulse is emitted, and the time it takes for the light to reach the object and return to the sensor is measured. This required time of flight is directly proportional to the distance to the object. The scene can be illuminated, for example, with an internal or external infrared light source. As a result, such a TAF camera delivers three-dimensional images in the form of a depth map, which contains a distance measurement for each of a number of elements or pixels arranged in a raster or image-like pattern.

[0107] Alternatively, a 3D image acquisition unit can be provided, which projects structured light patterns onto a surface, such as a user's hand. These patterns are then captured and digitized as three-dimensional images, for example, as three-dimensional point clouds. This can be done either by evaluating stereo images generated with two cameras or by detecting disturbances in the light pattern caused by objects in the reflected image.

[0108] The advantage of this compact hardware design is that it can also be integrated into a mobile device such as a smartphone or tablet, thus eliminating the need for bulky three-dimensional scanning sensors. Therefore, such a fingerprint recognition system can also be mobile and portable.

[0109] The detection areas of the 2D image acquisition unit 1 and the 3D image acquisition unit 3 of a fingerprint recognition arrangement 100 according to the invention overlap and can be particularly advantageously essentially identical, as is the case in the exemplary embodiment shown in Fig. 1. Since the detection areas of the 2D image acquisition unit 1 and the 3D image acquisition unit 3 are essentially identical in the exemplary embodiment, both detection areas are referred to collectively as the detection area 10 of the fingerprint recognition arrangement 100 in the following.

[0110] The acquisition times of the 2D image acquisition unit 1 and the 3D image acquisition unit 3 are synchronized, so that each two-dimensional image B is virtually paired with a three-dimensional image. The acquisition times of the two-dimensional and three-dimensional images are almost identical, so that the respective areas of the fingers Fj depicted are also almost identical. Any time differences of, for example, 10 to 20 milliseconds between the acquisition times can be disregarded, since experience has shown that the user's hand moves only minimally during this period.

[0111] The 2D image acquisition unit 1 and the 3D image acquisition unit 3 are calibrated to each other such that for every acquisition point from a three-dimensional image of the 3D image acquisition unit 3, a corresponding acquisition point in the two-dimensional image B of the 2D image acquisition unit 1 is known. These corresponding acquisition points change depending on the perspective. Calibration methods for determining such corresponding acquisition points are known from the prior art and are described, for example, in Rathnayaka, Pathum, Seung-Hae Baek, and Soon-Yong Park. “An Efficient Calibration Method for a Stereo Camera System with Heterogeneous Lenses Using an Embedded Checkerboard Pattern.” Edited by Oleg Sergiyenko. Journal of Sensors 2017 (September 2017): 6742615. https: / / doi.org / 10.1155 / 2017 / 6742615.

[0112] In the exemplary embodiment shown in Fig. 1, lighting elements 4, for example light-emitting diodes, are arranged around the 2D image acquisition unit 1 and the 3D image acquisition unit 3 to ensure sufficient illumination of the detection area 10. The detection area 10 is defined by the distances d min and d max limited, whereby the distance d max is chosen in such a way that the required resolution of the fingerprints of, for example, at least 500 DPI is guaranteed and the distance d min is chosen so that four fingers of a hand are still visible in image B at the same time.

[0113] In the embodiment shown in Fig. 2, a user holds their hand over the detection area 10 of the fingerprint recognition device 100. The fingerprint recognition device 100 automatically starts capturing fingerprints as soon as one or more of the user's fingers Fj are within the detection area 10. Hereinafter, i denotes an index ie{1 ,2, ... 10}, which identifies the user's finger F that is being captured or viewed. In the embodiment shown, four fingers F nF4 of the user, namely the index finger, middle finger, ring finger and little finger, are captured simultaneously. The user's thumb F5 is outside the capture area 10. To capture the fingers Fj, it is not necessary for the user to move their hand into the image capture area 10 in a predetermined position or orientation, but they can move it into the image capture area 10 in a comfortable, natural position and orientation.

[0114] To take a picture of a single finger Fj; F ..., F5, the user only needs to move it into the capture area 10 and rotate it so that the papillary lines are detectable for the image capture unit 1.

[0115] First, in step A, the 3D image acquisition unit 3 creates a three-dimensional image of the hand, including each individual finger Fj. Then, in step B, the individual fingers Fj, or rather the captured image of each finger Fj, are identified in the point cloud. This allows their position and / or orientation within the three-dimensional image, and thus their positioning or orientation relative to the 3D image acquisition unit 3, to be determined. This search for the images of the fingers Fj can be performed, for example, using established techniques such as threshold-based pattern matching. Pattern matching is a general form of image processing and is described, for example, in Christopher M. Bishop, Pattern Recognition and Machine Learning (Information Science and Statistics), January 1, 2007, Springer New York, NY.In this context, the 3D data can be divided into two different groups using a threshold, and the results can be searched for typical shapes that resemble fingertips.

[0116] In the case of 3D data, a distance value can be chosen as a threshold, and all pixels whose distance values ​​are greater than the threshold are assigned to a group such as "Background." All pixels whose distance values ​​are less than the threshold are assigned to a group called "Pattern." The pixels in the "Pattern" group can then be interpreted as patterns and matched accordingly using fingertips.

[0117] Instead of this approach, two thresholds can also be used, so that only pixels within a range between the first and second thresholds are assigned to the "Pattern" group and then matched accordingly. Alternatively, a search for images of the fingers Fj can be performed using a neural network, as is available, for example, as a programming interface under the TensorFlow Object Detection API, from https: / / github.com / fensorflow / modes / tree / master / research / objec.detectjon, last accessed on October 17, 2022.

[0118] In step B, the object distance dj between each identified finger Fj (or fingertip) and the 3D image acquisition unit 3 is then determined based on the 3D coordinates of the individual points of the identified image of finger Fj in the three-dimensional image. In most cases, the area of ​​the point cloud containing the values ​​originating from a single finger Fj is not characterized by a single distance value, but by several values ​​that differ at least slightly from one another. Therefore, to determine the object distance dj, for example, a mean or median of the relevant values ​​can be calculated, or a certain percentage of the nearest pixels from the relevant area can be used to determine a representative object distance dj.

[0119] In step C, the optics of the 2D image acquisition unit 1 are successively adjusted to the distance of the fingertips so that each fingertip can be sharply imaged by the 2D image acquisition unit 1. This means that the 2D image acquisition unit 1 is set to a focal plane with a corresponding focal plane distance value that corresponds to the determined object distance dj and the position and / or orientation of the respective finger Fj or fingertip in the three-dimensional image. In this set focal plane, at least one two-dimensional image B of the finger Fj, which is located within the detection area 100, is taken with the 2D image acquisition unit 1 in step D. In this two-dimensional image B, the fingertip is sharply imaged, so that the fingertip can be extracted as a fingerprint image FBjj or fingerprint FPjj, and thus as a fingerprint.

[0120] The previously mentioned calibration makes it possible to deduce the position of the fingertip in the two-dimensional image B from the three-dimensional image, specifically from the position and / or orientation of the fingertip in the three-dimensional image, since the corresponding pixels or voxels are known. In other words, the position and / or orientation of the determined image of finger Fj in the three-dimensional image is mapped onto the two-dimensional image B in step E. This eliminates the need to search for the fingertip in the two-dimensional image B; instead, the corresponding section can be selected directly.

[0121] Such mapping methods are also known as lookup tables (LUTs) in the prior art and are described, for example, at https: / / en.wjkipedja.org / wjki / Lookuo.tabte (last accessed on October 17, 2022). For a specific distance, the corresponding pixel in the two-dimensional image is determined for each pixel in the three-dimensional image and stored in a table. This process is then repeated for each relevant distance in suitable intermediate steps, e.g., all distance values ​​between 100 mm and 200 mm in 1 mm increments, and also stored in a table. These tables then constitute the lookup table (LUT). This means that, starting from a 3D pixel x,y,d (x-coordinate, y-coordinate, d-distance), it is possible to extract an x / y coordinate from the table and thus assign the point in the two-dimensional image.

[0122] Finally, in step F, a finger image FBjj and / or fingerprint FPjj is determined and extracted from the area of ​​the two-dimensional image B corresponding to the depicted position and / or orientation and stored in data storage 21. The result is sharp two-dimensional images of the fingertips in the form of a finger image FBjj or fingerprint FPjj, respectively, stored in data storage 21.

[0123] Figure 3 then describes further optional process steps: In step G, a predetermined quality measure is additionally determined for each finger F to ascertain whether the quality of the resulting fingerprint image FBjj and / or fingerprint FPjj is sufficiently high. Figure 2 shows how the image sharpness is determined for this purpose. In step H, it is then checked for each finger Fj whether the image sharpness is sufficiently high (indicated by the checkmark in Figure 2), and if so, the fingerprint images FBjj and / or fingerprints FPjj with the best quality are selected for each finger Fj (step I). If the image sharpness is not sufficiently high, indicated by the "X" in Figure 2, new two-dimensional and three-dimensional images are created, and the fingerprint image FBjj and / or fingerprint FPjj is extracted again.

[0124] The procedure described above is repeated for the remaining fingers Fj in detection area 10. That is, each finger Fj is individually determined in the three-dimensional image and the corresponding object distance d is recorded. h and its position and / or orientation is determined. Do the individual fingers Fj or fingertips require different focus plane distances a due to different object distances? k , so the 2D image capture unit 1 is adjusted to different focus planes and two-dimensional images B are created until an optimal image has been achieved for each fingertip.

[0125] From the individual two-dimensional images B (see Fig. 2a), a finger image FBjj is always created for the finger Fj for whose object distance dj and position and / or orientation the focus plane was selected, with which the two-dimensional image B was created. Fig. 2b shows an example of a captured finger image FBjj, which contains information about background and skin color. In Fig. 2a, an example with four fingers Fj ie{1 ,... ,4} is shown. The user's thumb was not fully captured and therefore not detected as an image of a finger. For the example in Fig. 1, four finger images FB^,..., FB are therefore created. 4ij the finger F4, ... , F4, i.e. the index finger, the middle finger, the ring finger and the little finger created.

[0126] Those parts or image areas of the two-dimensional image Bj in which no finger image of a finger was determined can optionally be discarded, so that the processing of the image Bj can be carried out faster and with less computational effort, while at the same time saving storage space.

[0127] In the illustrated embodiment, the fingerprints FPjj are extracted from the individual fingerprint images FBjj of fingers Fj in a further processing step, as shown, for example, in Liu, Xinwei, et al., "An improved 3-step contactless fingerprint image enhancement approach for minutiae detection." Visual Information Processing (EUVIP), 2016 6th European Workshop on Visual Information Processing, IEEE, 2016. This is optional, however. An example of such a fingerprint FPjj is shown in Fig. 2c. Optionally, the contrast of the ridges or papillary ridges can then be enhanced to improve the quality of the extracted fingerprints. This contrast enhancement can be achieved using mean variance, normalization, normalized box filters, and / or morphological operations.

[0128] Scaling

[0129] In the exemplary embodiment, the extracted fingerprints FPjj of the individual fingers F4,..., F4 are then scaled to a predetermined resolution of, for example, 500 DPI to enable easy comparison with existing fingerprint databases.

[0130] Since most fingerprint recognition algorithms require a standard resolution of at least 500 DPI to be compatible with the FBI specification given in Criminal Justice Information Services: Electronic Fingerprint Transmission Specification, May 2, 2005, Federal Bureau of Investigation, Criminal Justice Information Services Division, 1000 Custer Hollow Road, Clarksburg, L / L / V 26306, scaling the captured fingerprints FPjj to this resolution is advantageous.

[0131] Alternatively, such scaling can also be performed for fingerprint images in FBjj format. However, scaling fingerprint images in FBj or FPjj format is generally optional, as further processing is possible even without prior scaling.

[0132] There are two ways to convert image pixels into metric sizes: On the one hand, the scaling factor to be applied to a two-dimensional image B can be calculated using the following formula:

[0133] Here, G is the object size [mm] and B is the image size [mm] given by the 2D image acquisition unit 1 , d is the object distance [mm] and BW is the focal length [mm].

[0134] The calculation of pixels DPI [dots per inch] is based on the following relationship:

[0135] SenResi.,,,..

[0136] DPI = PPI = - -^-25, 4

[0137] (j

[0138] SenRes provides horThe horizontal pixel resolution of the sensor or image capture unit 1 is specified. In the context of the invention, DPI is used synonymously with PPI for resolution or pixel density.

[0139] Alternatively, if the technical specifications of the sensor or the 2D image acquisition unit 1 (e.g., for consumer market components) are not available, a one-time calibration for scaling can be performed. For this, a calibration pattern with known dimensions is used, and the same calibration process as previously described is followed. The image resolution, for example in DPI, is calculated using the known metric dimensions of the calibration pattern. This type of calibration thus enables the conversion of image pixels into metric units. The determined values ​​are stored in a scaling table. To determine the required scaling for the individual fingerprints FPjj of the individual fingers FT,..., F4 as quickly and easily as possible, a scaling matrix is ​​used in the illustrated embodiment for scaling the fingerprint images FBjj or fingerprints FPjj.To create the scaling matrix, a calibration pattern with known dimensions is used and the following calibration process is carried out:

[0140] During calibration, the lens elements are configured differently, i.e., an adjustment of the lens elements is made, e.g., by changing their shape and / or position. The corresponding focal plane distance value a k The calibration pattern is determined by moving it at varying distances through, or approaching, the detection area 10 of the 2D image acquisition unit 1 and the 3D image acquisition unit 3. During this process, two- and three-dimensional images of the calibration pattern are repeatedly taken, passing through the plane of focus.

[0141] The two-dimensional images B are divided into several segments, for example, by overlaying a rectangular grid onto the two-dimensional image. In this embodiment, the two-dimensional images B are divided into segments arranged in several columns and rows of equal length. The scaling matrix has a corresponding number of rows and columns.

[0142] Subsequently, in each of the two-dimensional images B created, those segments are sought whose quality is optimal, for example, those that are optimally sharp.

[0143] To determine the optimally sharp segment in each case, an image stack is first created from the acquired images B, je{1 ,... ,M}. The quality of the segments is evaluated individually, almost in real time. In the illustrated embodiment, a relative ranking of all segments of an image stack is performed for each segment individually based on its image sharpness. The calculation of image sharpness can be performed, for example, as described in Eskicioglu, Ahmet M., and Paul S. Fisher. "Image quality measures and their performance." IEEE Transactions on communications 43.12 (1995): 2959-2965, and is not considered an absolute measure, but rather serves to create a ranking of the segments. Optionally, it is of course also possible to use other measures of image quality, such as papillary ridge or ridge frequency, in addition to or as an alternative to image sharpness. For the segments identified as optimal, the image is then mapped or...The position and / or location of the three-dimensional image is determined, i.e., where and how it lies within the three-dimensional image, thus determining its positioning relative to the 3D image acquisition unit 3 of the area of ​​the calibration pattern depicted in the relevant segment. Furthermore, the object distance d corresponding to the "optimal" segment is determined.

[0144] The calculation of the image resolution for the segments identified as optimal, for example in DPI, is performed using the known metric dimensions of the calibration pattern. This type of calibration thus enables the conversion of image pixels into metric units.

[0145] The values ​​determined for the respective segment, i.e., the image resolution, the object distance d as the focus plane distance value a kThe position and / or orientation in the three-dimensional image are stored in the scaling matrix in the entry that corresponds to the respective segment. This is repeated until corresponding values ​​have been determined for all segments of the two-dimensional image Bj and the corresponding entries of the scaling matrix have been filled.

[0146] When a finger F is captured with the fingerprint capture device 100, it is possible to directly scale the portion of the two-dimensional image Bj that shows the captured finger Fj, based on the position and / or orientation of the image of finger Fj in the three-dimensional image and its object distance dj. Optionally, further processing, such as registration, of the finger images FBjj or fingerprints FPjj of a finger Fj can be performed after scaling. Of course, a reduction in the resolution of the image Bj can also be performed optionally before scaling.

[0147] Selection of the focus plane

[0148] To select the appropriate configuration for adjusting the lens elements and thus the focal plane, the focal plane distance value a is used. kTo simplify the process of determining the focal plane-object distance for a measured object distance dj, a focal plane-object distance matrix is ​​used in the exemplary embodiment. Such a focal plane-object distance matrix is ​​also created by a previously described calibration.

[0149] As previously described, the corresponding object distance d, as well as their position and / or orientation in the corresponding three-dimensional image, are determined for the optimally sharp segments of the two-dimensional images Bj. The focus plane-object distance matrix also has a number of rows and columns that correspond to the segments of the two-dimensional images. If the object distance d|* lies within the detection range 10, a focus plane with the focus plane distance value a is assigned to this configuration of lens elements. k= assigned to d and stored together with the position and / or location in the three-dimensional image in the focus plane-object distance matrix.

[0150] During calibration and subsequent operation, a large aperture is selected to ensure a shallow depth of field, i.e., a small extent of the sharply imaged area in the object space of the imaging optical system of the 2D image acquisition unit 1. In this way, the depth of field s, or the corresponding plane of focus, can be determined using the plane-of-focus distance value a. k At a specific position and / or location in the three-dimensional recording, exactly one distance value d between the recorded object or finger Fj and the 2D image recording unit 1 can be assigned.

[0151] Capturing the page areas

[0152] With a fingerprint recognition arrangement 100 according to the invention or a method according to the invention, it is advantageously possible to also detect the side areas of the fingers Fj or of the fingerprints, although the depth of field of the 2D image recording unit 1 can often only be 1 mm, so that the side areas of the fingers are not captured sharply.

[0153] To do this, the user simply needs to rotate their fingers Fj or fingertips laterally or forwards within the capture area 10 during the creation of two-dimensional and three-dimensional images, so that the fingers Fj are captured at different angles by the 3D image capture unit 3 and the 2D image capture unit 1. Depending on the angle of the fingertips, different fingerprint images FBjj or fingerprints FPjj are captured. From these individual images, a composite fingerprint image FB*j or fingerprint FP* is then created, which includes the fingerprint from different perspectives.

[0154] In the two-dimensional images B, the sharply focused areas corresponding to the position and / or orientation of the determined image of finger Fj in the three-dimensional image are then located and stitched together using, for example, a stitching method known from the prior art. Such a stitching method is described, for example, in Choi, Heeseung, Kyoungtaek Choi, and Jaihie Kim. “Mosaicing Touchless and Mirror-Reflected Fingerprint Images.” IEEE Transactions on Information Forensics and Security 5, no. 1 (March 2010): 52-61. https: / / doi.org / 10.1109 / TIFS.2009.2038758.

[0155] In this way, complete fingerprint images FB* j and complete fingerprint images FP* j can be created, which correspond to fingerprints rolled onto paper. This can increase the reliability of identification.

[0156] minute detection

[0157] Minutiae are endpoints and branches of the papillary ridges, which are particularly characteristic features for the reliable identification of fingerprints.

[0158] To identify minutiae with particular reliability, a fingerprint recognition arrangement 100 according to the invention can have various lighting options that illuminate the detection area 10, and thus also the fingers Fj within the detection area 10, from different directions, such as from left, right, above, below, half-left, half-right, etc., and combinations thereof. If three-dimensional and two-dimensional images are created during this alternating illumination, images of the fingers Fj are obtained with different lighting directions and consequently different shadows cast on the papillary ridges or finger grooves. If the images are taken in rapid succession, the offset is small enough to assign the fingers Fj in the different images.

[0159] The height of the finger grooves can be determined from this shadow, which contributes to improved minutiae recognition. The height of the finger grooves can be determined using a prior art method for calculating height from angle / shadow, for example, as described in Park, Ki-Hong, and Yang Sun Lee. “Simple Shadow Removal Using Shadow Depth Map and Illumination-Invariant Feature.” The Journal of Supercomputing 78, no. 3 (February 1, 2022): 4487–4502. https: / / doi.org / 10.1007 / s11227-021-04043-5.

[0160] Detecting fingerprint forgery

[0161] In addition, the embodiment shown in Fig. 1 allows for the illumination of the user's fingers Fj with light of different wavelengths during the creation of the two-dimensional and three-dimensional images of the fingers Fj. For this purpose, a lighting unit 4 is arranged in a ring around the 2D image acquisition unit 1 and the 3D image acquisition unit 3 in Fig. 1, so that the entire detection area 10 is illuminated by the lighting unit 4.

[0162] As an alternative to such a ring-shaped lighting unit 4, the lighting unit 4 can also be made up of different LED strips, i.e., one strip each on the left, right and above the fingertips.

[0163] The illumination unit 4, which comprises an array of LEDs, can be designed to emit infrared light with a wavelength of 750-960 nm, allowing it to penetrate deeper into the skin than the light from conventional light sources. This makes it possible to detect veins under the skin or to measure blood flow, using this information as an additional feature in assessing whether the object within detection range 10 is actually a human finger or merely a photograph or some other type of inanimate object—a process known as liveness detection.

[0164] Vein detection and blood flow measurement are described, for example, in Al-Tamimi, Mohammed Sabbih Hamoud. “A SURVEY ON THE VEIN BIOMETRIC RECOGNITION SYSTEMS: TRENDS AND CHALLENGES,” 2005, 18 and Liu, Chun-Yu, Shanq-Jang Ruan, Yu-Ren Lai, and Chih-Yuan Yao. “Finger-Vein as a Biometric-Based Authentication.” IEEE Consumer Electronics Magazine 8, no. 6 (November 2019): 29-34. https: / / doi.org / 10.1109 / MCE.2019.2941343. Finger / palm vein detection technology uses images captured by illuminating a finger or palm with near-infrared light, which easily penetrates the body. The hemoglobin in the veins of the palms absorbs this light, reducing the reflection rate and causing the veins to appear as a black pattern.Blood flow can be seen indirectly, as the veins pulsate rhythmically with the heartbeat (becoming larger or smaller) and this change between systole and diastole can be measured, as shown, for example, below. tesUnfo / rafgeber^unktionsweise / pulsmessung-mit-optoelektronischen-sensoren / , last accessed on 17.10.2022, is described.

[0165] However, the three-dimensional images can also be used in other ways to detect fingerprint forgery. For example, if a photograph of a finger or hand is held over the fingerprint recognition device 100 to deceive the user, the 3D data can be analyzed to determine whether it lies on the same plane, i.e., on the same photograph. This is done by analyzing the distance distribution between one or more fingers Fj and the 3D image acquisition unit 3, based on the 3D coordinates of the individual points of the detected image in the three-dimensional image. This easily reveals whether the object in the detection area 10 is a piece of paper with a photograph or a human hand with fingers, since a human hand has ridges and depressions between the fingers.A statistical analysis of the distance distribution, for example along a profile running perpendicular to the fingertips, reveals whether such bulges or depressions are present, or whether it is an approximately smooth profile like a photograph. For this purpose, pre-processed profiles of fingers from different users can be stored in memory 21 for comparison with the currently created profile. In the height profile of a human hand, a wave-like structure is visible across the fingers (Fj), resulting from the changing distance between the fingers; in a photograph, a uniform height profile is visible, more or less at a consistent level.

[0166] Furthermore, with a fingerprint recognition device 100 according to the invention, or a method according to the invention, it is possible to distinguish a hand replica with rigid fingers from human fingers, i.e., to detect so-called presentation attacks. For this purpose, the user moves their fingers within the detection area 10 during the creation of the three-dimensional and two-dimensional recording. Due to the measured object distances changing during the movement of the fingers and the changing position and orientation of the respective fingers in the three-dimensional recordings, it can be concluded that the user's hand or fingers are actually moving and belong to a real human hand.

[0167] Optionally, the user can also perform certain finger movements on command, which can then be checked against three-dimensional recordings of fingers of different users performing the same movements stored in data storage 21, in order to also conclude that a real human hand is present.

[0168] With a rigid hand replica, the individual fingers remain in the same relative position to each other, whereas with a real human hand, individual fingers can be moved, causing their relative positions to constantly change. Presentation attacks are carried out, for example, by first recording videos of the hands to be simulated and then holding these videos up to the recognition system using a monitor, tablet, or smartphone, thus creating the illusion of a "real" hand. The displays of these devices represent the image using pixel matrices, whose DPI values ​​are usually below 500 DPI. By appropriately selecting the 2D image capture unit, typical image artifacts are created in the recorded two-dimensional image, which differ from "real" live recordings.

[0169] Deep learning methods can be used as another way to detect presentation attacks. Specialized Generative Adversarial Networks (GANs) or Convolutional Neural Networks (CNNs) are used, for example. A large number of images are captured using presentation attacks. Various devices are used, such as Apple iPhones, iPads, or Samsung smartphones. These devices are fitted with different screen protectors to introduce varying reflection properties into the sample presentation attack images.

Claims

Patent claims 1. Method for capturing fingerprints with a fingerprint recognition arrangement (100) comprising a 2D image acquisition unit (1) and a 3D image acquisition unit (3), in particular a mobile portable arrangement, - wherein the 3D image acquisition unit (3) is designed to create three-dimensional images in the form of point clouds containing the 3D coordinates of the individual points, and - wherein the 2D image acquisition unit (1) can be focused on a multitude of focus planes, - wherein the detection range of the 2D image acquisition unit (1) and the detection range of the 3D image acquisition unit (3) overlap each other, in particular are essentially identical, - wherein at least one three-dimensional image of at least one finger (Fj) located in the detection range of the 3D image acquisition unit (3) is created using the 3D image acquisition unit (3), - whereby the image of the finger (Fj) is determined in the three-dimensional image, - wherein, based on the 3D coordinates of the individual points of the determined image of the finger (Fj) in the three-dimensional recording, the object distance (dj) between the 3D image acquisition unit (3) and the finger (Fj) is determined, - wherein the 2D image acquisition unit (1) is focused on a focus plane corresponding to the determined object distance (dj) and the position and / or orientation in the three-dimensional image with a focus plane distance value (a) k ) is set up, - wherein at least one two-dimensional image (Bj) of the finger (Fj) located within the detection range of the 2D image acquisition unit is created with the 2D image acquisition unit (1) in the previously set focus plane, - wherein the position and / or location of the determined image of the finger (Fj) in the three-dimensional image is mapped onto the two-dimensional image (Bj), and - wherein a finger image (FBjj) and / or fingerprint (FPjj) is determined from the area of ​​the two-dimensional image (Bj) corresponding to the depicted position and / or location.

2. Method according to claim 1, characterized in that a scaling of at least a part of the finger image (FBjj) and / or fingerprint (FPjj) is performed based on the focus plane distance value (a k ) and the position and / or location in the three-dimensional recording, in particular the positioning of at least one finger (Fj) in the detection area of ​​the 2D image acquisition unit (1) and / or the 3D image acquisition unit (3), is carried out, wherein the scaling to a predetermined resolution, in particular to a predetermined dot density, preferably specified in DPI, which is directly proportional to the focus plane distance value (a k ) is, is carried out.

3. Method according to claim 1 or 2, characterized in that the determined finger images (FBjj) are converted into a fingerprint format, wherein the fingerprints converted in this way (FPjj) represent the images of the papillae of the fingers (Fj), in particular in the form of a digital grayscale image.

4. Method according to one of the preceding claims, characterized in that - that the quality of the determined fingerprint image (FBjj) and / or fingerprint (FPjj) is checked against a predefined quality measure, in particular image sharpness, and - that at least one more two-dimensional image (B,) of the finger (Fj) is created with the 2D image acquisition unit (1) in the previously set focus plane if the quality of the finger image (FBjj) and / or fingerprint (FPjj) is recognized as not optimal based on the specified quality measure.

5. Method according to one of the preceding claims, characterized in that during the creation of the at least one three-dimensional image with the 3D image acquisition unit (3) and / or during the creation of the at least one two-dimensional image (B) with the 2D image acquisition unit (1), the at least one finger (Fj) that is located in the respective detection area is illuminated, in particular with infrared light, preferably with a wavelength of 750-960 nm.

6. Method according to one of the preceding claims, characterized in that during the creation of the at least one three-dimensional image with the 3D image acquisition unit (3) and / or during the creation of the at least one two-dimensional image (B) with the 2D image acquisition unit (1), the at least one finger (Fj) located in the respective detection area is illuminated from changing directions of illumination, wherein it is particularly provided that the height of the finger ridges is determined based on the finger images and / or fingerprints obtained under changing directions of illumination.

7. Method according to one of the preceding claims, characterized in that, - that a large number of three-dimensional images are created with the 3D image acquisition unit (3) and a large number of two-dimensional images (Bj) are created with the 2D image acquisition unit (1), - that during the creation of the three-dimensional images and / or during the creation of the two-dimensional images (Bj), at least one finger (Fj) located in the respective capture area is rotated and / or angled and - that those areas of the two-dimensional images which correspond to the depicted position and / or location of the determined image of the finger (Fj) in the three-dimensional image and which are sharply captured, are combined to form a total finger image (FB* j) and / or total fingerprint (FP*ij).

8. Method according to one of the preceding claims, characterized in that for the selection of the focus plane distance value (a k ) a focus plane-object distance matrix is ​​used, - wherein the focus plane-object distance matrix is ​​created by calibrating the fingerprint recognition arrangement (100) with a calibration pattern having known dimensions, - wherein for a multitude of known lens configurations of the 2D image acquisition unit (1) and the respective associated focal planes with their respective focal plane distance values ​​(a k ) respectively - the calibration pattern is moved through the detection range of the 2D image acquisition unit (1) and / or the 3D image acquisition unit (3) at different distances to the 3D image acquisition unit (3) and the 2D image acquisition unit (1), whereby two-dimensional and three-dimensional images of the calibration pattern are created in each case, - the two-dimensional images (Bj) are each divided into a number of, in particular square, segments, whereby the entries of the focus plane-object distance matrix correspond to the segments, - those segments of the individual two-dimensional images (Bj) are determined whose quality is recognized as optimal based on at least one predefined quality measure, in particular image sharpness, - for the segments of the individual two-dimensional images (Bj) identified as optimal, their position and / or location in the two-dimensional image (Bj) is mapped onto the three-dimensional image, and the object distance (d) between the 3D image acquisition unit (3) and the calibration pattern corresponding to the respective mapped position and / or location in the three-dimensional image is determined, and - the object distance (d) corresponding to each of the segments of the individual two-dimensional images (Bj) identified as optimal, as the focus plane distance value (a) k) the respective focal plane together with the position and / or orientation in the three-dimensional image and the associated lens configuration is stored in the focal plane-object distance matrix, so that the focal plane-object distance matrix for each segment of the two-dimensional image (Bj) contains the object distance (d ) as the focal plane distance value (a ). k ), which contains the position and / or location in the three-dimensional image and the associated lens configuration.

9. Method according to one of claims 2 to 8, characterized in that when using a fingerprint recognition arrangement (100) whose technical specification, in particular the camera parameters of the 2D image acquisition unit (1), is not known in advance, a scaling matrix is ​​used for scaling the finger images (FBj ) and / or fingerprints (FPjj), - wherein, to create the scaling matrix, a calibration of the fingerprint recognition arrangement (100) is carried out with a calibration pattern with known dimensions, - wherein for a multitude of lens configurations of the 2D image acquisition unit (1) and the respective associated focus planes with their respective focus plane distance values ​​(a k ) respectively - the calibration pattern is moved through the detection range of the 2D image acquisition unit (1) and / or the 3D image acquisition unit (3) at different distances to the 2D image acquisition unit (1) and the 3D image acquisition unit (3), whereby two-dimensional and three-dimensional images of the calibration pattern are created in each case, - the two-dimensional images (Bj) are each divided into a number of, in particular square, segments, whereby the entries of the scaling matrix correspond to the segments, - those segments of the individual two-dimensional images (Bj) are determined whose quality is recognized as optimal based on at least one predefined quality measure, in particular image sharpness, - for the segments of the individual two-dimensional images identified as optimal (Bj) - each of whose position and / or location in the two-dimensional image (Bj) is mapped onto the three-dimensional image, - the object distance (d) between the 3D image acquisition unit (3) and the calibration pattern corresponding to the respective depicted position and / or location in the three-dimensional image is determined, and - a calculation of the image resolution, especially in DPI, is performed based on the known dimensions of the calibration pattern, and - the image resolution determined in this way, especially in DPI, together with the object distance (d) belonging to the segment identified as optimal, as the focus plane distance value (a) k ) the respective focus plane and the position and / or location in the three-dimensional image is stored in the scaling matrix.

10. Method according to claim 8 or 9, characterized in that for the selection of the focus plane distance value based on the focus plane-object distance matrix and / or for the selection of the scaling based on the scaling matrix, a mean value of the relevant entries of the focus plane-object distance matrix and / or the scaling matrix is ​​determined if the image of the at least one finger (Fi) in the three-dimensional recording extends over several segments.

11. Method according to one of the preceding claims, characterized in that - that at least one three-dimensional image of several, in particular four, fingers (Fj) located in the detection range of the 3D image acquisition unit (3) is created, - that the image of the individual fingers (F) is determined in the three-dimensional recording, - that for each individual finger (F) the distance between the 3D image acquisition unit (3) and the respective finger (Fj) is determined based on the 3D coordinates of the individual points of the determined image of the respective finger (Fj) in the three-dimensional recording, - that the 2D image acquisition unit (1) is set to a focus plane corresponding to the determined distance of each finger (F) and that at least one two-dimensional image (B) of the respective finger (F) is created with the 2D image acquisition unit (1) in the previously set focus plane, - that the position and / or location of the determined images of the individual fingers (F) in the three-dimensional image are mapped onto the two-dimensional image (Bj) that was created in the focus plane previously set for the respective finger (Fj), and - that a finger image (FBjj) and / or fingerprint (FPjj) of the respective finger (Fj) is determined from the area of ​​the respective two-dimensional image (Bj) corresponding to the depicted position and / or location.

12. Method according to one of the preceding claims, characterized in that, based on the 3D coordinates of the individual points of the determined image of one or more fingers (Fj) in the three-dimensional recording, an analysis of the distance distribution between one or more fingers (Fj) and the 3D image recording unit (3) is carried out, wherein it is particularly provided that a statistical analysis of the distance distribution is carried out and / or that the distances along a profile running transversely to the fingertips are analyzed.

13. Method according to one of the preceding claims, characterized in that - that during the creation of three-dimensional images with the 3D image acquisition unit (3), predefined finger movements are performed with at least one finger (Fj) that is located in the respective detection area, - that the three-dimensional recordings created during the execution of the finger movements are compared with stored three-dimensional recordings of fingers that were created while the same movements were performed, in particular by applying neural networks, and - that the comparison leads to the conclusion that a real human hand is present.

14. Method according to one of the preceding claims, characterized in that the 3D image acquisition unit (3) is a time-of-flight sensor and that depth images are created as three-dimensional images from the 3D image acquisition unit (3), wherein it is particularly provided that depth images are created in the form of a 2D matrix whose individual entries include at least one object distance value (dj) for the individual object points in the recording area of ​​the 3D image acquisition unit (3).

15. Computer-readable storage medium comprising instructions which, when executed by a computer, in particular a processing unit of a fingerprint recognition arrangement (100), cause it to execute the method according to any one of claims 1 to 14.

16. Fingerprint recognition arrangement (100) for capturing fingerprints, in particular configured as a mobile portable terminal, preferably configured for carrying out a method according to one of claims 1 to 14 and / or comprising a data carrier according to claim 15, comprising a 2D image acquisition unit (1) and a 3D image acquisition unit (3) - wherein the 3D image acquisition unit (3) is designed to create three-dimensional images in the form of point clouds containing the 3D coordinates of the individual points, - wherein the 2D image acquisition unit (1) is capable of focusing on a multitude of focus planes, and - wherein the detection range of the 2D image acquisition unit (1) and the detection range of the 3D image acquisition unit (3) overlap each other, in particular are essentially identical, and - wherein a processing unit is provided which is designed to - to cause the 3D image acquisition unit (3) to create at least one three-dimensional image of at least one finger (Fi) located within the detection range of the 3D image acquisition unit (3), - to determine the image of the finger (Fj) in the three-dimensional image, - to determine the object distance (dj) between the 3D image acquisition unit (3) and the finger (F) based on the 3D coordinates of the individual points of the determined image of the finger (F) in the three-dimensional recording, - the 2D image acquisition unit (2) on at least one focal plane with a focal plane distance value (a) corresponding to the determined object distance (dj) and the position and / or orientation in the three-dimensional image k ) to set, and in this setting to create at least one two-dimensional image (Bj) of at least one finger (Fj) that is within the detection range of the 2D image capture unit (1), - to map the position and / or location of the determined image of the finger (F) in the three-dimensional image onto the two-dimensional image (B), and - to determine a finger image (FBjj) and / or fingerprint (FPj) from the area of ​​the two-dimensional image (B) corresponding to the depicted position and / or location.

Images