Method and device for biometric vascular recognition and/or identification

Abstract

The invention concerns a method and a biometric acquisition device for biometric vascular recognition and/or identification. The method comprising a step of capturing a plurality of veins images (116, 117, 118) of supposed subcutaneous veins (21) of a same inspecting portion (20) of a presented entity (2) from various converging orientations (113, 114, 115). The method further comprises a step of determine if said entity is a spoof based on estimated likelihood that said supposed subcutaneous veins within said plurality of veins images (116, 117, 118) are likely projections of solid veins (120).

Claims

1. Method of biometric vascular recognition and/or identification, comprising: presenting an entity to a biometric vascular recognition and/or identification system; capturing a plurality of veins images of supposed subcutaneous veins of a same inspecting portion of said entity from a plurality of converging orientations; estimating a likelihood that said supposed subcutaneous veins within said plurality of veins images are projections along said orientations of solid veins of the inspecting portion; deciding, based on said likelihood, whether the entity is a genuine body part or a spoof.

2. The method according to claim 1, wherein said step of estimating the likelihood comprises a step of comparing and/or matching a first pattern of supposed subcutaneous veins within one of said veins images with a second pattern of supposed subcutaneous veins within another of said veins images.

3. The method according to claim 1, wherein said step of estimating the likelihood comprises a step of determining a first portion within one of said veins images and a second portion within another one of said veins images, wherein the first and second portions comprise projections of a same solid vein portion along their respective converging orientations.

4. The method according to claim 1, comprising recognizing and/or identifying an individual based on one or more of said veins images and a reference image of subcutaneous veins.

5. The method according to claim 4, wherein recognizing and/or identifying an individual comprises separately comparing and/or matching each of the veins images within the reference image.

6. The method according to claim 5, wherein said step of recognizing and/or identifying an individual comprises steps of computing and/or estimating an image likelihood for each vein image by individually comparing each vein image with the reference image; and recognizing and/or identifying an individual based on an outcome computed from said computed image likelihoods.

7. The method according to claim 6, wherein computing the outcome comprises: quantizing the computed and/or estimated image likelihoods by comparing them with a given threshold, and applying a voting logic to said quantized likelihoods.

8. The method according to claim 4, wherein said step of recognizing and/or identifying an individual comprises: obtaining a biometric image by projecting the supposed sub-cutaneous veins each of each of the veins images on a same blending surface, the blending surface being a virtual flat plane; and comparing and/or matching said biometric image sub-cutaneous veins within the reference image.

9. The method according to claim 8, wherein said biometric image is obtained by blending the projected supposed sub-cutaneous veins together within a single image.

10. The method according to claim 1, wherein capturing a plurality of veins images comprises radiating said entity by a first electromagnetic radiation being a near-infrared radiation.

11. The method according to claim 10, wherein said step of deciding whether the entity is a genuine body part or a spoof also comprises: capturing an additional image of the entity being radiated by a second electromagnetic radiation that is distinct from the first electromagnetic radiation; and comparing and/or matching said additional image with one or more of said veins images.

12. The method according to claim 11, wherein said additional image is captured along one of said converging orientations.

13. The method according to claim 1, wherein a first and a second of said converging orientations are spatially arranged so as to form an angle in a range from 450 up to 135°.

14. The method according to claim 2, wherein said step of comparing and/or matching the first pattern with the second pattern comprises a step of assuming the entity being a fake if a similarity ratio between the said first and second pattern passes a pre-defined threshold ratio; said threshold ratio being 30%.

15. Biometric acquisition device for a biometric vascular recognition and/or identification system, the device comprising: a support for receiving an entity being a part of a body of an individual; a plurality of cameras configured to capture a plurality of images of subcutaneous veins of a same inspecting portion of the entity from various converging orientations, when said entity is received by said support; a computing unit configured to: estimate a likelihood that said supposed subcutaneous veins within said images are projections along said converging orientations of solid veins of the inspecting portion, and to deciding, based on said likelihood, whether the entity is a genuine body part or a spoof.

16. The biometric acquisition device according to claim 15, wherein the computing unit is configured, for estimating said likelihood, to: compare and/or match a pattern of subcutaneous veins within one image with a pattern of subcutaneous veins within another image.

17. The biometric acquisition device according to claim 15, wherein the computing unit is configured, for estimating said likelihood, to: determine a first portion within one of said veins images and a second portion within another one of said veins images, wherein the first and second portions comprise projections of a same solid vein portion along their respective converging orientations.

18. The biometric acquisition device according to claim 15, further comprising an irradiation source configured to provide a near-infrared radiation pointing to said entity; wherein said plurality of camera are configured to be sensitive to near-infrared radiation passing trough said entity.

19. The biometric acquisition device according to claim 15, wherein each camera is arranged to have a depth of field covering a subcutaneous region of the entity from a surface of the entity up to a depth of 5 mm, when the entity is received by said support, so as to be insensitive to objects located outside said subcutaneous region.

20. The biometric acquisition device according to claim 16, wherein the computing unit is configured to assume the entity being a fake if a similarity ratio between the said pattern passes a pre-defined threshold ratio; said threshold ratio being 30%.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention will be better understood with the aid of the description of an embodiment given by way of example and illustrated by the figures, in which:

(2) FIG. 1 shows a schematically view of a biometric vascular recognition and/or identification system;

(3) FIG. 2 shows a schematic representation of an exemplary embodiment of the proposed solution;

(4) FIG. 3 shows an embodiment of the proposed biometric acquisition device;

(5) FIG. 4 shows a schematic representation of a first variant of the proposed solution;

(6) FIGS. 5 and 6 show a schematic representation of a second variant of the proposed solution.

DETAILED DESCRIPTION OF POSSIBLE EMBODIMENTS OF THE INVENTION

(7) FIG. 1 shows an exemplary embodiment of a biometric vascular recognition and/or identification system 1 for recognize and/or identify an individual based on his vascular biometric of a part of his body. In this exemplary embodiment, the entity is a finger 2.

(8) The vascular biometric can also be a vein pattern of a hand, a wrist or a eye, or of another subcutaneous portion of (a preferably convex or at least flat part of) the body of the individual.

(9) The biometric vascular system 1 comprises:

(10) a biometric acquisition module 11 for acquiring a vascular biometric of a given entity (e.g. a finger 2, an hand, a wrist or an eye) of an individual;

(11) an anti-spoofing module 12 for deciding if the presented entity to the biometric acquisition module 11 is a genuine body part or a spoof;

(12) a recognition and/or identification module 13 for recognize and/or identify the individual based on the acquired vascular biometric so as to provide an indication 19 about an ID/recognition of the individual, when the presented entity is considered as a genuine body part (i.e. no spoof is detected).

(13) The biometric vascular recognition and/or identification system 1 can comprises a single device including the above listed modules 11, 112, 13. Alternatively, the biometric vascular recognition and/or identification system 1 can be implemented by a plurality of standalone, remotely connected devices.

(14) In the context of ID/recognition of individuals, a spoof is a falsified presentation of biometric pattern that is/was presented to the biometric vascular system 1 for forcing the system to identify and/or recognize another (absent and/or unwilling) individual.

(15) In the exemplary embodiment of the biometric vascular system 1, the system 1 is configured to capture a plurality of veins images 116, 117, 118 of supposed subcutaneous veins 21 (i.e. real or falsified subcutaneous veins) of a given entity being a distal portion of the finger 2, notably of the most distal phalanx of the finger. The term “image” indicates a two-dimensional representation of a particular sensed physical phenomenon.

(16) Each image is captured along a distinct and converging orientation 113, 114, 115, i.e. a non-parallel orientation tending to a same spatial region, even to a same point in the space (e.g. incident orientations). Preferably, the system 1 is configured so as the various converging orientations 113 converge within the biological entity (e.g. finger 2). Preferably, the converging orientations 113, 114, 115 spatially lay on a common plane being substantially perpendicular to the longitudinal (e.g. symmetry) axis of the given entity (e.g. distal portion of the finger), when the given entity is presented to the biometric system.

(17) Preferably, the images are simultaneously acquired or within a short delay (less than a 100 ms, preferably less than 10 ms) so as to avoid involuntary, or even voluntary, translations or rotations of the finger between images. In fact, these movements and/or rotations lead to a mismatch between captured portion of the finger within images as well as to a modification of the relative angular relationships between the converging orientations thereof. Such as mismatches can decrease the detection/identification rate of the system and/or the robustness against falsified presentations.

(18) In the exemplary embodiment of FIG. 2, the system 1 comprises a plurality of cameras 110, 111, 112 configured to provide the vein images 116, 117, 118 along a set of predefined converging orientations 123, 124, 125. Preferably, the converging orientations 113 of the set converge within the presented entity being considered as a finger or fingertips.

(19) According to the common sense, a camera is an optical device providing an image, i.e. a two-dimensional representation, of a particular sensed physical phenomenon, notably electromagnetically radiations. The orientation of the camera is defined as being the line 123, 124, 125 being substantially perpendicular to the surface of the image sensor 1100,1110,1120 that passes through the aperture of the optical system 1102,1112,1122 thereof (notably through the ideal pinhole of the camera).

(20) In the exemplary embodiment of FIG. 2, the veins images are provided by sensing near-infrared radiations passing through the entity being a distal portion of the finger. Sensed near-infrared radiation allows an acquisition of hemoglobin within subcutaneous veins while rejecting other disturbing elements such as skin, tissue, and bone.

(21) The illustrated biometric vascular recognition and/or identification system 1 is thus provided with an irradiation source 1192 configured to provide a near-infrared radiation so as to irradiate the entity (in this embodiment a distal portion of the finger), while the plurality of camera are configured to be sensitive to near-infrared radiation passing trough said portion of the biological entity.

(22) Advantageously, each camera can be further spatially arranged and configured so as to obtain a depth of field 1121 spatially coinciding with a 5 mm-depth subcutaneous region of the biological entity (i.e. a region spatially located between the surface of the entity up to a depth of 5 mm), preferably with a 3 mm-depth subcutaneous region. This arrangement permits to render the camera insensitive to (e.g. vascular) objects located outside said subcutaneous region.

(23) This configuration is advantageously used when the dimension and/or a relative position of the entity in the body is appropriate to let irradiations to pass through the entity, e.g. a finger, a hand, and a wrist of the individual.

(24) The captured vein images are then provided to the anti-spoofing module 12 for deciding (detecting) if the entity 2 being presented to the biometric acquisition module 11 is likely a fake (biological) entity, i.e. the entity is either a genuine body part or a spoof.

(25) The system 1 is configured so that each of the plurality of veins images comprises supposed subcutaneous veins of a same portion of the presented entity. This configuration permits the anti-spoofing module 12 to decide if the entity 2 is a genuine part or a spoof by checking if these veins are likely projections of hypothetically solid veins 120 (i.e. veins presenting a three dimensional structure or pattern) located inside the inspecting portion, along the converging orientations 113, 114, 115 of the veins images.

(26) This anti-spoofing approach provides a better robustness against malicious attempts based on presentations of falsified surfaces as these malicious presentations have limited capabilities to represented complex three dimensional patterns such as those of subcutaneous veins.

(27) The anti-spoofing module 12 is thus configured to decide (or determine) that the entity 2 being or not a spoof by estimating (e.g. computing) a likelihood that the supposed subcutaneous veins within the veins images 116, 117, 118 are projections of at least a solid vein 120 along the converging orientations 113, 114, 115 of the cameras.

(28) The likelihood can be a numerical probability, a (digital) hint, or a simple (digital) indication that the supposed subcutaneous veins within the veins images (likely) arose from either a falsified presentation of veins pattern or from genuine veins.

(29) The spoof likelihood can be estimated (e.g. computed) by taking advantage of the fact that projections of genuine subcutaneous veins along converging orientation would lead to observable, distinct two-dimensional patterns. The likelihood can be thus estimated or computed by comparing a pattern 126 of supposed subcutaneous veins within a first of the veins images 116 with a pattern 127, 128 of supposed subcutaneous veins within a second and/or a third of these veins image 117, 118. The aim of the comparison is to detect similarities between patterns of veins images of the same inspection portion along converging orientation, e.g. by matching these portions together, so as to detect a likely presentation of a falsified surface. The presented entity can be thus assumed to be a fake if the similarity ratio between the portions of distinct veins images (i.e. the spoof likelihood) passes a pre-defined threshold ratio (e.g. 30%).

(30) Alternatively or complementarily, the spoof likelihood can be estimated by determining that a portion 122 of a supposed subcutaneous vein within a first image 116 and a portion 122 of a supposed subcutaneous vein within a second and/or a third of these images 116, 117 are likely projections of a same hypothetically solid vein portion 121 (of the inspecting portion) along their respective converging orientations 113, 114,115.

(31) The solid vein portion 121 can be extrapolated, for example, from a disparity map or by triangulation using two images provided by two cameras having converging orientations.

(32) These approaches for estimating the likelihood, i.e. to decide that the presented entity is a fake, are particularly advantageous as they provide a detection of falsified presentations of vein pattern without reconstructing the entire 3D subcutaneous vein pattern of the inspecting portion, limiting thus the computational complexity required by the anti-spoofing function.

(33) When no spoof is detected or assumed, the system 1 is also be configured, e.g. by his recognition and/or identification module 13, to recognize and/or identify an individual based on his vascular biometric. An individual can thus be recognized and/or identified by comparing the already acquired subcutaneous veins within a reference image 138 and subcutaneous veins within at least a biometric image 130-134 obtained from one or more of said veins image 113, 114, 115.

(34) In the exemplary embodiment of FIG. 2, a biometric image generator 137 provides either one biometric image or a sequence of biometric images 130 by selecting one or more of the veins images that are successively compared with the reference image 138 so as to recognize and/or identify the individual.

(35) In a particular embodiment, the biometric images consist in one of the veins image. Preferably, the biometric images consist in the vein image provided by the camera placed in front of the entity (e.g. the flat of the finger, i.e. the convex surface of that is naturally devoid of fingernail). An individual can thus be recognized and/or identified based on an estimated or computed likelihood that the sub-cutaneous veins within this image likely match the subcutaneous veins within the reference image.

(36) The image reference 138 is advantageously a tagged image, i.e. provided with additional information concerning a (computer-readable) ID of an individual, and stored in local, shared or remote database 139. The database can be part of the biometric vascular system 1. Alternatively, the database can be located within another remotely connected device, such as a server.

(37) The comparison of subcutaneous veins within a reference image and sub-cutaneous veins within a biometric image can be based on (deformable or scale-invariant) template matching, veins features matching and/or by veins interest point detection and cross-verification.

(38) FIG. 3 shows an embodiment of a biometric acquisition device 5 for the biometric vascular recognition and/or identification system 1.

(39) The biometric acquisition device 5 comprises, in particular:

(40) a support 119 for receiving a given entity (in the exemplary embodiment the biological entity is a distal portion of a finger 2) of an individual;

(41) a plurality of cameras 110, 111, 112 configured to capture a plurality of images 116, 117, 118 of subcutaneous veins 21 of a same inspecting portion 20 of the presented entity 2) from a set of distinct and converging orientations 123, 124, 125, when the distal portion of the finger is received by the support; and

(42) a computing unit 50 configured to control and/or implement the biometric acquisition module 11, the anti-spoofing module 12 and/or the recognition and/or identification module 13.

(43) The computing unit 50 can be any type of electronic circuits such as a generic programmable circuit (e.g. an electronic microcontroller, a microprocessor, a DSP) or a dedicated electronic circuit.

(44) The support 119 is provided with surfaces destined to enter in contact with portions of the entity (finger) so as to provide a given spatially relationship between the biological entity and the cameras 110, 111, 112 when the entity is received by the support.

(45) Advantageously, the support 119 is dimensioned by considering a model of the biological entity. In case of a finger, the distal portion of the finger can be modelled (considered) as a cylinder having a given diameter representing a typical finger diameter of a given, adult population.

(46) The biometric acquisition device 5 comprises an irradiation source 1192 for providing a near-infrared radiation that point to the distal portion of the finger, notably oriented towards the fingernail 22. The illustrated irradiation source consists in a line of LEDs spatially aligned with the longitudinal axis of the finger when received by the support.

(47) The cameras of the biometric acquisition device 5 are thus configured to be sensitive to near-infrared radiation passing trough said distal portion of the finger.

(48) In order to avoid parasite irradiation, the support can be configured to obstruct irradiation provided by the source and passing outside the finger by using material not-transparent to NIR.

(49) Advantageously the biometric acquisition device 5 comprises a visible light source (not illustrated) configured to illuminate the distal portion of the finger when received by the support. The cameras are thus configured to be sensitive to visible radiation passing through or reflected on the distal portion of the finger.

(50) Biometric acquisition device can further be configured so as each camera has a depth of field covering a 5 mm-depth subcutaneous region of the most distal phalanx of the finger, preferably a 3 mm-depth subcutaneous region, when the distal portion of the finger is received by the support. Advantageously, the biometric acquisition device 5 can be configured so as the ⅗ mm-depth of field of each camera starts on the external surface of the cylinder representing the typical finger and used to dimension the support.

(51) In particular, the illustrated biometric acquisition device 5 comprises three distinct cameras radially located in front of the flat of the finger, when the distal portion is received by said support, so as to provide images of subcutaneous veins 21 from a set of three converging orientations 113, 114, 115.

(52) The three converging orientations 113, 114, 115 lay on a common plane being substantially perpendicular to the symmetry axis of the distal portion of the finger when received by the support. Advantageously, the biometric acquisition device 5 can be configured so as the three converging orientations converge to a point laying on the symmetry axis of the cylinder representing the typical finger and used to dimension the support.

(53) Preferably one of these cameras is located substantially in front of the flat of the finger (i.e. the camera is orientated toward and substantially perpendicular to the surface of the flat of the finger), when the finger is received by the support. The other two cameras are then laterally located with respect to this central located camera so as their orientations form an angle of 45°+/−10° with the orientation of the central positioned camera.

(54) The support 119 of the exemplary embodiment of FIG. 3 is in form of a cavity 1990 of the biometric acquisition device 5 having internal surfaces for receiving a distal end of a finger when inserted in the cavity. Advantageously, a NIR not-transparent, resilient material 1191 is placed on lateral sides of the internal walls of the cavity 1190, advantageously in form of foam-type material. This material is configured to enter in contact with lateral portion of the finger, when inserted in the cavity, so as to fill the empty space between the lateral surface of the finger and the lateral walls of the cavity.

(55) Advantageously, the computing unit 50 is configured to control the acquisition of the vein images and to estimate (e.g. to compute) the above described likelihood that the supposed subcutaneous veins within said images are likely projections of solid veins along said converging orientations.

(56) In particular, the computing unit 50 can be thus configured to compare and/or matching a pattern 126, 127, 128 of subcutaneous veins within one image with a pattern 126, 127, 128 of subcutaneous veins within another image. The computing unit 50 can be thus, alternatively or complementarily, configured to determine a portion 122, 123, 124 of supposed subcutaneous veins within each image that are likely projections of a same solid vein portion 121 along the set of converging orientations.

(57) The illustrated biometric acquisition device is a portable device advantageously provided with powering means, such as a power pack, for powering the active components thereof, notably the cameras, the computing unit and the radiations source(s).

(58) FIG. 4 shows a particular embodiment of the system 1, wherein an individual can be recognized and/or identified by weighting the likelihoods generated by individually comparing each veins image within the reference image.

(59) The biometric image generator 137 is thus configured to provide a sequence of biometric images 131-133 being a concatenation of the veins images 116, 117, 117 provided by the biometric acquisition module 11 and/or the anti-spoofing module 12. The illustrated embodiment further comprises a decision unit 136 configured to provide the indication 19 about an ID/recognition of the individual based on these comparisons.

(60) The decision unit 136 can provide an ID/recognition of the individual by weighting the single likelihoods provided by comparing each biometric image within the reference image.

(61) The decision unit 136 can, for example, provide an ID/recognition of the individual based on the maximal value and/or the minimal value of the single likelihoods provided by the biometric images.

(62) The decision unit 136 can, for example, provide an ID/recognition of the individual based on a voting logic applied to (binarized) likelihoods, e.g. 2 over 3. In this configuration, the likelihoods provided by the biometric images are binarized (i.e. represented in a two-level system) by individually comparing them with a thresholding reference. An individual can thus be recognized and/or identify by applying a voting logic to these binarized or quantized likelihoods, i.e. binary likelihoods. This approach permits to increase the recognition/identification rate and the robustness against falsified presentation.

(63) FIGS. 5 and 6 shows another particular embodiment of the system 1. In this embodiment, an individual can be recognized and/or identified by comparing the reference image 138 with a single biometric image 134 being obtained by processing the veins images 116, 117, 117 provided by the biometric acquisition module 11 and/or the anti-spoofing module 12.

(64) The biometric image generator 137 is thus configured to provide the biometric image 134 by:

(65) projecting of the supposed sub-cutaneous veins of each of the veins images 116, 117, 118 on a same blending surface 135; and then by

(66) blending the projected supposed sub-cutaneous veins 1350 together within a single image 134.

(67) The blending surface is a virtual flat plane, preferably spatially located so as to substantially touch or cross the inspecting portion 20 of said entity 2. Advantageously, the blending surface is parallel to the symmetry axis of the cylinder representing the typical finger and used to dimension the support.

(68) A step of correcting the optical distortions and aberrations introduced by each camera into the provided vein images can advantageously precede the step of projecting of the supposed subcutaneous veins on the same blending surface. A corrected image of the vein image can be thus created by warping it according to the intrinsic parameters of each camera so as to obtain a near-ideal pinhole image of the supposed sub-cutaneous veins.

(69) The blending operation can comprises a step of fusing the elements of the images, at the level of either the image's pixel or image's displayed objects.

(70) Alternatively, the blending operation can be executed, for example, by means of the Laplacian pyramid blending algorithm. The obtained blended image can thus be corrected using the contrast limited adaptive histogram equalization (CLAHE) method.

LIST OF REFERENCE NUMERALS

(71) 1 Vascular recognition or identification system 11 Biometric vein acquisition module 110,111,112 Camera 1100,1110,1120 Image sensor 1121 Depth of field 1102,1112,1122 Lens/optical system 113,114,115 Image acquisition orientation 116,117,118 Veins image of supposed subcutaneous veins 119 Support for a distal portion of a finger 1190 Cavity 1191 Resilient and foam-type material 1192 NIR source 12 Anti-spoofing module 120 Solid/3D veins 121 Portion of a solid/3D vein 122,123,124 Portion of supposed subcutaneous vein 126,127,128 Pattern of supposed subcutaneous veins 13 Recognition and/or identification module 130-134 Biometric image 135 Blending surface 1350 Projected sub-cutaneous veins 1353,1354,1355 Image with projected sub-cutaneous veins 136 Decision unit 137 Biometric image generator 138 Reference image 139 Database with references image 19 ID or recognition indication 2 Finger of an individual 20 Portion of the finger 21 Subcutaneous vein 22 Fingernail 5 Biometric acquisition device 50 Processing unit