Cross-terminal hand back vein living body detection method based on illumination response characteristics

Abstract

The invention discloses a cross-terminal hand back vein living body detection method based on an illumination response characteristic. The method comprises the following steps: S1, respectively collecting a multi-light-intensity vein image sequence of a target hand back and a to-be-detected hand back through a first terminal and a second terminal; s2, performing hand back contour extraction, direction adjustment and image normalization preprocessing on the two terminal images to ensure that the sizes and the orientations of the image contours are consistent; s3, extracting a hand back region of interest, and synchronously realizing region scaling and noise reduction by adopting mean filtering; s4 and S5 are combined with the camera response function and the inverse function of each terminal, illumination response offset values of pixel points in the region of interest are calculated, and a first offset feature map and a second offset feature map are generated respectively; and S6, selecting a matching template from the second offset feature map, performing sliding matching with the first offset feature map, calculating similarity through a Pearson's correlation coefficient, and comparing the similarity with a preset threshold to judge whether the to-be-detected hand back is a living body or not. According to the invention, cross-terminal hand back vein in-vivo detection with high adaptability and high accuracy is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biometric recognition, and particularly relates to a cross-terminal dorsal hand vein living body detection method based on illumination response characteristics. BACKGROUND

[0002] In the process of biometric recognition technology upgrading to the third generation of in-vivo features, dorsal hand vein recognition has been widely used in financial payment, access control security, public service and other fields due to its advantages of high concealment, high stability and living body correlation. With the diversified deployment of terminal devices (such as the same user needs to switch between mobile phone, computer and offline access terminal), cross-terminal dorsal hand vein living body detection has become a core technical bottleneck. The light source parameters (such as wavelength, light intensity distribution) and camera imaging performance (such as resolution, sensor sensitivity, camera response function) of different terminals are significantly different, resulting in a huge difference in gray scale distribution and texture definition of the vein images collected by the same dorsal hand in different terminals. For example, when the user registers a living dorsal hand vein image at terminal A in the office, the terminal B in the mall cannot determine whether the currently collected dorsal hand is a living body through traditional gray scale comparison when it is first verified at terminal B, because terminal B lacks the living body baseline data of the user. If a false body attack (such as a silicone imitation dorsal hand or a thin film printed with a vein pattern) or a video replay attack occurs at this time, the terminal will mistakenly determine the non-living body as a legal user, and all subsequent recognitions based on the terminal will continue to fail, which seriously threatens information security and property safety. In addition, the hardware differences between terminals also cause problems such as geometric distortion and uneven light intensity of the vein images, further increasing the difficulty of living body detection.

[0003] Traditional dorsal hand vein living body detection technology mainly focuses on two major ideas of "single-terminal baseline comparison" and "false body feature modeling", and both the technical solutions and limitations are very significant. In the single-terminal baseline comparison scheme, the technical logic is to pre-collect and store the user's living dorsal hand vein image on each terminal, and then compare the gray scale distribution and texture structure of the currently collected image with the baseline image stored in the terminal during subsequent detection. If the similarity exceeds the threshold, it is determined to be a living body. The advantages of this scheme are that the technology is simple to implement and does not require complex cross-device calibration, and it has high accuracy in a single-terminal scenario (such as enterprise exclusive access control). However, its disadvantages are extremely prominent: first, the cross-terminal applicability is extremely poor, and the user needs to register living body information on each terminal, which is tedious and has poor user experience; second, the terminal migration cost is high, and if the terminal hardware is replaced (such as camera upgrade, light source replacement), all user living body baseline data needs to be re-collected, and historical data cannot be reused; third, the attack resistance is limited, and if the baseline image stored in the terminal is leaked, the attacker can easily bypass the detection by forging a false body similar to the baseline image.

[0004] In the prosthesis feature modeling scheme, the technical logic is: by collecting a large number of static vein images of different materials of the prosthesis (such as silicone, resin, and paper), the common features of the prosthesis (such as gray uniformity, light reflectivity abnormal value, and texture edge blur) are extracted, and a prosthesis detection model is constructed. When detecting, it is determined whether the current image conforms to the prosthesis feature to determine the living body. The advantage of this scheme is that it does not need to be pre-registered by the user, and can be directly used in unknown terminals. However, its disadvantages are also fatal: first, the defense capability against unknown material prosthesis is weak. If the attacker uses a new material (such as biological gel and 3D printed bionic skin) that the model has not been trained on to make a prosthesis, the model cannot identify it; second, the model has poor generalization ability. The hardware differences of different terminals will cause the appearance of the prosthesis features to be different (such as the same silicone prosthesis showing high reflectivity on terminal A and low reflectivity on terminal B due to different light source wavelengths), and the detection accuracy of the model on new terminals will be greatly reduced; third, the calculation cost is high. The model needs to be continuously updated to cover new prosthesis types, and the maintenance difficulty and cost increase significantly with the application time.

[0005] With the increasing demand for cross-terminal identification, existing technologies have begun to break through the single-terminal limit, mainly forming two types of improved schemes: "texture structure invariant feature extraction" and "multi-modal fusion detection". In the texture structure invariant feature extraction scheme, the technical core is to extract the structural features (such as vein branch points, intersection points, and strike angles) in the vein texture that are not affected by light intensity and resolution through image preprocessing (such as scale-invariant feature transform SIFT and distortion correction algorithm). Cross-terminal detection only compares these structural features, ignoring gray differences. For example, some technologies extract the skeleton structure of the vein through edge detection algorithms, and then calculate the invariants such as Hu moment and Fourier descriptor of the skeleton, to realize texture matching between terminals. This scheme solves the problem of cross-terminal texture comparison to some extent, but its essence still focuses on "identification" rather than "liveness detection" - even if the extracted structural features are consistent with the known user, it cannot determine whether the current collected image comes from a living body, and is still vulnerable to prosthesis attacks (prostheses can imitate real vein texture structures).

[0006] In the multi-modal fusion detection scheme, the technical logic is: combining multiple biometric characteristics of the back of the hand (such as vein texture + palm temperature, vein texture + pulse fluctuation), collecting multi-dimensional data through multiple sensors, and comprehensively judging whether it is a living body. For example, some terminals are equipped with near-infrared cameras (collecting veins) and infrared temperature sensors (collecting back of hand temperature) at the same time. If the current collected back of hand temperature is within the normal body temperature range of the human body (36-37°C) and the vein texture matches, it is determined to be a living body. The advantages of this scheme are strong anti-attack ability, and it is difficult for a fake body to imitate multiple biometric characteristics at the same time. However, the disadvantages are high hardware cost, the need for additional deployment of temperature, pulse, and other sensors on the terminal, and the inability to adapt to existing ordinary vein recognition terminals. At the same time, the collection and fusion of multi-modal data increase the computational load of the terminal, leading to increased detection delay (usually more than 500ms), which cannot meet the needs of real-time recognition scenarios (such as subway gates and fast payment). In addition, the cross-terminal calibration of multi-modal data is more difficult - the accuracy of temperature sensors and the sensitivity of pulse detection vary between different terminals, and the data cannot be directly reused, and the core problem of cross-terminal living body detection has not been fundamentally solved.

[0007] Chinese patent (CN117152802A) discloses a finger vein recognition and detection method and device that fuses texture and living body features. The technical solution revolves around finger veins, and the core is to obtain different light intensity finger vein images on the same terminal through optical components such as beam splitters and mirrors, construct a transmission model, calculate the finger's transmission light offset image based on the camera's light response curve, and use it as a "texture + living body" fusion feature for judgment. Although the light response characteristic is introduced, it is only applicable to a single terminal and cannot solve the cross-terminal problem. The differences in optical components (beam splitter angle, mirror position) and camera response functions between different terminals will cause the calculation benchmarks of the transmission light offset image to be inconsistent, resulting in a large error when comparing across terminals, and making it impossible to accurately determine the living body. Moreover, the optical structure and processing logic designed for finger veins cannot be directly migrated to the back of the hand vein detection scenario. The imaging range and texture distribution complexity of the back of the hand vein are fundamentally different from those of the finger vein, and this technology cannot adapt to the cross-terminal detection needs of the back of the hand vein. SUMMARY

[0008] Based on the above technical problems, the present application discloses a cross-terminal back of the hand vein living body detection method based on the light response characteristic, specifically including:

[0009] S1, collecting a multi-light intensity living body back of the hand vein image sequence of a target and to be detected through a first terminal; the first terminal and the second terminal are different back of the hand vein recognition terminals, and the target back of the hand vein image collected by the first terminal has been confirmed to be a living body image;

[0010] S2. Preprocess the multi-intensity live dorsal hand vein image sequence acquired by the first terminal and the multi-intensity dorsal hand vein image sequence acquired by the second terminal.

[0011] S3. Extract the regions of interest from the preprocessed images of the backs of the hands of the two terminals, scale the regions of interest, and reduce image noise.

[0012] S4. Using the camera response function of the first terminal and combining the multi-intensity live hand back vein image sequence acquired by the first terminal, calculate the illumination response offset value of each pixel in the target hand back region of interest under the first terminal relative to the camera response function curve of the first terminal, and generate the first offset feature map.

[0013] S5. Using the camera response function of the second terminal and combining the multi-intensity back vein image sequence of the hand to be detected acquired by the second terminal, calculate the illumination response offset value of each pixel in the region of interest of the back of the hand to be detected under the second terminal relative to the camera response function curve of the second terminal, and generate the second offset feature map.

[0014] S6. Calculate the similarity between the first offset feature map and the second offset feature map, compare the obtained similarity result with the preset threshold, and determine whether the back of the hand to be detected is a living body.

[0015] Preferably, the step of acquiring the target and the multi-intensity live dorsal hand vein image sequence to be detected through the first terminal specifically includes:

[0016] The first terminal acquires a multi-intensity live hand dorsal vein image sequence with a duty cycle of { }collection Zhang's hand dorsal vein images are denoted as an image sequence. { },in This represents the total number of images captured by the first terminal. For the first terminal to collect the first Light source duty cycle when taking an image For the first terminal in duty cycle Images of the veins on the back of the hand acquired at that time, and All images in the dataset are free from overexposure and underexposure and meet pixel-level registration requirements.

[0017] The second terminal acquires a multi-intensity image sequence of the dorsal veins of the hand to be detected, specifically: the second terminal uses a duty cycle of { }collection The images of the veins on the back of the hand to be detected are denoted as the image sequence. { },in This represents the total number of images acquired by the second terminal. For the second terminal to collect the first Light source duty cycle when taking an image For the second terminal in duty cycle Images of the veins on the back of the hand acquired at that time, and All images in the dataset are also free from overexposure and underexposure and meet pixel-level registration requirements;

[0018] The light intensity of the light source of the first terminal at the minimum unit duty cycle is denoted as: The first terminal has a duty cycle of The actual light intensity of the light source at that time is denoted as ,satisfy The light intensity of the light source at the second terminal with a minimum unit duty cycle is denoted as . The second terminal has a duty cycle of The actual light intensity of the light source at that time is denoted as ,satisfy .

[0019] Preferably, the S2 preprocessing specifically includes:

[0020] S2.1 Extract the first terminal image sequence respectively Second terminal image sequence The outline of the back of the hand in each image is determined by a contour detection algorithm to define the boundary range of the back of the hand in the image.

[0021] S2.2 Adjust the image sequence of the second terminal The direction of the back of the middle hand makes The back of the middle hand facing and The back of the hand in the target image is always facing the same direction to ensure consistent hand posture across different devices.

[0022] S2.3 Calculate the size ratio of the back-of-hand contours of the first terminal and the second terminal: Let the length of the minimum bounding rectangle of the back-of-hand contour extracted by the first terminal be... Width is The length of the minimum bounding rectangle of the hand back contour extracted by the second terminal is Width is Then the length ratio Width ratio ;

[0023] S2.4. The first terminal image sequence is processed according to the stated size ratio. Perform scaling processing so that the scaled result is... The dimensions of the back of the hand and The dimensions of the back of the hand are consistent, thus completing the size normalization of images across terminals.

[0024] Preferably, in step S3, the region of interest is extracted by using the center of the smallest bounding rectangle of the preprocessed back of the hand contour. With the origin as the minimum bounding rectangle, the length of the outline is... Width The length and width of the region of interest are respectively taken as , , These are preset scaling factors, and the coordinate range of the region of interest is... ;

[0025] The region of interest is scaled. Let the original image of the region of interest be... The superimposed noise is The noisy image is Original image noise obey ; Use window size Step size is Mean filtering, Given a preset window size, the filtered image pixel values ​​are: The first term represents the mean of local blocks in the original image, and the second term represents the mean of local blocks in the noise. To output the filtered image in coordinates The pixel value at that location, and the mean of the local noise block follows The normal distribution To reduce the noise variance of the image before scaling, we aim to achieve the effect of scaling while simultaneously reducing the noise variance.

[0026] Preferably, the camera response function of the first terminal in S4 is denoted as... ,satisfy ,in The light intensity received by the camera sensor. The pixel grayscale values ​​output by the camera; Defined as the new camera response function of the first terminal ,Right now ;

[0027] For any pixel point in the region of interest on the back of the target hand under the first terminal Its reflectivity is denoted as The light intensity of this pixel under illumination by the first terminal light source Satisfy the formula By combining the camera response function, the grayscale value of the pixel is calculated. The formula is:

[0028]

[0029] in, The duty cycle of the light source. This represents the light intensity corresponding to the smallest unit duty cycle of the first terminal light source.

[0030] Preferably, the calculation of the illumination response offset value in step S4 specifically involves:

[0031] The first terminal new camera response function The inverse function is denoted as The first terminal will collect the first data. Pixels in the image grayscale value Substitute into the inverse function Combined with formula Rearranging the terms, we get:

[0032]

[0033] The average value of the results is taken to obtain the pixel points. The estimated value of the illumination response shift is given by the formula:

[0034]

[0035] The first offset feature map is obtained by assembling the estimated illumination response offsets of all pixels within the region of interest on the back of the target hand into a matrix. ,and .

[0036] Preferably, the camera response function of the second terminal in S5 is denoted as... ,satisfy ,in The light intensity received by the camera sensor. The pixel grayscale values ​​output by the camera; Defined as the new camera response function of the second terminal ,Right now ;

[0037] For any pixel in the region of interest on the back of the hand to be detected under the second terminal Its reflectivity is denoted as Calculate the light intensity of a pixel under illumination by the second terminal light source. The formula is: The grayscale value of a pixel is calculated by combining the camera response function, using the following formula:

[0038]

[0039] in, The duty cycle of the light source. This represents the light intensity corresponding to the minimum unit duty cycle of the second terminal light source.

[0040] Preferably, the calculation of the illumination response offset value in step S5 specifically involves:

[0041] The second terminal new camera response function The inverse function is denoted as The second terminal will collect the first Pixels in the image grayscale value Substitute into the inverse function Combined with formula Rearranging the terms, we get:

[0042]

[0043] The average value of the results is taken to obtain the pixel points. The estimated value of the illumination response shift is given by the formula:

[0044]

[0045] The second offset feature map is formed by assembling the estimated illumination response offsets of all pixels within the region of interest on the back of the hand. ,and .

[0046] Preferably, in step S6, the similarity between the first offset feature map and the second offset feature map is calculated specifically using the Pearson correlation coefficient combined with sliding template matching.

[0047] Second offset feature map Select the central region as the matching template , The size is , Template height This is the template width;

[0048] make First offset feature map Slide up pixel by pixel, when Top left corner is located of When in position, The covered sub-region is denoted as Calculate at this time and Pearson correlation coefficient:

[0049]

[0050] in, for The average pixel value, for The average pixel value;

[0051] The correlation coefficients of all sliding positions constitute the matching matrix. ,Pick The maximum value in the result is used as the similarity result. ,Right now .

[0052] Preferably, the S6 preset threshold The value is set to 0.6, and the determination rule is as follows:

[0053]

[0054] When similarity results Greater than or equal to When the second terminal collects data on the back of the hand to be detected, it determines that the person is alive; when Less than At that time, the back of the hand to be tested was determined to be a prosthesis, and the liveness detection of the back veins of the hand across the terminal was completed.

[0055] Compared with the prior art, the technical solution of this application has the following technical effects:

[0056] This invention solves the problem of liveness detection adaptation caused by hardware differences between different terminals. By exploring the cross-terminal consistency of illumination response characteristics, the liveness illumination response of unknown terminals (such as the second terminal) can be estimated by utilizing the liveness illumination response characteristics of the back of the hand of a known terminal (such as the first terminal). There is no need to pre-register or train a model on the second terminal, thus completely getting rid of the limitations of terminal hardware on liveness detection and achieving efficient cross-terminal adaptation.

[0057] This invention uses the inherent light response of the back of the hand as the core criterion for judgment. The light reflection and absorption characteristics of the veins and non-vein areas on the back of the hand of a living person are stable and unique. It is difficult for a prosthesis to imitate its pixel-level light response offset distribution. By calculating the similarity of the offset feature map, the living person can be judged. No matter what unknown material the prosthesis is made of, as long as its light response is different from that of a living person, it can be accurately identified, which greatly improves the active defense capability and avoids continuous identification errors caused by prosthesis attacks.

[0058] This invention uses the live offset feature map registered by the user on any terminal as a "global baseline" that can be directly reused on other terminals without repeated acquisition. At the same time, the technology does not rely on specific optical components or additional sensors on the terminal, and can be adapted to various terminals with multi-intensity adjustment functions, reducing hardware modification and maintenance investment and significantly reducing the cost of large-scale deployment of the back of the hand vein recognition system.

[0059] This invention achieves simultaneous scaling and noise reduction of the region of interest through mean filtering, simplifying the preprocessing process. The offset feature map calculation and similarity comparison logic are concise, requiring no complex model training and iteration. The overall computational load is small, enabling rapid liveness detection and meeting the speed requirements of real-time recognition scenarios such as subway turnstiles and financial payments. While ensuring accuracy, it also considers the user experience.

[0060] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the preferred embodiments of this application are described in detail below with reference to the accompanying drawings.

[0061] The above and other objects, advantages and features of this application will become more apparent to those skilled in the art from the following detailed description of specific embodiments in conjunction with the accompanying drawings. Attached Figure Description

[0062] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0063] Based on the description of the figures and their corresponding technical content in the document, the titles of the figures are as follows:

[0064] The following are accurate and detailed names proposed for each attached diagram, based on the technical solution, which not only reflect the process / data content but also relate to the technical scenario:

[0065] Figure 1 This is a schematic diagram showing the translation relationship between the camera response function curve across terminals and the illumination response curve of two points in a live hand image.

[0066] Figure 2 This is a flowchart of cross-terminal hand dorsal vein detection based on light response characteristics. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. In the following description, specific details such as specific configurations and components are provided merely to help fully understand the embodiments of this application. Therefore, those skilled in the art should understand that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. In addition, for clarity and brevity, descriptions of known functions and structures are omitted in the embodiments.

[0068] It should be understood that the phrase "an embodiment" or "this embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "an embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0069] Furthermore, reference numerals and / or letters may be repeated in different examples within this application. Such repetition is for the purpose of simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or settings discussed.

[0070] In this article, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article describes another type of relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the related objects before and after it are in an "or" relationship.

[0071] In this article, the term "at least one" is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, "at least one of A and B" can mean: A exists alone, A and B exist simultaneously, or B exists alone.

[0072] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion.

[0073] Example 1

[0074] This embodiment uses two different sets of dorsal hand vein recognition terminals (first terminal and second terminal) as examples to describe in detail the cross-terminal dorsal hand vein liveness detection method based on illumination response characteristics of this application. Specifically:

[0075] A hand back vein image was registered and acquired in terminals A and B respectively. P multi-intensity live hand back vein image sequences were acquired in terminal A, denoted as P. The B terminal acquired Q images of the dorsal veins of the hand with high light intensity, denoted as Q. It can now be confirmed that the image captured on terminal A is of a live hand, but this hand has never been captured or registered on terminal B. Now, it is necessary to analyze the live multi-intensity image sequence captured on terminal A. Extract the live light response features of the back of the hand to determine the first time the image was collected on the B terminal. Is it a living organism?

[0076] The response function of each pixel photoelectric sensor in a camera is identical and consistent, called the Camera Response Function (CRF), denoted as . The value can be calculated in advance; this property is stable and invariant, requiring only one calculation. The function is monotonically increasing and has an inverse function. The inverse function is also a monotonically increasing function. Input is light intensity The output is pixel grayscale. Typically, grayscale is quantized to 8-bit depth. In cross-device hand dorsal vein acquisition, The CRF of the terminal is denoted by equation (1):

[0077]

[0078] In acquisition terminal A, let's assume that one pixel in the acquired image of the back of the hand... grayscale value At that time, the light intensity received by the sensor pixel is , denoted as (2):

[0079]

[0080] For a point on the back of a living hand, its reflectance is , Each point on the back of the hand has a fixed and stable reflectivity, so there's no need to... The upper right corner is labeled A or B because the reflectivity of the back of the hand is the same in different terminals, and the light source emits uniform light intensity. Under the illumination, it reflects onto the camera sensor. Light intensity on pixels It can be written as formula (3):

[0081]

[0082] In the formula This indicates that the duty cycle of the uniform light source in terminal A is... The intensity of light emitted at that time Its actual duty cycle is Let the uniform light source in terminal A have a minimum unit duty cycle of . The light intensity at that time was Then equation (3) can be written as equation (4):

[0083]

[0084] Substituting equation 4 into terminal A in equation (1), we obtain equation (5):

[0085]

[0086] Light intensity Taking the logarithm, we get equation (6):

[0087]

[0088] here Is the bottom The logarithm of equation (6) is used to expand the logarithm to obtain equation (7):

[0089]

[0090] Will Considered as a new camera response function for * After substitution, equation (7) can be transformed into:

[0091]

[0092] In equation (8), the intermediate term Treated as an independent variable, it varies with the constant current source duty cycle of the driving light source. The change, that is, the change in light intensity, The term is related to the light source characteristics of terminal A and is a constant term; Item at point Once the position of the back of the hand is determined, it is also a constant term, so this point... The light response curve is The shift of the function curve is obviously due to the reflection coefficient. Less than 1, If it is negative, the back of the hand point The light response curve is given by equation (8).

[0093] The following is a brief analysis of the translation relationship between the camera response function curve and the illumination response curve of the back of the hand pixels. This analysis was conducted when a mirror-like back of the hand was directly illuminated by a uniform light source in the terminal, with a reflectivity of 1 (here, the reflectivity of each pixel on the mirror-like back of the hand is 1, i.e.) This is equivalent to a light source shining directly on a camera, allowing direct observation of the camera's response function curve. With a uniform light source, the light intensity is the same at every point, and the camera response function of each photoelectric sensor is also the same. Therefore, the position (i,j) can be omitted. The pixel's illumination response curve under direct light can be considered as the camera's response function curve. Substituting into equation (8), we obtain the camera response function curve (9):

[0094]

[0095] Will Treating it as the independent variable, comparing the right side of equation (9) with that of equation (8), we find that a certain pixel after the light is reflected on the back of the hand... Light response curve It is the camera response function curve. Shift to the right distance,( (For the absolute value calculation), here we denote the rightward translation of this point as... As in equation (10):

[0096]

[0097] Obviously in equation (10) So, the back of the hand The light response curve can be represented by equation (11):

[0098]

[0099] Similarly, when using terminal A to acquire a live image of the veins on the back of Zhang San's hand, another pixel on that image... The light response curve can be represented by equation (12):

[0100]

[0101] Comparing the translation terms of equations (11) and (12), the intermediate term... As the independent variable, the term and This is a constant term, also known as the offset, between these two locations in terminal A. and The shift difference between the light response curves is denoted as :

[0102]

[0103] These two points and reflectivity and The values ​​remain unchanged in terminals A and B, and are eliminated during the translation distance calculation. Similarly, when switching to terminal B, these two points... and The translation distance between the light response curves can also be eliminated. The term is (14):

[0104]

[0105] Clearly, the following equation (15) holds true:

[0106]

[0107] Equation (15) illustrates two points. The translation distance of the illumination response curve and the camera response function Shape-independent, and the minimum unit luminous intensity of a uniform light source in the terminal. , It's irrelevant, such as Figure 1 As shown, the camera response function curves of terminals A and B are as follows: Figure 1 In , The light response curves of point A and point B in terminals A and B are respectively and , The light response curves of point A and point B in terminals A and B are respectively and , Figure 1 Translation distance .

[0108] A sequence of vein images on the back of a live hand was acquired in terminal A. Estimate each pixel Relative to camera response function translation of the curve The offset distance between any two points in the illumination response curves of the comprehensive formulas (13)(14)(15) is only related to the reflectivity of these two points. This is achieved by using the live hand back vein image acquired in terminal A. Calculate the offset of the illumination response curve of each point within the contour of the back of the hand relative to the camera response function curve. This can be denoted as a feature offset map. Similarly, the multi-intensity image sequence of the veins on the back of the hand acquired in the B terminal was used. It also calculates the illumination response offset map for each point, denoted as . ,like Figure 1 Two points in The difference in the offset values ​​of the illumination response curves is a constant, and equation (16) holds:

[0109]

[0110] In formula (16) To and and The relevant constants, due to the presence of noise, and They will not differ by a constant in every situation. However, it is possible to perform a similarity measurement of the offset map across terminals, and then take a fixed threshold for the similarity value to determine whether the back of the hand registered on terminal B is a live body. Alternatively, the offset values ​​of some points with relatively stable reflectivity can be selected as live feature vectors for comparison.

[0111] This embodiment details how unknown terminals can be determined using the live offset feature map of the first terminal, which greatly improves cross-terminal adaptability. It can accurately identify various prostheses by relying on the inherent light response characteristics of the back of the live hand, and enhances anti-attack capabilities. At the same time, scaling and noise reduction are completed simultaneously through mean filtering, which simplifies the process and ensures real-time performance, significantly reduces terminal deployment and maintenance costs, and adapts to the needs of multiple scenarios.

[0112] Based on Example 1, this example details the translation relationship and offset consistency verification of the dorsal hand vein illumination response curve across terminals, specifically as follows:

[0113] Taking two sets of hand-back acquisition terminals A and B as examples, their camera response functions are denoted as: and The light intensity distribution illuminating the back of the hand is uniform. Terminals A and B can rapidly switch light intensities to simultaneously acquire pixel-level registered sequences of hand vein images under corresponding light intensities. These hand vein images should be free of overexposure and underexposure. Let's assume Zhang San's hand is illuminated in terminal A with a duty cycle of { P images of dorsal hand veins were collected. A ={ }, the same hand on terminal B with a duty cycle { Q images of dorsal hand veins were acquired. B ={ These images of veins on the back of the hand were neither overexposed nor underexposed, and were pixel-level registered.

[0114] In practical applications, the flowchart for cross-terminal dorsal hand vein liveness detection is as follows: Figure 2 As shown, Figure 2 Step (1) is the data collection process. Figure 2 In step (2), the problem of different resolutions and different actual pixel sizes of the back of the hand is encountered when collecting data across terminals. It is necessary to calculate the outline of the back of the hand collected by terminals A and B and adjust the direction so that the back of the hand faces the same direction across terminals. Figure 2 In step (3), the ratio of the back of the hand contour of the A and B terminal cameras is estimated using the back of the hand contour. Using this ratio, the back of the hand contour image captured by the A terminal can be scaled to the back of the hand contour size of the B terminal, i.e., normalized size. Figure 2After unifying the size of the back of the hand image in step (4), the region of interest (ROI) of the back of the hand can be extracted. The ROI is a rectangle, and its center can be the center of the smallest bounding rectangle of the back of the hand outline. Its size is about 80% of the back of the hand outline. Figure 2 Step (5) is due to the limited number of hand dorsal vein image sequences acquired, typically 5-7 images, and the noise in near-infrared hand dorsal vein angiography. The area is relatively large, so we need to reduce the computational load. Here, we further reduce the width and height of the ROI region by 1 / 4 and assume noise... It follows a mean of 0 and a variance of . The Gaussian distribution; the acquired noisy image is ,in , To produce a noise-free image; when scaling the width and height of the ROI region, a step size of 4 is used. Mean filtering, as shown in equation (17):

[0115]

[0116] Equation (17) expanded into equation (18):

[0117]

[0118] In equation (18), the first term is the mean of a local patch in the real image, and the second term, when accumulating noise, is denoted as noise because the noise is independently and identically distributed. Its distribution satisfies This means that during the process of reducing the size of the vein image on the back of the hand, the influence of noise was reduced to 1 / 16 of the original.

[0119] Figure 2 Step (6) involves using the grayscale values ​​from multiple illumination imaging operations at each point to calculate the offset of the illumination response curve for that point. The estimation; every point without overexposure or underexposure can be obtained through the inverse function of the camera response function. Estimate its offset Equation (8) is substituted into the image points acquired by the P light intensity acquisitions. This yields the following equation (19):

[0120]

[0121] As can be seen from equation (10) in the embodiment, a pixel on the back of the hand The offset of the illumination response curve relative to the camera response function curve It is its reflectivity Taking the absolute value of the logarithm directly solves for this value, which is difficult. In this invention, the sum of the logarithm of the logarithm of the unit illuminance can be used as the actual offset characteristic value. The left side of equation (19) is substituted into the inverse function of the camera response function. Eliminate on the right side We obtain the following equation (20):

[0122]

[0123] Rearranging terms, we get the following equation (21), with the left side... This is the offset estimate that can be calculated by the present invention;

[0124]

[0125] because Equation (21) can calculate P offset values. Summing and averaging these values ​​can provide a more accurate estimate. As shown in equation (22):

[0126]

[0127] Similarly, by acquiring Q images of the dorsal veins of the hand in terminal B, equation (23) is calculated:

[0128]

[0129] Using this method, you can obtain... and Estimate and When calculating similarity later, use and .

[0130]

[0131]

[0132] Figure 2 For steps (7) to (12) on the B terminal, please refer to the above. Figure 2 Steps (1) to (6) are as follows;

[0133] Figure 2 In step (13) The center of the partition is approximately 80% of the area used as a template. That is, in Divide into ;

[0134] Figure 2 In step (14), Figure 2 Steps (6) and (12) can be obtained and Estimate and , needs to be calculated and The similarity is calculated using the Pearson correlation coefficient method, and equation (26) holds:

[0135]

[0136] In equation (26) This involves calculating the Pearson correlation coefficient, and then using a sliding template for matching calculations. exist Slide in the middle, set The size is , The size is , , The top left pixel is Slide to When in position, cover The area is denoted as At this point, the Pearson correlation coefficient Calculated by the following formula (27):

[0137]

[0138] In equation (27), for The mean, For subgraph The mean of . The matching matrix obtained after pixel-by-pixel sliding matching is denoted as . ,

[0139]

[0140] Figure 2 In step (15), the final matching score is taken. The maximum value in the formula is obtained, as shown in equation (29):

[0141]

[0142] ​ In step (16), finally utilize With threshold Comparison to determine the veins in the back of the hand image sequence Set B Whether the samples were collected from the same dorsal vein of the hand in a live specimen, in practical applications 0.6 is acceptable;

[0143]

[0144] This embodiment details how the illumination response curves of the same pixel on the back of the hand under different terminals are derived, clarifying their translational quantization relationship with the camera response function curve. It verifies that the translational distance of the illumination response curves of any two points on the back of the hand remains consistent across terminals, unaffected by differences in terminal camera response functions or minimum unit light intensity of the light source. This conclusion provides core theoretical support for cross-terminal liveness detection, ensuring that similarity comparison based on illumination response offset features is effective across terminals, avoiding judgment errors caused by hardware differences, and further solidifying the reliability and adaptability of the technical solution.

[0145] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any changes, modifications, substitutions, integrations, and parameter changes made to these embodiments within the spirit and principles of the present invention, without departing from the principles and spirit of the present invention, through conventional substitutions or to achieve the same function, fall within the scope of protection of the present invention.

Claims

1. A cross-terminal live hand dorsal vein detection method based on illumination response characteristics, characterized in that, The application relates to a method for detecting a living body through a hand dorsal vein image, and the method comprises the following steps: S1, collecting a target and a multi-light-intensity living body hand dorsal vein image sequence through a first terminal; the first terminal and the second terminal are different hand dorsal vein identification terminals, and the target hand dorsal vein image collected by the first terminal has been confirmed as a living body image; S2, pre-processing the multi-light-intensity living body hand dorsal vein image sequence collected by the first terminal and the multi-light-intensity hand dorsal vein image sequence collected by the second terminal; S3, extracting a region of interest of the hand dorsal vein image of the two terminals after pre-processing, performing scaling processing on the region of interest, and reducing image noise; S4, calculating a light response offset value of each pixel point in the target hand dorsal vein region of interest relative to a camera response function curve of the first terminal through the camera response function of the first terminal in combination with the multi-light-intensity living body hand dorsal vein image sequence collected by the first terminal, and generating a first offset feature map; S5, calculating a light response offset value of each pixel point in the hand dorsal vein region of interest to be detected relative to a camera response function curve of the second terminal through the camera response function of the second terminal in combination with the multi-light-intensity hand dorsal vein image sequence to be detected collected by the second terminal, and generating a second offset feature map; S6, calculating the similarity of the first offset feature map and the second offset feature map, comparing the obtained similarity result with a preset threshold, and determining whether the hand dorsal vein to be detected is a living body.

2. The cross-terminal method for detecting the living being based on the light response characteristics of the dorsal venous network of the hand according to claim 1, characterized in that, The method for collecting a target and a multi-light-intensity living body hand dorsal vein image sequence through the first terminal is specific to: The multi-intensity living body dorsal vein image sequence collected by the first terminal is specifically: the first terminal collects target dorsal vein images with a duty cycle } of 50% to 90% , and records the image sequence as all images in the sequence have no overexposure, no underexposure and satisfy pixel-level registration.​​​​​​​ The second terminal acquires a multi-intensity image sequence of the dorsal veins of the hand to be detected, specifically: the second terminal uses a duty cycle of { }collection The images of the veins on the back of the hand to be detected are denoted as the image sequence. { },in This represents the total number of images acquired by the second terminal. For the second terminal to collect the first Light source duty cycle when taking an image For the second terminal in duty cycle Images of the veins on the back of the hand acquired at that time, and All images in the dataset are also free from overexposure and underexposure and meet pixel-level registration requirements; The light intensity of the light source of the first terminal at the minimum unit duty cycle is denoted as The actual light intensity of the light source of the first terminal at the duty cycle is denoted as , satisfying The light intensity of the light source of the second terminal at the minimum unit duty cycle is denoted as The actual light intensity of the light source of the second terminal at the duty cycle is denoted as , satisfying .

3. The cross-terminal method of detecting a live human dorsal hand vein based on the light response characteristics according to claim 1, characterized in that, The S2 pre-processing specifically comprises: S2.1, extracting a first terminal image sequence and a second terminal image sequence the back of the hand contour in each image of the first and second terminal image sequences, respectively, by a contour detection algorithm to determine the boundary range of the back of the hand in the image; S2.2, adjusting the second terminal image sequence the orientation of the back of the hand, such that the orientation of the back of the hand is consistent with the orientation of the back of the hand is consistent with the orientation of the target back of the hand, ensuring uniformity of the back of the hand pose across terminal images; S2.3, calculating the size ratio of the back of the hand contour of the first terminal and the second terminal: assuming that the length of the minimum circumscribed rectangle of the back of the hand contour extracted by the first terminal is , the width is , the length of the minimum circumscribed rectangle of the back of the hand contour extracted by the second terminal is , the width is , then the length direction ratio is , and the width direction ratio is ; S2.4, scaling the first terminal image sequence according to the size ratio S2.4, scaling the first terminal image sequence according to the size ratio S2.4, scaling the first terminal image sequence according to the size ratio S2.4, scaling the first terminal image sequence according to the size ratio 4. The cross-terminal method of detecting a live human dorsal hand vein based on the light response characteristics according to claim 1, characterized in that, The S3 extracts a region of interest, specifically: taking the center of the minimum circumscribed rectangle of the preprocessed hand back contour as the origin , the length of the minimum circumscribed rectangle of the contour is , and the width is , the length and width of the region of interest are respectively , , , and the coordinate range of the region of interest is ; The scaling processing is performed on the region of interest, and the original image of the region of interest is denoted as , the superimposed noise is denoted as , and the image with noise is denoted as , the noise of the original image is denoted as , and the noise of the original image is subject to ; the mean filtering is performed with a window size of and a step size of , and is a preset window size, and the pixel value of the filtered image is: , wherein the former term is the local block mean value of the original image, and the latter term is the noise local block mean value, is the pixel value of the output image after filtering at the coordinates , and the noise local block mean value is subject to a normal distribution of , and is the noise variance of the image before being scaled, and the effect of scaling and reducing the noise variance is realized.

5. The cross-terminal method of detecting a live human dorsal hand vein based on the illumination response characteristics according to claim 1, characterized in that, The camera response function of the first terminal in S4 is denoted as , satisfying , where is the light intensity received by the camera sensor, is the pixel gray value output by the camera; and is defined as the new camera response function of the first terminal , that is ; For any pixel point in the target hand back interested region of the first terminal The reflectivity of which is denoted as The light intensity of the pixel point under the irradiation of the first terminal light source Satisfies the formula In combination with the camera response function, the gray value of the pixel point is calculated The formula is: wherein, is the duty cycle of the light source, is the light intensity corresponding to the minimum unit duty cycle of the first terminal light source.

6. The cross-terminal method of detecting a live human dorsal vein based on the light response characteristics according to claim 1 or 5, characterized in that, The S4 calculation of the light response offset value is specific to: The first terminal new camera response function The inverse function is denoted as The first terminal will collect the first data. Pixels in the image grayscale value Substitute into the inverse function Combined with formula Rearranging the terms, we get: The results are averaged to obtain the illumination response offset estimate for the pixel point , which is given by the equation: A matrix is formed by the illumination response offset estimation values of all pixel points in the target hand back region of interest, that is, the first offset feature map , and .

7. The cross-terminal method of detecting a live human dorsal hand vein based on the illumination response characteristics according to claim 1, characterized in that, The camera response function of the second terminal in S5 is denoted as , satisfying , where is the light intensity received by the camera sensor, is the pixel gray value output by the camera; and is defined as the new camera response function of the second terminal , that is, ; For the second terminal under the detection of any pixel point in the region of interest of the back of the hand , the reflectivity is recorded as , the light intensity of the pixel point under the irradiation of the second terminal light source is calculated , the formula is: , combined with the camera response function, the gray value of the pixel point is calculated, and the formula is: wherein, is the duty cycle of the light source, is the light intensity corresponding to the minimum unit duty cycle of the second terminal light source.

8. The cross-terminal dorsal venous network liveness detection method based on the photoresponse characteristics according to claim 1 or 7, characterized in that, The S5 calculation of the light response offset value is specific to: The second terminal new camera response function The inverse function of the second terminal new camera response function is denoted as The gray value of the pixel point in the first image collected by the second terminal is substituted into the inverse function Combining the formula , and moving the term, the following formula is obtained:​​​ The results are averaged to obtain the illumination response offset estimate for the pixel point is the average of the pixel points. A matrix is formed by all the pixel points in the region of interest on the back of the hand to be detected, and the illumination response offset estimation value is the second offset feature map , and .

9. The cross-terminal method of detecting a live human dorsal hand vein based on illumination response characteristics according to claim 1, wherein, The S6 calculation of the similarity of the first offset feature map and the second offset feature map is specific to the calculation through a Pearson correlation coefficient in combination with a sliding template matching, and is specific to: In the second offset feature map The center region is selected as the matching template , The size of the center region is , The template height is The template width is Make In the first offset feature map Sliding pixel by pixel, when The upper left corner is located The Position, The upper covered sub-region is recorded as , Calculate the Pearson correlation coefficient of And At this time wherein, is the pixel mean of is the pixel mean of is the pixel mean of is the pixel mean of The correlation coefficients of all sliding positions constitute a matching matrix , take the maximum value in as the similarity result , that is .

10. The cross-terminal method of detecting a live human dorsal vein based on the illumination response characteristics according to claim 1 or 9, characterized in that, The S6 preset threshold Taking 0.6, the determination rule is: When the similarity result is greater than or equal to , it is determined that the hand back collected by the second terminal is a living body; when is less than , it is determined that the hand back to be detected is a fake body, and the cross-terminal hand back vein living body detection is completed.

Images