High-sensitivity and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method

Abstract

The invention provides a high-sensitivity and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method, which comprises the following steps of: constructing an ultrasonic transducer for fingerprint acquisition, placing the ultrasonic transducer on an optical sensor, and acquiring a sound field signal and a fingerprint grayscale image of a fingerprint; preprocessing the fingerprint grayscale image, and establishing a fingerprint speed model; extracting a first-motion wave time domain signal segment in the sound field signal of the fingerprint, and establishing a frequency domain sound field matrix; establishing an offline training data set of fingerprint imaging, and obtaining a descent gradient matrix through offline training; a post-processing model for image inpainting and detail matching is built and trained; performing online inversion through the descent gradient matrix to obtain a fingerprint inversion image of the test case; inputting the fingerprint inversion image into the trained post-processing model to obtain a fingerprint matching result; and calculating personal authentication confidence by using a fingerprint matching result, and identifying the fingerprint to be authenticated. According to the invention, the form of the fingertip epithelium can be effectively discriminated from the sound field response even under the condition of environmental pollution, and the method has the characteristics of zero occupation of the space under the ultrasonic transducer screen, high-sensitivity imaging, authentication and anti-counterfeiting.

Description

Technical Field

[0001] The present invention relates to the technical field of biometric authentication, and in particular to an ultrasonic guided wave fingerprint imaging and authentication method. Background Art

[0002] In recent years, biometric authentication technology has developed rapidly in the fields of consumer electronics and public security, which not only enhances system security, but also significantly optimizes the user experience, and has developed a series of representative technologies such as fingerprint recognition, facial recognition, voice recognition, palm print recognition, iris and retina recognition, and vein recognition. By 2023, fingerprint authentication technology will account for 34.8% of the global biometric authentication market, showing an absolute dominant position, with a market value of up to US$21.6 billion.

[0003] The combination of fingerprint scanning technology and authentication technology has been widely used in various electronic devices such as access control systems, mobile devices, personal computers, and wearable devices. As users' requirements for the comfort and universality of fingerprint authentication are increasing, as well as the demand for identity security and personal privacy protection is constantly upgrading, fingerprint authentication technology is gradually developing towards large scanning domain texture authentication such as multi-fingerprint or palmprint authentication, providing an opportunity for further innovation in biometric texture authentication technology.

[0004] At present, the under-screen fingerprint scanning technology is systematically divided into three categories: capacitive, optical and ultrasonic. Among them, capacitive scanning is widely used in mobile phone unlocking and the technology is relatively mature. The working principle of this technology is to characterize the fingerprint pattern by detecting the capacitance difference between the fingerprint ridges and the transducer. However, since it is difficult for capacitors to effectively distinguish the dielectric constant of the epithelial layer from water or oil, this technology is susceptible to interference from environmental pollution.

[0005] Optical scanning technology is commonly used in fingerprint authentication devices in public security fields such as banks and police, and has the ability to produce high-resolution images. This technology uses scattered light from the ridges and valleys of fingerprints to collect fingerprints, but the imaging results are often 2D light and dark patterns, which are vulnerable to authentication attacks. In addition, since the miniaturization of charge-coupled devices is a technical challenge, this technology has certain limitations in terms of device thickness, so its application scenarios are limited.

[0006] Ultrasonic scanning technology has higher robustness. By utilizing the pulse echo flight time of sound waves, the 3D information of the fingerprint can be characterized, showing significant resistance to environmental pollution such as moisture or dust. However, since the pulse echo involves vertical signal excitation and reception, the fingerprint scanning domain is often limited to a small range and requires dense transducer array support. Faced with the demand for large scanning domain texture authentication, this technology requires the deployment of a large area of ​​under-screen transducer array, which not only increases costs, but also has a negative impact on the practicality of the device.

[0007] Traditional fingerprint scanning technology relies on signal acquisition perpendicular to the screen, which is difficult to meet the large scanning area texture authentication requirements of diverse electronic devices in the fields of healthcare, banking, security, and attendance. Therefore, it has become a top priority to propose a new method to break the reliance on under-screen transducers.

[0008] Ultrasonic guided waves are elastic waves that propagate along the surface or inside of a waveguide structure. Their propagation behavior is affected by boundary conditions and exhibits diverse propagation modes under different structural thicknesses and excitation frequencies. This dispersion phenomenon is extremely sensitive to small changes in the surface of the structure, making it possible to distinguish different waveguide materials by analyzing the acoustic field response characteristics. If a series of transducers are arranged to form an array surrounding the target area and combined with tomography methods to detect changes in the acoustic field, the transducer network can be used to characterize structural changes in the scanned domain. This method can achieve target imaging without moving the transducers for scanning one by one.

[0009] For example, the invention patent with application number: 202311018944.7 involves the field of ultrasonic fingerprint recognition technology, and in particular, an ultrasonic fingerprint imaging method, a fingerprint recognition device, and an electronic device. The method includes the following steps: setting the ultrasonic transducers to be distributed in a uniform rectangular array on the sensor, dividing each n×n ultrasonic transducer array into a group, and transmitting multiple ultrasonic waves to the object by the ultrasonic transducer, receiving and outputting multiple ultrasonic signals reflected from the object; based on the bionic principle of image convolution, the amplitude of each pixel in the ultrasonic transducer array is adjusted individually based on the value of each point in the convolution kernel; based on the principle of beamforming, the phase of each pixel in the ultrasonic transducer array is adjusted individually; through beamforming technology, the sound wave intensity of the surrounding pixels is focused above the central pixel; and the fingerprint image of the object is obtained at the central pixel. However, the above application uses an under-screen ultrasonic transducer array to characterize the fingerprint morphology by measuring the flight time and intensity of the pulse echo reaching the fingerprint. Although it has the inherent robustness of ultrasonic technology in resisting environmental pollution, its scanning domain is basically the same size as the transducer array, which limits its application in fingerprint characterization in a large scanning domain. It has certain limitations in fingerprint authentication at any location, multiple fingerprints, and palmprint authentication. Summary of the invention

[0010] In view of the technical problems that the existing fingerprint scanning technology relies on complex under-screen devices, resulting in a limited scanning range, and that the capacitive and optical types are susceptible to false texture deception and environmental pollution interference, the present invention proposes a highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method, which utilizes signal acquisition in the direction of the plate plane to detect the sound field changes in the scanning range, and uses ultrasonic guided wave tomography based on the supervised descent method to characterize submillimeter textures.

[0011] In order to achieve the above object, the technical solution of the present invention is implemented as follows: a highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method, the steps of which include:

[0012] S1: Build an ultrasonic transducer for fingerprint collection and place it on an optical sensor. The ultrasonic transducer collects the acoustic field signal of the fingerprint and the optical sensor collects the grayscale image of the fingerprint.

[0013] S2: preprocessing the fingerprint grayscale image, and calculating the phase velocity interval used to characterize the fingerprint based on the global matrix method, and establishing a fingerprint velocity model according to the preprocessed fingerprint grayscale image;

[0014] S3: extract the first arrival wave time domain signal fragment of the mode corresponding to the phase velocity interval in the sound field signal of the fingerprint, extract the frequency domain energy flow amplitude to obtain the spectrum amplitude, and establish the frequency domain sound field matrix;

[0015] S4: Using the methods of steps S2 and S3, establish an offline training data set for fingerprint imaging, obtain a descent gradient matrix through offline training; build a post-processing model for image restoration and detail matching and perform training;

[0016] S5: The acoustic field signal of the fingerprint to be authenticated is collected by an ultrasonic transducer, and the grayscale image of the fingerprint to be authenticated is collected by an optical sensor. The fingerprint velocity model and the frequency domain acoustic field matrix of the test case are obtained by using the methods of steps S2 and S3. The fingerprint inversion image of the test case is obtained by online inversion of the descending gradient matrix; the fingerprint inversion image is input into the trained post-processing model to obtain the fingerprint matching result;

[0017] S6: Calculate personal authentication confidence using the fingerprint matching result, and identify the fingerprint to be authenticated through the personal authentication confidence.

[0018] Preferably, the ultrasonic transducer includes a plurality of ultrasonic transducer array elements arranged in a circular shape, the ultrasonic transducer array element includes a piezoelectric ceramic sheet group, a positive copper foil and a negative copper foil are respectively provided on the upper and lower sides of the piezoelectric ceramic sheet group, a glass plate is provided on the lower side of the negative copper foil and the upper side of the positive copper foil, and the positive copper foil and the negative copper foil are both connected to the ultrasonic instrument; an insulating layer is laid between the positive copper foil and the negative copper foil, and a sound field boundary absorption layer is provided at the outermost edge between the two glass plates.

[0019] Preferably, the piezoelectric ceramic sheet group consists of at least one circular PZT-5H sheet, the positive copper foil and the negative copper foil are respectively fixedly connected to the PZT-5H sheet by conductive silver glue, and the negative copper foil is fixed to the glass plate on the lower side of the negative copper foil by an epoxy resin adhesive.

[0020] Preferably, the PZT-5H sheet is made by cutting and polishing a PZT-5H raw material, and the positive electrode copper foil and the negative electrode copper foil are both obtained by laser cutting;

[0021] The positive copper foil and the negative copper foil both include circular endpoints with the same diameter as the PZT-5H sheet, and leads are arranged on the circular endpoints. The leads of the negative copper foil are connected to the annular copper foil, and the annular copper foil is grounded; the outermost part of the negative copper foil is directly connected to a square copper foil, and the four corners of the square copper foil are divided into straight copper foil and right-angle copper foil by breakpoints. The straight copper foil is folded outward by 90 degrees, and the copper-clad surface is turned downward to the same side as the copper-clad surface of the positive copper foil, and the right-angle copper foil is reserved as a debugging contact point;

[0022] The positive electrode copper foil is divided into four equal parts, and the copper-clad surface of each part of the positive electrode copper foil faces downward; each part of the positive electrode copper foil is fixed with a polydimethylsiloxane film; the outermost part of the glass plate is surrounded by plasticine to serve as a sound field boundary absorption layer;

[0023] A layer of transparent polydimethylsiloxane film is laid between the surface glass of the optical sensor and the glass plate of the ultrasonic transducer.

[0024] Preferably, the method for preprocessing the fingerprint grayscale image is: tracking the fingerprint and extracting preliminary features to obtain preliminary features of the fingerprint image; extracting the target area in the preliminary features of the fingerprint image through threshold control; converting the extracted target area into a low dynamic range image using a bilateral filtering algorithm, and eliminating the connected areas in the fingerprint details through a spatial domain filtering algorithm with a convolution kernel to obtain a fingerprint ridge and valley image.

[0025] Preferably, the method for establishing the fingerprint velocity model is: using the thickness of the fingertip epithelium, the height difference between the ridge and the valley, the epithelial density, the Young's modulus and the Poisson's ratio as inputs, and using the global matrix method to calculate the phase velocity dispersion curve of the fingertip ridge before and after the glass plate of the ultrasonic transducer is attached, and at the same frequency, the A0 mode with a shorter wavelength and carrying more sound field information is selected; compared with the glass structure of the glass plate, the material properties of the epithelial layer on the fingertip are closer to the fluid, and the fluid-solid coupling theory shows that after the ridge is attached to the glass, the sound field energy flow is mainly concentrated in the adjacent mode below the phase velocity dispersion curve of the glass plate. At the center frequency of the ultrasonic transducer, the attachment of the ridge to the glass plate causes the phase velocity of the A0 mode to drop from 2.91 mm / μs to 2.55 mm / μs, and the phase velocity interval is [2.55, 2.91] mm / μs.

[0026] The fingerprint ridge-valley image is mapped to the phase velocity interval [2.55, 2.91] mm / μs to obtain the fingerprint velocity model.

[0027] Preferably, the method for extracting the first arrival wave time domain signal fragment of the mode corresponding to the phase velocity interval in the sound field signal of the fingerprint is: using the conversion relationship between the phase velocity and the group velocity to calculate the group velocity dispersion curve before and after the ridge-bonded glass plate, at the center frequency of the ultrasonic transducer, there is an obvious mode separation phenomenon between the A0 mode and other modes, and using the group velocity difference and the window function, the time domain signal of the A0 mode is extracted from the time domain signal of the collected sound field signal;

[0028] The method for extracting the frequency domain energy flow amplitude to obtain the spectrum amplitude is as follows: converting the time domain signal fragment of each ultrasonic transducer array element into a frequency domain signal by frequency domain transformation, smoothing the frequency domain signal, extracting the peak value of the frequency domain signal spectrum after smoothing to obtain the spectrum amplitude, writing the extracted spectrum amplitude into the corresponding point of the frequency domain sound field matrix, and mapping the frequency domain sound field matrix to the interval range of [1,2].

[0029] Preferably, the supervised descent method used in the offline training is:

[0030] R=(Δs T Δs+β 2 I) -1 ·(Δs T Δm);

[0031] Where R is the descending gradient matrix obtained through offline training, Δs=s0-s1 is the difference in the sound field matrix, s0 and s1 represent the frequency domain sound field matrix before and after fingertip pressing, respectively; β is the regularization weight, Δm=m0-m1 is the difference in the fingerprint velocity model, m0 and m1 represent the fingerprint velocity model before and after fingertip pressing, respectively; the velocity sound field matrix s and the fingerprint velocity model m are both data matrices containing all examples;

[0032] The online inversion is performed by descending the gradient matrix R test case to obtain the imaging result M of the test case. test =m0+Δs test ·R;Δs test It is the frequency domain sound field matrix corresponding to the test case.

[0033] Preferably, the method for tracking and extracting preliminary features of fingerprints is as follows: in the grayscale image of the fingerprint, the point with the largest grayscale value is tracked, and the grayscale image within the range of 36mm×36mm around the point with the largest grayscale value is extracted, and the grayscale image is inverted to extract the range with grayscale values ​​below 110; the threshold control method is to discretely invert the grayscale image with a grid spacing, and retain the part with grayscale average value below 80 in the grid;

[0034] The smoothing process adopts Gaussian weighted moving average filtering algorithm;

[0035] The method for establishing an offline training data set is: collecting sound field signals and fingerprint grayscale images of fingertip epithelium in different dry states, at different positions, and with different pressing forces, and obtaining a fingerprint velocity model and sound field signal obtained for each fingertip pressing through the methods of steps S2 and S3 to form an example.

[0036] Preferably, the post-processing model includes a DLCNN model for image restoration and a MaskR-CNN model for fingerprint matching. The DLCNN model adopts a small-scale, deep-layer architecture, all convolutional layers adopt a 3×3 dimension, and the network depth of the DLCNN model is 20 layers; the Mask R-CNN model includes a ResNet101 feature extraction network, a feature pyramid network, a region proposal network, an area of ​​interest arrangement, and a full convolutional network connected in sequence; the DLCNN model is used to obtain a restored fingerprint grayscale image, and the restored fingerprint grayscale image is converted into a binary form of a fingerprint image through an even-symmetric Gabor filter and then input into the Mask R-CNN model, and the restored fingerprint grayscale image is passed through the Mask R-CNN model to obtain a fingerprint matching result image;

[0037] The training data set of the DLCNN model includes a fingerprint grayscale image of the artifact removal part and a grayscale image of the natural scenery image of the image denoising part. Each example of the artifact removal part includes an artifact-containing fingerprint inversion image and a corresponding pure fingerprint image; each example of the image denoising part includes a natural scenery image containing Gaussian noise and a corresponding pure image.

[0038] The training data set of the Mask R-CNN model includes the binary form of the grayscale images of fingerprints of multiple registered fingertips collected by the optical sensor and attached with fingerprint marks;

[0039] A large-scale marking method covering multiple fingerprint details was adopted, marking 8 large-scale fingerprint details;

[0040] The method for calculating the personal authentication confidence is: the uniqueness of the fingerprint details of the registered fingertips is divided into three levels, the fingerprint details of level 1 have the highest uniqueness and are selected as the key features representing personal information, the uniqueness of the fingerprint details of level 2 is second, and the uniqueness of the fingerprint details of level 3 is relatively weak; then the personal authentication confidence is:

[0041]

[0042] In the formula, ξ k is the compensation factor, the compensation factor is an empirical value related to the marked fingerprint details, which varies with the number of fingerprint details k∈[0,8] recognized by the Mask R-CNN model, and b is the weight of levels 1 to 3.

[0043] Compared with the prior art, the invention has the following beneficial effects: an ultrasonic transducer array is arranged on the upper surface of the edge of the large scanning area, so that the ultrasonic transducer does not occupy the space under the screen; the signal acquisition in the direction of the plate plane is used to detect the change of the sound field in the scanning area, and ultrasonic guided wave tomography based on the supervised descent method is used to characterize the submillimeter texture. An advanced post-processing model is introduced for image restoration and detail matching, improving the reconstructed fingerprint detail morphology and giving personal authentication anti-counterfeiting properties.

[0044] The ultrasonic guided wave fingerprint imaging and authentication method proposed in the present invention has the characteristics of zero space occupation under the transducer screen, high-sensitivity imaging and authentication and anti-counterfeiting; the collected signal in the direction of the plate plane uses the Lamb wave dispersion effect generated when the fingertip is pressed on the glass plate to characterize the submillimeter texture details, and learns the nonlinear mapping relationship between the sound field change and the fingerprint morphology change, so that even in the case of environmental pollution, the morphology of the epithelial layer on the fingertip can be effectively identified and recognized from the sound field response. It has the characteristics of zero space occupation under the transducer screen, high-sensitivity imaging and authentication and anti-counterfeiting, which not only provides an effective solution for large scanning domain fingerprint characterization, but also indicates its development prospects in multi-fingerprint characterization and palmprint authentication technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a flow chart of the present invention.

[0047] Figure 2 The figure is a schematic diagram of fingerprint collection operation according to an embodiment of the present invention.

[0048] Figure 3 This is the electrical interconnection method of the PZT-5H ultrasonic transducer array elements proposed by the present invention.

[0049] Figure 4 A schematic diagram of the structure of the fingerprint acquisition acoustic wave transducer developed for the present invention.

[0050] Figure 5 The invention provides a fingerprint velocity model establishment method.

[0051] Figure 6 A side view schematic diagram of the sound field propagation when a fingertip presses a glass plate developed for the present invention.

[0052] Figure 71 is a Lamb wave phase velocity dispersion curve before and after fingertip pressing according to an embodiment of the present invention, wherein (a) is a Lamb wave phase velocity dispersion curve in the frequency range of 0-7 MHz, and (b) is an enlarged view within the dotted box in (a).

[0053] Figure 8 This is a schematic diagram of the frequency domain sound field matrix proposed in the present invention, wherein (a) is the extracted A0 mode time domain signal, (b) is the A0 mode frequency domain energy flow amplitude, and (c) is the frequency domain sound field matrix.

[0054] Fig. 9 1 and 2 are Lamb wave group velocity dispersion curves before and after fingertip pressing according to an embodiment of the present invention, wherein (a) is a Lamb wave group velocity dispersion curve within a frequency range of 0-7 MHz, and (b) is an enlarged view within the dotted box in (a).

[0055] Fig.10 This is a processing flow chart of the image post-processing model proposed in the present invention.

[0056] Fig.11 Schematic diagram of imaging and authentication of a clean registered fingertip and authentication attack result of a clean unregistered fingertip in an embodiment of the present invention.

[0057] Fig.12 The figure is a schematic diagram of imaging and authentication of a registered fingertip with moisture and a registered fingertip with dust in an embodiment of the present invention.

[0058] In the figure, 1 is a piezoelectric ceramic sheet group, 2 is a positive copper foil, 3 is a negative copper foil, 4 is a glass plate, 5 is an epoxy resin binder, 6 is a conductive silver glue, 7 is a polydimethylsiloxane film, 8 is an insulating layer, 9 is an acoustic field boundary absorption layer, and 10 is a debugging contact point. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] like Figure 1As shown in the figure, a highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method is developed to achieve the three goals of zero space occupation under the transducer screen, high-sensitivity imaging and authentication and anti-counterfeiting. A fingerprint acquisition ultrasonic transducer with a φ100mm scanning area is developed, and a complete fingerprint acquisition system is built. The data acquisition and signal processing methods required for fingerprint imaging and authentication are combined. By obtaining the frequency domain sound field difference and velocity model difference caused by fingertip pressing, an offline training data set for fingerprint imaging is established. The nonlinear mapping relationship between the two differences is learned using the supervised descent method, and it is used for fast imaging of test case fingerprints. In the image post-processing model, the DLCNN (Deep Learning Convolutional Neural Network) model and the Mask R-CNN model (Mask Region-based Convolutional Neural Network) are used. The DLCNN model is used for artifact removal and image denoising to improve the reconstructed fingerprint detail morphology and significantly improve the reconstruction accuracy; the Mask R-CNN model is used for detail matching to give authentication and anti-counterfeiting properties. The specific steps are as follows:

[0061] S1: construct a fingerprint collection ultrasonic transducer and place the fingerprint collection ultrasonic transducer on the optical sensor. The fingerprint collection ultrasonic transducer collects the acoustic field signal of the fingerprint, and the optical sensor collects the grayscale image of the fingerprint.

[0062] like Figure 2 As shown, the ultrasonic guided wave fingerprint imaging method detects the changes in the sound field within the scanning domain through the signals collected in the direction of the plate plane, which is used to characterize sub-millimeter textures such as fingerprints. In order to meet the verification requirements of ultrasonic guided wave fingerprint imaging, a fingerprint acquisition transducer device with a φ100mm scanning domain is designed and developed in this embodiment. Among them, the range within the ultrasonic transducer array is called the scanning domain, and the diameter of the scanning domain is determined according to the size and number of the ultrasonic transducers. The ultrasonic transducer is obtained by grinding the PZT-5H material embryo. In order to ensure the effective displacement in the z direction (perpendicular to the glass plate), it cannot be made too small, so the diameter of the transducer array element processed in this embodiment is 2mm, such as Figure 3 As shown in the figure, in order to ensure sufficient fingerprint inversion image resolution, the number of transducers cannot be too small. This embodiment uses 128 transducers. If there are too many transducers, the amount of calculation will be too large. Therefore, the diameter of the scanning domain is not absolute. It can be made small or larger. The attenuation of the guided wave field needs to be considered. At present, the experimental and simulation results show that there is no problem with a laptop screen this large.

[0063] Specifically, Figure 3As shown, the fingerprint collection ultrasonic transducer includes a plurality of ultrasonic transducer array elements arranged in a circular ring shape, such as Figure 4 As shown, the ultrasonic transducer array element includes a circular piezoelectric ceramic sheet group 1, and the upper and lower sides of the piezoelectric ceramic sheet group 1 are respectively provided with a positive copper foil 2 and a negative copper foil 3, and the lower side of the negative copper foil 3 and the upper side of the positive copper foil 2 are both provided with a glass plate 4, and the outermost edge between the two glass plates is provided with a sound field boundary absorption layer 9, and the positive copper foil 2 is connected to the ultrasonic instrument through an interconnection module, and the negative copper foil 3 is connected to the ultrasonic instrument grounding end. The interconnection module is a signal acquisition and data transfer interface, which realizes the generation of ultrasonic transducer array element excitation signals and the collection of guided wave signals.

[0064] The piezoelectric ceramic sheet group 1 is composed of at least one PZT-5H sheet, which is made by cutting and polishing a PZT-5H original blank with a center frequency of 5 MHz. The PZT-5H sheet is a circular sheet with a diameter and thickness of approximately 2 mm and 0.5 mm, respectively.

[0065] The positive copper foil 2 and the negative copper foil 3 are obtained by laser cutting. The positive copper foil 2 and the negative copper foil 3 both include a circular end point with a diameter of 2 mm, and a lead with a width of 0.7 mm is arranged on the circular end point. Each piezoelectric ceramic sheet group 1 corresponds to a positive copper foil 2 and a negative copper foil 3. The positive copper foil 2 and the negative copper foil 3 are fixedly connected to the PZT-5H sheet by a conductive silver glue 6 without affecting the conductivity. The negative copper foil 3 is fixed to the glass plate on the lower side of the negative copper foil 3 by an epoxy resin adhesive 5. The adhesive plays the role of acoustic field coupling and ensures the acoustic field transmission between the ultrasonic transducer and the glass plate. After the transducer is electrically interconnected, the center frequency decreases due to the overall thickening. In this embodiment, the average center frequency of the 128-channel transducer is determined to be approximately 4.29 MHz through impedance analysis.

[0066] In this embodiment, 128 PZT-5H sheets are used to form a transducer array. The channel numbers are numbered in a counterclockwise increasing manner starting with the ultrasonic transducer array element in the positive direction of the x-axis. The PZT-5H sheets are arranged to form a ring. During use, multiple ultrasonic transducer array elements send excitation signals in the order of numbers, and all array elements collect guided wave signals at the same time. The glass plate size is 120mm×120mm×0.55mm, the model is Schott B270 soda-lime glass, and the density ρ is 2.56g / cm 3, Young's modulus E is 71.10GPa, and Poisson's ratio ν is 0.22. The lead ends of the negative copper foil 3 are all connected to a ring-shaped copper foil with a radius of 60mm, and the ring-shaped copper foil is grounded. At the outermost periphery of the negative copper foil 3, it is directly connected to a square copper foil with a side length of 120mm. The four corners of the square copper foil are divided into straight copper foil and right-angle copper foil by breakpoints. Among them, the straight copper foil is folded 90° outward, and the copper-clad surface is flipped to the bottom, on the same side as the copper-clad surface of the positive copper foil, and connected to the interconnection module together. The right-angle copper foil is reserved as the debugging contact point 10 of the device.

[0067] The positive copper foil is divided into four equal parts, each part is responsible for the electrical interconnection of 32 channels, and the four parts are connected to the ultrasonic instrument from four directions, with the copper-coated surface facing down, and each part is connected to the interconnection module together with a straight copper foil of the flipped negative copper foil 3, and finally connected to the ultrasonic instrument. In order to ensure the stability and flexibility of the positive copper foil, each part of the copper foil is fixed with a polydimethylsiloxane film 7 to form a bus. An insulating layer 8 is laid between the positive copper foil and the negative copper foil. The outermost periphery of the glass plate is surrounded by plasticine, which acts as a sound field boundary absorption layer 9.

[0068] The key to fingerprint imaging is to learn the nonlinear mapping relationship between fingerprint morphology and sound field changes when the fingertip is pressed through offline training. To this end, in addition to the ultrasonic instrument, a device that can accurately capture the fingerprint morphology is also required. In an embodiment of the present invention, a fingerprint acquisition ultrasonic transducer is placed above an optical sensor. The optical sensor model is EPSONV600, and the image storage and resolution are set to grayscale and 1200dpi respectively. In order to reduce the potential interference of the surface glass of the optical sensor to the sound field of the fingerprint acquisition ultrasonic transducer device, a layer of transparent polydimethylsiloxane film is laid between the two.

[0069] The optical sensor is used to collect the fingerprint shape when the fingertip is pressed, and uses the Lamb wave dispersion effect to map it to the corresponding phase velocity interval to obtain the fingerprint velocity model. The fingerprint collection ultrasonic transducer is used to collect the sound field signal when the fingertip is pressed. Compared with the no-load state of the fingertip without pressing, the change in the sound field signal caused by the fingertip pressing is called the sound field change. The fingerprint velocity model and sound field signal obtained from each fingertip press constitute a calculation example.

[0070] S2: Preprocess the fingerprint grayscale image, calculate the phase velocity interval used to characterize the fingerprint based on the global matrix method, and establish a fingerprint velocity model.

[0071] like Figure 5 As shown, the fingerprint velocity model is a reference standard for fingerprint morphology, and the fingerprint grayscale image is collected by an optical sensor placed under the fingerprint collection transducer device.

[0072] The fingerprint grayscale image obtained in step S1 is preprocessed. The fingerprint image is a grayscale image with a grayscale range of 0-255. In fingerprint image processing, first, the fingerprint is tracked and preliminary feature extraction is performed to obtain preliminary features of the fingerprint image. The specific method for tracking and preliminary feature extraction of the fingerprint is as follows: in the fingerprint grayscale image, within the range corresponding to the fingerprint acquisition ultrasonic transducer, track the point with the largest grayscale value, that is, the brightest point of the fingerprint. With the point with the largest grayscale value as the center, extract the grayscale image within the range of 36mm×36mm around it, invert the grayscale image to convert the fingerprint pixels from white to black, and extract the range with grayscale values ​​below 110. Since the white part in the grayscale image after inversion processing is basically glass, the part close to pure white in the figure is omitted. Next, the target area is extracted from the preliminary features of the grayscale image after inversion processing by threshold control. The grayscale image after inversion processing is discretized with a grid spacing of 0.4mm, and the part with a grayscale average value of less than 80 in the grid is retained to resist the interference of dust impurities on the fingerprint image and further reduce the range of valid fingerprints. In order to reduce the complexity and contrast of fingerprint images while maintaining the quality of fingerprint images, the extracted target area image is converted into a low dynamic range (LDR) image using a bilateral filtering algorithm. Finally, a spatial filtering algorithm with a convolution kernel is introduced. The spatial filtering algorithm is mainly used to eliminate the connected areas in the fingerprint details and process the low dynamic range image into a clear fingerprint ridge and valley image, thereby reducing the local image contrast difference caused by uneven tip pressure.

[0073] like Figure 6 As shown in the figure, fingerprints are composed of ridges and valleys. When the fingertip is pressed on the glass plate, the fingerprint ridges will adhere to the glass material, while the fingerprint valleys will not adhere to it. The adhesion between the ridges and the glass plate will form a double-layer structure, and the dispersion effect of this structure is completely different from that of a single-layer glass plate. Figure 6 The ultrasonic transducer array elements in the system send out different excitations respectively, and the waveform changes of the guided waves generated by the excitations before and after encountering fingertip pressure are collected to obtain the sound field signal.

[0074] In this embodiment, the thickness of the fingertip epithelium is defined as 1.5 mm, the height difference between the ridge and the valley is 0.25 mm, and the density ρ is 1.02 g / cm 3 , Young's modulus E is 2.19 GPa, and Poisson's ratio ν is 0.4. The thickness of the fingertip epithelium, the height difference between the ridge and the valley, the epithelial density, Young's modulus and Poisson's ratio are used as inputs to calculate the phase velocity dispersion curves before and after the ridge is attached to the glass plate using the global matrix method. The results are shown in Figure 7The gray solid curve represents the dispersion curve of the glass plate before the fingertip presses, and the black solid curve represents the dispersion curve of the double-layer material composed of the ridge and the glass after the fingertip presses. The operating frequency of the fingerprint collection ultrasonic transducer is 4.29MHz. In the figure, c p Represents the phase velocity. In the figure, A and S represent the antisymmetric mode and symmetric mode of the Lamb wave, respectively. In the frequency range of 0-7MHz, the A mode and S mode appearing from left to right are named A0, A1, S0 and S1 modes, respectively. At the same frequency, the A0 mode with a shorter wavelength will theoretically carry more sound field information. Compared with the glass structure of the glass plate, the material properties of the epithelium on the fingertips are closer to fluids. The fluid-solid coupling theory shows that after the ridges are bonded to the glass, the energy flow of the sound field is mainly concentrated in the adjacent mode below the phase velocity dispersion curve of the glass plate. At the operating frequency of 4.29MHz, the bonding of the ridges to the glass plate causes the phase velocity of the A0 mode to drop from 2.91mm / μs to 2.55mm / μs, that is, the sound speed drops by 12.37%.

[0075] By mapping the fingerprint ridge and valley image to the phase velocity interval [2.55,2.91] mm / μs, the fingerprint velocity model is obtained, which can characterize the thickness of the fingertip pressing the glass plate in the form of phase velocity, that is, converting 3D information into a 2D image, thereby improving the calculation speed of the fingerprint imaging method.

[0076] S3: Extract the first arrival time domain signal fragment of the phase velocity interval corresponding mode in the fingerprint sound field signal, perform frequency domain energy flow amplitude extraction to obtain the spectrum amplitude, and establish the frequency domain sound field matrix.

[0077] In this embodiment, the excitation signal uses a sine wave signal with a center frequency of 4.29 MHz, an amplitude of 10 V, and 10 cycles. Figure 8 The time domain signal between two transducers on the diameter of the ultrasonic transducer array is displayed. In the step of extracting the frequency domain energy flow amplitude of the A0 mode, the Gaussian weighted moving average filtering algorithm is used for smoothing.

[0078] Using the conversion relationship between phase velocity and group velocity, the group velocity dispersion curves before and after the ridge is bonded to the glass plate are calculated, as shown in Fig. 9 As shown in the figure, c g Represents group velocity. At the operating frequency of 4.29MHz, the group velocity of the A0 mode is 3.39mm / μs. There is an obvious mode separation phenomenon between the A0 mode, the A1 mode and the S0 mode, indicating that the A0 mode is easy to extract and has less overlap with the time domain signals of the A1 mode and the S0 mode. Using the difference in group velocity and the window function, the time domain signal of the A0 mode is extracted from the time domain signal of the collected sound field signal.

[0079] To further improve the calculation speed, the time domain signal collected by the ultrasonic instrument is processed into a frequency domain signal. The time domain signal of the extracted A0 mode is converted into a frequency domain signal by fast Fourier transform. To reduce the complexity of the frequency domain signal, the Gaussian weighted moving average filter algorithm is used to smooth the spectrum. The smoothed spectrum shows that fingertip pressing causes the amplitude of the acoustic field energy flow between the two transducers on the diameter to drop from a relative value of 3.47 to 2.35, and the amplitude reaches 32.28%, indicating that the fit between the fingerprint ridge and the glass plate has a significant impact on the acoustic field, and the acoustic field impact is theoretically proportional to the imaging quality. The spectrum amplitude is obtained by extracting the peak of the spectrum, and the smoothed spectrum amplitude extraction is written into the corresponding point of the frequency domain acoustic field matrix with a dimension of 128×128, and the frequency domain acoustic field matrix is ​​mapped to the interval range of [1,2] to avoid non-zero values ​​caused by the appearance of the value 0 in the imaging calculation.

[0080] The transducer array of the ultrasonic transducer adopts a method in which a single array element is excited in a counterclockwise direction and the entire array element array receives simultaneously. Therefore, using the above-mentioned frequency domain energy flow amplitude extraction method, a frequency domain sound field matrix with a dimension of 128×128 can be established. In order to avoid non-numerical problems caused by the existence of zero values ​​during the calculation process, the sound field energy flow amplitude is mapped to the interval [1,2].

[0081] The pure black grayscale image of the optical sensor when it is unloaded and the sound field collected by the fingerprint collection ultrasonic transducer when it is unloaded together constitute the baseline signal for obtaining the velocity model difference and the frequency domain sound field difference.

[0082] S4: Establish an offline training data set for fingerprint imaging, and obtain a descent gradient matrix through offline training; the fingerprint to be tested obtains the fingerprint velocity model and frequency domain acoustic field matrix of the test case through steps S1-S3, and obtains the fingerprint inversion image of the test case through online inversion of the descent gradient matrix.

[0083] In order to improve the breadth and diversity of the data set, an offline training data set is established in this embodiment, which includes imaging examples of fingertip epithelium in different dry states, different positions, and different pressing forces, with a total of 2048 examples. Each example includes a pair of corresponding fingerprint velocity models and frequency domain sound field matrices.

[0084] Fingerprint imaging offline training is performed on an HP workstation equipped with 32 Intel Xeon Gold 6230 CPUs, and the operating environment is MATLAB R2020a. The core formula of the supervised descent method used in offline training is:

[0085] R=(Δs T Δs+β 2 I) -1 ·(Δs T Δm);

[0086] Where R is the descending gradient matrix obtained through offline training, Δs=s0-s1 is the difference in the sound field matrix, s0 and s1 represent the frequency domain sound field matrices before and after fingertip pressing, respectively. β is a user-defined regularization weight. In this embodiment, the value of this variable is 1×10 -3 I is the unit matrix, Δm=m0-m1 is the fingerprint velocity model difference, m0 and m1 represent the fingerprint velocity model before and after fingertip pressing, respectively. In order to improve the computational efficiency of offline training, the velocity sound field matrix s and fingerprint velocity model m are both large matrices containing all examples, which enables the supervised descent method to output the descent gradient matrix R obtained through offline training in one iteration, that is, after all examples are trained once.

[0087] The fingerprint imaging method of the present invention concentrates most of the calculations in the offline training stage. The descent gradient matrix R obtained through offline training will be directly used for the online inversion of the test case:

[0088] M test =m0+Δs test ·R;

[0089] Where M test is the imaging result of the test case, Δs test It is the frequency domain sound field matrix corresponding to the test case.

[0090] S5: Build and train the post-processing model for image restoration and detail matching. The fingerprint inversion image of the test case obtains the fingerprint matching result through the trained post-processing model.

[0091] The post-processing model includes the DLCNN model for image restoration and the Mask R-CNN model for fingerprint matching, such as Fig.10As shown in the figure, the model includes a deep learning convolutional neural network (DLCNN) model for image restoration and a Mask R-CNN model for fingerprint matching, which are used to improve the quality of fingerprint reconstruction images and provide authentication and anti-counterfeiting properties. The DLCNN model adopts a small-scale, deep-layer architecture. All convolutional layers use a 3×3 dimension, the network depth is 20 layers, and linear rectifier units are set after the 1st to 19th convolutional layers. Batch normalization is set between the 2nd to 19th convolutional layers and linear rectifier units. This architecture is suitable for image processing. The linear rectifier unit is an activation function, which mainly increases the nonlinear characteristics of the network, allowing the network to learn more complex features and improve the training efficiency of the model. Batch normalization is a technology used to improve training speed and stability. By normalizing each small batch of data, the output distribution of the middle layer of the network is more stable and the problem of internal covariate shift is reduced. The fingerprint inversion image of the test case is converted into a grayscale image and input into the input layer of the DLCNN model through image cropping. The cropping of the input image is to adapt to the network structure, improve processing efficiency and restoration effect. The main functions are to adapt to the network input size, reduce the amount of calculation, avoid boundary effects, and improve the generalization ability of the model. The cropped image is input into each convolution layer of the middle layer through the image input layer in turn, and the output of the middle layer is regressed to obtain the grayscale image of the restored fingerprint.

[0092] In this embodiment, the training data set of the DLCNN model is divided into two parts, one part is used to train the artifact removal ability, and the other part is used to train the image denoising ability. Each example in the training data set includes a noisy image containing artifacts or noise and a pure image. The training data set includes a grayscale image of a fingerprint and a grayscale image of a natural scenery image, which is used to train the artifact removal performance and image denoising performance. The two parts of the data set contain 13195 and 150 examples respectively, and each example contains two grayscale images, the artifact removal part: an artifact-containing fingerprint inversion image and a corresponding pure fingerprint image; the image denoising part: a natural scenery image containing Gaussian noise and a corresponding pure image. Artifact removal is the main function, denoising is a secondary function, and there are few denoising examples. Moreover, denoising training uses a natural scenery image with high complexity, which can improve the comprehensiveness of denoising as much as possible with fewer examples. The training data set is divided into a training set and a test set in a ratio of 8:2. The training process was carried out in the Deep Learning Toolbox of MATLAB R2020a and accelerated using an NVIDIA GeForce RTX 2080Ti GPU. The optimizer used was SGDM (stochastic gradient descent momentum), and the gradient threshold and momentum parameters were set to 5×10 -3and 0.9, the batch size, maximum number of iterations, initial learning rate, and L2 trade-off parameters are set to 32, 30, 0.1, and 0.1×10, respectively. -3 .

[0093] The DLCNN model is used to obtain the repaired fingerprint grayscale image, which is converted into a binary form of the fingerprint image through an even-symmetric Gabor filter, and the binary form of the fingerprint image is input into the Mask R-CNN model to enhance the ridge and valley information of the fingerprint. The repaired fingerprint grayscale image is passed through the Mask R-CNN model to obtain the fingerprint matching result image.

[0094] The Mask R-CNN model consists of five parts: ResNet101 feature extraction network, feature pyramid network, region proposal network, region of interest arrangement and full convolution network. The ridge end and ridge bifurcation of the fingerprint are key details representing personal information. In order to integrate the relative position information of multiple fingerprint details, a large-scale marking method covering multiple fingerprint details is adopted. A total of 8 large-scale fingerprint details are marked. Fingerprint marking is a manual operation, and different registered fingerprints have different marking results. The marked details are as follows: Fig.10 The fingerprint details are marked on the top right. The fingerprint marking uses the Labelme 5..2.1 plug-in, and the output is a .json file containing the labeling information, which is directly input into the Mask R-CNN model for training. That is, the binary data set is marked and converted into a .json file before starting training.

[0095] In this embodiment, the fingertips of the second inventor's right hand are used as registered fingertips. The training data set of the Mask R-CNN model is the fingerprint image of the registered fingertips and its detail marks. Through training, the Mask R-CNN model can recognize the classification, regression analysis and mask of these details, thereby achieving detail matching of the test example fingerprint and outputting its fingerprint matching information and image.

[0096] The training dataset of the Mask R-CNN model includes 4680 binary grayscale images of fingerprints collected by optical sensors, with fingerprint labels attached. The training dataset is divided into training set and test set in a ratio of 9:1. The training is carried out in the Tensorflow2.6.0 environment, the code editor is PyCharm, and the NVIDIA GeForce RTX 2080Ti GPU is used for acceleration. In the training parameters, the batch size and initial learning rate are set to 16 and 0.02 respectively.

[0097] S6: Establish a personal authentication confidence algorithm and use the fingerprint matching result to calculate the personal authentication confidence for fingerprint recognition.

[0098] Among the eight large-scale fingerprint details marked, the simple and symmetrical types have higher uniqueness. For this reason, the uniqueness of the fingerprint details of the registered fingertips is divided into three levels, such as Fig.10 As shown. In dense fingerprint morphology, simple and symmetrical details have higher uniqueness. Level 1 fingerprint details have the highest uniqueness and are selected as key features to represent personal information. Level 2 fingerprint details have the second highest uniqueness, while level 3 fingerprint details have relatively weak uniqueness. For this registered fingertip, the personal authentication confidence calculation formula is established:

[0099]

[0100] In the formula, c f To authenticate the trust of the representative individual. k is the compensation factor, which is an empirical value related to the marked fingerprint details and varies with the number of fingerprint details k∈[0,8] recognized by the Mask R-CNN model. The values ​​of ξ0 to ξ8 are 0, 3, 3, 15 / 7, 5 / 3, 10 / 7, 5 / 4, 10 / 9 and 1 respectively. b is the weight, and the weight b values ​​corresponding to the fingerprint details of levels 1 to 3 are 1 / 6, 2 / 15 and 1 / 10 respectively. When all the 8 large-scale fingerprint details of the mark are recognized, the personal authentication confidence c f The value of is 0.99. Compensation factor ξ k The values ​​of and b can be adjusted flexibly.

[0101] S7: Anti-counterfeiting testing of fingerprint representation using descent gradient matrix and post-processing model.

[0102] First, test the registered fingertip. Press the fingertip on the fingerprint collection ultrasonic transducer and perform the operation of step S3 to obtain the frequency domain acoustic field matrix difference generated after the fingertip is pressed. Then, download the descending gradient matrix generated by offline training in step S4, perform online inversion of the fingerprint, and obtain the imaging result, such as Fig.11 As shown in the figure, the imaging range is 120 mm × 120 mm. To evaluate the imaging quality, the operation of step S2 is performed to obtain the fingerprint velocity model. Taking the fingerprint velocity model as the benchmark, the root mean square error (RMSE) value of the imaging result is 8.65 × 10 -3 , indicating that the imaging accuracy is high.

[0103] The imaging range is tracked to the 36mm×36mm area around the fingerprint, and the DLCNN model built in step S5 is used for image restoration. After restoration, the artifacts in the image are effectively removed. Subsequently, the MaskR-CNN model built in step S5 is used for fingerprint detail matching, and 5 large-scale details are successfully identified. Using the personal authentication confidence algorithm established in step S6, the authentication confidence of the registered fingertip example is calculated to be 0.94, and the personal authentication is passed.

[0104] To test the anti-counterfeiting performance, the authentication attack was performed using the fingertips of others. The test steps were the same as those for the registered fingertips. Within the imaging range of 120 mm × 120 mm, the RMSE value was 1.35 × 10 -2 After image restoration by the DLCNN model and fingerprint detail matching by the Mask R-CNN model, the calculated authentication confidence is 0.30, which fails personal authentication.

[0105] The results of the above tests show that, through offline training, the ultrasonic guided wave fingerprint imaging method proposed in the present invention effectively learns the nonlinear mapping relationship between the fingerprint morphology change caused by fingertip pressing and the acoustic field matrix change, thereby accurately characterizing the ridge and valley distribution of the fingerprint. In addition, the image post-processing model has the ability to repair images and match fingerprint details, which gives authentication and anti-counterfeiting properties.

[0106] S8: High sensitivity test of fingerprint characterization, resulting in a high sensitivity fingerprint authentication method.

[0107] like Fig.12 As shown, when the fingertips contain moisture or dust, the quality of the fingerprint image collected by the optical sensor is significantly affected, showing a decrease in the contrast between ridges and valleys or discontinuous ridge and valley morphology, which indicates that the traditional optical fingerprint scanning method is easily interfered by environmental pollution.

[0108] In contrast, after repeating the same operation of registering the fingertip test in step S7, the inverted fingerprint morphology shows the ability to resist interference from moisture or dust, and the RMSE values ​​are 1.76×10 -2 and 1.09×10 -2 ,The prediction accuracy is high. After image restoration and fingerprint minutiae matching, the authentication confidence is 0.73 and 0.80 respectively, both passing personal authentication.

[0109] The test results under the above environmental pollution conditions show that the ultrasonic guided wave fingerprint imaging method proposed in the present invention can effectively resist the influence of environmental pollution on imaging and authentication. While achieving zero occupancy of the space under the transducer screen, it also ensures high-sensitivity imaging capability and authentication anti-counterfeiting performance.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method, characterized in that: The steps include: S1: Build an ultrasonic transducer for fingerprint collection and place it on the optical sensor. The ultrasonic transducer collects the acoustic field signal of the fingerprint and the optical sensor collects the grayscale image of the fingerprint. S2: preprocessing the fingerprint grayscale image, and calculating the phase velocity interval used to characterize the fingerprint based on the global matrix method, and establishing a fingerprint velocity model according to the preprocessed fingerprint grayscale image; S3: extract the first arrival time domain signal fragment of the phase velocity interval corresponding mode in the sound field signal of the fingerprint, extract the frequency domain energy flow amplitude to obtain the spectrum amplitude, and establish the frequency domain sound field matrix; S4: Using the methods of steps S2 and S3, establish an offline training data set for fingerprint imaging, obtain a descent gradient matrix through offline training; build a post-processing model for image restoration and detail matching and perform training; S5: The acoustic field signal of the fingerprint to be authenticated is collected by an ultrasonic transducer, and the grayscale image of the fingerprint to be authenticated is collected by an optical sensor. The fingerprint velocity model and the frequency domain acoustic field matrix of the test case are obtained by using the methods of steps S2 and S3. The fingerprint inversion image of the test case is obtained by online inversion of the descending gradient matrix; the fingerprint inversion image is input into the trained post-processing model to obtain the fingerprint matching result; S6: Calculate personal authentication confidence using the fingerprint matching result, and identify the fingerprint to be authenticated through the personal authentication confidence.

2. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 1 is characterized in that: The ultrasonic transducer comprises a plurality of ultrasonic transducer array elements arranged in a circular ring shape, the ultrasonic transducer array element comprising a piezoelectric ceramic sheet group (1), a positive copper foil (2) and a negative copper foil (3) are respectively arranged on the upper side and the lower side of the piezoelectric ceramic sheet group (1), a glass plate (4) is arranged on the lower side of the negative copper foil (3) and the upper side of the positive copper foil, and the positive copper foil (2) and the negative copper foil (3) are both connected to an ultrasonic instrument; an insulating layer (8) is laid between the positive copper foil (2) and the negative copper foil (3), and a sound field boundary absorption layer (9) is arranged at the outermost edge between the two glass plates.

3. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 2 is characterized in that: The piezoelectric ceramic sheet group (1) is composed of at least one circular PZT-5H thin sheet, the positive copper foil (2) and the negative copper foil (3) are respectively fixedly connected to the PZT-5H thin sheet by means of a conductive silver glue (6), and the negative copper foil (3) is fixed to a glass plate (4) on the lower side of the negative copper foil (3) by means of an epoxy resin adhesive (5).

4. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 3 is characterized in that: The PZT-5H thin sheet is made by cutting and polishing a PZT-5H raw material, and the positive electrode copper foil (2) and the negative electrode copper foil (3) are both obtained by laser cutting; The positive electrode copper foil (2) and the negative electrode copper foil (3) both include circular endpoints with the same diameter as the PZT-5H sheet, and leads are arranged on the circular endpoints. The leads of the negative electrode copper foil (3) are connected to the annular copper foil, and the annular copper foil is grounded; the outermost part of the negative electrode copper foil (3) is directly connected to a square copper foil, and the four corners of the square copper foil are divided into straight copper foil and right-angle copper foil by breakpoints. The straight copper foil is folded outward by 90 degrees, and the copper-clad surface is turned downward to the same side as the copper-clad surface of the positive electrode copper foil (2), and the right-angle copper foil is reserved as a debugging contact point (10); The positive electrode copper foil (2) is equally divided into four parts, and the copper-clad surface of each part of the positive electrode copper foil faces downward; each part of the positive electrode copper foil is fixed by a polydimethylsiloxane film (7); the outermost part of the glass plate (4) is surrounded by plasticine to serve as a sound field boundary absorption layer (9); A layer of transparent polydimethylsiloxane film is laid between the surface glass of the optical sensor and the glass plate (4) of the ultrasonic transducer.

5. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to any one of claims 1 to 4, characterized in that: The method for preprocessing the fingerprint grayscale image is as follows: tracking the fingerprint and extracting preliminary features to obtain preliminary features of the fingerprint image; extracting the target area in the preliminary features of the fingerprint image by threshold control; converting the extracted target area into a low dynamic range image by using a bilateral filtering algorithm, and eliminating the connected areas in the fingerprint details by using a spatial domain filtering algorithm with a convolution kernel to obtain a fingerprint ridge and valley image.

6. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 5, characterized in that: The method for establishing the fingerprint velocity model is as follows: taking the thickness of the fingertip epithelium, the height difference between the ridge and the valley, the epithelium density, the Young's modulus and the Poisson's ratio as inputs, using the global matrix method to calculate the phase velocity dispersion curve of the fingertip ridge before and after the glass plate of the ultrasonic transducer is attached, and selecting the A0 mode with a shorter wavelength and more acoustic field information at the same frequency; compared with the glass structure of the glass plate, the material properties of the fingertip epithelium are closer to the fluid, and the fluid-solid coupling theory shows that after the ridge is attached to the glass, the acoustic field energy flow is mainly concentrated in the adjacent mode below the phase velocity dispersion curve of the glass plate; at the center frequency of the ultrasonic transducer, the ridge and the glass plate are attached to each other, resulting in the phase velocity of the A0 mode decreasing from 2.91 mm / μs to 2.55 mm / μs, and the phase velocity interval is [2.55, 2.91] mm / μs; The fingerprint ridge-valley image is mapped to the phase velocity interval [2.55, 2.91] mm / μs to obtain the fingerprint velocity model.

7. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 6, characterized in that: The method for extracting the first arrival wave time domain signal fragment of the mode corresponding to the phase velocity interval in the sound field signal of the fingerprint is: using the conversion relationship between the phase velocity and the group velocity to calculate the group velocity dispersion curve before and after the ridge-bonded glass plate, at the center frequency of the ultrasonic transducer, there is an obvious mode separation phenomenon between the A0 mode and other modes, and using the group velocity difference and the window function, the time domain signal of the A0 mode is extracted from the time domain signal of the collected sound field signal; The method for extracting the frequency domain energy flow amplitude to obtain the spectrum amplitude is as follows: converting the time domain signal fragment of each ultrasonic transducer array element into a frequency domain signal by frequency domain transformation, smoothing the frequency domain signal, extracting the peak value of the frequency domain signal spectrum after smoothing to obtain the spectrum amplitude, writing the extracted spectrum amplitude into the corresponding point of the frequency domain sound field matrix, and mapping the frequency domain sound field matrix to the interval range of [1,2].

8. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 5 or 6, characterized in that: The supervised descent method used in the offline training is: R=(Δs T Δs+β 2 I) -1 ·(Δs T Δm); Where R is the descending gradient matrix obtained through offline training, Δs=s0-s1 is the difference in the sound field matrix, s0 and s1 represent the frequency domain sound field matrix before and after fingertip pressing, respectively; β is the regularization weight, Δm=m0-m1 is the difference in the fingerprint velocity model, m0 and m1 represent the fingerprint velocity model before and after fingertip pressing, respectively; the velocity sound field matrix s and the fingerprint velocity model m are both data matrices containing all examples; The online inversion is performed by descending the gradient matrix R test case to obtain the imaging result M of the test case test =m0+Δs test ·R;Δs test It is the frequency domain sound field matrix corresponding to the test case.

9. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 8, characterized in that: The method for tracking and extracting preliminary features of fingerprints is as follows: in the grayscale image of the fingerprint, the point with the largest grayscale value is tracked, and the grayscale image within the range of 36mm×36mm around the point with the largest grayscale value is extracted, and the grayscale image is inverted to extract the range with grayscale values ​​below 110; the threshold control method is to discretely invert the grayscale image with grid spacing, and retain the part with grayscale average value below 80 in the grid; The smoothing process adopts Gaussian weighted moving average filtering algorithm; The method for establishing an offline training data set is: collecting sound field signals and fingerprint grayscale images of fingertip epithelium in different dry states, at different positions, and with different pressing forces, and obtaining a fingerprint velocity model and sound field signal obtained for each fingertip pressing through the methods of steps S2 and S3 to form an example.

10. The highly sensitive and anti-counterfeiting ultrasonic guided wave fingerprint imaging and authentication method according to claim 9, characterized in that: The post-processing model includes a DLCNN model for image restoration and a Mask R-CNN model for fingerprint matching. The DLCNN model adopts a small-scale, deep-layer architecture, all convolutional layers adopt a 3×3 dimension, and the network depth of the DLCNN model is 20 layers; the Mask R-CNN model includes a ResNet101 feature extraction network, a feature pyramid network, a region proposal network, an area of ​​interest arrangement, and a full convolutional network connected in sequence; the DLCNN model is used to obtain a restored fingerprint grayscale image, and the restored fingerprint grayscale image is converted into a binary form of a fingerprint image through an even-symmetric Gabor filter and then input into the Mask R-CNN model, and the restored fingerprint grayscale image is passed through the Mask R-CNN model to obtain a fingerprint matching result image; The training data set of the DLCNN model includes a fingerprint grayscale image of the artifact removal part and a grayscale image of the natural scenery image of the image denoising part. Each example of the artifact removal part includes an artifact-containing fingerprint inversion image and a corresponding pure fingerprint image; each example of the image denoising part includes a natural scenery image containing Gaussian noise and a corresponding pure image. The training data set of the Mask R-CNN model includes the binary form of the grayscale images of fingerprints of multiple registered fingertips collected by the optical sensor and attached with fingerprint marks; A large-scale marking method covering multiple fingerprint details was adopted, marking 8 large-scale fingerprint details; The method for calculating the personal authentication confidence is: the uniqueness of the fingerprint details of the registered fingertips is divided into three levels, the fingerprint details of level 1 have the highest uniqueness and are selected as the key features representing personal information, the uniqueness of the fingerprint details of level 2 is second, and the uniqueness of the fingerprint details of level 3 is relatively weak; then the personal authentication confidence is: In the formula, ξ k is the compensation factor, the compensation factor is an empirical value related to the marked fingerprint details, which varies with the number of fingerprint details k∈[0,8] recognized by the Mask R-CNN model, and b is the weight of levels 1 to 3.

Images