Paper currency digital anti-counterfeiting method based on multi-element fusion tactile reproduction

Abstract

The present application relates to a kind of paper currency digital anti-counterfeiting method based on multi-element fusion haptic reproduction, belong to the cross field of information security and man-machine interaction.It includes the extraction of haptic feature, the haptic feature " formula " is adopted by BP neural network model, the decryption of user, the rendering model information and formula information obtained are sent to haptic feature rendering unit, the weight in formula information is distributed to rendering model information, the rendering model information with weight is obtained, then the rendering model information with weight obtained is sent to signal driving unit, the haptic driving signal that can realize haptic reproduction in haptic interaction interface is obtained, the function of haptic reproduction is realized, to realize the multi-element fusion haptic anti-counterfeiting.The beneficial effect is to propose a kind of " formula " based on shape, texture, temperature and other haptic fusion, give the most real naked finger haptic feedback, can be applied in the anti-counterfeiting of digital renminbi, the authenticity of digital album, the authenticity of digital ticket and other directions.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of information security and human-computer interaction, and specifically relates to a digital anti-counterfeiting method for banknotes based on multi-dimensional fusion tactile reproduction. Background Technology

[0002] With the rapid development of the information age, the internet is constantly changing the way Chinese people live, work, and learn, and is increasingly integrated into all aspects of their lives. The rapid popularization of the internet has drastically changed how people obtain information and has also had a significant impact on their sense of participation in social life. The rise of the internet signifies a shift in the mediating mechanism of human thought, evolving from behavioral and linguistic mediating to digital mediating. In the binary digital mediating system, thought has become behavioralized; thought has become a tangible system, just like a behavioral process—a visible and repeatable "software." The visualization of thought, with the emergence and development of graphical browsers, has promoted the popularization of computer network applications and made science and technology more accessible and applicable to people. However, the internet not only brings convenience but also drawbacks, namely, a gradual decrease in the sense of participation in social life, especially for those special groups who need interactive means to understand the world.

[0003] Humans interact with the world through sight, hearing, touch, and smell. Take the blind as an example: they cannot see, so they rely on sound, touch, and smell to perceive things. When walking with a cane, they identify the tactile paving by the different sounds they make when tapping on the ground. When acquiring knowledge, they decipher Braille by touch. When picking up food, they can distinguish it by its different smells. Similarly, the deaf and mute perceive the world only through sight, touch, and smell. When they encounter visually similar objects, they cannot judge their authenticity with their eyes and must rely on touch and smell. In situations where they can only see, their situation becomes precarious. For them, their perception of society relies entirely on interaction with things. However, this interaction is decreasing. For example, with the rise of online payments, few people use paper money for transactions. While this brings convenience, it also reduces the sense of participation. For most people, receiving payment can be determined by seeing the information displayed on a mobile phone and hearing the sound. However, for certain groups, some individuals lack both visual and auditory senses simultaneously, which can cause inconvenience in their lives and reduce their sense of participation in social life. Furthermore, the act of touching the surface of banknotes to determine their authenticity has gradually disappeared. Therefore, to improve the sense of participation of these groups in social life, this paper proposes a multimodal tactile anti-counterfeiting method based on tactile reproduction technology.

[0004] Haptic reproduction is a human-computer interaction technology based on the human tactile perception mechanism. It can also be viewed as a digital technology of touch, achieving the perception and reproduction of surface texture information of distant or virtual objects by simulating and calculating the tactile perception process. For a long time, research on tactile reproduction did not receive much attention. In recent years, the rapid development of software and hardware technologies in the IT field, along with the emergence of the concept of the "metaverse" with the development of VR / AR, has brought tactile reproduction into the spotlight. The technology has matured rapidly, resulting in a large number of tactile reproduction technologies, especially those for bare-finger interaction. This makes it theoretically feasible for this invention to achieve anti-counterfeiting based on tactile reproduction on smart terminals.

[0005] In applications where existing smart devices can reproduce various tactile sensations, such as shape, texture, and temperature, tactile reproduction devices based on a single feedback mechanism cannot effectively integrate and coordinate to achieve the most ideal tactile perception. Therefore, it is necessary to combine multiple tactile sensations to achieve the most realistic feel. However, when multiple tactile reproduction devices operate simultaneously, the intensity of one or more sensations may become too high, masking the perception of others, thus failing to accurately reproduce the original tactile sensation.

[0006] To summarize the points mentioned above, in order to achieve multimodal tactile anti-counterfeiting based on tactile reproduction technology, it is necessary to solve the problem of how to coordinate multiple tactile reproduction devices to obtain the most realistic virtual tactile sensation.

[0007] (1) Due to the limitations of the dynamic range and dimension of single tactile reproduction feedback force, current rendering methods can often only reproduce a certain surface attribute of the visual object, which affects the realism and richness of the tactile reproduction effect.

[0008] (2) In the context of multiple tactile sensations being reproduced together, how can a mathematical model be found to arrange the various tactile reproduction devices so that the user can get the best tactile sensation?

[0009] Based on the above background analysis, the arrival of the digital age, by virtualizing some everyday things (such as RMB), has brought many conveniences to social life. However, for certain groups, their sense of experience in life has also decreased, and it may even cause them a lot of trouble, especially when they face digital goods, they find it difficult to distinguish between genuine and counterfeit products. Their means of judging authenticity are limited to real-world items, but digital goods do not possess this attribute, which reduces their sense of participation in social life. Summary of the Invention

[0010] This invention provides a digital anti-counterfeiting method for banknotes based on multi-dimensional fusion tactile reproduction. It proposes to provide virtual objects and banknotes with virtual tactile feedback to special groups, that is, to provide them with a real sense of experience in the digital age, namely tactile sensation. They can judge the authenticity of an object by their experience with real objects, thereby enhancing their sense of participation in social life.

[0011] The technical solution adopted by this invention includes the following steps:

[0012] Step 1: Extraction of tactile features, including shape tactile feature extraction, texture tactile feature extraction, and temperature tactile feature extraction;

[0013] Step 2: Use a BP neural network model to "recipe" the tactile features. The process of the BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer. The second stage is the backward propagation of the error, from the output layer to the hidden layer, and finally to the input layer. The weights and biases from the hidden layer to the output layer and from the input layer to the hidden layer are adjusted in sequence.

[0014] Step 3: User's decryption;

[0015] Step 4: Tactile Reproduction. The obtained rendering model information and recipe information are sent to the tactile feature rendering unit. The weights in the recipe information are assigned to the rendering model information to obtain weighted rendering model information. The weighted rendering model information is then sent to the signal driving unit to obtain a tactile driving signal that can realize tactile reproduction in the tactile interactive interface, thereby realizing the tactile reproduction function and achieving multi-dimensional tactile anti-counterfeiting.

[0016] The shape tactile feature extraction method based on the image shape height model in step 1 of this invention is as follows:

[0017] First, a weighted average method is used to convert the color image into a luminance image:

[0018]

[0019] in This is the converted image brightness value matrix. This is the red value matrix of the original image. This is the green value matrix of the original image. This is the blue value matrix of the original image;

[0020] Based on the light source incidence model, obtain the reflection function of the image:

[0021]

[0022] In the formula, and It is a pixel. The gradients that vary along the X and Y directions at the given locations, This is the reflection function, which contains information about the light source on the object's surface. and The angles between the direction of the light source and the x-axis and z-axis are called the light source tilt angle and deflection angle, respectively, and the light source vector gradient. , , The angle of inclination, Angle of elevation;

[0023] Using the finite difference method of the latter term, the independent variable of the reflection function is... and Discretization:

[0024]

[0025]

[0026] in The height matrix of the object shape;

[0027] Rearranging the above equation, we get: in Indicates that the independent variable is , , , The quaternion function will about Performing a Taylor expansion, we get:

[0028] Where n is the number of iterations.

[0029] use Alternative ,

[0030] and Simplifying the above equation, we get:

[0031]

[0032] in, It can be represented as:

[0033]

[0034] Let the initial value Substituting the above formula into the formula above that, and iterating n times, the corresponding pixel is calculated. height value Then, a normalization method is used to obtain the image shape and height model. Record the shape rendering model information and record the extracted shape tactile features.

[0035] The texture tactile feature extraction method based on the image texture height model in step 1 of this invention:

[0036] First, a weighted average method is used to convert the color image into a luminance image:

[0037]

[0038] Then adopt window Extract grayscale image Local image texture characteristics, As a periodic sequence, As a point A univariate function centered at a period of 8. for The corresponding local Fourier transform coefficient matrix, These are frequency domain coefficients;

[0039] get

[0040]

[0041] for Given an image of a certain size, obtain the set of local Fourier coefficients of the image;

[0042]

[0043] Among them, when When taking a specific value, This can represent the image texture height model of the corresponding frequency domain coefficients, denoted here as... This information is recorded as the texture rendering model information, and the extracted texture tactile features are also recorded.

[0044] The method for extracting temperature-tactile features based on image tone information in step 1 of this invention:

[0045] The specific steps to convert a color image based on the RGB color space into an image using the HSV model are as follows:

[0046] set up , , For point The RGB color values ​​at each point are normalized to be between 0 and 1, resulting in the points. Normalized RGB color values This is recorded as temperature rendering model information, and the extracted temperature tactile features are also recorded.

[0047] The recorded information—shape rendering model information, texture rendering model information, and temperature rendering model information—is stored in a secure storage chip.

[0048] The specific process of obtaining the "recipe" using a BP neural network in step 2 of this invention is as follows:

[0049] (1) Data Collection: Tactile data of the banknote's temperature, texture, and shape under normal temperature conditions are collected using the method in step 1. These individual tactile sensations can then be reproduced using the corresponding tactile reproduction device. The intensity of each of the three tactile sensations is randomly assigned (from 0 to 1). The intensity of the tactile reproduction device is set to [value missing]. It collects tactile perceptions of different individuals at random intensities, recording whether the tactile sensation at these intensities corresponds to the feel of banknotes. Simultaneously, it sets a machine learning objective: to subjectively judge whether the tactile sensation is real or inauthentic, and then integrates the data with... and The random intensity data of real touch will be set to 1, and otherwise set to 0, to form a data training library and be used to train the model;

[0050] (2) Selection of excitation function:

[0051] (3) Output of hidden layer: Data obtained from the database and Substitute middle;

[0052] Where f is the activation function in (2), N is the number of nodes in the input layer, N takes the value of 3, and i takes the values ​​of 1, 2, and 3, which correspond to temperature, texture, and shape respectively; while j takes the values ​​of 1, 2, and 3, which represent the input layer, hidden layer, and output layer respectively. The bias from the input layer to the hidden layer; The weights from the input layer to the hidden layer; This is the output of the hidden layer;

[0053] (4) Output of the output layer:

[0054] Where l is the number of nodes in the hidden layer, and l takes the value of 3; k takes the values ​​of 1, 2, and 3; The bias from the hidden layer to the output layer; The bias from the hidden layer to the output layer; For the output of the output layer;

[0055] (5) Initialize the weights and biases, and set the learning rate. Start calculating the error;

[0056] error:

[0057] The expected output is The error is E; the output of the output layer is

[0058] (6) Continuously update the weights and biases based on the error:

[0059] Weight update:

[0060] Bias update:

[0061] in For learning rate;

[0062] (7) Continue until the optimal solution is obtained, and record the obtained weights. ( , , The "recipe" is defined by the weights of temperature, texture, and shape, in that order.

[0063] Obtained weight records ( , , The recipe information is recorded in a secure storage chip.

[0064] In step 3 of this invention, the user enters a private key on the human-computer interaction interface, and decrypts it according to the corresponding public key stored in the secure storage chip. If decryption fails, the human-computer interaction interface will prompt "Error, re-enter"; if the input is correct, the rendering model information and recipe information will be extracted from the secure storage chip.

[0065] The secure storage chip stores: rendering model information, party information, and the public key corresponding to the private key entered by the user; the user enters the private key on the haptic interface to perform security verification.

[0066] The air pressure film tactile reproduction method based on the image shape height model in step 4 of this invention:

[0067] Based on the shape rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. The gradient matrix of the image shape can be obtained. :

[0068]

[0069] in, and Represent matrices respectively exist and Partial derivatives in direction, gradient matrix This indicates the rate of edge transformation at each point, and follows the same trend as the magnitude of the tangential force. The tangential force matrix can be used to represent this rate of change. Represented as:

[0070]

[0071] in, This refers to the tangential force felt when a finger swipes across the screen without haptic feedback. The maximum tangential force that electrostatic tactile sensation can increase. This represents the maximum tangential force that the air pressure membrane can reduce tactile energy.

[0072] in, This represents the normal force applied by the finger, and the tangential force matrix is ​​represented by the air pressure film matrix. :

[0073] The relationship between the driving voltage and the electrostatic feedback force and the air pressure membrane feedback force was determined based on human-computer interaction experiments:

[0074] in, The voltage amplitude of the air diaphragm drive signal. For air pressure diaphragm feedback force, This is the proportionality coefficient between the air pressure diaphragm drive signal and the feedback force;

[0075] From the above four equations, the amplitudes of the electrostatic force drive signal and the air pressure diaphragm drive signal can be obtained:

[0076]

[0077] in for The amplitude of the air pressure film driving signal applied at the point is denoted as the shape tactile driving signal;

[0078] After obtaining the shape tactile drive signal, it is sent to the tactile interaction interface. The shape tactile sensation can be reproduced using tactile reproduction technology. If a tactile reproduction device is required, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the drive signal according to the mechanical vibration tactile needs. Then, the amplitude value of the signal can be amplified by the power amplifier to realize the shape tactile sensation reproduction.

[0079] The electrostatic tactile reproduction method based on the image texture height model in step 4 of this invention:

[0080] Based on the texture rendering model information obtained in step 1 The "recipe" yields weighted texture rendering model information. The image texture gradient matrix can be calculated. :

[0081]

[0082] in, and Represent matrices respectively exist and Partial derivative in direction;

[0083] Normalizing the texture gradient matrix, we get:

[0084]

[0085] in, For the normalized texture gradient matrix, and These represent the minimum and maximum values ​​in the texture height matrix, and the texture gradient matrix, respectively. This represents the local concavity and convexity of an image; that is, the smaller the gradient value, the smoother the texture surface, and the weaker the tactile sensation when a finger glides across it. The normalized texture gradient matrix is ​​mapped according to the specified ratio. Matching electrostatic force magnitude:

[0086]

[0087] The mapping relationship between electrostatic driving voltage and electrostatic force:

[0088]

[0089] Obtain the amplitude matrix of the electrostatic force driving signal, denoted as the texture tactile driving signal;

[0090] After obtaining the texture tactile driving signal, it is sent to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. Then, the amplitude value of the signal can be amplified by the power amplifier, thus realizing the reproduction of texture tactile sensation.

[0091] The temperature-sensitive haptic rendering method based on image tone information in step 4 of this invention:

[0092] Based on the temperature rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. For point ,set up:

[0093]

[0094]

[0095] in, and Normalized color values , , The maximum and minimum values;

[0096] point Image tone at the location It can be determined by the following method:

[0097]

[0098] point Image saturation at the location It can be determined by the following method:

[0099]

[0100] point Image brightness at the location It can be determined by the following method:

[0101]

[0102] From this, we obtain Value at Within the interval, The value range is divided into four intervals: red ,yellow ,green ,blue Then, based on the four predefined zones, the color tones falling into different zones are... The system is designed with four pure colors: red, yellow, green, and blue. The corresponding RGB values ​​are returned and recorded as temperature-sensitive tactile driving signals. After obtaining the temperature-sensitive tactile driving signals, they are sent to the tactile interaction interface, and the texture tactile sensation can be reproduced using tactile reproduction technology.

[0103] A device for digital anti-counterfeiting of banknotes based on multi-source fusion tactile reproduction includes a tactile interactive interface, a secure storage chip, a signal rendering unit, and a signal driving unit, wherein:

[0104] The haptic interface is used by users to enter their private keys for security verification.

[0105] The secure storage chip stores: rendering model information, recipe information, a public key corresponding to the user-input private key, and the resulting weight record. ( , , These are shape rendering model information, texture rendering model information, and temperature rendering model information, which are recorded in a secure storage chip as recipe information. The rendering model information obtained from the secure storage chip is sent to the signal driving unit after being weighted by the recipe.

[0106] After receiving the tactile driving signal, the signal driving unit sends it to the tactile interactive interface;

[0107] The tactile interactive interface integrates a tactile reproduction device to achieve tactile reproduction, thereby realizing multi-faceted tactile anti-counterfeiting, specifically including:

[0108] After receiving the shape tactile drive signal, the tactile interaction interface sends it to the tactile interaction interface. The shape tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the mechanical vibration tactile requirements, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the drive signal. Then, the power amplifier can amplify the amplitude value of the signal to realize the shape tactile sensation reproduction.

[0109] After receiving the texture tactile driving signal, the tactile interaction interface sends it to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. The power amplifier can then amplify the amplitude of the signal to achieve texture tactile reproduction.

[0110] After receiving the temperature-sensitive tactile drive signal, the tactile interaction interface sends it to the interface. The texture tactile sensation can then be reproduced using tactile reproduction technology. This function requires a tactile reproduction device, including a digital-to-analog converter (DAC), a power amplifier, and a temperature measurement circuit. The DAC receives the stimulation pattern sent by the microcontroller and generates a corresponding heating or cooling drive signal. The drive signal is then amplified by the power amplifier to obtain the output temperature-sensitive tactile drive signal. The temperature measurement circuit monitors the temperature of the semiconductor cooling chip surface and transmits the real-time temperature back to the microcontroller. The microcontroller calculates the required stimulation pattern based on the reference temperature, temperature difference, and current temperature duration, and then sends the signal back to the DAC to achieve temperature-sensitive tactile reproduction.

[0111] The beneficial effects of this invention are as follows: It proposes a "recipe" based on the fusion of shape, texture, and temperature tactile sensations. In application scenarios where existing smart devices can achieve various tactile reproductions, this "recipe" can effectively control the weights of various virtual tactile attributes, providing users with the most realistic bare-finger tactile feedback. It also proposes tactile reproduction methods for shape, texture, and temperature. According to different implementation principles, tactile reproduction devices can be divided into mechanical tactile reproduction, vibration tactile reproduction, array tactile reproduction, and friction-controlled tactile reproduction types. For example, an air-pressure membrane tactile reproduction device can be used to achieve shape tactile reproduction, an electrostatic tactile reproduction device can achieve texture tactile reproduction, and image tone information can achieve temperature tactile reproduction. The proposed digital anti-counterfeiting method device based on multi-element fusion tactile reproduction includes: a tactile interactive interface, a secure storage chip, a signal rendering unit, and a signal driving unit. This invention solves the anti-counterfeiting problem by implementing a unique virtual tactile sensation on a smart terminal. Its applications include: anti-counterfeiting of digital RMB, authenticity verification of digital albums, and authenticity verification of digital tickets. Attached Figure Description

[0112] Figure 1 This is a block diagram of a tactile reproduction device for anti-counterfeiting of digital currency;

[0113] Figure 2 This is a flowchart of the present invention. Detailed Implementation

[0114] This invention achieves tactile rendering from the perspective of tactile information that affects the touch of bare fingers, such as temperature, shape, and texture, and completes a multimodal tactile anti-counterfeiting method based on tactile reproduction technology.

[0115] To enable those skilled in the art to better understand the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and a set of hardware facilities (which are only for better illustrating the functions implemented by this patent and are not unique).

[0116] Includes the following steps:

[0117] Step 1: Extraction of tactile features

[0118] (1) Shape tactile feature extraction method based on image shape height model:

[0119] First, a weighted average method is used to convert the color image into a luminance image:

[0120]

[0121] in This is the converted image brightness value matrix. This is the red value matrix of the original image. This is the green value matrix of the original image. This is the blue value matrix of the original image;

[0122] Based on the light source incidence model, obtain the reflection function of the image:

[0123]

[0124] In the formula, and It is a pixel. The gradients that vary along the X and Y directions at the given locations, This is the reflection function, which contains information about the light source on the object's surface. and The angles between the direction of the light source and the x-axis and z-axis are called the light source tilt angle and deflection angle, respectively, and the light source vector gradient. , , The angle of inclination, Angle of elevation;

[0125] Using the finite difference method of the latter term, the independent variable of the reflection function is... and Discretization:

[0126]

[0127]

[0128] in The height matrix of the object shape;

[0129] Rearranging the above equation, we get: in Indicates that the independent variable is , , , The quaternion function will about Performing a Taylor expansion, we get:

[0130] Where n is the number of iterations.

[0131] use Alternative ,

[0132] and Simplifying the above equation, we get:

[0133]

[0134] in, It can be represented as:

[0135]

[0136] Let the initial value Substituting the above formula into the formula above that, and iterating n times, the corresponding pixel is calculated. height value Then, a normalization method is used to obtain the image shape and height model. Record the shape rendering model information and record the extracted shape tactile features;

[0137] (2) Texture tactile feature extraction method based on image texture height model:

[0138] First, a weighted average method is used to convert the color image into a luminance image:

[0139]

[0140] Then adopt window Extract grayscale image Local image texture characteristics, As a periodic sequence, As a point A univariate function centered at a period of 8. for The corresponding local Fourier transform coefficient matrix, These are frequency domain coefficients;

[0141] get

[0142]

[0143] for Given an image of a certain size, obtain the set of local Fourier coefficients of the image;

[0144]

[0145] Among them, when When taking a specific value, This can represent the image texture height model of the corresponding frequency domain coefficients, denoted here as... This information is recorded as the texture rendering model information, and the extracted texture tactile features are also recorded.

[0146] (3) Temperature-tactile feature extraction method based on image tone information:

[0147] The specific steps to convert a color image based on the RGB color space into an image using the HSV model are as follows:

[0148] set up , , For point The RGB color values ​​at each point are normalized to be between 0 and 1, resulting in the points. Normalized RGB color values This is recorded as temperature rendering model information, and the extracted temperature tactile features are also recorded.

[0149] The recorded information—shape rendering model information, texture rendering model information, and temperature rendering model information—is stored in a secure storage chip.

[0150] Step 2: Recipeting Tactile Features

[0151] BP Neural Network Model: The process of a BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer. The second stage is the backward propagation of the error, from the output layer to the hidden layer, and finally to the input layer, adjusting the weights and biases from the hidden layer to the output layer and from the input layer to the hidden layer in sequence.

[0152] The specific process of obtaining the "recipe" using a BP neural network is as follows:

[0153] (1) Data Collection: Tactile data of the temperature, texture, and shape of RMB banknotes under normal temperature conditions are collected using the method in step 1. These individual tactile sensations can then be reproduced using a corresponding tactile reproduction device (the implementation method will be given in subsequent steps). The intensity of the three tactile sensations is randomly assigned (from 0 to 1). The intensity of the tactile reproduction device is set to [value missing]. It collects tactile perceptions of different individuals at random intensities, recording whether the tactile sensation at these intensities corresponds to the feel of RMB banknotes. Simultaneously, it sets a machine learning objective: to subjectively judge whether the tactile sensation is real or inauthentic, and then integrates the data with... and The random intensity data of real touch will be set to 1, and otherwise set to 0, to form a data training library and be used to train the model;

[0154] (2) Selection of excitation function:

[0155] (3) Output of hidden layer: Data obtained from the database and Substitute middle;

[0156] Where f is the activation function in (2), N is the number of nodes in the input layer, N takes the value of 3, and i takes the values ​​of 1, 2, and 3, which correspond to temperature, texture, and shape respectively; while j takes the values ​​of 1, 2, and 3, which represent the input layer, hidden layer, and output layer respectively. The bias from the input layer to the hidden layer; The weights from the input layer to the hidden layer; This is the output of the hidden layer;

[0157] (4) Output of the output layer:

[0158] Where l is the number of nodes in the hidden layer, and l takes the value of 3; k takes the values ​​of 1, 2, and 3; The bias from the hidden layer to the output layer; The bias from the hidden layer to the output layer; For the output of the output layer;

[0159] (5) Initialize the weights and biases, and set the learning rate. Start calculating the error;

[0160] error:

[0161] The expected output is The error is E; the output of the output layer is

[0162] (6) Continuously update the weights and biases based on the error:

[0163] Weight update:

[0164] Bias update:

[0165] in For learning rate;

[0166] (7) Continue until the optimal solution is obtained, and record the obtained weights. ( , , The "recipe" consists of weights for temperature, texture, and shape, respectively. This is a "virtual tactile sensation" obtained by a BP neural network that is closest to reality. In the context of digital RMB, the information in the "recipe" corresponds to the tactile sensation of real RMB.

[0167] Obtained weight records ( , , The recipe information is recorded in a secure storage chip.

[0168] Step 3: User's decryption

[0169] The user enters a private key into the human-computer interaction interface, which is then decrypted using the corresponding public key stored in the secure storage chip. If decryption fails, the human-computer interaction interface will display "Error, please re-enter"; if the input is correct, the rendered model information and recipe information will be extracted from the secure storage chip.

[0170] The secure storage chip stores: 1. rendering model information; 2. recipe information; 3. a public key corresponding to the private key input by the user; the user inputs the private key on the haptic interface for security verification;

[0171] Step 4: Tactile Reproduction

[0172] The rendering model information and recipe information obtained in step 3 are sent to the tactile feature rendering unit. The weights in the recipe information are assigned to the rendering model information to obtain weighted rendering model information. The weighted rendering model information is then sent to the signal driving unit to obtain a tactile driving signal that can reproduce tactile sensation in the tactile interaction interface.

[0173] (1) Air pressure film tactile reproduction method based on image shape height model:

[0174] Based on the shape rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. The gradient matrix of the image shape can be obtained. :

[0175]

[0176] in, and Represent matrices respectively exist and Partial derivatives in direction, gradient matrix This indicates the rate of edge transformation at each point, and follows the same trend as the magnitude of the tangential force. The tangential force matrix can be used to represent this rate of change. Represented as:

[0177]

[0178] in, This refers to the tangential force felt when a finger swipes across the screen without haptic feedback. The maximum tangential force that electrostatic tactile sensation can increase. This represents the maximum tangential force that the air pressure membrane can reduce tactile energy.

[0179] in, This represents the normal force applied by the finger, and the tangential force matrix is ​​represented by the air pressure film matrix. :

[0180] The relationship between the driving voltage and the electrostatic feedback force and the air pressure membrane feedback force was determined based on human-computer interaction experiments:

[0181]

[0182] in, The voltage amplitude of the air diaphragm drive signal. For air pressure diaphragm feedback force, This is the proportionality coefficient between the air pressure diaphragm drive signal and the feedback force;

[0183] From the above four equations, the amplitudes of the electrostatic force drive signal and the air pressure diaphragm drive signal can be obtained:

[0184]

[0185] in for The amplitude of the air pressure film driving signal applied at the point is denoted as the shape tactile driving signal;

[0186] After obtaining the shape-based tactile driving signal, it is sent to the tactile interaction interface, and the shape-based tactile sensation can be reproduced using tactile reproduction technology. This function requires a tactile reproduction device. Taking the provided hardware as an example, it includes a digital-to-analog converter and a power amplifier. Based on the requirements of mechanical vibration tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. The power amplifier then amplifies the amplitude of this signal, thus achieving shape-based tactile reproduction.

[0187] (2) Electrostatic tactile reproduction method based on image texture height model:

[0188] Based on the texture rendering model information obtained in step 1 The "recipe" yields weighted texture rendering model information. The image texture gradient matrix can be calculated. :

[0189]

[0190] in, and Represent matrices respectively exist and Partial derivative in direction;

[0191] Normalizing the texture gradient matrix, we get:

[0192]

[0193] in, For the normalized texture gradient matrix, and These represent the minimum and maximum values ​​in the texture height matrix, and the texture gradient matrix, respectively. This represents the local concavity and convexity of an image; that is, the smaller the gradient value, the smoother the texture surface, and the weaker the tactile sensation when a finger glides across it. The normalized texture gradient matrix is ​​mapped according to the specified ratio. Matching electrostatic force magnitude:

[0194]

[0195] The mapping relationship between electrostatic driving voltage and electrostatic force:

[0196]

[0197] Obtain the amplitude matrix of the electrostatic force driving signal, denoted as the texture tactile driving signal;

[0198] After obtaining the texture tactile driving signal, it is sent to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. Then, the power amplifier can amplify the amplitude value of the signal to realize the texture tactile sensation reproduction.

[0199] (3) Temperature-based haptic rendering method based on image tone information:

[0200] Based on the temperature rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. For point ,set up:

[0201]

[0202]

[0203] in, and Normalized color values , , The maximum and minimum values;

[0204] point Image tone at the location It can be determined by the following method:

[0205]

[0206] point Image saturation at the location It can be determined by the following method:

[0207]

[0208] point Image brightness at the location It can be determined by the following method:

[0209]

[0210] From this, we obtain Value at Within the interval, The value range is divided into four intervals: red ,yellow ,green ,blue Then, based on the four predefined zones, the color tones falling into different zones are... The system is configured with four pure colors: red, yellow, green, and blue. The corresponding RGB values ​​are returned and recorded as temperature tactile driving signals.

[0211] After receiving the temperature-sensitive tactile drive signal, it is sent to the tactile interaction interface, where tactile reproduction technology can be used to reproduce the texture tactile sensation. This function requires a tactile reproduction device, including a digital-to-analog converter (DAC), a power amplifier, and a temperature measurement circuit. The DAC receives the stimulation pattern sent by the microcontroller and generates a corresponding heating or cooling drive signal. This drive signal is then amplified by the power amplifier to obtain the output temperature-sensitive tactile drive signal. The temperature measurement circuit monitors the temperature of the semiconductor cooling chip surface and transmits the real-time temperature back to the microcontroller. The microcontroller calculates the required stimulation pattern based on the reference temperature, temperature difference, and the duration of the current temperature, and then sends the signal back to the DAC to achieve temperature-sensitive tactile reproduction.

[0212] The rendering model information obtained from the secure storage chip is sent to the signal driving unit after being weighted by the recipe to obtain the tactile driving signal, and then sent to the tactile interactive interface (the tactile reproduction device is integrated on the tactile interactive interface) to realize the tactile reproduction function, thereby realizing multi-dimensional tactile anti-counterfeiting.

[0213] A digital anti-counterfeiting device for banknotes based on multi-element fusion tactile reproduction includes a tactile interactive interface, a secure storage chip, a signal rendering unit, and a signal driving unit, wherein:

[0214] The haptic interface is used by users to enter their private keys for security verification.

[0215] The secure storage chip stores: rendering model information, recipe information, a public key corresponding to the user-input private key, and the resulting weight record. ( , , These are shape rendering model information, texture rendering model information, and temperature rendering model information, which are recorded in a secure storage chip as recipe information. The rendering model information obtained from the secure storage chip is sent to the signal driving unit after being weighted by the recipe.

[0216] After receiving the tactile driving signal, the signal driving unit sends it to the tactile interactive interface;

[0217] The tactile interactive interface integrates a tactile reproduction device to achieve tactile reproduction, thereby realizing multi-faceted tactile anti-counterfeiting, specifically including:

[0218] After receiving the shape tactile drive signal, the tactile interaction interface sends it to the tactile interaction interface. The shape tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the mechanical vibration tactile requirements, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the drive signal. Then, the power amplifier can amplify the amplitude value of the signal to realize the shape tactile sensation reproduction.

[0219] After receiving the texture tactile driving signal, the tactile interaction interface sends it to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. The power amplifier can then amplify the amplitude of the signal to achieve texture tactile reproduction.

[0220] After receiving the temperature-sensitive tactile drive signal, the tactile interaction interface sends it to the interface. The texture tactile sensation can then be reproduced using tactile reproduction technology. This function requires a tactile reproduction device, including a digital-to-analog converter (DAC), a power amplifier, and a temperature measurement circuit. The DAC receives the stimulation pattern sent by the microcontroller and generates a corresponding heating or cooling drive signal. The drive signal is then amplified by the power amplifier to obtain the output temperature-sensitive tactile drive signal. The temperature measurement circuit monitors the temperature of the semiconductor cooling chip surface and transmits the real-time temperature back to the microcontroller. The microcontroller calculates the required stimulation pattern based on the reference temperature, temperature difference, and current temperature duration, and then sends the signal back to the DAC to achieve temperature-sensitive tactile reproduction.

Claims

1. A digital anti-counterfeiting method for banknotes based on multi-source fusion tactile reproduction, characterized in that, Includes the following steps: Step 1: Extraction of tactile features, including shape tactile feature extraction, texture tactile feature extraction, and temperature tactile feature extraction; Step 2: Use a BP neural network model to "recipe" the tactile features. The process of the BP neural network is mainly divided into two stages. The first stage is the forward propagation of the signal, from the input layer through the hidden layer, and finally to the output layer. The second stage is the backpropagation of the error, from the output layer to the hidden layer, and finally to the input layer, adjusting the weights and biases from the hidden layer to the output layer, and from the input layer to the hidden layer in sequence. The specific process of obtaining the "recipe" using a BP neural network is as follows: (1) Data Collection: Tactile data of the banknote's temperature, texture, and shape under normal temperature conditions are collected using the method in step 1. These individual tactile sensations can then be reproduced using the corresponding tactile reproduction device. The intensity of each of the three tactile sensations is randomly assigned (from 0 to 1). The intensity of the tactile reproduction device is set to [value missing]. It collects tactile perceptions of different individuals at random intensities, recording whether the tactile sensation at these intensities corresponds to the feel of banknotes. Simultaneously, it sets a machine learning objective: to subjectively judge whether the tactile sensation is real or inauthentic, and then integrates the data with... and The random intensity data of real touch will be set to 1, and otherwise set to 0, to form a data training library and be used to train the model; (2) Selection of excitation function: ; (3) Output of hidden layer: Data obtained from the database and Substitute middle; Where f is the activation function in (2), N is the number of nodes in the input layer, N takes the value of 3, and i takes the values ​​of 1, 2, and 3, which correspond to temperature, texture, and shape respectively; while j takes the values ​​of 1, 2, and 3, which represent the input layer, hidden layer, and output layer respectively. The bias from the input layer to the hidden layer; The weights from the input layer to the hidden layer; This is the output of the hidden layer; (4) Output of the output layer: ; Where l is the number of nodes in the hidden layer, and l takes the value of 3; k takes the values ​​of 1, 2, and 3; The bias from the hidden layer to the output layer; The bias from the hidden layer to the output layer; For the output of the output layer; (5) Initialize the weights and biases, and set the learning rate. Start calculating the error; error: ; The expected output is The error is E; the output of the output layer is ; (6) Continuously update the weights and biases based on the error: Weight update: ; Bias update: ; in For learning rate; (7) Continue until the optimal solution is obtained, and record the obtained weights. ( , , The formula is defined as follows: the weights of temperature, texture, and shape are listed in order. The obtained weight records ( , , The recipe information is recorded in a secure storage chip. Step 3: User's decryption; Step 4: Tactile Reproduction. The obtained rendering model information and recipe information are sent to the tactile feature rendering unit. The weights in the recipe information are assigned to the rendering model information to obtain weighted rendering model information. The weighted rendering model information is then sent to the signal driving unit to obtain the tactile driving signal that realizes tactile reproduction in the tactile interaction interface, thereby realizing the tactile reproduction function and achieving multi-dimensional tactile anti-counterfeiting.

2. The digital anti-counterfeiting method for banknotes based on multi-element fusion tactile reproduction according to claim 1, characterized in that, The method for extracting shape tactile features based on the image shape height model in step 1 is as follows: First, a weighted average method is used to convert the color image into a luminance image: ； in This is the converted image brightness value matrix. This is the red value matrix of the original image. This is the green value matrix of the original image. This is the blue value matrix of the original image; Based on the light source incidence model, obtain the reflection function of the image: ； In the formula, and It is a pixel. The gradients that vary along the X and Y directions at the given locations, This is the reflection function, which contains information about the light source on the object's surface. and The angles between the direction of the light source and the x-axis and z-axis are called the light source tilt angle and deflection angle, respectively, and the light source vector gradient. , , It is the angle of inclination. Angle of elevation; Using the finite difference method of the latter term, the independent variable of the reflection function is... and Discretization: ； ； in The height matrix of the object shape; Rearranging the above equation, we get: ; in Indicates that the independent variable is , , , The quaternion function will about Performing a Taylor expansion, we get: ； Where n is the number of iterations, used Alternative ,and Simplifying the above equation, we get: ； in, It can be represented as: ； Let the initial value Substituting the above formula into the formula above that, and iterating n times, the corresponding pixel is calculated. height value Then, a normalization method is used to obtain the image shape and height model. Record the shape rendering model information and record the extracted shape tactile features.

3. The digital anti-counterfeiting method for banknotes based on multi-source fusion tactile reproduction as described in claim 1, characterized in that, The texture tactile feature extraction method based on the image texture height model in step 1: First, a weighted average method is used to convert the color image into a luminance image: ； Then adopt window Extract grayscale image Local image texture characteristics, As a periodic sequence, As a point A univariate function centered at a period of 8. for The corresponding local Fourier transform coefficient matrix, These are frequency domain coefficients; get ; ； for Given an image of a certain size, obtain the set of local Fourier coefficients of the image; ； Among them, when When taking a specific value, The image texture height model representing the corresponding frequency domain coefficients is denoted here as... This information is recorded as the texture rendering model information, and the extracted texture tactile features are also recorded.

4. The digital anti-counterfeiting method for banknotes based on multi-element fusion tactile reproduction according to claim 1, characterized in that, The temperature-tactile feature extraction method based on image tone information in step 1: The specific steps to convert a color image based on the RGB color space into an image using the HSV model are as follows: set up , , For point The RGB color values ​​at each point are normalized to be between 0 and 1, resulting in the points. Normalized RGB color values The temperature rendering model information is recorded, and the extracted temperature tactile features are also recorded. The recorded information, namely shape rendering model information, texture rendering model information, and temperature rendering model information, is stored in a secure storage chip.

5. The digital anti-counterfeiting method for banknotes based on multi-element fusion tactile reproduction according to claim 1, characterized in that, In step 3, the user enters a private key on the human-computer interaction interface, and decrypts it according to the corresponding public key stored in the secure storage chip. If decryption fails, the human-computer interaction interface will prompt "Error, re-enter"; if the input is correct, the rendering model information and recipe information will be extracted from the secure storage chip. The secure storage chip stores: rendering model information, party information, and the public key corresponding to the private key input by the user; Users enter their private key on the haptic interface for security verification.

6. The digital anti-counterfeiting method for banknotes based on multi-source fusion tactile reproduction as described in claim 1, characterized in that, The air pressure membrane tactile reproduction method based on the image shape height model in step 4: Based on the shape rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. Find the gradient matrix of the image shape. : ； in, and Represent matrices respectively exist and Partial derivatives in direction, gradient matrix This indicates the rate of edge transformation at each point, and follows the same trend as the magnitude of the tangential force. The tangential force matrix can be used to represent this rate of change. Represented as: ； in, This refers to the tangential force felt when a finger swipes across the screen without haptic feedback. The maximum tangential force that electrostatic tactile sensation can increase. This represents the maximum tangential force that the air pressure membrane can reduce tactile energy. in, This represents the normal force applied by the finger, and the tangential force matrix is ​​represented by the air pressure film matrix. : ; The relationship between the driving voltage and the electrostatic feedback force and the air pressure membrane feedback force was determined based on human-computer interaction experiments: ; in, The voltage amplitude of the air diaphragm drive signal. For air pressure diaphragm feedback force, This is the proportionality coefficient between the air pressure diaphragm drive signal and the feedback force; The amplitudes of the electrostatic force drive signal and the air pressure diaphragm drive signal can be obtained from the above formulas: ； in for The amplitude of the air pressure film driving signal applied at the point is denoted as the shape tactile driving signal; After obtaining the shape tactile drive signal, it is sent to the tactile interaction interface. The shape tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device. According to the mechanical vibration tactile requirements, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the drive signal. Then, the amplitude value of the signal can be amplified by the power amplifier, thus realizing the shape tactile sensation reproduction.

7. The digital anti-counterfeiting method for banknotes based on multi-source fusion tactile reproduction as described in claim 1, characterized in that, The electrostatic tactile reproduction method based on the image texture height model in step 4: Based on the texture rendering model information obtained in step 1 After "recipe", weighted texture rendering model information is obtained. Find the image texture gradient matrix. : ； in, and Represent matrices respectively exist and Partial derivative in direction; Normalizing the texture gradient matrix, we get: ； in, For the normalized texture gradient matrix, and These represent the minimum and maximum values ​​in the texture height matrix, and the texture gradient matrix, respectively. This represents the local concavity and convexity of an image; that is, the smaller the gradient value, the smoother the texture surface, and the weaker the tactile sensation when a finger glides across it. The normalized texture gradient matrix is ​​mapped according to the specified ratio. Matching electrostatic force magnitude: ； The mapping relationship between electrostatic driving voltage and electrostatic force: ； Obtain the amplitude matrix of the electrostatic force driving signal, denoted as the texture tactile driving signal; After obtaining the texture tactile driving signal, it is sent to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. Then, the amplitude value of the signal can be amplified by the power amplifier, thus realizing the reproduction of texture tactile sensation.

8. The digital anti-counterfeiting method for banknotes based on multi-element fusion tactile reproduction according to claim 1, characterized in that, The temperature-sensitive tactile rendering method based on image tone information in step 4: Based on the temperature rendering model information obtained in step 1 After "recipe", weighted shape rendering model information is obtained. For point ,set up: ； ； in, and Normalized color values , , The maximum and minimum values; point Image tone at the location It can be determined by the following method: ； point Image saturation at the location It can be determined by the following method: ； point Image brightness at the location It can be determined by the following method: ； From this, we obtain Value at Within the interval, The value range is divided into four intervals: red ,yellow ,green ,blue Then, based on the four predefined zones, the color tones falling into different zones are... The system is designed with four pure colors: red, yellow, green, and blue. The corresponding RGB values ​​are returned and these RGB values ​​are recorded as temperature tactile driving signals. After receiving the temperature-sensitive tactile driving signal, it is sent to the tactile interaction interface, and the texture tactile sensation can be reproduced using tactile reproduction technology.

9. An apparatus for implementing the digital anti-counterfeiting method for banknotes based on multi-element fusion tactile reproduction as described in claim 1, characterized in that, It includes a tactile interaction interface, a secure storage chip, a signal rendering unit, and a signal driving unit, wherein: The haptic interface is used by users to enter their private keys for security verification. The secure storage chip stores: rendering model information, recipe information, a public key corresponding to the user-input private key, and the resulting weight record. ( , , These are shape rendering model information, texture rendering model information, and temperature rendering model information, which are recorded in a secure storage chip as recipe information. The rendering model information obtained from the secure storage chip is sent to the signal driving unit after being weighted by the recipe. After receiving the tactile driving signal, the signal driving unit sends it to the tactile interactive interface; The tactile interactive interface integrates a tactile reproduction device to achieve tactile reproduction, thereby realizing multi-faceted tactile anti-counterfeiting, specifically including: After receiving the shape tactile drive signal, the tactile interaction interface sends it to the tactile interaction interface. The shape tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the mechanical vibration tactile requirements, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the drive signal. The power amplifier can then amplify the amplitude of the signal to achieve shape tactile reproduction. After receiving the texture tactile driving signal, the tactile interaction interface sends it to the tactile interaction interface. The texture tactile sensation can be reproduced using tactile reproduction technology. The function requires the use of a tactile reproduction device, including a digital-to-analog converter and a power amplifier. According to the requirements of electrostatic tactile sensation, the microcontroller sends a control signal to the digital-to-analog converter to adjust the frequency and waveform of the driving signal. The power amplifier can then amplify the amplitude of the signal to achieve texture tactile reproduction. After receiving the temperature-sensitive tactile drive signal, the tactile interaction interface sends it to the interface. The texture tactile sensation can then be reproduced using tactile reproduction technology. This function requires a tactile reproduction device, including a digital-to-analog converter (DAC), a power amplifier, and a temperature measurement circuit. The DAC receives the stimulation pattern sent by the microcontroller and generates a corresponding heating or cooling drive signal. The drive signal is then amplified by the power amplifier to obtain the output temperature-sensitive tactile drive signal. The temperature measurement circuit monitors the temperature of the semiconductor cooling chip surface and transmits the real-time temperature back to the microcontroller. The microcontroller calculates the required stimulation pattern based on the reference temperature, temperature difference, and current temperature duration, and then sends the signal back to the DAC to achieve temperature-sensitive tactile reproduction.

Images