Multi-modal biological characteristic bill anti-counterfeiting identification method

Abstract

The invention discloses a multi-modal biological characteristic bill anti-counterfeiting identification method, and relates to the technical field of bill anti-counterfeiting identification, and the method comprises the steps: collecting multi-modal biological characteristics and environmental parameters, calculating the dynamic weight of each biological characteristic through a plurality of dimensions, carrying out the dynamic fusion of the biological characteristics through a weighted fusion algorithm, and outputting an optimal characteristic combination. Performing three-layer cascade anti-counterfeiting coding on the fused features, including basic feature coding, associated feature coding and time sequence feature coding, and performing secure storage through distributed storage and block chain technology; a real-time monitoring mechanism is established, multiple features are extracted to construct a behavior model, the abnormal degree of a current behavior is evaluated, and when the abnormal degree exceeds a threshold value, a grading alarm mechanism and a processing strategy are automatically triggered. According to the invention, the recognition result is dynamically corrected in combination with the environmental parameters, and the accuracy of bill authenticity recognition is remarkably improved. Multi-level alarm is carried out through quantitative evaluation of the abnormal degree, and differentiated processing measures are taken for bills of different risk levels.

Description

Technical Field

[0001] The invention relates to the technical field of bill anti-counterfeiting identification, in particular to a multi-modal biometric bill anti-counterfeiting identification method. Background Art

[0002] With the continuous development of bill electronicization and anti-counterfeiting technology, bills play an increasingly important role in financial transactions and commercial activities. However, traditional anti-counterfeiting identification systems often use fixed discrimination criteria, lack comprehensive analysis of bill features and dynamic consideration of environmental factors, resulting in limited recognition accuracy. This fixed standard detection method cannot adapt to the recognition needs in different scenarios, and it is difficult to cope with the natural loss and feature attenuation that may occur in the circulation of bills. In practical applications, the anti-counterfeiting features of bills are affected by many factors, such as service life, storage conditions, and human damage. These factors will cause the anti-counterfeiting features to change to varying degrees, and the fixed standard detection method is difficult to effectively adapt and adjust these changes. In particular, under different environmental conditions, changes in factors such as light intensity, temperature, and humidity will significantly affect the acquisition quality of bill features. For example, in an environment with insufficient light, the recognition effect of fluorescent anti-counterfeiting features will be greatly reduced; in a high temperature and high humidity environment, the physical features of the bill may be slightly deformed; in an environment with large vibration, the clarity of image acquisition will be affected. However, the existing system fails to fully consider the impact of these environmental factors, lacks a dynamic adjustment mechanism for environmental adaptability, and easily causes instability in recognition results.

[0003] In addition, the existing system has a relatively simple processing mechanism after discovering suspicious bills, which usually only gives a binary judgment result of "true" or "false", and lacks a quantitative assessment of the degree of suspicion. This simple processing mechanism cannot meet the needs of differentiated treatment of different risk levels in actual business. At the same time, the system also lacks the ability to fully record and trace the anti-counterfeiting identification process, making it difficult to provide effective data support for subsequent fake bill traceability analysis and anti-counterfeiting technology optimization. With the increasingly frequent circulation of bills today, organized bill counterfeiting activities have shown the characteristics of ganging and specialization. In this case, traditional static anti-counterfeiting identification methods have been unable to cope with the increasingly severe bill anti-counterfeiting challenges. Therefore, there is an urgent need for an intelligent anti-counterfeiting identification solution that can comprehensively consider multi-dimensional features, environmental factors and historical data to improve the accuracy and reliability of bill anti-counterfeiting identification. Summary of the invention

[0004] In view of the problems existing in the existing multi-modal biometric bill anti-counterfeiting identification method, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to improve the accuracy and reliability of bill authenticity identification in combination with multi-dimensional features under different environmental conditions.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a multimodal biometric bill anti-counterfeiting identification method, which includes collecting multimodal biometrics and environmental parameters, calculating the dynamic weight of each biometric through three dimensions of feature quality score, environmental adaptability score and feature stability score, using a weighted fusion algorithm to dynamically fuse multiple biometrics, and outputting the optimal feature combination; implementing three-layer cascade anti-counterfeiting coding for the fused features, the first layer performs basic feature coding, and hashes the fused features with timestamps and random numbers; the second layer performs associated feature coding, and encrypts the basic code with block information and identity identification; the third layer performs time series feature coding, digitally signs the associated code with time series information and metadata, and securely stores them through distributed storage and blockchain technology; a real-time monitoring mechanism is established based on the three-layer cascade anti-counterfeiting coding, time features, location features, frequency features and operation features are extracted to construct a behavior model, and a multi-dimensional distance calculation method is used to evaluate the abnormality of the current behavior. When the abnormality exceeds the system preset threshold, a hierarchical alarm mechanism and an emergency response strategy are automatically triggered to achieve safe monitoring of the entire process of bill use.

[0007] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method described in the present invention, the biometric features include fingerprint features, palm print features, iris features and voice print features; the biometric features are extracted to obtain biometric image data, including grayscale information, resolution information and acquisition time information of the features; feature point extraction and feature vector conversion are performed on the biometric image data to obtain a feature vector.

[0008] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method of the present invention, the feature point extraction adopts a method combining Gabor filter enhancement preprocessing and Harris corner point detection, and Gabor filter is used to enhance the biometric image data. The expression of Gabor filter is: ; ; in, is the filter direction; is the filter center frequency, is the spatial aspect ratio; is the standard deviation of the Gaussian envelope; Apply the Harris corner detection algorithm to the image enhanced by Gabor filtering, including: Calculate the image and Directional gradient and ; Calculate the second-order matrix M of the gradient: ; in, is a Gaussian window function centered at the point to be detected; calculate the corner point response value : ; in, is the determinant of the matrix M, represents the trace of the matrix M, is an empirical constant; Corner point response value Perform non-maximum suppression and extract local maximum points as feature points; The feature points are extracted and expressed as feature vectors, which are expressed as: ; in, Indicates A set of feature vectors of biometric features, is the number of feature points, For the feature points coordinate, For the feature points coordinate, For the The gray value of feature points, For the The direction angle of the feature points, For the The regional attribute value of feature points.

[0009] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method of the present invention, wherein: based on the Euclidean distance of the feature point Similarity with attributes Calculate feature point matching , the calculation formula is: ; ; ; in, is the Euclidean distance between feature points, is the attribute difference value of the feature point, and is the standard deviation parameter of distance and attribute; According to the matching degree of feature points Calculating Biometric Quality Score , expressed as: ; in, is the feature point weight coefficient, is the total number of feature points; the environmental adaptability score is calculated based on standardized environmental parameters , expressed as: ; in, , , , are the influencing factors of different biological characteristics, Represents the light intensity parameter, represents the temperature parameter, Represents humidity parameters, Represents noise parameters; combined with quality score and environmental adaptability score Generate fusion weight coefficients , expressed as: ; in, and is the adjustment coefficient; according to the fusion weight coefficient For the feature vector Perform weighted fusion to obtain fused feature data , expressed as: ; in, Represents fused feature data.

[0010] As a preferred solution of the multi-modal biometric bill anti-counterfeiting identification method of the present invention, the three-layer cascade anti-counterfeiting coding includes: With timestamp , random numbers The basic feature code is calculated by the SHA-256 hash algorithm ; Extract block information , including block height, block hash value, and the identity , encode the basic features With block information , Identity Perform RSA encryption to generate associated feature codes , expressed as: ; in, Represents the RSA public key encryption function, Contains block height and block hash ; A unique identifier for the user; Get the code generation time and the validity period of the bill , combined to form timing information ; Get geographic location information , including longitude, latitude and altitude information; obtain device identification information , including device serial number and hardware feature code; combining geographic location information and device identification information to form metadata ; The timing information , metadata and associated feature encoding Generate a digital signature using the ECDSA algorithm , the calculation formula is: ; ; ; in, Represents the ECDSA private key signature function, Contains the timestamp of creation and validity period ; Contains location coordinates and equipment ; and is a signature value pair; is the hash value of the message to be signed; the digital signature Write to the preset distributed storage node to generate the storage address , where the address is stored Generated through IPFS content addressing, expressed as: ; in, Indicates the addition operation of the IPFS network. Encode addresses in Base58; create forgery-proof transactions on blockchain networks , the storage address , basic feature encoding , associated feature coding Packaging and writing anti-counterfeiting transactions ; Anti-counterfeiting transactions through broadcasting Go to the blockchain network and wait for the block to be confirmed before getting the transaction hash value ; Based on transaction hash value , storage address Generate QR code , the QR code Write the ticket, where the QR code format is: ; in, Represents a QR code generation function.

[0011] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method of the present invention, wherein: collecting bill usage behavior characteristics, including timestamp , geographic coordinates , operation sequence And use interval , forming the feature vector ; Based on the feature vector , through the combination of multiple Gaussian components, construct a behavioral model , calculate the probability distribution of normal behavior, expressed as: ; in, is the weight coefficient, is the Gaussian component, represents the number of Gaussian distributions; the probability density calculation formula of the Gaussian component is: ; in, yes dimensional feature vector, It is Quantity dimensional mean vector, yes × dimensional covariance matrix, is the determinant of the covariance matrix, is the vector transpose, is the inverse of the covariance matrix.

[0012] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method of the present invention, for each Gaussian component, K-means clustering is used to determine the initial weight coefficient , given N training samples, train the behavior model, determine the log-likelihood function of the behavior model, and optimize the behavior model parameters. The log-likelihood function is expressed as: ; in, is the number of training samples, is the number of model parameters, is the likelihood function value; the Bayesian Information Criterion BIC is used to select the optimal number of components , expressed as: ; in, is the optimal number of components; by maximizing the log-likelihood function, the EM algorithm is used to iteratively optimize the behavior model parameters to obtain a behavior model that describes the normal behavior pattern. , calculate the occurrence probability of any feature vector.

[0013] As a preferred solution of the multimodal biometric bill anti-counterfeiting identification method of the present invention, wherein: based on the feature vector , calculate multi-dimensional distance , including time distance , spatial distance , sequence distance and frequency distance , expressed as: ; ; ; ; ; in, is the feature weight coefficient, , , , , , is the reference value, is a sequence edit distance function; based on the multidimensional distance and behavioral models The probability distribution is combined to calculate the degree of abnormality , expressed as: ; ; in, is the adaptive weight factor, is the adjustment coefficient, is the distance threshold, is the probability of the behavior model; according to the degree of abnormality Set the alarm threshold , , }, when the abnormal degree When the corresponding threshold is exceeded, the corresponding level of alarm signal is triggered , specifically: If 0≤ < , it means that the system is in normal monitoring state, does not trigger the alarm mechanism, continues to accumulate behavior data, and updates behavior model parameters; like ≤ < , then set the alarm level = 1, trigger the response mechanism, increase the feature sampling frequency, record abnormal behavior characteristics, and prepare possible upgrade plans; like ≤ < , then set the alarm level = 2, deploy enhanced protection strategies, call behavioral models for in-depth analysis, establish behavioral trajectory tracking, and prepare emergency treatment plans; like ≥ , then set the alarm level = 3, execute emergency response, implement the most advanced protection strategy, start comprehensive monitoring mechanism, call historical data for analysis, and activate emergency response plan.

[0014] In the second aspect, an embodiment of the present invention provides a multimodal biometric bill anti-counterfeiting identification system, which includes: an acquisition module, which is used to collect multimodal biometrics and environmental parameters, calculate the dynamic weight of each biometric through three dimensions of feature quality score, environmental adaptability score and feature stability score, and dynamically fuse multiple biometrics using a weighted fusion algorithm to output the optimal feature combination; an encryption module, which is used to implement three-layer cascade anti-counterfeiting coding for the fused features, the first layer performs basic feature coding, and hashes the fused features with timestamps and random numbers; the second layer performs associated feature coding, and encrypts the basic code with block information and identity identification; the third layer performs time series feature coding, digitally signs the associated code with time series information and metadata, and securely stores them through distributed storage and blockchain technology; an early warning module, which establishes a real-time monitoring mechanism based on the anti-counterfeiting code, extracts time features, location features, frequency features and operation features to construct a behavior model, and uses a multi-dimensional distance calculation method to evaluate the abnormality of the current behavior. When the abnormality exceeds the system preset threshold, it automatically triggers a hierarchical alarm mechanism and an emergency response strategy to achieve safe monitoring of the entire process of bill use.

[0015] The beneficial effect of the present invention is that the present invention establishes a standardized feature collection environment and introduces an environmental adaptability assessment mechanism. On the basis of accurately extracting the multi-dimensional features of the bill, the recognition results are dynamically corrected in combination with environmental parameters, thereby significantly improving the accuracy of bill authenticity identification. It can not only accurately characterize the standard feature distribution of the bill, but also conduct multi-level alarms through quantitative assessment of the degree of abnormality, and take differentiated treatment measures for bills of different risk levels. At the same time, the key information of the entire identification process is recorded through blockchain technology, and a complete traceability system is established to provide reliable data support for the traceability of fake bills and the optimization of anti-counterfeiting technology. It effectively solves the problems of insufficient recognition accuracy, lack of environmental adaptability, and difficulty in tracing in complex environments of traditional methods, which is of great significance to improving the overall effectiveness of bill anti-counterfeiting identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. 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 creative work. Among them: Figure 1 The figure is a flow chart of a multimodal biometric bill anti-counterfeiting recognition method. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0020] Example 1: Reference Figure 1, which is the first embodiment of the present invention, provides a multi-modal biometric bill anti-counterfeiting identification method, comprising: S1: Collect multimodal biometric features and environmental parameters, calculate the dynamic weight of each biometric feature through three dimensions: feature quality score, environmental adaptability score and feature stability score, use weighted fusion algorithm to dynamically fuse multiple biometric features, and output the optimal feature combination.

[0021] Collect the user's fingerprint features, palm print features, iris features and voice print features. The feature collection obtains biometric image data through the corresponding biometric collector ( ),in Represents fingerprint features, Indicates palm print characteristics. Represents iris features, Represents voiceprint features, biometric image data Contains the feature's grayscale information, resolution information, and acquisition time information.

[0022] Biometric image data Perform feature point extraction and feature vector conversion to obtain feature vector , the eigenvector Contains the 2D coordinate values ​​of feature points And the gray value of the feature point , Direction value , Area attribute value .

[0023] Among them, the feature point extraction adopts the method of combining Gabor filter enhancement preprocessing and Harris corner point detection, and uses Gabor filter to extract the biometric image data. After enhancement preprocessing, the expression of Gabor filter is: ; ; in, is the filter direction; is the filter center frequency, is the spatial aspect ratio; is the standard deviation of the Gaussian envelope.

[0024] Apply the Harris corner detection algorithm to the image enhanced by Gabor filtering, including: Calculate the image and Directional gradient and ; Calculate the second-order matrix M of the gradient: ; in, is a Gaussian window function centered at the point to be detected.

[0025] Calculate corner response values : ; in, is the determinant of the matrix M, represents the trace of the matrix M, is an empirical constant.

[0026] Corner point response value Perform non-maximum suppression and extract local maximum points as feature points.

[0027] The feature points are extracted and expressed as feature vectors, which are expressed as: ; in, Indicates A set of feature vectors of biometric features, is the number of feature points, For the feature points coordinate, For the feature points coordinate, For the The gray value of feature points, For the The direction angle of the feature points, For the The regional attribute value of feature points.

[0028] Collect the light intensity parameters of the current environment through the environmental sensor , Temperature parameters , humidity parameters And noise parameters The environmental parameters are normalized by the maximum and minimum values ​​to obtain the standardized environmental parameters , , , .

[0029] The feature vector Match with the preset biometric template library respectively and calculate the matching degree of feature points , feature point matching degree Euclidean distance based on feature points Similarity with attributes The calculation formula is: ; ; ; in, is the Euclidean distance between feature points, is the attribute difference value of the feature point, and are the distance and standard deviation parameters of the attributes.

[0030] According to the matching degree of feature points Calculating Biometric Quality Score , expressed as: ; in, is the feature point weight coefficient, is the total number of feature points.

[0031] Feature point weight coefficient It is positively correlated with the importance of the area where the feature point is located. The weight range of the core area is [0.8, 1.0], the weight range of the transition area is [0.5, 0.8], and the weight range of the edge area is [0.2, 0.5].

[0032] Calculate environmental suitability scores based on standardized environmental parameters , expressed as: ; in, , , , are the influencing factors of different biological characteristics, and =1; Represents the light intensity parameter, represents the temperature parameter, Represents humidity parameters, represents the noise parameter, represents the noise parameter.

[0033] The impact factor is set according to the sensitivity of biological characteristics to environmental factors. The value range is [0.4-0.5], the voiceprint feature The value range is [0.4-0.5], and other factors are in the range of [0.1,0.3] according to the actual degree of influence.

[0034] Specifically, the iris feature is more sensitive to light intensity. The larger the value, the more sensitive the voiceprint feature is to environmental noise. The value is larger.

[0035] Combined quality score and environmental adaptability score Generate fusion weight coefficients , expressed as: ; in, and is the adjustment coefficient, which is adaptively adjusted through the environmental condition threshold. When the environmental condition is better than the threshold, The value range is [0.6-0.8], otherwise The value range is [0.6-0.8].

[0036] Among them, the setting of environmental condition thresholds takes into account that the light index reaches more than 75% of the standard value, the temperature deviation is within 75% of the standard range, the humidity deviation is within 75% of the standard range, and the noise level is within 75% of the allowable range.

[0037] According to the fusion weight coefficient For the feature vector Perform weighted fusion to obtain fused feature data , expressed as: ; By fusion weight coefficient Feature data obtained by weighted fusion , giving full play to the role of complementary feature advantages, retaining the advantageous information of each modal feature, weakening the limitations of a single feature, and significantly improving the robustness of feature expression.

[0038] S2: Implement three-layer cascade anti-counterfeiting coding for fusion features. The first layer performs basic feature coding, and hashes the fusion features with timestamps and random numbers. The second layer performs associated feature coding, and encrypts the basic code with block information and identity identification. The third layer performs time series feature coding, and digitally signs the associated code with time series information and metadata, and stores them securely through distributed storage and blockchain technology.

[0039] Based on the fusion features With timestamp , random numbers The basic feature code is calculated by the SHA-256 hash algorithm ; Extract block information , including block height, block hash value, and the identity , encode the basic features With block information , Identity Perform RSA encryption to generate associated feature codes , expressed as: ; in, Represents the RSA public key encryption function, using a 2048-bit key pair. Contains block height and block hash , used to anchor time points; It is the user's unique identifier, assigned by the identity authentication system; the encryption result The length is the same as the RSA key length.

[0040] Get the code generation time and the validity period of the bill , combined to form timing information ; Get geographic location information , including longitude, latitude and altitude information; obtain device identification information , including device serial number and hardware feature code; combining geographic location information and device identification information to form metadata ; The timing information , metadata and associated feature encoding Generate a digital signature using the ECDSA algorithm , the calculation formula is: ; ; ; in, Indicates the ECDSA private key signature function, using the secp256k1 curve. Contains the timestamp of creation and validity period , used to control the timeliness of bills; Contains location coordinates and equipment , provide environmental information; and It is a signature value pair, each of which is 256 bits; The hash value of the message to be signed.

[0041] Digital signature Write to the preset distributed storage node to generate the storage address , where the address is stored Generated through IPFS content addressing, expressed as: ; in, Indicates the addition operation of the IPFS network. It is a 46-byte Base58 encoded address. IPFS uses the DHT network for data distribution and storage.

[0042] Creating counterfeit-proof transactions on the blockchain network , the storage address , basic feature encoding , associated feature coding Packaging and writing anti-counterfeiting transactions , where the anti-counterfeiting transaction data structure is: ; Anti-counterfeiting transactions through broadcasting Go to the blockchain network and wait for the block to be confirmed before getting the transaction hash value , the number of block confirmations is set to 6 to ensure the finality of the transaction.

[0043] Based on transaction hash , storage address Generate QR code , the QR code Write the ticket, where the QR code format is: ; in, Represents the QR code generation function, using Error correction level, the QR code size is 256×256 pixels and contains version information for compatibility processing.

[0044] Through distributed storage and blockchain technology, data can be permanently stored. As a carrier, it can effectively transmit anti-counterfeiting information. Through the multi-level progressive coding scheme, it not only improves the reliability of anti-counterfeiting, but also facilitates subsequent verification and traceability.

[0045] S3: Establish a real-time monitoring mechanism based on anti-counterfeiting coding, extract time features, location features, frequency features and operation features to build a behavior model, and use a multi-dimensional distance calculation method to evaluate the abnormality of the current behavior. When the abnormality exceeds the preset threshold of the system, it automatically triggers a graded alarm mechanism and emergency response strategy to achieve safe monitoring of the entire process of ticket use.

[0046] Collect the usage behavior characteristics of the ticket and set the timestamp , geographic coordinates , operation sequence And use interval Composition feature vector , expressed as: ; in, The UTC timestamp when the ticket is used, accurate to milliseconds. is the three-dimensional geographic coordinate ( , ), indicating the location of use; is the operation sequence vector, recording the user interaction behavior, The time interval between two consecutive uses.

[0047] Based on the feature vector , through the combination of multiple Gaussian components, construct a behavioral model , calculate the probability distribution of normal behavior, expressed as: ; in, is the weight coefficient, is the Gaussian component, represents the number of Gaussian distributions; the probability density calculation formula of the Gaussian component is: ; in, yes dimensional feature vector, It is Quantity dimensional mean vector, yes × dimensional covariance matrix, is the determinant of the covariance matrix, is the vector transpose, is the inverse of the covariance matrix.

[0048] Calculate the eigenvector by the probability density of each Gaussian component In the The probability density under a Gaussian distribution reflects the degree to which the behavior matches the normal pattern of this type.

[0049] Furthermore, for each Gaussian distribution, K-means clustering is used to determine the initial weight coefficient , given N training samples, train the behavior model, determine the log-likelihood function of the behavior model, and optimize the behavior model parameters. The log-likelihood function is expressed as: ; in, is the number of training samples, is the number of model parameters, is the likelihood function value.

[0050] Use Bayesian Information Criterion BIC to select the optimal number of components , expressed as: ; in, is the optimal number of components.

[0051] By maximizing the log-likelihood function and using the EM algorithm to iteratively optimize the above parameters, a probability model describing the normal behavior pattern is obtained. , calculate the occurrence probability of any feature vector, and provide a theoretical basis for anomaly detection.

[0052] For the feature vector Calculating multi-dimensional distances , including time distance , spatial distance , sequence distance and frequency distance , expressed as: ; ; ; ; ; in, is the feature weight coefficient, , , , , , is the reference value, is the sequence edit distance function.

[0053] According to the collected feature vector A multi-dimensional comparison is performed with the normal behavior benchmark, and a comprehensive distance metric is obtained through weighted combination to reflect the degree of deviation of the current behavior from the normal pattern.

[0054] Calculate the degree of abnormality , multi-dimensional distance and behavioral models The probability distribution of is combined and expressed as: ; ; in, is the adaptive weight factor, is the adjustment coefficient, is the distance threshold, is the behavior model probability.

[0055] According to the degree of abnormality Set the alarm threshold , , }, when the abnormal degree When the corresponding threshold is exceeded, the corresponding level of alarm signal is triggered , specifically: If 0≤ < , it means that the system is in normal monitoring state, does not trigger the alarm mechanism, continues to accumulate behavior data, and updates the behavior model parameters.

[0056] like ≤ < , then set the alarm level = 1, trigger the response mechanism, increase the feature sampling frequency, record abnormal behavior characteristics, and prepare possible upgrade plans.

[0057] like ≤ < , then set the alarm level = 2, deploy enhanced protection strategies, call behavioral models for in-depth analysis, establish behavioral trajectory tracking, and prepare emergency response plans.

[0058] like ≥ , then set the alarm level = 3, execute emergency response, implement the most advanced protection strategy, start comprehensive monitoring mechanism, call historical data for analysis, and activate emergency response plan.

[0059] If the alarm level is determined, a monitoring log is generated to record the degree of abnormality , mark the alarm level , save the feature vector , add timestamp information, track changes in the degree of anomalies, the effectiveness of protection strategies, and update the system protection status.

[0060] In summary, the present invention establishes a standardized feature collection environment and introduces an environmental adaptability assessment mechanism. On the basis of accurately extracting the multi-dimensional features of the bill, the recognition results are dynamically corrected in combination with environmental parameters, thereby significantly improving the accuracy of bill authenticity identification. It can not only accurately characterize the standard feature distribution of the bill, but also conduct multi-level alarms through quantitative assessment of the degree of abnormality, and take differentiated treatment measures for bills of different risk levels. At the same time, the key information of the entire identification process is recorded through blockchain technology, and a complete traceability system is established to provide reliable data support for the traceability of fake bills and the optimization of anti-counterfeiting technology. It effectively solves the problems of insufficient recognition accuracy, lack of environmental adaptability, and traceability difficulties of traditional methods in complex environments, which is of great significance to improving the overall effectiveness of bill anti-counterfeiting identification.

[0061] This embodiment further provides a multi-modal biometric bill anti-counterfeiting identification system, including: The acquisition module is used to collect multimodal biometric features and environmental parameters, calculate the dynamic weight of each biometric feature through three dimensions: feature quality score, environmental adaptability score, and feature stability score, and dynamically fuse multiple biometric features using a weighted fusion algorithm to output the optimal feature combination; The encryption module is used to implement three-layer cascade anti-counterfeiting coding for fusion features. The first layer performs basic feature coding, and performs hash operations on fusion features, timestamps, and random numbers; the second layer performs associated feature coding, and encrypts the basic code, block information, and identity identification; the third layer performs time series feature coding, and digitally signs the associated code, time series information, and metadata, and stores them securely through distributed storage and blockchain technology; The early warning module establishes a real-time monitoring mechanism based on the anti-counterfeiting code, extracts time features, location features, frequency features and operation features to build a behavior model, and uses a multi-dimensional distance calculation method to evaluate the abnormality of the current behavior. When the abnormality exceeds the system preset threshold, it automatically triggers a graded alarm mechanism and emergency response strategy to achieve safe monitoring of the entire process of bill use.

[0062] This embodiment also provides a computer device, which is suitable for the case of a multimodal biometric bill anti-counterfeiting identification method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the multimodal biometric bill anti-counterfeiting identification method proposed in the above embodiment.

[0063] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0064] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the multimodal biometric bill anti-counterfeiting identification method proposed in the above embodiment is implemented.

[0065] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0066] Example 2: This example provides a multimodal biometric bill anti-counterfeiting identification method. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0067] In order to verify the effectiveness of the method of the present invention, three different methods were selected for comparative experiments, including a traditional single feature recognition method, a fixed threshold multi-feature recognition method and the method of the present invention.

[0068] The test sample is 1000 bills (including 800 genuine bills and 200 counterfeit bills); The tests were carried out in a standard environment (temperature 25°C, humidity 50%, light 500lux) and a complex environment (temperature fluctuation 15~35°C, humidity fluctuation 30%~80%, light fluctuation 100~1000lux).

[0069] The experimental results are shown in Table 1: Table 1 Comparative experimental results Test Method Environmental conditions Accuracy (%) False alarm rate (%) Missing reporting rate (%) Average recognition time (s) Traditional single feature method Standard environment 85.6 8.4 6.0 1.2 Traditional single feature method Complex environment 71.3 15.8 12.9 1.3 Fixed Threshold Multi-Feature Method Standard environment 91.2 5.1 3.7 2.1 Fixed Threshold Multi-Feature Method Complex environment 82.5 9.8 7.7 2.3 Method of the present invention Standard environment 97.8 1.3 0.9 2.5 Method of the present invention Complex environment 95.4 2.7 1.9 2.8 As shown in Table 1, the performance comparison results of the three bill anti-counterfeiting recognition methods under different environmental conditions are shown. From the data, it can be seen that the accuracy of the traditional single feature method is 85.6% in a standard environment, but it drops significantly to 71.3% in a complex environment, and the false alarm rate and missed alarm rate both increase significantly; the fixed threshold multi-feature method introduces multi-dimensional feature analysis, and the accuracy is increased to 91.2% in a standard environment and 82.5% in a complex environment, and the performance is relatively more stable.

[0070] The method of the present invention, by introducing an environmental adaptability assessment mechanism, achieves an accuracy rate of 97.8% and 95.4% in standard environments and complex environments, respectively, with both false alarm and missed alarm rates maintained at a low level. Although the average recognition time is slightly increased compared to traditional methods, this time overhead is completely acceptable considering the significant improvement in accuracy. In particular, the performance of the method of the present invention only slightly decreases in complex environments, which fully proves that the method has excellent adaptability and stability in response to environmental changes, and provides more reliable technical support for bill anti-counterfeiting recognition in practical applications.

[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A multi-modal biometric bill anti-counterfeiting identification method, characterized by: include, Collect multimodal biometric features and environmental parameters, calculate the dynamic weight of each biometric feature through three dimensions: feature quality score, environmental adaptability score, and feature stability score, use a weighted fusion algorithm to dynamically fuse multiple biometric features, and output the optimal feature combination; Implement three-layer cascade anti-counterfeiting coding for the fusion features. The first layer encodes the basic features and performs hash operations on the fusion features, timestamps, and random numbers. The second layer performs correlation feature coding, encrypting the basic code with block information and identity identification; the third layer performs time series feature coding, digitally signing the correlation code with time series information and metadata, and securely storing them through distributed storage and blockchain technology; A real-time monitoring mechanism is established based on the three-layer cascade anti-counterfeiting coding, and a behavior model is constructed by extracting time features, location features, frequency features and operation features. A multi-dimensional distance calculation method is used to evaluate the abnormality of the current behavior. When the abnormality exceeds the preset threshold of the system, a hierarchical alarm mechanism and emergency handling strategy are automatically triggered to achieve safe monitoring of the entire process of bill use.

2. The multimodal biometric bill anti-counterfeiting identification method according to claim 1, characterized in that: The biometric features include fingerprint features, palm print features, iris features and voice print features; Extracting the biometric feature to obtain biometric feature image data, including grayscale information, resolution information and acquisition time information of the feature; Feature point extraction and feature vector conversion are performed on the biometric feature image data to obtain a feature vector.

3. The multi-modal biometric bill anti-counterfeiting identification method according to claim 2, characterized in that: The feature point extraction adopts a method combining Gabor filter enhancement preprocessing and Harris corner point detection, and uses Gabor filter to enhance preprocess the biometric image data. The expression of Gabor filter is: ； ； in, is the filter direction; is the filter center frequency, is the spatial aspect ratio; is the standard deviation of the Gaussian envelope; Apply the Harris corner detection algorithm to the image enhanced by Gabor filtering, including: Calculate the image and Directional gradient and ; Calculate the second-order matrix M of the gradient: ； in, is a Gaussian window function centered at the point to be detected; Calculate corner response values : ； in, is the determinant of the matrix M, represents the trace of the matrix M, is an empirical constant; Corner point response value Perform non-maximum suppression and extract local maximum points as feature points; The feature points are extracted and expressed as feature vectors, which are expressed as: ； in, Indicates A set of feature vectors of biometric features, is the number of feature points, For the feature points coordinate, For the feature points coordinate, For the The gray value of feature points, For the The direction angle of the feature points, For the The regional attribute value of feature points.

4. The multimodal biometric bill anti-counterfeiting identification method according to claim 3, characterized in that: Euclidean distance based on feature points Similarity with attributes Calculate feature point matching , the calculation formula is: ； ； ； in, is the Euclidean distance between feature points, is the attribute difference value of the feature point, and is the standard deviation parameter of distance and attribute; According to the matching degree of feature points Calculating the biometric quality score , expressed as: ； in, is the feature point weight coefficient, is the total number of feature points; Calculate environmental suitability scores based on standardized environmental parameters , expressed as: ； in, , , , are the influencing factors of different biological characteristics, Represents the light intensity parameter, represents the temperature parameter, Represents humidity parameters, represents the noise parameter; Combined quality score and environmental adaptability score Generate fusion weight coefficients , expressed as: ； in, and is the adjustment coefficient; According to the fusion weight coefficient For the feature vector Perform weighted fusion to obtain fused feature data , expressed as: ； in, Represents fused feature data.

5. The multi-modal biometric bill anti-counterfeiting identification method according to claim 4, characterized in that: The three-layer cascade anti-counterfeiting code includes: Based on the fusion features With timestamp , random numbers The basic feature code is calculated by the SHA-256 hash algorithm ; Extract block information , including block height, block hash value, and the identity , encode the basic features With block information , Identity Perform RSA encryption to generate associated feature codes , expressed as: ； in, Represents the RSA public key encryption function, Contains block height and block hash ; A unique identifier for the user; Get the code generation time and the validity period of the bill , combined to form timing information ; Get geographic location information , including longitude, latitude and altitude information; obtain device identification information , including device serial number and hardware feature code; combining geographic location information and device identification information to form metadata ; The timing information , metadata and associated feature encoding Generate a digital signature using the ECDSA algorithm , the calculation formula is: ； ； ； in, Represents the ECDSA private key signature function, Contains the timestamp of creation and validity period ; Contains location coordinates and equipment ; and is a signature value pair; is the hash value of the message to be signed; Digital signature Write to the preset distributed storage node to generate the storage address , where the address is stored Generated through IPFS content addressing, expressed as: ； in, Indicates the addition operation of the IPFS network. Encode the address in Base58. Creating counterfeit-proof transactions on the blockchain network , the storage address , basic feature encoding , associated feature coding Packaging and writing anti-counterfeiting transactions ; Anti-counterfeiting transactions through broadcasting Go to the blockchain network and wait for the block to be confirmed before getting the transaction hash value ; Based on transaction hash , storage address Generate QR code , the QR code Write the ticket, where the QR code format is: ； in, Represents a QR code generation function.

6. The multimodal biometric bill anti-counterfeiting identification method according to claim 5, characterized in that: Collect ticket usage behavior characteristics, including timestamps , geographic coordinates , operation sequence And use interval , forming the feature vector ; Based on the feature vector , through the combination of multiple Gaussian components, construct a behavioral model , calculate the probability distribution of normal behavior, expressed as: ； in, is the weight coefficient, is the Gaussian component, represents the number of Gaussian distributions; The probability density calculation formula of the Gaussian component is: ； in, yes dimensional feature vector, It is Quantity dimensional mean vector, yes × dimensional covariance matrix, is the determinant of the covariance matrix, is the vector transpose, is the inverse of the covariance matrix.

7. The multi-modal biometric bill anti-counterfeiting identification method according to claim 6, characterized in that: For each Gaussian component, K-means clustering is used to determine the initial weight coefficient , given N training samples, train the behavior model, determine the log-likelihood function of the behavior model, and optimize the behavior model parameters. The log-likelihood function is expressed as: ； in, is the number of training samples, is the number of model parameters, is the likelihood function value; Use Bayesian Information Criterion BIC to select the optimal number of components , expressed as: ； in, is the optimal number of components; By maximizing the log-likelihood function, the EM algorithm is used to iteratively optimize the behavior model parameters to obtain a behavior model that describes the normal behavior pattern. , calculate the occurrence probability of any feature vector.

8. The multimodal biometric bill anti-counterfeiting identification method according to claim 7, characterized in that: Based on the feature vector , calculate multi-dimensional distance , including time distance , spatial distance , sequence distance and frequency distance , expressed as: ； ； ； ； ； in, is the feature weight coefficient, , , , , , is the reference value, is the sequence edit distance function; Based on the multi-dimensional distance and behavioral models The probability distribution of is combined to calculate the degree of abnormality , expressed as: ； ； in, is the adaptive weight factor, is the adjustment coefficient, is the distance threshold, is the behavior model probability; According to the degree of abnormality Set the alarm threshold , , }, when the abnormal degree When the corresponding threshold is exceeded, the corresponding level of alarm signal is triggered , specifically: If 0≤ < , it means that the system is in normal monitoring state, does not trigger the alarm mechanism, continues to accumulate behavior data, and updates behavior model parameters; like ≤ < , then set the alarm level = 1, trigger the response mechanism, increase the feature sampling frequency, record abnormal behavior characteristics, and prepare possible upgrade plans; like ≤ < , then set the alarm level = 2, deploy enhanced protection strategies, call behavioral models for in-depth analysis, establish behavioral trajectory tracking, and prepare emergency treatment plans; like ≥ , then set the alarm level = 3, execute emergency response, implement the most advanced protection strategy, start comprehensive monitoring mechanism, call historical data for analysis, and activate emergency response plan.

9. A multimodal biometric bill anti-counterfeiting identification system, based on the multimodal biometric bill anti-counterfeiting identification method according to any one of claims 1 to 7, characterized in that: include, The acquisition module is used to collect multimodal biometric features and environmental parameters, calculate the dynamic weight of each biometric feature through three dimensions: feature quality score, environmental adaptability score, and feature stability score, and dynamically fuse multiple biometric features using a weighted fusion algorithm to output the optimal feature combination; The encryption module is used to implement three-layer cascade anti-counterfeiting coding for the fusion features. The first layer performs basic feature coding and performs hash operations on the fusion features, timestamps, and random numbers. The second layer performs correlation feature coding, encrypting the basic code with block information and identity identification; the third layer performs time series feature coding, digitally signing the correlation code with time series information and metadata, and securely storing them through distributed storage and blockchain technology; The early warning module establishes a real-time monitoring mechanism based on the anti-counterfeiting code, extracts time features, location features, frequency features and operation features to build a behavior model, and uses a multi-dimensional distance calculation method to evaluate the abnormality of the current behavior. When the abnormality exceeds the system preset threshold, it automatically triggers a graded alarm mechanism and emergency response strategy to achieve safe monitoring of the entire process of bill use.

Images