Anti-counterfeiting verification method and system based on transparent laser spectrum characteristics

Abstract

This invention provides an anti-counterfeiting verification method and system based on transparent laser spectral features. The method includes: exciting a transparent laser verification area through mechanical variable-angle scanning, simultaneously acquiring the reflection spectrum, and constructing a spectral data matrix. By performing spectral differential analysis on this matrix, background noise is suppressed and characteristic wavelength trajectories are extracted. The instantaneous wavelength shift rate and spectral energy broadening of this trajectory are extracted and converted into a finite-state spectral state sequence. Initial screening is performed by matching a reference subsequence in this sequence using sliding hashing. If the match passes, the state transition probability matrix of the sequence is further constructed, and its relative entropy divergence with the reference matrix is ​​calculated to obtain the final authenticity determination result. This invention solves the technical problem of low verification reliability caused by the susceptibility of traditional methods to interference from the underlying background and local physical wear.

Description

Technical Field

[0001] This invention belongs to the field of transparent laser technology, specifically relating to an anti-counterfeiting verification method and system based on the spectral characteristics of transparent lasers. Background Technology

[0002] In the fields of commodity circulation and identity authentication, optical anti-counterfeiting features are commonly attached to the surface of target items to prevent counterfeiting and imitation. Among these, transparent laser anti-counterfeiting technology has gained some application because it can provide dynamic visual effects without obscuring the underlying printed information. Currently, the verification of such anti-counterfeiting features mainly relies on manual visual observation or the use of optical reading devices with fixed geometric configurations to collect optical images or spectral information at specific angles and compare them with pre-stored features of genuine product samples to determine their authenticity.

[0003] However, existing verification techniques primarily rely on matching deterministic optical characteristics of anti-counterfeiting features under single, static observation conditions. When the substrate background beneath the transparent laser verification area has complex, non-uniform colors or textures, the reflected light from the background will overlap with the diffracted light from the laser features, severely interfering with the verification signal and reducing the accuracy of feature extraction. Furthermore, items inevitably suffer from friction and scratches during circulation, causing localized, random physical damage to the microscopic grating structure of the transparent laser verification area. This damage causes deviations in its optical response in localized areas. If verification methods based on strict matching of deterministic feature points or global image templates are still used, genuine products with localized defects caused by normal wear and tear are easily misjudged as counterfeit products that do not meet standards. Summary of the Invention

[0004] This invention provides an anti-counterfeiting verification method and system based on transparent laser spectral characteristics to solve the above-mentioned technical problems.

[0005] In a first aspect, the present invention provides an anti-counterfeiting verification method based on transparent laser spectral characteristics, the method comprising the following steps: The mechanical variable angle scanning component drives the broadband light source to continuously scan and excite the transparent laser verification area on the surface of the target verification object at the incident angle, and simultaneously acquires the reflection spectrum signal at a fixed receiving angle to construct a spectral data matrix including time dimension, wavelength dimension and spectral intensity dimension. In the spectral data matrix, spectral difference analysis is performed on spectral slices of adjacent time dimensions to suppress non-frequency shift color background noise at the bottom of the target verification object, and the characteristic wavelength trajectory of the physical topology of the grating in the transparent laser verification area is extracted by tracing the coordinates of the difference extreme points. Extract the instantaneous wavelength shift rate and spectral energy broadening corresponding to the characteristic wavelength from the characteristic wavelength trajectory, and convert the continuous instantaneous wavelength shift rate and spectral energy broadening into finite state symbols according to the preset two-dimensional feature interval mapping rules to generate the spectral state sequence of the target verification object. The reference subsequence is obtained by cross-comparison of genuine samples under multi-substrate background interference by the cloud server, and sliding hash matching is performed in the spectral state sequence of the target verification object to perform the initial anti-counterfeiting screening verification. If the initial anti-counterfeiting screening verification fails, the blocking mechanism is triggered to determine that the verification has not passed. When the target verification object passes the initial anti-counterfeiting screening, the state transition probability matrix of the transparent laser verification area under local physical wear interference is constructed by traversing the statistical spectral state sequence and counting the frequency of transitions between all adjacent state symbols. Calculate the relative entropy divergence between the state transition probability matrix of the target verification object and the pre-stored baseline transition probability matrix, and obtain the authenticity judgment result of the target verification object based on the relative entropy divergence evaluation.

[0006] Optionally, the step of performing spectral difference analysis on spectral slices in adjacent time dimensions in the spectral data matrix to suppress non-frequency-shifted color background noise at the bottom of the target verification object, and extracting the characteristic wavelength trajectory of the physical topology of the transparent laser verification area grating by tracing the coordinates of the difference extremum points includes the following steps: Extract the first and second spectral slices from the spectral data matrix sequentially according to the time dimension. The first spectral slice and the second spectral slice are subtracted by the corresponding wavelength intensity to generate the original differential spectral data with suppressed non-frequency shift color background noise. The original differential spectral data is smoothed by using a Gaussian filter function to generate denoised differential spectral data. The coordinates of positive and negative difference extreme points are located in denoised differential spectral data using a derivative-based peak-finding algorithm. By connecting the coordinates of all positive and negative difference extreme points in all time dimensions in chronological order using curve fitting, the characteristic wavelength trajectory representing the physical topology of the grating in the transparent laser verification area is extracted.

[0007] Optionally, locating the coordinates of positive and negative difference extrema points in the denoised differential spectral data using a derivative-based peak-finding algorithm includes the following steps: Calculate the first derivative sequence of the denoised differential spectral data in the wavelength dimension and identify the zero-crossing points in the first derivative sequence; Calculate the second derivative of the denoised differential spectral data corresponding to the zero-crossing point location in the wavelength dimension; The sign of the second derivative value determines whether the zero-crossing point is a local maximum or a local minimum. The intensity of local maxima or local minima is compared with a preset noise floor removal threshold. Local maxima or local minima with intensity lower than the preset noise floor removal threshold are removed, while points with intensity higher than the preset noise floor removal threshold are retained as the coordinates of the positive and negative difference extreme points for localization.

[0008] Optionally, comparing the intensity of local maxima or minima with a preset noise floor removal threshold includes the following steps: Dark current reference data of the array-type spectral sensor was acquired when the broadband light source of the mechanical variable angle scanning component was not turned on; Calculate the standard deviation distribution matrix of dark current reference data in different time dimensions; The dynamic noise figure is calculated based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data; The dynamic noise figure is multiplied by a fixed empirical safety bias constant to generate a preset noise floor removal threshold. Extract the absolute intensity values ​​of local maxima or local minima and input them into a numerical comparator to compare them with a preset noise floor removal threshold.

[0009] Optionally, the calculation of the dynamic noise figure based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data includes the following steps: The environmental background noise intensity is obtained by integrating and summing the standard deviation distribution matrix along the time dimension. The global average energy is obtained by double integration along both the wavelength and time dimensions of the denoised differential spectral data. A signal-to-noise ratio (SNR) evaluation function is constructed that includes the ambient background noise intensity and the global average energy. A temperature compensation factor related to the motor temperature of the mechanical angle-scanning component is introduced into the SNR evaluation function. The real-time signal-to-noise ratio (SNR) value under the current test environment is calculated using the SNR evaluation function. A nonlinear logarithmic mapping model is used to map the real-time signal-to-noise ratio value to a normalized interval of zero to one, and the output result within the normalized interval is used as the dynamic noise coefficient.

[0010] Optionally, the step of extracting the instantaneous wavelength shift rate and spectral energy broadening corresponding to the feature wavelength trajectory, and converting the continuous instantaneous wavelength shift rate and spectral energy broadening into finite state symbols according to a preset two-dimensional feature interval mapping rule to generate the spectral state sequence of the target verification object includes the following steps: Perform a first-order differential operation on the characteristic wavelength trajectory in the time dimension to obtain the instantaneous wavelength offset rate corresponding to each time node; Locate the peak position of the characteristic wavelength trajectory in the spectral data matrix, and calculate the wavelength span when the intensity at the peak position drops to half to obtain spectral energy broadening; Construct a two-dimensional feature plane with instantaneous wavelength shift rate as the abscissa and spectral energy broadening as the ordinate; The preset two-dimensional feature interval mapping rules are applied to the two-dimensional feature plane, dividing the two-dimensional feature plane into multiple non-overlapping feature sub-intervals; Assign a unique and independent finite state symbol to each feature sub-interval; According to the chronological order of the time nodes, the characteristic sub-intervals where the instantaneous wavelength shift rate and spectral energy broadening are located at each time node are mapped to the corresponding finite state symbols, and combined to generate the spectral state sequence of the target verification object.

[0011] Optionally, the step of obtaining the reference subsequence selected by the cloud server based on cross-comparison of genuine samples under multi-substrate background interference, and performing sliding hash matching in the spectral state sequence of the target verification object for initial anti-counterfeiting screening verification, and triggering a blocking mechanism to determine that the verification has failed if the initial anti-counterfeiting screening verification fails, includes the following steps: Establish a secure communication connection with the cloud server and download the pre-compiled reference subsequence and the corresponding baseline hash signature; Set a sliding data window with the same length as the reference subsequence, and control the sliding data window to slide and traverse the spectral state sequence of the target verification object in steps of a single finite state symbol; At each position where the sliding data window stops, extract the data fragment within the window and calculate the local hash value. Then compare and match the local hash value with the base hash signature. When the sliding data window completes global traversal and all local hash values ​​fail to match the baseline hash signature, the blocking mechanism is triggered and an alarm signal indicating that the initial anti-counterfeiting verification has failed is output.

[0012] Optionally, when the target verification object passes the initial anti-counterfeiting screening, constructing the state transition probability matrix of the transparent laser verification area under local physical wear interference by traversing and statistically analyzing the frequency of transitions between all adjacent state symbols in the spectral state sequence includes the following steps: Receive the spectral state sequence that has passed the initial anti-counterfeiting screening and initialize an empty two-dimensional statistical counting matrix; Extract each pair of adjacent preceding and following state symbols from the spectral state sequence; Determine the coordinates of the corresponding elements in the two-dimensional statistical counting matrix based on the types of the preceding and following state symbols, and increment the transition frequency value under the corresponding element coordinates by one. Perform mathematical normalization on each row of the two-dimensional statistical counting matrix after traversal and statistical analysis. The normalized two-dimensional statistical counting matrix is ​​used as the state transition probability matrix of the transparent laser verification area under local physical wear interference.

[0013] In a second aspect, the present invention also provides an anti-counterfeiting verification system based on transparent laser spectral features, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the anti-counterfeiting verification method based on transparent laser spectral features as described in any one of the first aspects.

[0014] Thirdly, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the anti-counterfeiting verification method based on transparent laser spectral features according to any one of the first aspects.

[0015] The beneficial effects of this invention are: This invention employs an asymmetric optical configuration of mechanically variable-angle scanning excitation and fixed-angle acquisition to construct a spectral data matrix that comprehensively reflects the dynamic response of the physical topology of the transparent laser verification area grating with the incident angle. By performing spectral differential analysis on this data matrix, non-frequency shift interference introduced by the diverse color background of the verification object can be effectively suppressed, thereby extracting characteristic wavelength trajectories with high signal-to-noise ratio that directly characterize the grating diffraction properties. This method further extracts two higher-order dynamic features: instantaneous wavelength shift rate and spectral energy broadening, and transforms them into a finite sequence of state symbols. This transformation process reduces the dimensionality of the complex continuous spectral signal to a discrete state space that is easy to perform pattern matching and statistical analysis. Finally, through a dual verification mechanism of sliding hash initial screening and relative entropy divergence calculation of the state transition probability matrix, this invention ensures the existence of the core anti-counterfeiting features while also quantitatively evaluating the statistical stability of these features under random interference such as local physical wear, thus greatly improving the accuracy of anti-counterfeiting verification under complex substrate backgrounds or physical wear conditions. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an anti-counterfeiting verification method based on transparent laser spectral characteristics in one embodiment of this application.

[0017] Figure 2 This is a schematic diagram of the process of extracting characteristic wavelength trajectories through spectral difference analysis in one embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0019] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0020] Figure 1 This is a flowchart illustrating an anti-counterfeiting verification method based on transparent laser spectral features in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps. For example Figure 1 As shown, the anti-counterfeiting verification method based on transparent laser spectral characteristics disclosed in this invention specifically includes the following steps: S101. Control the mechanical variable angle scanning component to drive the broadband light source to continuously scan and excite the transparent laser verification area on the surface of the target verification object at the incident angle, and simultaneously acquire the reflection spectrum signal at a fixed receiving angle to construct a spectral data matrix including time dimension, wavelength dimension and spectral intensity dimension.

[0021] In the initial data acquisition phase of the transparent laser anti-counterfeiting verification system, the mechanical variable-angle scanning component drives a broadband light source stage via a precision stepper motor, causing the light source to generate a continuously and uniformly varying incident angle relative to the transparent laser verification area on the surface of the target verification object. The light emitted by the broadband light source covers the visible to near-infrared bands. When the beam illuminates the micro-nano grating structure of the transparent laser layer at different incident angles, due to the diffraction and interference effects of the grating, different wavelengths of light will be selectively enhanced or suppressed at specific angles. An array-type spectral sensor with a fixed receiving angle continuously records the reflected spectral signal. Each pixel of the sensor corresponds to a specific wavelength channel, and the sampling frequency is strictly synchronized with the mechanical scanning speed. The acquired raw data is organized into a three-dimensional tensor structure according to the three dimensions of timestamp, wavelength value, and light intensity, forming a spectral data matrix. ,in Indicates a time-dimensional index. Indicates wavelength dimension index, This represents the corresponding spectral intensity value. The temporal resolution of this matrix depends on the ratio of scan speed to sampling rate, while the wavelength resolution is determined by the grating line density of the spectral sensor. In a typical configuration, the temporal dimension contains hundreds of sampling points, and the wavelength dimension covers thousands of channels in the range of 400 to 1000 nanometers, providing a sufficient data foundation for subsequent differential analysis.

[0022] S102. In the spectral data matrix, perform spectral difference analysis operation on spectral slices of adjacent time dimensions to suppress non-frequency shift color background noise at the bottom layer of the target verification object, and extract the characteristic wavelength trajectory of the physical topology of the grating in the transparent laser verification area by tracing the coordinates of the difference extremum points.

[0023] After obtaining the complete spectral data matrix, it is necessary to extract the unique grating topological features of the transparent laser verification area while suppressing background interference from the underlying printed pattern or substrate color of the verification object. Following the natural order of time, two adjacent time slices are extracted sequentially, denoted as the first spectral slice. Second spectral slice The difference in incident angles between the two is minimal. Performing wavelength-by-wavelength intensity subtraction yields the original difference spectrum. Since the reflectivity of the underlying non-frequency-shifted colors remains essentially unchanged under minute angular changes, their contribution is effectively canceled out in the difference operation. However, the diffraction peak positions of the transparent laser grating exhibit wavelength shifts with changing incident angles, manifesting as alternating pairs of positive and negative extrema in the difference spectrum. To eliminate photoelectric noise and quantization errors, a Gaussian filter function is used to smooth the original difference spectrum. The filter kernel function is... ,in Using the filter standard deviation parameter, convolution operations generate denoised differential spectra. A derivative-based peak-finding algorithm is applied to the denoised differential spectrum to calculate the first derivative and identify zero-crossing points. The extremum type is determined by the sign of the second derivative. After eliminating false peaks with amplitudes below the dynamic threshold, the coordinates of all time-dimensional extremum points are connected sequentially to form a characteristic wavelength trajectory curve.

[0024] S103. Extract the instantaneous wavelength shift rate and spectral energy broadening corresponding to the characteristic wavelength trajectory, and convert the continuous instantaneous wavelength shift rate and spectral energy broadening into finite state symbols according to the preset two-dimensional feature interval mapping rules to generate the spectral state sequence of the target verification object.

[0025] The characteristic wavelength trajectory curve contains the dynamic response characteristics of the transparent laser grating, requiring further extraction of quantifiable feature parameters for subsequent encoding. A first-order numerical derivative is performed on the trajectory curve in the time dimension, and the instantaneous wavelength offset rate is calculated using a central difference scheme. This parameter reflects the sensitivity of the grating diffraction peak to changes in the incident angle and is closely related to the grating period and refractive index modulation depth. Simultaneously, the intensity peak positions corresponding to characteristic wavelengths are located in the spectral data matrix, and the wavelength span at which the peak intensity drops to half-maximum is measured as spectral energy broadening. This parameter characterizes the sharpness of the diffraction peak and is affected by grating uniformity and surface roughness. A two-dimensional feature plane is constructed with the instantaneous wavelength shift rate as the horizontal axis and spectral energy broadening as the vertical axis. The plane is divided into sections according to a preset mapping rule. There are three rectangular sub-intervals, each assigned a unique state symbol. The process iterates through all time points, assigning each point its... Coordinates are mapped to corresponding sub-intervals, the state symbols of those sub-intervals are extracted, and they are arranged in temporal order to form a spectral state sequence. ,in These are elements of a discrete symbol set. This symbolic encoding converts continuous spectral features into discrete sequences that facilitate pattern matching, while preserving the core information of the grating topology.

[0026] S104. Obtain the reference subsequence selected by the cloud server based on cross-comparison of genuine samples under multi-substrate background interference, and perform sliding hash matching in the spectral state sequence of the target verification object for initial anti-counterfeiting screening verification. If the initial anti-counterfeiting screening verification fails, the blocking mechanism is triggered to determine that the verification has not passed.

[0027] The first step in anti-counterfeiting verification is rapid initial screening based on reference subsequences. This is achieved through secure communication with a cloud server to obtain reference subsequences and their hash signatures, pre-extracted from a large number of genuine product samples. The reference subsequences are stable feature fragments selected after cross-comparison under various substrate background interference conditions, typically several tens of symbols in length. A sliding window of the same length as the reference subsequence is set, and it slides across the spectral state sequence of the target verification object with a single-symbol step size. Each time the window stops, a data fragment within the window is extracted, and its local hash value is calculated. To enhance anti-interference capabilities, volatile nodes located near the boundaries of feature sub-intervals within the window are identified, i.e., the Euclidean distance of each symbol from the sub-interval boundary in the two-dimensional feature plane is calculated. ,in The coordinates of the feature point corresponding to the symbol. is the boundary point of the subinterval. When If a node is identified as a critically volatile node, a generic placeholder is used to replace the generated mask data fragment, which is then input into a Locality Sensitive Hash Encoder to calculate a soft hash vector. The local hash value is then compared to each element in the baseline hash signature set using a Hamming distance calculation. If the minimum Hamming distance is less than a preset tolerance threshold, a match is considered successful. If no match is found after a global traversal, a blocking mechanism is triggered, outputting a verification failure alert to prevent further computational resource waste.

[0028] S105. When the target verification object passes the initial anti-counterfeiting screening verification, the state transition probability matrix of the transparent laser verification area under local physical wear interference is constructed by traversing the statistical spectral state sequence to count the frequency of transitions between all adjacent state symbols.

[0029] Among them, the spectral state sequences that pass the initial screening and verification enter the deep feature analysis stage, where a probability matrix is ​​constructed using statistical state transition modes to characterize the microscopic physical properties of the grating. Initialize a... The zero matrix as a statistical counting matrix ,in This represents the total number of state symbols. It iterates through each pair of adjacent symbols in the spectral state sequence. According to the preceding symbols The type determines the matrix row index, and the subsequent symbols. The type determines the column index, and the corresponding element... The value is increased by 1. After completing the full sequence traversal, the statistical counting matrix records the frequency of all possible state transitions. Normalization is performed on each row of the matrix, and the transition probability is calculated. The state transition probability matrix is ​​obtained. Each element of this matrix represents the state. Transition to state The conditional probability reflects the dynamic response law of the transparent laser grating under continuous angular scanning. Even if there is local wear or contamination in the verification area, as long as the main structure of the grating is not damaged, the statistical law of its state transition remains stable.

[0030] S106. Calculate the relative entropy divergence between the state transition probability matrix of the target verification object and the pre-stored baseline transition probability matrix, and obtain the authenticity judgment result of the target verification object based on the relative entropy divergence.

[0031] The final determination of authenticity is based on the similarity assessment between the state transition probability matrix of the target verification object and the pre-stored benchmark matrix. Relative entropy is used as the metric, and the calculation formula is as follows: ,in For the target matrix, The baseline matrix is ​​defined as follows. Relative entropy divergence quantifies the degree of difference between two probability distributions; a smaller value indicates greater similarity. The baseline matrix is ​​derived from the statistical average of genuine samples and represents the standard transfer pattern of a genuine transparent laser grating. The calculated relative entropy divergence is compared with a preset judgment threshold. If the divergence is less than the threshold, the product is judged to be genuine; otherwise, it is judged to be counterfeit or severely damaged. The advantage of this method lies in utilizing higher-order statistical features rather than simple sequence matching. Even if counterfeit products can mimic local spectral features, they cannot easily reproduce the complete transfer probability distribution of a genuine grating under full-angle scanning. Furthermore, relative entropy has a certain tolerance to small perturbations in matrix elements, making the verification results robust to measurement noise and environmental changes, ensuring the reliability and practicality of anti-counterfeiting verification.

[0032] In one embodiment, reference is made to Figure 2 In the spectral data matrix, spectral difference analysis is performed on spectral slices of adjacent time dimensions to suppress non-frequency-shifted color background noise at the bottom of the target verification object. The characteristic wavelength trajectory of the physical topology of the transparent laser verification area grating is extracted by tracing the coordinates of the difference extrema points, including the following steps: S201. Extract the first and second spectral slices from the spectral data matrix sequentially according to the time dimension; S202. Subtract the corresponding wavelength intensities from the first spectral slice and the second spectral slice to generate the original differential spectral data with suppressed non-frequency shift color background noise; S203. The original differential spectral data is smoothed by using a Gaussian filter function to generate denoised differential spectral data; S204. Locate the coordinates of positive and negative difference extreme points in denoised differential spectral data using a derivative-based peak-finding algorithm; S205. Connect the coordinates of all positive and negative difference extreme points in all time dimensions by curve fitting in time order, and extract the characteristic wavelength trajectory representing the physical topology of the grating in the transparent laser verification area.

[0033] In this embodiment, the spectral data matrix The time dimension includes multiple spectral snapshots acquired at fixed time intervals during the mechanical angle-shifting scan. Each snapshot corresponds to the complete wavelength-intensity distribution at a specific incident angle. The extraction operation starts from the initial index in the time dimension and reads adjacent time slices sequentially in ascending order. The first spectral slice... Corresponding time Spectral distribution, second spectral slice Corresponding time The spectral distribution is determined by the scanning speed and sampling frequency, with the time interval between two moments being determined by the scanning speed and sampling frequency. In a typical configuration, the time interval is approximately tens of milliseconds, corresponding to an incident angle change of less than 0.5 degrees. The extraction process is implemented through array indexing operations, cutting two-dimensional sub-matrices along the time axis from the three-dimensional tensor structure. Each sub-matrix has rows corresponding to wavelength channels and columns representing single time points, with the data format maintained as floating-point values ​​to ensure the accuracy of subsequent calculations. The selection of adjacent slices follows the principle of continuity, ensuring that the incident angle change is small enough to keep the underlying non-frequency-shifted background stable, while being large enough to produce a detectable wavelength shift in the grating diffraction peaks. After extraction, the two slices are temporarily stored in a memory buffer, awaiting the call of the difference operation. The entire extraction process is executed cyclically in the data processing pipeline until all time dimensions have been traversed.

[0034] Differential operations are a core step in suppressing background interference, based on the fundamental difference between the frequency-shift response characteristics of the transparent laser grating and the non-frequency-shift characteristics of the underlying printed pattern. The first spectral slice is then... With the second spectral slice Intensity subtraction is performed point-by-point at the same wavelength position to calculate the original difference spectrum. The diffuse reflectance spectrum of the underlying substrate remains almost unchanged with small angular changes, and its contribution approaches zero after subtraction, while the diffraction peaks of the transparent laser layer undergo wavelength shift due to the angle dependence of the Bragg condition. When the incident angle changes from... Change to When the wavelength that satisfies the diffraction condition is from Drift to This led to A negative difference peak appears at [location], at [location]. A positive difference peak appears. The spacing and amplitude of the positive and negative peak pairs in the difference spectrum are directly related to the grating period and modulation depth. The subtraction operation uses element-wise subtraction, directly subtracting the intensity values ​​of the corresponding wavelength channels, resulting in linear computational complexity. The difference result may contain positive, negative, and near-zero values. Positive regions represent wavelengths with increased intensity, negative regions represent wavelengths with decreased intensity, and zero regions correspond to an unchanged background. The original difference spectrum retains complete information about the grating's dynamic response but also contains interference components such as photoelectric noise, quantization error, and ambient light fluctuations.

[0035] High-frequency noise in the original differential spectral data can interfere with the accuracy of subsequent extreme point localization. Gaussian filtering achieves smooth noise reduction through weighted averaging. Gaussian filter function. A bell-shaped weight distribution centered at zero is defined, with a standard deviation parameter. The smoothness of the control filter is relatively large. The value produces a stronger smoothing effect but may lose detailed features. In practical applications... Typically, sampling intervals of 2 to 5 wavelengths are set to balance noise suppression and feature preservation. Convolution operation The filter kernel slides across the differential spectrum, and a weighted sum is calculated at each position as the output value. In practice, this is implemented using discrete convolution, where the sum is calculated for each wavelength point. The intensity value after denoising Summation range Take as The corresponding number of sampling points covers 99.7% of the Gaussian weights. A mirror expansion strategy is used for the convolution boundaries to avoid distortion caused by edge effects. The denoised differential spectral data preserves the positive and negative peak pairs structure generated by diffraction peak drift, while effectively suppressing random noise and spikes, providing a clean data foundation for subsequent derivative calculations and peak finding.

[0036] The derivative-based peak-finding algorithm accurately locates extreme points by analyzing the derivative characteristics of the denoised differential spectrum. It calculates the first-order derivative sequence of the denoised differential spectrum along the wavelength dimension using a central difference scheme. ,in The wavelength sampling interval is defined. The first derivative reflects the rate of change of spectral intensity. At extreme points, the derivative crosses zero; a change from positive to negative corresponds to a maximum, and a change from negative to positive corresponds to a minimum. Zero-crossing points are identified by detecting the sign changes of adjacent elements in the first derivative sequence, and the wavelength coordinates of these zero-crossing points are recorded. To distinguish between true extrema and pseudo-zeros in plateau regions, the second derivative at the zero-crossing point is calculated. A negative second derivative confirms that the point is a local maximum, while a positive second derivative confirms that the point is a local minimum. Further amplitude threshold filtering is applied, and the absolute intensity at the extreme points is calculated. Weak peaks below a dynamic threshold are removed by comparison with the dynamic threshold. The dynamic threshold is adaptively adjusted based on the global statistical characteristics of the differential spectrum, avoiding the adaptability problem of a fixed threshold across different samples. The coordinates of the positive and negative differential extreme points are stored in the form of wavelength-intensity pairs, forming the feature point set of this time slice.

[0037] The construction of characteristic wavelength trajectories organizes discrete extremum coordinates into continuous time-series curves, revealing the complete evolution of grating diffraction peaks with varying incident angles. Positive and negative difference extremum points located across all time dimensions are collected and categorized according to timestamp and extremum type. For the positive difference extremum point sequence, the peak wavelength coordinates at each time point are arranged in chronological order, forming an incremental time-wavelength data point set. Similarly, processing the negative difference extreme point sequence yields... Polynomial fitting or spline interpolation methods are used to connect discrete points into a smooth curve. The fitting order is chosen based on the data point density and curve complexity; typically, cubic spline interpolation can accurately follow data trends while maintaining smoothness. The fitting process minimizes the sum of squared residuals between the curve and the data points, generating a continuous function. and This is the characteristic wavelength trajectory. The positive and negative trajectories correspond to the enhancement and reduction sides of the diffraction peak shift, respectively, and the spacing between them reflects the peak width variation. The characteristic wavelength trajectory fully records the optical response fingerprint of the grating physical topology under angular scanning, and different grating periods, groove depths, and duty cycles produce unique trajectory shapes.

[0038] In one embodiment, locating the coordinates of positive and negative difference extrema points in denoised differential spectral data using a derivative-based peak-finding algorithm includes the following steps: Calculate the first derivative sequence of the denoised differential spectral data in the wavelength dimension and identify the zero-crossing points in the first derivative sequence; Calculate the second derivative of the denoised differential spectral data corresponding to the zero-crossing point location in the wavelength dimension; The sign of the second derivative value determines whether the zero-crossing point is a local maximum or a local minimum. The intensity of local maxima or local minima is compared with a preset noise floor removal threshold. Local maxima or local minima with intensity lower than the preset noise floor removal threshold are removed, while points with intensity higher than the preset noise floor removal threshold are retained as the coordinates of the positive and negative difference extreme points for localization.

[0039] In this embodiment, denoised differential spectral data The diffraction curve exhibits fluctuations along the wavelength dimension due to the peak shift of the transparent laser grating, with extreme points corresponding to the wavelengths where diffraction enhancement or attenuation is most significant. The first derivative is calculated using a numerical differentiation method for each wavelength sampling point. The central difference scheme is applied, which estimates the slope at a given location by dividing the intensity difference between two adjacent sampling points by the wavelength interval. A complete first-order derivative sequence is generated by traversing all wavelength sampling points; the sequence length is two endpoints shorter than the original data. Zero-crossing point identification is achieved by detecting the sign change of adjacent elements in the first-order derivative sequence. When, it means that the derivative is in and The wavelengths cross zero. Linear interpolation is used to accurately locate the wavelength coordinates of the zero-crossing points. The set of zero-crossing points contains all possible candidate locations for extrema, but it may contain spurious zeros from the platform region and false extrema caused by noise, requiring further screening and confirmation using the second derivative criterion.

[0040] The second derivative reflects the concavity and convexity of the denoised differential spectral curve, providing a mathematical basis for determining the type of extrema at the location of extrema points. For the identified zero-crossing point locations... Find the index of the sampling point closest to that location. The second derivative value at a given point is calculated. The calculation of the second derivative is based on the three-point difference formula, constructing a curvature estimate of a quadratic polynomial using the intensity values ​​of the current point and its two adjacent points. When the curve bends downwards at a point, the second derivative is negative, indicating a peak; when the curve bends upwards, the second derivative is positive, indicating a trough. In transparent laser anti-counterfeiting applications, positive difference extrema correspond to the wavelength position where the diffraction peak strengthens after drift, manifesting as a local maximum in the difference spectrum, with a negative second derivative value; negative difference extrema correspond to the wavelength position where the diffraction peak weakens before drift, manifesting as a local minimum in the difference spectrum, with a positive second derivative value. The absolute value of the second derivative also reflects the sharpness of the extremum; a larger absolute value indicates a sharper peak, typically corresponding to a deeper grating modulation depth or higher spectral resolution. The calculation process is performed on all zero-crossing points one by one, generating a list of second derivative values ​​corresponding to each zero-crossing point.

[0041] The extremum type determination is based on the sign information of the second derivative value, explicitly classifying zero-crossing points as either local maxima or local minima. The determination rule is: when the second derivative is constant, zero-crossing points... If a point is identified as a local maximum, record its wavelength coordinates and the corresponding differential spectral intensity value. A positive intensity value indicates a spectral enhancement relative to the previous moment. When the second derivative is zero, the zero-crossing point is identified as a local minimum, and its wavelength coordinates and corresponding differential spectral intensity value are recorded. A negative intensity value indicates a spectral weakening relative to the previous moment. The determination process is implemented using a sign function or conditional branch statement, which is computationally efficient and logically clear. In actual data processing, numerical calculation errors may lead to the second derivative being exactly zero. A very small tolerance value is set. When the absolute value of the second derivative is less than the tolerance, the zero-crossing point is considered to be located in a plateau region rather than a true extremum, and it is removed from the candidate list. After classification, two independent sets of extrema are obtained: the set of local maxima corresponds to positive peaks in the differential spectrum, and the set of local minima corresponds to negative peaks. The two sets usually show a paired relationship in the time series, reflecting the complete process of the diffraction peak drifting from one wavelength position to another.

[0042] Noise floor removal is a critical quality control step to ensure the reliability of extreme points. It filters out spurious extreme points caused by noise fluctuations by using an intensity threshold. The preset noise floor removal threshold is determined based on the noise level of the measurement system and the signal dynamic range. Setting the threshold too low will cause many noise peaks to be misidentified as valid features, while setting it too high may miss true but weak diffraction signals. For each extreme point of a identified type, the absolute value of the differential spectral intensity at the corresponding location is extracted. Compared with the preset noise removal threshold Perform numerical comparisons. When When the extreme point is determined to be a spurious peak caused by noise interference, it is removed from the extreme point list; when When an extreme point is confirmed as a valid diffraction feature, it is retained in the list and marked as the coordinates of the positive and negative difference extreme points. The threshold comparison uses absolute values ​​because the intensity of local minima is negative; after taking the absolute value, the signal strength is uniformly evaluated based on the amplitude. In a preferred embodiment, the noise floor removal threshold employs a dynamic adaptive strategy, adjusting in real time according to the global noise level of the current difference spectrum. The specific calculation method is described in detail in subsequent claims. After the removal operation is completed, the remaining extreme points constitute a high-confidence feature point set. The wavelength coordinates and intensity information are packaged and stored as the basic data points for constructing the feature wavelength trajectory, ensuring that subsequent anti-counterfeiting verification is based on reliable optical features.

[0043] In one embodiment, comparing the intensity of a local maximum or local minimum point with a preset noise floor removal threshold includes the following steps: Dark current reference data of the array-type spectral sensor was acquired when the broadband light source of the mechanical variable angle scanning component was not turned on; Calculate the standard deviation distribution matrix of dark current reference data in different time dimensions; The dynamic noise figure is calculated based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data; The dynamic noise figure is multiplied by a fixed empirical safety bias constant to generate a preset noise floor removal threshold. Extract the absolute intensity values ​​of local maxima or local minima and input them into a numerical comparator to compare them with a preset noise floor removal threshold.

[0044] In this embodiment, dark current reference data acquisition is performed before the formal spectral measurement to characterize the background noise characteristics of the array-type spectrometer under no-light conditions. The mechanical angle-changing scanning assembly remains stationary, and the broadband light source is completely off to ensure that no external light signals enter the sensor's photosensitive element. Even in this dark environment, the spectrometer still outputs a non-zero signal, primarily due to thermally excited dark current from the photodiode, electronic noise from the readout circuit, and quantization noise from the analog-to-digital converter. The acquisition process lasts for the same time span as the formal measurement, recording the output values ​​of each wavelength channel of the sensor at the same sampling frequency to generate a dark current reference data matrix. The time dimension of this matrix contains hundreds of sampling points, and the wavelength dimension covers the entire spectral channels of the sensor. Dark current values ​​are typically in the single-digit to tens of digits of the digital quantization unit, far lower than the intensity of normal spectral signals. The acquisition environment requires strict light shielding to avoid interference from stray light, while maintaining temperature stability because dark current is temperature-sensitive. Repeated acquisitions and averaging improve the statistical reliability of the benchmark data. The dark current benchmark data fully records the spatiotemporal distribution characteristics of the sensor's noise, providing a baseline reference for subsequent dynamic threshold calculations, allowing threshold settings to adapt to the noise level of specific hardware rather than relying on fixed empirical values.

[0045] The standard deviation distribution matrix quantifies the degree of fluctuation in dark current reference data over time, reflecting the randomness and stability of sensor noise. For each wavelength channel... Extract the numerical sequence of this channel at all time points from the dark current reference data matrix. Calculate the standard deviation of the sequence. ,in This is the time average of the channel. This represents the total number of time sampling points. A larger standard deviation indicates more severe noise fluctuations in that wavelength channel, which may be related to sensor pixel defects or circuit crosstalk. The same calculation is performed across all wavelength channels, and the standard deviation values ​​of each channel are arranged in wavelength order to form a one-dimensional standard deviation distribution vector. This is further extended to the time dimension, for each time slice... Calculate the standard deviation of all wavelength channel values ​​at that moment, generating a standard deviation sequence in the time dimension. The standard deviation information in both dimensions is combined to form a standard deviation distribution matrix. The element values ​​of this matrix represent the noise intensity at the corresponding spatiotemporal location. The standard deviation distribution matrix reveals the non-uniformity of sensor noise; certain wavelength regions or time periods may exhibit higher noise levels. This spatial and temporal difference in noise distribution is fully considered in dynamic threshold calculation to avoid generating too many false alarms in high-noise areas or missing weak signals in low-noise areas with a uniform threshold.

[0046] The calculation of the dynamic noise figure integrates the sensor's inherent noise characteristics with the actual measured signal strength, establishing an adaptive mechanism for signal-to-noise ratio sensing. First, the standard deviation distribution matrix... By performing double integration and summation along both the time and wavelength dimensions, a global estimate of the ambient background noise intensity is obtained. ,in This represents the total number of wavelength channels. This integral value reflects the accumulated noise level throughout the measurement process. It is also applied to the denoised differential spectral data. The global average energy is calculated by performing double integration across all time and wavelength dimensions. This parameter characterizes the average amplitude level of the differential spectral signal. A signal-to-noise ratio evaluation function is constructed. ,in To prevent extremely small positive numbers with a denominator of zero, a higher signal-to-noise ratio (SNR) indicates a more significant signal relative to noise, allowing for a lower relative threshold; conversely, a lower SNR requires a higher threshold to suppress noise interference. A temperature compensation factor is introduced. ,in The real-time temperature of the motor of the mechanical variable angle scanning component. For standard reference temperature, This is the temperature coefficient. The corrected signal-to-noise ratio is... Using a nonlinear logarithmic mapping model The signal-to-noise ratio is mapped to a normalized interval from zero to one, where and The normalized output is the dynamic noise figure, which is the mapping parameter.

[0047] The preset noise floor removal threshold is generated by multiplying the dynamic noise figure by a fixed empirical safety bias constant, achieving a balance between adaptability and robustness. Dynamic noise figure The value has been normalized to the range of zero to one. The closer the value is to one, the higher the noise level or the weaker the signal, requiring a higher rejection threshold. A fixed empirical safety bias constant is used. Based on extensive experimental statistics, the typical value range is a few percent of the differential spectral dynamic range; for example, in a 16-bit quantization system, it might be set to 100 to 500 quantization levels. The threshold calculation formula is as follows: This formula converts the dimensionless noise figure into a threshold value with spectral intensity dimensions. When the signal-to-noise ratio of the measurement environment is good, the dynamic noise figure is small, resulting in a low threshold and allowing the detection of weak diffraction features. When environmental noise increases or the signal weakens, the dynamic noise figure increases, and the threshold increases accordingly, preventing noise peaks from being misjudged as valid extrema. The introduction of a safety bias constant provides an additional safety margin; even if there are deviations in the dynamic calculation, the fixed bias still ensures basic noise suppression capability. The threshold generation process is executed once at the beginning of each measurement session, and the generated threshold remains unchanged throughout the scan, ensuring consistency of the judgment criteria. The threshold value is stored in a memory variable for subsequent extremum point screening steps.

[0048] The numerical comparison operation quantitatively compares the intensity of each candidate extremum point with a preset noise floor removal threshold, and performs a binary decision to retain or remove it. For local maxima of a determined type, the differential spectral intensity value at the corresponding wavelength position is extracted. Since the maxima point corresponds to a positive difference peak, the intensity value itself is positive and is directly used as the absolute intensity value. For local minima, the corresponding difference spectral intensity value is extracted as a negative number, and its absolute value is taken to obtain the absolute intensity value. The absolute intensity value is input into a numerical comparator circuit or a software comparison function and compared with a preset noise floor removal threshold. Perform a size comparison. The comparator outputs a binary logic signal, when... When the value is high, it outputs a high level or a true value, indicating that the extreme point has passed the threshold test and is retained in the list of valid feature points; when... A low-level or false value is output when the extreme point fails the test and is removed from the list. The comparison process is performed on all candidate extreme points one by one, and the comparison result of each extreme point is determined independently without affecting each other. In hardware implementation, a parallel comparator array can be used to improve processing speed, while in software implementation, it can be done through conditional statements or vectorized operations. After the comparison is completed, a set of filtered extreme points is generated. Each element in the set contains wavelength coordinates, intensity values, and extreme value type labels. These data points constitute high-confidence grating diffraction features, which are used for subsequent feature wavelength trajectory fitting and state sequence generation to ensure that the anti-counterfeiting judgment is based on reliable physical signals rather than noise interference.

[0049] In one implementation, calculating the dynamic noise figure based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data includes the following steps: The environmental background noise intensity is obtained by integrating and summing the standard deviation distribution matrix along the time dimension. The global average energy is obtained by double integration along both the wavelength and time dimensions of the denoised differential spectral data. A signal-to-noise ratio (SNR) evaluation function is constructed that includes the ambient background noise intensity and the global average energy. A temperature compensation factor related to the motor temperature of the mechanical angle-scanning component is introduced into the SNR evaluation function. The real-time signal-to-noise ratio (SNR) value under the current test environment is calculated using the SNR evaluation function. A nonlinear logarithmic mapping model is used to map the real-time signal-to-noise ratio value to a normalized interval of zero to one, and the output result within the normalized interval is used as the dynamic noise coefficient.

[0050] In this embodiment, the standard deviation distribution matrix It contains complete distribution information of sensor noise in both time and wavelength dimensions. Integrating and summing along the time dimension yields the global cumulative effect of the noise. The integration operation is implemented in a discrete data structure through summation, for each wavelength channel. The standard deviation of the channel at all time points is summed to obtain This value reflects the cumulative noise intensity of that wavelength channel throughout the entire measurement period. Further summing the cumulative values ​​for all wavelength channels yields the ambient background noise intensity. The ambient noise background intensity comprehensively considers the noise differences, temporal drift, and spatial non-uniformity of each sensor pixel, making it more representative of the overall noise level of the system than single-point noise measurement. Physically, this parameter is equivalent to the sum of the noise standard deviations of all spatiotemporal locations, similar to the noise power integral in signal processing. Its value is linearly affected by the number of sampling points and wavelength channels, therefore normalization is required when comparing measurement systems with different configurations. As the denominator in the signal-to-noise ratio calculation, the ambient noise background intensity directly affects the adaptive adjustment range of the threshold. Higher noise intensity results in a larger dynamic noise figure, ultimately leading to stricter extreme point selection criteria, effectively preventing noise peaks from being misjudged as valid diffraction features.

[0051] Denoising differential spectral data The dual integral calculation extracts the global energy features of the signal, which serve as the numerator for signal-to-noise ratio evaluation. First, integration is performed over each time slice along the wavelength dimension, calculating the sum of the absolute values ​​of the differential intensities of all wavelength channels at that moment. Absolute value calculations ensure that the contributions of both positive and negative difference peaks are included, avoiding energy underestimation caused by the cancellation of positive and negative peaks. The energy distribution of the time series is obtained by traversing all time slices. Furthermore, the energy sequence is summed over time and divided by the total number of sampling points to obtain the global average energy. This parameter characterizes the average amplitude level of the differential spectral signal, reflecting the intensity of the diffraction effect of the transparent laser grating. A larger global average energy value results from a larger grating modulation depth, a wider scanning angle range, or higher light source intensity; conversely, a lower energy value occurs when the grating is worn, contaminated, or under poor measurement conditions. The normalization process using dual integration eliminates the influence of data size, making measurement sequences of different lengths comparable. The ratio of global average energy to ambient background noise intensity forms the basis of the signal-to-noise ratio (SNR). High energy and low noise correspond to a high SNR, allowing for a lower detection threshold to capture more detailed features; low energy and high noise correspond to a low SNR, requiring a higher threshold to ensure reliable judgment.

[0052] The signal-to-noise ratio (SNR) evaluation function establishes a quantitative relationship between signal strength and noise level, and corrects for the influence of thermal effects on the mechanical system using a temperature compensation factor. The fundamental SNR is defined as the ratio of global average energy to the ambient background noise level. ,in Typical value To prevent division overflow when the noise level is zero. The stepper motor in the mechanical variable-angle scanning component generates heat during continuous operation, causing thermal expansion of the mechanical structure and refractive index drift of the optical elements, thus affecting the stability of spectral measurements. Temperature compensation factor. The effect of temperature deviation on the signal-to-noise ratio was quantified, where The real-time temperature of the motor is obtained through a thermistor or infrared temperature sensor. The standard reference temperature is usually set at 25 degrees Celsius. The temperature coefficient is determined based on system calibration experiments, with a typical value of [value missing]. to Every degree Celsius. When the temperature rises... Greater than 1, corrected signal-to-noise ratio The corresponding reduction reflects the signal quality degradation caused by thermal effects. The temperature compensation mechanism enables the threshold setting to adapt to thermal drift during the measurement process, which is especially important in long-term continuous scanning scenarios to avoid fluctuations in the false positive rate caused by temperature changes.

[0053] The real-time signal-to-noise ratio (SNR) value is obtained through direct calculation using the SNR evaluation function, comprehensively reflecting the relative relationship between signal and noise under the current test environment. The global average energy calculated in the preceding steps... Background noise intensity and the motor temperature collected in real time Substituting into the correction formula, first calculate the fundamental signal-to-noise ratio. This value is a dimensionless ratio, typically ranging from several to tens of times. Then, the temperature compensation factor is calculated. Finally, the corrected real-time signal-to-noise ratio is obtained. The physical meaning of the real-time signal-to-noise ratio (SNR) is the multiple of the signal amplitude relative to the noise standard deviation; a higher value indicates that the signal is more easily identified from noise. In anti-counterfeiting verification applications, genuine samples typically exhibit a high SNR due to their intact grating structure, while counterfeit or damaged products show a significantly lower SNR. The real-time SNR is also affected by measurement conditions; light source aging, optical path contamination, or environmental vibrations can all lead to a decrease in the value. As an intermediate variable in dynamic threshold calculation, the accuracy of this parameter directly affects the final extreme point selection effect; therefore, the calculation process must ensure numerical precision and avoid the accumulation of floating-point operation errors.

[0054] The nonlinear logarithmic mapping model converts the real-time signal-to-noise ratio (SNR) value into a normalized dynamic noise coefficient, establishing a nonlinear response relationship between the SNR and the threshold adjustment amplitude. The mapping function uses a Sigmoid curve. ,in and The mapping parameters are determined through statistical fitting of a large number of samples, with typical values... , Logarithmic operations The signal-to-noise ratio (SNR) multiple relationship is transformed into an additive relationship, compressing the numerical dynamic range for easier subsequent linear processing. The exponential function and sigmoid structure achieve a smooth nonlinear mapping; when the SNR is high, the logarithmic term is positive and has a large absolute value, while the exponential term approaches zero. A value close to zero indicates that the noise effect is negligible; when the signal-to-noise ratio is very low, the logarithmic term is negative, and the exponential term tends to infinity. A threshold close to 1 indicates noise dominance and requires a high threshold. The intermediate signal-to-noise ratio region has the steepest slope on the mapping curve, making it sensitive to changes in signal-to-noise ratio and allowing for fine-tuning of the threshold. Normalization interval. The output is easy to multiply with a fixed safety bias constant to generate the final threshold. The dimensionless nature of the mapping model makes it universal and applicable to measurement systems with different quantization bits and dynamic ranges.

[0055] In one embodiment, the instantaneous wavelength shift rate of the characteristic wavelength trajectory and the spectral energy broadening corresponding to the characteristic wavelength are extracted, and the continuous instantaneous wavelength shift rate and spectral energy broadening are converted into finite state symbols according to a preset two-dimensional feature interval mapping rule to generate the spectral state sequence of the target verification object, including the following steps: Perform a first-order differential operation on the characteristic wavelength trajectory in the time dimension to obtain the instantaneous wavelength offset rate corresponding to each time node; Locate the peak position of the characteristic wavelength trajectory in the spectral data matrix, and calculate the wavelength span when the intensity at the peak position drops to half to obtain spectral energy broadening; Construct a two-dimensional feature plane with instantaneous wavelength shift rate as the abscissa and spectral energy broadening as the ordinate; The preset two-dimensional feature interval mapping rules are applied to the two-dimensional feature plane, dividing the two-dimensional feature plane into multiple non-overlapping feature sub-intervals; Assign a unique and independent finite state symbol to each feature sub-interval; According to the chronological order of the time nodes, the characteristic sub-intervals where the instantaneous wavelength shift rate and spectral energy broadening are located at each time node are mapped to the corresponding finite state symbols, and combined to generate the spectral state sequence of the target verification object.

[0056] In this embodiment, the characteristic wavelength trajectory and The wavelength evolution of the diffraction peaks of a transparent laser grating as a function of the incident angle is described. The instantaneous rate of wavelength change can be extracted by performing a first-order differential of the trajectory in the time dimension. The instantaneous wavelength shift rate quantitatively characterizes the grating's spectral response sensitivity to angular perturbations. This parameter satisfies the differential relationship of the Bragg diffraction equation with the grating period, diffraction order, and incident light wavelength. Numerical differentiation methods are used to calculate the wavelength shift for each time node. Application Center Differential Scheme ,in and These are the characteristic wavelength values ​​at adjacent time points. The time sampling interval is defined as . Central difference has higher numerical accuracy than one-sided difference, with the error order being the square of the time interval. For the beginning and end endpoints of the trajectory, due to the lack of bilateral neighbors, forward or backward difference schemes are used for calculation. All time points of the characteristic wavelength trajectory are traversed to generate an instantaneous wavelength offset sequence. The offset rate is measured in nanometers per second and can be positive or negative. A positive value indicates a shift in wavelength towards longer wavelengths, while a negative value indicates a shift towards shorter wavelengths. The absolute value of the offset rate reflects the grating structure parameters; gratings with smaller periods are more sensitive to angular changes and produce larger offset rates.

[0057] Spectral energy broadening quantifies the sharpness characteristics of diffraction peaks, reflecting the grating surface quality and structural uniformity. In the spectral data matrix... In the middle, characteristic wavelength trajectory corresponding time and wavelength The coordinates of the peak position in the matrix were determined. The spectral intensity value at that position was extracted as the peak intensity. Calculate half-strength Searching along the wavelength dimension on both sides of the peak position, we find the two wavelength points where the intensity first drops to half the height, denoted as the left wavelength. and the wavelength on the right Spectral energy broadening is defined as the wavelength span between two half-maximum points. This parameter is called the full width at half maximum (FWHM) in spectroscopy. The search process achieves precise positioning through linear interpolation. When the intensity values ​​of two adjacent sampling points are above and below the FWHM, the wavelength coordinates of the intersection point are calculated through interpolation. The physical significance of spectral energy broadening is related to the coherence length and angular resolution of the grating. Increased surface roughness or non-uniformity of the grating lines will lead to increased broadening, resulting in a wider and shorter peak. Typical transparent laser gratings broaden in the range of a few nanometers to tens of nanometers. High-quality gratings have smaller broadening and exhibit sharp peaks, while low-quality or worn gratings show increased broadening and flattened peaks. The spectral energy broadening is calculated by traversing all time points to generate a broadening sequence. It is used as a second feature dimension paired with the instantaneous wavelength offset rate.

[0058] The construction of the two-dimensional feature plane organizes two independent temporal feature parameters into a planar coordinate system, facilitating subsequent interval partitioning and sign mapping. A rectangular coordinate system is established, with the horizontal axis representing the instantaneous wavelength shift rate. The vertical axis represents spectral energy broadening. The range of values ​​for the horizontal axis is determined based on the statistical analysis of the offset rates at all time points, with the minimum value recorded. and maximum value Set the coordinate axis range to ,in To allow for margin expansion at the boundaries, boundary points are prevented from falling on the endpoints of the coordinate axes. The range of values ​​for the ordinate is also determined based on the statistical range of the widening sequence. Each time point Corresponding feature parameter pairs This is represented as a coordinate point on a two-dimensional plane. Plotting the coordinates of all time points on the plane forms a feature point cloud distribution. The distribution pattern of the point cloud reflects the comprehensive response characteristics of the grating during the angle scanning process. Genuine samples typically exhibit a clustered structure in their point cloud, while counterfeit or damaged samples show scattered or deviated point clouds from the normal area.

[0059] The pre-defined two-dimensional feature interval mapping rule discretizes a continuous feature plane into a finite number of non-overlapping sub-intervals, achieving a symbolic representation of the features. The mapping rule employs a rectangular grid partitioning strategy, equally dividing the offset rate range along the horizontal axis. There are intervals, each with a width of . Divide the expansion range equally along the vertical axis. There are intervals, each with a height of . The intersection of the horizontal and vertical dividing lines forms... There are three rectangular sub-intervals, each sub-interval is defined by its lower left corner coordinates. and the coordinates of the upper right corner The only certainty is that among them , Sub-intervals share boundaries but do not overlap, covering the entire feature plane. Partition parameters. and The choice of partitioning balances encoding precision and the number of states. Too small a partition leads to feature loss, while too large a partition generates too many states, increasing computational complexity. A typical configuration is... , This generates 48 sub-intervals. The mapping rule can employ a non-uniform partitioning strategy, subdividing the mesh to improve resolution in dense feature point cloud regions and coarsening the mesh to reduce redundancy in sparse regions. The geometry of the sub-intervals can also be polar coordinate sector shapes or irregular polygons, optimizing the partitioning effect based on the geometric characteristics of the feature distribution.

[0060] The allocation of finite state symbols establishes a unique identifier for each feature sub-interval, realizing the mapping from the feature space to the symbol space. An alphanumeric combination encoding scheme is adopted for... The sub-intervals are numbered in row-major or column-major order, and are the [number]th [interval]. Sub-interval allocation symbols Symbols can be represented by single characters, such as A to Z covering 26 states, or by combinations of two characters, such as A1, A2, and extended to any number. In one implementation, a Base64 encoded character set is used, containing 64 symbols including uppercase and lowercase letters, numbers, and special characters, sufficient to cover the number of common sub-intervals. Symbol assignment follows the principle of uniqueness; different sub-intervals correspond to different symbols, and the same sub-interval always corresponds to the same symbol. The symbol set constitutes the state space of the finite state machine, where each symbol represents the physical state of the grating under a specific combination of offset rate and stretching. The mapping relationship between symbols and sub-intervals is stored in a lookup table, where the row and column indices correspond to the sub-interval coordinates, and the table elements are the assigned symbols. The generation of the spectral state sequence converts continuous feature parameters into discrete symbol strings in chronological order, completing the encoding from the physical domain to the information domain. All time nodes are traversed. Extract the corresponding instantaneous wavelength shift rate and spectral energy broadening Locate coordinate points on a two-dimensional feature plane The sub-interval in which it is located.

[0061] The determination method is to check whether the coordinate points satisfy the boundary conditions of the subinterval. and After finding the subinterval containing the coordinate point, the corresponding finite state symbol is extracted from the symbol lookup table. Arrange the symbols at each moment in natural order to generate a spectral state sequence. The length of the state sequence is equal to the number of time sampling points. Each symbol in the sequence records the transient response state of the grating at that moment. The state sequence retains the sequential information of the time dimension, and the transition relationship between adjacent symbols reflects the dynamic evolution law of the grating state. The state sequence of genuine samples exhibits a regular transition pattern, with smooth symbol changes and high repeatability; the state sequence of counterfeit samples is chaotic and disordered, with frequent symbol jumps and a lack of statistical regularity.

[0062] In one implementation, a reference subsequence is obtained by cross-referencing genuine samples under multi-substrate background interference by a cloud server, and sliding hash matching is performed in the spectral state sequence of the target verification object for initial anti-counterfeiting screening. If the initial anti-counterfeiting screening fails, a blocking mechanism is triggered to determine that the verification has not passed, including the following steps: Establish a secure communication connection with the cloud server and download the pre-compiled reference subsequence and the corresponding baseline hash signature; Set a sliding data window with the same length as the reference subsequence, and control the sliding data window to slide and traverse the spectral state sequence of the target verification object in steps of a single finite state symbol; At each position where the sliding data window stops, extract the data fragment within the window and calculate the local hash value. Then compare and match the local hash value with the base hash signature. When the sliding data window completes global traversal and all local hash values ​​fail to match the baseline hash signature, the blocking mechanism is triggered and an alarm signal indicating that the initial anti-counterfeiting verification has failed is output.

[0063] In this embodiment, the anti-counterfeiting verification terminal device establishes an encrypted communication channel with the cloud server through a network interface, employing the Transport Layer Security (TLS) 1.3 protocol to ensure the confidentiality and integrity of data transmission. During the handshake process, the server's digital certificate is verified, and a session key is negotiated after confirming the server's legitimacy. After establishing the connection, a download request is sent, containing the product category code and substrate type identifier of the target verification item. The cloud server retrieves the corresponding set of reference subsequences from its database based on the request parameters. These reference subsequences are stable feature fragments extracted from the spectral state sequences of a large number of genuine samples. Through cross-comparison and screening under multi-substrate background interference, core state transition patterns that consistently appear under different underlying printing patterns and substrate colors are retained. Each reference subsequence is typically 15 to 30 symbols long, covering the typical response of the grating within a specific angular range. The server also returns a pre-calculated set of baseline hash signatures, with each reference subsequence corresponding to one or more hash signatures. The hash algorithm uses Locality Sensitive Hashing (LSH) to ensure that similar sequences produce similar hash values. The downloaded data packet is digitally signed to ensure it has not been tampered with, and after decryption, it is stored in the secure memory area of ​​the terminal device.

[0064] The sliding data window mechanism enables a global search and matching of the reference subsequence within the spectral state sequence of the target verification material. The window length is set to... The number of symbols is strictly consistent with that of the reference subsequence to ensure that the data fragments within the window are comparable to the reference template. During initialization, the window is positioned on the spectral state sequence. The starting position, window overlay symbol The sliding operation increments by a single symbol, and after each slide, the window moves one position to the right, covering the symbol. ,in This is the starting index of the current window. The sliding process continues until the right boundary of the window reaches the end of the sequence, and a total of [number] operations are performed. The single-symbol step-size fine-grained sliding ensures that no possible matching positions are missed, even if the starting position of the reference subsequence in the target sequence is slightly offset. The sliding window technique, derived from time-frequency analysis methods in signal processing, is innovatively applied to pattern matching of discrete symbol sequences in this scheme, achieving global similarity evaluation through segment-by-segment comparison of local features. The window sliding is implemented by a loop control structure, updating the window boundary index and triggering subsequent hash calculations and matching determination processes in each iteration.

[0065] The calculation of local hash values ​​and the matching determination are performed independently at each window's dwell position, enabling fast similarity assessment. Data fragments within the current window are extracted. The segment is of length The data segment is input into a locality-sensitive hash encoder, which first converts the symbol sequence into a numerical vector, mapping each symbol to the center coordinates of its corresponding sub-interval. The feature vector is flattened into a one-dimensional feature vector. Dimensionality reduction is performed using a random projection matrix, and the projection result is binarized using a sign function to generate a hash code. The core characteristic of Locality Sensitive Hashing (LSH) is that similar inputs produce similar outputs; hash codes with small Hamming distances correspond to neighboring data points in the feature space. The calculated local hash values ​​are fixed-length binary strings, typically 64 to 128 bits long. The local hash value is then compared element-wise with each element in the baseline hash signature set to calculate the Hamming distance. ,in The length of the hash code. For XOR operation, and The local hash value and the base hash signature are respectively the first Bits. When the minimum Hamming distance is less than the preset hash matching tolerance threshold. When a match is found at the current window position, a matching flag is recorded, and the sliding can be selectively terminated to pass the initial screening ahead of time. The tolerance threshold is set to balance the false positive rate and the false negative rate. A typical value is 10% to 20% of the hash code length, allowing a small amount of bit difference to tolerate measurement noise and local wear.

[0066] The blocking mechanism is triggered when the sliding window completes a global traversal and finds no matching position, performing a fast rejection decision to avoid wasting subsequent computational resources. The criterion for completing the traversal is the starting index of the window. achieve At this point, all possible window positions have been checked. The matching results for all stopping positions are statistically analyzed. If the number of successful matches is zero, it indicates that the spectral state sequence of the target verification object is dissimilar to any known genuine reference subsequence, and is highly likely to be a counterfeit or severely damaged product. The blocking mechanism immediately halts subsequent state transition probability matrix calculations and relative entropy divergence assessments; these in-depth analysis steps are computationally complex and only meaningful for samples that have passed the initial screening. When blocking is triggered, an alarm signal indicating that the initial anti-counterfeiting screening verification failed is generated. The signal includes a timestamp, device identifier, and failure reason code. The alarm notifies the operator via audible and visual prompts, is recorded in a local log file, and can optionally be uploaded to a cloud server for statistical analysis. The blocking mechanism is designed according to the fail-fast principle, eliminating obviously abnormal samples with low computational cost in the early stages of the verification process, and concentrating limited computational resources on in-depth analysis of samples with lower suspicion levels.

[0067] In one implementation, at each sliding stop position of the sliding data window, a data segment within the window is extracted and a local hash value is calculated. The local hash value is then compared and matched with the base hash signature, including the following steps: Extract data segments within the sliding data window, and simultaneously extract the absolute intensity peaks and spectral energy broadening corresponding to each finite state symbol within the data segment from the spectral data matrix; The boundary distance Euclidean margin of each finite state symbol in the characteristic sub-interval corresponding to the mapping rule of the two-dimensional characteristic interval is calculated by using the absolute intensity peak value and spectral energy broadening. Finite state symbols whose boundary distance Euclidean margin is lower than the preset jitter tolerance threshold are identified as optically critically volatile nodes affected by mechanical scanning divergence angle interference. The optical critically volatile nodes in the data segment are replaced with universal placeholders to generate masked anti-disturbance data segments, and the masked anti-disturbance data segments are input into a preset local sensitive hash encoder to calculate and generate local soft hash feature vectors as local hash values. Calculate the Hamming distance between the local hash value and each baseline hash signature in the baseline hash signature set issued by the cloud server. When the minimum Hamming distance is less than the hash matching tolerance threshold, it is confirmed that the local hash value of the sliding data window at the current sliding position is successfully matched with the baseline hash signature.

[0068] In this embodiment, the data segment covered by the sliding data window at its current position. Includes There are a finite number of state symbols, each corresponding to a characteristic state at a specific time point in the spectral state sequence. To enhance the robustness of hash matching, it is necessary to backtrack to the original spectral data to extract richer physical information. For each symbol in the data segment... The spectral data matrix is ​​located by the timestamp index at the time of symbol generation. The corresponding time slice In this time slice, the wavelength coordinates are recorded based on the characteristic wavelength trajectory. The spectral intensity peak at this location is extracted as the absolute intensity peak. Simultaneously extract the calculated spectral energy broadening at that moment. This parameter was used as the ordinate during the state symbol generation phase. By traversing all symbols within the data segment, an absolute intensity peak sequence is constructed. and spectral energy broadening sequence The absolute intensity peak reflects the strength of the diffraction signal and is affected by the power of the light source, the efficiency of the grating, and the receiving angle; the spectral energy broadening reflects the sharpness of the peak shape and is affected by the quality of the grating and the surface condition.

[0069] Boundary distance Euclidean margin quantifies the distance of a feature point from the boundary of its sub-interval in the two-dimensional feature plane, identifying unstable symbols located near the classification boundary. For symbols in a data segment... The feature coordinates corresponding to this symbol during generation are: The boundaries of the subinterval are defined by four straight lines, namely the left boundary. Right boundary Lower boundary and upper boundary Calculate the perpendicular distances from the feature points to the four boundaries, which are respectively... , , and The boundary distance Euclidean margin is defined as the distance to the nearest boundary. The smaller this value, the closer the feature point is to the sub-interval boundary, and the lower the certainty of symbol classification. Physically, feature points near the boundary correspond to critical states where even small fluctuations in measurement parameters can cause symbol jumps. For example, angular jitter of the mechanical scanning component, fluctuations in light source intensity, or changes in ambient temperature can all cause feature points to cross the boundary into adjacent sub-intervals, resulting in different symbol outputs. The boundary distance Euclidean margin provides a quantitative assessment of symbol stability. Symbols with a large margin correspond to feature points located in the central region of the sub-interval, resulting in robust classification results; symbols with a small margin correspond to critical boundary states and are easily affected by interference, leading to misclassification. The boundary distance Euclidean margin is calculated for each symbol within the data segment, generating a margin sequence. This provides a basis for subsequent identification of volatile nodes.

[0070] The identification of optically critically volatile nodes is based on a comparison between the boundary distance Euclidean margin and a preset jitter tolerance threshold, filtering out symbols with high classification uncertainty. A jitter tolerance threshold is then set. This threshold is determined based on a combination of the angular repeatability of the mechanical scanning assembly and the measurement noise level of the spectral sensor, typically ranging from 5% to 15% of the sub-interval width or height. For symbols in the data segment... If the corresponding boundary distance has Euclidean margin If a symbol is found to be optically critically volatile, it is identified as an unreliable symbol. The physical causes of volatile nodes include minute deviations in the incident angle due to mechanical scanning divergence, uneven illumination caused by light source beam divergence, and fluctuations in diffraction conditions due to local thickness variations in the transparent laser layer. These factors manifest as random symbol jumps at feature points near the boundary, and the same sample may switch between adjacent symbols in repeated measurements. By identifying volatile nodes and reducing their weight or masking them in subsequent processing, the tolerance of hash matching to measurement uncertainties can be improved. Threshold comparisons are performed on all symbols within the data segment to statistically analyze the number and location distribution of volatile nodes.

[0071] The generation of masked, anti-interference data segments eliminates the interference of volatile nodes on hash calculations through symbol substitution operations, achieving robust feature encoding. For data segments... The symbols marked as optically critically volatile nodes are replaced with predefined universal placeholders to generate masked anti-interference data fragments. Universal placeholders use special characters such as asterisks or question marks, which do not conflict with any valid state symbols in the symbol set. The replacement operation preserves the length of the data segment and the position of non-volatile symbols, only replacing the identifier of the volatile symbol with a placeholder. The masked, anti-scratching data segment is input into a locality-sensitive hash encoder. The encoder employs a special mapping strategy when processing placeholders, setting the feature vector corresponding to the placeholder to a zero vector or a random vector to minimize its contribution to the final hash code. The locality-sensitive hash encoder uses a random projection method to map the high-dimensional symbol sequence to a low-dimensional binary hash code. The randomness of the projection matrix ensures that similar inputs produce similar outputs. The encoding process first converts the symbol sequence into numerical feature vectors, mapping each symbol to the center coordinates of its corresponding sub-interval. Placeholders are mapped to zero coordinates. After being flattened into a one-dimensional vector, it is multiplied by a random projection matrix. The projection result is binarized using a sign function, mapping positive values ​​to 1 and negative values ​​to 0, generating a fixed-length local soft hash feature vector. This vector serves as the local hash value, typically 64 to 128 bits in length, preserving the main features of the data fragment while maintaining high computational and storage efficiency.

[0072] Hamming distance calculation is used to evaluate the similarity between local hash values ​​and a baseline hash signature set, determining whether a match is successful. The baseline hash signature set distributed by the cloud server contains... Each hash code corresponds to a standard feature of a reference subsequence under specific conditions. The local hash value is calculated for the current window position. Each element in the base hash signature set is compared one by one. Calculate Hamming distance ,in , Let be the length of the hash code. The Hamming distance counts the number of distinct bits in two binary strings, achieved through XOR operations and summation, with a linear computational complexity. Find the minimum Hamming distance among all possible values. This minimum distance represents the degree of difference between the local hash value and the most similar benchmark signature. The minimum Hamming distance is then compared with a hash matching tolerance threshold. When comparing, When the local hash value at the current sliding window position matches the baseline hash signature, a matching flag is recorded, and the matching baseline signature index and Hamming distance value can be selectively recorded. The tolerance threshold is set to take into account the effects of masking operations and measurement noise, typically ranging from 10% to 20% of the hash code length. A successful match indicates that the spectral response characteristics of the target verification object within the current angular range are highly similar to those of a known genuine sample, passing the initial anti-counterfeiting screening verification and allowing it to proceed to the subsequent in-depth analysis stage.

[0073] In one implementation, when the target verification object passes the initial anti-counterfeiting screening, the state transition probability matrix of the transparent laser verification area under local physical wear interference is constructed by traversing the frequency of transitions between all adjacent state symbols in the statistical spectral state sequence. This includes the following steps: Receive the spectral state sequence that has passed the initial anti-counterfeiting screening and initialize an empty two-dimensional statistical counting matrix; Extract each pair of adjacent preceding and following state symbols from the spectral state sequence; Determine the coordinates of the corresponding elements in the two-dimensional statistical counting matrix based on the types of the preceding and following state symbols, and increment the transition frequency value under the corresponding element coordinates by one. Perform mathematical normalization on each row of the two-dimensional statistical counting matrix after traversal and statistical analysis. The normalized two-dimensional statistical counting matrix is ​​used as the state transition probability matrix of the transparent laser verification area under local physical wear interference.

[0074] In this embodiment, the spectral state sequence verified through anti-counterfeiting initial screening. It has been confirmed that the sample contains core feature fragments similar to those of the genuine sample, and has entered the deep statistical analysis stage to extract a more refined physical topological fingerprint. The construction of the state transition probability matrix is ​​based on Markov chain theory, modeling the dynamic response of the grating under continuous angular scanning as a stochastic transition process between discrete states. Initialization operation creates... Two-dimensional statistical counting matrix ,in The cardinality of the finite set of state symbols is the total number of subintervals divided by the two-dimensional feature plane. The row indices of the matrix represent the preceding states, and the column indices represent the following states. Each element... Used to accumulate from state Transition to state The number of occurrences. During initialization, all elements of the matrix are assigned zero, and integers are used for storage to ensure counting precision. The dimension of the matrix depends on the granularity of the feature plane partitioning; if... If a rectangular grid is used to generate 48 state symbols, then the statistical counting matrix is: There are a total of 2304 elements.

[0075] The extraction of adjacent state symbol pairs involves traversing the spectral state sequence to capture the temporal relationships of state transitions. Starting from the beginning of the sequence, each pair of adjacent symbols is extracted sequentially. ,in Precedence state symbol Corresponding time Grating response state, subsequent state symbol Corresponding time The states are defined, with the time interval between them equal to the sampling period. State transitions reflect the dynamic evolution of the grating under minute changes in the incident angle. Due to the continuity of mechanical scanning, states at adjacent moments usually remain unchanged or transition to neighboring states; large-scale state jumps are relatively rare. The extraction process is implemented through iterative loops, with each iteration reading the symbols at the current and next positions, forming a transition pair. A total of [number] symbols are extracted. For each transition pair, a longer sequence length provides a richer statistical sample and a more accurate transition probability estimate. Physically, state transitions correspond to the trajectory movement of feature points on a two-dimensional feature plane. Smooth trajectories produce concentrated transition patterns, while erratic trajectories produce dispersed transition patterns. Genuine samples, due to their regular grating structure, exhibit strong regularity in state transitions, with some transition paths occurring significantly more frequently than others. Counterfeit samples, due to structural defects or imitation errors, show chaotic state transition patterns, with a more uniform distribution of transition frequencies.

[0076] The statistical accumulation of transition frequencies maps the extracted state transition pairs to corresponding elements of the statistical counting matrix and increments the count. For each transition pair... Determine the preceding state symbol Index in the symbol set Determine the subsequent state symbol index The index is determined using a symbol lookup table, which stores the mapping between symbols and sub-interval numbers. The row index in the location statistics counting matrix is... Column index is elements This increments the value of the element by 1, i.e. The accumulation operation applies to all... Each transition pair is executed sequentially, with each transition pair affecting only one element in the matrix. After the traversal is complete, the matrix elements are... The numerical value represents the state. Transition to state The total number of occurrences during the entire scan. Row sum of the matrix. Equal to state The total number of occurrences of the preceding state reflects the frequency of access to that state. Diagonal elements The value represents the number of self-transitions of the state, i.e., maintaining the same state for two consecutive time steps. A larger value indicates higher stability of the state. Off-diagonal elements represent transitions between different states. The distribution pattern of element values ​​characterizes the statistical regularity of the grating's dynamic response, constituting a higher-order feature that distinguishes true from false states.

[0077] Normalization transforms the statistical counting matrix into a probability matrix, eliminating the influence of sequence length differences and satisfying the constraints of probability axioms. For each row of the matrix... Perform normalization independently and calculate the sum of all elements in the row. The sum represents the value from the state The total number of transitions that occurred. If Divide each element of the row by the row sum to obtain the normalized value. ;like Representing state If a row has never appeared as a preceding state in the sequence, set all elements in that row to zero or a uniform distribution. To avoid division by zero errors in subsequent calculations. The normalized matrix. Satisfies the property of a probability matrix, each element And the sum of the elements in each row is 1, that is .element Indicates the state Transition to state under conditions The conditional probability reflects the relative tendency of state transitions. The normalization operation eliminates the influence of absolute frequency, making the probability matrices generated by measurement sequences of different lengths comparable, which facilitates similarity assessment with a pre-stored benchmark matrix.

[0078] State transition probability matrix As a statistical fingerprint of the transparent laser verification area under local physical wear interference, it encodes the global dynamic characteristics of the grating's micro-topology. Even if there are local scratches, stains, or wear in the verification area, as long as the main structure of the grating is not damaged, the statistical law of state transition remains relatively stable. Local defects may cause state anomalies at individual moments, but under long-term statistical averaging, the influence of abnormal states is diluted, and the dominant transition mode still reflects the essential characteristics of the genuine product. The sparsity and non-uniformity of the probability matrix reflect the degree of determinism of the grating response. The genuine product matrix usually presents blocky or strip-shaped high-probability regions, corresponding to regular state evolution paths; the counterfeit product matrix tends to be uniformly distributed and lacks obvious structural features. The probability matrix, as a high-dimensional feature expression, contains... Each parameter has an independent parameter, and its information capacity far exceeds that of simple sequence matching.

[0079] The present invention also discloses an anti-counterfeiting verification system based on transparent laser spectral features, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the anti-counterfeiting verification method based on transparent laser spectral features as described above.

[0080] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it.

[0081] The memory can be an internal storage unit of a computer device, such as a hard disk or RAM, or an external storage device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) provided on the computer device. Furthermore, the memory can be a combination of internal storage units and external storage devices of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0082] The present invention also discloses a computer-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the anti-counterfeiting verification method based on transparent laser spectral features described in any of the above embodiments.

[0083] The computer program can be stored in a machine-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The machine-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the machine-readable medium includes, but is not limited to, the above-mentioned components.

[0084] The anti-counterfeiting verification method based on transparent laser spectral features described in the above embodiments is stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above method.

[0085] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0086] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A method for anti-counterfeiting verification based on transparent laser spectral characteristics, characterized in that, Includes the following steps: The mechanical variable angle scanning component drives the broadband light source to continuously scan and excite the transparent laser verification area on the surface of the target verification object at the incident angle, and simultaneously acquires the reflection spectrum signal at a fixed receiving angle to construct a spectral data matrix including time dimension, wavelength dimension and spectral intensity dimension. In the spectral data matrix, spectral difference analysis is performed on spectral slices of adjacent time dimensions to suppress non-frequency shift color background noise at the bottom of the target verification object, and the characteristic wavelength trajectory of the physical topology of the grating in the transparent laser verification area is extracted by tracing the coordinates of the difference extreme points. Extract the instantaneous wavelength shift rate and spectral energy broadening corresponding to the characteristic wavelength from the characteristic wavelength trajectory, and convert the continuous instantaneous wavelength shift rate and spectral energy broadening into finite state symbols according to the preset two-dimensional feature interval mapping rules to generate the spectral state sequence of the target verification object. The reference subsequence is obtained by cross-comparison of genuine samples under multi-substrate background interference by the cloud server, and sliding hash matching is performed in the spectral state sequence of the target verification object to perform the initial anti-counterfeiting screening verification. If the initial anti-counterfeiting screening verification fails, the blocking mechanism is triggered to determine that the verification has not passed. When the target verification object passes the initial anti-counterfeiting screening, the state transition probability matrix of the transparent laser verification area under local physical wear interference is constructed by traversing the statistical spectral state sequence and counting the frequency of transitions between all adjacent state symbols. Calculate the relative entropy divergence between the state transition probability matrix of the target verification object and the pre-stored baseline transition probability matrix, and obtain the authenticity judgment result of the target verification object based on the relative entropy divergence evaluation.

2. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 1, characterized in that, The step of performing spectral difference analysis on spectral slices in adjacent time dimensions within the spectral data matrix to suppress non-frequency-shifted color background noise at the bottom of the target verification object, and extracting the characteristic wavelength trajectory of the physical topology of the transparent laser verification area grating by tracing the coordinates of the difference extrema points, includes the following steps: Extract the first and second spectral slices from the spectral data matrix sequentially according to the time dimension. The first spectral slice and the second spectral slice are subtracted by the corresponding wavelength intensity to generate the original differential spectral data with suppressed non-frequency shift color background noise. The original differential spectral data is smoothed by using a Gaussian filter function to generate denoised differential spectral data. In denoised differential spectral data, the coordinates of positive and negative difference extrema points are located using a derivative-based peak-finding algorithm. By connecting the coordinates of all positive and negative difference extreme points in all time dimensions in chronological order using curve fitting, the characteristic wavelength trajectory representing the physical topology of the grating in the transparent laser verification area is extracted.

3. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 2, characterized in that, The process of locating the coordinates of positive and negative difference extrema points in denoised differential spectral data using a derivative-based peak-finding algorithm includes the following steps: Calculate the first derivative sequence of the denoised differential spectral data in the wavelength dimension and identify the zero-crossing points in the first derivative sequence; Calculate the second derivative of the denoised differential spectral data corresponding to the zero-crossing point in the wavelength dimension; The sign of the second derivative value determines whether the zero-crossing point is a local maximum or a local minimum. The intensity of local maxima or local minima is compared with a preset noise floor removal threshold. Local maxima or local minima with intensity lower than the preset noise floor removal threshold are removed, while points with intensity higher than the preset noise floor removal threshold are retained as the coordinates of the positive and negative difference extreme points for localization.

4. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 3, characterized in that, The step of comparing the intensity of local maxima or local minima with a preset noise floor removal threshold includes the following steps: Dark current reference data of the array-type spectral sensor was acquired when the broadband light source of the mechanical variable angle scanning component was not turned on; Calculate the standard deviation distribution matrix of dark current reference data in different time dimensions; The dynamic noise figure is calculated based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data; The dynamic noise figure is multiplied by a fixed empirical safety bias constant to generate a preset noise floor removal threshold. Extract the absolute intensity values ​​of local maxima or local minima and input them into a numerical comparator to compare them with a preset noise floor removal threshold.

5. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 4, characterized in that, The calculation of the dynamic noise figure based on the standard deviation distribution matrix and the global average energy of the denoised differential spectral data includes the following steps: The environmental background noise intensity is obtained by integrating and summing the standard deviation distribution matrix along the time dimension. The global average energy is obtained by double integration along both the wavelength and time dimensions of the denoised differential spectral data. A signal-to-noise ratio (SNR) evaluation function is constructed that includes the ambient background noise intensity and the global average energy. A temperature compensation factor related to the motor temperature of the mechanical angle-scanning component is introduced into the SNR evaluation function. The real-time signal-to-noise ratio (SNR) value under the current test environment is calculated using the SNR evaluation function. A nonlinear logarithmic mapping model is used to map the real-time signal-to-noise ratio value to a normalized interval of zero to one, and the output result within the normalized interval is used as the dynamic noise coefficient.

6. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 1, characterized in that, The process of extracting the instantaneous wavelength shift rate and spectral energy broadening corresponding to the characteristic wavelength trajectory, and converting the continuous instantaneous wavelength shift rate and spectral energy broadening into finite state symbols according to a preset two-dimensional feature interval mapping rule to generate the spectral state sequence of the target verification object includes the following steps: Perform a first-order differential operation on the characteristic wavelength trajectory in the time dimension to obtain the instantaneous wavelength offset rate corresponding to each time node; Locate the peak position of the characteristic wavelength trajectory in the spectral data matrix, and calculate the wavelength span when the intensity at the peak position drops to half to obtain spectral energy broadening; Construct a two-dimensional feature plane with instantaneous wavelength shift rate as the abscissa and spectral energy broadening as the ordinate; The preset two-dimensional feature interval mapping rules are applied to the two-dimensional feature plane, dividing the two-dimensional feature plane into multiple non-overlapping feature sub-intervals; Assign a unique and independent finite state symbol to each feature sub-interval; According to the chronological order of the time nodes, the characteristic sub-intervals where the instantaneous wavelength shift rate and spectral energy broadening are located at each time node are mapped to the corresponding finite state symbols, and combined to generate the spectral state sequence of the target verification object.

7. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 1, characterized in that, The process of obtaining reference subsequences selected by cross-comparison of genuine samples under multi-substrate background interference by a cloud server, and performing sliding hash matching in the spectral state sequence of the target verification object for initial anti-counterfeiting verification, and triggering a blocking mechanism to determine that the verification has failed if the initial anti-counterfeiting verification fails, includes the following steps: Establish a secure communication connection with the cloud server and download the pre-compiled reference subsequence and the corresponding baseline hash signature; Set a sliding data window with the same length as the reference subsequence, and control the sliding data window to slide and traverse the spectral state sequence of the target verification object in steps of a single finite state symbol; At each position where the sliding data window stops, extract the data fragment within the window and calculate the local hash value. Then compare and match the local hash value with the base hash signature. When the sliding data window completes global traversal and all local hash values ​​fail to match the baseline hash signature, the blocking mechanism is triggered and an alarm signal indicating that the initial anti-counterfeiting verification has failed is output.

8. The anti-counterfeiting verification method based on transparent laser spectral characteristics according to claim 1, characterized in that, When the target verification object passes the initial anti-counterfeiting screening verification, the process of constructing the state transition probability matrix of the transparent laser verification area under local physical wear interference by traversing and statistically analyzing the transition frequencies between all adjacent state symbols in the spectral state sequence includes the following steps: Receive the spectral state sequence that has passed the initial anti-counterfeiting screening and initialize an empty two-dimensional statistical counting matrix; Extract each pair of adjacent preceding and following state symbols from the spectral state sequence; Determine the coordinates of the corresponding elements in the two-dimensional statistical counting matrix based on the types of the preceding and following state symbols, and increment the transition frequency value under the corresponding element coordinates by one. Perform mathematical normalization on each row of the two-dimensional statistical counting matrix after traversal and statistical analysis. The normalized two-dimensional statistical counting matrix is ​​used as the state transition probability matrix of the transparent laser verification area under local physical wear interference.

9. A counterfeit verification system based on transparent laser spectral characteristics, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the anti-counterfeiting verification method based on transparent laser spectral features as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When executed by a processor, this instruction causes the processor to be configured to perform the anti-counterfeiting verification method based on transparent laser spectral features according to any one of claims 1 to 8.

Images