Packaging anti-counterfeiting system and anti-counterfeiting verification method based on frequency domain transformation

Abstract

The present application relates to a packaging anti-counterfeiting system based on frequency domain transformation, comprising a mobile terminal APP, an anti-counterfeiting outer package, a verification server and an anti-counterfeiting database. The present application also relates to an anti-counterfeiting verification method, comprising the steps of: collecting an original image of the anti-counterfeiting outer package; performing preprocessing, geometric correction processing and frequency domain decoding on the original image; comparing with standard information to complete the verification operation; and generating a digital certificate. The present application realizes visual invisibility of anti-counterfeiting information; does not need to change existing process parameters or add special materials; avoids moire and dot clustering problems; significantly improves the stability of hidden code embedding and extraction; adapts to the collection requirements of ordinary smart phones in complex scenes; effectively resists mainstream counterfeiting methods; balances the integrity of traceability and the convenience of verification without relying on a continuous network environment; simplifies the mobile terminal operation complexity, adapts to smart phones with different hardware configurations, balances the robustness of the algorithm and the decoding efficiency; and has the characteristics of resisting local contamination.

Description

Technical Field

[0001] This invention relates to the field of packaging anti-counterfeiting technology, and more specifically to a packaging anti-counterfeiting system and anti-counterfeiting verification method based on frequency domain transformation. Background Technology Background Technology In today's business world, packaging anti-counterfeiting is a core means of protecting brand rights and ensuring user safety. As counterfeiting technology advances, traditional anti-counterfeiting methods are gradually becoming ineffective, and digital anti-counterfeiting technology, which combines concealment, robustness, and convenience, has become the mainstream of development.

[0003] The core of product anti-counterfeiting lies in providing consumers and regulators with verifiable, traceable, and cost-effective means of identifying genuine products. However, with the rapid development of manufacturing and information technology, anti-counterfeiting processes and testing methods that were once considered high-barrier have gradually become more widespread. Counterfeiters can obtain relevant materials and equipment at a lower cost, leading to a continuous decline in the security threshold of many traditional anti-counterfeiting technologies, and their original "technical barriers" and "cost barriers" are being weakened.

[0004] At the same time, while many existing high-strength anti-counterfeiting technologies have good security, they also have problems such as high cost, reliance on special equipment, and complex application and operation. They are difficult to deploy on a large scale in general consumer goods with low average order value and high circulation frequency (such as paper towels, daily chemical products, and fast-moving consumer goods), and cannot meet the application requirements of "verification by everyone, verification at any time, and verification at low cost".

[0005] The existing packaging anti-counterfeiting systems and verification methods include the following mainstream approaches, and their advantages and disadvantages are analyzed below: 1. Traditional physical anti-counterfeiting technology Traditional physical anti-counterfeiting technologies are centered on physical media or special materials, mainly including laser holographic anti-counterfeiting, anti-counterfeiting inks (such as optically variable inks), and anti-counterfeiting paper. Laser holographic anti-counterfeiting uses the principle of laser interference to create three-dimensional markings, which are printed on the packaging surface for visual identification. Optically variable inks utilize photosensitive color-changing properties, displaying preset colors or patterns under specific light sources. They are compatible with common processes such as flexible packaging gravure printing and can be directly printed on packaging logos or key areas. Anti-counterfeiting paper adds special fibers, watermarks, and other physical structures to form anti-counterfeiting features that are visible to the naked eye or require auxiliary tools for identification. This type of technology creates specific optical effects through microstructure and diffraction effects, and initially had high recognizability and a certain technical threshold, making it suitable for products such as alcohol and cosmetics. Advantages: Plate-making and replication technologies have been widely adopted, and there are a large number of high-quality counterfeit holographic labels on the market. Individuals or small factories can achieve high-precision replication, which significantly reduces the anti-counterfeiting capabilities of genuine products, and the cost is still relatively high for low-priced goods; due to the low cost and intuitive recognition, no professional equipment is required; technologies such as optically variable inks can be adapted to existing gravure printing production lines, and the inks have strong adhesion and good storage stability.

[0006] Disadvantages: The anti-counterfeiting features are simple, and the laser holographic mark is easy to copy, making counterfeiting easy; the anti-counterfeiting ink relies on special materials, increasing production costs, and the optical variable effect is easily affected by color sequence and ink viscosity, posing a risk of color bleeding; the anti-counterfeiting information capacity is limited, making it impossible to achieve refined traceability.

[0007] 2. VOID peel-off anti-counterfeiting label technology This type of technology relies on the irreversible mark left after the label is peeled off to indicate whether it has been opened, and is suitable for sealing medicines, logistics seals, etc.

[0008] Disadvantages: This type of technology can only indicate whether the packaging has been opened, but cannot verify the source and authenticity of the product. It is also a one-time use technology with a single anti-counterfeiting dimension, making it difficult to cover the continuous verification needs of a wide range of products.

[0009] 3. Anti-counterfeiting technologies such as thermochromic and fluorescent inks These techniques enable simple verification through temperature changes or ultraviolet light irradiation.

[0010] Disadvantages: Since the relevant inks and processes are readily available, the threshold for counterfeiting has been significantly lowered. Moreover, the performance is greatly affected by the environment and material stability, resulting in limited actual anti-counterfeiting strength, and it tends to be more decorative or interactive.

[0011] 4. Digital watermarking anti-counterfeiting technology Digital watermarking technology achieves covert anti-counterfeiting by embedding anti-counterfeiting information (serial number, batch number, etc.) into the transform or spatial domain of packaging images. Mainstream technologies include algorithms based on Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), and Singular Value Decomposition (SVD). For example, embedding the watermark into low-frequency DCT coefficients of the image balances imperceptibility and robustness, or enhancing resistance to attacks through multi-transform domain fusion. Some solutions combine IoT and blockchain technologies to achieve full lifecycle traceability of anti-counterfeiting information, but the core still relies on watermark embedding and extraction algorithms.

[0012] Advantages: High imperceptibility, does not affect packaging appearance; large information capacity, can carry multi-dimensional data such as serial number and production batch; some algorithms can resist conventional image processing attacks such as compression and filtering.

[0013] Disadvantages: Insufficient robustness; low success rate of watermark extraction when faced with real-world scenarios such as printing distortion, shooting tilt / zoom, and changes in lighting; most detection methods require specialized software or equipment, making them inconvenient for users; lack of integration with printing process characteristics, easily leading to spectral conflicts with dot textures, resulting in watermark distortion or reduced printing quality.

[0014] 5. Halftone mesh anti-counterfeiting technology This technology combines anti-counterfeiting information with core printing processes, achieving information concealment by adjusting the distribution of halftone dots. Mainstream methods include the combined application of amplitude modulation (AM) and frequency modulation (FM) halftone dots. For example, AM halftone dots are used in the background area, while FM halftone dots are used in the warning information area and then superimposed and blended. During copying, the hidden information will be revealed to achieve an anti-copying effect. Alternatively, the dot angle can be adjusted dynamically using variable halftone dots, utilizing the principle of grating interference to reproduce the hidden graphic information, thus balancing the anti-counterfeiting characteristics of both digital processing and physical media.

[0015] Advantages: Deeply integrated into the printing process, requiring no additional materials, and compatible with existing plate-making and printing equipment; significant anti-copying effect, with easily identifiable hidden information in photocopies; combines the advantages of digital and analog anti-counterfeiting, with superior anti-copying capabilities compared to simple digital watermarks.

[0016] Disadvantages: Extraction of hidden information relies on specialized unlocking films or equipment, making it difficult for users to verify authenticity; dot adjustments can easily lead to printing moiré patterns, affecting the visual quality of the packaging; it has weak resistance to geometric transformations, and even slight deformation of the packaging can cause information extraction failure; it lacks a dedicated spectrum area, making it susceptible to interference from parameters such as printing screen angle and screen ruling.

[0017] 6. Spread spectrum modulation anti-counterfeiting technology By generating pseudo-random sequences through spread spectrum and encryption, and employing techniques such as quadrature phase shift modulation (QPSK), the spatial positions of the frequency-modulated dots in the printed material are modulated, causing the dots to redistribute according to rules to carry anti-counterfeiting information. Some schemes embed the spread spectrum sequence into the image transparency channel to enhance the watermark's anti-interference capability and improve the accuracy of authenticity verification.

[0018] Advantages: Strong anti-interference ability, resists attacks such as scanning and photographic copying; anti-counterfeiting information is distributed throughout the entire page, making it resistant to breakage, and information can still be extracted from fragments; the spread spectrum sequence encoding has high complexity, making it difficult for counterfeiters to crack the sequence generation rules.

[0019] Disadvantages: It does not take into account the main lobe characteristics of halftone dots in different printing processes (gravure printing, offset printing, etc.), which can easily lead to spectral conflicts; halftone dot position modulation may cause printing clustering, affecting the appearance of packaging; decoding requires distinguishing between frequency modulation and amplitude modulation halftone dots, which requires high resolution of the acquisition equipment and has poor adaptability to mobile devices.

[0020] 7. Explicit variable code anti-counterfeiting technology Represented by variable QR codes / barcodes and one-item-one-code systems, each package is printed with a unique QR code using inkjet printing technology. Users scan the code to access a server and verify the product information, while also supporting supply chain traceability. This technology relies on a robust database and stable servers, with the core strengths being the uniqueness of the code value and the security of the query system.

[0021] Advantages: Convenient user identification, requiring only mobile phone scanning; enables end-to-end traceability and supports brand-user interaction; strong code uniqueness, providing significant initial anti-counterfeiting effect; low cost; large information capacity; easy to deploy; Disadvantages: QR codes are visible identifiers, disrupting the integrity of packaging design; they are easily copied, photographed, and illegally reused, resulting in weak anti-counterfeiting capabilities; the coding speed (60-80 m / min) is incompatible with high-speed gravure printing production lines (200 m / min), affecting production efficiency; they rely on a network environment and cannot be verified without a network; their security depends entirely on the backend system and management mechanism, and without a supporting dynamic mechanism or higher structural design, static code images alone do not provide reliable anti-counterfeiting capabilities. 8. Radio Frequency Identification (RFID / NFC) Anti-counterfeiting Technology Embedding RFID / NFC chips within packaging stores unique product identification information, and the chip data is read using dedicated reading and writing devices to achieve anti-counterfeiting. This technology allows for identification without visual contact, is suitable for rapid batch verification, and is often used in conjunction with supply chain management systems.

[0022] Advantages: Fast recognition speed, supports batch verification; large information storage capacity, can be dynamically updated; no visual contact required, adaptable to harsh environments.

[0023] Disadvantages: The high cost of chips and reading / writing devices makes them unsuitable for low-cost FMCG packaging; chips are easily disassembled and transferred to counterfeit products, resulting in insufficient anti-counterfeiting reliability; chip embedding is difficult in flexible packaging production and is susceptible to failure due to printing tension and temperature.

[0024] 9. AI physical fingerprint / random texture recognition technology This type of technology uses random textures, creases, etc. on the surface of packaging or labels as "physical fingerprints" and verifies them through image algorithms.

[0025] Disadvantages: Although this type of technology has high theoretical security and anti-copying capabilities, it has high requirements for image acquisition quality, algorithm computing power and system construction, and the implementation cost is relatively high. It is more suitable for specific high-value categories such as jewelry and art.

[0026] In summary, existing technologies still have some common shortcomings, mainly reflected in the following undisclosed and urgently needing solutions: 1. The need to balance visual concealment and printing compatibility: The anti-counterfeiting information must be invisible to the naked eye (ΔEab≤1), without changing the parameters of existing printing processes (such as gravure printing and offset printing), without adding special materials or equipment, and be compatible with high-speed production lines (≥200m / min) to avoid moiré patterns or dot clustering.

[0027] 2. Frequency domain resource collaborative optimization requirements: It is necessary to plan a dedicated security spectrum corridor based on the blue noise distribution characteristics of the printing dots, avoid interference from the main frequency of the CMYK color separation channel, and solve the problem of conflict between anti-counterfeiting signals and printing texture spectrum in the existing technology.

[0028] 3. Robust decoding requirements for mobile devices: It needs to be compatible with ordinary smartphone acquisition scenarios, resist complex conditions such as tilt (±15°), low light (≥200 lux), and zoom (±8%), and achieve quality prediction and dynamic guidance by acquiring metadata (IMU, exposure parameters, etc.), with a decoding success rate of ≥95% and a latency of ≤300ms.

[0029] 4. Strong anti-counterfeiting capability requirement: It is necessary to build a multi-dimensional verification mechanism (such as printing fiber characteristics, display refresh rate detection) to target mainstream counterfeiting methods such as screen copying and photo forgery, with an anti-counterfeiting rate of ≥99% to prevent anti-counterfeiting information from being illegally reused.

[0030] 5. The need for integration of end-to-end traceability and convenient verification: It is necessary to bind metadata such as collected device information, timestamps, and geographical location with anti-counterfeiting information to achieve end-to-end traceability from production to consumption; at the same time, no professional equipment is required, and users can quickly complete the verification and certification through a mobile APP.

[0031] 6. Balancing the need for robustness and decoding efficiency: Techniques such as logarithmic augmentation, cyclic correlation operations, and LDPC error correction coding are needed to simplify computational complexity on mobile devices while ensuring anti-interference capabilities and adapting to smartphones with different hardware configurations. Summary of the Invention

[0032] This invention addresses the aforementioned problems by providing a packaging anti-counterfeiting system and verification method based on frequency domain transformation. Its purpose is to achieve visual invisibility of anti-counterfeiting information; without altering existing gravure or offset printing process parameters or adding special materials; avoid moiré patterns and dot clustering issues; effectively avoid interference from the main frequency of printed textures, resolving the conflict between anti-counterfeiting signals and dot spectrum in existing technologies, and significantly improving the stability of hidden code embedding and extraction; adapt to the acquisition needs of ordinary smartphones in complex scenarios; effectively resist mainstream counterfeiting methods, preventing the illegal reuse of anti-counterfeiting information, and addressing the pain points of easily copied explicit codes and weak anti-attack capabilities of traditional watermarks; achieve end-to-end traceability from production to consumption, balancing traceability integrity and verification convenience without relying on a continuous network environment; simplify mobile terminal computational complexity while ensuring anti-interference capabilities, adapting to smartphones with different hardware configurations, balancing algorithm robustness and decoding efficiency; and possess resistance to localized soiling.

[0033] To solve the above problems, the technical solution provided by the present invention is as follows: A packaging anti-counterfeiting system based on frequency domain transformation includes a mobile APP, anti-counterfeiting outer packaging, a verification server, and an anti-counterfeiting database; the anti-counterfeiting outer packaging includes a CMYK color separation dot structure; the CMYK color separation dot structure is adapted to the dot reference parameters of offset printing, gravure printing, flexographic printing, or digital printing processes; the dot reference parameters include screen ruling, screen angle, dot shape, and dot gain curve; The power spectrum of the CMYK color dot structure exhibits a blue noise distribution and satisfies the following conditions: low-frequency energy is recessed, and mid-to-high-frequency energy is concentrated in a predetermined radius band; a secure spectrum corridor for carrying hidden information is constructed outside the main lobe and harmonics of the predetermined radius band. The secure spectrum corridor is injected with a spread spectrum hidden code; the spread spectrum hidden code is obtained by encoding and spreading modulation of the target anti-counterfeiting information; the injection strength of the spread spectrum hidden code is lower than the human visual invisibility threshold preset by humans; the appearance of the anti-counterfeiting outer packaging is indistinguishable from that of an outer packaging with the same design without the spread spectrum hidden code in human visual perception. The mobile app is installed on a mobile phone, tablet, or laptop; the mobile app is electrically coupled to the camera sensor of the mobile phone, tablet, or laptop. The verification server is used to perform verification operations; The anti-counterfeiting database stores standard information for comparing the bitstream of the target anti-counterfeiting information during the verification operation.

[0034] Preferably, the blue noise distribution is as follows: low-frequency energy ≤ 10% of mid-to-high frequency main energy; wherein the low-frequency energy corresponds to a frequency range of 0 to 0.125 cycles / pixel, and the mid-to-high frequency main energy corresponds to a frequency range of 0.25 to 0.375 cycles / pixel; The predetermined radius band corresponds to the main lobe radius of the dot matrix, using Indicates the radius of the main lobe of the dot matrix. To adapt to printing processes, match the following parameters: Offset printing corresponds to 9-20 pixels, or equivalent to 120-266 LPI; Gravure printing corresponds to 12-20 pixels, or equivalent to 120-200 LPI; Flexographic printing corresponds to 12-24 pixels, or 100-200 LPI; Digital printing corresponds to 10-16 pixels, or 150-240 LPI; The radial range of the secure spectrum corridor is the radius of the main lobe of the dot. The bandwidth is 1.2 to 1.8 times that of the standard bandwidth; the secure spectrum corridor includes 3 to 6 annular sector sub-bands; the angle width of a single annular sector sub-band ranges from 4° to 10°; each annular sector sub-band is arranged in a staggered manner to avoid the main frequency and harmonics of the CMYK color separation channel.

[0035] Preferably, after the target anti-counterfeiting information is converted into a bitstream, the encoding process uses RS encoding, LDPC encoding, or Polar encoding; the code rate range of RS encoding, LDPC encoding, or Polar encoding is [value missing]. This enables error correction; where the symbols The code rate is used to characterize the encoding method; after the encoding is completed, the bit order is shuffled by an interleaver. The bitstream after error correction is spread spectrum modulated using a PN sequence; the length of the PN sequence is no less than 128 bits. The target anti-counterfeiting information, after the spread spectrum modulation, is injected into the security spectrum corridor in an amplitude-weighted manner; the weighting coefficients of the amplitude-weighted method are adaptively calculated by the human eye contrast sensitivity function and the background texture; the range of the injection intensity of the amplitude-weighted method is... ; among which symbols The injection intensity is used to characterize the magnitude weighting method.

[0036] Preferably, when the suitable printing process is offset printing, the CMYK color separation dot structure satisfies the following conditions: The mesh angle is C15° / M75° / Y0° / K45°; The range of the number of network lines is 120 to 266 LPI; The spread spectrum hidden code is preferentially carried in version Y or version K; The dot gain rate in the dot gain curve is no higher than 3%; The overprinting error is no higher than 0.1mm; When the compatible printing process is gravure printing, the CMYK color separation dot structure corresponds to the following cell parameters: The depth of the holes ranges from 10 to 30 μm; The opening range of the mesh is 8-20 μm; The engraving pulse energy ranges from 0.1 to 0.5 mJ; The scanning spacing ranges from 5 to 10 μm; The spread spectrum hidden code is carried in the overprinting area of ​​the C and M plates; The ink used is water-based ink, and its adhesion is not less than 4B.

[0037] An anti-counterfeiting verification method utilizing the aforementioned packaging anti-counterfeiting system includes the following steps: S100: The mobile APP collects the original image of the anti-counterfeiting outer packaging and simultaneously acquires image acquisition metadata at the time of acquisition; the image acquisition metadata is a digital description of the image acquisition behavior, including device hardware parameters, image acquisition parameters, posture and environment parameters, and identification and time parameters; wherein: The device hardware parameters include mobile device model, camera sensor model, lens focal length, camera pixels, and IMU parameters; the IMU includes accelerometer parameters and gyroscope parameters. The image acquisition parameters include exposure value, ISO sensitivity, shutter speed, white balance, focus mode, and image resolution; The attitude and environmental parameters include real-time IMU data, tilt angle at the time of acquisition, and ambient light intensity; the real-time IMU data includes acceleration and angular velocity. The identifier and time parameters include a collection timestamp, a unique device identifier, and geographic location information; wherein, the accuracy of the collection timestamp is milliseconds; S200: The mobile APP performs preprocessing, geometric correction processing, and frequency domain decoding on the original image of the anti-counterfeiting outer packaging collected in sequence; then, based on a manually preset decoding threshold, it extracts the spread spectrum hidden code to obtain the bit stream of the target anti-counterfeiting information. S300: The mobile APP uploads the extracted target anti-counterfeiting information bitstream to the verification server; the verification server compares it with the standard information in the preset anti-counterfeiting database to complete the verification operation; S400: After the verification operation is successful, a digital certificate containing the verification result, timestamp, geographical location information and device unique identifier is generated, and then the digital certificate is sent back to the mobile APP; the mobile APP displays the corresponding anti-counterfeiting verification result based on the digital certificate.

[0038] Preferably, the image acquisition metadata in S100 further includes an image quality score calculated by the mobile app based on the image acquisition metadata, denoted by a symbol. express; It is a 100-point scale, and only when... If the time is right, then proceed with the subsequent S200; otherwise, the user is prompted by the mobile app to adjust the tilt angle at the time of acquisition, or adjust the ambient light intensity, or adjust the acquisition distance, and then S100 is executed again. The dynamic interactive guidance includes arrows, distance circles, and stability indicators; the acquisition distance varies depending on the type of lens used, specifically: When the lens type used for data acquisition is an ultra-wide-angle lens, the acquisition distance ranges from 1 to 4 cm; When the lens type used for data acquisition is a close-up telephoto lens, the acquisition distance ranges from 20 to 30 cm. Preferably, the geometric correction in S200 specifically includes the following steps: S220. Extract candidate regions of the spreading hidden code using a lightweight CNN; S221. Solve the homography matrix based on corner detection or feature point matching to achieve perspective correction and affine transformation; S222. Perform rotation compensation and scaling error compensation; wherein: the rotation compensation range is -15° to +15°; the scaling error compensation range is -8% to +8%; S223. Frequency domain filtering combined with multi-frame HDR synthesis is used to suppress reflective interference and low-light interference.

[0039] Preferably, the frequency domain decoding in S200 specifically includes the following steps: S230. Perform a log-polar transform on the original image after the geometric correction to achieve rotation invariance; S231: Perform a two-dimensional fast Fourier transform on the original image after log-polar transformation to extract the annular sector sub-band corresponding to the safe spectrum corridor; S232: Perform cyclic correlation operation between the PN sequence and the amplitude spectrum of the annular sector sub-band, detect the correlation peak, and obtain the soft decision metric; The length of the PN sequence is 512 bits, and a synchronization flag bit is introduced when the PN sequence is generated; the proportion of the synchronization flag bit in the PN sequence is no more than 5%; the code pattern of the synchronization flag bit is a known fixed bit combination. The primitive polynomial of the PN sequence is: and ; The preset sequence of the PN sequence stored in the mobile app is completely consistent with the spread spectrum sequence used in the pre-press embedding stage, and the sequence phase is pre-calibrated to eliminate synchronization deviation; The cyclic related operations specifically include the following steps: S232.1. Normalize the amplitude spectrum data of each of the said annular sector sub-bands to obtain the sub-band normalized amplitude vector, denoted by the symbol... This indicates that the normalization process specifically involves mapping the amplitude spectrum data to the 0-1 interval. S232.2. Periodically expand the preset PN sequence, and the length of the expanded sequence is proportional to the normalized amplitude vector of the subband. The lengths are consistent, thus obtaining the spread spectrum sequence vector, denoted by the symbol... express; S232.3. Perform the aforementioned loop-related operation, expressed as follows: in: The length of the vector; This is a shift parameter, and its value range is... The cyclic correlation operation is adapted to the spin invariance characteristic after log-polar transformation; The result of the cyclic correlation operation; The detection of relevant peak values ​​specifically includes the following steps: S232.4. Calculate the result of the cyclic correlation operation. The noise variance, denoted by the symbol Indicate; based on The criterion sets an adaptive peak detection threshold, expressed as follows: in: The adaptive peak detection threshold; This represents the mean of the cyclic correlation results; Standard deviation; S232.5. Filter out those values ​​greater than the adaptive peak detection threshold. Local maximum values ​​are used as candidate peak values, and the candidate peak values ​​must satisfy a peak-to-second peak ratio of not less than 3, thereby eliminating false peak values ​​caused by noise interference; S232.6. Based on the candidate peak values ​​detected in the ring sector sub-band, perform the following operations: If multiple candidate peaks are detected in a single annular sector sub-band, the one with the largest amplitude is taken as the effective peak of the annular sector sub-band. If a valid peak value meeting the condition is detected in two or more consecutive annular sector sub-bands, it is determined that the signal of the spreading code exists. Obtaining the soft decision metric specifically includes the following steps: S232.7. For each of the aforementioned annular sector sub-bands, calculate the ratio of the peak amplitude to the noise variance of the effective peak value, expressed by the following formula: in: This is the ratio of peak amplitude to noise variance. The peak amplitude; S232.8. Calculate the soft decision metric, expressed as follows: in: This refers to the soft decision metric; The direction of the spreading symbol corresponding to the peak value; The function is used to determine the positive and negative logic of the spread spectrum symbol; S232.9. Based on whether the effective peak value is detected in the annular sector sub-band, the following operations are performed: If no valid peak is detected in the ring sector sub-band, the soft decision metric of the ring sector sub-band is set to 0 and marked as an invalid sub-band, so that the data of the ring sector sub-band is automatically removed during subsequent decoding. Before the cyclic correlation operation is executed, noise suppression preprocessing is also included, specifically comprising the following: The amplitude spectrum data of the annular sector sub-band is subjected to sliding window mean filtering, with the window size being 3×3 or 5×5, thereby filtering out high-frequency noise; The amplitude spectrum data is weighted and enhanced based on the human eye contrast sensitivity function, thereby increasing the weight of the frequency corresponding to the spread spectrum hidden code signal and suppressing the weight of background texture noise.

[0040] S233: Merge the soft decision metrics of multiple ring sector sub-bands, input them into the corresponding error correction decoder, and then output the bit stream of the target anti-counterfeiting information; the error correction decoder corresponds to the encoding method used in the encoding process; the bit error rate of the bit stream of the target anti-counterfeiting information is represented by the symbol... It indicates that, and satisfies .

[0041] Preferably, S200 further includes anti-replay detection, specifically comprising the following: Check whether the refresh rate of the monitor used in the image acquisition is 50-120Hz; Verify whether the printed microstructure conforms to the pre-defined statistical characteristics; the printed microstructure includes paper fibers and cells; The mobile terminal uses micro-modulation of supplementary light (10-20Hz) to detect whether the reflection response meets the preset reflection response standard, thereby filtering out images that are copied from the screen or forged from photographs; the frequency range of the mobile terminal's supplementary light micro-modulation is 10-20Hz. If any of the above three detection items result in a negative result, the anti-replay detection is determined to be unsuccessful. Then, the user is prompted to re-execute S100 through dynamic interaction via the mobile APP.

[0042] Preferably, in S300, if the mobile APP is in a network-free environment, ECDSA or Ed25519 signature verification is performed based on the locally cached whitelist and revocation list, and a temporary pre-verification result is output. In S400, the digital certificate also includes production batch information and logistics link information corresponding to the anti-counterfeiting outer packaging.

[0043] Preferably, S231 performs the two-dimensional fast Fourier transform, expressed as follows: in: The frequency domain complex matrix is ​​obtained by transforming the image after log-polar transformation through the two-dimensional fast Fourier transform. This is the horizontal frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the horizontal direction of the spatial domain image is as follows: A larger value indicates a higher frequency in the horizontal direction; This is the vertical frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the vertical direction of the spatial domain image is as follows: A larger value indicates a higher frequency in the vertical direction; These are the spatial domain image pixel values ​​after log-polar transformation; The radial coordinates are in the log-polar coordinate system, corresponding to the scaling scale of the spatial domain image, in pixels; The polar angle in the log-polar coordinate system corresponds to the rotation angle of the spatial domain image, in radians; Pi is a constant. The imaginary unit is used to describe phase information in the frequency domain and is adapted to the complex form of the Fourier transform. The horizontal dimension of the zero-padded image, in pixels, with a value of 512 or 1024, used to reduce spectral leakage; The vertical dimension of the zero-padded image, in pixels, is used for isotropic frequency domain transformation, and its value is related to... Consistent; The matching coefficient between the horizontal frequency component and the spatial domain pixel position represents the horizontal frequency. radial coordinates in the spatial domain Contribution weight at each location; The matching coefficient between the vertical frequency component and the spatial domain pixel position represents the vertical frequency. radial coordinates in the spatial domain Contribution weight at each location; This is a double summation symbol, and the range of summation is expressed by the following formula: It is used to perform an accumulation operation on all pixels of the zero-padded image, completing the conversion from the spatial domain to the frequency domain.

[0044] Compared with the prior art, the present invention has the following advantages: 1. Because this invention uses blue noise dot generation technology combined with frequency domain security spectrum corridor planning, the spread spectrum hidden code is injected into the mid-to-high frequency ring area of ​​the printing color separation channel according to the CSF weighting rule. Therefore, the anti-counterfeiting information is visually invisible, and there is no need to change the existing gravure printing, offset printing and other process parameters, no need to add special materials, and it is compatible with high-speed production lines of ≥200m / min. It avoids moiré patterns and dot aggregation problems, and takes into account both concealment and printing compatibility. 2. Because the present invention adopts the radial range of the security spectrum corridor calculated based on the main lobe radius of the printed halftone dots, and the avoidance design that makes the sub-band polar angle deviate from the main frequency of the CMYK color separation channel by ≥10°, it effectively avoids the main frequency interference of the printed texture, solves the problem of anti-counterfeiting signal and halftone spectrum conflict in the prior art, and significantly improves the stability of hidden code embedding and extraction. 3. Because this invention uses log-polar transform to achieve rotation invariance, combines image acquisition metadata for Q-Score quality scoring and dynamic guidance, and extracts feature points through a lightweight CNN to solve the homography matrix for geometric correction, it is suitable for the acquisition needs of ordinary smartphones in complex scenarios such as low light, tilt, and scaling. The decoding success rate is ≥95%, and the latency is ≤300ms, which greatly improves the convenience and robustness of user verification. 4. Because this invention uses a high-complexity PN sequence for spread spectrum modulation to carry anti-counterfeiting information, and combines a multi-dimensional anti-attack mechanism of printing medium microstructure detection and display refresh frequency recognition, it effectively resists mainstream counterfeiting methods such as screen copying and photo forgery, with an anti-counterfeiting rate of ≥99%, avoids the illegal reuse of anti-counterfeiting information, and solves the pain points of explicit codes being easy to copy and traditional watermarks having weak anti-attack capabilities. 5. Because this invention binds product serial numbers, production batches, and metadata such as fingerprints, timestamps, and geographical locations of the collection devices, and combines a dual mechanism of offline pre-verification on mobile devices and verification on cloud servers, it achieves full-chain traceability from production to consumption. Users can quickly complete verification and export encrypted evidence packages through a mobile APP, which balances traceability integrity and verification convenience, without relying on a continuous network environment. 6. Because this invention uses two-dimensional fast Fourier transform, logarithmic enhancement and cyclic correlation operations to improve signal extraction efficiency, and combines LDPC error correction coding with a code rate of 2 / 3 to reduce the bit error rate, it simplifies the computational complexity of mobile terminals while ensuring anti-interference capability, adapts to smartphones with different hardware configurations, and balances algorithm robustness and decoding efficiency. 7. Because the present invention adopts a multi-subband staggered layer distribution and weighted merging strategy, the hidden code information is distributed and carried in multiple frequency domain subbands. Therefore, it has the characteristic of resisting local contamination. Even if some areas of the packaging are damaged, effective information can still be extracted from the remaining subbands, which further improves the reliability of the technical solution in practical applications. Attached Figure Description

[0045] Figure 1a This is a schematic diagram of the frequency domain amplitude spectrum plane according to a specific embodiment of the present invention; Figure 1b This is a three-dimensional schematic diagram of the frequency domain amplitude spectrum of a specific embodiment of the present invention; Figure 2 This is a schematic diagram of the test printing packaging according to a specific embodiment of the present invention; Figure 3 This is a schematic diagram of the anti-counterfeiting verification method according to a specific embodiment of the present invention; Figure 4 This is a schematic diagram illustrating the dynamic guidance of a mobile app during the verification process, according to a specific embodiment of the present invention. Figure 5 This is a schematic diagram showing the prompt after verification by a mobile app in a specific embodiment of the present invention; Figure 6 A screenshot of a digital certificate according to a specific embodiment of the present invention; Figure 7 This is a schematic diagram of the evidence package export interface on a mobile APP according to a specific embodiment of the present invention. Detailed Implementation

[0046] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading the present invention, any modifications of the present invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0047] It should be noted beforehand that this specific embodiment uses a flexible outer packaging printed in a single test in the applicant's laboratory as the application object. The substrate is BOPP, and it is produced using gravure printing technology, specifically laser engraving roller. It needs to achieve the "one item, one code" anti-counterfeiting and traceability function, that is, users can verify authenticity by scanning the LOGO area of ​​the packaging with their mobile phones, distributors can trace the production batch, and regulatory authorities can access the entire chain of data.

[0048] In addition, the packaging design requires no visible markings, such as QR codes; furthermore, the appearance of the embedded code is indistinguishable from ordinary packaging, with a decoding success rate of ≥95%; the environmental requirements are low light and slightly tilted scenes, and the anti-screen copying counterfeiting rate is ≥99%.

[0049] It should be noted that, in order to facilitate understanding of the technical solution of this invention, the applicant will first briefly explain the technical principles, advantages, and comparison with the prior art of this technical solution to facilitate the reader's understanding. A detailed explanation will then follow, using data from a specific experiment as an example: It's important to understand that the principle behind this technical solution is to achieve a highly secure, low-visibility anti-counterfeiting mark by embedding a hidden image in the frequency domain of the image. The core steps are as follows: 1. Fourier transform of the original image: transforms the image from the spatial domain to the frequency domain to obtain its spectral distribution; 2. Hidden Image Frequency Modulation: The hidden image is encoded into specific frequency components (usually mid-to-high frequencies) and embedded into the original image spectrum; 3. Inverse Fourier Transform to Generate Composite Images: By using the inverse Fourier transform, an image containing hidden information but visually indistinguishable is generated; 4. Verification end spectrum extraction: Use a mobile phone or scanning device to perform frequency analysis on the image and extract the hidden image for authenticity verification.

[0050] Because this invention employs a frequency domain transform steganography technique that is not currently used in existing technologies, it possesses the following technical advantages: 1. Concealment: High-frequency information is not sensitive to the human eye, and hiding the image does not affect the visual effect of the original image; 2. Tamper resistance: Frequency domain embedding is inherently difficult to remove through pixel-level editing or compression; 3. Visual verification: Spectral features can be extracted via mobile devices to quickly verify authenticity; 4. Strong image compatibility: It can be embedded in any image, such as artwork, packaging, and identification documents; 5. No internet connection required: The verification process can be completed locally, making it suitable for offline scenarios.

[0051] Based on the above description, the advantages of this invention compared to existing packaging anti-counterfeiting technologies are as follows: 1. Existing steganography techniques are pixel-level steganography, which extracts steganographic information by pixel difference analysis and embeds it in the spatial domain. Therefore, they have many drawbacks such as weak resistance to compression and easy detection. 2. The steganography technique of the present invention is frequency domain steganography. The method of extracting steganographic information is Fourier spectrum extraction. Its embedding domain is the frequency domain. Therefore, it has many advantages such as strong resistance to compression and almost invisible in terms of visibility.

[0052] To help readers better understand the design key points of this technical solution, the key design points are summarized as follows: • Frequency selection: Avoid low-frequency regions that affect the main structure of the image, and prioritize the use of mid-to-high frequency regions.

[0053] • Modulation method: Information is embedded using phase modulation, amplitude modulation, or frequency perturbation.

[0054] • Robust design: Embedded images can employ redundant coding to improve stability against compression and changes in shooting angle.

[0055] Therefore, some typical application scenarios of this invention can be summarized, including but not limited to the following scenarios: • Artwork authentication: Embed the artist's signature or number in the spectrum of the artwork.

[0056] • Packaging anti-counterfeiting: Embed batch information or authenticity codes in the brand image.

[0057] • Document security: Embed identity information in photos to prevent forgery.

[0058] • Blockchain binding: Spectral characteristics can be bound to on-chain data to achieve image immutability.

[0059] Based on the above brief description, the following is a detailed description of the technical solution of the present invention by the applicant: A packaging anti-counterfeiting system based on frequency domain transformation includes a mobile APP, anti-counterfeiting outer packaging, a verification server, and an anti-counterfeiting database; the anti-counterfeiting outer packaging includes a CMYK color separation dot structure; the CMYK color separation dot structure is adapted to the dot reference parameters of offset printing, gravure printing, flexographic printing, or digital printing processes; the dot reference parameters include screen ruling, screen angle, dot shape, and dot gain curve.

[0060] The power spectrum of the CMYK color-separated dot structure exhibits a blue noise distribution and satisfies the following conditions: low-frequency energy is recessed, mid-to-high-frequency energy is concentrated in a predetermined radius band; a secure spectrum corridor for carrying hidden information is constructed outside the main lobe and harmonics of the predetermined radius band.

[0061] In this specific embodiment, this part pertains to the pre-press embedding process in the preparation of anti-counterfeiting outer packaging, specifically frequency domain coordination and spreading code loading; the specific parameters are as follows: Original artwork for printing: CMYK four-color layered file, logo area size 8cm×5cm, resolution 600dpi.

[0062] Gravure printing process baseline parameters: equivalent screen ruling 150 LPI, screen angle C15° / M75° / Y0° / K45°, water-based ink (adhesion 4B), printing speed 200 m / min, tension fluctuation control ±2%; these values ​​will be further mentioned and used in the following documents.

[0063] In this specific embodiment, the distribution of blue noise is as follows: low-frequency energy ≤ 10% of mid-to-high frequency main energy; wherein the low-frequency energy corresponds to a frequency range of 0 to 0.125 cycles / pixel, and the mid-to-high frequency main energy corresponds to a frequency range of 0.25 to 0.375 cycles / pixel.

[0064] In this specific embodiment, the predetermined radius band corresponds to the main lobe radius of the dot matrix, using... Indicates the radius of the main lobe of the dot. To adapt to printing processes, match the following parameters: Offset printing corresponds to 9-20 pixels, or 120-266 LPI.

[0065] Gravure printing corresponds to 12-20 pixels, or equivalent to 120-200 LPI.

[0066] Flexographic printing corresponds to 12-24 pixels, or 100-200 LPI.

[0067] Digital printing corresponds to 10-16 pixels, or 150-240 LPI.

[0068] It should be noted that the values ​​mentioned above in this specific embodiment are calculated based on an image resolution of 2400 dpi.

[0069] In this specific embodiment, the radial range of the secure spectrum corridor is the radius of the main lobe of the dot. The bandwidth is 1.2 to 1.8 times that of the standard; the secure spectrum corridor contains 3 to 6 ring sector sub-bands; the angle width of a single ring sector sub-band ranges from 4° to 10°; Each ring sector sub-band is arranged in a staggered manner to avoid the main frequency and harmonics of the CMYK color separation channel.

[0070] The security spectrum corridor is injected with a spread spectrum hidden code; the spread spectrum hidden code is obtained by encoding and spreading modulation of the target anti-counterfeiting information; the injection strength of the spread spectrum hidden code is lower than the human visual invisibility threshold preset by humans; the appearance of the anti-counterfeiting outer packaging is indistinguishable from the outer packaging with the same design without the spread spectrum hidden code in human visual perception.

[0071] The mobile app is installed on a mobile phone, tablet, or laptop; the mobile app is electrically coupled to the camera sensor of the mobile phone, tablet, or laptop.

[0072] The verification server is used to perform verification operations.

[0073] The anti-counterfeiting database contains standard information about the bitstream used to compare the target anti-counterfeiting information during verification operations.

[0074] In this specific embodiment, after the target anti-counterfeiting information is converted into a bitstream, the encoding process uses RS encoding, LDPC encoding, or Polar encoding; the bitrate range of RS encoding, LDPC encoding, or Polar encoding is [missing information]. This enables error correction; where the symbols The bit rate is used to characterize the encoding method; after encoding is completed, the bits are shuffled by an interleaver.

[0075] In this specific embodiment, the bitstream after error correction is spread spectrum modulated using a PN sequence; the length of the PN sequence is no less than 128 bits.

[0076] In this specific embodiment, the target anti-counterfeiting information after spread spectrum modulation is injected into the security spectrum corridor in an amplitude-weighted manner; the weighting coefficients of the amplitude-weighted method are adaptively calculated by the human eye contrast sensitivity function (CSF) and the background texture; the range of the injection intensity of the amplitude-weighted method is... ; among which symbols Used to characterize the injection intensity in amplitude-weighted methods.

[0077] In this specific embodiment, when the adapted printing process is offset printing, the CMYK color separation dot structure satisfies the following conditions: The mesh angles are C15° / M75° / Y0° / K45°.

[0078] The range of network cable count is 120 to 266 LPI.

[0079] Spread spectrum hidden codes are preferentially carried on the Y or K version.

[0080] The dot gain rate in the dot gain curve should not exceed 3%.

[0081] The overprinting error is no higher than 0.1mm.

[0082] When the compatible printing process is gravure printing, the CMYK color separation dot structure corresponds to the following cell parameters: The depth of the holes ranges from 10 to 30 μm.

[0083] The range of the mesh openings is 8-20 μm.

[0084] The range of engraving pulse energy is 0.1–0.5 mJ.

[0085] The scanning spacing ranges from 5 to 10 μm.

[0086] The spread spectrum hidden code is carried in the overprinted area of ​​the C and M plates.

[0087] The ink used is water-based ink, and its adhesion is not less than 4B.

[0088] In this specific embodiment, the gravure printing process specifically adopts the technical solution of laser engraving roller.

[0089] In this specific embodiment, the blue noise dots are generated by using an error diffusion algorithm combined with a blue noise threshold matrix to quantize the C channel image under the CMYK color separation channel, so that the dot power spectrum exhibits the characteristics of low-frequency energy depression and mid-to-high frequency ring distribution, and the dots do not cluster.

[0090] like Figures 1a-1b As shown, it can be clearly seen that the characteristics of the frequency domain amplitude spectrum are: the center is a low-frequency region, which is a black depression; the middle ring region is a mid-to-high frequency region, which is bright white, corresponding to the characteristics of blue noise. The outer edge of the ring is a high-frequency noise region, which appears gray.

[0091] Main lobe radius of the dot Then via the radial energy curve The calculation yields the following main lobe radius at 150 LPI: The value is 25, which is the radial distance from the center of the frequency domain to the peak of the main lobe.

[0092] The specific parameter values ​​for the secure spectrum corridor are set as follows: Radial range 1.2 × 25 = 30 pixels The resolution is 1.6 × 25 = 40 pixels, divided into 4 annular sector sub-bands, with a single sector angle width of 8° (converted to 1.396 radians). The initial phase angle is... With a polar angle of 0°, the polar angle ranges of the four annular sector sub-bands are [0°, 8°], [270°, 278°], [180°, 188°], and [90°, 98°], respectively, thereby achieving a staggered arrangement of the four annular sector sub-bands.

[0093] Furthermore, the main frequency avoidance parameters are as follows: the polar angle of the main frequency of channel C. The dominant frequency polar angle of the M channel is 15°. The angle is 75°, and the deviation between the sub-band polar angle and the main frequency is not less than 10°, thereby avoiding interference.

[0094] With the above preparations completed, the embedding of the spreading code can then proceed. Specifically: In this specific embodiment, the plaintext of the target anti-counterfeiting information to be embedded is as follows: Product serial number "SN2025080100001" Production batch "B20250801" The two strings above, when converted to a 64-bit bitstream, are represented in hexadecimal as: 0x534E32303235303830313030303031423230323530383031.

[0095] Error correction encoding is performed on the above bitstream; in this specific embodiment, LDPC encoding is used, with a code rate of The 64-bit bitstream is encoded and then extended to 96 bits.

[0096] Then, interleaving is performed. Specifically, the bits are interleaved using an 8×12 matrix to shuffle the bit order (to resist local contamination).

[0097] Next, spread spectrum modulation is performed. Specifically, a PN sequence with a length of 512 bits is used, consisting of two coprime primitive polynomials. and The generated LFSR sequence is XORed to obtain the sequence containing a 5% synchronization flag bit, which is set to 01101001 in this specific embodiment.

[0098] Then frequency domain injection is performed, specifically: injection intensity. The value is set to 0.025, the weighting coefficient. It is calculated using the human eye contrast sensitivity function CSF (5-10 cycles / degree, with a frequency weight of 1.2, and the rest being 0.8).

[0099] The injection formula is expressed as equation (1): (1) Next, we perform the inverse FFT transformation, specifically: Performing an inverse 2D FFT yields an image with the hidden code in the C channel. After CSF mask optimization, it becomes... This makes it invisible to the naked eye.

[0100] The specific parameters for laser engraving roller plate making and printing in this particular embodiment are as follows: Cell parameters: depth 20μm, opening 15μm, engraving pulse energy 0.3mJ, scanning spacing 8μm, no electroplating / chemical etching.

[0101] Color separation output: Outputs a 1-bit TIFF color separation file. The C version carries the spread spectrum hidden code, the M version carries redundant information, and the K version limits perturbation (preserving text sharpness).

[0102] Printing production: water-based ink (VOCs≤40g / L), printing speed 200m / min, registration error ≤0.08mm, dot gain 2.5%.

[0103] like Figure 2 As shown, the finished product's outer packaging is a rectangular test printed package, measuring 20cm × 15cm. The front features the brand logo, while the back displays the ingredient list. The logo area measures 8cm × 5cm and contains a hidden spread spectrum code. The appearance is consistent with the packaging without the hidden code, with no visible particles or stripes.

[0104] An anti-counterfeiting verification method utilizing the aforementioned packaging anti-counterfeiting system; firstly, the specific equipment and parameters used in this specific embodiment are as follows: Data acquisition device: Huawei Mate60 Pro, which is equipped with an IMX800 camera sensor with the following parameters: 24mm focal length and 50 megapixels.

[0105] Data collection scenario: Simulating a low-light environment on a supermarket shelf, with a specific light intensity of 250 lux, a shooting distance of 30cm, and a phone tilt angle of 7°.

[0106] Image acquisition metadata: ISO 400, shutter speed 1 / 60s, exposure value The IMU's angular velocity is 0.8 rad / s, the device fingerprint (i.e., the device's unique identifier) ​​is the string "868123060000123", and the collection timestamp is "2025-08-10 15:30:22.123".

[0107] like Figure 3 As shown, the anti-counterfeiting verification method includes the following steps: S100: The mobile app captures the original image of the anti-counterfeiting outer packaging and simultaneously obtains image acquisition metadata at the time of acquisition; the image acquisition metadata is a digital description of the image acquisition behavior, including device hardware parameters, image acquisition parameters, posture and environmental parameters, and identification and time parameters; among which: The device hardware parameters include the mobile device model, camera sensor model, lens focal length, camera pixels, and IMU parameters; the IMU includes accelerometer parameters and gyroscope parameters.

[0108] Image acquisition parameters include exposure value, ISO sensitivity, shutter speed, white balance, focus mode, and image resolution.

[0109] Attitude and environmental parameters include IMU real-time data, tilt angle at the time of acquisition, and ambient light intensity; IMU real-time data includes acceleration and angular velocity.

[0110] The identification and time parameters include the collection timestamp, the device's unique identifier, and geographical location information; among them, the accuracy of the collection timestamp is in milliseconds.

[0111] In this specific embodiment, the image acquisition metadata in S100 also includes an image quality score calculated by the mobile APP based on the image acquisition metadata, denoted by symbols. express; It is a 100-point scale, and only when... Only when the time is right will the subsequent S200 be executed; otherwise, the user will be prompted by the mobile APP to adjust the tilt angle at the time of acquisition, or adjust the ambient light intensity, or adjust the acquisition distance, and then S100 will be executed again.

[0112] Dynamic interactive guidance includes arrows, distance circles, and stabilization indicators; the acquisition distance varies depending on the type of lens used, specifically: When using an ultra-wide-angle lens for data acquisition, the acquisition distance ranges from 1 to 4 cm. When the lens used for data acquisition is a lens that can capture both close-up and telephoto shots, the acquisition distance range is 20–30cm. In this specific embodiment, the applicant's experiments show that the actual sampling distance range of 1 to 40 cm can achieve good results, which can well cover the sampling needs under different lens configurations.

[0113] In this specific embodiment, quality scoring The score calculated based on metadata is 68 (≥55). The mobile app displays dynamic guidance "Stay stable, recognizing", and there is no need to retake the photo.

[0114] like Figure 4 As shown, the top of the mobile app features a camera viewfinder displaying the logo area, the middle displays dynamic guidance text "Stay stable, recognizing," and the bottom shows a quality score of 68 points and a "Verifying" progress bar. like Figure 5 As shown, after verification, a "Genuine Product" label and a digital certificate entry will be displayed; S200: The mobile APP performs preprocessing, geometric correction, and frequency domain decoding on the original image of the collected anti-counterfeiting outer packaging in sequence; then, based on the manually preset decoding threshold, it completes the extraction of the spread spectrum hidden code to obtain the bit stream of the target anti-counterfeiting information.

[0115] In this specific embodiment, the preprocessing uses 5-frame HDR synthesis to suppress reflections and 3×3 sliding window mean filtering to reduce noise.

[0116] In this specific embodiment, the geometric correction process in S200 specifically includes the following steps: S220. Extract candidate regions for spread spectrum hidden codes using a lightweight CNN.

[0117] S221. Based on corner detection or feature point matching, solve the homography matrix to achieve perspective correction and affine transformation.

[0118] S222. Perform rotation compensation and scaling error compensation; wherein: the compensation range for rotation compensation is -15° to +15°; the compensation range for scaling error compensation is -8% to +8%.

[0119] S223. Frequency domain filtering combined with multi-frame HDR synthesis is used to suppress reflective interference and low-light interference.

[0120] In this specific embodiment, the geometric correction process extracts candidate points of the LOGO region using a lightweight CNN, selecting a total of 16 feature points. The homography matrix is ​​then solved using equation (2): (2) In this specific embodiment, the actual measurement , The 7° tilt error was compensated to obtain a corrected image with a size of 512×512 pixels.

[0121] In this specific embodiment, the frequency domain decoding in S200 specifically includes the following steps: S230. Perform a log-polar transform on the original image after geometric correction to achieve rotation invariance.

[0122] In this specific embodiment, the corrected image is converted to the log-polar domain with a radial scaling factor of 0.8 and an angular step of 0.01 rad to achieve rotation invariance and adapt to ±8% scaling error.

[0123] S231: Perform a two-dimensional fast Fourier transform on the original image after log-polar transformation to extract the annular sector sub-band corresponding to the safe spectrum corridor.

[0124] In this specific embodiment, when extracting the annular sector sub-band corresponding to the secure spectrum corridor, a 2DFFT is performed on the log-polar transform image with the same parameters as the preprint, a size of 512×512, and log-enhanced after centering.

[0125] Next is the polar coordinate transformation. According to the conversion formula between polar coordinates and rectangular coordinates, the transformed polar coordinates can be easily obtained, expressed as in equation (3): (3) It should be noted that 256 here refers to the center coordinates in the frequency domain.

[0126] At this point, the rectangular coordinates have been transformed. Changed to polar coordinates .

[0127] The next step is to extract the ring sector sub-bands corresponding to the secure spectrum corridor. The specific parameters are as follows: Pixels Four annular sector sub-bands form four amplitude vectors. , , , Each vector has a length of 32.

[0128] In this specific embodiment, S231 performs a two-dimensional fast Fourier transform, expressed as in equation (4): (4) in: This is the frequency domain complex matrix obtained by transforming the image after log-polar transformation into a two-dimensional fast Fourier transform.

[0129] It should be noted that, The calculations describe the core data of the image frequency distribution, where each element contains the amplitude and phase information of the corresponding frequency component.

[0130] This is the horizontal frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the horizontal direction of the spatial domain image is as follows: A higher value indicates a higher frequency in the horizontal direction.

[0131] This is the vertical frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the vertical direction of the spatial domain image is as follows: A larger value indicates a higher frequency in the vertical direction.

[0132] It should be noted that a centralized phase adjustment factor also needs to be defined here, written as Obviously, when If the value is even, the value of the centralized phase adjustment factor is 1, while when... If the value is odd, the value of the centralized phase adjustment factor is -1. In this way, by flipping the frequency domain matrix through the centralized phase adjustment factor, the low-frequency components can be concentrated in the center of the frequency domain image, which corresponds to the smooth features such as the overall brightness and large areas of color blocks of the image, while the high-frequency components can be distributed at the edges, which corresponds to the abrupt features such as the edges, details, and halftone textures of the image.

[0133] It should be further noted that two amplitude spectra need to be calculated: The frequency domain complex matrix after two-dimensional fast Fourier transform The amplitude spectrum, written as This is a unitless value that only represents relative intensity and is used to reflect the energy level of the corresponding frequency component. It is the core carrier of the hidden code signal.

[0134] The amplitude spectrum after logarithmic enhancement is written as This is also a unitless value, which enhances the visibility of low-energy components and highlights the spread spectrum hidden code signal concealed in the background noise.

[0135] These are the spatial domain image pixel values ​​after log-polar transformation.

[0136] The radial coordinates are in the log-polar coordinate system, corresponding to the scaling scale of the spatial domain image, in pixels.

[0137] The polar angle in the log-polar coordinate system corresponds to the rotation angle of the spatial domain image, in radians.

[0138] Pi is a constant. The imaginary unit is used to describe phase information in the frequency domain and is adapted to the complex form of the Fourier transform.

[0139] The horizontal dimension of the image after zero-padding, in pixels, with a value of 512 or 1024, is used to reduce spectral leakage.

[0140] The vertical dimension of the zero-padded image, in pixels, is used for isotropic frequency domain transformation, and its value is related to... Consistent.

[0141] The matching coefficient between the horizontal frequency component and the spatial domain pixel position represents the horizontal frequency. radial coordinates in the spatial domain The contribution weight of each location.

[0142] The matching coefficient between the vertical frequency component and the spatial domain pixel position represents the vertical frequency. radial coordinates in the spatial domain The contribution weight of each location.

[0143] The summation symbol is a double summation symbol, and the range of summation is expressed by equation (5): (5) It is used to perform an accumulation operation on all pixels of the zero-padded image, completing the conversion from the spatial domain to the frequency domain.

[0144] S232: Perform cyclic correlation operation between the PN sequence and the amplitude spectrum of the ring sector sub-band, detect the correlation peak, and obtain the soft decision metric.

[0145] The length of the PN sequence is 512 bits, and a synchronization flag bit is introduced when the PN sequence is generated; the proportion of the synchronization flag bit in the PN sequence is no more than 5%; the code pattern of the synchronization flag bit is a known fixed bit combination.

[0146] The primitive polynomial of the PN sequence is and ; The preset sequence of the PN sequence stored in the mobile app is completely consistent with the spread spectrum sequence used in the pre-press embedding stage, and the sequence phase is pre-calibrated to eliminate synchronization deviation; Loop-related operations specifically include the following steps: S232.1. Normalize the amplitude spectrum data of each ring sector sub-band to obtain the sub-band normalized amplitude vector, denoted by the symbol... This indicates that normalization specifically involves mapping the amplitude spectrum data to the 0-1 interval.

[0147] S232.2. Periodically extend the preset PN sequence. The length of the extended sequence is related to the normalized amplitude vector of the subband. The lengths are consistent, thus obtaining the spread spectrum sequence vector, denoted by the symbol... express.

[0148] It should be noted that the PN sequence stored in the mobile app is the same as that in the pre-press version, with a length of 256 bits and a period extended to 32 bits to obtain the spread spectrum sequence vector. .

[0149] In this specific embodiment, the normalization process specifically involves... Mapped to the 0-1 interval, the normalized vector's specific value is .

[0150] S232.3. Perform loop-related operations, expressed as in equation (6): (6) in: The length of the vector; This is a shift parameter, and its value range is... Cyclic correlation operations are adapted to the spin invariance characteristic after log-polar transformation; This is the result of a loop-related operation.

[0151] The detection of relevant peak values ​​specifically includes the following steps: S232.4. Calculate the results of cyclic correlation operations. The noise variance, denoted by the symbol Indicate; based on The criterion sets an adaptive peak detection threshold, expressed as in equation (7): (7) in: An adaptive peak detection threshold; This represents the mean of the cyclic correlation results; The standard deviation is denoted as .

[0152] In this specific embodiment, the parameter values ​​for the cyclic correlation operation are as follows: Thus, the calculation is obtained The mean is Standard deviation .

[0153] S232.5. Filter out those exceeding the adaptive peak detection threshold. Local maxima are used as candidate peak values, and the candidate peak values ​​must satisfy the peak-to-second peak ratio of not less than 3, thereby eliminating false peak values ​​caused by noise interference.

[0154] S232.6. Based on the candidate peak values ​​detected in the ring sector sub-band, perform the following operations: If multiple candidate peaks are detected in a single ring sector sub-band, the one with the largest amplitude is taken as the valid peak of this ring sector sub-band.

[0155] If valid peak values ​​that meet the conditions are detected in two or more consecutive ring sector sub-bands, then the signal is determined to be a spread spectrum hidden code.

[0156] In this specific embodiment, the adaptive peak detection threshold Filter out hour A peak value greater than the adaptive peak detection threshold and with a peak-to-second peak ratio of 3.5 (≥3) is considered a valid peak value.

[0157] Obtaining the soft decision metric specifically involves the following steps: S232.7. For each ring sector sub-band, calculate the ratio of the effective peak amplitude to the noise variance, expressed as in equation (8): (8) in: This is the ratio of peak amplitude to noise variance. This represents the peak amplitude.

[0158] S232.8. Calculate the soft decision metric, expressed as in equation (9): (9) in: For soft decision metrics; The direction of the spreading symbol corresponding to the peak value; The function is used to determine the positive and negative logic of the spread spectrum symbol.

[0159] In this specific implementation, the soft decision metric is calculated. The sign is positive, corresponding to bit "1".

[0160] Then repeat the above operation for the remaining three sub-bands to obtain the soft decision metrics. , , .

[0161] S232.9. Based on whether a valid peak value is detected in the ring sector sub-band, perform the following operations: If no valid peak is detected in the ring sector subband, the soft decision metric of this ring sector subband is set to 0 and marked as an invalid subband, so that the data of this ring sector subband will be automatically removed during subsequent decoding.

[0162] Before the execution of loop-related operations, noise suppression preprocessing is also included, which specifically includes the following: The amplitude spectrum data of the ring sector sub-band is subjected to sliding window mean filtering, with the window size being 3×3 or 5×5, thereby filtering out high-frequency noise.

[0163] The amplitude spectrum data is weighted and enhanced based on the human eye contrast sensitivity function, thereby increasing the weight of the frequency corresponding to the spread spectrum hidden code signal and suppressing the weight of background texture noise.

[0164] S233: The soft-decision metric of multiple ring sector subbands is merged, input into the corresponding error correction decoder, and then output as the bit stream of the target anti-counterfeiting information; the error correction decoder corresponds to the encoding method used in the encoding process; the bit error rate of the target anti-counterfeiting information bit stream is represented by the symbol... It indicates that, and satisfies .

[0165] In this specific embodiment, since LDPC is used for encoding, LDPC is also used for decoding. The input decoder outputs a 64-bit bitstream that matches the product serial number and batch number embedded in the pre-press assembly. The bit error rate is... (≤5×10) -4 ).

[0166] In this specific embodiment, the soft decision metrics of multiple ring sector sub-bands are merged. A weighted merging of four soft decision metrics is adopted. Since the four soft decision metrics have the same status, the weight of each one is 0.25, resulting in a merged metric of 50.4.

[0167] In this specific embodiment, S200 further includes anti-replay detection, which specifically includes the following: Check whether the refresh rate of the monitor used to capture the image is 50-120Hz.

[0168] Verify whether the printed microstructure conforms to the statistical characteristics preset by humans; the printed microstructure includes paper fibers and cells.

[0169] The mobile device uses micro-modulation of supplementary light (10-20Hz) to detect whether the reflection response meets the preset reflection response standard, thereby filtering out images that are copied from the screen or forged from photos; the frequency range of the mobile device's supplementary light micro-modulation is 10-20Hz.

[0170] If any of the above three tests result in a negative result, the anti-replay test is deemed to have failed. Then, the user is prompted to re-execute S100 via dynamic interaction through the mobile APP.

[0171] In this specific embodiment, the anti-replay detection results are as follows: Display refresh rate detection: If the captured image does not have 60Hz / 120Hz stripe features, it is determined that it is not a screen copy.

[0172] Microstructure verification: Fiber characteristics of the packaging paper were extracted, with an average fiber width of 15 μm and a distribution density of 20 fibers / mm. 2 The discrepancy between the discrepancy and the planar features of the screen image indicates that it is a genuine printed product.

[0173] S300: The mobile APP uploads the extracted target anti-counterfeiting information bitstream to the verification server; the verification server compares it with the standard information in the preset anti-counterfeiting database to complete the verification operation.

[0174] In this specific embodiment, in S300, if the mobile APP is in a network-free environment, ECDSA or Ed25519 signature verification is performed based on the locally cached whitelist and revocation list, and a temporary pre-verification result is output.

[0175] S400: After the verification operation is successful, a digital certificate containing the verification result, timestamp, geographical location information and device unique identifier is generated, and then the digital certificate is sent back to the mobile APP; the mobile APP displays the corresponding anti-counterfeiting verification result based on the digital certificate.

[0176] In this specific embodiment, in S400, the digital certificate also includes production batch information and logistics link information corresponding to the anti-counterfeiting outer packaging; in addition, the mobile APP can further export an evidence package containing original image fragments and verification information.

[0177] In this specific embodiment, the server verification and digital certificate issuance results are as follows: First, the data upload process is as follows: the mobile APP uploads the 64-bit bitstream, device fingerprint, timestamp, and geographical location (latitude 39.9°, longitude 116.4°) to the Alibaba Cloud server.

[0178] Then comes the offline pre-verification, which is as follows: the product batch is included in the local cache whitelist of the mobile APP, the Ed25519 signature verification is executed, and the temporary pre-verification result "preliminary verification passed" is output.

[0179] Next is cloud verification, specifically: the server compares the database, stores 10,000 serial numbers for this batch, confirms that "SN2025080100001" is genuine, and associates it with production data: production date 2025-08-01, production workshop number 3, logistics link: factory → Beijing warehouse → Chaoyang supermarket; all of the above information is preset manually as experimental material.

[0180] Finally, digital certificate generation involves the server issuing a digital certificate containing verification results (genuine product), timestamp, geolocation, device fingerprint, and logistics information.

[0181] like Figure 6 As shown, the digital certificate is a rectangular electronic document with the brand logo at the top, product information in the middle including serial number, batch number, and production date, and a green "Genuine Product" label and certificate number "Cert20250810153022123" at the bottom, which is then sent back to the mobile device.

[0182] like Figure 7 As shown, the evidence package export process is as follows: After the verification process is completed and a digital certificate is generated, the user can further click "Export Evidence Package" to obtain an encrypted file containing the original image fragments, signature information, and timestamps.

[0183] This concludes one experimental procedure of the present invention. Based on the above description, the technical effects of the present invention can be verified as follows: Appearance inspection: In a blind test with 100 people, no one could distinguish between packaging with and without hidden codes. (DeltaE^*ab=0.8) (≤1, meeting the requirement of being visually invisible).

[0184] Decoding success rate: Under conditions of light intensity of 200-1000 lux, tilt angle of 0° to 15°, and shooting distance of 20 to 40cm, the decoding success rate was 96.8% after 1000 tests.

[0185] Anti-forgery capability: It has a 99.5% detection rate against 200 attacks involving screen capture and photo forgery, with no false positives.

[0186] Decoding latency: The average latency from data acquisition to result display on the mobile device is 280ms (≤300ms).

[0187] Industrial compatibility: It can be directly deployed on existing gravure printing production lines without the need for additional equipment, and its production efficiency is consistent with that of ordinary packaging (200m / min).

[0188] Finally, to further demonstrate the technical effects of the present invention and its advantages over the prior art, the applicant compares the advantages and technical support of the present invention with the disadvantages of the prior art, as shown in Table 1: Table 1. Comparison of the technical advantages of this invention and existing technologies In the above detailed description, various features are combined together in a single embodiment to simplify this disclosure. This approach to disclosure should not be construed as reflecting an intention that embodiments of the claimed subject matter require more features than are explicitly stated in each claim. Rather, as reflected in the appended claims, the invention is presented with fewer features than all of the features of the single disclosed embodiment. Therefore, the appended claims are hereby explicitly incorporated into the detailed description, wherein each claim stands alone as a preferred embodiment of the invention.

[0189] The disclosed embodiments have been described above to enable any person skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the spirit and scope of this disclosure. Therefore, this disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0190] The foregoing description includes examples of one or more embodiments. It is certainly impossible to describe all possible combinations of components or methods in order to describe the above embodiments, but those skilled in the art will recognize that further combinations and arrangements of the various embodiments are possible. Therefore, the embodiments described herein are intended to cover all such changes, modifications, and variations that fall within the scope of the appended claims. Furthermore, the term "comprising" as used in the specification or claims is interpreted in a manner similar to the term "including," as interpreted when used as a conjunction in the claims. Additionally, the use of any term "or" in the specification of the claims is intended to mean "non-exclusive or."

[0191] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A packaging anti-counterfeiting system based on frequency domain transformation, characterized in that: It includes a mobile app, anti-counterfeiting outer packaging, verification server, and anti-counterfeiting database; the anti-counterfeiting outer packaging includes a CMYK color separation dot structure; the CMYK color separation dot structure is adapted to the dot reference parameters of offset printing, gravure printing, flexographic printing, or digital printing processes; the dot reference parameters include screen ruling, screen angle, dot shape, and dot gain curve. The power spectrum of the CMYK color dot structure exhibits a blue noise distribution and satisfies the following conditions: low-frequency energy is recessed, and mid-to-high-frequency energy is concentrated in a predetermined radius band; a secure spectrum corridor for carrying hidden information is constructed outside the main lobe and harmonics of the predetermined radius band. The secure spectrum corridor is injected with a spreading code; The spread spectrum hidden code is obtained by encoding and spreading modulation of the target anti-counterfeiting information; The injection strength of the spread spectrum code is lower than the human visual invisibility threshold preset by humans; the appearance of the anti-counterfeiting outer packaging is indistinguishable from that of the same design outer packaging without the spread spectrum code injected by humans. The mobile app is installed on a mobile phone, tablet, or laptop; the mobile app is electrically coupled to the camera sensor of the mobile phone, tablet, or laptop. The verification server is used to perform verification operations; The anti-counterfeiting database stores standard information for comparing the target anti-counterfeiting information in the verification operation; The blue noise distribution is as follows: low-frequency energy ≤ 10% of mid-to-high frequency main energy; wherein the low-frequency energy corresponds to a frequency range of 0 to 0.125 cycles / pixel, and the mid-to-high frequency main energy corresponds to a frequency range of 0.25 to 0.375 cycles / pixel; The predetermined radius band corresponds to the main lobe radius of the dot matrix, using Indicates the radius of the main lobe of the dot matrix. To adapt to printing processes, match the following parameters: Offset printing corresponds to 9-20 pixels, or equivalent to 120-266 LPI; Gravure printing corresponds to 12-20 pixels, or equivalent to 120-200 LPI; Flexographic printing corresponds to 12-24 pixels, or 100-200 LPI; Digital printing corresponds to 10-16 pixels, or 150-240 LPI; The radial range of the secure spectrum corridor is the radius of the main lobe of the dot. The bandwidth is 1.2 to 1.8 times that of the standard bandwidth; the secure spectrum corridor includes 3 to 6 annular sector sub-bands; the angle width of a single annular sector sub-band ranges from 4° to 10°; each annular sector sub-band is arranged in a staggered manner to avoid the main frequency and harmonics of the CMYK color separation channel.

2. The packaging anti-counterfeiting system based on frequency domain transformation according to claim 1, characterized in that: After the target anti-counterfeiting information is converted into a bitstream, the encoding process uses RS encoding, LDPC encoding, or Polar encoding; the code rate range for RS encoding, LDPC encoding, or Polar encoding is [value missing]. This enables error correction; where the symbols The code rate is used to characterize the encoding method; after the encoding is completed, the bit order is shuffled by an interleaver. The bitstream after error correction is spread spectrum modulated using a PN sequence; the length of the PN sequence is no less than 128 bits. The target anti-counterfeiting information, after the spread spectrum modulation, is injected into the security spectrum corridor in an amplitude-weighted manner; the weighting coefficients of the amplitude-weighted method are adaptively calculated by the human eye contrast sensitivity function and the background texture; the range of the injection intensity of the amplitude-weighted method is... ; among which symbols The injection intensity is used to characterize the magnitude weighting method.

3. The packaging anti-counterfeiting system based on frequency domain transformation according to claim 2, characterized in that: When the compatible printing process is offset printing, the CMYK color separation dot structure satisfies the following conditions: The mesh angle is C15° / M75° / Y0° / K45°; The range of the number of network lines is 120 to 266 LPI; The spread spectrum hidden code is preferentially carried in version Y or version K; The dot gain rate in the dot gain curve is no higher than 3%; The overprinting error is no higher than 0.1mm; When the compatible printing process is gravure printing, the CMYK color separation dot structure corresponds to the following cell parameters: The depth of the holes ranges from 10 to 30 μm; The opening range of the mesh is 8-20 μm; The engraving pulse energy ranges from 0.1 to 0.5 mJ; The scanning spacing ranges from 5 to 10 μm; The spread spectrum hidden code is carried in the overprinting area of ​​the C and M plates; The ink used is water-based ink, and its adhesion is not less than 4B.

4. A method for verifying anti-counterfeiting using the packaging anti-counterfeiting system described in claim 3, characterized in that: Includes the following steps: S100: The mobile APP collects the original image of the anti-counterfeiting outer packaging and simultaneously acquires image acquisition metadata at the time of acquisition; the image acquisition metadata is a digital description of the image acquisition behavior, including device hardware parameters, image acquisition parameters, posture and environment parameters, and identification and time parameters; wherein: The device hardware parameters include mobile device model, camera sensor model, lens focal length, camera pixels, and IMU parameters; the IMU includes accelerometer parameters and gyroscope parameters. The image acquisition parameters include exposure value, ISO sensitivity, shutter speed, white balance, focus mode, and image resolution; The attitude and environmental parameters include real-time IMU data, tilt angle at the time of acquisition, and ambient light intensity; the real-time IMU data includes acceleration and angular velocity. The identifier and time parameters include a collection timestamp, a unique device identifier, and geographic location information; wherein, the accuracy of the collection timestamp is milliseconds; S200: The mobile APP performs preprocessing, geometric correction processing, and frequency domain decoding on the original image of the anti-counterfeiting outer packaging collected in sequence; then, based on a manually preset decoding threshold, it extracts the spread spectrum hidden code to obtain the bit stream of the target anti-counterfeiting information. S300: The mobile APP uploads the extracted target anti-counterfeiting information bitstream to the verification server; the verification server compares it with the standard information in the preset anti-counterfeiting database to complete the verification operation; S400: After the verification operation is successful, a digital certificate containing the verification result, timestamp, geographical location information and device unique identifier is generated, and then the digital certificate is sent back to the mobile APP; the mobile APP displays the corresponding anti-counterfeiting verification result based on the digital certificate.

5. The anti-counterfeiting verification method according to claim 4, characterized in that: The image acquisition metadata in S100 also includes an image quality score calculated by the mobile app based on the image acquisition metadata, denoted by symbols. express; It is a 100-point scale, and only when... Only then will the subsequent S200 be executed; Otherwise, the user is prompted by the mobile APP to adjust the tilt angle at the time of collection, or to adjust the ambient light intensity, or to adjust the collection distance, and then S100 is executed again. The dynamic interactive guidance includes arrows, distance circles, and stability indicators; the acquisition distance varies depending on the type of lens used, specifically: When the lens type used for data acquisition is an ultra-wide-angle lens, the acquisition distance ranges from 1 to 4 cm; When the lens type used for data acquisition is a close-up to telephoto lens, the acquisition distance ranges from 20 to 30 cm.

6. The anti-counterfeiting verification method according to claim 5, characterized in that: The geometric correction in S200 specifically includes the following steps: S220. Extract candidate regions of the spreading hidden code using a lightweight CNN; S221. Solve the homography matrix based on corner detection or feature point matching to achieve perspective correction and affine transformation; S222. Perform rotation compensation and scaling error compensation; wherein: the rotation compensation range is -15° to +15°; the scaling error compensation range is -8% to +8%; S223. Frequency domain filtering combined with multi-frame HDR synthesis is used to suppress reflective interference and low-light interference.

7. The anti-counterfeiting verification method according to claim 6, characterized in that: The frequency domain decoding in S200 specifically includes the following steps: S230. Perform a log-polar transform on the original image after the geometric correction to achieve rotation invariance; S231: Perform a two-dimensional fast Fourier transform on the original image after log-polar transformation to extract the annular sector sub-band corresponding to the safe spectrum corridor; S232: Perform cyclic correlation operation between the PN sequence and the amplitude spectrum of the annular sector sub-band, detect the correlation peak, and obtain the soft decision metric; The length of the PN sequence is 512 bits, and a synchronization flag bit is introduced when the PN sequence is generated; the proportion of the synchronization flag bit in the PN sequence is no more than 5%; the code pattern of the synchronization flag bit is a known fixed bit combination. The primitive polynomial of the PN sequence is: and ; The preset sequence of the PN sequence stored in the mobile app is completely consistent with the spread spectrum sequence used in the pre-press embedding stage, and the sequence phase is pre-calibrated to eliminate synchronization deviation; The cyclic related operations specifically include the following steps: S232.

1. Normalize the amplitude spectrum data of each of the said annular sector sub-bands to obtain the sub-band normalized amplitude vector, denoted by the symbol... This indicates that the normalization process specifically involves mapping the amplitude spectrum data to the 0-1 interval. S232.

2. Periodically expand the preset PN sequence, and the length of the expanded sequence is proportional to the normalized amplitude vector of the subband. The lengths are consistent, thus obtaining the spread spectrum sequence vector, denoted by the symbol... express; S232.

3. Perform the aforementioned loop-related operation, expressed as follows: in: The length of the vector; This is a shift parameter, and its value range is... The cyclic correlation operation is adapted to the spin invariance characteristic after log-polar transformation; The result of the cyclic correlation operation; The detection of relevant peak values ​​specifically includes the following steps: S232.

4. Calculate the result of the cyclic correlation operation. The noise variance, denoted by the symbol Indicate; based on The criterion sets an adaptive peak detection threshold, expressed as follows: in: The adaptive peak detection threshold; This represents the mean of the cyclic correlation results; Standard deviation; S232.

5. Filter out those values ​​greater than the adaptive peak detection threshold. Local maximum values ​​are used as candidate peak values, and the candidate peak values ​​must satisfy a peak-to-second peak ratio of not less than 3, thereby eliminating false peak values ​​caused by noise interference; S232.

6. Based on the candidate peak values ​​detected in the ring sector sub-band, perform the following operations: If multiple candidate peaks are detected in a single annular sector sub-band, the one with the largest amplitude is taken as the effective peak of the annular sector sub-band. If a valid peak value meeting the condition is detected in two or more consecutive annular sector sub-bands, it is determined that the signal of the spreading code exists. Obtaining the soft decision metric specifically includes the following steps: S232.

7. For each of the aforementioned annular sector sub-bands, calculate the ratio of the peak amplitude to the noise variance of the effective peak value, expressed by the following formula: in: It is the ratio of peak amplitude to noise variance; The peak amplitude; S232.

8. Calculate the soft decision metric, expressed as follows: in: For the soft decision metric; The direction of the spreading symbol corresponding to the peak value; The function is used to determine the positive and negative logic of the spread spectrum symbol; S232.

9. Based on whether the effective peak value is detected in the annular sector sub-band, the following operations are performed: If no valid peak is detected in the ring sector sub-band, the soft decision metric of the ring sector sub-band is set to 0 and marked as an invalid sub-band, so that the data of the ring sector sub-band is automatically removed during subsequent decoding. Before the cyclic correlation operation is executed, noise suppression preprocessing is also included, specifically comprising the following: The amplitude spectrum data of the annular sector sub-band is subjected to sliding window mean filtering, with the window size being 3×3 or 5×5, thereby filtering out high-frequency noise; The amplitude spectrum data is weighted and enhanced based on the human eye contrast sensitivity function, thereby increasing the weight of the frequency corresponding to the spread spectrum hidden code signal and suppressing the weight of background texture noise. S233. Merge the soft decision metrics of multiple ring sector sub-bands, input them into the corresponding error correction decoder, and then output the bit stream of the target anti-counterfeiting information; the error correction decoder corresponds to the encoding method used in the encoding process; the bit error rate of the bit stream of the target anti-counterfeiting information is represented by the symbol It indicates that, and satisfies .

8. The anti-counterfeiting verification method according to claim 7, characterized in that: The S200 also includes replay protection detection, which specifically includes the following: Check whether the refresh rate of the monitor used in the image acquisition is 50-120Hz; Verify whether the printed microstructure conforms to the pre-defined statistical characteristics; the printed microstructure includes paper fibers and cells; The mobile terminal uses micro-modulation of supplementary light (10-20Hz) to detect whether the reflection response meets the preset reflection response standard, thereby filtering out images that are copied from the screen or forged from photographs; the frequency range of the mobile terminal's supplementary light micro-modulation is 10-20Hz. If any of the above three detection items has a negative result, the anti-replay detection is determined to be unsuccessful. Then, the user is prompted to re-execute S100 through dynamic interaction via the mobile APP.

9. The anti-counterfeiting verification method according to claim 8, characterized in that: In S300, if the mobile APP is in a network-free environment, ECDSA or Ed25519 signature verification is performed based on the locally cached whitelist and revocation list, and a temporary pre-verification result is output. In S400, the digital certificate also includes production batch information and logistics link information corresponding to the anti-counterfeiting outer packaging.

10. The anti-counterfeiting verification method according to claim 9, characterized in that: S231 performs the two-dimensional fast Fourier transform, expressed as follows: in: The frequency domain complex matrix is ​​obtained by transforming the image after log-polar transformation through the two-dimensional fast Fourier transform. This is the horizontal frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the horizontal direction of the spatial domain image is as follows: A larger value indicates a higher frequency in the horizontal direction; This is the vertical frequency index in a frequency-domain rectangular coordinate system, unitless, with a value range of [value missing]. And the corresponding frequency change rate in the vertical direction of the spatial domain image is as follows: A larger value indicates a higher frequency in the vertical direction; These are the spatial domain image pixel values ​​after log-polar transformation; The radial coordinates are in the log-polar coordinate system, corresponding to the scaling scale of the spatial domain image, in pixels; The polar angle in the log-polar coordinate system corresponds to the rotation angle of the spatial domain image, in radians; Pi is a constant. The imaginary unit is used to describe phase information in the frequency domain and is adapted to the complex form of the Fourier transform. The horizontal dimension of the zero-padded image, in pixels, with a value of 512 or 1024, used to reduce spectral leakage; The vertical dimension of the zero-padded image, in pixels, is used for isotropic frequency domain transformation, and its value is related to... Consistent; The matching coefficient between the horizontal frequency component and the spatial domain pixel position represents the horizontal frequency. radial coordinates in the spatial domain Contribution weight at each location; The matching coefficient between the vertical frequency component and the spatial domain pixel position represents the vertical frequency. radial coordinates in the spatial domain Contribution weight at each location; This is a double summation symbol, and the range of summation is expressed by the following formula: It is used to perform an accumulation operation on all pixels of the zero-padded image, completing the conversion from the spatial domain to the frequency domain.

Images