A Method for Reading a Physically Unclonable Anti-Counterfeiting Mark with a Flower-Like Dendritic Pattern

Abstract

The present invention proposes a method for reading a physically unclonable anti-counterfeiting label with a flower-shaped dendritic pattern. First, a corresponding reading method is proposed for the flower-shaped dendritic pattern to achieve inter-class judgment of different types of flower-shaped dendritic patterns. For the same type of flower-shaped dendritic pattern, a feature description operator is constructed using the pattern shape features, and a corresponding feature vector is established, enabling intra-class matching recognition. Compared with machine learning, the matching of eigenvalues and feature vectors greatly reduces the computational difficulty, and there is no need to store the entire image. Only the image features need to be stored, which can greatly reduce the memory occupancy, relieve the pressure on the processor and memory, and reduce the requirements for software and hardware configurations.

Description

Technical Field

[0001] The present invention relates to pattern recognition technology, and particularly to a method for reading a transparent flower-shaped dendritic pattern anti-counterfeiting mark. Background Art

[0002] Using physical unclonable function (PUF) to construct anti-counterfeiting marks is an important means to resist counterfeit products. Such marks are generally prepared by precise and controlled complex technological processes and have diverse random patterns, which are not only difficult to reverse-engineer and replicate, but also the reading process is quite challenging.

[0003] The existing reading methods for physical unclonable anti-counterfeiting marks such as item fingerprints, randomly distributed nanoparticles, and random wrinkles all adopt deep learning methods to achieve precise discrimination of similar patterns through large-scale training by collecting pictures. However, deep learning as a reading method for anti-counterfeiting marks has the following problems: First, for deep learning to achieve high-precision reading, it needs to rely on a large number of training samples to provide a sufficient number of data sets, which is not applicable to high-value and high-precision products with small quantities, customized products, and small-batch production. Second, deep learning is based on neural network learning of images and stores a large number of original images, which occupies a high memory of the computer. Third, deep learning requires a long training process in the early stage, increasing the time cost, and the training process has high requirements for computing power. As the complexity of the graph model increases, the time complexity of the algorithm increases sharply, and ordinary CPUs cannot meet the requirements of deep learning. GPUs and TPUs need to be used, which have high requirements for hardware configuration. Finally, deep learning cannot estimate the regularity of data without bias, and the interpretability of the recognition results is not high.

[0004] Feature matching is a recognized highly reliable algorithm with the ability to explain recognition results in pattern recognition and has an unshakable position in fields related to national security such as military, national defense, and currency. The parameter matching method for constructing feature descriptors does not depend on the number of samples and can achieve pattern recognition for small sample groups. The feature parameter matching method stores the features of images, which are a series of specific numerical values, and is convenient to store and occupies little memory. Moreover, the parameter matching method has low requirements for computing power and does not require high configurations of CPUs, GPUs, and TPUs, greatly reducing the configuration requirements for arithmetic units and processors. Most importantly, pattern recognition based on the feature parameter matching method has the right to speak in the determination of results and can optimize the accuracy of recognition results by continuously correcting the representation form of feature vectors. However, for unclonable patterns with complex styles, how to construct feature descriptors that can reflect the subtle differences of flower-shaped dendritic patterns not only requires a series of image processing but also needs to discard invalid information and extract useful information from a large amount of fuzzy information. Currently, no relevant solutions have been proposed. SUMMARY OF THE INVENTION

[0005] To solve the problems existing in the prior art, the present invention proposes a method for reading a physically unclonable anti-counterfeiting label with a flower-shaped dendritic pattern.

[0006] The flower-shaped dendritic pattern is formed by regulating the process of an aqueous sodium silicate solution, and presents six flower-shaped dendritic patterns of "cellular", "multi-branched and variegated flower-shaped", "cross-shaped", "bacterial-shaped", "three-branched-shaped" and "densely branched-shaped" under a scanning electron microscope or an optical microscope.

[0007] In actual use, for a genuine anti-counterfeiting label, it is necessary to read the flower-shaped dendritic pattern in the anti-counterfeiting label, including inter-class feature matching and intra-class matching recognition, establish a feature vector corresponding to the anti-counterfeiting label, and the feature vectors of all genuine anti-counterfeiting labels form a feature vector library. When judging the label to be recognized, first judge whether there is a flower-shaped dendritic pattern in the label and what specific flower-shaped dendritic pattern it is through inter-class feature matching, and then perform intra-class matching recognition with the feature vector library to judge whether it is a genuine anti-counterfeiting label.

[0008] The technical solution of the present invention is as follows:

[0009] A method for reading a flower-shaped dendritic pattern, comprising the following steps:

[0010] Step 1: Take a microscopic image of the flower-shaped dendritic pattern under a microscope, and preprocess the image, including mean filtering, binarization, morphological opening processing, contour extraction, and connected domain parameter calculation;

[0011] The flower-shaped dendritic pattern is divided into six types: "cellular", "multi-branched and variegated flower-shaped", "cross-shaped", "bacterial-shaped", "three-branched-shaped" and "densely branched-shaped";

[0012] Step 2: Judge the number of contours of the microscopic image of the flower-shaped dendritic pattern. If the number of contours reaches the set threshold, it is judged as a "cellular" pattern; otherwise, further judge the number of straight lines. If the number of straight lines reaches the set threshold, it is judged as a "densely branched-shaped" pattern; otherwise, further draw a polar coordinate diagram D–θ of the contour, extract the number of wave peaks and wave valleys in the polar coordinate diagram D–θ of the contour. A pattern with three wave peaks and three wave valleys is a "three-branched" pattern, and a pattern with four wave peaks and four wave valleys is a "cross-shaped" pattern. If the number of wave peaks and wave valleys does not meet the requirements, calculate the shape parameter SP according to the contour length P and the contour area A:

[0013]

[0014] The pattern with the SP value in the range of [3.3, 7.4] is a "bacterial-shaped" pattern, and the pattern with the SP value in the range of [19.4, 27.1] is a "multi-branched and variegated flower-shaped" pattern.

[0015] Further, in step 2, the threshold for the number of contours is set to 15, and the threshold for the number of straight lines is set to 15.

[0016] Further, the flower-like dendrite pattern is obtained through the following steps:

[0017] Step a: Pretreat the sodium silicate aqueous solution to obtain a solution for use. The pretreatment process includes static settlement, ultrasonic oscillation, and ball milling.

[0018] Step b: Coat the solution for use pretreated in step a on the surface of an object to obtain a smooth, flat, uniform coating film without depressions and defects on the surface of the object.

[0019] Step c: Cure the coating film treated in step b to obtain an amorphous silicon polymer with a flower-like dendrite pattern.

[0020] Among them: In step a, room temperature static settlement pretreatment is adopted. In step c, the curing conditions are: keep warm at a temperature of 280 - 288K for not less than 6h to obtain a flower-like dendrite pattern of "cellular" pattern.

[0021] In step a, room temperature static settlement pretreatment is adopted. In step c, the curing conditions are: keep warm at a temperature of 297 - 306K for not less than 6h to obtain a flower-like dendrite pattern of "multi-branched and miscellaneous flower-like" pattern.

[0022] In step a, room temperature static settlement pretreatment is adopted. In step c, the curing conditions are: first keep warm at a temperature of 280 - 288K for 0.3 - 0.7h, then heat up to 297 - 306K at a heating rate of 5 - 12K / min and keep warm for not less than 6h to obtain a flower-like dendrite pattern of "cross-shaped" pattern.

[0023] In step a, ultrasonic oscillation pretreatment is adopted. In step c, the curing conditions are: first keep warm at a temperature of 297 - 306K for 1 - 2h, then heat up to 315 - 335K at a heating rate of 5 - 12K / min and keep warm for not less than 6h to obtain a flower-like dendrite pattern of "bacteria-like" pattern.

[0024] In step a, ultrasonic oscillation pretreatment is adopted. In step c, the curing conditions are: keep warm at a temperature of 297 - 306K for not less than 6h to obtain a flower-like dendrite pattern of "three-branch-shaped" pattern.

[0025] In step a, forward and reverse alternating ball milling pretreatment is adopted. In step c, the curing conditions are: keep warm at a temperature of 297 - 306K for not less than 6h to obtain a flower-like dendrite pattern of "densely branched" pattern.

[0026] A method for establishing a database of standard characteristic parameters of an anti-counterfeiting label based on a flower-like dendrite pattern includes the following steps:

[0027] Step A: For the prepared anti-counterfeiting label based on the flower-like dendrite pattern, extract the outer contour of the product anti-counterfeiting label under a microscope, select the corresponding magnification according to the flower-like dendrite pattern corresponding to the anti-counterfeiting label, take the microscopic image of the flower-like dendrite pattern, and preprocess the image, including mean filtering, binarization, morphological opening processing, contour extraction, and connected component parameter calculation;

[0028] Step B: According to the flower-like dendrite pattern corresponding to the anti-counterfeiting label, establish a Q×2-dimensional feature matrix corresponding to the anti-counterfeiting label; the Q×2-dimensional feature matrix is composed of two Q-dimensional vectors. For the "dense branch-like" pattern, the elements of the two Q-dimensional vectors are respectively the length ratio and slope of the longest straight line segment in their respective Q sub-images of the pattern; for the "cellular", "multi-branch and miscellaneous flower-like", "cross-like", "bacterial-like", "three-branch-like" patterns, the elements of the two Q-dimensional vectors are respectively the contour area ratio in their respective Q sub-images of the pattern, and the ratio of the number of their respective contours to the total number of contours in the pattern;

[0029] Step C: Repeat Step A and Step B for all prepared anti-counterfeiting labels based on the flower-like dendrite pattern to obtain their respective corresponding Q×2-dimensional feature matrices. The Q×2-dimensional feature matrices of all anti-counterfeiting labels form a feature parameter database.

[0030] Further, the value range of Q is an even number within the range of 4 to 16.

[0031] Further, the magnification of the "cellular" pattern is 2000 times, the magnification of the "multi-branch and miscellaneous flower-like" pattern is 100 times, the magnification of the "cross-like" pattern is 300 times, the magnification of the "bacterial-like" pattern is 300 times, the magnification of the "three-branch-like" pattern is 100 times, and the magnification of the "dense branch-like" pattern is 300 times.

[0032] A method for reading a physically unclonable anti-counterfeiting label with a flower-like dendrite pattern includes the following steps:

[0033] Step 1: Extract the outer contour of the anti-counterfeiting label to be identified under a microscope, and determine whether there is a flower-like dendrite pattern in the anti-counterfeiting label to be identified. If there is no flower-like dendrite pattern, it is determined that the anti-counterfeiting label to be identified is false. If there is a flower-like dendrite pattern, further use the reading method of the flower-like dendrite pattern to determine the type of the flower-like dendrite pattern in the anti-counterfeiting label to be identified;

[0034] Step 2: Establish a Q×2-dimensional feature matrix corresponding to the anti-counterfeiting label to be recognized according to the types of flower-like dendritic patterns in the anti-counterfeiting label to be recognized; the Q×2-dimensional feature matrix consists of two Q-dimensional vectors. For the "dense branch-like" pattern, the elements of the two Q-dimensional vectors are respectively the length and slope of the longest straight line segment in the K-part images of the pattern. For the "cellular", "multi-branch and miscellaneous flower-like", "cross-like", "bacterial-like", and "three-branch-like" patterns, the elements of the two Q-dimensional vectors are respectively the proportion of the contour area in the Q-part images of the pattern and the proportion of the number of their contours in the total number of contours of the pattern;

[0035] Step 3: Calculate the Euclidean distance between the Q×2-dimensional feature matrix of the anti-counterfeiting label to be recognized and each feature matrix in the standard feature parameter database of the genuine anti-counterfeiting label. If the obtained minimum Euclidean distance is less than the set threshold, it is determined that the anti-counterfeiting label to be recognized is a genuine anti-counterfeiting label; otherwise, it is a false anti-counterfeiting label.

[0036] Beneficial effects

[0037] The present invention first proposes a corresponding reading method for flower-like dendritic patterns, realizing the inter-class judgment of different types of flower-like dendritic patterns. For the same type of flower-like dendritic patterns, a feature description operator is constructed using the pattern shape features, and the corresponding feature vectors are established, so as to be able to perform intra-class matching recognition. Compared with machine learning, the matching of eigenvalues and feature vectors greatly reduces the operation difficulty, and there is no need to store the entire image, only the image features need to be stored, which can greatly reduce the memory occupancy, relieve the pressure on the processor and memory, and reduce the requirements for software and hardware configurations.

[0038] The additional aspects and advantages of the present invention will be partly given in the following description, partly will become obvious from the following description, or will be understood through the practice of the present invention. Description of the drawings

[0039] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, in which:

[0040] Figure 1 is the overall flowchart of the method of the present invention.

[0041] Figure 2 is the central point positioning diagram of the anti-counterfeiting label in the microscope in the method of the present invention.

[0042] Figure 3 is the image preprocessing flowchart in the method of the present invention.

[0043] Figure 4 is the flowchart of the classification and recognition process of six types of patterns in the method of the present invention.

[0044] Figure 5 It is the contour polar coordinate diagram in the method of the present invention.

[0045] Figure 6 Schematic diagram of the parameter solution of the sextuplicate image in the method of the present invention. (a) Cellular; (b) “Three-branched”; (c) “Multi-branched flower”; (d) “Cross”; (e) “Bacteria”; (f) “Densely branched”. DETAILED DESCRIPTION

[0046] The present invention is to read the physical non-clonable anti-counterfeiting mark based on the flower-like dendrite pattern, which is formed by the sodium silicate aqueous solution through process regulation, including precise regulation of the initial state and curing process of the inorganic sodium silicate aqueous solution (water glass), mainly considering the pretreatment, curing temperature and curing time, to achieve directional control of the microscopic morphology after curing. It includes the following steps:

[0047] Step a: pre-treating the sodium silicate aqueous solution to obtain a stand-by solution, wherein the pre-treatment process includes standing, ultrasonic oscillation, and ball milling;

[0048] For pretreatment at room temperature, it is required to stand at room temperature for no less than 3 days to achieve natural aging. Under the condition that the solution does not deteriorate, the higher the aging degree, the better;

[0049] For ultrasonic oscillation pretreatment, the ultrasonic oscillation should be carried out for no less than 10 minutes in an ultrasonic cleaning machine;

[0050] For ball milling, the balls should be rotated alternately in forward and reverse directions in a planetary ball mill for no less than 0.5 h.

[0051] Of course, different equipment suppliers have different parameters such as equipment power, frequency, speed, etc., and the pretreatment time can be changed accordingly without limitation. Under the premise of ensuring efficiency and cost, increasing the time of static treatment, ultrasonic treatment and ball milling treatment as appropriate can obtain a more ideal sodium silicate or water glass solution.

[0052] Step b: respectively transfer the standby solutions after different pretreatment processes in step a to their respective air spray guns, adjust the spraying distance, and start spraying on the surface of the object respectively. Each spraying should ensure that the droplets are evenly distributed on the surface of the object. Use 5-8 times of thin coating to make the coating on the surface of the object smooth and flat, without bubbles, defects and impurities; of course, other uniform coating methods in the air such as spin coating and painting can also be used to ensure that the coating surface is smooth and flat without depressions and defects.

[0053] Step c: Curing the coating film processed in step b respectively, including different combination methods such as isothermal curing, stepwise temperature rise curing, continuous temperature rise curing, etc., to obtain an amorphous silicon polymer with variable flower shapes. Here, the variable flower shapes are amorphous cage polymer structures formed by different silicate polymerization states; the flower-shaped dendritic patterns in the obtained amorphous silicon polymer include six types: "cellular", "multi-branched and miscellaneous flower-shaped", "cross-shaped", "bacterial-shaped", "three-branched-shaped" and "densely branched-shaped".

[0054] Specifically, the pretreatment, curing temperature and curing time combination schemes in Table 1 are adopted to fully cure the coating film.

[0055] Table 1 Preparation conditions for six types of flower shapes

[0056]

[0057]

[0058] For the recognition method of the above flower-shaped dendritic patterns, it includes the following steps:

[0059] Step 1: Place the product with the anti-counterfeiting label of the flower-shaped dendritic pattern in an optical microscope or an electron microscope, extract the outer contour of the entire anti-counterfeiting label, calculate the centroid position according to the contour, take the centroid as the center of magnification and reduction, look up the magnification table of six types of patterns, and take a microscopic image of the flower-shaped dendritic pattern. The setting principle of the magnification is to ensure that there is at least one complete flower-shaped pattern in the field of view and the pattern occupies no less than 50×50pt in pixels.

[0060] Table 2 Magnification ratios of anti-counterfeiting labels of six types of flower-shaped dendritic patterns

[0061] Flower-like dendritic pattern Magnification ratio Cellular 2000x Multi-branched and variegated flower-like 100x Cross-shaped 300x Bacteria-like 300x Three-branched 100x Densely branched 300x

[0062] Perform preprocessing on the image, including mean filtering, binarization, morphological opening processing, contour extraction, and connected component parameter calculation;

[0063] Mean filtering: Perform mean filtering on the image to remove noise points. The specific operation is to form a filtering template with a certain pixel point and its 8 adjacent pixels nearby, and then use the average value of all pixels in the template to replace the pixel value of the central pixel point, that is

[0064] Binarization: Statistically calculate the gray value of each point in the image and draw a gray distribution histogram; because the gray value contrast between the micro-pattern and the background is obvious, showing a bimodal feature, mark the vertices of the two peaks as H max1 and H max2 , corresponding to the gray values T1 and T2; find the threshold T within the interval [T1, T2] so that min = H T ,, make H < H T, T = 0; H ≥ H T , T = 1.

[0065] Morphological opening operation: First erode and then dilate, and the erosion and dilation unit is a 3×3 pixel block.

[0066] Contour extraction: Traverse the binary image after the morphological opening operation, and make a judgment in a line-by-line scanning manner. The points that meet the rule of j ≤ t, i = 1; j > t, i = 0 are determined as contour points.

[0067] Calculation of connected component parameters: The perimeter and area are calculated using the arcLength and contourArea functions in OpenCV respectively.

[0068] Step 2: As Figure 4 shown, classify and recognize the microscopic image of the flower-like dendritic pattern:

[0069] Use the findContours function in OpenCV to find the contours in the image. This function will return the number of contours in the image and the coordinate array of the points that make up each contour curve. Patterns with the number of contours greater than the threshold (K = 15) are "cellular" patterns;

[0070] Use the Hough line detection algorithm HoughLines to count the number of lines in the image. Patterns with the number of lines greater than the threshold (K = 15) are "dense-branched" patterns;

[0071] Using the moment function, the moment of each contour can be calculated, and then the centroid coordinates of each contour can be obtained. By traversing, the pixel distance from each point on the contour to the centroid can be calculated, and the contour polar coordinate graph D–θ can be drawn. As Figure 5 shown, the pattern with three peaks and three valleys is "three-branched", and the pattern with four peaks and four valleys is "cross-shaped";

[0072] The drawing method of the contour polar coordinate graph D–θ is to start from a point on the contour, calculate the distance ri from this point to the center point of the pattern, rotate clockwise for one week and calculate the distance D between the points on the contour and the center point of the pattern in turn. Using the distance D as the ordinate and the rotation angle θ as the abscissa to draw a graph, which is the contour polar coordinate graph D–θ.

[0073] Using the arcLength and contourArea functions, the perimeter P and area A of each contour can be calculated respectively, and the shape parameter SP can be calculated:

[0074]

[0075] Those with SP values in the range of [3.3, 7.4] are "bacteria-like", and those with SP values in the range of [19.4, 27.1] are "multi-branched and miscellaneous flowers" patterns.

[0076] For the anti-counterfeiting labels based on flower-like dendrite patterns obtained in batches, a standard feature parameter database for anti-counterfeiting labels based on flower-like dendrite patterns is established by constructing feature descriptors, including the following steps:

[0077] Step A: For the prepared anti-counterfeiting label based on the flower-like dendrite pattern, extract the outer contour of the product anti-counterfeiting label under a microscope, select the corresponding magnification according to the flower-like dendrite pattern corresponding to the anti-counterfeiting label, take a microscopic image of the flower-like dendrite pattern, and preprocess the image, including mean filtering, binarization, morphological opening processing, contour extraction, and connected domain parameter calculation;

[0078] Step B: According to the flower-like dendrite pattern corresponding to the anti-counterfeiting label, establish a Q×2-dimensional feature matrix corresponding to the anti-counterfeiting label; the Q×2-dimensional feature matrix is composed of two Q-dimensional vectors, and the value range of Q is an even number in the range of 4 to 16. In this embodiment, Q = 6 is taken. That is, the image processed in Step A is equally divided into 6 parts, evenly divided into 3 parts in the X direction and 2 parts in the Y direction, and the six equal parts of the image are numbered #1 to #6 in the order from left to right and from top to bottom.

[0079] For the "cellular", "multi-branched and variegated flower-like", "cross-shaped", "bacterial-like", and "three-branched" patterns, calculate the number of pixels with a pixel value of 1 in images #1 to #6 in turn, and then use the findContours function to calculate the number of contour connected domains in images #1 to #6, as Figure 6 shown.

[0080] For the "dense-branched" pattern, use the HoughLines algorithm to detect the lines contained in images #1 to #6, obtain the two endpoints (x min , y min ) and (x max , y max ) on each line, calculate the length of each line, select the longest line segment in each of the #1 to #6 equal division regions, and calculate the length and slope of the longest line segment, as Figure 6 shown.

[0081] The statistical methods for the "cellular", "multi-branched and variegated flower-like", "cross-shaped", "bacterial-like", and "three-branched" patterns are the same. Specifically, calculate the area ratio of the points with a pixel value of 1 in the 1 / 6 region in images #1 to #6 and the ratio of the number of connected domains in images #1 to #6 to the total number of connected domains, which are represented by two six-dimensional vectors, n = [n x1 , n x2 , n x3 , n x4 , n x5 , n x6 T and s = [s​x1 , s x2 , s x3 , s x4 , s x5 , s x6 T 。

[0082] For the "dense branches" pattern, first calculate the ratio of the length of the longest straight line segment in Images #1 - #6 to the diagonal of the 1 / 6 region, and then calculate the slope of the longest straight line segment in Images #1 - #6. Represent it with two six-dimensional vectors l = [l x1 , l x2 , l x3 , l x4 , l x5 , l x6 T and k = [k x1 , k x2 , k x3 , k x4 , k x5 , k x6 T 。

[0083] Form a matrix with the two six-dimensional vectors, and the finally formed 6×2 feature matrix is:

[0084]

[0085] Each flower-like dendritic crystal pattern corresponds to a 6×2 feature matrix. Solve the feature matrices of all six types of flower-like dendritic crystal patterns and establish six feature matrix databases respectively.

[0086] Based on the above method, for a security label to be recognized, it is read through the following steps to determine whether it is a genuine security label belonging to the feature matrix database:

[0087] Step 1: Extract the outer contour of the security label to be recognized under a microscope, and determine whether there is a flower-like dendritic crystal pattern in the security label to be recognized. If there is no flower-like dendritic crystal pattern, it is determined that the security label to be recognized is false. If there is a flower-like dendritic crystal pattern, further use the reading method of the flower-like dendritic crystal pattern to determine the type of the flower-like dendritic crystal pattern in the security label to be recognized;

[0088] Step 2: According to the type of the flower-like dendritic crystal pattern in the security label to be recognized, establish a 6×2-dimensional feature matrix corresponding to the security label to be recognized;

[0089] Step 3: Calculate the Euclidean distance between the 6×2-dimensional feature matrix of the security label to be recognized and each feature matrix in the standard feature parameter database of the genuine security label:

[0090] ​​​

[0091]

[0092] If the obtained minimum Euclidean distance is less than the set threshold value, it is determined that the anti-counterfeiting label to be recognized is a genuine anti-counterfeiting label; otherwise, it is a false anti-counterfeiting label.

[0093] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A method for identifying a flower-shaped dendritic pattern, characterized in that: It includes the following steps: Step 1: Take a microscopic image of the flower-like dendritic pattern under a microscope and preprocess the image, including mean filtering, binarization, morphological opening, contour extraction, and connected component parameter calculation; The flower-like dendritic patterns are divided into six types: "cellular", "multi-branched and variegated flower-like", "cross-shaped", "bacterial-like", "three-branched", and "densely branched"; The flower-like dendritic pattern is obtained through the following steps: Step a: Pretreat the sodium silicate aqueous solution to obtain a solution for use. The pretreatment process includes standing, ultrasonic oscillation, and ball milling; Step b: Coat the solution for use pretreated in step a on the surface of an object to obtain a smooth, flat, non-sunken and intact uniform coating film on the object surface; Step c: Cure the coating film treated in step b to obtain an amorphous silicon polymer with a flower-like dendritic pattern; Through the pretreatment process and the curing process, the directional control of the microstructure after curing is realized; Among them: In step a, room temperature standing pretreatment is adopted, and the curing conditions in step c are: Keep warm at a temperature of 280 - 288K for no less than 6h to obtain a "cellular" pattern of the flower-like dendritic pattern; In step a, room temperature standing pretreatment is adopted, and the curing conditions in step c are: Keep warm at a temperature of 297 - 306K for no less than 6h to obtain a "multi-branched and variegated flower-like" pattern of the flower-like dendritic pattern; In step a, room temperature standing pretreatment is adopted, and the curing conditions in step c are: First, keep warm at a temperature of 280 - 288K for 0.3 - 0.7h, then heat up to 297 - 306K at a heating rate of 5 - 12K / min and keep warm for no less than 6h to obtain a "cross-shaped" pattern of the flower-like dendritic pattern; In step a, ultrasonic oscillation pretreatment is adopted, and the curing conditions in step c are: First, keep warm at a temperature of 297 - 306K for 1 - 2h, then heat up to 315 - 335K at a heating rate of 5 - 12K / min and keep warm for no less than 6h to obtain a "bacterial-like" pattern of the flower-like dendritic pattern; In step a, ultrasonic oscillation pretreatment is adopted, and the curing conditions in step c are: Keep warm at a temperature of 297 - 306K for no less than 6h to obtain a "three-branched" pattern of the flower-like dendritic pattern; In step a, forward and reverse alternating ball milling pretreatment is adopted, and the curing conditions in step c are: Keep warm at a temperature of 297 - 306K for no less than 6h to obtain a "densely branched" pattern of the flower-like dendritic pattern; Step 2: Judge the number of contours of the microscopic image of the flower-like dendritic pattern. If the number of contours reaches the set threshold, it is judged as a "cellular" pattern; Otherwise, further judge the number of straight lines. If the number of straight lines reaches the set threshold, it is judged as a "densely branched" pattern; Otherwise, further draw the contour polar coordinate diagram D–θ, extract the number of peaks and valleys in the contour polar coordinate diagram D–θ. The pattern with three peaks and three valleys is a "three-branched" pattern, and the pattern with four peaks and four valleys is a "cross-shaped" pattern. If the number of peaks and valleys does not meet the requirements, calculate the shape parameter SP according to the contour length P and the contour area A: Those with SP values in the range of [3.3, 7.4] are "bacterial" patterns, and those with SP values in the range of [19.4, 27.1] are "multi-branched and variegated flower" patterns.

2. The method for reading a flower-shaped dendritic pattern according to claim 1, characterized in that: In step 2, the threshold for the number of contours is set to 15, and the threshold for the number of straight lines is set to 15.

3. A method for establishing a standard feature parameter database of an anti-counterfeiting mark based on a flower-like dendritic pattern, characterized in that: It includes the following steps: Step A: For the anti-counterfeiting label with the flower-like dendritic pattern prepared according to claim 1, extract the outer contour of the product anti-counterfeiting label under a microscope, select the corresponding magnification according to the flower-like dendritic pattern corresponding to the anti-counterfeiting label, take a microscopic image of the flower-like dendritic pattern, and preprocess the image, including mean filtering, binarization, morphological opening processing, contour extraction, and calculation of connected domain parameters; Step B: According to the flower-like dendritic pattern corresponding to the anti-counterfeiting label, establish a Q×2-dimensional feature matrix corresponding to the anti-counterfeiting label; the Q×2-dimensional feature matrix consists of two Q-dimensional vectors. For the "dense branch" pattern, the elements of the two Q-dimensional vectors are respectively the length ratio and slope of the longest straight line segment in each Q-equal part image of the pattern; for the "cellular", "multi-branched and variegated flower", "cross-shaped", "bacterial", "three-branched" patterns, the elements of the two Q-dimensional vectors are respectively the ratio of the contour area of each in the Q-equal part image of the pattern to the area of the 1 / Q region of the pattern, and the ratio of the number of their respective contours to the total number of contours in the pattern; Step C: Repeat step A and step B for all prepared anti-counterfeiting labels with flower-like dendritic patterns to obtain their respective corresponding Q×2-dimensional feature matrices, and the Q×2-dimensional feature matrices of all anti-counterfeiting labels constitute a feature parameter database.

4. The method for establishing a standard feature parameter database of an anti-counterfeiting label based on a flower-shaped dendrite pattern according to claim 3, wherein: The value range of Q is an even number within the range of 4 to 16.

5. The method for establishing a standard feature parameter database of an anti-counterfeiting label based on a flower-shaped dendrite pattern according to claim 3, wherein: The magnification of the "cellular" pattern is 2000 times, the magnification of the "multi-branched and variegated flower" pattern is 100 times, the magnification of the "cross-shaped" pattern is 300 times, the magnification of the "bacterial" pattern is 300 times, the magnification of the "three-branched" pattern is 100 times, and the magnification of the "dense branch" pattern is 300 times.

6. A method for reading a physically unclonable anti-counterfeiting mark with a flower-shaped dendritic pattern, characterized in that: It includes the following steps: Step 1: Extract the outer contour of the anti-counterfeiting label to be recognized under a microscope, and judge whether there is a flower-like dendritic pattern in the anti-counterfeiting label to be recognized. If there is no flower-like dendritic pattern, judge that the anti-counterfeiting label to be recognized is fake. If there is a flower-like dendritic pattern, further use the recognition method of the flower-like dendritic pattern according to any one of claims 1 to 2 to determine the type of the flower-like dendritic pattern in the anti-counterfeiting label to be recognized; Step 2: According to the type of the flower-like dendritic pattern in the anti-counterfeiting label to be recognized, establish a Q×2-dimensional feature matrix corresponding to the anti-counterfeiting label to be recognized; the Q×2-dimensional feature matrix consists of two Q-dimensional vectors. For the "dense branch" pattern, the elements of the two Q-dimensional vectors are respectively the length and slope of the longest straight line segment in each Q-equal part image of the pattern; for the "cellular", "multi-branched and variegated flower", "cross-shaped", "bacterial", "three-branched" patterns, the elements of the two Q-dimensional vectors are respectively the ratio of the contour area of each in the Q-equal part image of the pattern to the area of the 1 / Q region of the pattern, and the ratio of the number of their respective contours to the total number of contours in the pattern; Step 3: Calculate the Euclidean distance between the Q×2-dimensional feature matrix of the anti-counterfeiting label to be recognized and each feature matrix in the standard feature parameter database of the genuine anti-counterfeiting label. If the obtained minimum Euclidean distance is less than the set threshold, then determine that the anti-counterfeiting label to be recognized is a genuine anti-counterfeiting label; otherwise, it is a fake anti-counterfeiting label.

Images