Method and device for verifying the authenticity of a product

Abstract

The present invention relates to a method and device for verifying the authenticity of a product, wherein a product identification of a product has randomly distributed microdots, and the method comprises: extracting microdot features on the product identification from an image of the product identification of the verified product; extracting image features of at least a portion of the product identification from the image using a machine learning algorithm; and verifying the authenticity of the product identification of the verified product based on the extracted microdot features and image features. The method and device provided by the embodiments of the present invention can improve the accuracy of verifying the authenticity of a product.

Description

Technical Field

[0001] The present invention relates to a method and a device for verifying the authenticity of a product. Background Art

[0002] Counterfeit and shoddy products cause huge losses to both producers and consumers, so they need to be controlled by using safe and reliable anti-counterfeiting technologies. Existing anti-counterfeiting technologies for products include digital anti-counterfeiting technology and texture anti-counterfeiting technology.

[0003] Digital anti-counterfeiting technology uses barcodes or QR codes to give products a unique identity (ID) for anti-counterfeiting verification and traceability functions. This digital anti-counterfeiting technology is easy to copy and has poor security.

[0004] Texture anti-counterfeiting technology uses randomly generated natural textures as anti-counterfeiting features. This texture is physically non-replicable and has non-reproducible characteristics. However, existing texture anti-counterfeiting technology lacks the ability to automatically identify anti-counterfeiting features. The automatic identification capability requires that the anti-counterfeiting features have visual identifiability, or rely on the addition of fiber materials during the production process to form anti-counterfeiting features, which leads to increased costs of anti-counterfeiting products and inconvenience in production.

[0005] At present, a new technology combining barcodes or QR codes with printed micro-dot features has emerged to further improve the anti-counterfeiting performance of product identification, while simplifying the production process of anti-counterfeiting products and reducing production costs. However, in the process of verifying the product identification, it is first necessary to obtain an image of the product identification of the verified product, for example, the user needs to use a mobile phone or digital camera to take an image of the product identification. However, due to differences in the shooting function of the mobile phone or camera, the shooting environment (such as light) and the shooting level (such as shooting angle, shooting distance, camera stability) and other factors, the image quality is affected to varying degrees, resulting in deviations in the verification results, such as verifying that the image of the product identification of the authentic product is a counterfeit, or verifying that the image of the product identification of the counterfeit product is a genuine product. Summary of the invention

[0006] In view of at least one of the above problems in the prior art, embodiments of the present invention provide a method and apparatus for verifying the authenticity of a product, which can improve the accuracy of verifying the authenticity of a product.

[0007] An embodiment of the present invention provides a method for verifying the authenticity of a product, wherein the product identification of the product has randomly distributed microdots, and the method comprises: extracting microdot features on the product identification of the verified product from an image of the product identification of the verified product; extracting image features of at least a portion of the product identification from the image using a machine learning algorithm; and verifying the authenticity of the product identification of the verified product based on the extracted microdot features and the image features.

[0008] An embodiment of the present invention provides a device for verifying the authenticity of a product, wherein the product identification of the product has randomly distributed microdots, and the device comprises: a microdot feature extraction module, used to extract microdot features on the product identification of the verified product from an image of the product identification of the verified product; an image feature extraction module, used to extract image features of at least a part of the product identification from the image using a machine learning algorithm; and a verification module, used to verify the authenticity of the product identification of the verified product based on the extracted microdot features and the image features.

[0009] An embodiment of the present invention provides a device for verifying the authenticity of a product, wherein the product identification of the product has randomly distributed micro-dots, and the device comprises: a memory for storing instructions; and a processor coupled to the memory, wherein the instructions, when executed by the processor, cause the processor to execute the method according to the above embodiment.

[0010] An embodiment of the present invention further provides a computer-readable storage medium on which executable instructions are stored. When the executable instructions are executed by a computer, the computer executes the method of the above embodiment.

[0011] According to the scheme of the embodiment of the present invention, when verifying the authenticity of a product logo, not only the micro-dot features on the product logo are utilized, but also the image features extracted from the image of the product logo through a machine learning algorithm are utilized, thereby improving the accuracy of verifying the authenticity of the product logo. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Other features, characteristics, benefits and advantages of the present invention will become more apparent from the detailed description taken in conjunction with the following drawings, in which:

[0013] Figure 1 A flow chart of a method for verifying the authenticity of a product according to a first embodiment of the present invention is shown;

[0014] Figure 2 is a schematic diagram of embedding micro-dot features into a product QR code in an embodiment;

[0015] FIG3(a) and FIG3(b) show an image of a probability density function when a uniform distribution function is used as a random distribution function of micro-dots, and a distribution diagram of micro-dots sampled from a random distribution;

[0016] Figure 4 A flow chart showing a method for verifying the authenticity of a product according to a second embodiment of the present invention;

[0017] FIG. 5A to FIG. 5CThree exemplary convolutional neural network structures used in the method for extracting image features of product logos in an embodiment of the present invention are shown; and

[0018] Figure 6 A structural block diagram of a device for verifying the authenticity of a product according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0019] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0020] Figure 1 A flow chart of a method 100 for verifying the authenticity of a product according to a first embodiment of the present invention is shown. Figure 1 The method 100 shown can be implemented by any computing device having computing capabilities, such as, but not limited to, a desktop computer, a laptop computer, a tablet computer, a server, or a smart phone.

[0021] like Figure 1 As shown, the verification method 100 includes: extracting micro-dot features on the product logo from the image of the product logo of the verified product (step 101); extracting image features of at least a portion of the product logo from the image of the product logo using a machine learning algorithm (step 102); and verifying the authenticity of the product logo of the verified product based on the extracted micro-dot features and image features (step 103).

[0022] In an embodiment of the present invention, the verification step 103 includes: using a classifier trained by a machine learning algorithm to verify the authenticity of the product identification of the verified product.

[0023] In an embodiment of the present invention, the machine learning algorithm used to extract image features is a convolutional neural network. The machine learning algorithm used to train the classifier is a machine learning algorithm that can classify feature vectors, such as a support vector machine (SVM) or a boost tree. A convolutional neural network is a type of feedforward neural network that includes convolution calculations and has a deep structure. It has representational learning capabilities and can perform translation-invariant classification of input information according to its hierarchical structure. The convolutional neural network is constructed by imitating the visual perception mechanism of organisms, and can perform supervised learning and unsupervised learning. The convolution kernel parameter sharing within the hidden layer and the sparsity of the inter-layer connections enable the convolutional neural network to learn grid features (such as pixels and audio) with a smaller amount of computation.

[0024] In an embodiment of the present invention, the verification method may further include: training a convolutional neural network and a classifier by using a plurality of authentic identification images as positive samples and a plurality of counterfeit identification images as negative samples.

[0025] In an embodiment of the present invention, step 102 of extracting image features includes: extracting image features from the image using a trained convolutional neural network to output a feature vector describing the image features.

[0026] In an embodiment of the present invention, the classifier includes a first classifier, and the extracted image features include at least a printing feature related to the printing of at least a part of the product identification; the first classifier distinguishes the authenticity of the product identification of the verified product based on the printing feature. The first classifier can be trained using positive samples and negative samples of the product identification. The verification step 103 may include: based on the extracted printing features, using the trained first classifier to output the probability that the product identification of the verified product is true; and / or based on the extracted printing features, using the trained first classifier to output the probability that the product identification of the verified product is false.

[0027] In an embodiment of the present invention, the printing characteristics of the product identification of the genuine product are characteristics associated with at least one of the paper, ink, and printing equipment used in the printing process of the product identification of the genuine product. Product identification printing can be printing a digital file on physical paper or other carriers. When the same digital file is printed, due to different printer settings, different types of printing machines, different inks or toners or colorants, different paper properties, and other complex combinations, the details of the same digital image may be different after printing. Such details reflect the printing characteristics. For example, due to the different paper, ink, or printing equipment used, the printed lines may have slight differences, such as slight jagged parts with different shapes or arrangements on the edges. For example, the two-dimensional code in the product identification contains multiple black blocks and white blocks, and all black and white boundaries may be different under different printing conditions. In addition, affected by the paper, ink, or printing equipment, the printed color or grayscale may also be different. Such printing differences are distributed in the entire printing area of ​​the two-dimensional code. And using such differences, the printing characteristics of the product identification can be extracted. The printed features of the product logo of a counterfeit product produced by copying technology are different from those of the genuine product logo. The convolutional neural network and the first classifier can be trained with a sufficient number of positive samples and negative samples. The convolutional neural network can learn the printed features in the positive samples and the printed features in the negative samples that are different from the positive samples during the training process; the trained convolutional neural network will have the ability to extract the printed features in the verified product logo. The trained first classifier can compare the printed features contained in the image features extracted by the convolutional neural network with the printed features contained in the image features of the genuine product to output the authenticity probability of the product logo of the verified product.

[0028] In an embodiment of the present invention, the classifier may further include a second classifier, which determines the authenticity of the product identification based on the authenticity probability and micro-dot features of the verified product identification output by the first classifier. The second classifier can be trained using positive samples and negative samples of the product identification. The verification step 103 may also include: comparing the extracted micro-dot features with the micro-dot features of the product identification pre-saved during or after the production of the product; forming a description vector about the product identification based on the comparison result and the authenticity probability of the product identification output by the first classifier; and using the second classifier to determine the authenticity of the product identification of the verified product based on the description vector.

[0029] In an embodiment of the present invention, the description vector includes data related to at least one of the following items: the matching rate between the extracted micro-dot features and the pre-saved micro-dot features, the statistical parameters of the pixel distance of the matched micro-dots from the pre-saved micro-dots in the image coordinate system, the number of micro-dots in the image of the product logo of the verified product that do not match the pre-saved micro-dot features, and the image quality of the acquired product logo.

[0030] In an embodiment of the present invention, step 101 of extracting micro-dot features may include: extracting at least one of shape features, position features, grayscale features, and color features of the micro-dots from the image using image processing technology.

[0031] In an embodiment of the present invention, the product identification may include at least one of a barcode and a two-dimensional graphic code.

[0032] Figure 22 is a schematic diagram of embedding micro-dot features into a product two-dimensional code in an embodiment, wherein the micro-dot feature 202 is not shown in detail in the figure because of its small size. In the process of generating micro-dot features, a specific high-dimensional random distribution map 201 of micro-dots is first generated by an algorithm, as a distribution characteristic of at least one of the position distribution, grayscale distribution, color distribution and microscopic morphology of all micro-dot features, and products of the same type or the same batch can follow a certain distribution characteristic, wherein each product has other different micro-dot features to distinguish. For example, different batches of products can use different random distribution maps, and different products of the same batch use different micro-dots. Then, the random distribution map of micro-dots is sampled using the algorithm, and a micro-dot feature 202 with unique identification is generated for each product (or product identification or label), and then the generated micro-dot feature is embedded in the digital two-dimensional identification 203 of the product (such as a quick response matrix code, i.e., a two-dimensional code) according to a predetermined avoidance rule, and the two-dimensional code embedded with the micro-dot feature is printed on the surface of the product or the surface of the product packaging as a product identification, or printed on the surface of the product label to form a digital product identification (ID) with micro-dots. The avoidance rules can limit at least one of the specific position distribution, grayscale distribution and color distribution of the micro-dots. For example, the position distribution avoidance rule can ensure that only black or dark micro-dots are generated in the white module of the QR code, and the grayscale distribution or color distribution avoidance rule can ensure that the grayscale or color of the micro-dots meets certain grayscale and saturation restrictions and does not interfere with the white module of the QR code. These avoidance rules work together to ensure that the reading of the QR code itself will not be affected by the embedded micro-dot features, and that the QR code still meets the corresponding national standards and / or international standards after the micro-dot features are added.

[0033] In some embodiments, the white micro-dot feature 202 can also be embedded in the black module of the two-dimensional code 203, and the avoidance rule limits the micro-dots to be generated only in the black module of the two-dimensional code, so that the two-dimensional code still meets the corresponding national standards and / or international standards after adding the micro-dot feature. The white micro-dots maintain the highest contrast in the black module of the two-dimensional code, and the white micro-dots are generated by short pauses in the printing inkjet during the printing process.

[0034] The composition of micro-dot features includes the most basic two-dimensional coordinates (X, Y) as position features, and can also include other optional features such as color, grayscale, shape, etc. Usually, the non-reproducibility and anti-counterfeiting performance of micro-dot features are first achieved through the random distribution of the two-dimensional positions of the micro-dots. The color, grayscale or shape characteristics of the micro-dots can be used to further improve the anti-counterfeiting performance of the product. Randomly distributed micro-dot features can also form randomly distributed micro-dot texture features.

[0035] After the product logo is made or during the production process, the micro-dot feature information on the product logo needs to be saved in a database for subsequent product authenticity verification. The saved micro-dot feature information includes, for example, randomly distributed position features, and other features such as color, grayscale or shape.

[0036] As an example of micro-dot features, Figures 3(a) and 3(b) show the image of the probability density function when the uniform distribution function is used as the random distribution function of the micro-dots, and the micro-dot distribution diagram sampled from the random distribution. The probability density function of the uniform distribution function is:

[0037] PDF(x,y)=const

[0038] In the probability density function image of Figure 3(a), the Z-axis coordinate is the probability density, and the horizontal coordinate X and the vertical coordinate Y indicate the position (x, y) of the micro-dot. The micro-dot distribution map of Figure 3(b) is sampled from the random distribution map of Figure 3(a) when generating the micro-dot coordinates (x, y).

[0039] Figure 4 A flow chart of a method 400 for verifying the authenticity of a product according to a second embodiment of the present invention is shown. In method 400, an image or picture of a product identification of a verified product is first obtained (step 401). For example, after purchasing a product, a user can take a photo of the product identification portion containing a barcode or a QR code, and transmit the image of the product identification obtained by taking the photo to a verification party or a verification device, so as to verify the authenticity of the obtained product identification image. In this embodiment, the processing of an image containing a QR code includes two parts, namely, a convolutional neural network algorithm processing part (including steps 402-405), and a micro-point processing part (including steps 406-409). In these two processing parts, the image needs to be preprocessed first (steps 402 and 406), for example, the partial area containing effective features (such as a QR code) in the image is subjected to light and dark adjustment, effective part interception, contrast enhancement, image sharpening, picture normalization and other common preprocessing methods in image processing technology.

[0040] In the convolutional neural network algorithm processing part, after the preprocessing step 402, the convolutional neural network algorithm is used to extract image features (step 403), wherein the image processed by the preprocessing step 402 is used as input, and the convolutional neural network is used as an overall algorithm module to finally output a feature vector, that is, the image containing the two-dimensional code is quantized into a feature vector. The feature vector may include k floating point numbers (for example, a floating point number sequence of 1x512), also known as a k-dimensional feature vector, which is used to describe the printing features, that is, the unique and subtle features of the printed product logo caused by the use of physical paper, ink, printing equipment, etc. during the printing process, which can be reflected in the image. In an embodiment of the present invention, the first classifier can be trained using positive samples and negative samples of multiple product logos and the printing features extracted therefrom. The pre-trained first classifier can analyze the product identification of the verified product based on the image features extracted in step 403 (step 404). For example, a convolutional neural network algorithm can be used to compare and analyze the extracted image features with the printed features of the positive sample and the printed features of the negative sample, thereby outputting the probability that the product identification of the verified product is true or false (step 405).

[0041] The convolutional neural network may include layers such as linear1, ReLU, Dropout(), Linear2, and Linear3. The Linear3 layer output is used in the end. Since we are only concerned with true and false classification here, the output is a 2D vector, where p1 represents the probability of being true and p2 represents the probability of being false.

[0042] The training uses the cross entropy loss function:

[0043] H(y,p)=-∑ i y i log(p i )

[0044] Among them, y is the true value of the target [y1,y2], the true label is [1.0,0.0], and the pseudo label is [0.0,1.0].

[0045] During the training process of the first classifier, the loss can be calculated based on the known true value of the sample. When the loss (‖y-y'‖, that is, the absolute value of the difference between the true value and the predicted result output by the classifier at this time) is judged to be less than the predetermined threshold, the convolutional neural network training is stopped; when the loss is judged to be greater than the predetermined threshold, the neural network parameters and the parameters of the first classifier can be updated according to the loss value. Then, the updated convolutional neural network and the updated first classifier continue to extract image features and judge their true and false probabilities. When the loss value continues to decrease and stabilizes to a relatively low loss value (i.e., the threshold), it can be considered that the first classifier has been trained.

[0046] Positive samples can be multiple QR code labels of genuine products, and negative samples can be copies of these positive samples obtained by various means. Negative samples have the same QR code labels, but their printed features differ in details from those of genuine QR code labels. After the convolutional neural network is initialized and before the first classifier is trained, the printed features of the product logo cannot be extracted. When there are differences in printed features between the positive and negative samples, and other image details are the same, these positive and negative samples can be used to train the convolutional neural network to identify the printed features of the positive samples and the printed features of the negative samples, such as the type of printed features (lines, colors or grayscale, etc.), location, degree of difference, etc. After continuous training, the convolutional neural network has the ability to quickly and accurately extract the printed features of the verified product logo.

[0047] In the micro-dot processing part, after the pre-processing step 406, the micro-dot extraction algorithm is used to extract the micro-dot features in the image (step 407), wherein the image processing technology can be used to read the randomly distributed micro-dot features in the area where the product logo to be verified is located, including statistical data based on at least one of the position, size, color or grayscale of the micro-dots. For example, by counting the size of each micro-dot area (such as the number of pixels contained in each micro-dot), or by counting the average RGB three-channel value of each area, the grayscale information of the micro-dot area in the image is obtained. Then, the corresponding micro-dot features of the authentic product logo pre-saved are extracted from the database, and the read micro-dot features are compared with the micro-dot features in the database (step 408), thereby outputting the result of the micro-dot feature comparison (step 409) as one of the bases for judging the authenticity of the product logo.

[0048] Then, the authenticity probability of the product identification outputted in step 405 and the result of the micro-dot feature comparison outputted in step 409 (where different statistical data are normalized) are used to form a description vector X (step 410). For example, the authenticity probability of the outputted product identification is used as a feature dimension such as x1. The result of the micro-dot feature comparison may include several statistical data obtained after the matching quantification processing of the micro-dot features in the comparison, such as the percentage of micro-dots found in the target two-dimensional code matching the micro-dots of the corresponding two-dimensional code in the database as x2, the statistical parameters of the pixel distance of the matched micro-dots from the micro-dots in the database in the image coordinate system (such as the mean and variance) as x3 and x4, the penalty for the micro-dots that are not matched (mismatch) as x5, etc. The above information can be used to form a 5-dimensional description vector about the verified product identification.

[0049] In an embodiment of the present invention, all collected positive and negative samples of product identification images can be processed to obtain corresponding authenticity probabilities and micro-point statistical features, and then obtain corresponding description vectors as sample data sets, of which a part (such as 80%) can be used as a training set for training the second classifier; the other part (such as 20%) can be used as a test set. Based on the sample data set, a relatively popular machine learning algorithm can be used to train and test the second classifier, and the selectable classifier types include support vector machine (SVM), boost tree (Boost Tree), decision tree, shallow neural network, k nearest neighbor algorithm, random forest, etc. The pre-trained second classifier can discriminate and classify the verified product identification based on the description features in the description vector obtained in step 410 (including: the authenticity probability of the verified product identification output by the first classifier, and the micro-point statistical features) (step 411), thereby outputting the authenticity determination result of the verified product identification (step 412).

[0050] In the process of preparing training image samples, you can use a variety of different mobile phones on the market to take photos of several authentic product logos under different lighting environments, and use the resulting images as positive samples; use a variety of different mobile phones on the market to take photos of the produced non-authentic labels under different lighting environments, and use the resulting images as negative samples; then, randomly divide the positive and negative samples into training sample sets and test sample sets according to the proportion.

[0051] The first classifier and the second classifier can be continuously trained using the accumulated product identification samples, so as to provide more accurate authenticity discrimination.

[0052] FIG. 5A to FIG. 5C Three examples of convolutional neural network structures used in the method for extracting image features of product logos in an embodiment of the present invention are shown. Figure 5A Showing the VGG network, Figure 5B Shows the ResNet network structure, Figure 5C The Inception network structure is shown, and three ways of extracting image features using a convolutional neural network are shown respectively. However, the present invention is not limited to these three network structures.

[0053] exist FIG. 5A to FIG. 5CIn the figure, 501 represents the input layer, in which the image of the product logo after preprocessing is input. 502 represents the convolution layer, which is used to extract features from the input image data. It contains multiple convolution kernels, and each element of the convolution kernel corresponds to a weight coefficient and a bias, which is similar to a neuron in a feedforward neural network. Each neuron in the convolution layer is connected to multiple neurons in the area with a similar position in the previous layer. In the convolution layer, features are extracted for each small area in the input image, and convolution is performed using multiple filters to obtain multiple feature maps. 503 represents the pooling layer. After the convolution layer 502 performs feature extraction, the output feature map will be passed to the pooling layer 503 for feature selection and information filtering. The pooling layer 503 contains a preset pooling function, and its function is to replace the result of a single point in the feature map with the feature map statistics of its adjacent area. 504 represents the fully connected layer, which is equivalent to the hidden layer in the traditional feedforward neural network. The fully connected layer is located at the last part of the hidden layer of the convolutional neural network and only transmits signals to other fully connected layers. The function of the fully connected layer is to perform nonlinear combination of the extracted features to obtain the output. The fully connected layer itself does not have the ability to extract features, but attempts to use the existing high-order features to achieve the learning goal. 505 represents a residual network module, which includes a combination of multiple convolutional layers with jump connections as the building unit of the ResNet network structure. 506 represents an Inception module, which is a hidden layer constructed by stacking multiple convolutional layers and pooling layers. Specifically, an Inception module will contain multiple different types of convolution and pooling operations at the same time, and use the same padding to make the above operations obtain feature maps of the same size, and then superimpose the channels of these feature maps in the array and pass through the excitation function. 510 represents an output layer, which outputs the image features extracted by the convolutional neural network.

[0054] The device for verifying the authenticity of a product provided according to an embodiment of the present invention may include: a memory for storing instructions; and a processor coupled to the memory, and the processor may execute the method according to the above embodiment of the present invention when executing the stored instructions. The memory may also store a database, which contains micro-dot features of the authentic product identification saved during or after the production of the product. The micro-dot features may include at least one of the shape features, position features, grayscale features, and color features of the micro-dots.

[0055] The memory of this embodiment may also store a sample library, which includes a plurality of genuine product identification images as positive samples and a plurality of counterfeit product identification images as negative samples. The processor is configured to use at least a portion of the samples in the sample library to train a convolutional neural network, and a first classifier and a second classifier for verifying product identification.

[0056] Figure 6The structure block diagram of the device 600 for verifying the authenticity of a product according to an embodiment of the present invention is shown. The device 600 comprises: a micro-dot feature extraction module 601, which is used to extract micro-dot features on the product logo from the image of the product logo of the verified product; an image feature extraction module 602, which is used to extract image features of at least a part of the product logo from the image using a machine learning algorithm; and a verification module 603, which is used to verify the authenticity of the product logo of the verified product based on the extracted micro-dot features and image features. Figure 6 The device 600 shown can be implemented by software, hardware or a combination of software and hardware, and can be designed to include corresponding modules to implement the above-mentioned various method embodiments for verifying the authenticity of a product of the present invention.

[0057] The product verification method and device provided in accordance with the above-mentioned embodiments of the present invention, using a combination of product identification including, for example, a QR code or a barcode, micro-dot features, and image features, greatly improves the accuracy of verification of authentic products, allowing users who purchase products to use various mobile phones or cameras under various lighting conditions to capture product identification images and perform accurate verification.

[0058] The embodiments of the present invention disclosed above are exemplary rather than restrictive. Those skilled in the art should understand that the embodiments disclosed above can be modified, altered and changed in various ways without departing from the essence of the invention, and these modifications, alterations and changes should fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for verifying the authenticity of a product, wherein the product logo has randomly distributed micro dots, the method include: Extracting micro-dot features on the product logo from an image of the product logo of the product to be verified; Extracting image features of at least a portion of the product logo from the image using a machine learning algorithm, wherein the extracted image features at least include printing features related to printing of at least a portion of the product logo, wherein the printing features refer to unique features generated during the printing process of the printed product logo and can be reflected in the image; and Verifying the authenticity of the product identification of the verified product based on the extracted micro-dot features and the image features; Wherein, verifying the authenticity of the product identification of the verified product includes: A classifier trained by a machine learning algorithm is used to verify the authenticity of the product identification of the verified product. The classifier includes a first classifier and a second classifier. The first classifier distinguishes the authenticity of the product identification of the verified product based on the printed features, and the second classifier determines the authenticity of the product identification based on the authenticity probability of the product identification of the verified product output by the first classifier and the micro-dot features.

2. The method according to claim 1, in, The machine learning algorithm used to extract the image features is a convolutional neural network; the machine learning algorithm used to train the classifier is a machine learning algorithm that can classify feature vectors.

3. According to the method of claim 2, further include: The convolutional neural network and the classifier are trained by using a plurality of authentic identification images as positive samples and a plurality of counterfeit identification images as negative samples.

4. The method according to claim 3, in, The step of extracting the image features comprises: The image features are extracted from the image using a trained convolutional neural network to output a feature vector describing the image features.

5. The method according to claim 1; in, Using positive samples and negative samples of product identification to train the first classifier; Wherein, verifying the authenticity of the product identification of the verified product includes: Based on the printed features, using the trained first classifier to output a probability that the product identification of the verified product is true; and / or Based on the printed features, the trained first classifier is used to output a probability that the product identification of the verified product is fake.

6. The method according to claim 1, in, The printing feature is a feature associated with at least one of paper, ink, and printing equipment used in the printing process of the product logo of the positive sample.

7. The method according to claim 5, in, Using positive samples and negative samples of product identification to train the second classifier; Wherein, verifying the authenticity of the product identification of the verified product further includes: comparing the extracted micro-dot features with micro-dot features of the product identification that are pre-saved during or after the production of the anti-counterfeit product; Based on the comparison result and the authenticity probability of the product identification output by the first classifier, a description vector about the product identification is formed; and Based on the description vector, the second classifier is used to determine the authenticity of the product identification of the verified product.

8. The method according to claim 7, in, The description vector includes data related to at least one of the following: a matching rate between the extracted micro-dot features and pre-saved micro-dot features, statistical parameters of the pixel distance of the matched micro-dots from the pre-saved micro-dots in the image coordinate system, the number of micro-dots in the image of the product logo of the verified product that do not match the pre-saved micro-dot features, and the quality of the image.

9. The method according to claim 1, in, The step of extracting the micro-point features comprises: At least one of the shape feature, position feature, grayscale feature and color feature of the micro-dot is extracted from the image using image processing technology.

10. The method according to claim 1, in, The product identification includes at least one of a bar code and a two-dimensional graphic code.

11. A device for verifying the authenticity of a product, wherein the product logo has randomly distributed micro dots, and the device include: A micro-dot feature extraction module, used to extract micro-dot features on the product logo from the image of the product logo of the verified product; an image feature extraction module, configured to extract image features of at least a portion of the product logo from the image using a machine learning algorithm, wherein the extracted image features at least include printing features associated with printing of at least a portion of the product logo, wherein the printing features refer to unique features generated during the printing process of the printed product logo and which can be reflected in the image; and A verification module, used to verify the authenticity of the product identification of the verified product based on the extracted micro-dot features and the image features; Wherein, the verification module is further used for: A classifier trained by a machine learning algorithm is used to verify the authenticity of the product identification of the verified product. The classifier includes a first classifier and a second classifier. The first classifier distinguishes the authenticity of the product identification of the verified product based on the printed features, and the second classifier determines the authenticity of the product identification based on the authenticity probability of the product identification of the verified product output by the first classifier and the micro-dot features.

12. A device for verifying the authenticity of a product, wherein the product logo has randomly distributed micro dots, and the device include: a memory for storing instructions; as well as A processor coupled to the memory, the instructions, when executed by the processor, cause the processor to perform the method according to any one of claims 1 to 10.

13. The device according to claim 12, in, The memory also stores a database; the database contains micro-dot features of product identification saved during or after the production of the product, and the micro-dot features include: at least one of the shape features, position features, grayscale features, and color features of the micro-dots.

14. The device according to claim 12, in, The memory also stores a sample library, which includes a plurality of authentic product identification images as positive samples and a plurality of counterfeit product identification images as negative samples; the processor is configured to use at least a portion of the samples in the sample library to train a convolutional neural network, and a first classifier and a second classifier for verifying product identification.

15. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a computer, cause the computer to perform the method according to any one of claims 1 to 10.

Images