Computer vision-based automatic identification system and method for infringing and counterfeit goods
Through an automatic identification system based on computer vision, combining brand identification and workmanship details, infringing and counterfeit products are identified, the problem of insufficient identification efficiency and accuracy in the existing technology is solved, efficient identification of authenticity and false goods is achieved, and the brand and Consumer safety.
Patent Information
- Application Number
- CN202410951992.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-16
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-07-16
AI Technical Summary
The existing technology is difficult to effectively identify and distinguish infringing and counterfeit goods, resulting in damage to brand image, threats to consumers' health and safety, and insufficient identification efficiency and accuracy.
An automatic recognition system based on computer vision is used to extract high-dimensional spatial features of product images, and combine the details of brand logos and the quality of workmanship details to determine whether the product is an infringing counterfeit product. The system includes a module for product image collection, feature encoding and identification result generation, and uses deep learning and neural network technology to improve recognition efficiency and accuracy.
It realizes the automated identification of infringing and counterfeit products, improves identification efficiency and accuracy, reduces misjudgments and misjudgments, and protects the brand image and consumer health and safety.
Smart Images

Figure CN118552791B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intellectual property protection, and more specifically, to a computer vision-based automatic identification system and method for infringing and counterfeit goods. Background Art
[0002] Infringing and counterfeit goods refer to goods that infringe upon the legitimate rights and interests of others and impersonate others' products. For regular brands, the existence of infringing and counterfeit goods is not only an economic loss, but also a serious erosion of brand image and reputation. When the market is flooded with a large number of low-quality infringing and counterfeit products, consumers may have a negative impression of the entire brand because they cannot distinguish between the real and the fake, which will lead to a decline in brand value. In the long run, it will weaken the brand's market competitiveness and consumer loyalty. Infringing and counterfeit products are often of poor quality, may contain harmful ingredients or fail to meet safety standards, posing a threat to consumer health.
[0003] Therefore, there is a need for an automatic identification solution for infringing and counterfeit goods based on computer vision. Summary of the invention
[0004] In view of the shortcomings of the prior art, this application provides a computer vision-based automatic identification system and method for infringing and counterfeit goods, which uses computer vision technology to extract features in high-dimensional space from the image of the target identification product to obtain the production workmanship detail quality features and brand logo detail features of the target identification product, and combines these two features to determine whether the target identification product is an infringing and counterfeit product. In this way, the automatic identification of infringing and counterfeit goods can be achieved, which is conducive to improving the efficiency and accuracy of identification.
[0005] According to one aspect of the present application, a computer vision-based automatic identification system for infringing and counterfeit goods is provided, which includes:
[0006] A product image collection module is used to obtain images of target identification products;
[0007] A commodity image feature encoding module, used for performing feature encoding on the image of the target identification commodity to obtain a target identification commodity feature map;
[0008] The commodity identification result generating module is used to judge whether the target identification commodity is an infringing or counterfeit commodity based on the target identification commodity feature map.
[0009] According to another aspect of the present application, a method for automatically identifying infringing and counterfeit goods based on computer vision is provided, which includes:
[0010] Acquire an image of a target identification product;
[0011] Performing feature encoding on the image of the target identification commodity to obtain a target identification commodity feature map;
[0012] Based on the target identification commodity feature map, determine whether the target identification commodity is an infringing or counterfeit commodity.
[0013] This application has significant technical effects due to the adoption of the above technical solutions:
[0014] The computer vision-based automatic identification system and method for infringing and counterfeit goods provided in this application utilize computer vision technology to extract features in a high-dimensional space from the image of the target identification product to obtain the production workmanship detail quality features and brand logo detail features of the target identification product, and combine these two features to determine whether the target identification product is an infringing and counterfeit product. In this way, the automatic identification of infringing and counterfeit goods can be achieved, which is conducive to improving the efficiency and accuracy of identification. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0016] Figure 1 4 is a system block diagram of a computer vision-based automatic identification system for infringing and counterfeit goods according to an embodiment of the present application.
[0017] Figure 2 4 is a block diagram of a commodity image feature encoding module in a computer vision-based automatic identification system for infringing and counterfeit commodities according to an embodiment of the present application.
[0018] Figure 3 The block diagram is a unit for generating product brand identification detail features in a computer vision-based automatic identification system for infringing and counterfeit products according to an embodiment of the present application.
[0019] Figure 4 The block diagram is a module for generating commodity identification results in a computer vision-based automatic identification system for infringing and counterfeit commodities according to an embodiment of the present application.
[0020] Figure 5 The present invention is a flowchart of a method for automatically identifying infringing and counterfeit goods based on computer vision according to an embodiment of the present application.
[0021] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.
[0023] As mentioned in the above background technology, infringing and counterfeit goods refer to goods that infringe the legitimate rights and interests of others and impersonate others' products. For regular brands, the existence of infringing and counterfeit goods profoundly and widely damages their carefully constructed brand image and valuable reputation. Take well-known clothing brands as an example. These brands often win the favor and trust of consumers with their unique design concepts, excellent quality assurance and profound cultural heritage. However, when a large number of crudely made and fake counterfeit products are flooding the market, this foundation of trust begins to shake. In the process of purchasing, consumers will inevitably have doubts when faced with goods that are difficult to distinguish between true and false, worrying that they will spend a lot of money but buy products that are not worthy of the name. This uncertainty not only hurts consumers' feelings, but also weakens their perception of the positive image of the brand. In the long run, the image of high quality, high-end and innovation represented by the brand will be greatly discounted, and it may even be labeled as "fakes rampant", seriously damaging the brand's intangible assets.
[0024] What is more serious is that counterfeit goods directly encroach on the market share of genuine brands through low-price competition strategies. These counterfeit products often do not have formal production costs, quality control and after-sales service systems, so they can be sold at prices far lower than genuine products, attracting some price-sensitive consumers. This unfair competition method not only makes genuine brands lose their due sales opportunities, but may also cause their sales to decline and profits to decrease, which in turn affects the brand's R&D investment, market expansion and supply chain stability, forming a vicious circle. In addition, due to the lack of strict quality control in the production process, the use of inferior materials and processes, and even may contain harmful ingredients or fail to meet safety standards, these quality problems not only damage the economic interests of consumers, but also pose a direct threat to their health and safety. Therefore, an automatic identification solution for infringing and counterfeit goods based on computer vision is expected.
[0025] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for the automatic identification of infringing and counterfeit goods based on computer vision.
[0026] Figure 1FIG. 1 is a system block diagram of an automatic identification system for infringing and counterfeit goods based on computer vision according to an embodiment of the present application. Figure 1 As shown, in the computer vision-based infringing and counterfeit goods automatic identification system 100, it includes: a product image collection module 110, used to obtain the image of the target identification product; a product image feature encoding module 120, used to feature encode the image of the target identification product to obtain a target identification product feature map; a product identification result generation module 130, used to determine whether the target identification product is an infringing and counterfeit product based on the target identification product feature map.
[0027] Specifically, in the technical solution of the present application, the image of the target identification product is first obtained. It should be well understood that the target identification product image contains brand logo information. Most products, especially well-known brand products, will have a brand logo printed on a prominent position on the product. This logo is an important symbol of brand identity and has a unique design and characteristics. Counterfeit products often try to imitate this logo, but due to technology, cost or other reasons, the imitated logo often has slight differences from the genuine product. The target identification product image also contains workmanship details information. The workmanship details of the product, such as stitching, pattern alignment, color uniformity, etc., are important indicators for measuring the quality of the product. Genuine products usually have higher standards in these aspects, while counterfeit products often reveal their flaws in these details due to different production processes and raw materials. In general, the brand logo information and workmanship details information contained in the image of the target identification product are important bases for identifying infringing counterfeit products. Through computer vision technology and machine learning algorithms, this information can be effectively used to improve the accuracy and efficiency of recognition. The image data of the target identification product can be obtained from some e-commerce platforms.
[0028] Figure 2 FIG. 1 is a block diagram of a commodity image feature encoding module in a computer vision-based automatic identification system for infringing and counterfeit commodities according to an embodiment of the present application. Figure 2 As shown, the commodity image feature encoding module 120 includes: a commodity brand logo detail feature generation unit 121, which is used to extract the features of the brand logo in the image of the target identification commodity to obtain a brand logo detail feature map; a commodity workmanship detail quality feature generation unit 122, which is used to extract the features of the workmanship details in the image of the target identification commodity to obtain a workmanship detail quality feature map; a commodity multi-angle identification feature fusion unit 123, which is used to fuse the brand logo detail feature map and the workmanship detail quality feature map to obtain the target identification commodity feature map.
[0029] Figure 3FIG. 1 is a block diagram of a product brand identification detail feature generation unit in a computer vision-based infringing and counterfeit product automatic identification system according to an embodiment of the present application. Figure 3 As shown, the product brand logo detail feature generation unit 130 includes: a brand logo area selection subunit 131, which is used to pass the image of the target identification product through a brand logo target detection network to obtain a brand logo area of interest map; a brand logo feature extraction subunit 132, which is used to pass the brand logo area of interest map through a brand logo feature encoder to obtain the brand logo detail feature map.
[0030] Next, the image of the target identification product is passed through the brand logo target detection network to obtain a brand logo area of interest map. The brand logo is an important identifier of the product identity and has unique visual features. By inputting the image into the brand logo target detection network, the brand logo in the image can be accurately located and identified. This positioning capability is crucial for subsequent identification of infringing and counterfeit goods because it allows key areas in the image to be quickly found and focused on. In addition, directly identifying infringement and counterfeiting products on the entire image may consume a lot of computing resources and time. By first locating the logo in the image through the brand logo target detection network, the amount of data processed subsequently can be greatly reduced, thereby improving the overall recognition efficiency. The brand logo target detection network in this application is an anchor window-based target detection network.
[0031] Anchor is a commonly used concept in target detection. It refers to a series of rectangular boxes (or anchor boxes) of fixed size and shape preset on the image. These anchor boxes are used as candidate regions for target detection. By predicting the offset between the anchor box and the true target box, the specific position of the target in the image can be obtained. The anchor window matches and regresses the target box through the preset anchor box to determine the position and size of the target. This method can cover targets of different scales and shapes in the image and improve the flexibility of detection. Anchor-free does not rely on the preset anchor box, but directly predicts the position of the target center point and other geometric properties of the bounding box (such as size and direction). This method simplifies the detection process, but may require more training data and computing resources when dealing with complex scenes.
[0032] In an embodiment of the present application, one possible implementation method of passing the image of the target identification product through a brand logo target detection network to obtain a brand logo region of interest map may be: using the brand logo target detection network to process the image of the target identification product using the following formula to obtain the brand logo region of interest map;
[0033] Wherein, the formula is: ;in, is a map of the area of interest of the brand logo, is the anchor frame, an image identifying the item for the target, Indicates classification, Indicates regression.
[0034] Then, the brand logo area of interest map is passed through the brand logo feature encoder to obtain a brand logo detail feature map. The brand logo feature encoder can automatically extract high-level features from the brand logo area of interest map, which can better represent the uniqueness and difference of the logo than the original pixel value. These high-level features include but are not limited to shape, texture, color distribution, etc. In this application, the brand logo feature encoder is a convolutional neural network model using convolution kernels with different void rates. It extracts and transforms the input brand logo area of interest map through a series of convolution layers, pooling layers, activation functions and other components, and finally outputs a feature map containing high-level features. Compared with using ordinary convolution kernels to extract features from the brand logo area of interest map, convolution kernels with different void rates can capture the features of the brand logo at different scales. Smaller void rates can focus on detail information, while larger void rates can capture more extensive structural information. This multi-scale feature extraction capability helps to understand the features of the logo more comprehensively, thereby improving the accuracy of recognition and judgment.
[0035] In an embodiment of the present application, a possible implementation method of passing the brand logo area of interest map through a brand logo feature encoder to obtain a brand logo detail feature map may be: using each layer of the first convolutional neural network model in the brand logo feature encoder to respectively perform convolution processing based on the first void rate void convolution kernel, pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward pass of the layer so that the first scale brand logo detail feature map is output by the last layer of the first convolutional neural network model in the brand logo feature encoder; using each layer of the second convolutional neural network model in the brand logo feature encoder to respectively perform convolution processing based on the second void rate void convolution kernel, pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward pass of the layer so that the second scale brand logo detail feature map is output by the last layer of the second convolutional neural network model in the brand logo feature encoder; fusing the first scale brand logo detail feature map and the second scale brand logo detail feature map to obtain the brand logo detail feature map.
[0036] Workmanship details are one of the important bases for judging the authenticity of goods. Counterfeit goods often have significant differences from genuine goods in terms of manufacturing process, material usage, etc. In order to capture these detail differences, the image of the target identification product is passed through a workmanship detail quality feature encoder to obtain a workmanship detail quality feature map. Through the workmanship detail quality feature encoder, you can focus on the key details in the image, such as sutures, pattern alignment, etc. In this application, the workmanship detail quality feature encoder is a convolutional neural network model using a spatial attention mechanism. The spatial attention mechanism is a method that imitates the visual attention mechanism in the human visual system. When processing an image, it assigns different importance weights to different spatial positions in the image, so that the model can focus on key areas for information extraction and processing. Specifically, the spatial attention mechanism enhances or suppresses the feature representation of the corresponding position or area by calculating the attention weight of each position or area in the image. And because the spatial attention mechanism enables the model to focus on the key areas in the image, it reduces the amount of calculation on irrelevant areas and improves the efficiency of feature extraction. Through weighted processing, the spatial attention mechanism can enhance the feature representation of key areas, making them more prominent and significant, which helps to understand and use these features more accurately in the future.
[0037] In an embodiment of the present application, a possible implementation method of passing the image of the target identification product through a workmanship detail quality feature encoder to obtain a workmanship detail quality feature map can be: using each layer of the workmanship detail quality feature encoder to perform the following on the input data in the forward pass of the layer: convolution processing on the input data based on the convolution kernel to obtain a convolution feature map; passing the convolution feature map through a spatial attention module to obtain a spatial attention score matrix; multiplying the spatial attention score matrix and each feature matrix of the convolution feature map along the channel dimension by the spatial attention feature map at each position point; performing pooling processing on the spatial attention feature map based on a local feature matrix to obtain a pooling feature map; performing nonlinear activation on the pooling feature map to obtain an activation feature map; wherein the input of the first layer of the workmanship detail quality feature encoder is the image of the target identification product, and the output of the last layer of the workmanship detail quality feature encoder is the workmanship detail quality feature map.
[0038] The brand logo detail feature map mainly contains the detailed feature information of the brand logo on the product, while the workmanship detail quality feature map focuses on the details of the product such as the stitching and texture. In order to capture the feature information of the product more comprehensively, the brand logo detail feature map and the workmanship detail quality feature map are integrated in the technical solution of this application to obtain the target identification product feature map. The brand logo detail feature map and the workmanship detail quality feature map complement each other when identifying infringing and counterfeit goods, and together constitute a complete basis for judging the authenticity of the goods. The brand logo feature provides the brand ownership information of the product, while the workmanship detail quality feature further verifies the quality and production process of the product. Only when both meet the standards of genuine products can the product be judged as authentic; otherwise, there may be a risk of infringement and counterfeiting. That is, the fusion of the two feature maps can comprehensively consider the information of both brand logo and workmanship details, reduce the misjudgment and missed judgment that may be caused by a single feature, and thus improve the accuracy of recognition.
[0039] In an embodiment of the present application, a possible implementation method of fusing the brand logo detail feature map and the workmanship detail quality feature map to obtain a target identification commodity feature map may be: fusing the brand logo detail feature map and the workmanship detail quality feature map using the following fusion formula to obtain the target identification commodity feature map; wherein the fusion formula is: ;in, identifying a product feature map for the target, is the brand logo detail feature map, This is the workmanship detail quality feature diagram. It represents the addition of the elements at the corresponding positions of the brand logo detail feature map and the workmanship detail quality feature map, and It is a weighting parameter used to control the balance between the brand logo detail feature map and the workmanship detail quality feature map in the target identification product feature map.
[0040] Figure 4 FIG. 1 is a block diagram of a commodity identification result generation module in a computer vision-based infringing and counterfeit commodity automatic identification system according to an embodiment of the present application. Figure 4 As shown, the commodity identification result generation module includes: a target identification commodity feature map dimensionality reduction unit 131, which is used to expand the target identification commodity feature map to obtain a target identification commodity feature vector; a target identification commodity feature vector optimization unit 132, which is used to reconstruct the target identification commodity feature vector in a high-dimensional feature space based on the self-correlation dimension to obtain an optimized target identification commodity feature vector; a feature vector classification unit 133, which is used to pass the optimized target identification commodity feature vector through a commodity classifier to obtain a classification result, and the classification result is used to indicate whether the target identification commodity is an infringing or counterfeit commodity.
[0041] Next, the target identification commodity feature map is expanded to obtain a target identification commodity feature vector. It should be understood that the target identification commodity feature map is high-dimensional data, which will increase the complexity of calculation when performing model calculation and training. In order to simplify the calculation process and improve the processing speed, it is necessary to reduce the dimension of the target identification commodity feature map, that is, expand the three-dimensional feature map into a one-dimensional feature vector. It is also worth mentioning that in the process of expanding the feature map into a feature vector, the key features of the commodity contained therein are not lost.
[0042] In particular, the logo designs of different brands are diverse. Even for the same brand, the size, color, shape, etc. of the logo may be different on different products, which will cause local differences when the feature encoder processes it. At the same time, the workmanship details of the product vary depending on factors such as production batches and process levels. The slight changes in these details will be reflected in the workmanship detail feature map. When the brand logo detail feature map and the workmanship detail quality feature map are fused, the weight distribution and fusion method of different features will affect the distribution of the final feature vector, resulting in local distribution differences in the target identification product vector. Due to the local distribution differences of the feature vector, the product classifier may be biased towards certain specific features when classifying, rather than fully considering all features, resulting in a deviation in the class harmony of the predetermined class label. Therefore, it is necessary to reconstruct the target identification product feature vector in a high-dimensional feature space based on the self-correlation dimension.
[0043] Specifically, in an embodiment of the present application, the target identification product feature vector is reconstructed in a high-dimensional feature space based on the autocorrelation dimension to obtain an optimized target identification product feature vector, including: calculating a target identification product feature autocorrelation association matrix between the target identification product feature vector and its transposed vector; matrix multiplying a weight parameter matrix of the product classifier with the target identification product feature autocorrelation association matrix to obtain a weight parameter modulated classification autocorrelation association matrix; calculating the correlation perception factor between any two row vectors in the weight parameter modulated classification autocorrelation association matrix to obtain a metric correlation perception matrix; multiplying the metric correlation perception matrix with the target identification product feature vector to obtain the optimized target identification product feature vector.
[0044] In an embodiment of the present application, another expression of reconstructing the target identification commodity feature vector in a high-dimensional feature space based on the self-correlation dimension to obtain an optimized target identification commodity feature vector may be: processing the target identification commodity feature vector with the following formula to obtain the optimized target identification commodity feature vector; wherein the formula is: ;in, is the target identification product feature vector, represents the transposed vector, is the target identification commodity feature autocorrelation matrix, is the weight parameter matrix of the commodity, is the weight parameter modulated classification autocorrelation matrix, The weight parameter modulates the classification autocorrelation matrix. row vectors, The weight parameter modulates the classification autocorrelation matrix. row vectors, is the total number of rows in the autocorrelation matrix of the classification modulated by the weight parameter, yes and The correlation perception coefficient between is the bi-norm of the eigenvector, is the metric correlation perception matrix, is the set of real numbers, is the optimization target identification product feature vector, Represents matrix multiplication.
[0045] In the technical solution of the present application, it is considered that there are local distribution differences in the feature vectors of the target identification products, which leads to the problem that when the product is classified and judged, there will be a class harmony offset relative to the predetermined class label of the classifier, which affects the classification iteration effect. Based on this, in the technical solution of the present application, based on the autocorrelation dimension of the feature vectors of the target identification products, the spatial sub-dimensional complexity of the high-dimensional feature space of the feature distribution is used to perform association expansion to obtain the autocorrelation association matrix of the target identification product features. Furthermore, the target identification product feature autocorrelation association matrix is subjected to metric correlation perception based on weight learning parameters. By introducing weight learning parameters, the local distribution differences of different target identification product feature vectors can be effectively quantified and adjusted, thereby reducing the class harmony offset caused by these local distribution differences when the product is classified and judged.
[0046] Finally, the optimized target identification product feature vector is passed through the product classifier to obtain a classification result, and the classification result is used to indicate whether the target identification product is an infringing or counterfeit product. The optimized target identification product feature vector has already contained rich and representative product feature information through the previous processing steps (such as the fusion of the workmanship detail quality feature encoder and the brand logo detail feature map). However, this information still needs to be further analyzed and interpreted to draw a conclusion on whether the product is infringing or counterfeit. The product classifier can map complex feature vectors to specific classification results, that is, to determine whether the product is an infringing or counterfeit product. The product classifier is a machine learning model that receives the feature vector of the product as input and outputs a classification label to represent the attributes of the product, that is, "authentic" or "counterfeit". The classification result can directly tell the user whether it is an infringing or counterfeit product, providing an important basis for subsequent decision-making and actions.
[0047] In an embodiment of the present application, a possible implementation method for passing the optimized target identification commodity feature vector through a commodity classifier to obtain a classification result indicating whether the target identification commodity is an infringing or counterfeit commodity may be: using the fully connected layer of the commodity classifier to fully connect encode the optimized target identification commodity feature vector to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the Softmax classification function of the commodity classifier to obtain the classification result.
[0048] In summary, the computer vision-based infringing and counterfeit goods automatic identification system 100 according to the embodiment of the present application is explained, which uses computer vision technology to extract features of the target identification product in a high-dimensional space to obtain the production workmanship detail quality features and brand logo detail features of the target identification product, and combines these two features to determine whether the target identification product is an infringing and counterfeit product. In this way, the automatic identification of infringing and counterfeit goods can be achieved, which is conducive to improving the efficiency and accuracy of identification.
[0049] As described above, the computer vision-based automatic identification system 100 for infringing and counterfeit goods according to the embodiment of the present application can be implemented in various terminal devices, such as a server for automatic identification of infringing and counterfeit goods based on computer vision. In one example, the computer vision-based automatic identification system 100 for infringing and counterfeit goods according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the computer vision-based automatic identification system 100 for infringing and counterfeit goods can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the computer vision-based automatic identification system 100 for infringing and counterfeit goods can also be one of the many hardware modules of the terminal device.
[0050] Alternatively, in another example, the computer vision-based infringing and counterfeit goods automatic identification system 100 and the terminal device may also be separate devices, and the computer vision-based infringing and counterfeit goods automatic identification system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.
[0051] Figure 5 FIG. 1 is a flow chart of a method for automatically identifying infringing and counterfeit goods based on computer vision according to an embodiment of the present application. Figure 5 As shown, in the automatic identification method of infringing and counterfeit goods based on computer vision, it includes: S110, obtaining an image of the target identification product; S120, feature encoding the image of the target identification product to obtain a target identification product feature map; S130, based on the target identification product feature map, judging whether the target identification product is an infringing and counterfeit product.
[0052] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned method for automatically identifying infringing and counterfeit goods based on computer vision have been described in the above reference. Figures 1 to 4 It has been introduced in detail in the description of the computer vision-based automatic identification system for infringing and counterfeit goods, and therefore, its repeated description will be omitted.
[0053] In summary, the automatic identification method of infringing and counterfeit goods based on computer vision in the embodiment of the present application is explained, which uses computer vision technology to extract features in high-dimensional space from the image of the target identification product to obtain the production workmanship detail quality features and brand logo detail features of the target identification product, and combines these two features to determine whether the target identification product is an infringing and counterfeit product. In this way, the automatic identification of infringing and counterfeit goods can be achieved, which is conducive to improving the efficiency and accuracy of identification.
[0054] Below, reference Figure 6 To describe an electronic device according to an embodiment of the present application. Figure 6 1 is a block diagram of an electronic device 10 according to an embodiment of the present application. For example, the electronic device 10 may be provided as a server or a terminal device. Figure 6 The electronic device 10 includes a processing component 11, which further includes one or more processors, and a memory resource represented by a memory 12 for storing instructions executable by the processing component 11, such as an application. The application stored in the memory 12 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 11 is configured to execute instructions to perform the above method.
[0055] The electronic device 10 may further include a power supply component 13 configured to perform power management of the electronic device 10, a wired or wireless network interface 14 configured to connect the electronic device 10 to a network, and an input / output (I / O) interface 15. The electronic device 10 may operate based on an operating system stored in the memory 12, such as Windows Server™, Mac OSX™, Unix™, Linux™, FreeBSD™ or the like.
[0056] In an embodiment of the present application, a computer-readable storage medium is also provided, such as a memory 12 including computer program instructions. The computer program instructions can be executed by a processing component 11 of an electronic device 10 to complete the above method.
[0057] The present application may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer program instructions for causing a processor to implement various aspects of the present application.
[0058] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.
[0059] The computer program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer program instructions from the network and forwards the computer program instructions for storage in a computer-readable storage medium in each computing / processing device.
[0060] The computer program instructions for performing the operation of the present application can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The computer program instructions can be executed completely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet). In some embodiments, by using the state information of the computer program instructions to personalize an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA) or a programmable logic array (PLA), the electronic circuit can execute the computer program instructions, thereby realizing various aspects of the present application.
[0061] Various aspects of the present application are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each box in the flowchart and / or block diagram and the combination of boxes in the flowchart and / or block diagram can be implemented by computer program instructions.
[0062] These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.
[0063] Computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.
Claims
1. A computer vision-based automatic identification system for infringing and counterfeit goods, characterized in that: include: A product image collection module is used to obtain images of target identification products; A commodity image feature encoding module, used for performing feature encoding on the image of the target identification commodity to obtain a target identification commodity feature map; A commodity identification result generating module, used to determine whether the target identification commodity is an infringing or counterfeit commodity based on the target identification commodity feature map; Wherein, the commodity identification result generating module includes: A target identification commodity feature graph dimension reduction unit, used for expanding the target identification commodity feature graph to obtain a target identification commodity feature vector; A target identification commodity feature vector optimization unit, used for reconstructing the target identification commodity feature vector in a high-dimensional feature space based on the self-correlation dimension to obtain an optimized target identification commodity feature vector; A feature vector classification unit, used for passing the optimized target identification commodity feature vector through a commodity classifier to obtain a classification result, wherein the classification result is used to indicate whether the target identification commodity is an infringing or counterfeit commodity; Wherein, the target identification commodity feature vector optimization unit is used to: Calculate the target identification product feature autocorrelation matrix between the target identification product feature vector and its transposed vector; Performing matrix multiplication of the weight parameter matrix of the commodity classifier and the target identification commodity feature autocorrelation matrix to obtain a weight parameter modulation classification autocorrelation matrix; Calculating the correlation perception factor between any two row vectors in the weight parameter modulation classification autocorrelation association matrix to obtain a metric correlation perception matrix; The metric correlation perception matrix is multiplied by the target identification commodity feature vector to obtain the optimized target identification commodity feature vector.
2. The computer vision-based automatic identification system for infringing and counterfeit goods according to claim 1 is characterized in that: The commodity image feature encoding module comprises: A product brand logo detail feature generation unit, used for extracting features of the brand logo in the image of the target identification product to obtain a brand logo detail feature map; A product workmanship detail quality feature generation unit, used to extract features of the workmanship details in the image of the target identification product to obtain a workmanship detail quality feature map; The commodity multi-angle identification feature fusion unit is used to fuse the brand logo detail feature map and the workmanship detail quality feature map to obtain the target identification commodity feature map.
3. The computer vision-based automatic identification system for infringing and counterfeit goods according to claim 2 is characterized in that: The commodity brand identification detail feature generating unit includes: A brand logo region selection subunit, used for passing the image of the target identification product through a brand logo target detection network to obtain a brand logo region of interest map; The brand logo feature extraction subunit is used to pass the brand logo region of interest map through a brand logo feature encoder to obtain the brand logo detail feature map.
4. The computer vision-based automatic identification system for infringing and counterfeit goods according to claim 3 is characterized in that: The commodity workmanship detail quality feature generation unit is used to: pass the image of the target identification commodity through a workmanship detail quality feature encoder to obtain the workmanship detail quality feature map.
5. The computer vision-based automatic identification system for infringing and counterfeit goods according to claim 4 is characterized in that: The brand logo target detection network is an anchor window-based target detection network, and the brand logo feature encoder is a convolutional neural network model using convolution kernels with different void rates.
6. The computer vision-based automatic identification system for infringing and counterfeit goods according to claim 5 is characterized in that: The workmanship detail quality feature encoder is a convolutional neural network model using a spatial attention mechanism.
7. A method for automatically identifying infringing and counterfeit goods based on computer vision, characterized in that: include: Acquire an image of a target identification product; Performing feature encoding on the image of the target identification commodity to obtain a target identification commodity feature map; Based on the target identification commodity feature map, determining whether the target identification commodity is an infringing or counterfeit commodity; Wherein, judging whether the target identification commodity is an infringing or counterfeit commodity based on the target identification commodity feature map includes: Expanding the target identification commodity feature graph to obtain a target identification commodity feature vector; Reconstructing the target identification commodity feature vector in a high-dimensional feature space based on the autocorrelation dimension to obtain an optimized target identification commodity feature vector; Passing the optimized target identification commodity feature vector through a commodity classifier to obtain a classification result, wherein the classification result is used to indicate whether the target identification commodity is an infringing or counterfeit commodity; The target identification commodity feature vector is reconstructed in a high-dimensional feature space based on the self-correlation dimension to obtain an optimized target identification commodity feature vector, including: Calculate the target identification product feature autocorrelation matrix between the target identification product feature vector and its transposed vector; Performing matrix multiplication of the weight parameter matrix of the commodity classifier and the target identification commodity feature autocorrelation matrix to obtain a weight parameter modulation classification autocorrelation matrix; Calculating the correlation perception factor between any two row vectors in the weight parameter modulation classification autocorrelation association matrix to obtain a metric correlation perception matrix; The metric correlation perception matrix is multiplied by the target identification commodity feature vector to obtain the optimized target identification commodity feature vector.
8. The method for automatically identifying infringing and counterfeit goods based on computer vision according to claim 7, characterized in that: Characteristic encoding is performed on the image of the target identification commodity to obtain a target identification commodity characteristic map, including: Extracting features of the brand logo in the image of the target identification product to obtain a brand logo detail feature map; Extracting features of workmanship details in the image of the target identification product to obtain a workmanship detail quality feature map; The brand logo detail feature map and the workmanship detail quality feature map are fused to obtain the target identification product feature map.
Citation Information
Cited By
Intelligent identification system for infringement and counterfeit commodities based on dynamic spectrogram topology
CN122176356A
