Ontology-based identification-point object segmentation and text recognition method for determining authenticity of luxury goods, and system for performing method

Abstract

The present invention relates to an ontology-based identification-point object segmentation and text recognition method for determining the authenticity of luxury goods, and a system for performing the method. The ontology-based identification-point object segmentation and text recognition method for determining the authenticity of luxury goods may comprise steps in which: an identification-point setting unit of a luxury goods authenticity determination device sets an identification point; an ontology setting unit of the luxury goods authenticity determination device sets an identification-point-based ontology in consideration of the identification point; a hierarchical object segmentation unit of the luxury goods authenticity determination device generates, on the basis of the identification-point-based ontology, an object segmentation image for a captured image of luxury goods to be evaluated; and an authenticity determination unit of the luxury goods authenticity determination device evaluates, on the basis of the object segmentation image, whether the luxury goods to be evaluated is genuine.

Description

Ontology-based identification point object segmentation and text recognition method for determining the authenticity of luxury goods and a system for performing such method

[0001] The present invention relates to an ontology-based identification point object segmentation and text recognition method for determining the authenticity of a luxury item, and a system for performing such a method. More specifically, the invention relates to an ontology-based identification point object segmentation and text recognition method for determining the authenticity of a luxury item, which defines an ontology structure for determining the authenticity of a luxury item and performs the determination of the authenticity of a luxury item based on the ontology structure.

[0002] Today, the luxury market has continuously grown into a global market worth over 300 trillion won. Of course, due to the aftermath of the COVID-19 pandemic in 2020, the global luxury market recorded a decline for the first time in 12 years, but starting in 2021, it has shown a rapid recovery and is growing steeply, and is expected to grow into a global market worth over 550 trillion won by 2025.

[0003] However, at the same time, the counterfeit market has grown rapidly. The counterfeit market has become a massive problem in the global market, accounting for 7% of world trade volume. There are two main reasons for the expansion of the counterfeit market, as outlined below. First, technological advancements have made it easier to manufacture counterfeits. Consequently, it has become possible to produce more fakes with less cost and effort. Second, the rapid increase in online shopping malls and direct overseas purchases compared to the past is the cause. Customers frequently purchase luxury goods online, yet unknowingly end up buying counterfeits.

[0004] The growth of this counterfeit market is causing various problems. It poses a significant issue not only to customers who unknowingly purchased counterfeit goods but also to global luxury companies. As the counterfeit market expands, the intellectual property rights of the aforementioned luxury firms are being severely infringed, leading to declining sales and causing tens of billions of dollars in global losses. A more critical issue is that if these losses accumulate annually, this massive global industry, worth over 500 trillion won, will eventually collapse, potentially leading to unemployment and a global economic recession. Various systems utilizing artificial intelligence (e.g., deep learning) to determine the authenticity of luxury products based solely on images are currently being researched.

[0005] The present invention aims to solve all of the aforementioned problems.

[0006] In addition, the present invention aims to define an ontology structure for identification points for determining the authenticity of luxury goods and to perform object segmentation and text recognition based on the defined ontology structure to determine the authenticity of luxury goods more accurately.

[0007] Furthermore, the present invention enables systematic classification and management of identification points through the definition of an identification point-based ontology, and facilitates easy system expansion when expanding brands or product groups. Additionally, it aims to improve accuracy based on associations between identification points, enhance model reusability, and improve learning efficiency when applying new brands or products.

[0008] A representative configuration of the present invention for achieving the above objective is as follows.

[0009] According to one embodiment of the present invention, an ontology-based identification point object segmentation and text recognition method for determining the authenticity of a luxury item may include the steps of: an identification point setting unit of a luxury item authenticity determination device setting an identification point; an ontology setting unit of the luxury item authenticity determination device setting an identification point-based ontology considering the identification point; a hierarchical object segmentation unit of the luxury item authenticity determination device generating an object segmentation image for an image of a luxury item to be determined based on the identification point-based ontology; and an authenticity determination unit of the luxury item authenticity determination device determining whether the luxury item to be determined is genuine based on the object segmentation image.

[0010] Meanwhile, the above identification points may be images set by brand, product category, product, and material to determine the authenticity of the luxury goods subject to judgment.

[0011] In addition, the above identification point-based ontology can be integrated for brands and products and generated hierarchically.

[0012] According to another embodiment of the present invention, in a device for determining the authenticity of a luxury item, the device may include an identification point setting unit implemented to set an identification point, an ontology setting unit implemented to set an identification point-based ontology considering the identification point, a hierarchical object segmentation unit implemented to generate an object segmentation image of an image of a luxury item to be judged based on the identification point-based ontology, and an authenticity determination unit implemented to determine whether the luxury item to be judged is genuine based on the object segmentation image.

[0013] Meanwhile, the above identification points may be images set by brand, product category, product, and material to determine the authenticity of the luxury goods subject to judgment.

[0014] In addition, the above identification point-based ontology can be integrated for brands and products and generated hierarchically.

[0015] According to the present invention, an ontology structure for identification points for determining the authenticity of a luxury item is defined, and object segmentation and text recognition are performed based on the defined ontology structure, thereby enabling more accurate determination of the authenticity of the luxury item.

[0016] Furthermore, according to the present invention, systematic classification and management of identification points are possible through the definition of an identification point-based ontology, and easy system expansion is possible when expanding brands or product groups. In addition, accuracy based on associations between identification points is improved, model reusability is enhanced, and learning efficiency can be improved when applying new brands or products.

[0017] FIG. 1 is a conceptual diagram showing a device for determining the authenticity of a luxury item according to an embodiment of the present invention.

[0018] FIG. 2 is a conceptual diagram showing the operation of an ontology setting unit according to an embodiment of the present invention.

[0019] FIG. 3 is a conceptual diagram illustrating a method for defining relationships between classes in an identification point-based ontology according to an embodiment of the present invention.

[0020] FIG. 4 is a conceptual diagram showing the operation of a hierarchical object partitioning unit according to an embodiment of the present invention.

[0021] FIG. 5 is a conceptual diagram showing the operation of a text recognition unit according to an embodiment of the present invention.

[0022] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be modified from one embodiment to another without departing from the spirit and scope of the invention. It should also be understood that the location or arrangement of individual components within each embodiment may be modified without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not meant to be limiting, and the scope of the invention should be understood to encompass the scope claimed by the claims and all equivalents thereof. Similar reference numerals in the drawings indicate identical or similar components across various aspects.

[0023] Hereinafter, in order to enable a person skilled in the art to easily practice the present invention, various preferred embodiments of the present invention will be described in detail with reference to the attached drawings.

[0024]

[0025] FIG. 1 is a conceptual diagram showing a device for determining the authenticity of a luxury item according to an embodiment of the present invention.

[0026] In FIG. 1, a device for determining the authenticity of a luxury item is disclosed.

[0027] Referring to FIG. 1, the genuine product authenticity determination device may include an identification point setting unit (100), an ontology setting unit (110), a hierarchical object segmentation unit (120), a text recognition unit (130), an authenticity determination unit (140), and a processor (150).

[0028] The identification point setting unit (100) can be implemented to set identification points for determining the authenticity of luxury goods. The identification points may include various points that can be utilized for determining the authenticity of luxury goods, such as by brand, by product category, by product, and by material. The identification point setting unit (100) can set such identification points for determining the authenticity of luxury goods.

[0029] The ontology setting unit (110) can set an identification point-based ontology for determining the authenticity of luxury goods based on identification points. According to an embodiment of the present invention, the identification point-based ontology set by the ontology setting unit (110) can be implemented as a single integrated entity for a brand / product.

[0030] By using an integrated identification point-based ontology by brand, the ontology can be constructed without duplicate identification points, and even when new brands or products are created, they can be added to the single identification point-based ontology and used integrally for authenticity verification.

[0031] By utilizing an integrated identification point-based ontology for these brands and products, the difficulties in integrated management that arise when categories (e.g., bags) are defined with different structures for each brand can be resolved. Furthermore, adding new brands or products eliminates the need to modify the entire system, and prevents data quality degradation caused by the lack of consistent standards during identification point labeling. In addition, it avoids model performance variability resulting from individual training for each brand or category, and enables object detection that considers the relationships between identification points, rather than simple object detection that disregards such connections.

[0032] A hierarchical object segmentation unit (120) may be implemented to perform object segmentation on a captured luxury image using an identification point-based ontology and to generate an object segmentation image. The object segmentation image may be used to determine the authenticity of the luxury. For example, if the luxury item subject to the authenticity determination (hereinafter referred to as the luxury item subject to determination) is a bag, the object segmentation image may be an image segmented for identification points for determining the authenticity of the luxury, such as an image of the logo part of the bag, an image of the handle part, and an image of the stitching part, by segmenting the captured image of the luxury item subject to determination.

[0033] The hierarchical object segmentation unit (120) can perform object segmentation on the captured image of a luxury item to be judged by specifying the location of the identification points using an identification point-based ontology. In addition, the hierarchical object segmentation unit can perform multi-scale feature extraction and perform object segmentation through learning the relationships between identification points. Furthermore, the hierarchical object segmentation unit (120) can segment the luxury item image through context-aware object segmentation.

[0034] More specifically, the hierarchical object segmentation unit (120) can determine the location of an ontology-based identification point based on a hierarchical object segmentation engine, divide the captured image into a grid, analyze the whole at once, and then calculate the probability that there is an identification point in each grid cell.

[0035] In addition, the hierarchical object segmentation unit (120) can perform hierarchical object segmentation based on the principle of verifying from multiple angles by using a pyramidal network that analyzes captured images in a multi-layered structure to extract multi-scale features based on the hierarchical object segmentation engine. That is, the object segmentation engine can generate an object segmentation image by performing hierarchical object segmentation based on far distance and near distance using a pyramidal network that analyzes in a multi-layered structure to generate an object segmentation image.

[0036] In addition, the hierarchical object segmentation unit (120) can learn the relationships between identification points based on a hierarchical object segmentation engine. By setting reference points for identification points, it can learn the relative positional relationships of key identification points and calculate the distance, angle, and size ratio between each identification point through association analysis. Furthermore, a procedure can be performed to select only the most accurate one when the same part is detected multiple times through duplicate removal. That is, through the above-mentioned learning of relationships between identification points, overlapping object segmentation images can be removed, and the authenticity of the luxury item can be determined based only on the object segmentation images that are of the highest resolution and can be analyzed accurately.

[0037] The text recognition unit (130) can be implemented for recognizing text on a luxury image. The text recognition unit (130) can generate a text recognition result by performing recognition of text located on the luxury image through recognition of unique fonts and engraving styles by brand, string pattern matching and verification, and OCR results and ontology mapping.

[0038] The authenticity determination unit (140) can determine the authenticity of the luxury item based on the object segmentation image and the text recognition result of the text recognition unit.

[0039] A processor (150) can be implemented to control the operation of an identification point setting unit (100), an ontology setting unit (110), a hierarchical object partitioning unit (120), a text recognition unit (130), and a truth determination unit (140).

[0040] The components according to the embodiments of the present invention may be implemented based on hardware (e.g., a computer server), and the operation of the components described above may be performed on the hardware (e.g., a computer server). Although referred to as components for convenience of explanation, the operations of the components may perform an automated process that did not exist before using computer program code based on hardware components such as a CPU (central processing unit), a GPU (graphic processing unit), and memory.

[0041] By performing this computer program code-based automation through hardware, previously unknown processes are automated. Compared to non-automated procedures, an ontology structure for identification points for determining the authenticity of luxury goods is defined, and object segmentation and text recognition are performed based on this defined ontology structure, thereby enabling more accurate determination of the authenticity of luxury goods. Furthermore, according to the present invention, systematic classification and management of identification points are possible through the definition of an identification point-based ontology, and the system can be easily expanded when expanding brands or product groups. Additionally, accuracy based on the association between identification points is improved, model reusability is enhanced, and learning efficiency can be improved when applying new brands or products.

[0042] More specifically, the processor used in an embodiment of the present invention can execute software (e.g., a program) to control at least one other component (e.g., a hardware or software component) of an electronic device connected to the processor, and can perform various data processing or operations. According to one embodiment of the present invention, as at least part of the data processing or operations, the processor can store commands or data received from other components (e.g., a sensor module or a communication module) in volatile memory, process the commands or data stored in volatile memory, and store the resulting data in non-volatile memory. According to one embodiment, the processor may include a main processor (e.g., a central processing unit or an application processor) or an auxiliary processor that can operate independently or together with it (e.g., a graphics processing unit, a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor). For example, if the electronic device includes a main processor and an auxiliary processor, the auxiliary processor may be configured to use lower power than the main processor or to be specialized for a designated function. The auxiliary processor may be implemented separately from the main processor or as part thereof.

[0043] According to one embodiment, an auxiliary processor (e.g., a neural network processing unit) may include a hardware structure specialized for processing the artificial intelligence model disclosed in the present invention. The artificial intelligence model may be generated through machine learning. Such learning may be performed, for example, on the electronic device itself where the artificial intelligence model is executed, or through a separate server (e.g., a server). The learning algorithm may include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but is not limited to the examples described above. The artificial intelligence model may include a plurality of artificial neural network layers. An artificial neural network may be a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), a deep Q-network, or a combination of two or more of the above, but is not limited to the examples described above. In addition to the hardware structure, the artificial intelligence model may include a software structure, either additionally or substantially.

[0044]

[0045] FIG. 2 is a conceptual diagram showing the operation of an ontology setting unit according to an embodiment of the present invention.

[0046] FIG. 2 discloses a method for defining an identification point-based ontology in an ontology setting unit.

[0047] Referring to FIG. 2, the identification point-based ontology (200) can be configured by considering identification points. The identification point-based ontology (200) can be configured hierarchically based on classes.

[0048] An identification point-based ontology (200) can be set up in a hierarchical structure such as a superclass (group (e.g., meta, super, part, etc.)) and a subclass (attribute (logo, pattern, text, etc.)).

[0049] In addition, ontology relationship definitions can be performed in an identification point-based ontology (200). Ontology relationship definitions can be performed such as relationships between groups and attributes, hierarchical relationships between components, and association relationships between attributes.

[0050] According to an embodiment of the present invention, a hierarchical structure between classes, such as a superclass and a subclass, can be defined in various ways. For example, the superclass may be defined as a specific brand, or it may be defined based on a specific product, such as a bag. That is, the superclass is not defined as a non-overlapping concept, but can also be defined within a concept that overlaps with it.

[0051] Subclasses can be determined based on a superclass. Subclasses can also be defined hierarchically. If the superclass is a bag, subclasses can be defined by type or by parts of the bag (handles, materials, etc.). If the superclass is a brand (e.g., Hermès), subclasses can be defined by bags, clothes, shoes, etc. sold by Hermès.

[0052] Classes like the above can be added in various ways depending on the addition of identification points. For example, if a new identification point for a luxury item is added, the corresponding class can be defined and added to the identification point-based ontology (200).

[0053] In addition, the definition of relationships between classes can be performed after the hierarchical definition of classes. For example, after a bag is defined as a top-level class, brand-specific bags such as Gucci bags, Chanel bags, and Hermès bags can be defined as sub-classes. These brand-specific bags can have a relationship with the bag that is a sub-class when the brand is defined as the top-level class. That is, in the case of the identification point-based ontology (200) according to the embodiment of the present invention, conceptual nesting between top-level classes is possible, and a specific sub-class can be defined as a conceptually higher concept than a specific top-level class.

[0054] Through the definition of such identification point-based ontology (200), systematic classification and management of identification points are possible, and the system can be easily expanded when expanding brands / product groups. In addition, accuracy based on associations between identification points is improved, model reusability is enhanced, and learning efficiency can be improved when applying new brands / products.

[0055] In this way, relationships between classes can be defined on an identification point-based ontology (200). Relationships between classes can be defined in various ways, which will be described later.

[0056] By defining the relationships between these classes, when an identification point-related class is added to a single class, it affects the definitions of other related classes associated with that class, allowing it to be utilized later for object partitioning based on identification points.

[0057]

[0058] FIG. 3 is a conceptual diagram illustrating a method for defining relationships between classes in an identification point-based ontology according to an embodiment of the present invention.

[0059] FIG. 3 discloses a method for defining relationships between classes in an identification point-based ontology.

[0060] Referring to FIG. 3, in order to define relationships between classes in an identification point-based ontology (350), relationship definitions can be performed in various types.

[0061] The first relationship type (300) is an equiclass relationship. An equiclass relationship is one in which the concepts defined in the classes are identical. For example, a class called Hermès bag may be located within various superclasses or on various classes in a hierarchical structure. In this case, Hermès bag defined on the ontology may be the same class.

[0062] The second relationship type (320) is a dependent class relationship. For example, the class called Material can be defined as a subclass for various products such as clothes, bags, etc. In this case, the class called Material can be defined as a subclass dependent on the class for various products such as clothes, bags, etc. As another example, types of Hermès bags such as Birkin bags, Picotin bags, and Herbags can be defined as subclasses of the class called Hermès bags.

[0063] The third relationship type (330) is a pseudo-class relationship. For example, Hermès bags, Chanel bags, Gucci bags, etc. may have a pseudo-class relationship with each other.

[0064] An identification point-based ontology (350) is defined based on these first relationship type (320), second relationship type (320), and third relationship type (340), and the identification point-based ontology (350) can be adaptedly changed according to the addition of new identification points or changes in identification points.

[0065] For example, it can be assumed that an identification point for an Hermès bag is added. In this case, information about the identification point for identifying the Hermès bag can be added to the ontology corresponding to the Hermès bag class. That is, for the same class, if it is associated with a specific identification point, the corresponding information can be updated for all of the same class.

[0066] As another example, it can be assumed that an identification point for a new material is added for the purpose of identifying luxury goods. In this case, a new material for the bag is used, and a judgment regarding this material can be added as a new identification point. In this case, information regarding the corresponding identification point can be added to the class corresponding to the material dependent on the bag. That is, when an identification point corresponding to a specific superclass is added based on a dependency relationship, a subclass corresponding to the dependency relationship can be newly defined in the ontology.

[0067] As another example, when a change occurs to a specific class in an ontology within a similar class relationship, the change can be propagated within the ontology based on the characteristics of the change and the similarity between classes. For example, if an identification point is added to identify the authenticity of an Hermès bag, the change in the class can be propagated based on whether the identification point can be applied to bags of other brands and the similarity between brands.

[0068] For example, if the identification point is the printing state of the Hermès logo, this is an identification point characteristic that cannot be applied to anything other than Hermès bags and may not propagate to other similar classes. As another example, if the identification point is capable of determining the quality of the leather material, this may be a characteristic that can be propagated to other similar classes that sell leather bags.

[0069] In this way, when a change occurs on a specific class, propagation to similar classes takes place, and an identification point-based ontology (350) can be newly defined.

[0070]

[0071] FIG. 4 is a conceptual diagram showing the operation of a hierarchical object partitioning unit according to an embodiment of the present invention.

[0072] FIG. 4 discloses a method for performing object segmentation on a captured luxury goods image using an identification point-based ontology and generating an object segmentation image.

[0073] Referring to FIG. 4, the hierarchical object segmentation unit can determine identification points on an identification point-based ontology according to the object for object segmentation and generate an object segmentation image for the determined identification points.

[0074] More specifically, the class corresponding to the object can be determined as the target class.

[0075] For example, if the luxury item subject to authenticity determination (hereinafter the luxury item subject to determination (400)) is a Birkin bag sold by Hermès, a class corresponding to the Hermès bag on the identification point-based ontology (410) may be determined as the target class (420). For example, the target class (420) may be a class related to the luxury item subject to determination, such as a bag, Hermès bag, material, logo, Birkin bag, etc.

[0076] The identification point (430) can be defined based on the target class (420). The identification point (430) may be defined for each target class (420), and a reference object segmentation image (440) corresponding to the identification point (430) may be defined.

[0077] An identification point (430) is defined for each of the entire target classes, and an object segmentation image (450) can be determined by considering a reference object segmentation image (440) for each of the identification points (430) and dividing the captured image.

[0078] At this time, the identification points (430) and the overlapping reference object segmentation images (440) on the target class (420) are excluded, and a reference object segmentation image (440) for determining the authenticity of the luxury item can be determined. Subsequently, the captured image of the luxury item to be judged can be segmented considering the reference object segmentation image (440) to generate a plurality of target object segmentation images (450).

[0079] Subsequently, the authenticity determination unit can perform authenticity determination by considering the object segmentation image (450) corresponding to the luxury item to be determined. At this time, importance and priority for determining the authenticity of the luxury item for determination can be set for each reference object segmentation image (440), and according to the importance and priority of these reference object segmentation images, the weight of the object segmentation image corresponding to the reference object segmentation image (440) can be set and used to determine the reliability of the authenticity determination of the luxury item.

[0080] The object segmentation image weight can basically be set relatively higher as the importance and priority of the corresponding reference object segmentation image (440) are relatively higher. In addition, the object segmentation image weight can also be adjusted based on the resolution of the object segmentation image itself. If the resolution of the object segmentation image is above a threshold resolution, it is not adjusted, but if the resolution of the object segmentation image is below the threshold resolution, the object segmentation image weight can be adjusted to a relatively lower value as the resolution becomes relatively lower.

[0081] According to an embodiment of the present invention, the reliability of determining the authenticity of a luxury item can be determined by considering the weight of each of the plurality of object segmentation images (450). The more object segmentation images (450) with relatively high weights there are relatively, the higher the reliability of determining the authenticity of a luxury item can be.

[0082] Additionally, according to an embodiment of the present invention, priority and importance may be set by considering the importance of each of the plurality of reference object segmentation images (440), and a decision may be made on whether to request additional photos or set an additional class as a target class by considering the existence of an object segmentation image (450) corresponding to each of the plurality of reference object segmentation images (440) in the actual masterpiece to be judged.

[0083] The priority and importance of these multiple standard object segmentation images (440) can be individually set for each individual product / product category.

[0084] For example, it may be assumed that, for the accuracy of authenticity determination, an object segmentation image (450) corresponding to a reference object segmentation image (440) corresponding to five top priorities is basically required. However, it may be assumed that in the uploaded image of the luxury item to be judged, there is no object segmentation image (450) corresponding to a reference object segmentation image (440) corresponding to a second priority or a third priority. In this case, an additional image of the luxury item to be judged may be requested to generate a reference object segmentation image (440) corresponding to a second priority or a third priority.

[0085] Alternatively, according to an embodiment of the present invention, if it is difficult to obtain an image of an additional luxury item to be judged, the accuracy of the judgment can be increased by setting an additional associated class associated with the image of the currently available luxury item to be judged as an additional target class to further increase reliability.

[0086] The associated class can be determined from a similar class / dependent class related to the luxury item being judged. For example, a class related to the finishing condition of a bag of another brand is set as the associated class, and a reference object segmentation image (440) corresponding to the associated class is additionally determined, and an object segmentation image (450) corresponding to the reference object segmentation image (440) can be additionally segmented and generated.

[0087] Reliability can be increased in the manner described above, and if it falls below the set threshold reliability, the authenticity of the luxury item under judgment can be determined as indisputable.

[0088]

[0089] FIG. 5 is a conceptual diagram showing the operation of a text recognition unit according to an embodiment of the present invention.

[0090] In FIG. 5, a method for recognizing text in a text recognition unit is disclosed.

[0091] Referring to Fig. 5, the text recognition unit can recognize text through learning various characters representing a brand and generate a text recognition result.

[0092] A captured image containing text among the captured images can be input to the text recognition unit as a text captured image.

[0093] Through the photo filtering unit (510) of the text recognition unit, an image among the text images that can be used to determine the authenticity of a luxury item can be determined as a target text image. The text image may include an image containing a distinguishing text (e.g., brand name, product name) that can be used to determine the authenticity of a luxury item, and an image containing a non-distinguishing text (e.g., washing method) that cannot be used to determine the authenticity of a luxury item.

[0094] The photo filtering unit (510) can filter the target text image to extract the target text image containing the identification text.

[0095] Afterwards, a text recognition image can be generated by extracting text portions from a target text image through the photo preprocessing unit (520) of the text recognition unit.

[0096] Next, the text recognition unit can input the text recognition image into the OCR unit (530) to perform primary recognition of the characters.

[0097] A first text recognition procedure is performed to recognize characters recognized by the OCR unit (530) of the text recognition unit, and a first text recognition result based on the first text recognition procedure can be generated.

[0098] The first text recognition procedure may be a procedure that performs a spelling correctness determination. For example, in the case of Chanel, a judgment on whether the text 'chanel' is recognized is performed through the first text recognition procedure, and if the text is judged to be incorrect, such as 'channel', the judgment is stopped in the first text recognition procedure and it may be judged to be a counterfeit. For the first text recognition procedure, the spelling of the word used for each brand may be registered and utilized.

[0099] For the text recognition images of the remaining luxury goods subject to judgment that are not judged as counterfeit through the first text recognition procedure, the second text recognition procedure and the third text recognition procedure can be performed in the second text recognition unit (540) and the third text recognition unit (550), respectively.

[0100] The secondary text recognition procedure can perform a judgment on the overall design of the text. The entire text is recognized as a single image, and the authenticity of the luxury item in question can be determined by judging differences within the image. For this judgment, reference text images extracted for each brand may exist, and a judgment on the differences from these reference text images can be performed. The reference text image may be the entire image containing the text. For example, in the case of Hermès, it may be an image that includes the text "HERMES" and the shapes surrounding the text. The judgment on the differences from the reference text image can be performed based on judgment factors that determine the entire text image, such as the overall outline character ratio, character spacing, distance from shapes, and ratio between characters. Alternatively, the judgment on the differences from the reference text image can be performed through AI-based learning of the reference text image itself during the secondary text recognition procedure.

[0101] The third-level text recognition procedure may involve judging the design of each individual character within the text. Individual characters within the text are recognized as images, and a determination of the authenticity of the luxury item under evaluation can be made based on an assessment of differences within these images. The third-level text recognition procedure may involve judging the completeness of individual characters and the font. The completeness of individual characters refers to an assessment of the printing condition, while the assessment of the font refers to an assessment of the font design itself. These judgments can be performed based on standard text images established for each brand. The weighting for judging the printing condition, such as the completeness of individual characters, can be adjusted by considering the category of the luxury item under evaluation, the printing location, and the age of the item. For example, if the category of the luxury item under evaluation is one where the printing condition may change due to washing, such as with clothing, the weighting for the completeness of individual characters may be adjusted to a relatively lower level. Additionally, if the printing location is prone to damage from friction or if the luxury item under evaluation is old, the weighting for the completeness of individual characters may be adjusted to a relatively lower level. Alternatively, in a third-order text recognition procedure, a judgment on the difference from the reference text image can be performed through AI-based learning on the reference text image itself.

[0102] The first text recognition result through the first text recognition procedure, the second text recognition result through the second text recognition procedure, and the third text recognition result through the third text recognition procedure can be input into the authenticity determination unit and used to determine the authenticity of the luxury goods subject to judgment.

[0103]

[0104] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0105] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or instruct the processing unit independently or collectively. Software and / or data may be embodied in any type of machine, component, physical device, computer storage medium, or device so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0106] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. In this case, the medium may continuously store a program executable by a computer, or temporarily store it for execution or download. Additionally, the medium may be various recording or storage means in the form of a single or several hardware combined, and may not be limited to a medium directly connected to a computer system but may exist distributed over a network. Examples of media may include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and media configured to store program instructions, including ROM, RAM, and flash memory. Additionally, other examples of media may include recording or storage media managed by app stores that distribute applications or sites and servers that supply or distribute various other software.

[0107] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results can be achieved even if the described techniques are performed in a different order than described, and / or the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0108] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

1. The ontology-based identification point object segmentation and text recognition method for determining the authenticity of luxury goods is, A step in which the identification point setting unit of the luxury goods authenticity determination device sets an identification point; The ontology setting unit of the above-mentioned luxury goods authenticity determination device sets an identification point-based ontology considering the identification points; A step in which a hierarchical object segmentation unit of the above-described luxury goods authenticity determination device generates an object segmentation image for an image of a luxury goods to be determined based on the identification point-based ontology; and A method characterized by including a step in which the authenticity determination unit of the above-described luxury goods authenticity determination device determines whether the luxury goods to be determined are genuine based on the object segmentation image.

2. In Paragraph 1, A method characterized in that the above identification points are images set by brand, product category, product, and material to determine whether the above-mentioned luxury goods are genuine.

3. In Paragraph 2, A method characterized by the above identification point-based ontology being integrated for brands and products and generated hierarchically.

4. In a device for determining the authenticity of luxury goods, The above luxury goods authenticity verification device An identification point setting unit implemented to set an identification point; An ontology setting unit implemented to set an identification point-based ontology considering the above identification points; A hierarchical object segmentation unit implemented to generate an object segmentation image for an image of a luxury item to be judged based on the above identification point-based ontology; and A luxury goods authenticity determination device characterized by including an authenticity determination unit implemented to determine whether the luxury goods subject to judgment are genuine based on the object segmentation image.

5. In Paragraph 4, A luxury goods authenticity determination device characterized by the fact that the above-mentioned identification points are images set by brand, product category, product, and material to determine whether the luxury goods subject to judgment are genuine.

6. In Paragraph 5, A luxury goods authenticity determination device characterized by the above-mentioned identification point-based ontology being integrated for brands and products and generated hierarchically.

Images