Appearance patent similarity determination method based on comparative learning and terminal equipment

Abstract

The invention is suitable for the technical field of information processing, and provides an appearance patent similarity determination method based on comparative learning and terminal equipment, and the method comprises the steps: obtaining a to-be-compared first appearance patent file and a to-be-compared second appearance patent file; determining classification number similarity and name similarity of the first appearance patent document and the second appearance patent document; pairing the plurality of product views in the first appearance patent file with the plurality of product views in the second appearance patent file to obtain a plurality of view pairs; determining the similarity of each view pair, and calculating the view similarity of the first appearance patent file and the second appearance patent file according to the preset adaptive weight of each view learned in advance; and the weighted sum of the classification number similarity, the name similarity and the view similarity is determined as the overall similarity of the first appearance patent file and the second appearance patent file, so that the labor cost of appearance patent similarity determination can be reduced, and the accuracy of appearance patent similarity determination is improved.

Description

Technical Field

[0001] The present application belongs to the field of information processing technology, and in particular, relates to a method and terminal device for determining the similarity of appearance patents based on comparative learning. Background Art

[0002] Designs usually involve multiple elements, such as shape, color, and texture, and the combination of these elements makes the similarity determination of designs extremely complicated. The relevant technology usually determines whether the designs in two design patents are similar by visual observation. However, this method of determining the similarity of design patents is greatly affected by subjective factors and is prone to overlooking small but important differences between different designs, which cannot ensure that the similarity determination of design patents is always highly accurate. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a method and a terminal device for determining the similarity of appearance patents based on comparative learning to solve the technical problem of low accuracy of the existing methods for determining the similarity of appearance patents.

[0004] In a first aspect, the present application provides a method for determining the similarity of appearance patents based on comparative learning, comprising:

[0005] Obtaining a first appearance patent document and a second appearance patent document to be compared; wherein the first appearance patent document and the second appearance patent document both include: a classification number, a product name, a product purpose, and a key view that best illustrates the key points of the design of the product to be protected described in words, and multiple product views from different perspectives represented by images; the key view is one of the product views;

[0006] Determining the similarity of the classification numbers of the first appearance patent document and the second appearance patent document according to the classification numbers in the first appearance patent document and the second appearance patent document;

[0007] Determining the similarity between the names of the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document;

[0008] According to the product uses and the names of the product views in the first appearance patent document and the second appearance patent document, the multiple product views in the first appearance patent document are paired with the multiple product views in the second appearance patent document to obtain multiple view pairs;

[0009] Determine the similarity of each of the view pairs, and calculate the view similarity between the first appearance patent document and the second appearance patent document according to the similarity of each of the view pairs and a preset adaptive weight; wherein the preset adaptive weight is determined according to the key view, and the preset adaptive weight of the view pair including the key view is greater than the preset adaptive weight of the view pair not including the key view;

[0010] The weighted sum of the classification number similarity, the name similarity and the view similarity is determined as the overall similarity between the first appearance patent document and the second appearance patent document; wherein the preset weight of the classification number similarity and the preset weight of the name similarity are both smaller than the preset weight of the view similarity.

[0011] In an optional implementation of the first aspect, the first appearance patent document and the second appearance patent document further include: design features of the product to be protected described in words; correspondingly, determining the similarity of each of the view pairs includes:

[0012] In the case where the key design features in the first appearance patent document or the second appearance patent document include shapes, determining the shape similarity of each of the view pairs, and determining the overall feature similarity of each of the view pairs through a feature similarity determination model, and determining the weighted sum of the shape similarity and the overall feature similarity of each of the view pairs as the similarity of each of the view pairs;

[0013] When the design points in the first appearance patent document or the second appearance patent document do not include shapes, the overall feature similarity of each of the view pairs is determined by the feature similarity determination model, and the overall feature similarity of each of the view pairs is determined as the similarity of each of the view pairs.

[0014] In an optional implementation manner of the first aspect, determining the shape similarity of each of the view pairs includes:

[0015] Extracting the contours of each product view in each of the view pairs, and determining the shape invariants of each product view according to the contours of each product view; the shape invariants are used to describe the shape characteristics of the product views, and specifically to describe the property that the shape of the product to be protected in the product views remains unchanged under geometric transformation;

[0016] Determine, based on the shape invariants of each product view in each of the view pairs, a relative difference, a square difference, and an absolute difference between two shape invariants of each of the view pairs;

[0017] According to the relative difference, square difference and absolute difference between the two shape invariants of each view pair, the shape similarity of each view pair is calculated using a preset shape similarity calculation formula; the preset shape similarity calculation formula is:

[0018]

[0019] Among them, S shape is the shape similarity of any one of the view pairs; d1 is the relative difference between the two shape invariants of any one of the view pairs, min(d1) is the minimum value among the relative differences between the two shape invariants of all view pairs, and max(d1) is the maximum value among the relative differences between the two shape invariants of all view pairs; d2 is the square difference between the two shape invariants of any one of the view pairs, min(d2) is the minimum value among the square differences between the two shape invariants of all view pairs, and max(d2) is the maximum value among the square differences between the two shape invariants of all view pairs; d3 is the absolute difference between the two shape invariants of any one of the view pairs, min(d3) is the minimum value among the absolute differences between the two shape invariants of all view pairs, and max(d3) is the maximum value among the absolute differences between the two shape invariants of all view pairs.

[0020] In an optional implementation of the first aspect, the feature similarity determination model is obtained by training a preset convolutional neural network using a deep learning method based on a preset sample set;

[0021] The preset sample set includes a plurality of sample data, wherein the sample data includes positive sample data and negative sample data; each of the positive sample data includes a front view pair and the overall feature similarity of the front view pair, and the overall feature similarity of the front view pair is 1; each of the negative sample data includes a negative view pair and the overall feature similarity of the negative view pair, and the overall feature similarity of the negative view pair is 0;

[0022] When training the preset convolutional neural network, each view pair in the sample data is used as the input of the preset convolutional neural network, and the overall feature similarity of each view pair in the sample data is used as the output of the preset convolutional neural network. The trained preset convolutional neural network is the feature similarity determination model.

[0023] In an optional implementation of the first aspect, the classification number includes a major category number and a minor category number; the major category number is used to indicate a broad product category to which the function or use of the product to be protected belongs, and the minor category number is used to indicate a subdivided product category of the product to be protected under the broad product category;

[0024] Determining the similarity of the classification numbers of the first appearance patent document and the second appearance patent document according to the classification numbers in the first appearance patent document and the second appearance patent document includes:

[0025] In the case where the classification number in the first appearance patent document is exactly the same as the classification number in the second appearance patent document, the classification number similarity is determined to be 1; the classification numbers are exactly the same when the major category numbers and the minor category numbers in the two classification numbers are the same;

[0026] In the case where the classification number in the first appearance patent document is partially identical to the classification number in the second appearance patent document, the classification number similarity is determined to be 0.5; the classification number being partially identical means that the major category numbers in the two classification numbers are the same and the minor category numbers are different;

[0027] In the case where the classification number in the first appearance patent document is completely different from the classification number in the second appearance patent document, the classification number similarity is determined to be 0; completely different classification numbers refer to different major category numbers in the two classification numbers.

[0028] In an optional implementation of the first aspect, determining the name similarity between the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document includes:

[0029] Perform word segmentation processing on the product name in the first appearance patent document and the product name in the second appearance patent document to obtain a first name keyword set corresponding to the first appearance patent document and a second name keyword set corresponding to the second appearance patent document respectively; the first name keyword set and the name keyword set both include one or more keywords from the product name;

[0030] Based on the first name keyword set and the second name keyword set, the name similarity is calculated by a preset set similarity calculation formula; the preset set similarity calculation formula is:

[0031]

[0032] Among them, |WordSet1∩WordSet2| is the number of keywords included in the intersection of the first name keyword set and the second name keyword set, and |WordSet1∪WordSet2| is the number of keywords included in the union of the first name keyword set and the second name keyword set.

[0033] In a second aspect, an embodiment of the present application provides a terminal device, including:

[0034] A file acquisition unit is used to acquire a first appearance patent document and a second appearance patent document to be compared; wherein the first appearance patent document and the second appearance patent document both include: a classification number, a product name, a product purpose and a key view that best illustrates the key points of the design of the product to be protected described in words, and a plurality of product views from different perspectives represented by images; the key view is one of the product views;

[0035] A first determining unit, configured to determine the similarity of the classification codes of the first appearance patent document and the second appearance patent document according to the classification codes in the first appearance patent document and the second appearance patent document;

[0036] A second determining unit is used to determine the name similarity between the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document;

[0037] a view pairing unit, configured to pair the plurality of product views in the first appearance patent document with the plurality of product views in the second appearance patent document, respectively, according to the product uses and the names of the respective product views in the first appearance patent document and the second appearance patent document, to obtain a plurality of view pairs;

[0038] a third determining unit, configured to determine the similarity of each of the view pairs, and calculate the view similarity of the first appearance patent document and the second appearance patent document according to the similarity of each of the view pairs and a preset adaptive weight; wherein the preset adaptive weight of the view pair including the key view is greater than the preset adaptive weight of the view pair not including the key view;

[0039] The fourth determination unit is used to determine the weighted sum of the classification number similarity, the name similarity and the view similarity as the overall similarity between the first appearance patent document and the second appearance patent document; wherein the preset weight of the classification number similarity and the preset weight of the name similarity are both smaller than the preset weight of the view similarity.

[0040] In a third aspect, an embodiment of the present application provides another terminal device, comprising a memory and a computer program stored in the memory and executable on a processor, wherein the processor implements the method described in any optional implementation of the first aspect when executing the computer program.

[0041] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any optional implementation of the first aspect above is implemented.

[0042] The method for determining the similarity of appearance patents based on comparative learning, the terminal device, and the computer-readable storage medium provided in the embodiments of the present application have the following beneficial effects:

[0043] The method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application simultaneously analyzes the text information and view information in the first appearance patent document and the second appearance patent document, and determines the classification number similarity and name similarity of the first appearance patent document and the second appearance patent document respectively according to the classification number and product name in the text information; according to the product use in the text information and the name of each product view in the view information, the first appearance patent document and the product view of the same perspective in the second appearance patent document are paired to obtain multiple view pairs, and the similarity of each view pair is calculated, and the view similarity of the first appearance patent document and the second appearance patent document is calculated in combination with the preset adaptive weights of each view pair obtained according to the key view; and then the overall similarity of the first appearance patent document and the second appearance patent document is comprehensively determined according to the classification number similarity, name similarity and view similarity of the first appearance patent document and the second appearance patent document. Since the overall similarity is obtained by comprehensively combining the text information and all product views in the appearance patent document, the subtle differences between different appearance patent documents can be taken into account, and the calculation of the overall similarity is not affected by human subjective factors, thereby improving the accuracy of the determination of the appearance patent similarity and reducing the labor cost of the appearance patent similarity determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 A schematic flow chart of a method for determining the similarity of appearance patents based on comparative learning provided in an embodiment of the present application;

[0046] Figure 2 A schematic diagram of an interface of a preset similarity determination tool provided in an embodiment of the present application;

[0047] Figure 3A specific implementation flow chart of S105 in a method for determining appearance patent similarity based on comparative learning provided in an embodiment of the present application;

[0048] Figure 4 A schematic diagram of the structure of a terminal device provided in an embodiment of the present application;

[0049] Figure 5 A schematic diagram of the structure of a terminal device provided in another embodiment of the present application. DETAILED DESCRIPTION

[0050] The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0051] In the description of the embodiments of the present application, the technical terms "include", "comprise", "have" and any of their variations mean "including but not limited to", unless otherwise specifically emphasized in other ways. In the description of the embodiments of the present application, unless otherwise specified, the technical term "multiple" refers to two or more than two, and the technical terms "at least one", "one or more" refer to one, two or more. The technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. The technical term "and / or" is only a description of the association relationship of associated objects, indicating that there can be three relationships, such as A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0052] The execution subject of the method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application can be a terminal device. The terminal device can be included in a mobile phone, a tablet computer or a desktop computer, etc. In a specific application, a target script file can be configured for the terminal device, and the target script file describes the method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application, so that the terminal device executes the target script file when it needs to determine the similarity of the appearance patent, and then executes each step of the method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application.

[0053] See also Figure 1 , is a schematic flow chart of a method for determining the similarity of appearance patents based on comparative learning provided in an embodiment of the present application. Figure 1 As shown, the method for determining the similarity of appearance patents based on comparative learning may include S101 to S106, which are described in detail as follows:

[0054] S101, obtaining a first appearance patent document and a second appearance patent document to be compared.

[0055] The first appearance patent document and the second appearance patent document may be any two appearance patent documents to be compared. The first appearance patent document and the second appearance patent document may both be derived from a patent database. For example, the format of the first appearance patent document and the format of the second appearance patent document may both be a portable document format (PDF) or an extensible markup language (XML), etc. The embodiment of the present application does not specifically limit the format of the appearance patent document.

[0056] Optionally, both the first appearance patent document and the second appearance patent document may include: the classification number, product name, product use, design points and key views that best illustrate the design points of the product to be protected described in words, as well as multiple product views represented by images. That is to say, the classification number, product name, product use, design points and key views that best illustrate the design points of the product to be protected in the first appearance patent document and the second appearance patent document are in text form; the multiple product views in the first appearance patent document and the second appearance patent document are in image form. Among them, the key view can be a product view that is specified from multiple product views and best illustrates the design points of the product to be protected.

[0057] In a specific application, a preset similarity determination tool may be installed on the terminal device. The terminal device may execute the above target script file by running the preset similarity determination tool, and then execute each step of the method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application.

[0058] For example, see Figure 2 , is a schematic diagram of an interface of a preset similarity determination tool provided in an embodiment of the present application. Figure 2As shown, in an optional implementation, the interface of the preset similarity determination tool may include a file adding button 201. When the user needs to determine the similarity of any two appearance patent documents, the user can add the appearance patent document to be compared by clicking the file adding button 201. The terminal device can display one or more storage paths when detecting that the file adding button 201 is clicked, and obtain the first appearance patent document and the second appearance patent document to be compared from the storage path selected by the user. In another optional implementation, the interface of the preset similarity determination tool may include a file adding area 202. When the user needs to determine the similarity of any two appearance patent documents, the user can drag the appearance patent document to be compared to the file adding area 202. The terminal device can determine any two appearance patent documents dragged by the user to the file adding area 202 as the first appearance patent document and the second appearance patent document to be compared.

[0059] S102, determining the similarity of the classification codes of the first appearance patent document and the second appearance patent document according to the classification codes in the first appearance patent document and the second appearance patent document.

[0060] After the terminal device obtains the first appearance patent document and the second appearance patent document, it can parse the classification number, product name, product use, design points and key views that best illustrate the design points of the product to be protected from the first appearance patent document and the second appearance patent document through a parsing tool corresponding to the format of the patent document, and parse multiple product views of the product to be protected. For example, when the first appearance patent document and the second appearance patent document are both files in PDF format, the corresponding parsing tool can be a PDF parser. For another example, when the first appearance patent document and the second appearance patent document are both files in XML format, the corresponding parsing tool can be an XML parser.

[0061] The classification number of the product to be protected in the appearance patent document may include a major category number and a minor category number. The major category number may be used to indicate the broad product category to which the function, use or industry attribute of the product to be protected belongs. The minor category number may be used to indicate the subdivided product category of the product to be protected under the broad product category. The major category number and the minor category number may be connected by a separator. For example, the separator may be “-”.

[0062] For example, if the classification number of the product to be protected in a certain appearance patent document is 02-04, then this classification number can be used to indicate that the product to be protected is a shoe (subclass number 04) in clothing (major class number 02). For another example, if the classification number of the product to be protected in a certain appearance patent document is 12-16, then this classification number can be used to indicate that the product to be protected is a car tire (subclass number 16) in a transportation or lifting device (major class number 12).

[0063] Based on this, in an optional implementation, the terminal device may compare the classification number in the first appearance patent document with the classification number in the second appearance patent document, and may determine the similarity of the classification numbers of the first appearance patent document and the second appearance patent document based on the following strategy:

[0064] In the case where the classification number in the first appearance patent document is exactly the same as the classification number in the second appearance patent document, the terminal device can determine that the classification number similarity between the first appearance patent document and the second appearance patent document is 1. The classification number being exactly the same can refer to the two classification numbers having the same major category number and the same minor category number. For example, in the case where the classification number in the first appearance patent document and the classification number in the second appearance patent document are both 12-16, the terminal device can determine that the classification number of the first appearance patent document is exactly the same as the classification number of the second appearance patent document.

[0065] In the case where the classification number in the first appearance patent document is partially identical to the classification number in the second appearance patent document, the terminal device can determine that the classification number similarity between the first appearance patent document and the second appearance patent document is 0.5. The same classification number may refer to the same major category number and different minor category numbers in the two classification numbers. For example, in the case where the classification number in the first appearance patent document is 12-08 and the classification number in the second appearance patent document is 12-16, the terminal device can determine that the classification number of the first appearance patent document is partially identical to the classification number of the second appearance patent document.

[0066] In the case where the classification number in the first appearance patent document is completely different from the classification number in the second appearance patent document, the terminal device can determine that the similarity between the classification numbers of the first appearance patent document and the second appearance patent document is 0. Wherein, completely different classification numbers can refer to different major category numbers in the two classification numbers. For example, in the case where the classification number in the first appearance patent document is 12-16 and the classification number in the second appearance patent document is 14-03, the terminal device can determine that the classification number of the first appearance patent document is completely different from the classification number of the second appearance patent document; for another example, in the case where the classification number in the first appearance patent document is 12-03 and the classification number in the second appearance patent document is 14-03, the terminal device can also determine that the classification number of the first appearance patent document is completely different from the classification number of the second appearance patent document. It can be understood that when the major category numbers in the two classification numbers are different, it means that the two products to be protected belong to completely different product categories. At this time, regardless of whether the minor category numbers in the two classification numbers are the same, the terminal device can determine that the similarity between the classification numbers of the first appearance patent document and the second appearance patent document is 0.

[0067] S103, determining the similarity between the names of the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document.

[0068] In an optional implementation, the terminal device may perform word segmentation processing on the product name in the first appearance patent document and the product name in the second appearance patent document, and obtain a first name keyword set corresponding to the first appearance patent document and a second name keyword set corresponding to the second appearance patent document, respectively. Among them, the first name keyword set and the second name keyword set both include one or more keywords from the corresponding product name, and the keywords may all be nouns. Exemplarily, assuming that the product name in the first appearance patent document is "car tire", the first name keyword set obtained by the terminal device after the word segmentation processing of "car tire" may be [car, tire]. Assuming that the product name in the second appearance patent document is "Bluetooth headset component (Bluetooth headset and Bluetooth headset box)", the second name keyword set obtained by the terminal device after the word segmentation processing of "Bluetooth headset component (Bluetooth headset and Bluetooth headset box)" may be [Bluetooth, headset, component, headset box, Bluetooth headset, Bluetooth headset box].

[0069] After the terminal device obtains the first name keyword set and the second name keyword set, it can calculate the name similarity between the first appearance patent document and the second appearance patent document by using a preset set similarity calculation formula. Exemplarily, the preset set similarity calculation formula can be:

[0070]

[0071] Among them, |WordSet1∩WordSet2| is the number of keywords included in the intersection of the first name keyword set and the second name keyword set, and |WordSet1∪WordSet2| is the number of keywords included in the union of the first name keyword set and the second name keyword set.

[0072] For example, assuming that the first name keyword set is [car, tire] and the second name keyword set is [Bluetooth, earphone, component, earphone box, Bluetooth earphone, Bluetooth earphone box], the intersection of the first name keyword set and the second name keyword set is an empty set, and the union of the first name keyword set and the second name keyword set is [car, tire, Bluetooth, earphone, component, earphone box, Bluetooth earphone, Bluetooth earphone box]. Since the number of keywords included in the intersection is 0 and the number of keywords included in the union is 8, the name similarity between the first appearance patent document and the second appearance patent document calculated by the terminal device based on the above preset set similarity calculation formula is 0 / 8=0.

[0073] For another example, assuming that the first name keyword set is [car, tire] and the second name keyword set is [bicycle, tire], the intersection of the first name keyword set and the second name keyword set is [tire], and the union of the first name keyword set and the second name keyword set is [car, bicycle, tire]. Since the number of keywords included in the intersection is 1 and the number of keywords included in the union is 3, the name similarity between the first appearance patent document and the second appearance patent document calculated by the terminal device based on the above preset set similarity calculation formula is 1 / 3=0.33.

[0074] For another example, assuming that the first name keyword set and the second name keyword set are both [car, tire], then the intersection and union of the first name keyword set and the second name keyword set are both [car, tire]. Since the number of keywords included in the intersection and the union are both 2, the name similarity between the first appearance patent document and the second appearance patent document calculated by the terminal device based on the above preset set similarity calculation formula is 2 / 2=1.

[0075] S104, according to the product uses and the names of each product view in the first appearance patent document and the second appearance patent document, the multiple product views in the first appearance patent document are paired with the multiple product views in the second appearance patent document to obtain multiple view pairs.

[0076] It is understandable that, in specific applications, the product to be protected in the appearance patent document may be a single product, or a complete set of products, a component product, or a product with multiple changing states, etc. In the case where the product to be protected in the appearance patent document is a single product, the appearance patent document usually only includes a set of product views of the product to be protected, and the set of product views usually includes multiple product views of the single product from different perspectives. Exemplarily, in the case where the product to be protected is a car tire, a set of product views of the car tire usually includes a stereoscopic view, a front view, a rear view, a left view, a right view, a top view, and a top view of the car tire, etc. In the case where the product to be protected in the appearance patent document is a complete set of products, a component product, or a product with multiple changing states, the appearance patent document usually includes multiple sets of product views of the product to be protected, each set of product views may include multiple product views from different perspectives, and different sets of product views may correspond to different sets in the complete set of products, or correspond to different components in the component product, or correspond to products in different states of the product with multiple changing states.

[0077] Since the appearance patent document includes multiple groups of product views, the name of each group of product views will not directly reflect the real name of the corresponding kit / component / state, and the correspondence between each group of product views and the real name of the kit / component / state is usually described in the text part of the product use. Therefore, the terminal device can pair the multiple product views in the first appearance patent document with the multiple product views in the second appearance patent document based on the product uses and the names of each product view in the first appearance patent document and the second appearance patent document, so as to form a pair of product views of the same perspective of the same product / kit / component / state to be protected in the two appearance patent documents, and obtain multiple view pairs. In other words, each view pair includes two product views of the same perspective of the same product / kit / component / state to be protected, which are from the first appearance patent document and the second appearance patent document, respectively.

[0078] S105, determining the similarity of each view pair, and calculating the view similarity between the first appearance patent document and the second appearance patent document according to the similarity of each view pair and the preset adaptive weight of each view pair learned in advance.

[0079] The similarity of a view pair refers to the similarity between two product views contained in the view pair.

[0080] It is understandable that the design points of the product to be protected in the appearance patent document may lie not only in shape, but also in color, pattern, texture, material and / or local features, etc. Since the shape feature is the most basic feature of the appearance design, it directly affects the overall visual effect and functionality of the product to be protected, and the number of appearance patent documents whose design points lie in shape is far greater than the number of appearance patent documents whose design points lie in shape. Therefore, in order to adapt to various appearance patent documents with different design points and improve the accuracy of determining the similarity of each view pair, for appearance patent documents whose design points lie in shape and appearance patent documents whose design points lie in shape, the terminal device may adopt different similarity determination strategies to determine the similarity of each view pair. Exemplarily, in an optional implementation method, the terminal device may determine the similarity of each view pair by, for example, Figure 3 S1051 to S1052 shown are used to determine the similarity of each view pair, as described in detail as follows:

[0081] S1051, when the key design features in the first appearance patent document or the second appearance patent document include shape, determine the shape similarity of each view pair, and determine the overall feature similarity of each view pair through a trained feature similarity determination model, and determine the weighted sum of the shape similarity and the overall feature similarity of each view pair as the similarity of each view pair.

[0082] The shape similarity of the view pair can be used to describe the similarity between the two product views in the view pair in terms of shape. The overall feature similarity of the view pair can be used to describe the similarity between the two product views in the view pair in terms of overall image features.

[0083] In this implementation, when the design points of at least one of the first appearance patent document and the second appearance patent document include shape, the terminal device can simultaneously determine the shape similarity and the overall feature similarity of each view pair, and determine the similarity of each view pair by combining the shape similarity and the overall feature similarity of each view pair.

[0084] For example, in a specific implementation, the terminal device may determine the shape similarity of each view pair by following steps 1.1 to 1.3, as described in detail below:

[0085] Step 1.1, extract the contours of each product view in each view pair, and determine the shape invariant of each product view according to the contours of each product view.

[0086] Shape invariants can be used to describe the shape characteristics of a product view. Specifically, they can be used to describe the property that the shape of the product to be protected in the product view remains unchanged under geometric transformations such as rotation, scaling or mirroring.

[0087] In an optional implementation, the terminal device may use an edge detection algorithm to extract the outline of each product view, and may use a shape invariant function to determine the shape invariants of each product view based on the outline of each product view. Exemplarily, the shape invariant function may be the cv2.matchShapes() function in the OpenCV library, and the shape invariant determined based on the cv2.matchShapes() function may be a sequence of 7 invariant moments, which are characterized by remaining unchanged for some common geometric transformations (such as scaling, rotation, or mirroring). For example, assuming that the product to be protected in a product view is rotated 90 degrees, the shape invariant of the product to be protected in the product view can still remain unchanged, so the shape invariant of the product view can be reliably used for shape feature matching.

[0088] Step 1.2, based on the shape invariants of each product view in each view pair, determine the relative difference, square difference and absolute difference between two shape invariants of each view pair.

[0089] The two shape invariants of the view pair may refer to the shape invariants corresponding to the two product views in the view pair.

[0090] In a specific implementation, the terminal device may determine the relative difference between the two shape invariants of each view pair based on the following relative difference determination formula:

[0091]

[0092] Wherein, I1 is the relative difference between the two shape invariants of any view pair, n is the number of invariant moments contained in the shape invariant of each product view, M1[i] is the value of the i-th invariant moment in the shape invariant of one product view in any view pair, and M2[i] is the value of the i-th invariant moment in the shape invariant of the other product view in any view pair.

[0093] It should be noted that the smaller the relative difference between the two shape invariants of the view pair is, the closer the shape invariants of the two product images in the view pair are (that is, the more similar the shapes are).

[0094] In a specific implementation, the terminal device may determine the square difference between two shape invariants of each view pair based on the following square difference determination formula:

[0095]

[0096] Where I2 is the squared difference between the two shape invariants of any view pair.

[0097] It should be noted that the smaller the square difference between the two shape invariants of the view pair is, the closer the shape invariants of the two product images in the view pair are (that is, the more similar the shapes are).

[0098] In a specific implementation, the terminal device may determine the absolute difference between the two shape invariants of each view pair based on the following absolute difference determination formula:

[0099]

[0100] Where I3 is the absolute difference between the two shape invariants of any view pair.

[0101] It should be noted that the smaller the relative difference between the two shape invariants of the view pair is, the closer the shape invariants of the two product images in the view pair are (that is, the more similar the shapes are).

[0102] Step 1.3, according to the relative difference, square difference and absolute difference between the two shape invariants of each view pair, a preset shape similarity calculation formula is used to calculate the shape similarity of each view pair.

[0103] Exemplarily, the preset shape similarity calculation formula may be:

[0104]

[0105] Among them, S shape is the shape similarity of any view pair, and its value range is [0,1]; d1 is the relative difference between the two shape invariants of any view pair, min(d1) is the minimum value of the relative difference between the two shape invariants of all view pairs, and max(d1) is the maximum value of the relative difference between the two shape invariants of all view pairs; d2 is the square difference between the two shape invariants of any view pair, min(d2) is the minimum value of the square difference between the two shape invariants of all view pairs, and max(d2) is the maximum value of the square difference between the two shape invariants of all view pairs; d3 is the absolute difference between the two shape invariants of any view pair, min(d3) is the minimum value of the absolute difference between the two shape invariants of all view pairs, and max(d3) is the maximum value of the absolute difference between the two shape invariants of all view pairs.

[0106] It can be understood that, since the relative difference can be used to measure the relative difference between the two shape invariants of the view pair, when the value of the shape invariant of any product view is very small (for example, close to 0), the relative difference can also capture the subtle difference between the two shape invariants of the view pair. Since the square difference will amplify the difference between the two shape invariants of the view pair, the square difference can capture the larger difference between the two shape invariants of the view pair. Since the absolute difference is used to reflect the direct difference between the two shape invariants of the view pair, the absolute difference can be used to measure the difference between the two shape invariants of the view pair more accurately. Based on this, this embodiment comprehensively determines the shape similarity of each view pair through the relative difference, square difference and absolute difference between the two shape invariants of the view pair, and can comprehensively evaluate the shape difference between the two product views in any view pair from multiple angles, thereby improving the accuracy of determining the shape similarity of the view pair.

[0107] In another specific implementation manner, the terminal device may determine the overall feature similarity of each view pair by using a trained feature similarity determination model.

[0108] Exemplarily, the feature similarity determination model can be obtained by training a preset convolutional neural network (CNN) based on a preset sample set using a deep learning method.

[0109] The preset convolutional neural network can be, for example, a residual network (ResNet) or a visual geometry group (VGG) network, etc. The embodiment of the present application does not limit the specific type of the preset convolutional neural network.

[0110] The preset sample set may include several positive sample data and several negative sample data, each of which may include a front view pair and the overall feature similarity of the front view pair, and each of which may include a negative view pair and the overall feature similarity of the negative view pair. The two product views in the front view pair come from two appearance patent documents with exactly the same classification number, and the overall feature similarity of the front view pair is 1; the two product views in the negative view pair come from two appearance patent documents with completely different classification numbers, and the overall feature similarity of the negative view pair is 0.

[0111] Based on this, when training the preset convolutional neural network, the view pair in each sample data can be used as the input of the preset convolutional neural network, and the overall feature similarity of the view pair in each sample data can be used as the output of the preset convolutional neural network. In this way, through training, the preset convolutional neural network can learn the ability to determine the overall feature similarity of the view pair. The terminal device can determine the trained preset convolutional neural network as a feature similarity determination model.

[0112] Exemplarily, the feature similarity determination model may include a feature extraction network and a similarity determination network. Among them, the input of the feature extraction network can be used as the input of the feature similarity determination model, the output of the feature extraction network can be connected to the input of the similarity determination network, and the output of the similarity determination network can be used as the output of the feature similarity determination model. Specifically, the feature extraction network can be used to extract high-dimensional image features of each product view in each view pair, and the high-dimensional image features are a comprehensive reflection of the features such as shape, color and texture of the product to be protected in the product view. Exemplarily, the dimension of the high-dimensional image features can be 2048 dimensions. The similarity determination network can be used to calculate the overall feature similarity of each view pair based on the high-dimensional image features of each product view in each view pair.

[0113] In this embodiment, after the terminal device determines the shape similarity and the overall feature similarity of each view pair, the weighted sum of the shape similarity and the overall feature similarity of each view pair can be determined as the similarity of each view pair. The sum of the weight of the shape similarity and the weight of the overall feature similarity is 1, and the weight of the shape similarity and the weight of the overall feature similarity can be set according to actual needs. For example, the weight of the shape similarity and the weight of the overall feature similarity can both be 0.5. For another example, the weight of the shape similarity can be 0.3, and the weight of the overall feature similarity can be 0.7.

[0114] S1052, when the design key points in the first appearance patent document or the second appearance patent document do not include shape, the overall feature similarity of each view pair is determined by the trained feature similarity determination model, and the overall feature similarity of each view pair is determined as the similarity of each view pair.

[0115] It should be noted that the specific method for determining the overall feature similarity of the view pair in this embodiment is the same as the specific method for determining the overall feature similarity of the view pair in S1051. For the specific process of determining the overall feature similarity of the view pair, please refer to the relevant description in S1051, which will not be repeated here.

[0116] In the embodiment of the present application, after the terminal device determines the similarity of each view pair, the view similarity of the first appearance patent document and the second appearance patent document can be calculated based on the similarity of each view pair and the preset adaptive weight. The preset adaptive weight of the view pair can be determined by the terminal device based on the key views in the first appearance patent document and the second appearance patent document, and the sum of the preset adaptive weights of all view pairs is 1.

[0117] It is understandable that since the key views recorded in the appearance patent documents that best illustrate the design points of the product to be protected often best reflect the core design points of the product to be protected and are also the key points that distinguish the product to be protected from other products, the terminal device can assign a greater weight to the view pairs containing the key views and a smaller weight to the view pairs not containing the key views based on the key views recorded in the text part of the first appearance patent document and the second appearance patent document. The weight assigned to the view pairs is the preset adaptive weight of the view pairs, that is, the preset adaptive weight of the view pairs containing the key views is greater than the preset adaptive weight of the view pairs not containing the key views.

[0118] Optionally, when the key view in the first appearance patent document and the key view in the second appearance patent document are not product views from the same perspective, it means that the key view in the first appearance patent document and the key view in the second appearance patent document belong to different view pairs, that is, there are two view pairs containing key views. Therefore, the terminal device can adaptively assign the same weight to the view pair to which the key view in the first appearance patent document belongs and the view pair to which the key view in the second appearance patent document belongs, and ensure that the weights assigned to the two view pairs to which the key views in the first appearance patent document and the second appearance patent document belong are greater than the weights assigned to other view pairs.

[0119] Exemplarily, assuming that the terminal device obtains five view pairs by pairing each product view in the first appearance patent document and the second appearance patent document, namely: a view pair corresponding to the stereoscopic view, a view pair corresponding to the front view, a view pair corresponding to the left view, a view pair corresponding to the right view, and a view pair corresponding to the top view, and the key view recorded in the first appearance patent document is a stereoscopic view, and the key view recorded in the second appearance patent document is a front view, then the terminal device can assign a weight of 0.35 to the view pair corresponding to the stereoscopic view, a weight of 0.35 to the view pair corresponding to the front view, a weight of 0.1 to the view pair corresponding to the left view, a weight of 0.1 to the view pair corresponding to the right view, and a weight of 0.1 to the view pair corresponding to the top view.

[0120] Optionally, when the key view in the first appearance patent document and the key view in the second appearance patent document are product views from the same perspective, it means that the key view in the first appearance patent document and the key view in the second appearance patent document belong to the same view pair, that is, there is only one view pair containing the key view. Therefore, the terminal device can adaptively assign the highest weight to the view pair containing the key view.

[0121] Exemplarily, assuming that the terminal device obtains five view pairs by pairing each product view in the first appearance patent document and the second appearance patent document, namely: a view pair corresponding to the stereoscopic view, a view pair corresponding to the front view, a view pair corresponding to the left view, a view pair corresponding to the right view, and a view pair corresponding to the top view, and the key views recorded in the first appearance patent document and the key views recorded in the second appearance patent document are both stereoscopic views, then the terminal device can assign a weight of 0.6 to the view pair corresponding to the stereoscopic view, a weight of 0.1 to the view pair corresponding to the front view, a weight of 0.1 to the view pair corresponding to the left view, a weight of 0.1 to the view pair corresponding to the right view, and a weight of 0.1 to the view pair corresponding to the top view.

[0122] The terminal device may determine the weighted sum of the similarities of all view pairs as the view similarity between the first appearance patent document and the second appearance patent document.

[0123] S106, determining the weighted sum of the classification number similarity, the name similarity and the view similarity as the overall similarity between the first appearance patent document and the second appearance patent document.

[0124] The preset weights of classification number similarity, name similarity and view similarity may all be pre-configured, and the sum of the preset weights of classification number similarity, name similarity and view similarity is 1.

[0125] It is understandable that, since the protection scope of the appearance patent is based on the appearance design of the product to be protected represented in each view, in order to more accurately reflect the similarity between the first appearance patent document and the second appearance patent document, the preset weight of the view similarity can be greater than the preset weight of the classification number similarity and the preset weight of the name similarity. Exemplarily, the preset weight of the view similarity can be 0.6, and the preset weights of the classification number similarity and the name similarity can both be 0.2.

[0126] As can be seen from the above, the method for determining the similarity of appearance patents based on comparative learning provided in the embodiment of the present application simultaneously analyzes the text information and view information in the first appearance patent document and the second appearance patent document, and determines the classification number similarity and name similarity of the first appearance patent document and the second appearance patent document respectively according to the classification number and product name in the text information; according to the product use in the text information and the name of each product view in the view information, the first appearance patent document and the product view of the same perspective in the second appearance patent document are paired to obtain multiple view pairs, and the similarity of each view pair is calculated, and the view similarity of the first appearance patent document and the second appearance patent document is calculated in combination with the preset adaptive weights of each view pair obtained according to the key view; and then the overall similarity of the first appearance patent document and the second appearance patent document is comprehensively determined according to the classification number similarity, name similarity and view similarity of the first appearance patent document and the second appearance patent document. Since the overall similarity is obtained by comprehensively combining the text information and all product views in the appearance patent document, the subtle differences between different appearance patent documents can be taken into account, and the calculation of the overall similarity is not affected by human subjective factors, which can improve the accuracy of the determination of the appearance patent similarity and reduce the labor cost of the appearance patent similarity determination.

[0127] It is understandable that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0128] Based on the method for determining the similarity of appearance patents based on comparative learning provided in the above embodiment, the present application further provides an embodiment of a terminal device that implements the above method embodiment. Figure 4 , is a schematic diagram of the structure of a terminal device provided in an embodiment of the present application. For ease of description, only the parts related to this embodiment are shown. Figure 4 As shown, the terminal device 40 may include: a file acquisition unit 401, a first determination unit 402, a second determination unit 403, a view pairing unit 404, a third determination unit 405, and a fourth determination unit 406. Among them:

[0129] The file acquisition unit 401 is used to acquire a first appearance patent document and a second appearance patent document to be compared; wherein the first appearance patent document and the second appearance patent document both include: a classification number, a product name, a product purpose and a key view that best illustrates the design points of the product to be protected described in words, and a plurality of product views from different perspectives represented by images; the key view is one of the product views.

[0130] The first determining unit 402 is used to determine the similarity of the classification codes of the first appearance patent document and the second appearance patent document according to the classification codes in the first appearance patent document and the second appearance patent document.

[0131] The second determining unit 403 is used to determine the name similarity between the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document.

[0132] The view pairing unit 404 is used to pair the multiple product views in the first appearance patent document with the multiple product views in the second appearance patent document according to the product uses and the names of each product view in the first appearance patent document and the second appearance patent document, so as to obtain multiple view pairs.

[0133] The third determination unit 405 is used to determine the similarity of each of the view pairs, and calculate the view similarity between the first appearance patent document and the second appearance patent document based on the similarity of each of the view pairs and the preset adaptive weight; wherein the preset adaptive weight of the view pair including the key view is greater than the preset adaptive weight of the view pair not including the key view.

[0134] The fourth determination unit 406 is used to determine the weighted sum of the classification number similarity, the name similarity and the view similarity as the overall similarity between the first appearance patent document and the second appearance patent document; wherein the preset weight of the classification number similarity and the preset weight of the name similarity are both smaller than the preset weight of the view similarity.

[0135] Optionally, the first appearance patent document and the second appearance patent document further include: design features of the product to be protected described in words; correspondingly, the third determining unit 405 is specifically used to:

[0136] In the case where the key design features in the first appearance patent document or the second appearance patent document include shapes, determining the shape similarity of each of the view pairs, and determining the overall feature similarity of each of the view pairs through a feature similarity determination model, and determining the weighted sum of the shape similarity and the overall feature similarity of each of the view pairs as the similarity of each of the view pairs;

[0137] When the design points in the first appearance patent document or the second appearance patent document do not include shapes, the overall feature similarity of each of the view pairs is determined by the feature similarity determination model, and the overall feature similarity of each of the view pairs is determined as the similarity of each of the view pairs.

[0138] Optionally, the third determining unit 405 is further configured to:

[0139] Extracting the contours of each product view in each of the view pairs, and determining the shape invariants of each product view according to the contours of each product view; the shape invariants are used to describe the shape characteristics of the product views, and specifically to describe the property that the shape of the product to be protected in the product views remains unchanged under geometric transformation;

[0140] Determine, based on the shape invariants of each product view in each of the view pairs, a relative difference, a square difference, and an absolute difference between two shape invariants of each of the view pairs;

[0141] According to the relative difference, square difference and absolute difference between the two shape invariants of each view pair, the shape similarity of each view pair is calculated using a preset shape similarity calculation formula; the preset shape similarity calculation formula is:

[0142]

[0143] Among them, S shapeis the shape similarity of any one of the view pairs; d1 is the relative difference between the two shape invariants of any one of the view pairs, min(d1) is the minimum value among the relative differences between the two shape invariants of all view pairs, and max(d1) is the maximum value among the relative differences between the two shape invariants of all view pairs; d2 is the square difference between the two shape invariants of any one of the view pairs, min(d2) is the minimum value among the square differences between the two shape invariants of all view pairs, and max(d2) is the maximum value among the square differences between the two shape invariants of all view pairs; d3 is the absolute difference between the two shape invariants of any one of the view pairs, min(d3) is the minimum value among the absolute differences between the two shape invariants of all view pairs, and max(d3) is the maximum value among the absolute differences between the two shape invariants of all view pairs.

[0144] Optionally, the feature similarity determination model is obtained by training a preset convolutional neural network using a deep learning method based on a preset sample set;

[0145] The preset sample set includes a plurality of sample data, wherein the sample data includes positive sample data and negative sample data; each of the positive sample data includes a front view pair and the overall feature similarity of the front view pair, and the overall feature similarity of the front view pair is 1; each of the negative sample data includes a negative view pair and the overall feature similarity of the negative view pair, and the overall feature similarity of the negative view pair is 0;

[0146] When training the preset convolutional neural network, each view pair in the sample data is used as the input of the preset convolutional neural network, and the overall feature similarity of each view pair in the sample data is used as the output of the preset convolutional neural network. The trained preset convolutional neural network is the feature similarity determination model.

[0147] Optionally, the classification number includes a major category number and a minor category number; the major category number is used to indicate a broad product category to which the function or purpose of the product to be protected belongs, and the minor category number is used to indicate a subdivided product category of the product to be protected under the broad product category; the first determining unit 402 is specifically used to:

[0148] In the case where the classification number in the first appearance patent document is exactly the same as the classification number in the second appearance patent document, the classification number similarity is determined to be 1; the classification numbers are exactly the same when the major category numbers and the minor category numbers in the two classification numbers are the same;

[0149] In the case where the classification number in the first appearance patent document is partially identical to the classification number in the second appearance patent document, the classification number similarity is determined to be 0.5; the classification number being partially identical means that the major category numbers in the two classification numbers are the same and the minor category numbers are different;

[0150] In the case where the classification number in the first appearance patent document is completely different from the classification number in the second appearance patent document, the classification number similarity is determined to be 0; completely different classification numbers refer to different major category numbers in the two classification numbers.

[0151] Optionally, the second determining unit 403 is specifically used to: perform word segmentation processing on the product name in the first appearance patent document and the product name in the second appearance patent document, and respectively obtain a first name keyword set corresponding to the first appearance patent document and a second name keyword set corresponding to the second appearance patent document; the first name keyword set and the name keyword set both include one or more keywords from the product name;

[0152] Based on the first name keyword set and the second name keyword set, the name similarity is calculated by a preset set similarity calculation formula; the preset set similarity calculation formula is:

[0153]

[0154] Among them, |WordSet1∩WordSet2| is the number of keywords included in the intersection of the first name keyword set and the second name keyword set, and |WordSet1∪WordSet2| is the number of keywords included in the union of the first name keyword set and the second name keyword set.

[0155] It should be noted that the information interaction, execution process and other contents between the above-mentioned units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be specifically referred to the method embodiment part and will not be repeated here.

[0156] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units is used as an example for illustration. In actual applications, the above-mentioned functions can be assigned to different functional units as needed, that is, the internal structure of the terminal device can be divided into different functional units to complete all or part of the functions described above. The functional units in the embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of each unit in the above-mentioned terminal device can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0157] See also Figure 5 , Figure 5 This is a schematic diagram of the structure of a terminal device provided in another embodiment of the present application. Figure 5 As shown, the terminal device 5 provided in this embodiment may include: a processor 50, a memory 51, and a computer program 52 stored in the memory 51 and executable on the processor 50, such as a program corresponding to the method for determining the similarity of appearance patents based on comparative learning. When the processor 50 executes the computer program 52, the steps in the embodiment of the method for determining the similarity of appearance patents based on comparative learning are implemented, such as Figure 1 Alternatively, when the processor 50 executes the computer program 52, the functions of each module / unit in the above terminal device embodiment are realized, for example Figure 4 The functions of the units 401 to 406 are shown.

[0158] Exemplarily, the computer program 52 may be divided into one or more modules / units, one or more modules / units are stored in the memory 51 and executed by the processor 50 to complete the present application. The one or more modules / units may be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program 52 in the terminal device 5. For example, the computer program 52 may be divided into a file acquisition unit, a first determination unit, a second determination unit, a view pairing unit, a third determination unit, and a fourth determination unit. For the specific functions of each unit, please refer to Figure 4 The relevant descriptions in the corresponding embodiments are not repeated here.

[0159] Those skilled in the art will understand that Figure 5 It is only an example of the terminal device 5 and does not constitute a limitation on the terminal device 5. The terminal device 5 may include more or less components than shown in the figure, or a combination of certain components, or different components.

[0160] The processor 50 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0161] The memory 51 may be an internal storage unit of the terminal device 5, such as a hard disk or memory of the terminal device 5. The memory 51 may also be an external storage device of the terminal device 5, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, or a flash card, etc., equipped on the terminal device 5. Further, the memory 51 may also include both an internal storage unit of the terminal device 5 and an external storage device. The memory 51 is used to store computer programs and other programs and data required by the terminal device. The memory 51 may also be used to temporarily store data that has been output or is to be output.

[0162] The embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, each step of the method for determining the similarity of appearance patents based on comparative learning in the above method embodiment is implemented.

[0163] An embodiment of the present application provides a computer program product. When the computer program product is executed on a terminal device, the terminal device implements the steps in the above-mentioned various method embodiments.

[0164] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0165] It should be noted that, unless otherwise specified, all technical terms used in the embodiments of the present application have the same meaning as those commonly understood by technicians in the technical field of the present application. The technical terms used in the embodiments of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0166] The "embodiment" mentioned in the description of the embodiment of the present application means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0167] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0168] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for determining the similarity of appearance patents based on contrastive learning, characterized in that: include: Obtaining a first appearance patent document and a second appearance patent document to be compared; wherein the first appearance patent document and the second appearance patent document both include: a classification number, a product name, a product purpose, and a key view that best illustrates the key points of the design of the product to be protected described in words, and multiple product views from different perspectives represented by images; the key view is one of the product views; Determining the similarity of the classification numbers of the first appearance patent document and the second appearance patent document according to the classification numbers in the first appearance patent document and the second appearance patent document; Determining the similarity between the names of the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document; According to the product uses and the names of the product views in the first appearance patent document and the second appearance patent document, the multiple product views in the first appearance patent document are paired with the multiple product views in the second appearance patent document to obtain multiple view pairs; Determine the similarity of each of the view pairs, and calculate the view similarity between the first appearance patent document and the second appearance patent document according to the similarity of each of the view pairs and a preset adaptive weight; wherein the preset adaptive weight is determined according to the key view, and the preset adaptive weight of the view pair including the key view is greater than the preset adaptive weight of the view pair not including the key view; The weighted sum of the classification number similarity, the name similarity and the view similarity is determined as the overall similarity between the first appearance patent document and the second appearance patent document; wherein the preset weight of the classification number similarity and the preset weight of the name similarity are both smaller than the preset weight of the view similarity.

2. The method according to claim 1, characterized in that The first appearance patent document and the second appearance patent document also include: the design features of the product to be protected described in words; correspondingly, determining the similarity of each pair of views includes: In the case where the key design features in the first appearance patent document or the second appearance patent document include shapes, determining the shape similarity of each of the view pairs, and determining the overall feature similarity of each of the view pairs through a feature similarity determination model, and determining the weighted sum of the shape similarity and the overall feature similarity of each of the view pairs as the similarity of each of the view pairs; When the design points in the first appearance patent document or the second appearance patent document do not include shapes, the overall feature similarity of each of the view pairs is determined by the feature similarity determination model, and the overall feature similarity of each of the view pairs is determined as the similarity of each of the view pairs.

3. The method according to claim 2, characterized in that Determining the shape similarity of each of the view pairs comprises: Extracting the contours of each product view in each of the view pairs, and determining the shape invariants of each product view according to the contours of each product view; the shape invariants are used to describe the shape characteristics of the product views, and specifically to describe the property that the shape of the product to be protected in the product views remains unchanged under geometric transformation; Determine, based on the shape invariants of each product view in each of the view pairs, a relative difference, a square difference, and an absolute difference between two shape invariants of each of the view pairs; According to the relative difference, square difference and absolute difference between the two shape invariants of each view pair, the shape similarity of each view pair is calculated using a preset shape similarity calculation formula; the preset shape similarity calculation formula is: Among them, S shape is the shape similarity of any one of the view pairs; d1 is the relative difference between the two shape invariants of any one of the view pairs, min(d1) is the minimum value among the relative differences between the two shape invariants of all view pairs, and max(d1) is the maximum value among the relative differences between the two shape invariants of all view pairs; d2 is the square difference between the two shape invariants of any one of the view pairs, min(d2) is the minimum value among the square differences between the two shape invariants of all view pairs, and max(d2) is the maximum value among the square differences between the two shape invariants of all view pairs; d3 is the absolute difference between the two shape invariants of any one of the view pairs, min(d3) is the minimum value among the absolute differences between the two shape invariants of all view pairs, and max(d3) is the maximum value among the absolute differences between the two shape invariants of all view pairs.

4. The method according to claim 2, characterized in that: The feature similarity determination model is obtained by training a preset convolutional neural network based on a preset sample set using a deep learning method; The preset sample set includes a plurality of sample data, wherein the sample data includes positive sample data and negative sample data; each of the positive sample data includes a front view pair and the overall feature similarity of the front view pair, and the overall feature similarity of the front view pair is 1; each of the negative sample data includes a negative view pair and the overall feature similarity of the negative view pair, and the overall feature similarity of the negative view pair is 0; When training the preset convolutional neural network, each view pair in the sample data is used as the input of the preset convolutional neural network, and the overall feature similarity of each view pair in the sample data is used as the output of the preset convolutional neural network. The trained preset convolutional neural network is the feature similarity determination model.

5. The method according to any one of claims 1 to 4, characterized in that: The classification number includes a major category number and a minor category number; the major category number is used to indicate the broad product category to which the function or use of the product to be protected belongs, and the minor category number is used to indicate the subdivided product category of the product to be protected under the broad product category; Determining the similarity of the classification numbers of the first appearance patent document and the second appearance patent document according to the classification numbers in the first appearance patent document and the second appearance patent document includes: In the case where the classification number in the first appearance patent document is exactly the same as the classification number in the second appearance patent document, the classification number similarity is determined to be 1; the classification numbers are exactly the same when the major category numbers and the minor category numbers in the two classification numbers are the same; In the case where the classification number in the first appearance patent document is partially identical to the classification number in the second appearance patent document, the classification number similarity is determined to be 0.5; the classification number being partially identical means that the major category numbers in the two classification numbers are the same and the minor category numbers are different; In the case where the classification number in the first appearance patent document is completely different from the classification number in the second appearance patent document, the classification number similarity is determined to be 0; completely different classification numbers refer to different major category numbers in the two classification numbers.

6. The method according to any one of claims 1 to 4, characterized in that: Determining the similarity between the names of the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document includes: The product name in the first appearance patent document and the product name in the second appearance patent document are segmented to obtain a first name keyword set corresponding to the first appearance patent document and a second name keyword set corresponding to the second appearance patent document; the first name keyword set and the name keyword set both include one or more keywords from the product name. word; Based on the first name keyword set and the second name keyword set, the name similarity is calculated by a preset set similarity calculation formula; the preset set similarity calculation formula is: Among them, WordSet1WordSet2 is the number of keywords included in the intersection of the first name keyword set and the second name keyword set, and WordSet1WordSet2 is the number of keywords included in the union of the first name keyword set and the second name keyword set.

7. A terminal device, characterized in that: include: A file acquisition unit is used to acquire a first appearance patent document and a second appearance patent document to be compared; wherein the first appearance patent document and the second appearance patent document both include: a classification number, a product name, a product purpose and a key view that best illustrates the key points of the design of the product to be protected described in words, and a plurality of product views from different perspectives represented by images; the key view is one of the product views; A first determining unit, configured to determine the similarity of the classification codes of the first appearance patent document and the second appearance patent document according to the classification codes in the first appearance patent document and the second appearance patent document; A second determining unit is used to determine the name similarity between the first appearance patent document and the second appearance patent document according to the product names in the first appearance patent document and the second appearance patent document; a view pairing unit, configured to pair the plurality of product views in the first appearance patent document with the plurality of product views in the second appearance patent document, respectively, according to the product uses and the names of the respective product views in the first appearance patent document and the second appearance patent document, to obtain a plurality of view pairs; a third determining unit, configured to determine the similarity of each of the view pairs, and calculate the view similarity of the first appearance patent document and the second appearance patent document according to the similarity of each of the view pairs and a preset adaptive weight; wherein the preset adaptive weight of the view pair including the key view is greater than the preset adaptive weight of the view pair not including the key view; The fourth determination unit is used to determine the weighted sum of the classification number similarity, the name similarity and the view similarity as the overall similarity between the first appearance patent document and the second appearance patent document; wherein the preset weight of the classification number similarity and the preset weight of the name similarity are both smaller than the preset weight of the view similarity.

8. The terminal device according to claim 7, characterized in that: The first appearance patent document and the second appearance patent document also include: the design features of the product to be protected described in words; Correspondingly, the third determining unit is specifically configured to: In the case where the key design features in the first appearance patent document or the second appearance patent document include shapes, determining the shape similarity of each of the view pairs, and determining the overall feature similarity of each of the view pairs through a feature similarity determination model, and determining the weighted sum of the shape similarity and the overall feature similarity of each of the view pairs as the similarity of each of the view pairs; When the design points in the first appearance patent document or the second appearance patent document do not include shapes, the overall feature similarity of each of the view pairs is determined by the feature similarity determination model, and the overall feature similarity of each of the view pairs is determined as the similarity of each of the view pairs.

9. A terminal device, characterized in that: The method comprises a memory and a computer program stored in the memory and executable on a processor, wherein the processor implements the method according to any one of claims 1 to 6 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images