Asset identification method, apparatus, device, and storage medium

Abstract

The application belongs to the technical field of artificial intelligence, and discloses an asset identification method, device, equipment and storage medium; the method comprises the following steps: acquiring a reference icon image of a target enterprise, and expanding the reference icon image to obtain a reference icon image set; extracting reference information features of each reference icon image, matching the reference information features with web features of each web image in a preset web image set to obtain a target image; acquiring a feature vector in the target image, inputting the feature vector into a preset web classification neural network model to obtain a web category; and obtaining assets of the target enterprise according to the web category; the application expands enterprise icons to obtain a more comprehensive enterprise icon set, matches feature values of icons in the enterprise icon set with web icon features in the preset web image set, identifies content in a reference web page that is successfully matched, judges the relationship between the reference web page and the enterprise, and realizes more accurate search and classification of enterprise assets.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an asset identification method, apparatus, device, and storage medium. Background Technology

[0002] Currently, the most common asset search methods on the market include syntax search, icon search, and batch search. Icon search is a search method based on image feature extraction and matching. The common implementation method is to calculate the icon hash value using the Message-digest Algorithm 5 (MD5) to accurately search for assets that match a specified icon.

[0003] However, during the process of searching for corresponding assets based on icons, it's possible that while the company's homepage has undergone changes, similar changes on the websites of its subsidiaries (such as branches or branches) may lag behind. These lagging websites, however, are still considered direct assets of the company and may not be found in the search. While a company's icon can be used to search for all sites using the same icon, theoretically, any website can use a company's icon. Therefore, when searching for assets based on a specific company icon, non-official assets may be collected—for example, systems released by partners may also use the company's icon, and these systems may have serious problems due to lack of online testing, potentially affecting the company. Summary of the Invention

[0004] The main objective of this invention is to provide an asset identification method, apparatus, device, and storage medium, aiming to solve the technical problem that existing technologies cannot exclude the collection of non-corporate official assets, which may lead to misidentification as the company's own assets, or ignore the failure of subordinate companies to update their icons in a timely manner, resulting in inaccurate identification of corporate assets.

[0005] To achieve the above objectives, the present invention provides an asset identification method, the method comprising the following steps:

[0006] Obtain reference icon images of the target company, and expand the reference icon images to obtain a reference icon image set;

[0007] Extract reference information features from each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching results;

[0008] The feature vector in the target image is obtained, and the feature vector is input into a preset website classification neural network model to obtain the website category of the target image.

[0009] The assets of the target company are obtained based on the website category of the target image.

[0010] Optionally, the step of expanding the reference icon image to obtain a reference icon image set includes:

[0011] The reference icon images are expanded according to the target company's preset icon deformation rules to obtain a first set of icon images;

[0012] The reference icon image is expanded according to a preset conventional icon deformation rule to obtain a second set of icon images;

[0013] Obtain a reference icon image containing hidden information, and expand the reference icon image containing hidden information to obtain a third set of icon images;

[0014] The first set of icon images, the second set of icon images, and the third set of icon images are used as reference icon image sets.

[0015] Optionally, the step of expanding the reference icon image according to the target enterprise's preset icon deformation rules to obtain the first icon image set includes:

[0016] The reference icon image is input into the pyramid network, which includes a first stage, a second stage, a third stage, and a fourth stage.

[0017] In the first stage, the reference icon image is iterated multiple times to obtain a low-resolution image;

[0018] In the second stage, features of the low-resolution image are extracted through convolutional layers to obtain multiple similar images;

[0019] In the third stage, the dimensions of the multiple similar images are defined to obtain multiple reference images of different sizes;

[0020] In the fourth stage, the size of the reference image is adjusted by a preset training model to obtain a first set of icon images.

[0021] Optionally, the step of obtaining a reference icon image including hidden information, and expanding the reference icon image including hidden information to obtain a third set of icon images, includes:

[0022] Obtain a reference icon image including hidden information, and perform inverse embedding processing on the hidden information to obtain the hidden information;

[0023] Obtain a standard icon image, embed the hidden information into the hidden blocks of the standard icon image, and obtain a third icon image set;

[0024] The step of embedding the hidden information into the hidden blocks of the standard icon image to obtain the third icon image set includes:

[0025] The white or black pixels in the standard image are flipped using a graph topology algorithm to obtain hidden blocks.

[0026] The hidden information is embedded into the hidden block to obtain a third set of icon images.

[0027] Optionally, the reference information features include hidden information features and physical information features; the web page features of each web page image include hidden information features and physical information features of each web page image.

[0028] The step of extracting reference information features from each reference icon image in the reference icon image set, matching the reference information features with the web page features of each web page image in the preset web page image set, and obtaining the target image based on the matching result includes:

[0029] Extract the hidden information features and physical information features of each reference icon image in the reference icon image set;

[0030] The hidden information features are matched with the hidden information features of each webpage image in the preset webpage image set. When the hidden information features are successfully matched with the hidden information of each webpage image, the successfully matched webpage image is taken as the target image.

[0031] The physical information features are matched with the physical information features of each webpage image in the preset webpage image set. When the physical information features match the physical information features of each webpage image, the successfully matched webpage image is taken as the target image.

[0032] Optionally, obtaining the feature vector in the target image includes:

[0033] The target image is input into a preset target detection model to obtain the category features and location features in the target image;

[0034] Semantic recognition is performed on the target image to obtain the business attribute features of the target image;

[0035] The feature vector of the target image is obtained by combining the category features, the location features, and the business attribute features.

[0036] Optionally, the website categories include the target company's current website, the target company's historical websites, and websites not belonging to the target company;

[0037] The step of obtaining the target company's assets based on the website category of the target image includes:

[0038] The domain name information of the target company's current website and the domain name information of the target company's historical websites are regarded as a type of asset of the target company;

[0039] Based on a preset enterprise relationship database, it is determined whether the target enterprise and the non-target enterprise website have a cooperative relationship. The domain name information of the non-target enterprise website that has a cooperative relationship with the target enterprise is regarded as the second type of asset of the target enterprise.

[0040] The assets of the target enterprise are determined based on the first type of assets and the second type of assets.

[0041] Furthermore, to achieve the above objectives, the present invention also proposes an asset identification device, the asset identification device comprising:

[0042] The image acquisition module is used to acquire reference icon images of the target enterprise and expand the reference icon images to obtain a reference icon image set;

[0043] The feature matching module is used to extract reference information features of each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching result.

[0044] The website classification module is used to obtain feature vectors from the target image, input the feature vectors into a preset website classification neural network model, and obtain the website category of the target image.

[0045] The asset identification module is used to obtain the assets of the target enterprise based on the website category of the target image.

[0046] Furthermore, to achieve the above objectives, the present invention also proposes an asset identification device, the asset identification device comprising: a memory, a processor, and an asset identification program stored in the memory and executable on the processor, the asset identification program being configured to implement the steps of the asset identification method as described above.

[0047] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing an asset identification program, which, when executed by a processor, implements the steps of the asset identification method as described above.

[0048] This invention expands enterprise icons to obtain a more comprehensive set of enterprise icons, matches the feature values ​​of icons in the enterprise icon set with the features of web page icons in a preset web page image set, identifies the content of successfully matched reference web pages, determines the relationship between the reference web page and the enterprise, and achieves more accurate search and classification of enterprise assets. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the structure of the asset identification device in the hardware operating environment involved in the embodiments of the present invention;

[0050] Figure 2 This is a flowchart illustrating the first embodiment of the asset identification method of the present invention;

[0051] Figure 3 This is a schematic diagram illustrating the process of extracting image features and hidden features according to an embodiment of the asset identification method of the present invention;

[0052] Figure 4 This is a schematic diagram of a target detection algorithm according to an embodiment of the asset identification method of the present invention;

[0053] Figure 5 This is a flowchart illustrating the second embodiment of the asset identification method of the present invention;

[0054] Figure 6 This is a pyramid network structure diagram of an embodiment of the asset identification method of the present invention;

[0055] Figure 7 This is a schematic diagram of an expanded reference icon image according to an embodiment of the asset identification method of the present invention;

[0056] Figure 8 This is a flowchart illustrating the third embodiment of the asset identification method of the present invention;

[0057] Figure 9 This is a complete flowchart of an embodiment of the asset identification method of the present invention;

[0058] Figure 10 This is a structural block diagram of the first embodiment of the asset identification device of the present invention.

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0061] Reference Figure 1 , Figure 1 This is a schematic diagram of the asset identification device structure in the hardware operating environment involved in the embodiments of the present invention.

[0062] like Figure 1As shown, the asset identification device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0063] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the asset identification device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0064] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and an asset identification program.

[0065] exist Figure 1 In the asset identification device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the asset identification device of the present invention can be set in the asset identification device. The asset identification device calls the asset identification program stored in the memory 1005 through the processor 1001 and executes the asset identification method provided in the embodiment of the present invention.

[0066] This invention provides an asset identification method, referring to... Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of an asset identification method according to the present invention.

[0067] In this embodiment, the asset identification method includes the following steps:

[0068] Step S10: Obtain the reference icon image of the target enterprise, and expand the reference icon image to obtain a reference icon image set.

[0069] Understandably, if the goal is to identify a company's assets, then that company is considered the target company.

[0070] Understandably, the reference icon image can be a corporate icon image that the target company is using, provided by the target company.

[0071] It should be emphasized that there can be multiple reference icon images or just one; an icon is also called an icon.

[0072] It should be noted that when the target company uses the reference icon image, or when other related companies use the reference icon image, or when the reference icon image is stolen, the image may be altered, such as by changing the shape, color, clarity, or by other means. The altered image is not easy to identify, and when searching for and identifying the target company's assets, it is easy to miss or make mistakes.

[0073] It should be further explained that expanding the reference icon image can be understood as performing image processing such as changing the shape, color, and clarity of the reference icon image in advance to generate multiple images similar to the reference icon image. The step of generating similar images can be understood as expansion. Each reference icon image obtained after expansion and its expanded image constitute a reference icon image set.

[0074] It should be further emphasized that there may be one or multiple reference icon image sets.

[0075] Step S20: Extract the reference information features of each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching result.

[0076] It should be noted that extracting the reference information features of each reference icon image in the reference icon image set can involve extracting features related to the hidden information from each reference icon image in the reference icon image set. In specific implementations, this can be done by calculating the hash value of each reference icon image separately. Simultaneously, image features (image features can be shape features or other features with visually perceptible attributes) of each reference icon image in the reference icon image set can be extracted. The extraction of image features and hidden features can refer to... Figure 3 .

[0077] Understandably, the reference information features may include the hash values ​​of each reference icon image and the image features of each reference icon image.

[0078] Understandably, the asset identification of a target company can be achieved by searching through a vast number of web pages, obtaining web page icons from multiple websites, and then combining the web page images from multiple websites into a preset web page image set.

[0079] It should be understood that the reference information features are matched with the web page features of each web page image in the preset web page image set. If the match is successful, the matched web page can be regarded as the relevant asset of the target enterprise.

[0080] It should be noted that a match can be considered successful if the reference information features and webpage features meet the preset similarity threshold. The preset similarity threshold can be 80% or 90%, etc., and can be limited according to the actual situation.

[0081] It should be emphasized that the reference icon image is derived from the same icon image, and there may be one webpage feature and multiple reference information features that match successfully.

[0082] Step S30: Obtain the feature vector in the target image, input the feature vector into a preset website classification neural network model, and obtain the website category of the target image.

[0083] Understandably, the target image can be a successfully matched webpage image, and the preset website classification neural network model can be a pre-trained neural network model that can obtain the website category of the target image based on the feature vector of the target image.

[0084] It should be noted that the website category of the target image can include the target company's current website, the target company's historical websites, and websites other than the target company's website.

[0085] It should be emphasized that while a company's website homepage may have undergone some changes, similar changes on the websites of the company's subordinate agencies (such as branches or subsidiaries) may lag behind. However, these websites that have lagged behind in their changes are also direct assets of the company and need to be identified and statistically analyzed.

[0086] The neural network model can identify specific website categories based on the input feature vectors. Specifically, it can use the category and location information of objects appearing in screenshots of the target company's current website homepage, historical website homepages, and non-company website homepages, along with feature vectors obtained from combinations of domain name, registration number, keywords, icons other than those matching the hash value or image features of similar icons, contact persons, and business attributes, as training samples. Furthermore, the training samples are also labeled with "target company's current website," "target company's historical website," or "non-target company," and are obtained through supervised training.

[0087] It should be further explained that obtaining the feature vector in the target image can be:

[0088] The target image is input into a preset target detection model to obtain the category features and location features in the target image. The preset target detection model can be understood as a model used to identify web page information in the target image.

[0089] In practice, the target image is input into a deep learning network for target detection to obtain the category and location information of the target object (such as icon, title, text content, etc.) detected by the deep learning network from the target image.

[0090] The object detection task for images includes a classification branch and a regression branch. The classification branch is used for object recognition, that is, to determine the category of the object, while the regression branch is used for object localization, that is, to determine the location information of the object.

[0091] In this proposal, a region convolutional neural network feature algorithm can be used to achieve object detection. This object detection algorithm can be found at [reference needed]. Figure 4 .

[0092] It should be emphasized that, considering that different companies' websites often have different styles, the location and category of target objects such as icons, titles, and text content in the homepage screenshot can distinguish websites with different styles to a certain extent. Object detection can be used to obtain the location and category of target objects as a feature for identifying website categories. This feature is the category feature and location feature in the target image.

[0093] Furthermore, semantic recognition is performed on the target image to obtain the business attribute features of the target image.

[0094] Understandably, different websites target different categories and locations of objects.

[0095] It should be noted that from the screenshot of the website homepage, the following features were obtained using semantic recognition and other methods: domain name, filing number, keywords, other icons besides those that match the hash value or image features of the reference icon image, contact person, business, etc. These features were used as business attribute features.

[0096] Furthermore, the category features, location features, and business attribute features are combined or fused to obtain the feature vector of the target image.

[0097] Step S40: Obtain the assets of the target enterprise based on the website category of the target image.

[0098] Understandably, the target company's assets may include both Category I and Category II assets.

[0099] It should be noted that the domain name information of the target company's current website and the domain name information of the target company's historical websites are considered as the target company's first-class assets. A preset enterprise relationship database is used to determine whether the target company has a cooperative relationship with the websites of non-target companies, and the domain name information of the non-target company websites that have a cooperative relationship with the target company is considered as the target company's second-class assets. The assets of the target company are determined based on the first-class and second-class assets. The preset enterprise relationship database can be understood as a database that includes relevant information of other companies associated with the target company. If the data in the database contains the domain name information of the website in the target image, then the website in the target image can be considered to have a cooperative relationship with the current company.

[0100] Understandably, the website category for the target image may also include websites of non-target companies, which are not considered assets of the target company.

[0101] In practice, from the assets to be classified, the domain name information that belongs to the website categories such as "target company's current website" and "target company's historical website" is identified as the target company's direct assets, or can be called the first category of assets.

[0102] Furthermore, for the domain names identified as belonging to the "non-target enterprise website" category, based on the full name of the target enterprise and the full names of the non-target enterprises corresponding to the "non-target enterprise websites," the system searches online information or queries relevant databases storing company relationships to determine whether a designated partnership exists between them. The designated partnership mentioned here includes, but is not limited to, the non-target enterprise being a subsidiary or branch of the target enterprise, or the non-target enterprise being a holding company of the target enterprise.

[0103] Furthermore, for non-target companies that have a designated partnership with the target company, the domain name information of the non-target company's website included in the assets to be classified is identified as the target company's indirect assets or usable assets, or what may be called Category II assets. All other assets in the assets to be classified, excluding Category I and Category II assets, are identified as assets unrelated to the target company.

[0104] This embodiment expands the enterprise icon set to obtain a more comprehensive set of enterprise icons, matches the feature values ​​of icons in the enterprise icon set with the feature values ​​of web page icons in the preset web page image set, identifies the content of the successfully matched reference web page, determines the relationship between the reference web page and the enterprise, and achieves more accurate search and classification of enterprise assets.

[0105] refer to Figure 5 , Figure 5 This is a flowchart illustrating a second embodiment of an asset identification method according to the present invention.

[0106] Based on the first embodiment described above, the asset identification method of this embodiment includes the following in step S10:

[0107] Step S11: Expand the reference icon image according to the target company's preset icon deformation rules to obtain a first icon image set.

[0108] Understandably, preset deformation rules can predetermine how to perform graphic transformations on reference icon images.

[0109] It should be understood that the first set of icon images can be similar images obtained by transforming each reference icon image using preset icon deformation rules.

[0110] It should be emphasized that the first set of icon images can be the first similar image of the reference icon image, and there are also a second set of icon images and a third set of icon images.

[0111] It should be noted that the reference icon image is input into the pyramid network, which includes a first stage, a second stage, a third stage, and a fourth stage. The pyramid network structure can be referenced from... Figure 6 In the diagram, G represents the generator, which generates icon image files to be trained; D represents the discriminator, which is used to discriminate the usability of the training data; and Noise represents the noise needed during training to ensure that the reference icon image files tend to be realistic and usable.

[0112] It should be emphasized that the principle of the pyramid network algorithm is to train it in a multi-stage manner using a pyramid structure, starting from a reference icon image. In each stage, the image features and color dimensions of the reference icon image are learned. During the stage learning stage, the size of the reference icon image is adjusted to control the process, thereby continuously generating similar icon images based on the reference icon image file.

[0113] Furthermore, in the first stage, the reference icon image is iterated multiple times to obtain a low-resolution image;

[0114] In the second stage, features of the low-resolution image are extracted through convolutional layers to obtain multiple similar images. Convolutional layers are added to increase the generator size, and each stage takes the icon features of the previous stage as input. Each stage trains images of different sizes to generate any number of similar images.

[0115] In the third stage, the dimensions of the multiple similar images are defined to obtain multiple reference images of different sizes. The dimensions of the similar images at each stage can be defined with reference to the following formula:

[0116] x n =x N ×r ((N-1) / log(N))*log(N-n)+1

[0117] n = 0, ..., N-1

[0118] Where n represents the stage, N represents the total number of stages, and r represents a scalar, which is defined here as r = 0.55. When there are 3 stages, the sample icon image sizes are 25*34, 32*42, 42*56, and 63*84.

[0119] In the fourth stage, the size of the reference image is adjusted by a preset training model to obtain a first set of icon images. The reference images are obtained by continuously training the model through a Generative Adversarial Network (GAN, which consists of a generator network and a discriminator network) by continuously changing the size parameters. There can be multiple reference images.

[0120] In practical implementation, the reference image obtained after pyramid network image expansion can be used for reference. Figure 7 , Figure 7 The first row contains multiple reference icon images; the second row contains image files, which are the first set of icon images obtained by expanding according to the first type of expansion rules.

[0121] Step S12: Expand the reference icon image according to the preset conventional icon deformation rules to obtain a second icon image set.

[0122] It should be noted that the preset regular icon transformation rules can be based on historical experience and are common transformation rules for icons, such as changing the color of the standard icon.

[0123] It should be understood that the same reference icon image can be used to obtain a second set of icon images through preset regular icon deformation rules, or it can be extended through preset icon deformation rules.

[0124] Step S13: Obtain a reference icon image including hidden information, and expand the reference icon image including hidden information to obtain a third icon image set.

[0125] Understandably, the reference icon image containing hidden information stored in the target company's application system is obtained from the application system's backend, in order to obtain several images similar to the reference icon image.

[0126] It should be understood that not all reference icon images of target companies contain hidden information.

[0127] Understandably, after a target company publishes its standard icon, it can use a program to run this rule under specific triggering conditions to prevent malicious misuse of the icon. These specific triggering conditions could include, for example, detecting actions such as downloading, saving as, referencing, or crawling the icon from the target company's website.

[0128] It should be noted that the hidden information can be a unique identifier provided by the target company to represent the target company's official website, timestamp information, specific strings, etc.

[0129] To elaborate further, this hidden information can be attached to the icon image as a semi-transparent or invisible watermark, serving as part of the icon image.

[0130] It should be emphasized that obtaining a reference icon image containing hidden information and expanding the reference icon image containing hidden information to obtain a third set of icon images can be a process of obtaining a reference icon image containing hidden information and performing inverse embedding processing on the hidden information to obtain the hidden information. The inverse processing can be understood as the reverse operation of each step of embedding the hidden information into the image.

[0131] Obtain a standard icon image, embed the hidden information into the hidden blocks of the standard icon image, and obtain a third icon image set;

[0132] Furthermore, the step of embedding the hidden information into the hidden blocks of the standard icon image to obtain the third icon image set can be based on a graph topology algorithm to flip the white or black pixels in the standard image to obtain the hidden blocks; the hidden information is then embedded into the hidden blocks to obtain the third icon image set.

[0133] It is worth noting that the reference icon image may contain hidden information pre-added by the company, which acts as an "anti-counterfeiting mark". The search principle of the asset search engine may be to calculate the hash value of the reference icon image and search based on the hash value. If the target company allows its affiliated companies (such as subsidiaries, branches, partners) to use its icon, it will provide an icon carrying hidden information. Affiliated companies may modify or change the color of the icon carrying hidden information before using it.

[0134] In practice, the original image (the standard corporate icon image released by the target company) can be divided into blocks, and then those hidden blocks into which information can be embedded can be selected. The white (or black) pixels can be flipped using a hiding method based on the principle of graph topology, and the required hidden information can be embedded into the corresponding blocks. The embedding of hidden information can be selected according to the actual situation, and this embodiment does not limit it.

[0135] Step S14: Use the first icon image set, the second icon image set, and the third icon image set as a reference icon image set.

[0136] Understandably, each reference icon image can be expanded as described in steps S11-S13. Furthermore, each reference icon image can be composed of a first set of icon images, a second set of icon images, and a third set of icon images, each set consisting of images similar to the reference icon images.

[0137] It should be understood that merging the first set of icon images, the second set of icon images, and the third set of icon images yields a reference set of icon images.

[0138] This embodiment expands the reference icon image from three angles using preset icon deformation rules, preset regular icon deformation rules, and hidden information, resulting in a richer set of images similar to the target company's reference icon image. Based on a more comprehensive set of reference icon images, it can more accurately identify websites related to the target company in the massive web page features during matching, thereby enabling more accurate identification of the company's assets.

[0139] refer to Figure 8 , Figure 8 This is a flowchart illustrating a second embodiment of an asset identification method according to the present invention.

[0140] Based on the first embodiment described above, the asset identification method of this embodiment includes the following in step S20:

[0141] Step S21: Extract the hidden information features and physical information features of each reference icon image in the reference icon image set.

[0142] It should be noted that extracting the hidden information features of each reference icon image in the reference icon image set can be done by extracting the hash value of the reference icon image.

[0143] Furthermore, extracting the physical information features of each reference icon image in the reference icon image set can be done by extracting the image features of the reference icon images. The image features can be features with visually perceptible attributes, such as shape features.

[0144] Step S22: Match the hidden information features with the hidden information features of each webpage image in the preset webpage image set. When the hidden information features successfully match the hidden information of each webpage image, take the successfully matched webpage image as the target image.

[0145] It should be noted that the preset webpage image set can be obtained by the search device from a massive number of webpage icon images in the network space, stored, and used to create an image set for searching.

[0146] It should be emphasized that the domain information of the source websites of each webpage icon can be saved in advance as an asset to be classified, and the preset webpage image set can not only obtain webpage icons, but also capture screenshots of the homepage of the website corresponding to each webpage icon.

[0147] It is worth noting that, in order to avoid the situation where the number of webpage icon images is too large, resulting in an excessive workload for obtaining and storing massive amounts of webpage icon images, the acquisition range of massive amounts of webpage icon images can be preset according to actual needs before acquisition and storage.

[0148] Furthermore, matching the hidden information features with the hidden information features of each webpage image in the preset webpage image set can be done by calculating the hash value of each webpage image in the preset webpage image set; comparing the hidden information features with the hash values ​​of each webpage image, if the hash value of the webpage image successfully matches the hidden information feature of at least one reference icon image in the reference icon image set, then the webpage is considered to be a website related to the target enterprise, and the webpage image is taken as the target image.

[0149] Step S23: Match the physical information features with the physical information features of each webpage image in the preset webpage image set. When the physical information features match the physical information features of each webpage image successfully, use the successfully matched webpage image as the target image.

[0150] Understandably, matching physical information features with the physical information features of each webpage image in the preset webpage image set can be done by calculating the image features of each webpage image in the preset webpage image set; comparing the physical information features with the image features of each webpage image; if the image features of the webpage image successfully match the physical information features of at least one reference icon image in the reference icon image set, then the webpage is considered to be a website related to the target company, and the webpage image is taken as the target image.

[0151] Understandably, a webpage image that successfully matches physical information features and / or hidden information features can be used as a target image. Simply put, if the physical information features of a webpage image fail to match, but the hidden information features do match, that webpage image can still be used as the target image.

[0152] It should be noted that the target image that is successfully matched with physical information features may not contain hidden information, but there is a certain probability that it is an icon image used by the target company or related companies.

[0153] In practical implementation, the complete process from image extension to feature matching, and then to webpage classification and asset identification can be referenced. Figure 9 .

[0154] This embodiment matches the hidden information features and physical information features of each reference icon image in a pre-collected massive web page image and reference icon image set. If either the hidden information feature or the physical information feature is successfully matched, the corresponding web page image is taken as the target image. By identifying web pages that may be related to the target company from the massive web page images from both the hidden information feature and physical information feature perspectives, a preliminary selection is made, providing a more accurate target image for the next step of web page category identification, thereby accurately identifying the assets of the target company from the massive web pages.

[0155] Furthermore, embodiments of the present invention also propose a storage medium storing an asset identification program, which, when executed by a processor, implements the steps of the asset identification method described above.

[0156] Reference Figure 10 , Figure 10 This is a structural block diagram of the first embodiment of the asset identification device of the present invention.

[0157] like Figure 10 As shown, the asset identification device proposed in this embodiment of the invention includes:

[0158] Image acquisition module 10 is used to acquire reference icon images of the target enterprise and expand the reference icon images to obtain a reference icon image set;

[0159] The feature matching module 20 is used to extract reference information features of each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching result.

[0160] Website classification module 30 is used to obtain feature vectors in the target image, input the feature vectors into a preset website classification neural network model, and obtain the website category of the target image;

[0161] The asset identification module 40 is used to obtain the assets of the target enterprise based on the website category of the target image.

[0162] This embodiment expands the enterprise icon set to obtain a more comprehensive set of enterprise icons, matches the feature values ​​of icons in the enterprise icon set with the feature values ​​of web page icons in the preset web page image set, identifies the content of the successfully matched reference web page, determines the relationship between the reference web page and the enterprise, and achieves more accurate search and classification of enterprise assets.

[0163] In one embodiment, the image acquisition module 10 is further configured to expand the reference icon image according to the preset icon deformation rules of the target enterprise to obtain a first icon image set;

[0164] The reference icon image is expanded according to a preset conventional icon deformation rule to obtain a second set of icon images;

[0165] Obtain a reference icon image containing hidden information, and expand the reference icon image containing hidden information to obtain a third set of icon images;

[0166] The first set of icon images, the second set of icon images, and the third set of icon images are used as reference icon image sets.

[0167] In one embodiment, the image acquisition module 10 is further configured to input the reference icon image into a pyramid network, the pyramid network including a first stage, a second stage, a third stage, and a fourth stage;

[0168] In the first stage, the reference icon image is iterated multiple times to obtain a low-resolution image;

[0169] In the second stage, features of the low-resolution image are extracted through convolutional layers to obtain multiple similar images;

[0170] In the third stage, the dimensions of the multiple similar images are defined to obtain multiple reference images of different sizes;

[0171] In the fourth stage, the size of the reference image is adjusted by a preset training model to obtain a first set of icon images.

[0172] In one embodiment, the image acquisition module 10 is further configured to acquire a reference icon image including hidden information, and perform inverse embedding processing on the hidden information to obtain the hidden information;

[0173] Obtain a standard icon image, embed the hidden information into the hidden blocks of the standard icon image, and obtain a third icon image set;

[0174] The step of embedding the hidden information into the hidden blocks of the standard icon image to obtain the third icon image set includes:

[0175] The white or black pixels in the standard image are flipped using a graph topology algorithm to obtain hidden blocks.

[0176] The hidden information is embedded into the hidden block to obtain a third set of icon images.

[0177] In one embodiment, the feature matching module 20 is further configured to extract hidden information features and physical information features of each reference icon image in the reference icon image set;

[0178] The hidden information features are matched with the hidden information features of each webpage image in the preset webpage image set. When the hidden information features are successfully matched with the hidden information of each webpage image, the successfully matched webpage image is taken as the target image.

[0179] The physical information features are matched with the physical information features of each webpage image in the preset webpage image set. When the physical information features match the physical information features of each webpage image, the successfully matched webpage image is taken as the target image.

[0180] In one embodiment, the website classification module 30 is further configured to input the target image into a preset target detection model to obtain category features and location features in the target image;

[0181] Semantic recognition is performed on the target image to obtain the business attribute features of the target image;

[0182] The feature vector of the target image is obtained by combining the category features, the location features, and the business attribute features.

[0183] In one embodiment, the asset identification module 40 is further configured to regard the domain name information of the target enterprise's current website and the domain name information of the target enterprise's historical websites as a type of asset of the target enterprise;

[0184] Based on a preset enterprise relationship database, it is determined whether the target enterprise and the non-target enterprise website have a cooperative relationship. The domain name information of the non-target enterprise website that has a cooperative relationship with the target enterprise is regarded as the second type of asset of the target enterprise.

[0185] The assets of the target enterprise are determined based on the first type of assets and the second type of assets.

[0186] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0187] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0188] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0189] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0190] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0191] It should be understood that although the steps in the flowcharts of this application's embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.

[0192] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. An asset identification method, characterized in that, The asset identification method includes: Obtain reference icon images of the target company, and expand the reference icon images to obtain a reference icon image set; Extract reference information features from each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching results; The feature vector in the target image is obtained, and the feature vector is input into a preset website classification neural network model to obtain the website category of the target image. The assets of the target company are obtained based on the website category of the target image; The step of expanding the reference icon image to obtain a reference icon image set includes: The reference icon image is input into a pyramid network, which includes a first stage, a second stage, a third stage, and a fourth stage. In the first stage, the reference icon image is iterated multiple times to obtain a low-resolution image. In the second stage, features of the low-resolution image are extracted using convolutional layers to obtain multiple similar images. The generator size is increased by adding convolutional layers, and each stage uses the icon features from the previous stage as input. Each stage trains on images of different sizes to generate an arbitrary number of similar images. In the third stage, the dimensions of the multiple similar images are defined to obtain multiple reference images of different sizes. The size definition of the similar images in each stage can refer to the formula. , Indicates the stage, Indicates the total number of stages. To represent a scalar, defined as follows: When the stage is 3, the corresponding size is 25. 34, 32 42, 42 56, 63 84; In the fourth stage, the size of the reference image is adjusted by a preset training model to obtain a first set of icon images, wherein the size parameters are continuously changed and a generative adversarial network is used to train the model. The reference icon image is expanded according to a preset conventional icon deformation rule to obtain a second set of icon images; Obtain a reference icon image containing hidden information, and expand the reference icon image containing hidden information to obtain a third set of icon images; The first set of icon images, the second set of icon images, and the third set of icon images are used as reference icon image sets.

2. The asset identification method as described in claim 1, characterized in that, The step involves obtaining a reference icon image containing hidden information, and then expanding that reference icon image to obtain a third set of icon images, including: Obtain a reference icon image including hidden information, and perform inverse embedding processing on the hidden information to obtain the hidden information; Obtain a standard icon image, embed the hidden information into the hidden blocks of the standard icon image, and obtain a third icon image set; The step of embedding the hidden information into the hidden blocks of the standard icon image to obtain the third icon image set includes: The white or black pixels in the standard icon image are flipped using a graph topology algorithm to obtain hidden blocks. The hidden information is embedded into the hidden block to obtain a third set of icon images.

3. The asset identification method as described in claim 1, characterized in that, The reference information features include hidden information features and physical information features; the web page features of each web page image include hidden information features and physical information features of each web page image. The step of extracting reference information features from each reference icon image in the reference icon image set, matching the reference information features with the web page features of each web page image in the preset web page image set, and obtaining the target image based on the matching result includes: Extract the hidden information features and physical information features of each reference icon image in the reference icon image set; The hidden information features are matched with the hidden information features of each webpage image in the preset webpage image set. When the hidden information features are successfully matched with the hidden information of each webpage image, the successfully matched webpage image is taken as the target image. The physical information features are matched with the physical information features of each webpage image in the preset webpage image set. When the physical information features match the physical information features of each webpage image, the successfully matched webpage image is taken as the target image.

4. The asset identification method as described in claim 1, characterized in that, The step of obtaining the feature vector in the target image includes: The target image is input into a preset target detection model to obtain the category features and location features in the target image; Semantic recognition is performed on the target image to obtain the business attribute features of the target image; The feature vector of the target image is obtained by combining the category features, the location features, and the business attribute features.

5. The asset identification method as described in any one of claims 1-4, characterized in that, The website categories include the target company's current website, the target company's historical websites, and websites not belonging to the target company. The step of obtaining the target company's assets based on the website category of the target image includes: The domain name information of the target company's current website and the domain name information of the target company's historical websites are regarded as a type of asset of the target company; Based on a preset enterprise relationship database, it is determined whether the target enterprise and the non-target enterprise website have a cooperative relationship. The domain name information of the non-target enterprise website that has a cooperative relationship with the target enterprise is regarded as the second type of asset of the target enterprise. The assets of the target enterprise are determined based on the first type of assets and the second type of assets.

6. An asset identification device, characterized in that, The asset identification device includes: The image acquisition module is used to acquire reference icon images of the target enterprise and expand the reference icon images to obtain a reference icon image set; The feature matching module is used to extract reference information features of each reference icon image in the reference icon image set, match the reference information features with the web page features of each web page image in the preset web page image set, and obtain the target image based on the matching result. The website classification module is used to obtain feature vectors from the target image, input the feature vectors into a preset website classification neural network model, and obtain the website category of the target image. The asset identification module is used to obtain the assets of the target enterprise based on the website category of the target image; The image acquisition module is also used to input the reference icon image into the pyramid network, which includes a first stage, a second stage, a third stage, and a fourth stage. In the first stage, the reference icon image is iterated multiple times to obtain a low-resolution image; In the second stage, features of the low-resolution image are extracted through convolutional layers to obtain multiple similar images. The generator size is increased by adding convolutional layers, and each stage takes the icon features of the previous stage as input. Each stage trains images of different sizes to generate any number of similar images. In the third stage, the dimensions of the multiple similar images are defined to obtain multiple reference images of different sizes. The dimensions of the similar images at each stage can be defined with reference to the formula. , Indicates the stage, Indicates the total number of stages. To represent a scalar, defined as follows: When the stage is 3, the corresponding size is 25. 34, 32 42, 42 56, 63 84; In the fourth stage, the size of the reference image is adjusted by a preset training model to obtain a first set of icon images, wherein the size parameters are continuously changed and a generative adversarial network is used to train the model. The reference icon image is expanded according to a preset conventional icon deformation rule to obtain a second set of icon images; Obtain a reference icon image containing hidden information, and expand the reference icon image containing hidden information to obtain a third set of icon images; The first set of icon images, the second set of icon images, and the third set of icon images are used as reference icon image sets.

7. An asset identification device, characterized in that, The device includes: a memory, a processor, and an asset identification program stored in the memory and executable on the processor, the asset identification program being configured to implement the asset identification method as described in any one of claims 1 to 5.

8. A storage medium, characterized in that, The storage medium stores an asset identification program, which, when executed by a processor, implements the asset identification method as described in any one of claims 1 to 5.

Images