Website page monitoring method and device based on vectorization and multiple modes, equipment and medium

Abstract

The invention relates to a website page monitoring method, device and equipment based on vectorization and multiple modes and a medium, and is applied to the technical field of computers, the method comprises the following steps: obtaining update information of a target website in real time, and capturing a dynamic picture of the target website to obtain the dynamic picture; obtaining static processing information of the target website in real time, and capturing a static picture of the target website to obtain the static picture; performing steering quantity processing on the dynamic picture and the static picture to generate a dynamic picture vector and a static picture vector; calculating an Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector; judging whether the static picture meets a preset normal condition based on the Euclidean distance; if the preset normal condition is met, judging that the page of the static picture is normal; if the preset normal condition is not met, judging that the page of the static picture is abnormal; and processing the static picture based on a preset multi-modal model. The static picture detection method has the effect of efficiently and accurately detecting the static picture.

Description

Technical Field

[0001] The present application relates to the technical field of computers, and in particular to a method, device, equipment and medium for monitoring website pages based on vectorization and multimodality. Background Art

[0002] In order to improve user experience while reducing the processing burden of the server, most companies will choose to make their websites static. Website staticization refers to the process of converting web page content originally generated by dynamic scripts into static HTML files.

[0003] After traditional corporate websites are static, in order to ensure the normal operation of the website, the website needs to be checked and maintained. The general processing method is to manually click to check the pages one by one, and feedback to the technical staff for processing when there are problems with the pages, or use scripts to read and detect the status code and the last modification time according to the URL in the sitemap using the header method. If the status code is 200 and the last modification time is today, it can be determined that the staticization has been completed normally.

[0004] Currently, both commonly used solutions have obvious problems. The former can be completed relatively easily and quickly for corporate websites with fewer pages, but for some customer websites with a large number of pages and multi-language modes, the number of pages that need to be checked may exceed 10,000, which is difficult to handle manually. The latter status code and last modification time method can only ensure that the staticization is executed on this page, but it cannot effectively detect whether there are style problems in the page. There is an urgent need for a technology that can efficiently and accurately perform static image detection. Summary of the invention

[0005] In order to achieve efficient and accurate static image detection, the present application provides a website page monitoring method, device, equipment and medium based on vectorization and multimodality.

[0006] In the first aspect, the present application provides a website page monitoring method based on vectorization and multimodality, which adopts the following technical solutions:

[0007] A website page monitoring method based on vectorization and multimodality, comprising:

[0008] Acquire update information of a target website in real time, and access and capture dynamic images of the target website based on the update information to obtain dynamic images;

[0009] Acquire static processing information of a target website in real time, and capture a static image of the target website based on the static processing information to obtain a static image;

[0010] Performing steering amount processing on the dynamic picture and the static picture respectively to generate a dynamic picture vector and a static picture vector;

[0011] Calculate the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector;

[0012] Determining whether the static image meets a preset normal condition based on the Euclidean distance;

[0013] If the static image meets the preset normal condition, determining that the page of the target website corresponding to the static image is normal;

[0014] If the static image does not meet the preset normal condition, determining that the page of the target website corresponding to the static image is abnormal;

[0015] The static image is processed based on a preset multimodal model to generate a processing result.

[0016] By adopting the above technical solution, before the page is staticized, the page image in the dynamic page state is captured to obtain a dynamic picture used as a reference. After the page is staticized, the page image in the static page state is captured to obtain a static picture for monitoring. The pictures obtained in the two states are respectively processed with steering quantities to obtain dynamic picture vectors and static picture vectors. The Euclidean distance between the vectors is practically compared to determine whether the static picture is in a normal state. If it is not in a normal state, a preset multimodal model will be used for further processing. The capture of dynamic images and static images is all performed by a computer, thereby realizing efficient and accurate static picture detection.

[0017] Optionally, the performing steering amount processing on the dynamic image and the static image respectively to generate a dynamic image vector and a static image vector comprises:

[0018] Acquire a standard zoom size, and modify the size of the dynamic image and the static image based on the standard zoom size to generate a modified dynamic image and a modified static image;

[0019] Performing adjustment processing on the modified dynamic image based on a preset convolution layer and a preset pooling layer to generate an adjusted dynamic image;

[0020] Normalizing and activating the dynamic image to generate a dynamic image processing result;

[0021] Characterizing the dynamic image processing result, extracting the steering amount, and generating a dynamic image vector;

[0022] Performing adjustment processing on the modified static image based on a preset convolution layer and a preset pooling layer to generate an adjusted static image;

[0023] Normalizing and activating the adjusted static image to generate a static image processing result;

[0024] The static image processing result is characterized to extract the steering amount and generate a static image vector.

[0025] Optionally, the calculating the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector includes:

[0026] Compressing the static image vector based on a preset algorithm to generate a static compressed vector;

[0027] Performing compression processing on the dynamic picture vector based on the preset algorithm to generate a dynamic compression vector;

[0028] Get the target calculation formula;

[0029] The Euclidean distance between the dynamic image and the static image is calculated based on the target calculation formula, the static compression vector and the dynamic compression vector.

[0030] Optionally, the processing the static image based on a preset multimodal model includes:

[0031] Based on the preset multimodal model, text recognition is performed on the static image and the dynamic image respectively to determine static text information and dynamic text information;

[0032] Preprocessing the static text information to generate static text to be used;

[0033] Characterize the static text to be used, extract the steering amount, and generate a static text vector;

[0034] Preprocessing the dynamic text information to generate dynamic text to be used;

[0035] Characterize the dynamic text to be used, extract the steering amount, and generate a dynamic text vector;

[0036] Obtain target similarity calculation rules;

[0037] Calculating page similarity between the static image and the dynamic image based on the target similarity calculation rule, the static text vector and the dynamic text vector;

[0038] If the page similarity is greater than a preset threshold, it is determined that the page of the target website corresponding to the static image is normal;

[0039] If the page similarity is less than a preset threshold, the step of processing the static image based on a preset multimodal model is repeatedly executed until the number of repeated executions reaches a preset repetition number threshold.

[0040] Optionally, after compressing the dynamic picture vector based on the preset algorithm to generate a dynamic compressed vector, the method further includes:

[0041] Acquire dynamic resource information of the dynamic image and static resource information of the static image;

[0042] generating a dynamic vector value based on the dynamic resource information, and generating a static vector value based on the static resource information;

[0043] Binding the compressed dynamic vector to the dynamic vector value to generate dynamic vector data;

[0044] Binding the compressed static vector to the static vector value to generate static vector data;

[0045] The dynamic vector data and the static vector data are stored in a preset storage system.

[0046] Optionally, after repeatedly executing the step of processing the static image based on the preset multimodal model in the base until the number of repeated executions reaches a preset repetition number threshold, the method further includes:

[0047] Acquire static related information of the static picture and dynamic related information of the dynamic picture;

[0048] generating a manual review request based on the static related information and the dynamic related information;

[0049] The manual review request is sent to the mobile terminal of the target work reviewer.

[0050] Optionally, after repeatedly executing the step of processing the static image based on the preset multimodal model in the base until the number of repeated executions reaches a preset repetition number threshold, the method further includes:

[0051] Get the first processing time and each repeated processing time;

[0052] generating an execution record based on the first processing time, the repeated processing time, the dynamic related information and the static related information;

[0053] The execution record is written into a preset feedback data table.

[0054] In a second aspect, the present application provides a website page monitoring device based on vectorization and multimodality, which adopts the following technical solution:

[0055] A website page monitoring device based on vectorization and multimodality, comprising:

[0056] A dynamic image generation module is used to obtain the update information of the target website in real time, and to access and capture the dynamic images of the target website based on the update information to obtain the dynamic images;

[0057] A static image generation module is used to obtain static processing information of a target website in real time, and to capture a static image of the target website based on the static processing information to obtain a static image;

[0058] A picture vector generation module, used for performing steering amount processing on the dynamic picture and the static picture respectively to generate a dynamic picture vector and a static picture vector;

[0059] A Euclidean distance calculation module, configured to calculate the Euclidean distance between the dynamic image and the static image based on the dynamic image vector and the static image vector;

[0060] A preset condition judgment module, used for judging whether the static image meets the preset normal condition based on the Euclidean distance;

[0061] A page normality determination module, used to determine whether the page of the target website corresponding to the static image is normal;

[0062] A page anomaly determination module, used to determine if a page of a target website corresponding to the static image is abnormal;

[0063] The abnormal page processing module is used to process the static image based on a preset multimodal model and generate a processing result.

[0064] By adopting the above technical solution, before the page is staticized, the page image in the dynamic page state is captured to obtain a dynamic picture used as a reference. After the page is staticized, the page image in the static page state is captured to obtain a static picture for monitoring. The pictures obtained in the two states are respectively processed with steering quantities to obtain dynamic picture vectors and static picture vectors. The Euclidean distance between the vectors is practically compared to determine whether the static picture is in a normal state. If it is not in a normal state, a preset multimodal model will be used for further processing. The capture of dynamic images and static images is all performed by a computer, thereby realizing efficient and accurate static picture detection.

[0065] In a third aspect, the present application provides an electronic device, which adopts the following technical solution:

[0066] An electronic device comprises a processor, wherein the processor is coupled to a memory;

[0067] The processor is used to execute the computer program stored in the memory so that the electronic device executes the computer program of the website page monitoring method based on vectorization and multimodality as described in any one of the first aspects.

[0068] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution:

[0069] A computer-readable storage medium stores a computer program that can be loaded by a processor and execute the website page monitoring method based on vectorization and multimodality as described in any one of the first aspects.

[0070] In summary, this application includes the following beneficial technical effects:

[0071] Before the page is staticified, the page image in the dynamic page state is captured to obtain a dynamic image used as a reference. After the page is staticified, the page image in the static page state is captured to obtain a static image for monitoring. The images obtained in the two states are respectively processed with steering quantities to obtain dynamic image vectors and static image vectors. The Euclidean distance between the vectors is practically compared to determine whether the static image is in a normal state. If it is not a normal state, a preset multimodal model will be used for further processing. The capture of dynamic images and static images is all performed by a computer, thereby achieving efficient and accurate static image detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a flow chart of a website page monitoring method based on vectorization and multimodality provided in an embodiment of the present application.

[0073] Figure 2 It is a structural block diagram of a website page monitoring device based on vectorization and multimodality provided in an embodiment of the present application.

[0074] Figure 3 It is a structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0075] The present application is further described in detail below in conjunction with the accompanying drawings.

[0076] The embodiment of the present application provides a website page monitoring method based on vectorization and multimodality, which can be executed by an electronic device, which can be a server or a terminal device, wherein the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smart phone, a tablet computer, a desktop computer, etc., but is not limited thereto.

[0077] Figure 1 A flowchart of a website page monitoring method based on vectorization and multimodality provided in an embodiment of the present application.

[0078] like Figure 1 As shown, the main process of the method is described as follows (steps S101 to S108):

[0079] Step S101, obtaining update information of a target website in real time, and accessing and capturing dynamic images of the target website based on the update information to obtain dynamic images.

[0080] In this embodiment, the target website is monitored and the update information of the target website is obtained in real time. After the target website is updated, the target website is visited and the page of the target website is captured to obtain a dynamic image. The dynamic image here refers to an image captured when the page of the target website is in a dynamic page state.

[0081] Generally, the pages of the target website to be monitored are pages that can be accessed through the homepage navigation bar. For example, the navigation bar of the homepage of an ordinary B2B type website consists of columns including homepage, solutions, videos, products, about us, corporate news, FAQ, contact us, etc. Among them, videos, products, corporate news, and FAQ are third-level pages, including their own column homepage, column list page, and column specific content details page, while homepage, solutions, about us, and contact us are separate pages. If it is a multilingual website, the minority language pages corresponding to each page are also the key objects to be monitored. After the website content maintenance is completed, all URL addresses of the current website will be reproduced and generated into the website's site map. The site map is the address of all pages to be monitored. After the site map is updated, a Google Chrome headless browser is started on the background server. The headless browser then visits all the addresses saved in the site map one by one. Each time an address is visited, after waiting for the address to be completely loaded, the browser screenshot function is called to automatically capture the content of the page and save it to the specified directory, thereby obtaining a dynamic image of the page.

[0082] Step S102, acquiring the static processing information of the target website in real time, and capturing the static image of the target website based on the static processing information to obtain the static image.

[0083] In this embodiment, the target website is also monitored, and the static processing information of the target website is obtained in real time. After the target website completes the static processing, the target website is accessed and the page of the target website is captured to obtain a static picture. The static picture here refers to a picture captured when the page of the target website is in a static page state.

[0084] Generally, SAAS websites have a static refresh function in the background, which automatically calls the static refresh function. This function will dynamically access this page through curl according to the address recorded in the previous site map, and then save the source code returned by this page as a new static file in the Save As method. This file address is the static URL accessed by the end customer and CDN. After saving, enable the chrome headless browser on the server side in the same way as the dynamic method, and let the headless browser request the static URL. After all loading is completed, call the browser screenshot function to automatically capture the content of the static page and save it to another specified directory, so as to obtain a static image of the page.

[0085] Step S103, performing steering amount processing on the dynamic image and the static image respectively to generate a dynamic image vector and a static image vector.

[0086] For step S103, a standard scaling size is obtained, and the dynamic image and the static image are resized based on the standard scaling size to generate a modified dynamic image and a modified static image; the modified dynamic image is adjusted based on the preset convolution layer and the preset pooling layer to generate an adjusted dynamic image; the adjusted dynamic image is normalized and activated by a function to generate a dynamic image processing result; the dynamic image processing result is characterized and a steering amount is extracted to generate a dynamic image vector; the modified static image is adjusted based on the preset convolution layer and the preset pooling layer to generate an adjusted static image; the adjusted static image is normalized and activated by a function to generate a static image processing result; the static image processing result is characterized and a steering amount is extracted to generate a static image vector.

[0087] In this embodiment, the conversion of image vectors is achieved through embedding technology. First, all dynamic images and static images are batch-scaled according to the standard scaling size, and are uniformly changed to modified dynamic images and modified static images with a length and width size of 224*224 pixels.

[0088] For each single picture, the same processing method is used to process the steering amount. The specific processing steps are as follows: the first step is to process it with a convolution layer with a step size of 2 and 7*7, and the processing is [1,64,112,112], that is, 64 pictures of 112*112 size, where 64 means that there are 64 feature extraction methods and a total of 64*112*112 feature points. Next, it is processed again through normalization and activation function operations, and a pooling layer with a step size of 2, and the output size is in the form of [1,256,56,56]. Similarly, it is processed again through normalization and activation function operations, and a pooling layer with a step size of 2, and the output size is [1,512,28,28], thereby obtaining the corresponding adjusted dynamic pictures and adjusted static pictures. After processing to this step, the steering amount operation is extracted for the representation of this layer to obtain the dynamic picture vector and static picture vector corresponding to the picture.

[0089] Step S104, calculating the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector.

[0090] For step S104, the static image vector is compressed based on a preset algorithm to generate a static compressed vector; the dynamic image vector is compressed based on a preset algorithm to generate a dynamic compressed vector; the target calculation formula is obtained; and the Euclidean distance between the dynamic image and the static image is calculated based on the target calculation formula, the static compressed vector and the dynamic compressed vector.

[0091] In this embodiment, in order to facilitate the storage of image vector data, all image vectors are compressed by the Brotli algorithm to obtain static compressed vectors and dynamic compressed vectors. When calculating the Euclidean distance between dynamic images and static images, it is necessary to calculate two images from the same URL. For example, if the URL of the dynamic image is 1, then it is necessary to calculate the Euclidean distance between the static image with the same URL of 1. When calculating, it can be automatically calculated through the target calculation formula set in numpy. The specific target calculation formula needs to be set according to actual needs, and it only needs to be able to calculate the Euclidean distance between two vectors. No specific limitation is made here.

[0092] After compressing the static image vector and the dynamic image vector, the dynamic resource information of the dynamic image and the static resource information of the static image are obtained; a dynamic vector value is generated based on the dynamic resource information, and a static vector value is generated based on the static resource information; the compressed dynamic vector is bound to the dynamic vector value to generate dynamic vector data; the compressed static vector is bound to the static vector value to generate static vector data; and the dynamic vector data and the static vector data are stored in a preset storage system.

[0093] In this embodiment, the dynamic resource information is the domain name and URL of the dynamic image, and the static resource information is the domain name and URL of the static image. The domain name and URL are used to perform hash operations to obtain dynamic vector values ​​and static vector values. For static images and static images with the same URL, the dynamic vector values ​​and static vector values ​​obtained are the same. The dynamic vector value is then bound to the compressed dynamic vector to generate dynamic vector data, so that the corresponding compressed dynamic vector can be quickly found through the dynamic vector value. Similarly, the static vector value is bound to the compressed static vector to generate static vector data, so that the corresponding compressed static vector can be quickly found through the static vector value. Finally, the dynamic vector data and static vector data are stored in a preset storage system. The preset storage system can be specifically Redis, or other storage systems. The specific storage system is not specifically limited here.

[0094] Step S105 , judging whether the static image meets the preset normal condition based on the Euclidean distance.

[0095] In this embodiment, the normal condition is preset and the Euclidean distance calculation is performed based on the image determined to be normal. The image determined to be normal is then abnormalized and the Euclidean distance of the image after abnormalization is calculated. Based on the difference between the normal image and the abnormalized image on the same website page, a value is determined and used as a threshold for determination. When the Euclidean distance is greater than or equal to the threshold, it is determined to meet the normal condition. When the Euclidean distance is less than the threshold, it is determined to not meet the normal condition.

[0096] Step S106: If the static image meets the preset normal condition, it is determined that the page of the target website corresponding to the static image is normal.

[0097] In this embodiment, if the calculated Euclidean distance is greater than or equal to the preset normal condition, the static image is determined to be a normal image, and then it is further determined that the website page corresponding to the static image is a normal page.

[0098] Step S107: if the static image does not meet the preset normal condition, it is determined that the page of the target website corresponding to the static image is abnormal.

[0099] In this embodiment, if the calculated Euclidean distance is less than the preset normal condition, the static image will be judged as an abnormal image, and then further determined that the website page corresponding to the static image is an abnormal page, that is, the page after static processing is an abnormal page and needs to be verified again.

[0100] Step S108: Process the static image based on the preset multimodal model to generate a processing result.

[0101] For step S108, text recognition is performed on static images and dynamic images respectively based on a preset multimodal model to determine static text information and dynamic text information; the static text information is preprocessed to generate a static text to be used; the static text to be used is characterized and a diversion amount is extracted to generate a static text vector; the dynamic text information is preprocessed to generate a dynamic text to be used; the dynamic text to be used is characterized and a diversion amount is extracted to generate a dynamic text vector; the target similarity calculation rule is obtained; the page similarity of the static image and the dynamic image is calculated based on the target similarity calculation rule, the static text vector and the dynamic text vector; if the page similarity is greater than a preset threshold, it is determined that the page of the target website corresponding to the static image is normal; if the page similarity is less than the preset threshold, the step of processing the static image based on the preset multimodal model is repeatedly executed until the number of repeated executions reaches the preset repetition number threshold.

[0102] In this embodiment, when the abnormal picture is subjected to secondary verification, the abnormal static picture and the dynamic picture matching it will be used for verification again, and the multimodal function of the large language model such as the CogVLM2 visual model will be used for the static picture and the dynamic picture respectively to perform text recognition operations on the picture, identify all the text information in the current picture, and obtain the static text information extracted from the static picture and the dynamic text information extracted from the dynamic picture, and then process the static text information and the dynamic text information.

[0103] The static text information is subjected to preprocessing steps such as removing punctuation marks, segmenting, removing stop words, removing common meaningless words, standardizing, removing redundant spaces, etc. to obtain the stand-by static text. After the preprocessing is completed, the static text to be used is characterized and the steering quantity is extracted by the method in the tfidfVectorizer class provided in the sklearn library in the Python language to obtain the static text vector. The dynamic text information is processed in the same manner as the static text information, and the dynamic text information is subjected to preprocessing steps such as removing punctuation marks, segmenting, removing stop words, removing common meaningless words, standardizing, removing redundant spaces, etc. to obtain the stand-by dynamic text. After the preprocessing is completed, the dynamic text to be used is characterized and the steering quantity is extracted by the method in the tfidfVectorizer class provided in the sklearn library in the Python language to obtain the dynamic text vector.

[0104] After obtaining the static text vector and the dynamic text vector, the similarity between the static text vector and the dynamic text vector is calculated using a target similarity calculation rule, wherein the target similarity calculation rule may specifically be cosine similarity or other similar similarity calculation formulas, which are not specifically limited here.

[0105] Then, a second judgment is made in the same way as the first judgment. Here, the setting method of the preset threshold used for similarity judgment is the same as above, that is, Euclidean distance calculation is performed based on the image judged to be normal, and then the image judged to be normal is abnormalized, and the similarity of the image after abnormalization is calculated. According to the difference between the normal image and the abnormalized image on the same website page, a value is determined, and the value is used as the threshold for judgment. When the similarity is greater than or equal to the threshold, it is judged to meet the normal conditions. When the similarity is less than the threshold, it is judged to not meet the normal conditions. When it is judged to not meet the normal conditions, the second verification step is repeated until the number of verifications reaches the preset repetition threshold.

[0106] In this embodiment, static related information of static pictures and dynamic related information of dynamic pictures are obtained; a manual review request is generated based on the static related information and the dynamic related information; and the manual review request is sent to a mobile terminal of a target work reviewer.

[0107] After reaching the preset repetition threshold, in order to ensure the normal operation of the target website, the static related information of the static image and the dynamic related information of the dynamic image are summed up to generate a manual review request, and the generated manual review request is sent to the mobile terminal of the target work reviewer to remind the target work reviewer to review the website pages, among which the static related information includes static images, domain names and URLs, etc., and the dynamic related information is consistent with the static related information, and the dynamic related information includes dynamic images, domain names and URLs, etc.

[0108] In this embodiment, the first processing time and each repeated processing time are obtained; an execution record is generated based on the first processing time, the repeated processing time, dynamic related information and static related information; and the execution record is written into a preset feedback data table.

[0109] In addition to the manual review request mentioned above, a recording function is also provided. The processing time for calculating the Euclidean distance is used as the first processing time, and the similarity processing time for each time is used as the repeated processing time. The first processing time, repeated processing time, dynamic related information and static related information are integrated and bound to generate an execution record. The execution record is written into the preset feedback data table to facilitate the staff to check the execution status of each time.

[0110] Figure 2 A structural block diagram of a website page monitoring device 200 based on vectorization and multimodality provided in an embodiment of the application.

[0111] like Figure 2 As shown, the website page monitoring device 200 based on vectorization and multimodality mainly includes:

[0112] The dynamic image generation module 201 is used to obtain the update information of the target website in real time, and access and capture the dynamic image of the target website based on the update information to obtain the dynamic image;

[0113] The static image generation module 202 is used to obtain the static processing information of the target website in real time, and to capture the static image of the target website based on the static processing information to obtain the static image;

[0114] The picture vector generation module 203 is used to perform steering amount processing on the dynamic picture and the static picture respectively to generate a dynamic picture vector and a static picture vector;

[0115] The Euclidean distance calculation module 204 is used to calculate the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector;

[0116] A preset condition judgment module 205 is used to judge whether the static image meets the preset normal condition based on the Euclidean distance;

[0117] A page normality determination module 206 is used to determine whether the page of the target website corresponding to the static image is normal;

[0118] A page anomaly determination module 207 is used to determine whether a page of a target website corresponding to a static image is abnormal;

[0119] The abnormal page processing module 208 is used to process the static image based on the preset multimodal model and generate a processing result.

[0120] As an optional implementation manner of this embodiment, the image vector generation module 203 is specifically used to obtain a standard scaling size, modify the size of the dynamic image and the static image based on the standard scaling size, and generate a modified dynamic image and a modified static image; adjust the modified dynamic image based on the preset convolution layer and the preset pooling layer to generate an adjusted dynamic image; normalize and activate the adjusted dynamic image to generate a dynamic image processing result; characterize and extract the steering amount of the dynamic image processing result to generate a dynamic image vector; adjust the modified static image based on the preset convolution layer and the preset pooling layer to generate an adjusted static image; normalize and activate the adjusted static image to generate a static image processing result; characterize and extract the steering amount of the static image processing result to generate a static image vector.

[0121] As an optional implementation of this embodiment, the Euclidean distance calculation module 204 is specifically used to compress the static image vector based on a preset algorithm to generate a static compression vector; compress the dynamic image vector based on a preset algorithm to generate a dynamic compression vector; obtain a target calculation formula; and calculate the Euclidean distance between the dynamic image and the static image based on the target calculation formula, the static compression vector and the dynamic compression vector.

[0122] As an optional implementation of this embodiment, the abnormal page processing module 208 is specifically used to perform text recognition on static images and dynamic images respectively based on a preset multimodal model to determine static text information and dynamic text information; preprocess the static text information to generate a static text to be used; characterize and extract the diversion amount of the static text to be used to generate a static text vector; preprocess the dynamic text information to generate a dynamic text to be used; characterize and extract the diversion amount of the dynamic text to be used to generate a dynamic text vector; obtain the target similarity calculation rule; calculate the page similarity of the static image and the dynamic image based on the target similarity calculation rule, the static text vector and the dynamic text vector; if the page similarity is greater than the preset threshold, it is determined that the page of the target website corresponding to the static image is normal; if the page similarity is less than the preset threshold, the step of processing the static image based on the preset multimodal model is repeated until the number of repeated executions reaches the preset repetition threshold.

[0123] As an optional implementation of this embodiment, the website page monitoring device 200 based on vectorization and multimodality also includes:

[0124] A resource information acquisition module, used to acquire dynamic resource information of dynamic images and static resource information of static images;

[0125] A vector value generation module, used to generate dynamic vector values ​​based on dynamic resource information and to generate static vector values ​​based on static resource information;

[0126] A dynamic data generation module, used for binding the compressed dynamic vector with the dynamic vector value to generate dynamic vector data;

[0127] A static data generation module, used for binding the compressed static vector with the static vector value to generate static vector data;

[0128] The vector data storage module is used to store dynamic vector data and static vector data into a preset storage system.

[0129] As an optional implementation of this embodiment, the website page monitoring device 200 based on vectorization and multimodality also includes:

[0130] A related information acquisition module, used to acquire static related information of static pictures and dynamic related information of dynamic pictures;

[0131] A review request generation module, used for generating a manual review request based on static related information and dynamic related information;

[0132] The review request sending module is used to send the manual review request to the mobile terminal of the target work reviewer.

[0133] As an optional implementation of this embodiment, the website page monitoring device 200 based on vectorization and multimodality also includes:

[0134] A processing time acquisition module is used to obtain the first processing time and each repeated processing time;

[0135] An execution record generation module, used to generate an execution record based on the first processing time, the repeated processing time, the dynamic related information and the static related information;

[0136] The execution record writing module is used to write the execution record into the preset feedback data table.

[0137] In one example, the module in any of the above devices can be one or more integrated circuits configured to implement the above methods, such as: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0138] For another example, when the modules in the device can be implemented in the form of a processing element scheduling program, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call a program. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0140] Figure 3 This is a structural block diagram of an electronic device 300 provided in an embodiment of the present application.

[0141] like Figure 3 As shown, the electronic device 300 includes a processor 301 and a memory 302 , and may further include an information input / information output (I / O) interface 303 , one or more of a communication component 304 , and a communication bus 305 .

[0142] The processor 301 is used to control the overall operation of the electronic device 300 to complete all or part of the steps of the above-mentioned vectorization and multimodal website page monitoring method; the memory 302 is used to store various types of data to support the operation of the electronic device 300, and these data may include, for example, instructions for any application or method operated on the electronic device 300, and application-related data. The memory 302 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), programmable read-only memory (Programmable Read-Only Memory, PROM), read-only memory (Read-Only Memory, ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0143] The I / O interface 303 provides an interface between the processor 301 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 304 is used for wired or wireless communication between the electronic device 300 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 104 can include: Wi-Fi components, Bluetooth components, NFC components.

[0144] The electronic device 300 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the vectorization and multimodal website page monitoring method given in the above embodiment.

[0145] The communication bus 305 may include a path to transmit information between the above components. The communication bus 305 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 305 may be divided into an address bus, a data bus, a control bus, etc.

[0146] The electronic device 300 may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc., and may also be servers, etc.

[0147] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned website page monitoring method based on vectorization and multimodality are implemented.

[0148] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0149] The terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article, or apparatus.

[0150] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of application involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the aforementioned application concept. For example, the above features are replaced with (but not limited to) technical features with similar functions applied in the present application.

Claims

1. A website page monitoring method based on vectorization and multimodality, characterized in that: include: Acquire update information of a target website in real time, and access and capture dynamic images of the target website based on the update information to obtain dynamic images; Acquire static processing information of a target website in real time, and capture a static image of the target website based on the static processing information to obtain a static image; Performing steering amount processing on the dynamic picture and the static picture respectively to generate a dynamic picture vector and a static picture vector; Calculate the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector; Determining whether the static image meets a preset normal condition based on the Euclidean distance; If the static image meets the preset normal condition, determining that the page of the target website corresponding to the static image is normal; If the static image does not meet the preset normal condition, determining that the page of the target website corresponding to the static image is abnormal; The static image is processed based on a preset multimodal model to generate a processing result.

2. The method according to claim 1, characterized in that The processing of the dynamic image and the static image respectively to generate a dynamic image vector and a static image vector comprises: Acquire a standard zoom size, and modify the size of the dynamic image and the static image based on the standard zoom size to generate a modified dynamic image and a modified static image; Performing adjustment processing on the modified dynamic image based on a preset convolution layer and a preset pooling layer to generate an adjusted dynamic image; Normalizing and activating the dynamic image to generate a dynamic image processing result; Characterizing the dynamic image processing result, extracting the steering amount, and generating a dynamic image vector; Performing adjustment processing on the modified static image based on a preset convolution layer and a preset pooling layer to generate an adjusted static image; Normalizing and activating the adjusted static image to generate a static image processing result; The static image processing result is characterized to extract the steering amount and generate a static image vector.

3. The method according to claim 1, characterized in that: The calculating the Euclidean distance between the dynamic picture and the static picture based on the dynamic picture vector and the static picture vector comprises: Compressing the static image vector based on a preset algorithm to generate a static compressed vector; Performing compression processing on the dynamic picture vector based on the preset algorithm to generate a dynamic compression vector; Get the target calculation formula; The Euclidean distance between the dynamic image and the static image is calculated based on the target calculation formula, the static compression vector and the dynamic compression vector.

4. The method according to claim 1, characterized in that: The processing of the static picture based on the preset multimodal model includes: Based on the preset multimodal model, text recognition is performed on the static image and the dynamic image respectively to determine static text information and dynamic text information; Preprocessing the static text information to generate static text to be used; Characterize the static text to be used, extract the steering amount, and generate a static text vector; Preprocessing the dynamic text information to generate dynamic text to be used; Characterize the dynamic text to be used, extract the steering amount, and generate a dynamic text vector; Obtain target similarity calculation rules; Calculating page similarity between the static image and the dynamic image based on the target similarity calculation rule, the static text vector and the dynamic text vector; If the page similarity is greater than a preset threshold, it is determined that the page of the target website corresponding to the static image is normal; If the page similarity is less than a preset threshold, the step of processing the static image based on a preset multimodal model is repeatedly executed until the number of repeated executions reaches a preset repetition number threshold.

5. The method according to claim 3, characterized in that: After the dynamic picture vector is compressed based on the preset algorithm to generate a dynamic compressed vector, the method further includes: Acquire dynamic resource information of the dynamic image and static resource information of the static image; generating a dynamic vector value based on the dynamic resource information, and generating a static vector value based on the static resource information; Binding the compressed dynamic vector to the dynamic vector value to generate dynamic vector data; Binding the compressed static vector to the static vector value to generate static vector data; The dynamic vector data and the static vector data are stored in a preset storage system.

6. The method according to claim 4, characterized in that The step of repeatedly performing the step of processing the static picture based on the preset multimodal model on the base until the number of repeated executions reaches a preset repetition number threshold, further comprising: Acquire static related information of the static picture and dynamic related information of the dynamic picture; generating a manual review request based on the static related information and the dynamic related information; The manual review request is sent to the mobile terminal of the target work reviewer.

7. The method according to claim 6, characterized in that The step of repeatedly performing the step of processing the static picture based on the preset multimodal model on the base until the number of repeated executions reaches a preset repetition number threshold, further comprising: Get the first processing time and each repeated processing time; generating an execution record based on the first processing time, the repeated processing time, the dynamic related information and the static related information; The execution record is written into a preset feedback data table.

8. A website page monitoring device based on vectorization and multimodality, characterized in that: include: A dynamic image generation module is used to obtain the update information of the target website in real time, and to access and capture the dynamic images of the target website based on the update information to obtain the dynamic images; A static image generation module is used to obtain static processing information of a target website in real time, and to capture a static image of the target website based on the static processing information to obtain a static image; A picture vector generation module, used for performing steering amount processing on the dynamic picture and the static picture respectively to generate a dynamic picture vector and a static picture vector; A Euclidean distance calculation module, configured to calculate the Euclidean distance between the dynamic image and the static image based on the dynamic image vector and the static image vector; A preset condition judgment module, used for judging whether the static image meets the preset normal condition based on the Euclidean distance; A page normality determination module, used to determine whether the page of the target website corresponding to the static image is normal; A page anomaly determination module, used to determine if a page of a target website corresponding to the static image is abnormal; The abnormal page processing module is used to process the static image based on a preset multimodal model and generate a processing result.

9. An electronic device, characterized in that: comprising a processor coupled to a memory; The processor is configured to execute a computer program stored in the memory, so that the electronic device executes the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The method comprises a computer program or an instruction, which, when executed on a computer, causes the computer to execute the method according to any one of claims 1 to 7.

Images