Page anomaly detection method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of computers, in particular to a page anomaly detection method and device, electronic equipment and a storage medium. The method comprises the following steps: according to different types of user portraits of a platform, forming user accounts corresponding to the different types of user portraits based on the platform; obtaining a display page after each user account regularly logs in the platform, and extracting an area image with image information in the display page; the regional image is preprocessed, and a plurality of image items are separated from the preprocessed regional image; and performing anomaly detection on the image item to obtain an anomaly detection result. The method is used for detecting abnormal display in the page, improving the user experience and improving the quality of platform or application recommendation.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and more specifically, to a method, apparatus, electronic device, and storage medium for detecting page anomalies. Background Art

[0002] At present, video live streaming or graphic community applications are widely used by a large number of users. The recommendation systems designed within video live streaming or graphic community applications can generate different personalized recommendation lists for users according to their preferences, recommend different contents, and improve the application experience of users. However, in actual applications, the quality of the recommendation work of the current recommendation systems is unstable, and problems such as black screens, white screens, and repeated recommendations often occur in the pictures on the display pages, which will seriously affect the user experience. However, these problems are difficult to be completely discovered and solved through conventional code analysis and testing means during the application development stage. Existing recommendation anomaly detection technologies mainly focus on the analysis of video content accessed by users and data stream processing. For example, by combining edge computing and cloud computing, abnormal events in the recommendation list are analyzed. Most of these existing solutions focus on underlying technology implementations such as object recognition and behavior analysis, which are commonly used in scenarios such as monitoring, and usually do not target display problems such as black screens, white screens, UI (User Interface) anomalies, or repeated recommendations on the display interface. Therefore, it is necessary to make necessary improvements to the method for detecting page anomalies. Summary of the Invention

[0003] The present invention aims to overcome at least one defect (insufficiency) of the above-mentioned prior art, and provides a method, apparatus, electronic device, and storage medium for detecting page anomalies, which are used to detect abnormal displays on the page, improve the user experience, and enhance the quality of application recommendations.

[0004] According to a first aspect of the present application, a method for detecting page anomalies is provided. The method includes:

[0005] Based on user portraits of different categories on the platform, user accounts corresponding to the user portraits of different categories are formed based on the platform;

[0006] Obtain the display pages after each user account logs in to the platform at regular intervals, and extract the regional images with image information in the display pages;

[0007] Preprocess the regional images, and separate a number of image items from the preprocessed regional images;

[0008] Perform anomaly detection on the image items to obtain an anomaly detection result.

[0009] It is understandable that in the prior art, for the display page, especially the anomaly detection of the recommendation list in the page, specific codes are often detected and adjusted during the development stage. However, the back-end code can often only detect and adjust the anomalies of incorrect recommendations, while it is often difficult to directly identify the anomalies of the user interface (UI) in the page and / or whether the recommended content conforms to the user profile based on the back-end processor's code. Through this application, unique-preference accounts are formed according to different user profiles, and the user interface (UI) in the display page of the accounts is detected, so that the anomalies of the user-side page can be quickly detected, thereby obtaining the anomaly detection results, effectively improving the accuracy and diversity of the page content, and ultimately improving the user's usage experience and the quality of platform or application recommendations.

[0010] Optionally, forming the user accounts corresponding to different categories of the user profiles based on the platform includes:

[0011] Creating corresponding user accounts according to different categories of the user profiles of the platform;

[0012] Regularly logging in the user accounts on the platform and interacting with the content related to the corresponding user profiles on the user accounts to form the user accounts corresponding to different categories of the user profiles.

[0013] It is understandable that in order to form the user accounts corresponding to the user profiles, it is necessary to regularly log in the user accounts on the platform and interact with the relevant content according to the corresponding user profiles, including simulating ordinary users to browse and access the relevant content, etc., to enhance the nature of the user profiles of the user accounts, and also provide a real detection environment for subsequent detection, as much as possible restoring the anomalies that occur when ordinary users use, so as to detect and correct them, and enhance the user's experience on the platform.

[0014] Optionally, the display page is a recommendation page, and specifically extracting the regional image with image information in the display page is: extracting the regional image corresponding to the image recommendation item in the recommendation page.

[0015] It is understandable that the page of the platform is the page displayed to the user side, including different interaction buttons, recommended content, page aesthetic design pictures and other elements. Specifically, detecting the anomalies of the recommendation page in the page can centrally locate the anomalies of the recommendation page. The recommendation page is the most attractive part of the entire page. Solving the anomalies of the recommendation items in the page can maximize the user's interest in using the platform and maximize the user's experience on the platform.

[0016] Optionally, the anomaly detection of the image item to obtain the anomaly detection result specifically includes:

[0017] Obtain recommendation information according to each of the described image recommendation items;

[0018] Preset the relevance scoring criteria and abnormal score thresholds between the image recommendation items and the user portrait;

[0019] Obtain the scoring result of the user account according to all the recommendation information of the user account, the user portrait corresponding to the user account, and the relevance scoring criteria;

[0020] If the scoring result is lower than the abnormal score threshold, it is determined that all the image recommendation items are abnormal, and the abnormal detection result of the image recommendation items is obtained according to the abnormality.

[0021] It can be understood that personalized content recommendation for the user's page according to the user portrait of the user account can enable the user to see as much content as possible that they are interested in, increase the browsing time and usage time of the user on the platform, and improve the user experience on the platform; introducing a scoring standard to score the relevance of the recommendation information can quantify the recommendation information and accurately evaluate and define the quality of the recommendation information and its adaptation degree to the user portrait; and introducing an abnormal score threshold can accurately judge whether the recommendation information has a recommendation abnormality, enabling more efficient detection of the situation where the recommendation information does not match the user portrait. After obtaining the detection result, the corresponding algorithm is adjusted according to the detection result to ensure as high a user experience on the platform as possible.

[0022] Optionally, the abnormal detection of the image items to obtain the abnormal detection result specifically includes:

[0023] Extract the recommended cover image and / or recommended copywriting in the image recommendation items;

[0024] Perform abnormal detection according to the recommended cover image and / or recommended copywriting in the previously obtained image recommendation items to obtain the abnormal detection result.

[0025] It can be understood that the image recommendation items include the recommended cover image and the recommended copywriting, and both the recommended cover image and the recommended copywriting may be abnormal. It is necessary to extract both of them and perform abnormal detection on the extracted recommended cover image and recommended copywriting to avoid missing different types of abnormal situations.

[0026] Optionally, the performing abnormal detection according to the recommended cover image and / or recommended copywriting in the previously obtained image recommendation items to obtain the abnormal detection result specifically includes:

[0027] According to the extracted recommended cover image, obtain the first occupation ratio of black pixels in the recommended cover image for each of the image recommendation items;

[0028] According to the extracted recommended cover image, obtain the second proportion of white pixels in each of the image recommendation items in the recommended cover image;

[0029] When the first proportion or the second proportion exceeds a preset threshold, it is determined that the corresponding image recommendation item is abnormal, and an abnormal detection result of the image recommendation item is obtained according to the abnormality.

[0030] It can be understood that by obtaining the proportions of black pixels and white pixels in the recommended cover image in each image recommendation item, it is possible to identify whether there is a black screen or a white screen in the recommended cover image according to the proportions. The situation of abnormal cover display is judged through simple picture pixel recognition, reducing the traceability of a large amount of access confirmation information, improving the detection efficiency, saving processing computing resources, and making the entire detection process more intuitive and simple.

[0031] Optionally, the abnormal detection result is obtained by performing abnormal detection according to the recommended cover image and / or recommended copywriting in the image recommendation item obtained in advance, specifically including:

[0032] Judge the extracted recommended cover image. If the recommended cover image of the image recommendation item is a default background image preset for loading failure, it is determined that the corresponding image recommendation item is abnormal, and an abnormal detection result of the image recommendation item is obtained according to the abnormality; and / or,

[0033] According to the extracted recommended cover image, extract the text information in the recommended cover image;

[0034] Judge the text information in the recommended cover image and the extracted recommended copywriting. If the corresponding text information in the recommended cover image overlaps with the recommended copywriting, and / or the display format of the recommended cover image is incorrect, and / or the display format of the recommended copywriting is incorrect, and / or the recommended cover image has quality problems, it is determined that the corresponding image recommendation item is abnormal, and an abnormal detection result of the image recommendation item is obtained according to the abnormality;

[0035] and / or,

[0036] Compare the recommended cover images extracted from any two of the image recommendation items, and at the same time compare the corresponding extracted recommended copywriting; if the recommended cover images of the two image recommendation items are the same and the recommended copywriting is the same, it is determined that the corresponding two image recommendation items are abnormal, and an abnormal detection result of the image recommendation item is obtained according to the abnormality.

[0037] It is understandable that if the recommended cover image of an image recommendation item cannot be obtained from the background or there are transmission or network anomalies during the acquisition, the image recommendation item directly uses the default background image stored in the cache in the UI user interface. According to the intercepted image recommendation item, it is simply detected by image recognition that the recommended cover image is the default background image, thereby detecting that the recommended cover image of the image recommendation item fails to load. This can avoid the background from confirming the information flow for obtaining the default background image, making the entire detection method more intuitive and simple; during the acquisition or transmission of the recommended cover image, information loss may occur; when the recommended cover image and recommended copy are displayed in the user interface UI, display format errors, overlaps, etc. may also occur. Detecting the recommended cover image and recommended copy of each image recommendation item can visually identify abnormal recommended cover images and recommended copy, generate corresponding detection results and then make corrections, avoiding the tracing of a large amount of access data and improving the detection efficiency; it is understandable that when executing the recommendation algorithm, each image recommendation item is recommended to the corresponding account, and the phenomenon that the same image item may be recommended multiple times to the same account may occur. Repeated image recommendation items may reduce the user's browsing experience and browsing boredom, so it is necessary to detect and identify them; comparing each image recommendation item to detect the same image item can be detected through simple image comparison, avoiding monitoring each link of the execution of the recommendation algorithm, reducing the computing resources for detection, and improving the detection efficiency.

[0038] According to a second aspect of the present application, there is provided a page anomaly detection system, the system comprising:

[0039] An account cultivation module for forming user accounts corresponding to different categories of the user portraits based on the platform according to different categories of user portraits of the platform;

[0040] An extraction module for obtaining the display pages after each of the user accounts logs in to the platform at regular intervals, and extracting the regional images with image information in the display pages;

[0041] A preprocessing module for preprocessing the regional images and separating a plurality of image items from the preprocessed regional images;

[0042] An anomaly detection module for performing anomaly detection on the image items to obtain an anomaly detection result.

[0043] According to a third aspect of the present application, there is provided an electronic device, comprising:

[0044] A memory for storing one or more computer programs;

[0045] A processor, when the one or more computer programs are executed by the processor, implements a page exception detection method described in the first aspect above.

[0046] According to a fourth aspect of the present application, there is provided a computer-readable storage medium storing computer instructions for causing a processor to implement a page exception detection method described in the first aspect above when executed.

[0047] Based on any of the above aspects, a page exception detection method, device, electronic device, and storage medium provided by embodiments of the present application can form user accounts corresponding to different categories of user portraits based on the platform by according to different categories of user portraits of the platform; obtain the display pages after each user account logs in to the platform regularly, extract the regional images with image information in the display pages; preprocess the regional images, and separate several image items from the preprocessed regional images; perform anomaly detection on the image items to obtain an anomaly detection result. The method can effectively automatically detect anomalies in pages, especially in the recommended list page, from the perspective of users, such as black screens, white screens, repeated recommendations, UI display anomalies, and inconsistent recommended content with user interests. This detection method can simulate the actual usage scenarios of ordinary users, discover and fix these problems in a timely manner, thereby significantly improving the stability of the application recommendation system and the user experience, ultimately improving the accuracy and diversity of the recommended content, and optimizing the overall quality of the platform or application. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0049] Figure 1 It is a schematic diagram of an application scenario of a page exception detection method provided in this embodiment.

[0050] Figure 2 It is a flowchart of a page exception detection method provided in this embodiment.

[0051] Figure 3 It is a flowchart of a method for forming user accounts provided in this embodiment.

[0052] Figure 4 It is a flowchart of a first anomaly detection method provided in this embodiment.

[0053] Figure 5Flowchart of an anomaly detection method two provided in this embodiment.

[0054] Figure 6 Schematic diagram of a page anomaly detection system provided in this embodiment.

[0055] Figure 7 Schematic diagram of the structure of an electronic device provided in this embodiment. Detailed implementation manners

[0056] The accompanying drawings of this application are only for illustrative purposes and should not be construed as limitations to this application. To better illustrate the following embodiments, some components in the drawings may be omitted, enlarged or reduced, which do not represent the dimensions of actual products; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0057] In order to enable those skilled in the art to better understand the solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, rather than all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this application.

[0058] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above accompanying drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such used data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these process, method, product or device.

[0059] Recently, graphic and text community applications or platforms have widely emerged in the public eye and are used by a large number of users. Improving the overall user experience of the application or platform can increase the user base and enhance user stickiness. The recommendation work in the application or platform is one of the important factors affecting the user experience. Understandably, for the same application or platform, including various application forms such as the APP (Application) mobile terminal, PC (Personal Computer) terminal application, and WEB (World Wide Web) web page used by users, different users will affect their browsing behavior of the application or platform according to their unique preferences, including being inclined to click on the content they are interested in, thus forming different user portrait accounts, such as accounts interested in games, accounts interested in movies, etc. Recommending corresponding relevant content according to the unique preferences of users can increase the usage duration and interest of users in this application or platform, thereby improving the user experience. However, in the existing technology, the recommendation effects and quality for different users by the recommendation system often cannot be guaranteed. Usually, anomalies such as black screens, white screens, UI (User Interface) user interface anomalies, and repeated recommendations will appear in the recommendation list or recommendation items on the display page, seriously affecting the user experience. However, these anomalies are difficult to be completely discovered and solved through conventional code analysis and testing means during the development stage. The existing anomaly detection technologies mainly focus on video stream analysis and data stream processing. For example, by combining edge computing and cloud computing to analyze abnormal events in the video stream, the problem of UI (User Interface) user interface anomalies often cannot be accurately detected. Moreover, most of the existing solutions focus on the underlying technology implementation, such as object recognition and behavior analysis, which are commonly used in monitoring scenarios and usually do not target the display problems of the UI (User Interface). Therefore, it is necessary to make necessary improvements to the detection method of the page, especially the recommendation list.

[0060] This embodiment provides a technical solution that can solve the above problems. The following will combine the accompanying drawings to detail the specific implementation manners of the present application.

[0061] Exemplarily, it is a schematic diagram of an application scenario of a page anomaly detection method provided by an embodiment of the present application. As Figure 1 shown, the application scenario at least includes a server 100 and a terminal 200 that can communicate with the server 100. The server 100 has a background algorithm processing function and can also have functions of identifying, alarming, and correcting anomalies; the terminal device 200 has a function of displaying the display page, especially the recommendation list, and can also have a function of screenshot recording.

[0062] It is understandable that the server 100 may be an independent electronic device or a cluster composed of multiple electronic devices; the terminal 200 may be a smart phone terminal, a personal computer, a tablet computer, a vehicle-mounted terminal, etc., but is not limited thereto.

[0063] In an implementable manner, the server 100 and the terminal 200 can respectively execute a page anomaly detection method provided by an embodiment of the present application. Alternatively, optionally, a part of the page anomaly detection method provided by an embodiment of the present application is executed in the server 100, and a part is executed in the terminal 200.

[0064] As Figure 2 shown, this embodiment provides a page anomaly detection method, and the method can be subdivided into the following steps:

[0065] S100. According to different categories of user portraits on the platform, form user accounts corresponding to different categories of the user portraits based on the platform;

[0066] In this embodiment, the user portrait refers to constructing a dedicated digital file of a user by analyzing the interaction behaviors of the user on the platform, such as browsing data or access data, for personalized recommendation and behavior prediction of different users. By obtaining different user portraits, it is possible to accurately obtain the user groups with different preferences and tags on the platform during actual use, accurately capture the preferences of each type of user without omission, and achieve the effect of detecting the recommendation quality of all user portraits, thereby completing anomaly correction; the corrected recommendation algorithm can cover as many user portraits as possible, avoiding the situation of missing recommendation for a certain user portrait or a certain recommendation item not being recommended.

[0067] Specifically, as Figure 3 shown, the forming of user accounts corresponding to different categories of the user portraits based on the platform includes the following steps:

[0068] S110. Create corresponding user accounts according to different categories of the user portraits on the platform;

[0069] S120. Regularly log in to the platform with the user accounts and interact with the content related to the corresponding user portraits on the user accounts to form user accounts corresponding to different categories of the user portraits.

[0070] In this embodiment, corresponding accounts are formed according to different categories of user portraits, which can simulate the interactive behaviors of ordinary users such as access and browsing. Therefore, as an application, it can detect materials during the actual operation stage, achieving the effect of logging in to the user accounts regularly and detecting them without affecting the use of ordinary users and protecting the privacy of ordinary users. It can be understood that during the formation stage of the user accounts, it is necessary to log in to the user accounts on the platform regularly and interact with the content related to the corresponding user portraits on the user accounts, including simulating the interactive behaviors of ordinary users such as browsing and access, and this interactive behavior is guided by the corresponding user portraits, so as to continuously strengthen and reinforce the corresponding user portraits.

[0071] S200. Obtain the display pages after each of the user accounts logs in to the platform at regular intervals, and extract the regional images with image information in the display pages.

[0072] In this embodiment, when extracting the regional images from the display pages, specifically, an image recognition algorithm is used to recognize the part with image information in the display page, and then a scrolling long screenshot of the recognized regional image is obtained manually using a screenshot tool or by using code to control the machine. Among them, the regional image with image information refers to the part of the display page with multiple picture elements, and also includes the text elements introducing the picture elements. The picture elements and the text elements together constitute the regional image with image information.

[0073] In this embodiment, the cultivated user accounts can restore and simulate the real usage situations of ordinary users as much as possible, and can truly reflect the real situations of ordinary users, so that the detection work can discover the abnormalities in the use of ordinary users. Obtaining the display pages after each of the user accounts logs in to the platform at regular intervals can obtain the display conditions of the display pages; extracting the regional images with image information in the display pages can narrow the target area of detection and improve the detection efficiency.

[0074] Preferably, the regional images with image information in the display pages are usually also associated with the corresponding user accounts and user images of the regional images. It is necessary to store the regional pictures, the corresponding user accounts, and the corresponding user images in the file server to archive the corresponding regional images for convenient subsequent tracing.

[0075] S300. Preprocess the regional images, and separate several image items from the preprocessed regional images.

[0076] In this embodiment, preprocessing the regional image includes using conventional picture processing steps such as geometric transformation, filtering, etc., which will not be elaborated here; several image items are separated from the preprocessed regional image, and each image item needs to be recognized from the regional image, where the image item includes a single picture element and a text element; for the recommendation page, the obtained regional image is a recommendation list, and the recommendation list includes multiple image recommendation items, which mainly have a single recommended cover picture and recommended text. It is necessary to use an algorithm to recognize the image recommendation items formed by the single recommended cover picture and recommended text and separate them one by one in the regional image to obtain a single image recommendation item; detecting the image recommendation item with a single recommended cover picture and recommended text can reduce the detected picture area, concentrate the target area of detection, and improve the detection efficiency; detecting this image recommendation item can directly obtain the display anomalies in the regional image;

[0077] S400. Perform anomaly detection on the image item to obtain an anomaly detection result.

[0078] Preferably, the anomaly detection is performed based on a large language model and traditional image algorithms as the underlying logic;

[0079] It can be understood that a large language model (LLM) is a natural language processing model based on deep learning technology. It is a deep learning model trained with a large amount of text data, aiming to understand and generate human language. It can learn language patterns from massive data, capture complex language relationships and structures, and then perform various language-related tasks. In this embodiment, the large language model can be used to recognize and understand the text in the image item, providing the effect of text recognition for the detection work;

[0080] Traditional image algorithms are developed based on computer vision and image processing technologies, and are a series of methods for performing feature extraction, filtering, compression, etc. on images, including image enhancement algorithms, image classification algorithms, etc.; in this embodiment, traditional image algorithms can be used to detect the color, pixels, format, etc. of the image item, so as to obtain the anomaly detection result of the image item.

[0081] In this embodiment, the anomaly detection result is obtained according to the anomalies that occur in the image item, and then the corresponding recommendation algorithm and / or UI user interface display code are corrected according to the anomaly detection result, so as to be able to repair the user interface UI, effectively improve the accuracy and diversity of the display page, and ultimately improve the user's usage experience and the quality of platform or application recommendations.

[0082] Specifically, the display page is a recommendation page, and extracting the regional image with image information in the display page specifically means: extracting the regional image corresponding to the image recommendation item in the recommendation page.

[0083] In this embodiment, the display page of the platform usually includes interactive buttons, a recommendation list, page aesthetic design pictures, etc. The above jointly constitute the final page displayed by the platform to users; whether the display page is user-friendly can affect the user experience of the platform; usually, the recommendation page in the display page is often the most eye-catching to users because the recommendation page always recommends relevant content according to the user portrait and the user's preferences. Therefore, performing anomaly detection on the recommendation page can maximize the user experience in the platform and improve the processing value of anomaly detection.

[0084] Specifically, as Figure 4 described, performing anomaly detection on the image item to obtain an anomaly detection result, and Method Flow 1 specifically includes the following steps:

[0085] S411. Obtain recommendation information according to each of the image recommendation items;

[0086] S412. Preset the correlation scoring criteria between the image recommendation item and the user portrait and the anomaly score threshold;

[0087] In this embodiment, in the recommendation mechanism, content is often personalized for users to recommend browsing according to the user portrait. However, a recommendation list will contain multiple image recommendation items. In the case where the content of each image recommendation item is as non-repetitive as possible, there may be individual image recommendation items that are not relevant to the user portrait due to incorrect recommendations or system judgment errors. Count the correlation degree between each image recommendation item and the user portrait, and obtain the correlation degree between all image recommendation items and the user portrait; assign different scores to different correlation degrees for subsequent scoring; and set an anomaly score threshold. All image recommendation items below this anomaly score threshold prove that the image recommendation items in the recommendation list have a low correlation with the corresponding user portrait and are determined to be recommendation anomalies.

[0088] S413. Obtain the scoring result of the user account according to all the recommendation information of the user account, the user portrait corresponding to the user account, and the correlation scoring criteria;

[0089] In this embodiment, it is necessary to score the recommendation information according to the correlation scoring criteria between the image recommendation item and the user portrait, obtain whether the recommendation information conforms to the corresponding user portrait, and judge whether it is a recommendation work that fits the corresponding user image;

[0090] S414. If the scoring result is lower than the abnormal score threshold, it is determined that all the image recommendation items are abnormal, and the abnormal detection result of the image recommendation items is obtained according to the abnormality.

[0091] In this embodiment, the scoring result is lower than the abnormal score threshold, which proves that the relevance between the recommended content and the user portrait is not high, resulting in low interest of the user when browsing the image recommendation item, and the user has no desire to continue browsing, reducing the user experience. Therefore, it is determined that all the image items are abnormal.

[0092] Specifically, the abnormal detection of the image items to obtain the abnormal detection result specifically includes:

[0093] Extract the recommended cover image and / or recommended copywriting in the image recommendation item;

[0094] Perform abnormal detection on the extracted recommended cover image and / or the extracted recommended copywriting in the image recommendation item to obtain the abnormal detection result.

[0095] It can be understood that the image recommendation item includes a recommended cover image and recommended copywriting, and both the recommended cover image and the recommended copywriting may be abnormal. It is necessary to extract both and perform abnormal detection on the extracted recommended cover image and recommended copywriting to avoid missing different types of abnormal situations.

[0096] Specifically, as Figure 5 shown, the method for performing abnormal detection on the extracted recommended cover image and / or recommended copywriting in the image recommendation item to obtain the abnormal detection result, and the second method flow specifically includes the following steps:

[0097] S421. According to the extracted recommended cover image, obtain the first occupation ratio of black pixels in the recommended cover image for each image recommendation item;

[0098] S422. According to the extracted recommended cover image, obtain the second occupation ratio of white pixels in the recommended cover image for each image recommendation item;

[0099] S423. When the first occupation ratio or the second occupation ratio exceeds the preset threshold, it is determined that the corresponding image recommendation item is abnormal, and the abnormal detection result of the image recommendation item is obtained according to the abnormality.

[0100] In this embodiment, the proportions of black pixels and white pixels in each image recommendation item in the entire recommended cover image frame can be obtained respectively. When the first proportion value is greater than a preset threshold, it can be determined that the recommended cover image corresponding to the image recommendation item is in a black screen state; when the second proportion value is greater than the preset threshold, it can be determined that the recommended cover image corresponding to the image recommendation item is in a white screen state; and both the black screen state and the white screen state of the recommended cover image indicate that the front end obtains an abnormal recommended cover image from the back end, resulting in a black screen and a white screen. Neither the black screen nor the white screen situation can enable the user to obtain the visual information of the image recommendation item, and the recommended information obtained only from the text in the image recommendation item cannot fully arouse the user's maximum browsing interest, thus reducing the user experience. Therefore, it is necessary to detect the black screen and white screen situations to obtain an abnormal detection result;

[0101] Preferably, the preset threshold is 90%;

[0102] In this embodiment, the first proportion value and the second proportion value need to be calculated for each cover image in each image recommendation item to facilitate obtaining the situation of each recommended cover image.

[0103] Specifically, the abnormal detection to obtain the abnormal detection result according to the recommended cover image and / or recommended text in the pre-obtained image recommendation item specifically includes:

[0104] Judge the extracted recommended cover image. If the recommended cover image of the image recommendation item is the default bottom image preset for failed loading, it is determined that the corresponding image recommendation item is abnormal, and the abnormal detection result of the image recommendation item is obtained according to the abnormality;

[0105] In this embodiment, the front-end processor will preset a default bottom image. If the front-end processor cannot obtain a normal recommended cover image due to other transmission problems such as network, the front-end processor will consider that the recommended cover image fails to load and display the default bottom image as the recommended cover image of the image recommendation item; the default bottom image is usually a white-bottom image with the words "loading failed", which will also reduce the user's browsing experience, so it needs to be detected.

[0106] Specifically, the method of abnormal detection includes:

[0107] According to the extracted recommended cover image, extract the text information in the recommended cover image;

[0108] Judge the text information in the recommended cover image and the extracted recommended copywriting. If the corresponding text information in the recommended cover image overlaps with the recommended copywriting, and / or the display format of the recommended cover image is incorrect, and / or the display format of the recommended copywriting is incorrect, and / or the recommended cover image has quality problems, it is determined that the corresponding image recommendation item is abnormal, and the abnormal detection result of the image recommendation item is obtained according to the abnormality;

[0109] In this embodiment, there is an overlap in the text between the text information in the recommended cover image of the image recommendation item and the recommended copywriting, which causes users to be unable to clearly obtain the recommended information. And / or, the display format of the recommended cover image of the image recommendation item is incorrect, and / or the display format of the recommended copywriting is incorrect, including that the recommended cover image and / or the recommended copywriting are not aligned on the display panel, exceed the boundary of the display panel, etc. And / or, the recommended cover image has quality problems, including abnormal flower blocks, abnormal screen flickering, overexposure, low picture contrast and other quality problems in the recommended cover image, resulting in unclear display of the recommended cover image. The above situations will cause users to be unable to accurately obtain the recommended information of the image recommendation item, thus reducing the user experience. Therefore, it is necessary to detect and process it.

[0110] Specifically, the method for abnormal detection includes:

[0111] Compare the recommended cover images extracted from any two of the image recommendation items, and at the same time compare the corresponding recommended copywriting; if the recommended cover images of the two image recommendation items are the same and the recommended copywriting are both the same, it is determined that the corresponding two image recommendation items are abnormal, and the abnormal detection result of the image recommendation item is obtained according to the abnormality.

[0112] In this embodiment, there will also be a situation where the content of two or more image recommendation items in the recommended list is exactly the same, including the same recommended cover images and the same recommended copywriting of two or more image recommendation items. The above situation will lead to the problem that the content of one image recommendation item is recommended on multiple image recommendation item layouts at the same time. This will reduce the recommended content obtained by users during browsing, and will also cause user browsing fatigue and reduce the user experience. Therefore, it is detected and processed.

[0113] As Figure 6 shown, the embodiment of the present application also provides a page abnormal detection system. Optionally, the system includes:

[0114] A number-raising module 511, an extraction module 512, a preprocessing module 513, and an abnormal detection module 514, where:

[0115] The account nurturing module 511 is used to form user accounts corresponding to different categories of user portraits on the platform according to different categories of user portraits on the platform;

[0116] In this embodiment, the account nurturing module 511 can be used to execute Figure 2 the steps S100 shown. The specific description of the account nurturing module 511 can refer to the description of the steps S100.

[0117] The extraction module 512 is used to obtain the display pages after each user account logs in to the platform at regular intervals, and extract the regional images with image information in the display pages;

[0118] In this embodiment, the extraction module 512 can be used to execute Figure 2 the steps S200 shown. The specific description of the extraction module 512 can refer to the description of the steps S200.

[0119] The preprocessing module 513 is used to preprocess the regional images and separate several image items from the preprocessed regional images;

[0120] In this embodiment, the preprocessing module 513 can be used to execute Figure 2 the steps S300 shown. The specific description of the preprocessing module 513 can refer to the description of the steps S300.

[0121] The anomaly detection module 514 is used to perform anomaly detection on the image items to obtain an anomaly detection result.

[0122] In this embodiment, the anomaly detection module 514 can be used to execute Figure 2 the steps 400 shown. The specific description of the anomaly detection module 514 can refer to the description of the steps S400.

[0123] This application embodiment also provides an electronic device, whose structure is as Figure 7 shown. The electronic device includes a memory 611, a processor 612, a communication module 613, an input / output interface 614, etc. Optionally, the memory 611, the processor 612, the communication module 613, and the input / output interface 614 can be connected and communicate through a bus 615.

[0124] The memory 611 is used to store one or more computer programs and transmit the codes of the computer programs to the processor 612; when the one or more computer programs are executed by the processor 612, a page anomaly detection method in this application embodiment is implemented.

[0125] Optionally, the electronic device can be connected to a network through the communication module 613 to communicate with other devices, such as terminals or servers, via the network to achieve data interaction. The electronic device can be various forms of digital computers, such as, by way of example, desktop computers, servers, workstations, mainframe computers or other types of computers. The electronic device can also be various forms of mobile terminals, such as, by way of example, smart phones, tablet computers, wearable devices (such as helmets, glasses, watches, etc.) and other similar mobile terminals.

[0126] Optionally, the electronic device can be connected to the required input / output devices, such as keyboards, display devices, etc., through the input / output interface 614. The electronic device itself can have a display device, and can also externally connect other display devices through the input / output interface 614. Optionally, a storage device, such as a hard disk, etc., can also be connected through the input / output interface 614, so that the data in the electronic device can be stored in the storage device, or the data in the storage device can be read, and the data in the storage device can also be stored in the memory 611. It can be understood that the input / output interface 614 can be a wired interface or a wireless interface. According to different actual application scenarios, the devices connected to the input / output interface 614 can be components of the electronic device or external devices connected to the electronic device when needed.

[0127] Optionally, the memory 611 can be a volatile memory and / or a non-volatile memory. The volatile memory can be a random access memory, etc., and the non-volatile memory can be a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory or a flash memory, etc.

[0128] Optionally, the computer program stored in the memory 611 can be divided into one or more modules. The one or more modules are stored in the memory 611 and executed by the processor 612 to complete the method provided by its own embodiment. The one or more modules can be a series of computer program instruction segments capable of completing specific functions, and the computer program instruction segments are used to describe the execution process of the computer program in the electronic device.

[0129] Optionally, the processor 612 may be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 612 include, but are not limited to, a central processing unit, a graphics processing unit, a digital signal processor, various dedicated artificial intelligence computing chips, various processors running machine learning model algorithms, and may also be any suitable controller, microcontroller, processor, etc. The processor 612 executes the various methods and processes of this embodiment. Exemplarily, such as a page exception detection method according to an embodiment of the present application.

[0130] Optionally, the bus 615 may include a path for transmitting information. The bus 615 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. According to different functions, the bus 615 may be divided into an address bus, a data bus, a control bus, etc.

[0131] In an alternative implementation, an embodiment of the present application further provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a computer, the computer can execute the methods of the above method embodiments. Part or all of the computer program may be loaded and / or installed on the memory 611 of the electronic device. When the computer program is executed by the processor 612, one or more steps of a page exception detection method according to an embodiment of the present application can be executed.

[0132] Optionally, the computer-readable storage medium may be a random access memory, a read-only memory, a programmable read-only memory, an erasable programmable read-only memory, an electrically erasable programmable read-only memory, etc.

[0133] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the technical solutions of the present invention, rather than limitations on the specific implementation manners of the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the claims of the present invention shall be included within the protection scope of the claims of the present invention.

Claims

1. A page anomaly detection method, characterized in that: The method comprises: According to different categories of user portraits on the platform, user accounts corresponding to the different categories of user portraits are formed based on the platform; Obtaining a display page after each user account logs into the platform regularly, and extracting a region image with image information in the display page; Preprocessing the regional image and separating a plurality of image items from the preprocessed regional image; Anomaly detection is performed on the image item to obtain an anomaly detection result.

2. A page anomaly detection method according to claim 1, characterized in that: The user accounts corresponding to the user portraits of different categories formed based on the platform include: Create corresponding user accounts according to the user portraits of different categories on the platform; The user account is logged in to the platform regularly, and interacts with the corresponding user portrait-related content on the user account to form user accounts corresponding to the user portraits of different categories.

3. A page anomaly detection method according to any one of claims 1-2, characterized in that: The display page is a recommendation page, and extracting the regional image with image information in the display page specifically includes: extracting the regional image corresponding to the image recommendation item in the recommendation page.

4. A page anomaly detection method according to claim 3, characterized in that: The performing anomaly detection on the image item to obtain an anomaly detection result specifically includes: Acquire recommendation information according to each of the image recommendation items; Preset the relevance scoring criteria and anomaly score threshold for image recommendation items and user portraits; Obtaining a scoring result of the user account according to all the recommendation information of the user account, the user portrait corresponding to the user account, and a relevance scoring standard; If the scoring result is lower than the abnormal score threshold, it is determined that all the image recommendation items are abnormal, and an abnormality detection result of the image recommendation item is obtained based on the abnormality.

5. A page anomaly detection method according to claim 3, characterized in that: The performing anomaly detection on the image item to obtain an anomaly detection result specifically includes: Extracting a recommended cover image and / or a recommended text from the image recommendation item; Anomaly detection is performed based on the recommended cover image extracted from the image recommendation item and / or the recommended text extracted to obtain anomaly detection results.

6. A page anomaly detection method according to claim 5, characterized in that: The performing anomaly detection according to the extracted recommended cover image and / or recommended text in the image recommendation item to obtain an anomaly detection result specifically includes: According to the extracted recommended cover image, obtaining a first proportion value of black pixels in the recommended cover image in each of the image recommendation items; According to the extracted recommended cover image, obtaining a second proportion value of white pixels in the recommended cover image in each of the image recommendation items; When the first proportion value or the second proportion value exceeds a preset threshold, it is determined that the corresponding image recommendation item is abnormal, and an abnormality detection result of the image recommendation item is obtained based on the abnormality.

7. A page anomaly detection method according to claim 5, characterized in that: The performing anomaly detection according to the extracted recommended cover image and / or recommended text in the image recommendation item to obtain an anomaly detection result specifically includes: The extracted recommended cover image is judged, and if the recommended cover image of the image recommendation item is a preset default background image that fails to load, it is determined that an abnormality occurs in the corresponding image recommendation item, and an abnormality detection result of the image recommendation item is obtained according to the abnormality; and / or, Extracting text information from the recommended cover image according to the extracted recommended cover image; The text information in the recommended cover image and the extracted recommended copy are judged. If the corresponding text information in the recommended cover image overlaps with the recommended copy, and / or the display format of the recommended cover image is incorrect, and / or the display format of the recommended copy is incorrect, and / or the recommended cover image has a quality problem, it is determined that the corresponding image recommendation item is abnormal, and an abnormality detection result of the image recommendation item is obtained according to the abnormality; and / or, The recommended cover images extracted from any two of the image recommendation items are compared, and the corresponding extracted recommended texts are compared at the same time; if the recommended cover images of the two image recommendation items are the same and the recommended texts are the same, it is determined that the corresponding two image recommendation items have an anomaly, and the anomaly detection result of the image recommendation item is obtained based on the anomaly.

8. A page anomaly detection system, characterized in that: The system comprises: An account maintenance module, used to generate user accounts corresponding to different categories of user portraits based on the platform according to different categories of user portraits of the platform; An extraction module, used to obtain the display page after each user account logs into the platform regularly, and extract the area image with image information in the display page; A preprocessing module, used for preprocessing the regional image and separating a plurality of image items from the preprocessed regional image; The anomaly detection module is used to perform anomaly detection on the image item to obtain an anomaly detection result.

9. An electronic device, characterized in that: include: a memory for storing one or more computer programs; A processor, when the one or more computer programs are executed by the processor, implements a page anomaly detection method as described in any one of claims 1-7.

10. A computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable a processor to implement a page exception detection method according to any one of claims 1 to 7 when executed.

Images