Multimedia content identification system and method

By conducting preliminary and in-depth identification of multimedia content, generating risk detailed reports and specific names, the problem of malicious promotion content recognition is solved, effective positioning and processing of malicious content is achieved, and the purification effect of the content ecosystem is improved.

CN118939968APending Publication Date: 2024-11-12GUANGDONG ELECTRIC SHOCK MEDIA TECH CO LTD
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202411015013.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

It is difficult for the prior art to effectively identify and locate multimedia content with malicious promotion nature, affecting the purification of the content ecology.

Method used

By obtaining the multimedia content to be identified from the media resource library, preliminary identification is carried out to generate a risk detailed report, including risk type, time node and screenshot, deeply identifying it according to the risk type, identify the specific name of the risk, and store the results in the database. Specific steps include building a risk type resource database, extracting and comparing image grayscale values, segmenting pixel blocks, calculating aberration features, generating matching degrees, and training recognition models based on multi-angle images.

Benefits of technology

It realizes the positioning and identification of malicious promotion-oriented multimedia content, facilitates user query and processing, and improves the purification efficiency of the content ecosystem.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939968A_ABST
    Figure CN118939968A_ABST
Patent Text Reader

Abstract

The invention discloses a multimedia content identification system and method, and the method comprises the steps: obtaining a to-be-identified multimedia content from a media resource library, carrying out the preliminary identification of the to-be-identified multimedia content, obtaining a risk detail report, and enabling the risk detail report to comprise a risk type, a time node of a risk, and a screenshot. And performing deep identification according to the risk type, and identifying the specific name of the risk, thereby realizing an effect of positioning the multimedia content with the malicious popularization property, and meanwhile, storing the risk detail report and the specific name of the risk in the database, so that a user can conveniently inquire, and the user experience is improved. Therefore, the multimedia content with the malicious popularization property can be conveniently processed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of multimedia content recognition, and in particular to a multimedia content recognition system and method. Background Art

[0002] With the vigorous development of video applications, the video content ecosystem has undergone tremendous changes. More and more multimedia content (such as videos, pictures, texts or audios, etc.) uploaded by users are facing the public, while a large amount of multimedia content with malicious promotional nature has seriously endangered the development of the content ecosystem. In order to purify the multimedia content ecological environment, it is necessary to identify multimedia content. Through identification, multimedia content with malicious promotional nature can be effectively located. Summary of the invention

[0003] In view of this, the present invention proposes a multimedia content identification system and method, which can effectively locate multimedia content with malicious promotion properties.

[0004] The technical solution of the present invention is achieved in this way:

[0005] A multimedia content recognition method, specifically comprising:

[0006] Acquire multimedia content to be identified from a media resource library;

[0007] Performing preliminary identification on the multimedia content to be identified and obtaining a risk detail report, wherein the risk detail report includes the risk type, the time point where the risk occurs, and a screenshot;

[0008] Conduct in-depth identification based on the risk type and identify the specific name of the risk;

[0009] The risk detail report and the specific name of the risk are stored in the database, thereby realizing the identification of multimedia content.

[0010] As a further optional solution of the multimedia content identification method, the preliminary identification of the multimedia content to be identified and obtaining a risk detail report specifically includes:

[0011] Build a risk type resource database;

[0012] Compare and analyze the multimedia content to be identified based on the risk type resource database to obtain the risk type;

[0013] Extract time nodes based on risk types to obtain the time nodes and screenshots where the risks are located;

[0014] Generate a risk detail report based on the risk type, time point where the risk occurs, and screenshots.

[0015] As a further optional solution of the multimedia content identification method, the risk type resource database is used to compare and analyze the multimedia content to be identified to obtain the risk type, which specifically includes:

[0016] Extracting the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generating a sequence of risk type images and a sequence of multimedia content to be identified;

[0017] scaling the risk type image sequence and the to-be-identified multimedia content sequence;

[0018] Dividing the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks;

[0019] Calculating the block pixel means of all pixels in the N pixel blocks;

[0020] Calculate the difference between the grayscale value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, mark the pixel as 1, if the difference is less than 0, mark the pixel as 0;

[0021] Generating N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result;

[0022] The matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence is calculated, thereby obtaining the risk type.

[0023] As a further optional solution of the multimedia content recognition method, the in-depth recognition is performed according to the risk type to identify the specific name of the risk, which specifically includes:

[0024] Get multi-angle images of multiple risk types;

[0025] Setting a coordinate system; collecting first feature points of a plurality of multi-angle images, and determining a first angle of a first feature of a person based on the coordinate system and the first feature points; and collecting second feature points of a plurality of multi-angle images, and determining a second angle of a second feature of the person based on the coordinate system and the second feature points;

[0026] Performing training based on the training image, the first angle of the first feature, and the second angle of the second feature to generate a risk identification model;

[0027] Input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

[0028] A multimedia content recognition system, comprising:

[0029] A first acquisition module, used to acquire multimedia content to be identified from a media resource library;

[0030] A preliminary identification module, used to perform preliminary identification on the multimedia content to be identified and obtain a risk detail report, wherein the risk detail report includes the risk type, the time node where the risk occurs, and a screenshot;

[0031] A deep identification module is used to perform deep identification based on the risk type and identify the specific name of the risk;

[0032] The storage module is used to store the risk detail report and the specific name of the risk in the database, so as to realize the identification of multimedia content.

[0033] As a further optional solution of the multimedia content recognition system, the preliminary recognition module includes:

[0034] A construction module is used to construct a risk type resource database;

[0035] A comparison module, used for comparing and analyzing the multimedia content to be identified based on the risk type resource database to obtain the risk type;

[0036] The extraction module is used to extract time nodes based on risk types and obtain the time nodes and screenshots where the risks are located;

[0037] The first generation module is used to generate a risk detail report based on the risk type, the time node where the risk occurs, and the screenshot.

[0038] As a further optional solution of the multimedia content recognition system, the comparison module includes:

[0039] An extraction module, used to extract the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generate a sequence of risk type images and a sequence of multimedia content to be identified;

[0040] A scaling module, used for scaling the risk type image sequence and the to-be-identified multimedia content sequence;

[0041] A segmentation module, used for segmenting the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks;

[0042] A first calculation module, used for calculating the block pixel means of all pixels in the N pixel blocks;

[0043] A second calculation module is used to calculate the difference between the gray value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, the pixel is marked as 1, if the difference is less than 0, the pixel is marked as 0;

[0044] A second generating module, configured to generate N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result;

[0045] The third calculation module is used to calculate the matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence, so as to obtain the risk type.

[0046] As a further optional solution of the multimedia content recognition system, the depth recognition module includes:

[0047] A second acquisition module is used to acquire multi-angle images of multiple risk types;

[0048] A setting module is used to set a coordinate system; collect first feature points of multiple multi-angle images, and determine a first angle of a first feature of a person based on the coordinate system and the first feature points; and collect second feature points of multiple multi-angle images, and determine a second angle of a second feature of the person based on the coordinate system and the second feature points;

[0049] A training module, used for performing training based on a training image, a first angle of a first feature, and a second angle of a second feature to generate a risk identification model;

[0050] The input module is used to input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

[0051] The beneficial effect of the present invention is as follows: by acquiring multimedia content to be identified from a media resource library, performing preliminary identification on the multimedia content to be identified, and obtaining a risk detail report, wherein the risk detail report includes the risk type, the time node where the risk is located, and a screenshot, and performing in-depth identification based on the risk type to identify the specific name of the risk, thereby achieving the effect of locating multimedia content with malicious promotion properties. At the same time, by storing the risk detail report and the specific name of the risk in a database, it is convenient for users to query, thereby facilitating the processing of multimedia content with malicious promotion properties. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A schematic diagram of a process of a multimedia content recognition method of the present invention;

[0054] Figure 2 The figure is a schematic diagram of the composition of a multimedia content recognition system of the present invention. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present invention are described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] refer to Figure 1-2 , a multimedia content recognition method, specifically comprising:

[0057] Acquire multimedia content to be identified from a media resource library;

[0058] Performing preliminary identification on the multimedia content to be identified and obtaining a risk detail report, wherein the risk detail report includes the risk type, the time point where the risk occurs, and a screenshot;

[0059] Conduct in-depth identification based on the risk type and identify the specific name of the risk;

[0060] The risk detail report and the specific name of the risk are stored in the database, thereby realizing the identification of multimedia content.

[0061] In this embodiment, by obtaining the multimedia content to be identified from the media resource library, the multimedia content to be identified is preliminarily identified to obtain a risk detail report, wherein the risk detail report includes the risk type, the time node where the risk is located, and a screenshot, and in-depth identification is performed based on the risk type to identify the specific name of the risk, thereby achieving the effect of locating the multimedia content with malicious promotion properties. At the same time, by storing the risk detail report and the specific name of the risk in a database, it can be convenient for users to query, thereby facilitating the processing of multimedia content with malicious promotion properties.

[0062] It should be noted that the risk types include risky persons or risky items.

[0063] Preferably, the preliminary identification of the multimedia content to be identified and obtaining a risk details report specifically includes:

[0064] Build a risk type resource database;

[0065] Compare and analyze the multimedia content to be identified based on the risk type resource database to obtain the risk type;

[0066] Extract time nodes based on risk types to obtain the time nodes and screenshots where the risks are located;

[0067] Generate a risk detail report based on the risk type, time point where the risk occurs, and screenshots.

[0068] Preferably, the risk type resource database is used to compare and analyze the multimedia content to be identified to obtain the risk type, which specifically includes:

[0069] Extracting the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generating a sequence of risk type images and a sequence of multimedia content to be identified;

[0070] scaling the risk type image sequence and the to-be-identified multimedia content sequence;

[0071] Dividing the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks;

[0072] Calculating the block pixel means of all pixels in the N pixel blocks;

[0073] Calculate the difference between the grayscale value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, mark the pixel as 1, if the difference is less than 0, mark the pixel as 0;

[0074] Generating N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result;

[0075] The matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence is calculated, thereby obtaining the risk type.

[0076] Preferably, the in-depth identification is performed based on the risk type to identify the specific name of the risk, specifically including:

[0077] Get multi-angle images of multiple risk types;

[0078] Setting a coordinate system; collecting first feature points of a plurality of multi-angle images, and determining a first angle of a first feature of a person based on the coordinate system and the first feature points; and collecting second feature points of a plurality of multi-angle images, and determining a second angle of a second feature of the person based on the coordinate system and the second feature points;

[0079] Performing training based on the training image, the first angle of the first feature, and the second angle of the second feature to generate a risk identification model;

[0080] Input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

[0081] In this embodiment, by setting a coordinate system, collecting the first feature points of multiple multi-angle images, judging the first angle of the first feature of a person based on the coordinate system and the first feature points, and collecting the second feature points of multiple multi-angle images, judging the second angle of the second feature of the person based on the coordinate system and the second feature points, training is performed based on training images, the first angle of the first feature, and the second angle of the second feature to generate a risk identification model, which can improve the recognition accuracy of the risk identification model and further improve the effect of locating multimedia content with malicious promotional properties.

[0082] A multimedia content recognition system, comprising:

[0083] A first acquisition module, used to acquire multimedia content to be identified from a media resource library;

[0084] A preliminary identification module, used to perform preliminary identification on the multimedia content to be identified and obtain a risk detail report, wherein the risk detail report includes the risk type, the time node where the risk occurs, and a screenshot;

[0085] A deep identification module is used to perform deep identification based on the risk type and identify the specific name of the risk;

[0086] The storage module is used to store the risk detail report and the specific name of the risk in the database, so as to realize the identification of multimedia content.

[0087] Preferably, the preliminary identification module includes:

[0088] A construction module is used to construct a risk type resource database;

[0089] A comparison module, used for comparing and analyzing the multimedia content to be identified based on the risk type resource database to obtain the risk type;

[0090] The extraction module is used to extract time nodes based on risk types and obtain the time nodes and screenshots where the risks are located;

[0091] The first generation module is used to generate a risk detail report based on the risk type, the time node where the risk occurs, and the screenshot.

[0092] Preferably, the comparison module comprises:

[0093] An extraction module, used to extract the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generate a sequence of risk type images and a sequence of multimedia content to be identified;

[0094] A scaling module, used for scaling the risk type image sequence and the to-be-identified multimedia content sequence;

[0095] A segmentation module, used for segmenting the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks;

[0096] A first calculation module, used for calculating the block pixel means of all pixels in the N pixel blocks;

[0097] A second calculation module is used to calculate the difference between the gray value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, the pixel is marked as 1, if the difference is less than 0, the pixel is marked as 0;

[0098] A second generating module, configured to generate N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result;

[0099] The third calculation module is used to calculate the matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence, so as to obtain the risk type.

[0100] Preferably, the depth recognition module includes:

[0101] A second acquisition module is used to acquire multi-angle images of multiple risk types;

[0102] A setting module is used to set a coordinate system; collect first feature points of multiple multi-angle images, and determine a first angle of a first feature of a person based on the coordinate system and the first feature points; and collect second feature points of multiple multi-angle images, and determine a second angle of a second feature of the person based on the coordinate system and the second feature points;

[0103] A training module, used for performing training based on a training image, a first angle of a first feature, and a second angle of a second feature to generate a risk identification model;

[0104] The input module is used to input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

[0105] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A multimedia content recognition method, characterized in that: Specifically include: Acquire multimedia content to be identified from a media resource library; Performing preliminary identification on the multimedia content to be identified and obtaining a risk detail report, wherein the risk detail report includes the risk type, the time point where the risk occurs, and a screenshot; Conduct in-depth identification based on the risk type and identify the specific name of the risk; The risk detail report and the specific name of the risk are stored in the database, thereby realizing the identification of multimedia content.

2. A multimedia content recognition method according to claim 1, characterized in that: The preliminary identification of the multimedia content to be identified is performed to obtain a risk detail report, which specifically includes: Build a risk type resource database; Compare and analyze the multimedia content to be identified based on the risk type resource database to obtain the risk type; Extract time nodes based on risk types to obtain the time nodes and screenshots where the risks are located; Generate a risk detail report based on the risk type, time point where the risk occurs, and screenshots.

3. A multimedia content recognition method according to claim 2, characterized in that: The risk type resource database is used to compare and analyze the multimedia content to be identified to obtain the risk type, specifically including: Extracting the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generating a sequence of risk type images and a sequence of multimedia content to be identified; scaling the risk type image sequence and the to-be-identified multimedia content sequence; Dividing the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks; Calculating the block pixel means of all pixels in the N pixel blocks; Calculate the difference between the grayscale value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, mark the pixel as 1, if the difference is less than 0, mark the pixel as 0; Generating N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result; The matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence is calculated, thereby obtaining the risk type.

4. A multimedia content recognition method according to claim 3, characterized in that: The in-depth identification is performed based on the risk type to identify the specific name of the risk, including: Get multi-angle images of multiple risk types; Setting a coordinate system; collecting first feature points of a plurality of multi-angle images, and determining a first angle of a first feature of a person based on the coordinate system and the first feature points; and collecting second feature points of a plurality of multi-angle images, and determining a second angle of a second feature of the person based on the coordinate system and the second feature points; Performing training based on the training image, the first angle of the first feature, and the second angle of the second feature to generate a risk identification model; Input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

5. A multimedia content recognition system, characterized in that: include: A first acquisition module, used to acquire multimedia content to be identified from a media resource library; A preliminary identification module, used to perform preliminary identification on the multimedia content to be identified and obtain a risk detail report, wherein the risk detail report includes the risk type, the time node where the risk occurs, and a screenshot; A deep identification module is used to perform deep identification based on the risk type and identify the specific name of the risk; The storage module is used to store the risk detail report and the specific name of the risk in the database, so as to realize the identification of multimedia content.

6. A multimedia content recognition system according to claim 5, characterized in that: The preliminary identification module comprises: A construction module is used to construct a risk type resource database; A comparison module, used for comparing and analyzing the multimedia content to be identified based on the risk type resource database to obtain the risk type; The extraction module is used to extract time nodes based on risk types and obtain the time nodes and screenshots where the risks are located; The first generation module is used to generate a risk detail report based on the risk type, the time node where the risk occurs, and the screenshot.

7. A multimedia content recognition system according to claim 6, characterized in that: The comparison module comprises: An extraction module, used to extract the grayscale value of the risk type image and the grayscale value of the multimedia content to be identified in the risk type resource database, and generate a sequence of risk type images and a sequence of multimedia content to be identified; A scaling module, used for scaling the risk type image sequence and the to-be-identified multimedia content sequence; A segmentation module, used for segmenting the scaled risk type image sequence and the to-be-identified multimedia content sequence into N pixel blocks; A first calculation module, used for calculating the block pixel means of all pixels in the N pixel blocks; A second calculation module is used to calculate the difference between the gray value of each pixel and the mean value of the block pixels, if the difference is greater than or equal to 0, the pixel is marked as 1, if the difference is less than 0, the pixel is marked as 0; A second generating module, configured to generate N block aberration features of the risk type image sequence and N block aberration features of the to-be-identified multimedia content sequence according to the marking result; The third calculation module is used to calculate the matching degree between the N block disparity features of the risk type image sequence and the N block disparity features of the to-be-identified multimedia content sequence, so as to obtain the risk type.

8. A multimedia content recognition system according to claim 7, characterized in that: The depth recognition module comprises: A second acquisition module is used to acquire multi-angle images of multiple risk types; A setting module is used to set a coordinate system; collect first feature points of multiple multi-angle images, and determine a first angle of a first feature of a person based on the coordinate system and the first feature points; and collect second feature points of multiple multi-angle images, and determine a second angle of a second feature of the person based on the coordinate system and the second feature points; A training module, used for performing training based on a training image, a first angle of a first feature, and a second angle of a second feature to generate a risk identification model; The input module is used to input the risk type into the risk identification model for in-depth identification to obtain the specific name of the risk.

Citation Information

Patent Citations

  • Character recognition method and device and storage device

    CN109784240A

  • Risk user identification method and device, storage medium and server

    CN110223158A

  • Video content comparison method and device based on perception and aberration

    CN110619362A

  • Detection method and device for malicious websites, equipment and storage medium

    CN112100559A

  • Urban management violation event information acquisition method and system

    CN114627420A