Search

CN-122024451-A - Intelligent security broadcast monitoring and early warning method, device and system based on AI pre-generated data model comparison

CN122024451ACN 122024451 ACN122024451 ACN 122024451ACN-122024451-A

Abstract

The invention discloses a security broadcasting intelligent monitoring and early warning method, device and system based on AI pre-generated data model comparison, comprising the steps of establishing a reference library containing expected data models; the method comprises the steps of collecting broadcast signals in real time to generate a real-time broadcast data model with the same dimension, comparing the two models in layers of data layers and content layers according to time stamps, judging abnormality and triggering an alarm when the comparison result exceeds a preset threshold value. The invention realizes the upgrade from signal monitoring to intelligent content understanding monitoring, the real-time fault discovery and the accurate monitoring judgment, greatly reduces the false alarm rate and the manual pressure, realizes the intelligent guarantee of the whole broadcasting process closed loop, and is suitable for the broadcasting signal monitoring of broadcasting and new media.

Inventors

  • WANG DONG
  • WANG SIYUAN

Assignees

  • 北京格非信息技术有限公司

Dates

Publication Date
20260512
Application Date
20260325

Claims (10)

  1. 1. The intelligent monitoring and early warning method for the safe broadcasting based on the comparison of the AI pre-generated data models is characterized by comprising the following steps: Step 1, monotonically extracting corresponding broadcasting material library files according to broadcasting arrangement series, extracting video key frame characteristic values and audio characteristic values of the material files, and generating expected data models of all broadcasting periods by combining time stamps of the series list and expected program identifiers; step 2, in the program broadcasting process, video and audio signals are collected and broadcast in real time, video characteristic values and audio characteristic values which are in the same dimension with the expected data model are extracted, and a real-time broadcasting data model is generated by combining broadcasting real-time stamps; Step 3, carrying out layered comparison on the real-time broadcasting data model and an expected data model of a corresponding period according to a time stamp, wherein the layered comparison comprises data layer comparison and content layer comparison; and step 4, abnormality judgment and alarm triggering, namely judging whether the comparison result exceeds a preset fault tolerance threshold, if so, judging that the broadcast is abnormal, triggering an alarm instruction, and executing operations of abnormal information pushing and abnormal fragment interception.
  2. 2. The method of claim 1, wherein in step 1, the series list contains broadcast material ID, material edit point, material broadcast time, channel name information, and the expected program identification includes station logo, advertisement tag information.
  3. 3. The method according to claim 1, wherein, in step 1 or step 2, The video characteristic value is a binary character string calculated based on image content, and the extraction process comprises the steps of reducing an image to a fixed size, converting the image into a gray level image, performing discrete cosine transform, reserving an 8X 8 matrix at the left upper corner of a transform result, calculating the average value of matrix coefficients, comparing the coefficients with the average value, binarizing the coefficients, and sequentially combining the coefficients to obtain the video characteristic value; The audio characteristic value is a hash value generated based on audio acoustic characteristics, and the extraction process comprises the steps of performing short-time Fourier transform on an audio fragment to generate a spectrogram, extracting energy peak characteristic points on the spectrogram, and generating the audio characteristic value through a hash function according to time and frequency information of the characteristic points.
  4. 4. The method according to claim 1, wherein in the step 3, the data layer comparison is a comparison code rate, IP stream information, and technical index of data in a vertical blanking period, and the content layer comparison includes comparing picture similarity, comparing audio feature values, and comparing consistency of the identified caption text and the expected text by using a deep hash algorithm or a convolutional neural network.
  5. 5. The method according to claim 1, wherein, in step 4, The preset fault tolerance threshold comprises one or more of picture similarity lower than 90%, audio silence exceeding 2 seconds, and discrepancy between the actual advertisement time interval and the pre-model advertisement time interval; the alarm instruction comprises audible and visual alarm, and the abnormal information pushing is used for pushing the abnormal data analysis result to the alarm terminal and the SDN centralized control system.
  6. 6. An AI-pre-generated data model comparison-based safe broadcasting intelligent monitoring and early warning device for realizing the method as set forth in any one of claims 1 to 5, comprising: The broadcasting control data interface module is used for interfacing a broadcasting control system and acquiring broadcasting serial single information and corresponding broadcasting files; The AI expected model generation service module is used for pre-generating video and audio characteristic values of materials of each broadcasting period according to the broadcasting serial connection list to form an expected data model and an abnormal broadcasting data model; The AI data model generation service module is used for collecting broadcast video and audio signals in real time, generating video and audio characteristic values of materials in each broadcast period, and forming a real-time broadcast data model; The comparison engine and the alarm service module are used for carrying out layered comparison on the real-time broadcasting data model and the expected data model in the corresponding period, judging whether the comparison result exceeds a preset fault tolerance threshold, judging that broadcasting is abnormal and triggering an alarm instruction if the comparison result exceeds the preset fault tolerance threshold, and recording a monitoring log; And the SDN interface module is used for interfacing with the SDN centralized control system and reporting alarm information and corresponding abnormal data analysis results.
  7. 7. The apparatus of claim 6, wherein a NVIDIA Jetson T-5000 series chip is used as a core processing chip of the apparatus, and is based on a Blackwell GPU architecture to support running of a large model of the primary flow of ilama 3, qwen, stable Diffusion.
  8. 8. The device of claim 7, wherein the comparison engine and the alarm service module comprise a data layer comparison unit, a content layer comparison unit and an abnormality determination unit, the content layer comparison unit is provided with a deep hash algorithm and operation logic of a convolutional neural network, the comparison engine and the alarm service module are connected with an alarm terminal, and the alarm terminal supports the functions of audible and visual alarm, abnormal fragment storage and abnormal information display.
  9. 9. The intelligent monitoring and early warning system for safety broadcasting based on AI pre-generated data model comparison is characterized by comprising the device of any one of claims 6-8, a broadcasting control system, an SDN centralized control system and a broadcasting material library which are in communication connection with the device; the broadcasting control system is used for issuing a broadcasting serial bill and a broadcasting control instruction to the intelligent monitoring and early warning device; the broadcasting material library is used for storing video and audio material files of programs, advertisements and gaskets for being called by the intelligent monitoring and early warning device; The SDN centralized control system is used for receiving alarm information and abnormal data analysis results reported by the intelligent monitoring and early warning device and realizing centralized control of the broadcasting system.
  10. 10. The system of claim 9, wherein the intelligent monitoring and early warning device can adopt a single-device or multi-device deployment mode according to the channel scale of the broadcasting system, and the single-device deployment integrates file monitoring and stream monitoring functions, and the multi-device deployment realizes the generation of the expected data model and the generation of the real-time broadcasting data model respectively.

Description

Intelligent security broadcast monitoring and early warning method, device and system based on AI pre-generated data model comparison Technical Field The invention relates to the technical field of broadcast television safety broadcasting, in particular to a safety broadcasting intelligent monitoring and early warning method, device and system based on AI pre-generated data model comparison. Background The safe broadcasting of the broadcast television is the core work of the broadcast television industry, and the signal monitoring is the key means for guaranteeing the safe broadcasting. The traditional signal monitoring method in the prior art mainly depends on threshold judgment, such as monitoring physical signal indexes of black fields, static frames, silence and the like, has the technical defects of high false alarm rate and serious missing report, for example, a wind scraping leaf shielding lens can be misjudged as the black field, the real signal static frames can be missing report due to static pictures, meanwhile, the traditional monitoring method can only realize the monitoring of a shallow physical layer and a transmission layer, and cannot understand the fact that the played content is not aligned, and the content identification and logic judgment capability are not provided. In the prior art, a monitoring mode of multipath signal comparison, such as main and standby path signal comparison, is also available, but the mode can only detect faults in the signal transmission process, and can not solve the difference problem between broadcast content and pre-arranged program content, namely, the error broadcast abnormality of 'the content A to be broadcast and the content B to be broadcast actually' can not be actively found, the relevance between the monitoring result and the expected broadcast content is lost, and the subjective judgment of an operator is excessively relied on. In addition, in the prior art, an alarm optimizing system and an emergency processing system which are designed aiming at safe broadcasting are adopted, the system is a fault post-processing scheme, abnormality can be detected and subsequent processing can be carried out only after faults such as signal interruption, static frames, black fields and the like occur, at the moment, the faults actually occur and influence spectators, the problem of monitoring lag can not be fundamentally solved, and the early pre-judging and immediate processing of broadcasting faults are difficult to realize. In summary, the existing broadcast television safety broadcasting monitoring technology has the technical problems of monitoring lag, content identification capability loss, high false alarm rate and excessive dependence on manual judgment, and a monitoring and early warning scheme capable of realizing intelligent understanding, real-time pre-judgment and accurate alarm of the content is needed. Disclosure of Invention Aiming at the defects existing in the prior art, the invention provides a safe broadcasting intelligent monitoring and early warning method, device and system based on AI pre-generated data model comparison, which solve the problems of monitoring lag and logic error of identification content in the prior art and realize the technical jump from 'signal monitoring' to 'intelligent understanding and monitoring' of the content. The invention discloses a security broadcasting intelligent monitoring and early warning method based on AI pre-generated data model comparison, which comprises the following steps: Step 1, an AI pre-generation reference data model base is established, namely, corresponding broadcasting material base files are singly fetched according to broadcasting arrangement series, video key frame characteristic values and audio characteristic values of the material files are extracted, and expected data models of broadcasting periods are generated by combining time stamps of serial lists and expected program identifiers, and abnormal broadcasting data models of static frames, black fields and gaskets are generated; step 2, real-time signal acquisition and characteristic analysis, namely acquiring broadcast video and audio signals in real time in the program broadcasting process, extracting video characteristic values and audio characteristic values which are in the same dimension as the expected data model, and generating a real-time broadcasting data model by combining broadcast real-time stamps; step 3, data model comparison, namely carrying out layered comparison on the real-time broadcasting data model and an expected data model of a corresponding period according to a time stamp, wherein the layered comparison comprises data layer comparison and content layer comparison; and step 4, abnormality judgment and alarm triggering, namely judging whether the comparison result exceeds a preset fault tolerance threshold, if so, judging that the broadcast is abnormal, triggering an alarm instruction, and executing operat