Search

CN-121984964-A - Intelligent short video library transmission system and method for maritime satellite network

CN121984964ACN 121984964 ACN121984964 ACN 121984964ACN-121984964-A

Abstract

The invention discloses a short video library intelligent transmission system and method for a maritime satellite network, which relate to the technical field of intelligent data transmission and comprise the steps of carrying out content complexity and network state joint evaluation on a service flow group, generating corresponding compression ratio and data transmission block length according to an evaluation result, carrying out difference recognition on data to be transmitted to generate a data unit set to be transmitted and a packet decision record file, selecting transmission content based on the data unit set to be transmitted and the decision record, adopting a multiple quality control mechanism to monitor QoS indexes in real time to generate a QoS evaluation report, and carrying out optimization updating on a service flow packet model based on the QoS evaluation report. The method and the device effectively improve the utilization efficiency of the maritime satellite bandwidth and enhance the transmission reliability in a complex network environment.

Inventors

  • YAO HU
  • LI WEN
  • Si Gangjun

Assignees

  • 北京数智星通科技有限公司

Dates

Publication Date
20260505
Application Date
20260121

Claims (10)

  1. 1. A marine satellite network-oriented short video library intelligent transmission method is characterized by comprising the steps of, Collecting metadata of each file in a short video library, performing traditional feature analysis on short video key frames to obtain standardized feature vectors, and simultaneously combining the metadata and the standardized feature vectors to generate a metadata library; Performing check sum cleaning on the metadata database to obtain a normalized feature set, constructing a content similarity matrix and a time overlap matrix to generate a feature association structure, and constructing a business flow grouping model by adopting an integrated learning framework; performing content complexity and network state joint evaluation on the service flow group, generating corresponding compression ratio and data transmission block length according to an evaluation result, and performing difference recognition on compressed data to be transmitted to generate a data unit set to be transmitted and a grouping decision record document; Selecting transmission content based on the data unit set to be transmitted and the decision record, and adopting a multiple quality control mechanism to monitor QoS indexes in real time to generate a QoS evaluation report; and based on the QoS evaluation report, optimizing and updating the service flow grouping model.
  2. 2. The method for intelligent transmission of short video library for marine satellite network according to claim 1, wherein the metadata comprises file size, format, duration, resolution, transmission priority, emergency level identification, version number and storage path.
  3. 3. The method for intelligent transmission of short video library for marine satellite network according to claim 2, wherein the generating metadata library comprises the specific steps of, Reading metadata of each short video, temporarily storing the metadata into a memory buffer area, and generating a metadata set; extracting short video key frames from the metadata set and performing traditional feature analysis to generate a feature multidimensional array; Performing normalization conversion on the feature multidimensional array, integrating the feature multidimensional array into floating point type vectors with uniform length, and generating standardized feature vectors; And combining the metadata set and the standardized feature vector to generate a structured data record, persisting the structured data record to a disk by adopting columnar storage, and constructing a memory index acceleration query based on a B+ tree to generate a metadata database.
  4. 4. The method for intelligent transmission of short video library for marine satellite network according to claim 3, wherein said generating feature association structure comprises the steps of, Carrying out integrity check and normalization processing on the metadata base to generate a normalized feature set; calculating content feature similarity between data to be transmitted based on the normalized feature set; constructing a content similarity matrix according to the content feature similarity, analyzing the overlapping relation of the execution time window, and establishing a time overlapping proportion matrix; and carrying out association analysis on the content similarity matrix and the time overlap matrix to generate a characteristic association structure.
  5. 5. The method for intelligent transmission of short video library for marine satellite network according to claim 4, wherein said constructing a traffic stream packet model by using an ensemble learning framework comprises the steps of, Selecting a random forest model under an integrated learning framework as a basis, mapping a feature association structure to an integrated learning input space, and establishing a feature selection strategy through a multi-stage filtering method and a recursive feature elimination method; Based on the feature selection strategy, a feature weight distribution mechanism is adopted to formulate a group boundary judgment condition and a priority processing principle, and a group rule is generated; And carrying out deep fusion on the random forest model under the integrated learning framework and the grouping rules, defining the input and output specifications of the model, establishing a parameter configuration interface, and generating a business flow grouping model.
  6. 6. The method for intelligent transmission of short video library for marine satellite network according to claim 5, wherein said generating corresponding compression ratio and data transmission block length comprises the steps of, Deep feature analysis is carried out on short video content in each service flow group, a content complexity evaluation report is obtained, and the bandwidth of a maritime satellite link is monitored, so that a network state evaluation report is obtained; combining the content complexity score report with the network state evaluation report to establish a two-dimensional decision matrix; Performing unit matching and transmission parameter analysis on the two-dimensional decision matrix to generate a transmission strategy document; compression ratio and data transmission block length are set for each traffic group according to the transmission policy document.
  7. 7. The method for intelligent transmission of short video library for marine satellite network according to claim 6, wherein said generating the set of data units to be transmitted and the grouping decision record document comprises the steps of, According to the set compression ratio, comparing the current version of short video data with the existing version of the receiving end in frame level, identifying the changed area, and performing block processing according to the set data transmission block length to generate an incremental update data block; the method comprises the steps of carrying out standardized packaging on incremental updating data blocks, adding metadata heads for each data block, judging content importance by analyzing decoding dependency relations and motion characteristics among short video key frames, and meanwhile, carrying out priority classification on the data blocks to generate a data unit set to be transmitted; Performing transmission sequence arrangement and capacity adaptation on a data unit set to be transmitted to generate a data transmission list; and calling an archiving interface based on the data transmission list to generate the grouping decision record document.
  8. 8. The method for intelligent transmission of short video library for marine satellite network according to claim 7, wherein said generating QoS assessment report comprises the specific steps of, Screening transmission contents according to the transmission priority and the execution sequence according to the data unit set to be transmitted and the grouping decision record file, and generating transmissible contents; based on the transmissible content, adopting a multiple quality control mechanism to monitor QoS index in real time; establishing a time sequence database by using the monitored QoS index, and extracting the historical variation trend of the QoS index from the time sequence database; and carrying out abnormal pattern recognition and performance degradation early warning on the historical variation trend of the QoS index by adopting a multidimensional evaluation method, and generating a QoS evaluation report.
  9. 9. The method for intelligent transmission of short video library for marine satellite network according to claim 8, wherein said optimizing updating of traffic flow grouping model based on QoS evaluation report comprises the specific steps of, Extracting performance indexes from the QoS evaluation report, and precisely matching the performance indexes of each group with the grouping decision record file to generate a grouping efficiency sample library; Dividing the grouping efficiency sample library into an excellent group and a poor group according to performance, and performing difference analysis on the excellent group and the poor group; Based on the difference analysis result, establishing a problem mode of the poor performance group corresponding to the evasion rule in the traffic flow grouping model, strengthening a grouping mode with good performance in the good group, and generating an optimization strategy; and carrying out optimization updating on the service flow grouping model according to the optimization strategy.
  10. 10. A marine satellite network-oriented short video library intelligent transmission system based on the marine satellite network-oriented short video library intelligent transmission method of any one of claims 1-9 is characterized by comprising, The processing analysis module is used for collecting the metadata of the short video library, extracting the traditional features of the key frames, generating standardized feature vectors, establishing a metadata library, and simultaneously performing data cleaning and normalization processing to form a standardized feature set; The intelligent grouping module is used for constructing a content similarity matrix and a time overlapping matrix according to the normalized feature set, generating a feature association structure, and training a service flow grouping model by adopting an integrated learning framework; The dynamic configuration module is used for carrying out content complexity and network state joint evaluation on the service flow group, generating a transmission parameter configuration scheme according to an evaluation result, and carrying out difference recognition on data to be transmitted to generate a data unit set to be transmitted and a grouping decision record document; The execution monitoring module is used for selecting transmission contents according to the data unit set to be transmitted and the decision record, implementing a multiple quality control mechanism, monitoring QoS indexes in real time and generating a QoS evaluation report; and the optimization updating module is used for extracting performance indexes based on the QoS evaluation report, generating an optimization strategy and continuously optimizing and updating the service flow grouping model.

Description

Intelligent short video library transmission system and method for maritime satellite network Technical Field The invention relates to the technical field of intelligent data transmission, in particular to a short video library intelligent transmission system and method for a maritime satellite network. Background With the rapid growth of maritime communication and multimedia service demands, short video content transmission technology oriented to maritime satellite networks has become a key field of research and application. Traditional maritime satellite communications are limited by inherent characteristics such as high latency, limited bandwidth, and channel instability, and data transmission efficiency and reliability pose serious challenges. In recent years, edge computing and machine learning are introduced to improve processing efficiency, but intelligent analysis of content, dynamic grouping of tasks and real-time network perception are not coordinated deeply at the edge side to form a complete resource optimization closed loop, so that service quality and transmission efficiency in a marine severe environment are difficult to ensure. Therefore, how to improve the service quality and the transmission efficiency of the intelligent transmission of the short video library facing the maritime satellite network by using an advanced technical means is one of the problems to be solved currently. In the field of intelligent transmission of short video libraries for maritime satellite networks, the traditional processing method lacks instant decision capability for deep fusion of short video content characteristics and real-time network states under an edge calculation paradigm, so that a transmission strategy is stiff, satellite bandwidth resource utilization efficiency is poor, a closed-loop optimization mechanism based on real-time service quality feedback acquired at an edge side and rapid model iteration is lacking, and long-term dynamic change of a maritime environment is difficult to deal with. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a short video library intelligent transmission method for a maritime satellite network, which solves the problems of insufficient collaborative decision-making capability and lack of optimization mechanism. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides a short video library intelligent transmission method for a marine satellite network, which comprises the steps of collecting metadata of each file in a short video library, carrying out traditional feature analysis on short video key frames to obtain standardized feature vectors, simultaneously combining the metadata with the standardized feature vectors to generate a metadata library, checking and cleaning the metadata library to obtain a standardized feature set, generating a feature association structure by constructing a content similarity matrix and a time overlapping matrix, constructing a service flow grouping model by adopting an integrated learning framework, carrying out content complexity and network state joint evaluation on the service flow group, generating corresponding compression ratio and data transmission block length according to an evaluation result, carrying out difference recognition on compressed data to be transmitted to generate a data unit set to be transmitted and a grouping decision record file, selecting transmission content based on the data unit set to be transmitted and the decision record, monitoring QoS indexes in real time by adopting a multiple quality control mechanism to generate a QoS evaluation report, and carrying out optimization updating on the service flow grouping model based on the QoS evaluation report. As an optimal scheme of the maritime satellite network-oriented short video library intelligent transmission method, the metadata comprises file size, format, duration, resolution, transmission priority, emergency degree identification, version number and storage path. As an optimal scheme of the short video library intelligent transmission method facing the marine satellite network, the method comprises the specific steps of generating a metadata library, Reading metadata of each short video, temporarily storing the metadata into a memory buffer area, and generating a metadata set; extracting short video key frames from the metadata set and performing traditional feature analysis to generate a feature multidimensional array; Performing normalization conversion on the feature multidimensional array, integrating the feature multidimensional array into floating point type vectors with uniform length, and generating standardized feature vectors; And combining the metadata set and the standardized feature vector to generate a structured data record, persisting the structured data record