A video data acquisition method and system for intelligent video generation

Through crawling technology and splitting algorithm, video data is converted into text and visual storyboard segment sets, which solves the demand for high-quality training data of intelligent video technology, and realizes efficient video data acquisition and intelligent video generation model support.

CN118646834BActive Publication Date: 2024-11-08HAOSHEN TECH (BEIJING) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411107806.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-08
Estimated Expiration
2044-08-13

AI Technical Summary

Technical Problem

Existing video data acquisition methods cannot effectively meet the demand for high-quality training data of intelligent video technology, resulting in high cost and low efficiency.

Method used

Using crawler technology to collect video data on the network platform, convert video data into text segmentation sets and visual segmentation sets through text segmentation algorithms and visual segmentation algorithms, and generate the final material set to realize the visual and text of video content. Split.

Benefits of technology

It realizes the generation of high-quality storyboard text and video material pairing data, providing reliable training data support for the intelligent video generation model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118646834B_ABST
    Figure CN118646834B_ABST
Patent Text Reader

Abstract

The present invention discloses a video data acquisition method and system for intelligent video generation, which relates to the field of video generation technology. The method includes: using crawler technology to collect video data on various network platforms; extracting audio tracks in video data and converting the audio tracks into text sequences; creating a text splitting algorithm, and converting the text sequence into a text storyboard segment set based on the text splitting algorithm; creating a visual splitting algorithm, and converting the frame sequence of video data into a visual storyboard segment set based on the visual splitting algorithm; generating a final material set based on the text storyboard segment set and the visual storyboard segment set. The beneficial effects achieved by the present invention are as follows: realizing visual and text splitting of video content, generating high-quality storyboard text and video material pairing data, and providing reliable data support for the training of intelligent video generation models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of video generation technology, and in particular to a video data acquisition method and system for intelligent video generation. Background Art

[0002] Currently, video data acquisition methods cover a variety of technologies from traditional analog closed-circuit television monitoring to modern digital, networked, and intelligent technologies. Intelligent video technology uses computer vision technology to understand, analyze, and process video information collected by cameras, thereby filtering and excluding irrelevant information and extracting useful information. However, intelligent video technology requires a large amount of high-quality training data. The current video data acquisition methods cannot fully meet the needs of intelligent video technology. Therefore, how to generate these high-quality training data at low cost and high efficiency is an urgent problem that people in this field need to solve. Summary of the invention

[0003] The present invention provides a video data acquisition method for generating smart videos, comprising:

[0004] Step 1: Use crawler technology to collect video data on various network platforms;

[0005] Step 2, extract the audio track from the video data and convert the audio track into a text sequence;

[0006] Step 3, create a text segmentation algorithm, and convert the text sequence into a text storyboard set based on the text segmentation algorithm;

[0007] Step 4, create a visual segmentation algorithm, and convert the frame sequence of the video data into a set of visual storyboard segments based on the visual segmentation algorithm;

[0008] Step 5. Generate the final material set based on the text storyboard segment set and the visual storyboard segment set.

[0009] The video data collection method for intelligent video generation described above, in which crawler technology is used to collect video data on various network platforms, specifically includes the following sub-steps:

[0010] Configure crawler tasks and obtain video links on the target network platform by executing crawler tasks;

[0011] Define pull conditions for video links and filter out video links that do not meet the pull conditions;

[0012] The video data is obtained through the video link and the video data is numbered and stored.

[0013] The video data acquisition method for intelligent video generation as described above, wherein a text sequence is converted into a text storyboard segment set based on a text segmentation algorithm, specifically comprises the following sub-steps:

[0014] Define focus topics based on user needs for smart video generation;

[0015] Calculate the degree of fit between each paragraph of the text sequence and different focus topics, and split the text sequence into multiple text storyboards based on the fit;

[0016] The text storyboards of various text sequences are organized and stored as a text storyboard segment set.

[0017] The video data acquisition method for intelligent video generation as described above, wherein the frame sequence of the video data is converted into a set of visual storyboard segments based on a visual segmentation algorithm, specifically comprises the following sub-steps:

[0018] Extracting frame data of each video to form multiple frame sequences;

[0019] Calculate the continuity index between each video frame and adjacent frames in the frame sequence, and split the frame sequence into multiple visual storyboards based on the continuity index;

[0020] The visual storyboards of each frame sequence are organized and stored as a visual storyboard clip set.

[0021] The video data acquisition method for intelligent video generation as described above, wherein the final material set is generated based on the text storyboard segment set and the visual storyboard segment set, specifically comprises the following sub-steps:

[0022] Align the text storyboard clips with the visual storyboard clips according to the source video numbers and timeline positions;

[0023] If the text clip is longer than the visual clip, remove the redundant part of the text clip;

[0024] If the visual storyboard is longer than the text storyboard, remove the redundant part of the visual storyboard;

[0025] Use the aligned text storyboard clips to annotate the visual storyboard clips and store them in the asset collection.

[0026] The present invention also provides a video data acquisition system for intelligent video generation, comprising: a video data acquisition module, a text storyboard splitting module, a visual storyboard splitting module and a material set generation module;

[0027] Video data collection module, used to collect video data on various network platforms using crawler technology;

[0028] A text storyboard splitting module is used to convert the audio track in the video data into a text sequence and split the text sequence into multiple text storyboards;

[0029] A visual storyboard splitting module, used for splitting a frame sequence of video data into visual storyboard segments;

[0030] The material set generation module is used to generate the final material set based on the text storyboard segment set and the visual storyboard segment set.

[0031] A video data acquisition system for intelligent video generation as described above, wherein the video data acquisition module includes a crawler task configuration submodule, a pull condition configuration submodule, and a video data storage submodule;

[0032] The crawler task configuration submodule is used to configure the crawler task and obtain the video link on the target network platform by executing the crawler task;

[0033] The pull condition configuration submodule is used to define pull conditions for video links and filter out video links that do not meet the pull conditions;

[0034] The video data storage submodule is used to obtain video data through a video link and store the video data in numbers.

[0035] A video data acquisition system for intelligent video generation as described above, wherein the text storyboard splitting module includes an audio-to-text submodule, a focus theme definition submodule, a fit calculation submodule, and a text storyboard annotation submodule;

[0036] The audio-to-text submodule is used to extract the audio track from the video data and convert the audio track into a text sequence;

[0037] A focus topic definition submodule is used to define a focus topic according to the user's needs for smart video generation;

[0038] A fit calculation submodule is used to calculate the fit between each paragraph of the text sequence and different focus topics, and split the text sequence into multiple text storyboards according to the fit;

[0039] The text storyboard annotation submodule is used to annotate the text storyboards and store them as a text storyboard segment set.

[0040] A video data acquisition system for intelligent video generation as described above, wherein the visual storyboard splitting module includes a frame sequence extraction submodule, a continuity index calculation submodule, and a visual storyboard annotation submodule;

[0041] A frame sequence extraction submodule is used to extract frame data of each video to form multiple frame sequences;

[0042] A continuity index calculation submodule, used to calculate the continuity index between each video frame and adjacent frames in the frame sequence, and split the frame sequence into multiple visual storyboards based on the continuity index;

[0043] The visual storyboard annotation submodule is used to annotate the visual storyboard data and store it as a visual storyboard segment set.

[0044] The beneficial effects achieved by the present invention are as follows: achieving visual and textual segmentation of video content, generating high-quality storyboard text and video material pairing data, and providing reliable data support for the training of intelligent video generation models. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] 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 recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0046] Figure 1 This is a flow chart of a video data acquisition method for intelligent video generation provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0047] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention. Embodiment 1

[0048] like Figure 1 As shown, the first embodiment of the present invention provides a video data acquisition method for smart video generation, including:

[0049] Step S10: using crawler technology to collect video data on various network platforms;

[0050] Crawler technology can quickly obtain a large amount of video data. During the collection process, we should also pay attention to a preliminary screening of the video data to remove invalid or poor quality video data. Specifically:

[0051] Step S11: configuring a crawler task, and obtaining a video link on a target network platform by executing the crawler task;

[0052] The configuration items of the crawler task include the link of the target network platform, the crawled content, the request frequency, the exception handling strategy, and the database used to store and access the crawled data. After the configuration is completed and tested, you can set a timer to execute the crawler task regularly.

[0053] Step S12: defining a pull condition for the video link, and filtering out video links that do not meet the pull condition;

[0054] You can obtain the desired video data by defining specific pull conditions such as release year, publisher, link format, video description, clarity, etc.

[0055] Step S13: Obtain video data through the video link, and store the video data by number;

[0056] The numbering is in the form of pull date-pull batch-auto-increment sequence, so that the data pulled at that time can be accurately located when a crawling exception occurs.

[0057] Step S20: extracting the audio track from the video data and converting the audio track into a text sequence;

[0058] Extract the audio track M from the video and convert the audio into text using speech recognition technology , the formula is as follows:

[0059] The converted text is preprocessed, including removing noise, stop words, punctuation marks, as well as word segmentation and paragraph tagging, to form a text sequence containing multiple word units. The text sequences generated by different videos are distinguished by the video numbers.

[0060] Step S30: creating a text segmentation algorithm, and converting the text sequence into a text storyboard segment set based on the text segmentation algorithm;

[0061] The purpose of the text segmentation algorithm is to identify semantically continuous but content-different segments by analyzing the text semantics of the audio in the video. Specifically:

[0062] Step S31: defining a focus topic according to the user's demand for generating a smart video;

[0063] The generation requirements of users of various smart video generation platforms are collected and organized into a data set, and then a cluster analysis is performed on the generation requirement data set to ensure that the gap between each cluster center and the most marginal data in the cluster analysis results is within 10%~15%, and then each cluster center is extracted as the feature vector of the focusing subject.

[0064] Step S32: calculating the degree of fit between each paragraph of the text sequence and different focus topics, and splitting the text sequence into multiple text storyboards according to the degree of fit;

[0065] Calculation paragraph A i With focus topic C j The fit formula is expressed as: ,in is the i-th paragraph A in the text sequence i With the jth focused topic C j The fit, represents the feature vector of the kth word in the i-th paragraph in the text sequence, express The frequency of occurrence in the i-th paragraph, N is the total number of all words in the entire text sequence, express The frequency of occurrence in the entire text sequence, and Respectively and The modulus of k ranges from 1 to Z, where Z represents the number of tokens contained in the i-th paragraph in the text sequence.

[0066] After calculating the paragraph and A i After comparing the fit with each focus topic, select the focus topic with the highest fit as paragraph A. i The topic of paragraph A is then calculated i+1 The degree of fit with each focus topic and paragraph A i+1 Assign themes until all paragraphs in the text sequence have been assigned themes. Paragraphs belonging to the same theme will be automatically grouped into one text storyboard.

[0067] Step S33: Arrange the text storyboards of each text sequence and store them as a text storyboard segment set;

[0068] The text storyboards split from each text sequence are labeled according to the theme + source video number + timeline position and stored in the text storyboard clip set.

[0069] Step S40: creating a visual segmentation algorithm, and converting a frame sequence of the video data into a set of visual storyboard segments based on the visual segmentation algorithm;

[0070] The purpose of the visual segmentation algorithm is to identify visually continuous but content-different segments by analyzing the continuity of video frames. Specifically:

[0071] Step S41: extracting frame data of each video to form multiple frame sequences;

[0072] Extracting frame sequences from a video , the formula is as follows:

[0073]

[0074] in represents the frame sequence extracted from the video V, Indicates the video frame extracted at timestamp t, where t ranges from 1 to L, and L is the total number of timestamps extracted from video V.

[0075] Step S42: calculating the continuity index between each video frame and adjacent frames in the frame sequence, and splitting the frame sequence into multiple visual storyboards based on the continuity index;

[0076] Use the formula: To calculate the continuity index between each video frame and adjacent frames in the frame sequence, Represents a video frame With video frame The continuous index between Represents a video frame In Location The value of the upper color channel r, Represents a video frame In Location The value of the upper color channel r, is an adjustable parameter used to control the color gradient. is a function that returns 1 if the argument in the parentheses is 0, otherwise it returns 0. Video frame The image coordinates of the pth random position in , p takes values ​​from 1 to W, W represents the number of random positions taken, r takes values ​​from 1 to R, R represents the number of color channels, Represents the judgment weight of the rth color channel on the continuity index.

[0077] When the continuity index mutates, it means that the video content has changed. Then the video frames between this mutation and the previous mutation are intercepted as a visual storyboard. The conditions for judging the mutation can be To indicate that Represents a video frame With video frame The continuous index between Represents video frames and video frames The continuous index between is the change threshold, when Greater than When , it is considered that the continuity index has undergone a mutation.

[0078] Step S43: Arranging the visual storyboards of each frame sequence and storing them as a visual storyboard segment set;

[0079] The visual storyboards split from each frame sequence are labeled according to the storyboard sequence number + source video number + timeline position, and stored in the time storyboard segment set.

[0080] Step S50: generating a final material set based on the text storyboard segment set and the visual storyboard segment set;

[0081] The generated material set can be used as training data for intelligent generation of large models or multimodal large models, or as a material library for large editing software. The specific generation process includes the following sub-steps:

[0082] Step S51: aligning the text storyboard segments and the visual storyboard segments according to the source video numbers and timeline positions;

[0083] Step S52: if the text storyboard segment is longer than the visual storyboard segment, remove the redundant part of the text storyboard segment;

[0084] Step S53: if the visual storyboard segment is longer than the text storyboard segment, removing the redundant portion of the visual storyboard segment;

[0085] Step S54: Use the aligned text storyboard segments to annotate the visual storyboard segments and store them in the material collection. Embodiment 2

[0086] Embodiment 2 of the present invention provides a video data acquisition system for intelligent video generation, including: a video data acquisition module, a text storyboard splitting module, a visual storyboard splitting module and a material set generation module;

[0087] (1) Video data acquisition module, which is used to collect video data on various network platforms using crawler technology, including crawler task configuration submodule, pull condition configuration submodule, and video data storage submodule;

[0088] 1. The crawler task configuration submodule is used to configure the crawler task and obtain the video link on the target network platform by executing the crawler task;

[0089] The configuration items of the crawler task include the link of the target network platform, the crawled content, the request frequency, the exception handling strategy, and the database used to store and access the crawled data. After the configuration is completed and tested, you can set a timer to execute the crawler task regularly.

[0090] 2. Pull condition configuration submodule, used to define pull conditions for video links and filter out video links that do not meet the pull conditions;

[0091] You can obtain the desired video data by defining specific pull conditions such as release year, publisher, link format, video description, clarity, etc.

[0092] 3. Video data storage submodule, used to obtain video data through video link and store the video data by number;

[0093] The numbering is in the form of pull date-pull batch-auto-increment sequence, so that the data pulled at that time can be accurately located when a crawling exception occurs.

[0094] (2) A text storyboard splitting module, which is used to convert the audio track in the video data into a text sequence and split the text sequence into multiple text storyboards, including an audio-to-text submodule, a focus theme definition submodule, a fit calculation submodule, and a text storyboard annotation submodule;

[0095] 1. Audio-to-text submodule, used to extract the audio track from the video data and convert the audio track into a text sequence;

[0096] Extract the audio track M from the video and convert the audio into text using speech recognition technology , the formula is as follows:

[0097] The converted text is preprocessed, including removing noise, stop words, punctuation marks, as well as word segmentation and paragraph tagging, to form a text sequence containing multiple word units. The text sequences generated by different videos are distinguished by the video numbers.

[0098] 2. Focus topic definition submodule, used to define focus topics according to user needs for smart video generation;

[0099] The generation requirements of users of various smart video generation platforms are collected and organized into a data set, and then a cluster analysis is performed on the generation requirement data set to ensure that the gap between each cluster center and the most marginal data in the cluster analysis results is within 10%~15%, and then each cluster center is extracted as the feature vector of the focusing subject.

[0100] 3. The fit calculation submodule is used to calculate the fit between each paragraph of the text sequence and different focus topics, and split the text sequence into multiple text storyboards according to the fit;

[0101] Calculation paragraph A i With focus topic C j The fit formula is expressed as: ,in is the i-th paragraph A in the text sequence i With the jth focused topic C j The fit, represents the feature vector of the kth word in the i-th paragraph in the text sequence, express The frequency of occurrence in the i-th paragraph, N is the total number of all words in the entire text sequence, express The frequency of occurrence in the entire text sequence, and Respectively and The modulus of k ranges from 1 to Z, where Z represents the number of tokens contained in the i-th paragraph in the text sequence.

[0102] After calculating the paragraph and A i After comparing the fit with each focus topic, select the focus topic with the highest fit as paragraph A. i The topic of paragraph A is then calculated i+1 The degree of fit with each focus topic and paragraph A i+1 Assign themes until all paragraphs in the text sequence have been assigned themes. Paragraphs belonging to the same theme will be automatically grouped into one text storyboard.

[0103] 4. A text storyboard annotation submodule, which is used to annotate the text storyboards and store them as a text storyboard segment set;

[0104] The text storyboards split from each text sequence are labeled according to the theme + source video number + timeline position and stored in the text storyboard clip set.

[0105] (3) A visual storyboard splitting module, which is used to split the frame sequence of video data into visual storyboard segments, including a frame sequence extraction submodule, a continuity index calculation submodule, and a visual storyboard annotation submodule;

[0106] 1. A frame sequence extraction submodule, used to extract frame data of each video to form multiple frame sequences;

[0107] Extracting frame sequences from a video , the formula is as follows:

[0108] in represents the frame sequence extracted from the video V, Indicates the video frame extracted at timestamp t, where t ranges from 1 to L, and L is the total number of timestamps extracted from video V.

[0109] 2. A continuity index calculation submodule, which is used to calculate the continuity index between each video frame and adjacent frames in the frame sequence, and split the frame sequence into multiple visual storyboards based on the continuity index;

[0110] Use the formula: To calculate the continuity index between each video frame and adjacent frames in the frame sequence, Represents a video frame In Location The value of the upper color channel r, Represents a video frame In Location The value of the upper color channel r, is an adjustable parameter used to control the color gradient. is a function that returns 1 if the argument in the parentheses is 0, otherwise it returns 0. Represents a video frame The image coordinates of the pth random position in , p takes values ​​from 1 to W, W represents the number of random positions taken, r takes values ​​from 1 to R, R represents the number of color channels, Represents the judgment weight of the rth color channel on the continuity index.

[0111] When the continuity index mutates, it means that the video content has changed. Then the video frames between this mutation and the previous mutation are intercepted as a visual storyboard. The conditions for judging the mutation can be To indicate that Represents a video frame With video frame The continuous index between Represents a video frame With video frame The continuous index between is the change threshold, when Greater than When , it is considered that the continuity index has undergone a mutation.

[0112] 3. A visual storyboard annotation submodule, which is used to annotate the visual storyboards and store them as a visual storyboard segment set;

[0113] The visual storyboards split from each frame sequence are labeled according to the storyboard sequence number + source video number + timeline position, and stored in the time storyboard segment set.

[0114] (4) a material set generation module, used to generate a final material set based on the text storyboard segment set and the visual storyboard segment set;

[0115] The generated material set can be used as training data for intelligent generation of large models or multimodal large models, or as a material library for large editing software. The specific generation process includes the following sub-steps:

[0116] I. Align the text storyboard clips and the visual storyboard clips according to the source video numbers and timeline positions;

[0117] II. If the text storyboard is longer than the visual storyboard, remove the redundant part of the text storyboard;

[0118] III. If the visual storyboard segment is longer than the text storyboard segment, remove the redundant portion of the visual storyboard segment;

[0119] IV. Use the aligned text storyboard segments to annotate the visual storyboard segments and store them in the material collection.

[0120] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.

Claims

1. A video data acquisition method for intelligent video generation, characterized in that: The method comprises: Step 1: Use crawler technology to collect video data on various network platforms; Step 2, extract the audio track from the video data and convert the audio track into a text sequence; Step 3: Create a text segmentation algorithm and convert the text sequence into a text storyboard set based on the text segmentation algorithm. The specific steps include the following: Define focus topics based on user needs for smart video generation; Calculate the degree of fit between each paragraph of the text sequence and different focus topics, and split the text sequence into multiple text storyboards based on the fit; The formula for calculating the fit between the i-th paragraph and the j-th focus topic is expressed as: ,in is the i-th paragraph A in the text sequence i With the jth focused topic C j The fit, represents the feature vector of the kth word in the i-th paragraph in the text sequence, express The frequency of occurrence in the i-th paragraph, N is the total number of all words in the entire text sequence, express The frequency of occurrence in the entire text sequence, and Respectively and The modulus of k ranges from 1 to Z, where Z represents the number of tokens contained in the i-th paragraph in the text sequence; Arrange the text storyboards of each text sequence and store them as a text storyboard segment set; Step 4, create a visual segmentation algorithm, and convert the frame sequence of the video data into a set of visual storyboard segments based on the visual segmentation algorithm; Step 5. Generate the final material set based on the text storyboard segment set and the visual storyboard segment set.

2. The video data acquisition method for intelligent video generation according to claim 1, characterized in that: Use crawler technology to collect video data on various network platforms, which includes the following sub-steps: Configure crawler tasks and obtain video links on the target network platform by executing crawler tasks; Define pull conditions for video links and filter out video links that do not meet the pull conditions; The video data is obtained through the video link and the video data is numbered and stored.

3. The video data acquisition method for intelligent video generation according to claim 1, characterized in that: The frame sequence of the video data is converted into a set of visual storyboard segments based on a visual segmentation algorithm, which specifically includes the following sub-steps: Extracting frame data of each video to form multiple frame sequences; Calculate the continuity index between each video frame and adjacent frames in the frame sequence, and split the frame sequence into multiple visual storyboards based on the continuity index; The visual storyboards of each frame sequence are organized and stored as a visual storyboard clip set.

4. The video data acquisition method for intelligent video generation according to claim 1, characterized in that: The final material set is generated based on the text storyboard segment set and the visual storyboard segment set, which specifically includes the following sub-steps: Align the text storyboard clips with the visual storyboard clips according to the source video numbers and timeline positions; If the text clip is longer than the visual clip, remove the redundant part of the text clip; If the visual storyboard is longer than the text storyboard, remove the redundant part of the visual storyboard; Use the aligned text storyboard clips to annotate the visual storyboard clips and store them in the asset collection.

5. A video data acquisition system for intelligent video generation, characterized in that: The system comprises: a video data acquisition module, a text storyboard splitting module, a visual storyboard splitting module and a material set generation module; Video data collection module, used to collect video data on various network platforms using crawler technology; A text storyboard splitting module is used to convert the audio track in the video data into a text sequence and split the text sequence into multiple text storyboards; it includes an audio-to-text submodule, a focus theme definition submodule, a fit calculation submodule, and a text storyboard annotation submodule; The audio-to-text submodule is used to extract the audio track from the video data and convert the audio track into a text sequence; A focus topic definition submodule is used to define a focus topic according to the user's needs for smart video generation; A fit calculation submodule is used to calculate the fit between each paragraph of the text sequence and different focus topics, and split the text sequence into multiple text storyboards according to the fit; The formula for calculating the fit between the i-th paragraph and the j-th focus topic is expressed as: ,in is the i-th paragraph A in the text sequence i With the jth focused topic C j The fit, represents the feature vector of the kth word in the i-th paragraph in the text sequence, express The frequency of occurrence in the i-th paragraph, N is the total number of all words in the entire text sequence, express The frequency of occurrence in the entire text sequence, and Respectively and The modulus of k ranges from 1 to Z, where Z represents the number of tokens contained in the i-th paragraph in the text sequence; A text storyboard annotation submodule is used to annotate the text storyboards and store them as a text storyboard segment set; A visual storyboard splitting module, used for splitting a frame sequence of video data into visual storyboard segments; The material set generation module is used to generate the final material set based on the text storyboard fragment set and the visual storyboard fragment set.

6. The video data acquisition system for intelligent video generation according to claim 5, characterized in that: The video data acquisition module includes a crawler task configuration submodule, a pull condition configuration submodule, and a video data storage submodule; The crawler task configuration submodule is used to configure the crawler task and obtain the video link on the target network platform by executing the crawler task; The pull condition configuration submodule is used to define pull conditions for video links and filter out video links that do not meet the pull conditions; The video data storage submodule is used to obtain video data through a video link and store the video data in numbers.

7. The video data acquisition system for intelligent video generation according to claim 5, characterized in that: The visual storyboard splitting module includes a frame sequence extraction submodule, a continuity index calculation submodule, and a visual storyboard annotation submodule; A frame sequence extraction submodule is used to extract frame data of each video to form multiple frame sequences; A continuity index calculation submodule, used to calculate the continuity index between each video frame and adjacent frames in the frame sequence, and split the frame sequence into multiple visual storyboards based on the continuity index; The visual storyboard annotation submodule is used to annotate the visual storyboard data and store it as a visual storyboard segment set.