A broadcast television news video auxiliary editing method and system

Abstract

The present invention relates to the field of content creation of radio and television news media assets, and discloses a method and system for auxiliary editing of radio and television news videos, wherein the method comprises: scene recognition: extracting the theme of a news manuscript, and identifying the scene paragraphs of the news manuscript based on the theme; shot script generation: segmenting the video material into video material segments, and marking the corresponding content tags and shooting technique tags; segmenting the identified scene paragraphs into segmented shot texts, and generating content scripts, picture descriptions, and shooting technique scripts for the shot texts; material recall: selecting suitable candidate video materials for the segmented shot texts; material assembly: assembling the candidate video materials into an edited video of the news manuscript. The present invention integrates advanced technologies in vision, natural language processing, and large models into the post-production stage of video production, and automatically edits the filmed materials in combination with the rules of audio-visual language knowledge, so as to ensure the rationality of editing and improve the quality of video editing and production.

Description

Technical Field

[0001] The present invention relates to the technical field of broadcasting and television news media content creation, and in particular to a broadcasting and television news video auxiliary editing method and system. Background Art

[0002] In the new media era, video has become a popular way of communication for the audience, and the video expression of news has of course become a necessity. In recent years, traditional mainstream media have launched a variety of new media forms, among which the video of news is the most representative. The dissemination of news videos and even news short videos can not only meet users' information needs and freshness, but also expand the influence of traditional media.

[0003] However, video production is very time-consuming. Generally speaking, video production can be divided into the pre-shooting and post-production stages. In the pre-shooting stage, after the shooting theme is determined, the shooting staff will go out to shoot a large amount of materials related to the theme for selection in the post-production stage; in the post-production stage, the editors need to select appropriate materials from a large number of alternative materials based on the news copy, and organize these materials into videos on the relevant editing software. This traditional purely manual video production is very time-consuming and labor-intensive, and it is easy to lose the timeliness of the news. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a method and system for auxiliary editing of radio and television news videos, which integrates advanced technologies in vision, natural language processing and large models into the post-production stage of video production, and automatically edits the filmed materials based on the rules of audio-visual language knowledge to shape the editing style, thereby ensuring the rationality of editing and improving the quality of video editing and production.

[0005] The technical solution adopted by the present invention is as follows:

[0006] On the one hand, the present invention provides a broadcast television news video auxiliary editing method, comprising:

[0007] Scene recognition: extract the subject of the news manuscript and identify the scene paragraphs of the news manuscript based on the subject;

[0008] Shot script generation: Segment the video material into video material segments and annotate the corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts;

[0009] Material recall: select suitable candidate video materials for the segmented shot text;

[0010] Material assembly: Assemble the candidate video materials into a video clip of the news manuscript.

[0011] Furthermore, the scene recognition comprises the following steps:

[0012] S11, extracting the subject of the news manuscript using subject extraction technology;

[0013] S12. Use text segmentation technology combined with the topics extracted in S11 to identify scene paragraphs in news manuscripts: generate the first sentence of each paragraph based on the segmentation capability of the large language model; use the text similarity algorithm to locate the position of the first sentence of each paragraph in the news manuscript, thereby identifying the scene paragraphs.

[0014] Furthermore, the shot script generation includes the following steps:

[0015] S21. Segment the video material into a plurality of video material segments by using visual technology, and label each video material segment with a corresponding content label and a shooting technique label;

[0016] S22. Generate a content script, a picture description and a shooting technique script for each shot text in the scene segment using content script generation technology, picture description generation technology and shooting technique script generation technology.

[0017] Furthermore, in S21, the step of using visual technology to segment the video material into a plurality of video material segments, and labeling each video material segment with a corresponding content label and a shooting technique label, includes the following sub-steps:

[0018] S211, dividing the video material into a plurality of video material segments by using a transition recognition technology;

[0019] S212, using background recognition technology, object recognition technology, face recognition technology, empty shot recognition technology, character feature generation technology and behavior action generation technology to annotate corresponding content tags for each video material segment in S211, including background, object, appearing character, empty shot, character feature and behavior action;

[0020] S213, using the lens scene recognition technology, the lens angle recognition technology, the shooting method recognition technology and the camera movement recognition technology to label each video material segment in S211 with a corresponding shooting technique label, including the scene, shooting angle, shooting method and camera movement.

[0021] Further, in S22, the process of generating a content script, a picture description and a shooting technique script for each shot text in the scene paragraph by using the content script generation technology, the picture description generation technology and the shooting technique script generation technology includes the following sub-steps:

[0022] S221, using a tag selection technique to select a set of characters, backgrounds, and objects tags related to the scene segment from the content tags in S21;

[0023] S222, combining the relevant character, background, and object label sets selected in S221 to recall the top K candidate content labels from the content labels corresponding to the video material clips in S21;

[0024] S223, using the shot segmentation technology to further segment the scene segments to obtain shot texts;

[0025] S224, using a content script generation technology to select a content tag as a shot content script for each shot based on the candidate content tags in S222;

[0026] S225, based on the shot text obtained in S223, the shot content script obtained in S224, and the scene paragraph obtained by the scene recognition, using a picture description generation technology to generate a corresponding picture description for each shot text;

[0027] S226. Using shooting technique script generation technology, based on the shot text obtained in S223, the shot content script obtained in S224, the scene segments obtained by the scene recognition, and the rules of audio-visual language knowledge, a shooting technique label is generated for each shot text as a shooting technique script.

[0028] Furthermore, the material recall includes the following steps:

[0029] S31, for each news manuscript, integrating and storing all video material clips obtained in the generation of the shot script to form a video material library;

[0030] S32. For the picture description generated for each shot text in the shot script generation, use cross-modal retrieval technology to find the top K candidate video material clips that best match the picture description from the video material library.

[0031] Furthermore, the material assembly includes the following steps:

[0032] S41, performing internal video splicing for each shot text using candidate video materials to generate a candidate video corresponding to the shot text;

[0033] S42. Execute S41 in sequence according to the order of the shot text in the news manuscript to obtain candidate videos of the shot text, and then perform video splicing of the edited videos of the shot text in the news manuscript to form an edited video of the news manuscript.

[0034] Furthermore, in S41, the internal video splicing is performed on each shot text using the candidate video materials to generate a candidate video corresponding to the shot text, including the following sub-steps:

[0035] S411, setting a scoring standard and a video duration required for the shot text; the scoring standard includes a weighted average of the matching degree of the candidate video material clip with the shooting technique script in the shot script generation, the matching degree with the shot text, and the matching degree with the rules of audio-visual language knowledge;

[0036] S412, for each shot text, the first K candidate video material segments retrieved from the material recall are used as the initial video bundle, and the video splicing process is started;

[0037] S413, using the scoring criteria in S411 to score each video material segment in the initial video bundle, selecting beam_size video material segments with the highest scores and connecting them to the video bundle, and then entering the next round of search; where beam_size is the maximum number of candidate items;

[0038] S414, during the splicing process, if the total duration of the spliced ​​video bundle exceeds the video duration required by the shot text, the video material segments are cropped to match the video duration requirement of the shot text; if the total duration of the video is not enough to meet the video duration required by the shot text, the cluster search strategy is continued to be used to select and splice the remaining candidate video materials into the video bundle;

[0039] S415. Once the total duration of the assembled video bundle reaches the video duration required by the shot text, the video assembly process of the video bundle is terminated; S413 to S414 are repeated until all video bundles meet the video duration requirements of the shot text; finally, candidate video bundles corresponding to beam_size shot texts are obtained as candidate videos of the shot text.

[0040] Furthermore, in S42, S41 is sequentially executed according to the order of the shot texts in the news manuscript to obtain candidate videos of the shot texts, and then the video editing of the shot texts in the news manuscript is performed to form the edited video of the news manuscript, including the following sub-steps:

[0041] S421, setting scoring criteria, including a weighted average of the matching degree of the candidate video with the shot text, the matching degree with the rules of audio-visual language knowledge, and the coherence between the shots;

[0042] S422, according to the order of the shot texts in the news manuscript, for each shot text, the candidate videos corresponding to the beam_size shot texts obtained in S41 are used as the initial editing video bundle, and the editing video assembly process is started;

[0043] S423, according to the scoring standard of S421, the candidate video of each shot text is combined with the current editing video bundle; the combined editing video bundles are scored, and beam_size editing video bundles with the highest scores are selected;

[0044] S424, marking the used video material segments in each clip video bundle to ensure that the used video material segments are no longer involved in subsequent searches;

[0045] S425. For each shot text, first execute S41 to obtain the candidate video corresponding to the shot text, and then execute S423 to S424 to obtain the edited video of the shot text, until the video assembly of all the shot texts in the news manuscript is completed; finally, beam_size candidate edited video bundles corresponding to the news manuscript are obtained, and the edited video bundle with the highest score is selected as the edited video of the news manuscript.

[0046] On the other hand, the present invention provides a broadcast television news video auxiliary editing system, comprising:

[0047] A scene recognition module is configured to extract the subject of the news manuscript and recognize the scene paragraphs of the news manuscript based on the subject;

[0048] The shot script generation module is configured to segment the video material into video material segments and annotate corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts;

[0049] A material recall module is configured to select suitable candidate video materials for the segmented shot texts;

[0050] The material assembly module is configured to assemble the candidate video materials into a clipped video of the news manuscript.

[0051] The beneficial effects of the present invention are:

[0052] (1) The present invention integrates advanced technologies in vision, natural language processing, and large models into the post-production stage of video production, and automatically edits the filmed materials based on the rules of audio-visual language knowledge to shape the editing style, thereby ensuring the rationality of the editing and improving the quality of video editing and production.

[0053] (2) The present invention can significantly reduce the time and effort required for manual editing through an automated editing process, thereby improving the efficiency of video production.

[0054] (3) The present invention utilizes cross-modal retrieval technology to more accurately match appropriate video materials and ensure that the edited video is highly consistent with the content of the news manuscript.

[0055] (4) The present invention combines audio-visual language knowledge to assemble a video with coherent content and harmonious audio-visual experience, further improving the audience's viewing experience.

[0056] (5) The present invention can improve the efficiency of news video production and the timeliness of news release, while ensuring the high quality and consistency of video content. By integrating advanced technology into news video editing, it can promote the accelerated transformation and upgrading of traditional media and lead and promote the deep integration and development of media. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 A flow chart of a broadcast and television news video auxiliary editing method of the present invention;

[0058] Figure 2 The present invention is a schematic diagram of a broadcast television news video auxiliary editing system. DETAILED DESCRIPTION

[0059] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, the specific implementation methods of the present invention are now described. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0060] Example 1

[0061] This embodiment provides a method for auxiliary editing of broadcast television news video. Figure 1 As shown, the following steps are included:

[0062] S1. Scene recognition: extract the theme of the news manuscript and identify the scene paragraphs of the news manuscript based on the theme;

[0063] S2, shot script generation: Segment the video material into video material segments and annotate the corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts;

[0064] S3, material recall: select suitable candidate video materials for the segmented shot text;

[0065] S4. Material assembly: Assemble the candidate video materials into a video clip of the news manuscript.

[0066] Preferably, step S1 includes the following sub-steps:

[0067] S11. Identify the subject of the news article using subject extraction technology;

[0068] S12. Use text segmentation technology combined with the topics extracted in S11 to identify scene paragraphs in the news manuscript.

[0069] More preferably, in step S12, the text segmentation technology is combined with the topics extracted in S11 to perform scene paragraph recognition on the news manuscript, including the following sub-steps:

[0070] S121. Generate the first sentence of each paragraph based on the segmentation capability of the large language model;

[0071] S122. Using a text similarity algorithm, locate the first sentence of each paragraph in the news manuscript, thereby identifying the scene paragraph.

[0072] Preferably, step S2 includes the following sub-steps:

[0073] S21. Segment the video material into a plurality of video material segments by using visual technology, and label each video material segment with a corresponding content label and a shooting technique label;

[0074] S22. Generate a content script, a picture description and a shooting technique script for each shot text in the scene segment using content script generation technology, picture description generation technology and shooting technique script generation technology.

[0075] More preferably, in step S21, the video material is segmented into a plurality of video material segments using visual technology, and a corresponding content label and shooting technique label are annotated for each video material segment, including the following sub-steps:

[0076] S211, dividing the video material into a plurality of video material segments by using a transition recognition technology;

[0077] S212, using background recognition technology, object recognition technology, face recognition technology, empty shot recognition technology, character feature generation technology, and behavior action generation technology to label each video material segment in S211 with corresponding content tags, including background, object, person appearing in the shot, empty shot, character feature, and behavior action;

[0078] S213, using the lens scene recognition technology, the lens angle recognition technology, the shooting method recognition technology, and the camera movement recognition technology to label each video material segment in S211 with a corresponding shooting technique label, including the scene, shooting angle, shooting method, and camera movement.

[0079] More preferably, in step S22, a content script generation technology, a picture description generation technology and a shooting technique script generation technology are used to generate a content script, a picture description and a shooting technique script for each shot text in the scene paragraph, including the following sub-steps:

[0080] S221, using a tag selection technique to select a set of characters, backgrounds, and objects tags related to the scene segment from the content tags in S21;

[0081] S222, using the scoring module in combination with the relevant character, background and object label sets selected in S221 to recall the top K candidate content labels from the content labels corresponding to the video material clips in S21;

[0082] S223, using the shot segmentation technology to further segment the scene segments to obtain shot texts;

[0083] S224, using a content script generation technology to select a content tag as a content script for each shot based on the candidate content tags in S222;

[0084] S225, based on the shot text in S223, the shot content script in S224, and the scene paragraph in S1, using the picture description generation technology to generate a corresponding picture description for each shot text;

[0085] S226. Using shooting technique script generation technology, based on the shot text of S223, the content script of S224, the scene segment of S1 and the rules of audio-visual language knowledge, a shooting technique label is generated for each shot text as its shooting technique script.

[0086] Preferably, step S3 includes the following sub-steps:

[0087] S31, for each news manuscript, integrating and storing all the video material clips obtained in S2 to form a video material library;

[0088] S32. For the picture description generated for each shot text in S2, use cross-modal retrieval technology to find the top K candidate video material clips that best match the picture description from the video material library.

[0089] Preferably, step S4 includes the following sub-steps:

[0090] S41, performing internal video splicing for each shot text using candidate video materials to generate a candidate video corresponding to the shot text;

[0091] S42. Execute S41 in sequence according to the order of the shot text in the news manuscript to obtain candidate videos of the shot text, and then perform video splicing of the edited videos of the shot text in the news manuscript to form an edited video of the news manuscript.

[0092] More preferably, in step S41, internal video splicing is performed on each shot text using candidate video materials to generate a candidate video corresponding to the shot text, including the following sub-steps:

[0093] S411, setting the scoring criteria and the video duration required for the shot text; wherein the scoring criteria is a weighted average of the matching degree of the candidate video material clip with the shooting technique script in S2, the matching degree with the shot text, and the matching degree with the rules of audio-visual language knowledge;

[0094] S412, for each shot text, the first K candidate video material segments retrieved from S3 are used as the initial video bundle to start the video splicing process;

[0095] S413, using the scoring criteria in S411 to score each video material segment in the initial video bundle, selecting beam_size video material segments with the highest scores and connecting them to the video bundle, and then entering the next round of search; where beam_size is the maximum number of candidate items;

[0096] S414, during the splicing process, if the total duration of the spliced ​​video bundle exceeds the video duration required by the shot text, the video material segments are cropped to match the video duration requirement of the shot text; if the total duration of the video is not enough to meet the video duration required by the shot text, the cluster search strategy is continued to be used to select and splice the remaining candidate video materials into the video bundle;

[0097] S415. Once the total duration of the assembled video bundle reaches the video duration required by the shot text, the video assembly process of the video bundle is terminated; S413 to S414 are repeated until all video bundles meet the video duration requirements of the shot text; finally, candidate video bundles corresponding to beam_size shot texts are obtained and used as candidate videos of the shot text.

[0098] More preferably, in step S42, S41 is executed in sequence according to the sequence of the shot texts in the news manuscript to obtain candidate videos of the shot texts, and then the video clips of the shot texts in the news manuscript are spliced ​​to form a clipped video of the news manuscript, including the following sub-steps:

[0099] S421, setting a scoring standard, the standard including a weighted average of the matching degree of the candidate video with the shot text, the matching degree with the rules of audio-visual language knowledge, and the coherence with the shots;

[0100] S422, according to the order of the shot texts in the news manuscript, for each shot text, the candidate videos corresponding to the beam_size shot texts obtained in S41 are used as the initial editing video bundle, and the editing video assembly process is started;

[0101] S423, according to the scoring standard of S421, the candidate video of each shot text is combined with the current editing video bundle; the combined editing video bundles are scored, and beam_size editing video bundles with the highest scores are selected;

[0102] S424, marking the used video material segments in each clip video bundle to ensure that the used video material segments are no longer involved in subsequent searches;

[0103] S425. For each shot text, first execute S41 to obtain the candidate video corresponding to the shot text, and then execute S423 to S424 to obtain the edited video of the shot text, until the video assembly of all the shot texts in the news manuscript is completed; finally, beam_size candidate edited video bundles corresponding to the news manuscript are obtained, and the edited video bundle with the highest score is selected as the edited video of the news manuscript.

[0104] Example 2

[0105] This embodiment is based on embodiment 1:

[0106] This embodiment provides a broadcast television news video auxiliary editing system, such as Figure 2 As shown, including:

[0107] A scene recognition module is configured to extract the subject of the news manuscript and recognize the scene paragraphs of the news manuscript based on the subject;

[0108] The shot script generation module is configured to segment the video material into video material segments and annotate corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts;

[0109] A material recall module is configured to select suitable candidate video materials for the segmented shot texts;

[0110] The material assembly module is configured to assemble the candidate video materials into a clipped video of the news manuscript.

[0111] Preferably, the scene recognition module includes a topic extraction unit and a text segmentation unit, wherein the topic extraction unit uses topic extraction technology to identify the topic of the news manuscript, and the text segmentation unit uses text segmentation technology combined with the extracted topic to perform scene paragraph recognition on the news manuscript.

[0112] Preferably, the shot script generation module includes a visual unit and a label selection unit, wherein the visual unit uses visual technology to divide the video material into multiple video material segments, and labels each video material segment with a corresponding content label and shooting technique label; the label selection unit uses content script generation technology, picture description generation technology and shooting technique script generation technology to generate a content script, picture description and shooting technique script for each shot text in the scene segment.

[0113] Preferably, the material recall module includes a cross-modal retrieval unit, which uses cross-modal retrieval technology to search for the top K candidate video material segments that best match the picture description generated by each shot text from the video material library.

[0114] Preferably, the material assembly module includes an audio-visual language knowledge rule unit and a beam search unit, wherein the audio-visual language knowledge rule unit is used to perform internal video assembly for each shot text using candidate video materials to generate candidate videos corresponding to the shot text; the beam search unit is used to obtain candidate videos of the shot text in the order of the shot text in the news manuscript, and perform video assembly of the edited videos of the shot text in the news manuscript to form an edited video of the news manuscript.

[0115] Example 3

[0116] This embodiment is based on embodiment 1:

[0117] This embodiment provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the broadcast television news video auxiliary editing method of embodiment 1 are implemented. The computer program may be in source code form, object code form, executable file, or some intermediate form.

[0118] Example 4

[0119] This embodiment is based on embodiment 1:

[0120] This embodiment provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the broadcast and television news video auxiliary editing method of embodiment 1. The computer program may be in source code form, object code form, executable file, or some intermediate form, etc. The storage medium includes: any entity or device capable of carrying computer program code, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content contained in the storage medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the storage medium does not include electric carrier signals and telecommunication signals.

[0121] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

Claims

1. A method for auxiliary editing of broadcast television news video, characterized in that: include: Scene recognition: extract the subject of the news manuscript and identify the scene paragraphs of the news manuscript based on the subject; Shot script generation: Segment the video material into video material segments and annotate the corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts; Material recall: select suitable candidate video materials for the segmented shot text; Material assembly: assemble candidate video materials into a video clip of the news manuscript; The lens script generation comprises the following steps: S21. Segment the video material into a plurality of video material segments by using visual technology, and label each video material segment with a corresponding content label and a shooting technique label; S22, generating a content script, a picture description and a shooting technique script for each shot text in the scene paragraph using a content script generation technology, a picture description generation technology and a shooting technique script generation technology; In S21, the step of using visual technology to segment the video material into a plurality of video material segments, and labeling each video material segment with a corresponding content label and a shooting technique label, includes the following sub-steps: S211, dividing the video material into a plurality of video material segments by using a transition recognition technology; S212, using background recognition technology, object recognition technology, face recognition technology, empty shot recognition technology, character feature generation technology and behavior action generation technology to annotate corresponding content tags for each video material segment in S211, including background, object, appearing character, empty shot, character feature and behavior action; S213, using the lens scene recognition technology, the lens angle recognition technology, the shooting method recognition technology and the camera movement recognition technology to label each video material segment in S211 with a corresponding shooting technique label, including the scene, shooting angle, shooting method and camera movement; In S22, the process of generating a content script, a picture description and a shooting technique script for each shot text in the scene paragraph by using the content script generation technology, the picture description generation technology and the shooting technique script generation technology includes the following sub-steps: S221, using a tag selection technique to select a set of characters, backgrounds, and objects tags related to the scene segment from the content tags in S21; S222, combining the relevant character, background, and object label sets selected in S221 to recall the top K candidate content labels from the content labels corresponding to the video material clips in S21; S223, using the shot segmentation technology to further segment the scene segments to obtain shot texts; S224, using a content script generation technology to select a content tag as a shot content script for each shot based on the candidate content tags in S222; S225, based on the shot text obtained in S223, the shot content script obtained in S224, and the scene paragraph obtained by the scene recognition, using a picture description generation technology to generate a corresponding picture description for each shot text; S226. Using shooting technique script generation technology, based on the shot text obtained in S223, the shot content script obtained in S224, the scene segments obtained by the scene recognition, and the rules of audio-visual language knowledge, a shooting technique label is generated for each shot text as a shooting technique script.

2. A broadcast and television news video auxiliary editing method according to claim 1, characterized in that: The scene recognition comprises the following steps: S11, extracting the subject of the news manuscript using subject extraction technology; S12. Use text segmentation technology combined with the topics extracted in S11 to identify scene paragraphs in news manuscripts: generate the first sentence of each paragraph based on the segmentation capability of the large language model; use the text similarity algorithm to locate the position of the first sentence of each paragraph in the news manuscript, thereby identifying the scene paragraphs.

3. A broadcast and television news video auxiliary editing method according to claim 1, characterized in that: The material recall includes the following steps: S31, for each news manuscript, integrating and storing all video material clips obtained in the generation of the shot script to form a video material library; S32. For the picture description generated for each shot text in the shot script generation, use cross-modal retrieval technology to find the top K candidate video material clips that best match the picture description from the video material library.

4. A broadcast and television news video auxiliary editing method according to claim 1, characterized in that: The material assembly comprises the following steps: S41, performing internal video splicing for each shot text using candidate video materials to generate a candidate video corresponding to the shot text; S42. Execute S41 in sequence according to the order of the shot text in the news manuscript to obtain candidate videos of the shot text, and then perform video splicing of the edited videos of the shot text in the news manuscript to form an edited video of the news manuscript.

5. A broadcast and television news video auxiliary editing method according to claim 4, characterized in that: In S41, the step of performing internal video splicing on each shot text using candidate video materials to generate a candidate video corresponding to the shot text includes the following sub-steps: S411, setting the video duration required for the scoring criteria and shot text; The scoring criteria include a weighted average of the matching degree of the candidate video material clip with the shooting technique script in the shot script generation, the matching degree with the shot text, and the matching degree with the rules of audio-visual language knowledge; S412, for each shot text, the first K candidate video material segments retrieved from the material recall are used as the initial video bundle, and the video splicing process is started; S413, using the scoring criteria in S411 to score each video material segment in the initial video bundle, selecting beam_size video material segments with the highest scores and connecting them to the video bundle, and then entering the next round of search; where beam_size is the maximum number of candidate items; S414, during the splicing process, if the total duration of the spliced ​​video bundle exceeds the video duration required by the shot text, the video material segments are cropped to match the video duration requirement of the shot text; if the total duration of the video is not enough to meet the video duration required by the shot text, the cluster search strategy is continued to be used to select and splice the remaining candidate video materials into the video bundle; S415. Once the total duration of the assembled video bundle reaches the video duration required by the shot text, the video assembly process of the video bundle is terminated; S413 to S414 are repeated until all video bundles meet the video duration requirements of the shot text; finally, candidate video bundles corresponding to beam_size shot texts are obtained as candidate videos of the shot text.

6. A broadcast and television news video auxiliary editing method according to claim 5, characterized in that: In S42, S41 is executed in sequence according to the sequence of the shot texts in the news manuscript to obtain candidate videos of the shot texts, and then the video editing of the shot texts in the news manuscript is performed to form the edited video of the news manuscript, including the following sub-steps: S421, setting scoring criteria, including a weighted average of the matching degree of the candidate video with the shot text, the matching degree with the rules of audio-visual language knowledge, and the coherence between the shots; S422, according to the order of the shot texts in the news manuscript, for each shot text, the candidate videos corresponding to the beam_size shot texts obtained in S41 are used as the initial editing video bundle, and the editing video assembly process is started; S423, according to the scoring standard of S421, the candidate video of each shot text is combined with the current editing video bundle; the combined editing video bundles are scored, and beam_size editing video bundles with the highest scores are selected; S424, marking the used video material segments in each clip video bundle to ensure that the used video material segments are no longer involved in subsequent searches; S425. For each shot text, first execute S41 to obtain the candidate video corresponding to the shot text, and then execute S423 to S424 to obtain the edited video of the shot text, until the video assembly of all the shot texts in the news manuscript is completed; finally, beam_size candidate edited video bundles corresponding to the news manuscript are obtained, and the edited video bundle with the highest score is selected as the edited video of the news manuscript.

7. A broadcast and television news video auxiliary editing system, based on the broadcast and television news video auxiliary editing method according to claim 1, characterized in that: include: A scene recognition module is configured to extract the subject of the news manuscript and recognize the scene paragraphs of the news manuscript based on the subject; The shot script generation module is configured to segment the video material into video material segments and annotate corresponding content tags and shooting technique tags; segment the identified scene segments into segmented shot texts, and generate content scripts, picture descriptions and shooting technique scripts for the shot texts; A material recall module is configured to select suitable candidate video materials for the segmented shot texts; The material assembly module is configured to assemble the candidate video materials into a clipped video of the news manuscript.

Images