Screen projection document switching identification method, program product, equipment and storage medium

Abstract

The invention provides a screen projection document switching identification method, a program product, electronic equipment and a storage medium. The method comprises the following steps: extracting a video frame from a conference video; obtaining document identification information corresponding to the video frame; and comparing the document identification information of the adjacent video frames, if the document identification information corresponding to the adjacent video frames is inconsistent, extracting a new video frame from the adjacent video frames, obtaining the document identification information corresponding to the new video frame, and continuously comparing the document identification information of the adjacent video frames for the new video frame. Obtaining a document switching adjacent frame until the time interval between two adjacent video frames with inconsistent document identification information is smaller than a preset precision threshold value; and determining the switching time of the screen projection document according to the time of the document switching adjacent frame in the conference video. According to the method, key video frames are extracted and processed in a targeted manner, the extraction range of the video frames is continuously reduced, resource consumption is reduced, and the processing efficiency of determining the switching time of the projection screen document is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of video processing technology, and more specifically, to a method, program product, device and storage medium for identifying switching of projection documents. Background Art

[0002] In the collaborative work of various industries, video conferencing has become a key means of teamwork and communication. Documents can be shared in video conferencing by projection, and one or more documents may be projected in a video conference. In some scenarios, it is necessary to locate a specific projection document from the conference video after the meeting and determine the switching time of the projection document in order to review and record the meeting. The current solution is to perform text area detection on the video frames of the conference video frame by frame, perform OCR (optical character recognition) analysis on the text area, and determine the switching time based on the analysis results. This method requires processing a large number of image frames, consumes a lot of computing resources, and has low processing efficiency. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a projection document switching recognition method, program product, electronic device and storage medium, which are used to improve the problem of low efficiency in projection document switching recognition processing.

[0004] In the first aspect, an embodiment of the present application provides a method for identifying screen projection document switching, including: extracting video frames from a conference video; obtaining document identification information corresponding to the video frames; the document identification information is used to determine the screen projection document to which the video frames belong; comparing the document identification information of adjacent video frames, if the document identification information corresponding to adjacent video frames is inconsistent, extracting new video frames from between adjacent video frames, obtaining the document identification information corresponding to the new video frames, and continuing to compare the document identification information of adjacent video frames for the new video frames until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, and obtaining document switching adjacent frames; determining the switching time of the screen projection document according to the time of the document switching adjacent frames in the conference video.

[0005] In the above implementation process, first, some video frames are extracted from the conference video for processing, and there is no need to process all the video frames of the conference video, so the processing efficiency can be improved. Then, for the extracted video frames, by comparing the document identification information of adjacent video frames, when the document identification information of adjacent video frames is inconsistent, new video frames continue to be extracted from the adjacent video frames, and new adjacent video frames are determined until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby determining the switching time of the document. By comparing the document identification information of adjacent video frames, the scope of video frame extraction is continuously narrowed, and more critical video frames are extracted and processed in a targeted manner. There is no need to perform frame-by-frame text area detection and OCR analysis on a large number of video frames, thereby reducing resource consumption and effectively improving the processing efficiency of determining the switching time of the projection document.

[0006] Optionally, in an embodiment of the present application, the document identification information includes a document URL or an empty identifier; obtaining the document identification information corresponding to the video frame includes: performing target detection on the video frame to obtain an address bar area in the video frame, the address bar area being used to display the document URL; performing text recognition on the address bar area to obtain text to be processed; using a preset regular expression template to match the text to be processed to obtain the document URL corresponding to the video frame; if the document URL corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is the empty identifier; the document identification information corresponding to adjacent video frames is inconsistent, including: the document URLs corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are a document URL and an empty identifier, respectively.

[0007] In the above implementation process, in some screen projection scenarios, the projection document includes a document URL. In this case, the document URL can be obtained through video frame target detection processing. In the subsequent process, the switching time of the projection document can be determined by comparing the document URL. Compared with full-text content recognition of video frames, the method of comparing document URLs requires less content to be recognized, so it is more convenient, saves computing power, and improves video frame processing efficiency.

[0008] Optionally, in an embodiment of the present application, the document identification information includes a document identification or an empty identification; the obtaining of the document identification information corresponding to the video frame includes: identifying the document content to be matched in the video frame, matching the document content to be matched with a preset video conferencing document, and using the document identification of the successfully matched target video conferencing document as the document identification corresponding to the video frame; if the document identification corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is the empty identification; the document identification information corresponding to adjacent video frames is inconsistent, including: the document identifications corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are respectively a document identification and an empty identification.

[0009] In the above implementation process, in other screen projection scenarios, the target video conference document can be first determined by the document content to be matched in the video frame, and then the document identifier of the target video conference document is used as the document identifier corresponding to the video frame.

[0010] Optionally, in an embodiment of the present application, the document identification information includes a document URL, a document identifier or an empty identifier; the obtaining of the document identification information corresponding to the video frame includes: when the document URL and the document identifier corresponding to the video frame are not obtained, the document identification information corresponding to the video frame is the empty identifier; the document identification information corresponding to adjacent video frames is inconsistent, including: the document URLs corresponding to adjacent video frames are different; or, the document identifiers corresponding to adjacent video frames are different; or, the document URLs and document identifiers corresponding to adjacent video frames are not information of the same document; or, the document identification information corresponding to adjacent video frames are respectively a document URL and an empty identifier; or, the document identification information corresponding to adjacent video frames are respectively a document identifier and an empty identifier.

[0011] In the above implementation process, considering that there may be some screens in the video conference that are not document projection, the document URL and document identifier cannot be obtained for these video frames, so the document identifier information is determined as an empty identifier. By comparing the document identifier information, the switching time from other screens to the projection document, or from the projection document to other screens can be determined, which is closer to the actual video conference scene and improves the accuracy of determining the switching time. Supports multiple projection methods, and can accurately determine the document switching time for multiple projection methods.

[0012] Optionally, in an embodiment of the present application, the document identification information of adjacent video frames is compared. If the document identification information corresponding to the adjacent video frames is inconsistent, a new video frame is extracted from between the adjacent video frames to obtain the document identification information corresponding to the new video frame, and the document identification information of the adjacent video frames is continued to be compared with the new video frame until the time interval between two adjacent video frames whose document identification information is inconsistent is less than a preset accuracy threshold, and a document switching adjacent frame is obtained, including: if the document identification information corresponding to the adjacent video frames is inconsistent, an intermediate frame between the adjacent video frames is determined; the adjacent video frames include a first video frame and a second video frame; the intermediate frame is identified to obtain the document identification information of the intermediate frame; if the document identification information of the intermediate frame is consistent with the document identification information of the first video frame, the first frame is updated according to the intermediate frame. a video frame, taking the updated first video frame and the second video frame as adjacent video frames, the updated first video frame being located between the middle frame and the second video frame; and jumping to the step of "if the document identification information corresponding to the adjacent video frames is inconsistent, then determining the middle frame between the adjacent video frames"; if the document identification information of the middle frame is consistent with the document identification information of the second video frame, then updating the second video frame according to the middle frame, taking the first video frame and the updated second video frame as adjacent video frames, the updated second video frame being located between the middle frame and the first video frame; and jumping to the step of "if the document identification information corresponding to the adjacent video frames is inconsistent, then determining the middle frame between the adjacent video frames"; until the time interval between the adjacent video frames is less than the preset accuracy threshold, the document switching adjacent frame is obtained.

[0013] In the above implementation process, for adjacent video frames with inconsistent document identification information, the middle frames are extracted from the conference video and the document identification information of the middle frames is obtained. The middle frames between adjacent video frames are used to gradually narrow the scope of video frame extraction, which effectively reduces the number of processed video frames and improves processing efficiency; and the time interval between the latest adjacent videos obtained is less than the preset accuracy threshold, so that the accuracy of the determined switching time better meets the requirements.

[0014] Optionally, in an embodiment of the present application, extracting video frames from a conference video includes: determining a frame extraction frequency based on the time when a screen projection document is projected in a historical conference video; and extracting the video frames from the conference video according to the frame extraction frequency.

[0015] In the above implementation process, when extracting video frames from the conference video, if the extraction time interval is too small, more video frames will be extracted and more video frames need to be processed, resulting in reduced processing efficiency; if the extraction time interval is too large, some screen projection documents with shorter projection times may be directly skipped, resulting in reduced accuracy in identifying switching time. The embodiment of the present application determines a more appropriate frame extraction frequency based on the projection time of the projection document, taking into account both recognition accuracy and processing efficiency.

[0016] Optionally, in an embodiment of the present application, after determining the switching time of the projection document according to the time when the document switches adjacent frames in the conference video, the method also includes: generating a document link according to the switching time of the conference video and the projection document; the document link is used to open the conference video and jump to the switching time of the projection document, and play the video clip corresponding to the projection document.

[0017] In the above implementation process, after determining the switching time of the projection document, a document link corresponding to the projection document is generated. By clicking the link, the conference video can be opened and jumped to the switching time of the projection document (start projection time), and the video clip corresponding to the projection document can be played. This greatly improves the user's operational efficiency in locating the projection document in the conference video and improves the user experience.

[0018] In the second aspect, an embodiment of the present application also provides a projection document switching identification device, including: a video frame extraction module, used to extract video frames from a conference video; an identification information acquisition module, used to obtain document identification information corresponding to the video frame; the document identification information is used to determine the projection document to which the video frame belongs; an iterative execution module, used to compare the document identification information of adjacent video frames, if the document identification information corresponding to the adjacent video frames is inconsistent, then extract a new video frame from between the adjacent video frames, obtain the document identification information corresponding to the new video frame, and continue to compare the document identification information of the adjacent video frames for the new video frame until the time interval between two adjacent video frames whose document identification information is inconsistent is less than a preset accuracy threshold, and the document switching adjacent frame is obtained; a determination module, used to determine the switching time of the projection document according to the time of the document switching adjacent frame in the conference video.

[0019] In a third aspect, an embodiment of the present application further provides a computer program product, including computer program instructions, which, when executed by a processor, execute the method provided by the first aspect or any one of the implementations of the first aspect.

[0020] In a fourth aspect, an embodiment of the present application further provides an electronic device, comprising: a processor and a memory, the memory storing computer program instructions, and the computer program instructions, when executed by the processor, execute the method provided by the first aspect or any one of the implementations of the first aspect.

[0021] In a fifth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method provided by the first aspect or any one of the implementations of the first aspect is executed.

[0022] The present application provides a projection document switching identification method, program product, electronic device and storage medium. By comparing the document identification information of adjacent video frames, when the document identification information of adjacent video frames is inconsistent, new video frames are continuously extracted from the adjacent video frames to determine new adjacent video frames until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby determining the document switching time. By comparing the document identification information of adjacent video frames, the scope of video frame extraction is continuously narrowed, and more critical video frames are extracted and processed in a targeted manner. There is no need to perform frame-by-frame text area detection and OCR analysis on a large number of video frames, thereby reducing resource consumption and effectively improving the processing efficiency of determining the projection document switching time. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1 A flowchart of a method for switching and identifying screen projection documents provided in an embodiment of the present application;

[0025] Figure 2 A schematic diagram of the structure of a projection document switching recognition device provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following embodiments of the technical solution of the present application are described in detail in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present application.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0029] In the description of the embodiments of the present application, the technical terms "first", "second", etc. are only used to distinguish different objects, and cannot be understood as indicating or implying relative importance or implicitly indicating the number, specific order or primary and secondary relationship of the indicated technical features. In the description of the embodiments of the present application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0030] Documents can be shared during a video conference by projecting the screen. If you need to review or summarize the meeting after the meeting, you can save the meeting video by recording the screen during the meeting, and then analyze or search the meeting video to determine the switching time of the projected document during the meeting. The switching time includes the start time and / or end time of the projected document.

[0031] The current solution for determining the switching time of the projected document is to perform text area detection on the video frames of the conference video frame by frame to obtain the text area in the video frame, then perform OCR analysis on the text area to determine the text content of each frame, and then determine which document the text of each frame belongs to based on the text content. For example, frames 1-N are document A, and frames N+1-M are document B. Then the time between frames N to N+1 is determined as the switching time for document A to switch to document B. This switching time can also be regarded as the end projection time of document A, or the start projection time of document B.

[0032] The above processing method requires processing a large number of video frames, including the use of convolutional neural networks (CNN) for feature extraction and text area detection, and the use of OCR technology for text recognition, which requires a lot of computing resources. In addition, to achieve high-precision document recognition, a large amount of labeled data is required for model training, which is a huge workload, complex models, and low processing efficiency.

[0033] The embodiment of the present application provides a method for identifying switching of projected documents, which extracts video frames from conference videos, and compares the document identification information of adjacent video frames. When the document identification information of adjacent video frames is inconsistent, new video frames are continuously extracted from the adjacent video frames to determine new adjacent video frames, until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby determining the switching time of the document. By comparing the document identification information of adjacent video frames, the scope of video frame extraction is continuously narrowed, and more critical video frames are extracted and processed in a targeted manner, without the need to perform frame-by-frame text area detection and OCR analysis on a large number of video frames, thereby reducing resource consumption and effectively improving processing efficiency.

[0034] See also Figure 1 A flowchart of a screen projection document switching identification method provided by an embodiment of the present application is shown. The screen projection document switching identification method provided by an embodiment of the present application can be applied to an electronic device, which may include a physical device such as a server, a PC, a tablet computer or a smart phone, or a virtual device such as a virtual machine or a container. The electronic device may be a single device, or a combination of multiple devices or a cluster of a large number of devices. The screen projection document switching identification method may include:

[0035] Step S110: extracting video frames from the conference video.

[0036] Step S120: Obtain document identification information corresponding to the video frame; the document identification information is used to determine the projection document to which the video frame belongs.

[0037] Step S130: Compare the document identification information of adjacent video frames. If the document identification information corresponding to adjacent video frames is inconsistent, extract a new video frame from between the adjacent video frames, obtain the document identification information corresponding to the new video frame, and continue to compare the document identification information of adjacent video frames for the new video frame until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby obtaining document switching adjacent frames.

[0038] Step S140: Determine the switching time of the projection document according to the time when the document switches adjacent frames in the conference video.

[0039] In step S110, video conferencing is a conference method for real-time video, document sharing and communication through a computer network. Conference video is a video saved by cloud recording or local recording during the video conference, and the conference video includes at least one screen projection document. Video frames can be extracted from the conference video randomly or at a certain extraction frequency, for example, one frame can be selected every one minute or every two minutes. It should be noted that the video frames here do not include a method of extracting every frame of the conference video.

[0040] In step S120, the conference video includes multiple screen projection methods, such as desktop projection, document projection, or mixed projection. Among them, desktop projection refers to the projection screen including the document URL; document projection refers to a projection method that displays a specific document file (such as word, PPT or Excel, etc.) separately on the conference display screen; mixed projection refers to a flexible projection method that combines multiple projection methods such as desktop projection and document projection. For example, in the same video conference, the first participant uses desktop projection to display document 1, and the second participant uses document projection to display document 2. For the entire conference video, this method can be called mixed projection.

[0041] For different screen projection methods, the video frame has corresponding document identification information, and the document identification information is an identifier that can determine a unique document. For example, for desktop projection, the video frame includes a document URL in the picture, and the document URL of the video frame can be obtained, and the document URL can be used as the document identification information of the video frame; wherein, the document URL can be a website address used to point to a web document, and the document URL can also be a document archive path used to point to a specific document file. For document projection, since the conference display only displays the content of the projection document and does not include the document URL, the document identification of the projection document can be obtained through the content of the projection document, and the document identification can be used as the document identification information. The document identification can be, for example, the document number or the unique ID of the document. The above document URL or document identification can determine the unique projection document to which the video frame belongs. The method of obtaining the document URL and document identification of the video frame will be described in detail later.

[0042] During a meeting, the projected screen may not always be the projected document. For example, when the meeting has just started recording and the document has not been opened, or when the projection is exited midway through the meeting, the screen in the meeting video may be, for example, a desktop pattern. For this type of video frame, the document URL and document ID cannot be identified. In this case, the document ID information can be set to an empty ID, indicating that the projected document to which the video frame belongs is empty, that is, there is no document used for the meeting in the video frame.

[0043] In step S130, for the extracted video frames, the document identification information of every two adjacent video frames is compared. For example, 5 video frames are extracted in total, the document identification information of the first video frame is compared with the second video frame, the document identification information of the second video frame is compared with the third video frame, ..., the document identification information of the fourth video frame is compared with the fifth video frame.

[0044] If the document identification information is consistent, it means that adjacent video frames belong to the same document. If the document identification information is inconsistent, it means that adjacent video frames do not belong to the same document. Depending on the screen projection method, there are many situations in which the document identification information is inconsistent. For example, the document URLs corresponding to adjacent video frames are different, the document identifications corresponding to adjacent video frames are different, the document URLs and document identifications corresponding to adjacent video frames are not the information of the same document, the document identification information corresponding to adjacent video frames are the document URL and the empty identification respectively, or the document identification information corresponding to adjacent video frames are the document identification and the empty identification respectively, etc. Each situation will be introduced later in combination with the screen projection method.

[0045] It should be noted that adjacent video frames have different document identification information types, which does not mean that their document identification information is necessarily inconsistent. For example, in adjacent video frames, the document identification information of one video frame is the document URL, and the document identification information of another video frame is the document identification, but the document URL and the document identification are information of the same document. In this case, even if the document identification information types are different, the document identification information of the adjacent video frames is consistent.

[0046] If it is determined through comparison that the document identification information corresponding to adjacent video frames is inconsistent, a new video frame is extracted from the adjacent video frames. The new video frame can be one frame or multiple frames. For example, if it is determined that the document identification information of the first video frame is inconsistent with that of the second video frame, one or more video frames are randomly extracted from the first video frame and the second video frame or according to a certain extraction rule as new video frames. The extraction rule may be, for example, extracting a frame at 1 / 2 of the time or 1 / 2 of the frame number between the first video frame and the second video frame, or extracting a frame at 1 / n of the time or 1 / n of the frame number between the first video frame and the second video frame, or extracting frame by frame, etc.

[0047] Then, the document identification information corresponding to the new video frame is obtained, and the document identification information of the adjacent video frames is continuously compared with the new video frame until the time interval between the two adjacent video frames whose document identification information is inconsistent is less than the preset accuracy threshold, and the document switching adjacent frames are obtained. Continuing with the above embodiment, if the new video frame extracted between the first video frame and the second video frame includes three frames: video frame A, video frame B and video frame C, then the first video frame, video frame A, video frame B, video frame C and the second video frame are recomposed into a video frame sequence, and for the video frame sequence, the step of "comparing the document identification information of the adjacent video frames, if the document identification information corresponding to the adjacent video frames is inconsistent, then extracting a new video frame from between the adjacent video frames, obtaining the document identification information corresponding to the new video frame, and continuing to compare the document identification information of the adjacent video frames with respect to the new video frame" is iteratively executed, that is, the first video frame is compared with video frame A, video frame A is compared with video frame B, video frame B is compared with video frame C, and so on. It is understandable that if it is found during the comparison process that the document identification information of adjacent video frames is inconsistent, there is no need to compare subsequent adjacent video frames. Of course, it is also possible to choose to complete the comparison of all the frames.

[0048] Iterative execution refers to repeating the above steps, and for the adjacent video frames whose document identification information is inconsistent, continuing to execute the step of extracting new videos. For example, if the document identification information of video frame A and video frame B is inconsistent, then continue to extract new video frames between video frame A and video frame B until the time interval between the two adjacent video frames whose document identification information is inconsistent is less than a preset accuracy threshold, and obtain document switching adjacent frames. For example, the video frames a and video frames b extracted between video frame A and video frame B are two adjacent video frames whose document identification information is inconsistent, and the time interval between video frame a and video frame b is less than the preset accuracy threshold, then the above video frames a and video frame b are used as document switching adjacent frames.

[0049] In an embodiment of the present application, if the document identification information of the first video frame is inconsistent with that of the second video frame, it is also possible to extract only one video frame between the first video frame and the second video frame as a new video frame. Then, during the iterative execution process, the new video frame needs to be compared with the first video frame and the second video frame for document identification information respectively. This scheme will be introduced later.

[0050] In step S140, the time when the document switches adjacent frames in the conference video is first determined, for example, it can be determined based on the video frame number or timestamp of the video frame. Thereafter, the time when the document switches to any one of the adjacent frames can be used as the switching time, and any moment in the time interval when the document switches adjacent frames can also be used as the switching time. It can be understood that video frame a and video frame b are used as adjacent frames for document switching, and the determined switching time refers to the time when the projection document corresponding to video frame a is switched to the projection document corresponding to video frame b.

[0051] In the above implementation process, first, some video frames are extracted from the conference video for processing, and there is no need to process all the video frames of the conference video, so the processing efficiency can be improved. Then, for the extracted video frames, by comparing the document identification information of adjacent video frames, when the document identification information of adjacent video frames is inconsistent, new video frames continue to be extracted from the adjacent video frames, and new adjacent video frames are determined until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby determining the switching time of the document. By comparing the document identification information of adjacent video frames, the scope of video frame extraction is continuously narrowed, and more critical video frames are extracted and processed in a targeted manner. There is no need to perform frame-by-frame text area detection and OCR analysis on a large number of video frames, thereby reducing resource consumption and effectively improving the processing efficiency of determining the switching time of the projection document.

[0052] Optionally, in an embodiment of the present application, in a desktop screen projection scenario, the document identification information includes a document URL or an empty identifier; obtaining the document identification information corresponding to the video frame includes:

[0053] Method for obtaining document URL: (1) Perform target detection on the video frame to obtain the address bar area in the video frame, which is used to display the document URL. A pre-trained target detection model can be used to perform target detection on the video frame. The target detection model generation process includes, for example, collecting video frame screenshots or images including the address bar, and using annotation tools to annotate the video frame screenshots or images, such as annotating the coordinates and area of ​​the address bar using a bounding box. Based on a deep learning framework (such as PyTorch or Keras, etc.), select a suitable target detection algorithm (such as YOLOV8 or SSD, etc.), use annotated data sets to train the target detection model, monitor the loss and accuracy during the training process, and adjust hyperparameters such as the learning rate in a timely manner.

[0054] As an implementation method, before performing target detection on a video frame, the input image may be preprocessed, including operations such as image scaling, normalization, de-averaging, and standardization. The preprocessed video frame is then input into a trained target detection model to obtain target detection model output information, such as address bar coordinates, which can be used to determine the address bar area in the video frame.

[0055] (2) Perform text recognition on the address bar area to obtain the text to be processed. First, the video frame can be cropped according to the above-mentioned address bar coordinates to obtain the address bar area, where the address bar area refers to the image corresponding to the address bar in the video frame. Before performing OCR text recognition on the address bar area, image preprocessing can also be performed, such as grayscale, binarization, etc. Grayscale refers to converting a color image into a grayscale image, and binarization refers to converting a grayscale image into a binary image. Of course, this preprocessing process can also be performed before cropping the video frame. The recognized text is output by the OCR engine, i.e., the text to be processed.

[0056] (3) Use a preset regular expression template to match the text to be processed and obtain the document URL corresponding to the video frame. The preset regular expression template may include content such as a protocol, a domain name, or an optional path. For example, the preset regular expression template may be in the form of url_pattern = re.compile(r'https?: / / [^\s<>"]+|www\.[^\s<>"]+'); the preset regular expression template may also include content such as a fixed format of a file path, which depends on the type of the document URL. The content that matches the preset regular expression template in the text to be recognized is identified as the document URL corresponding to the video frame. This step can remove characters other than the document URL contained in the text to be processed to a certain extent.

[0057] In the desktop projection scenario, if the document URL corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is an empty identifier. For example, when a video frame is just beginning to be recorded and the document has not yet been opened, the confidence information of the address bar output by the target detection model is lower than the preset confidence threshold, and it can be determined that the image frame may not have an address bar area; or if the document URL corresponding to the video frame cannot be obtained through text recognition, the document identification information of such video frames is determined to be an empty identifier.

[0058] In the desktop projection scenario, the document identification information corresponding to adjacent video frames is inconsistent, including: the document URLs corresponding to adjacent video frames are different (indicating that the two document URLs point to different documents respectively); or, the document identification information corresponding to adjacent video frames are the document URL and an empty identifier respectively (indicating that one of the adjacent video frames is the projection screen of the document, and the other is not the projection screen of the document).

[0059] During the implementation of the above embodiment: In some screen projection scenarios, the projection document includes a document URL, so the document URL can be obtained by detecting the video frame target. In the subsequent process, the switching time of the projection document can be determined by comparing the document URL. Compared with full-text content recognition of video frames, the method of comparing document URLs requires less content to be recognized, so it is more convenient, saves computing power, and improves video frame processing efficiency.

[0060] Optionally, in an embodiment of the present application, in a scenario of document projection, the document identification information includes a document identification or an empty identification; obtaining the document identification information corresponding to the video frame includes:

[0061] Methods for obtaining document identification: (1) Identify the content of the document to be matched in the video frame. For example, the OCR technology can be used to identify the pre-processed video frame to obtain the content of the document to be matched. Deep learning models such as convolutional neural networks (CNNs) can also be used to automatically learn the feature representation of characters and improve the recognition accuracy and generalization ability of OCR.

[0062] (2) Matching the content of the document to be matched with the preset video conference document, and using the document identifier of the successfully matched target video conference document as the document identifier corresponding to the video frame.

[0063] Among them, the preset video conference document includes all documents used for screen projection in the conference video. It can be directly obtained through the preset document path, or it can be directly requested by calling the interface to obtain the document URL, and then the preset video conference document can be obtained using the document URL. For example, when preparing for a meeting, the document URL used for the meeting can be filled in the conference software, and the interface of the conference software can be called to obtain the document URL. It can be understood that in the scenario of document projection, since the screen in the conference video does not have a document URL, the document URL cannot be directly used to determine the switching time. The document URL here is used to obtain the preset video conference document.

[0064] When directly requesting to obtain the document URL by calling the interface, the document URL source can be divided into three categories. The first category is that the required document link already exists in the RDC (remote data center) scenario. At this time, if the entered document is not projected, it is considered not entered, and the unentered document will not be used to match the content of the document to be matched. The second category is to enter the projected document URL in the meeting minutes. The third category is to enter the projected document URL in other entrances.

[0065] During the matching process, if there are multiple preset video conferencing documents, a similarity algorithm can be used to match the contents of the document to be matched with the preset video conferencing documents respectively, and the video conferencing document with the highest similarity and a similarity greater than a preset threshold is determined as the target video conferencing document that has been successfully matched, and then the document identifier of the target video conferencing document is used as the document identifier of the video frame.

[0066] If the document identifier corresponding to the video frame is not obtained, the document identifier information corresponding to the video frame is an empty identifier. For example, during the matching process, if the similarity data between the content of the document to be matched and each preset video conference document is lower than the threshold, it is possible that the content of the document to be matched is not the projection document in the conference video, but other screen content, then the document identifier information corresponding to the video frame is set to an empty identifier.

[0067] In the document projection scenario, the document identification information corresponding to adjacent video frames is inconsistent, including: the document identification information corresponding to adjacent video frames is different (indicating that the two document identifications correspond to different documents respectively); or the document identification information corresponding to adjacent video frames are a document identification and an empty identification respectively (indicating that one of the adjacent video frames is a document projection screen, and the other is not a document projection screen).

[0068] In the implementation process of the above embodiment: In some other screen projection scenarios, the target video conference document can be determined by the content of the document to be matched in the video frame, and then the document identifier of the target video conference document is used as the document identifier corresponding to the video frame. Multiple screen projection modes are supported, and the switching time of the document can be accurately determined for multiple screen projection modes.

[0069] Optionally, in an embodiment of the present application, in hybrid screen projection, the document identification information includes a document URL, a document identifier or an empty identifier; obtaining the document identification information corresponding to the video frame includes: when the document URL and document identifier corresponding to the video frame are not obtained, the document identification information corresponding to the video frame is an empty identifier. For example, when the above-mentioned method for obtaining the document URL is used for the video frame, the document URL cannot be obtained, and when the above-mentioned method for obtaining the document identifier is used, the document identifier cannot be obtained, then the document identification information corresponding to the video frame is an empty identifier.

[0070] In an optional embodiment, if it is uncertain whether the conference video is a desktop screen projection or a document screen projection, the document URL of the video frame can be obtained by using the method of obtaining the document URL. If the document URL is not obtained, the document identifier of the video frame can be obtained by using the method of obtaining the document identifier. If the document identifier is not obtained, it can be determined that the document identifier information of the video frame is an empty identifier. Of course, the method of obtaining the document identifier can also be used first, and then the method of obtaining the document URL can be used. This application does not limit this order.

[0071] Cases where the document identification information corresponding to adjacent video frames is inconsistent include: the document URLs corresponding to adjacent video frames are different, the document identifications corresponding to adjacent video frames are different, the document URLs and document identifications corresponding to adjacent video frames are not information of the same document, the document identification information corresponding to adjacent video frames are respectively a document URL and an empty identification, or the document identification information corresponding to adjacent video frames are respectively a document identification and an empty identification.

[0072] In the hybrid screen projection scenario, the inconsistent document identification information corresponding to adjacent video frames not only includes the four cases of document projection and desktop projection, but also the newly added case that the document URL and document identification corresponding to adjacent video frames are not the information of the same document. Because in the hybrid screen projection scenario, the document identification information of some video frames may include the document URL, and some document identification information is the document identification, so it is also necessary to determine whether the document identification information and the document URL are the information of the same document. If they are not the information of the same document, it is determined that the document identification information corresponding to adjacent video frames is inconsistent.

[0073] In the implementation process of the above embodiment: Considering that there may be some screens that are not document projection in the video conference, the document URL and document identification cannot be obtained for these video frames, so the document identification information is determined as an empty identification. By comparing the document identification information, the switching time from other screens to the projection document, or the switching time from the projection document to other screens can be determined, which is closer to the actual video conference scene and improves the accuracy of determining the switching time. Supports multiple projection methods, and the document switching time can be accurately determined for multiple projection methods.

[0074] Optionally, in the embodiment of the present application, step S130 may include the following steps:

[0075] If the document identification information corresponding to adjacent video frames is inconsistent, then the middle frame between the adjacent video frames is determined; the adjacent video frames include the first video frame and the second video frame. For example, in any screen projection scenario (desktop projection, document projection or mixed projection), if it is determined that the document identification information corresponding to adjacent video frames is inconsistent, then the middle frame between the adjacent video frames is determined. Among them, the middle frame refers to the video frame at halfway between adjacent videos. For example, the middle frame can be determined based on the video frame serial number or timestamp of the first video frame and the second video frame.

[0076] After the intermediate frame is determined, the intermediate frame is identified to obtain the document identification information of the intermediate frame. The document identification information of the intermediate frame can be a document URL, a document identifier or an empty identifier. The method for obtaining the document identification information of the intermediate frame is as described above.

[0077] If the document identification information of the middle frame is consistent with the document identification information of the first video frame, the first video frame is updated according to the middle frame, and the updated first video frame and the second video frame are used as adjacent video frames, and the updated first video frame is located between the middle frame and the second video frame; and jump to the step of "if the document identification information corresponding to the adjacent video frames is inconsistent, then determine the middle frame between the adjacent video frames". The first video frame is updated according to the middle frame, for example, a frame between the middle frame and the second video frame is used as the updated first video frame. Exemplarily, the next frame after the middle frame can be used as the updated first video frame. Thereafter, a (new) middle frame is extracted from between the updated first video frame and the second video frame, and the (new) middle frame is compared with the updated first video frame and the second video frame for file identification information respectively.

[0078] If the document identification information of the middle frame is consistent with the document identification information of the second video frame, the second video frame is updated according to the middle frame, and the first video frame and the updated second video frame are used as adjacent video frames, and the updated first video frame is located between the middle frame and the second video frame; and jump to the step of "if the document identification information corresponding to adjacent video frames is inconsistent, then determine the middle frame between adjacent video frames". The second video frame is updated according to the middle frame, for example, a frame between the middle frame and the second video frame is used as the updated first video frame. Exemplarily, the previous frame of the middle frame can be used as the updated second video frame. Thereafter, a (new) middle frame is extracted between the first video frame and the updated second video frame, and the (new) middle frame is compared with the first video frame and the updated second video frame for file identification information respectively.

[0079] Among them, the document identification information of the middle frame is consistent with the document identification information of the first video frame, including the following situations: first, the document URL of the middle frame is the same as the document URL of the first video frame; second, the document identification of the middle frame is the same as the document identification of the first video frame; third, the document URL of the middle frame and the document identification of the first video frame represent the same document; fourth, the document identification of the middle frame and the document URL of the first video frame represent the same document.

[0080] Repeat the above steps until the time interval between adjacent video frames is less than the preset accuracy threshold, and use the adjacent video frames as document switching adjacent frames. The preset accuracy threshold is, for example, a unit time such as seconds, milliseconds, microseconds, etc., and of course, it can also be 2 seconds or 5 milliseconds. For example, if the preset accuracy threshold is set to 2 seconds, then the switching time determined for document switching adjacent frames with a time interval less than the preset accuracy threshold of less than 2 seconds is acceptable. Through the above steps, it is determined that the time of two adjacent video frames with inconsistent document identification information in the conference video is 30 minutes and 1 second and 30 minutes and 2 seconds respectively, and the time interval is less than 2 seconds, then these two video frames are determined as document switching adjacent frames. Afterwards, the switching time of the projection document is determined according to the time of the document switching adjacent frames in the conference video. For example, 30 minutes and 1 second or 30 minutes and 2 seconds can be used as the switching time for the projection document of the first video frame to switch to the projection document of the second video frame, and any time between 30 minutes and 1 second and 30 minutes and 2 seconds can also be used as the switching time of the projection document.

[0081] During the implementation of the above embodiment: for adjacent video frames with inconsistent document identification information, the middle frame is extracted from the conference video and the document identification information of the middle frame is obtained, and the range of video frame extraction is gradually narrowed by using the middle frame between adjacent video frames, thereby effectively reducing the number of processed video frames and improving processing efficiency; and the time interval between adjacent videos used to determine the switching time is less than a preset accuracy threshold, so that the accuracy of the determined switching time better meets the requirements.

[0082] Optionally, in the embodiment of the present application, extracting video frames from the conference video includes:

[0083] The frame extraction frequency is determined based on the time when the screen projection document was projected in the historical conference video. The historical conference video can be a conference video of the same type as the conference video. For example, the conference video is a report video. The frame extraction frequency can be determined based on the time when the screen projection document was projected in the historical conference video of the report. It is found through the historical conference videos of the report that the screen projection time of each document is about 10 minutes. The frame extraction frequency can be determined to be half of this time, that is, 5 minutes / time. Of course, it can also be higher or lower. In principle, it is necessary to ensure that at least one video frame is extracted for each screen projection document.

[0084] If there is no obvious regularity in the projection time of the projection documents in the historical conference videos, the frame extraction frequency can be determined based on the average projection time or the shortest projection time of each projection document in the historical conference videos. For example, the frame extraction frequency can be the average projection time multiplied by a proportional coefficient less than 1. After the frame extraction frequency is determined, video frames are extracted from the conference video according to the frame extraction frequency.

[0085] In an optional embodiment, the average projection duration of each projection document in the conference video can be determined based on the total duration of the conference video itself and the number of projection documents in the conference video, and the frame extraction frequency can be determined based on the average projection duration. This method does not require the use of historical conference videos and can also obtain a more suitable frame extraction frequency.

[0086] In the implementation process of the above embodiment: when extracting video frames from the conference video, if the extraction time interval is too small, more video frames will be extracted and more video frames need to be processed, resulting in low processing efficiency; if the extraction time interval is too large, some screen projection documents with shorter projection time may be directly skipped, resulting in reduced accuracy in identifying switching time. The embodiment of the present application determines a more appropriate frame extraction frequency based on the projection time of the projection document, effectively improving processing efficiency without reducing recognition accuracy.

[0087] Optionally, in an embodiment of the present application, after determining the switching time of the projected document based on the time when the document switches adjacent frames in the conference video, the method also includes: generating a document link based on the switching time of the conference video and the projected document; the document link is used to open the conference video and jump to the switching time of the projected document, and play the video clip corresponding to the projected document.

[0088] The process of generating a document link is as follows: Get the original link of the conference video, which points to the location of the conference video file on the server. Generate a document link based on the original link of the conference video and the switching time of the projection document. For example, the switching time of the projection document can be added to the original link of the conference video by adding a query parameter after the original link of the conference video.

[0089] During the implementation of the above embodiment: after determining the switching time of the projection document, a document link corresponding to the projection document is generated. By clicking the link, the conference video can be opened and jumped to the switching time of the projection document (start projection time), and the video clip corresponding to the projection document can be played. This greatly improves the user's operational efficiency in locating the projection document in the conference video and improves the user experience.

[0090] See also Figure 2 The structure diagram of the screen projection document switching identification device provided in the embodiment of the present application is shown; the embodiment of the present application provides a screen projection document switching identification device 200, including:

[0091] The video frame extraction module 210 is used to extract video frames from the conference video;

[0092] The identification information acquisition module 220 is used to obtain the document identification information corresponding to the video frame; the document identification information is used to determine the screen projection document to which the video frame belongs;

[0093] The iterative execution module 230 is used to compare the document identification information of the adjacent video frames. If the document identification information corresponding to the adjacent video frames is inconsistent, a new video frame is extracted from between the adjacent video frames to obtain the document identification information corresponding to the new video frame, and the document identification information of the adjacent video frames is continuously compared with the new video frame until the time interval between two adjacent video frames whose document identification information is inconsistent is less than a preset accuracy threshold, thereby obtaining a document switching adjacent frame.

[0094] The determination module 240 is used to determine the switching time of the projection document according to the time when the adjacent frames of the document are switched in the conference video.

[0095] Optionally, in an embodiment of the present application, the projection document switching identification device 200, the document identification information includes a document URL or an empty identifier; the identification information acquisition module 220 is specifically used to perform target detection on the video frame to obtain the address bar area in the video frame, and the address bar area is used to display the document URL; perform text recognition on the address bar area to obtain the text to be processed; use a preset regular expression template to match the text to be processed to obtain the document URL corresponding to the video frame; if the document URL corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is the empty identifier; the document identification information corresponding to adjacent video frames is inconsistent, including: the document URLs corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are document URLs and empty identifiers, respectively.

[0096] Optionally, in an embodiment of the present application, the projection document switching identification device 200, the document identification information includes a document identification or an empty identification; the identification information acquisition module 220 is specifically used to identify the document content to be matched in the video frame, match the document content to be matched with a preset video conferencing document, and use the document identification of the successfully matched target video conferencing document as the document identification corresponding to the video frame; if the document identification corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is the empty identification; the document identification information corresponding to adjacent video frames is inconsistent, including: the document identifications corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are respectively a document identification and an empty identification.

[0097] Optionally, in an embodiment of the present application, the projection document switching identification device 200, the document identification information includes a document URL, a document identifier or an empty identifier; the identification information acquisition module 220 is specifically used for, when the document URL and document identifier corresponding to the video frame are not obtained, the document identification information corresponding to the video frame is the empty identifier; the document identification information corresponding to adjacent video frames is inconsistent, including: the document URLs corresponding to adjacent video frames are different; or, the document identifiers corresponding to adjacent video frames are different; or, the document URLs and document identifiers corresponding to adjacent video frames are not information of the same document; or, the document identification information corresponding to adjacent video frames are respectively a document URL and an empty identifier; or, the document identification information corresponding to adjacent video frames are respectively a document identifier and an empty identifier.

[0098] Optionally, in an embodiment of the present application, the projection document switching identification device 200 and the iterative execution module 230 are specifically used to determine an intermediate frame between adjacent video frames if the document identification information corresponding to adjacent video frames is inconsistent; the adjacent video frames include a first video frame and a second video frame; identify the intermediate frame to obtain document identification information of the intermediate frame; if the document identification information of the intermediate frame is consistent with the document identification information of the first video frame, update the first video frame according to the intermediate frame, and use the updated first video frame and the second video frame as adjacent video frames, and the updated first video frame is located between the intermediate frame and the second video frame; and jump Go to the step of "if the document identification information corresponding to adjacent video frames is inconsistent, then determine the intermediate frame between adjacent video frames"; if the document identification information of the intermediate frame is consistent with the document identification information of the second video frame, then update the second video frame according to the intermediate frame, and use the first video frame and the updated second video frame as adjacent video frames, and the updated second video frame is located between the intermediate frame and the first video frame; and jump to the step of "if the document identification information corresponding to adjacent video frames is inconsistent, then determine the intermediate frame between adjacent video frames"; until the time interval between adjacent video frames is less than the preset accuracy threshold, the document switching adjacent frames are obtained.

[0099] Optionally, in an embodiment of the present application, the projection document switching identification device 200 and the video extraction module 210 are used to determine the frame extraction frequency according to the projection time of the projection document in the historical conference video; and extract the video frames from the conference video according to the frame extraction frequency.

[0100] Optionally, in an embodiment of the present application, the projection document switching identification device 200 also includes a link generation module, which is used to generate a document link based on the switching time of the conference video and the projection document; the document link is used to open the conference video and jump to the switching time of the projection document, and play the video clip corresponding to the projection document.

[0101] It should be understood that the device corresponds to the above-mentioned projection document switching identification method embodiment and can execute the various steps involved in the above-mentioned method embodiment. The specific functions of the device can be found in the description above. To avoid repetition, the detailed description is appropriately omitted here. The device includes at least one software function module that can be stored in a memory in the form of software or firmware or solidified in the operating system (OS) of the device.

[0102] See also Figure 3 The electronic device 300 provided in the embodiment of the present application includes: a processor 310 and a memory 320, wherein the memory 320 stores machine-readable instructions executable by the processor 310, and when the machine-readable instructions are executed by the processor 310, the above method is executed.

[0103] Figure 3 Each component shown in can be implemented by hardware, software or a combination thereof. The electronic device 300 may be a physical device, such as a server, a PC, etc., or a virtual device, such as a virtual machine, a virtualized container, etc. Moreover, the electronic device 300 is not limited to a single device, but may also be a combination of multiple devices or a cluster consisting of a large number of devices.

[0104] An embodiment of the present application further provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above method is executed.

[0105] Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable red-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0106] In several embodiments provided by the embodiments of the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a part of the module, program segment or code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0107] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist separately, or two or more modules can be integrated to form an independent part.

[0108] The above description is only an optional implementation manner of the embodiments of the present application, but the protection scope of the embodiments of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed in the embodiments of the present application, which should be covered within the protection scope of the embodiments of the present application.

Claims

1. A method for identifying screen projection document switching, characterized in that: include: Extract video frames from conference videos; Obtain document identification information corresponding to the video frame; the document identification information is used to determine the screen projection document to which the video frame belongs; Comparing the document identification information of adjacent video frames, if the document identification information corresponding to the adjacent video frames is inconsistent, extracting a new video frame from between the adjacent video frames, obtaining the document identification information corresponding to the new video frame, and continuing to compare the document identification information of adjacent video frames for the new video frame, until the time interval between two adjacent video frames with inconsistent document identification information is less than a preset accuracy threshold, thereby obtaining a document switching adjacent frame; The switching time of the projection document is determined according to the time when the adjacent frames of the document are switched in the conference video.

2. The method according to claim 1, characterized in that The document identification information includes a document URL or an empty identifier; the obtaining of the document identification information corresponding to the video frame includes: Performing target detection on the video frame to obtain an address bar area in the video frame, where the address bar area is used to display the document URL; performing text recognition on the address bar area to obtain text to be processed; using a preset regular expression template to match the text to be processed to obtain the document URL corresponding to the video frame; In the case where the document URL corresponding to the video frame is not obtained, the document identification information corresponding to the video frame is the empty identification; The document identification information corresponding to adjacent video frames is inconsistent, including: The document URLs corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are the document URL and the empty identification respectively.

3. The method according to claim 1, characterized in that The document identification information includes a document identification or a null identification; and obtaining the document identification information corresponding to the video frame includes: Identify the content of the document to be matched in the video frame, match the content of the document to be matched with a preset video conference document, and use the document identifier of the successfully matched target video conference document as the document identifier corresponding to the video frame; In the case where the document identifier corresponding to the video frame is not obtained, the document identifier information corresponding to the video frame is the empty identifier; The document identification information corresponding to adjacent video frames is inconsistent, including: The document identifiers corresponding to the adjacent video frames are different; or, the document identification information corresponding to the adjacent video frames are a document identifier and a null identifier respectively.

4. The method according to claim 1, characterized in that: The document identification information includes a document URL, a document identifier or an empty identifier; and obtaining the document identification information corresponding to the video frame includes: In the case where the document URL and document identifier corresponding to the video frame are not obtained, the document identifier information corresponding to the video frame is the empty identifier; The document identification information corresponding to adjacent video frames is inconsistent, including: The document URLs corresponding to adjacent video frames are different; Alternatively, the document identifiers corresponding to adjacent video frames are different; Alternatively, the document URL and document identifier corresponding to the adjacent video frames are not information about the same document; Alternatively, the document identification information corresponding to the adjacent video frames are respectively a document URL and an empty identifier; Alternatively, the document identification information corresponding to the adjacent video frames are respectively a document identification and a null identification.

5. The method according to claim 1, characterized in that Comparing the document identification information of adjacent video frames, if the document identification information corresponding to the adjacent video frames is inconsistent, extracting a new video frame from between the adjacent video frames, obtaining the document identification information corresponding to the new video frame, and continuing to compare the document identification information of adjacent video frames for the new video frame until the time interval between two adjacent video frames whose document identification information is inconsistent is less than a preset accuracy threshold, and obtaining a document switching adjacent frame, including: If the document identification information corresponding to the adjacent video frames is inconsistent, determining an intermediate frame between the adjacent video frames; the adjacent video frames include a first video frame and a second video frame; Identify the intermediate frame and obtain document identification information of the intermediate frame; If the document identification information of the intermediate frame is consistent with the document identification information of the first video frame, the first video frame is updated according to the intermediate frame, and the updated first video frame and the second video frame are used as adjacent video frames, and the updated first video frame is located between the intermediate frame and the second video frame; and jump to the step of "if the document identification information corresponding to the adjacent video frames is inconsistent, determine the intermediate frame between the adjacent video frames"; If the document identification information of the intermediate frame is consistent with the document identification information of the second video frame, the second video frame is updated according to the intermediate frame, the first video frame and the updated second video frame are used as adjacent video frames, and the updated second video frame is located between the intermediate frame and the first video frame; and jump to the step of "if the document identification information corresponding to the adjacent video frames is inconsistent, determine the intermediate frame between the adjacent video frames"; Until the time interval between adjacent video frames is less than a preset accuracy threshold, the document switching adjacent frames are obtained.

6. The method according to claim 1, characterized in that The extracting of video frames from the conference video includes: Determine the frame extraction frequency based on the time when the screen document was projected in the historical conference video; The video frames are extracted from the conference video according to the frame extraction frequency.

7. The method according to claim 1, characterized in that After determining the switching time of the projection document according to the time of switching adjacent frames of the document in the conference video, the method further includes: A document link is generated according to the switching time of the conference video and the projection document; the document link is used to open the conference video and jump to the switching time of the projection document, and play the video clip corresponding to the projection document.

8. A computer program product, characterized in that The method comprises computer program instructions, and when the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is executed.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method according to any one of claims 1 to 7 is executed.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is executed.

Images