Online training video recording method and system

Abstract

The invention relates to the technical field of communication, and discloses an online training video recording method and system, and the method comprises the steps: obtaining a recorded video image of a target subject, and adding a timestamp to key contents in the recorded video image; identifying the target main body in each video frame in the recorded video image to obtain a first video picture; extracting features of the first video picture through a preset neural network model, and performing de-noising processing on the recorded video image based on the features to obtain a high-definition recorded video image picture; combining the high-definition recorded video image pictures according to a time sequence to obtain a high-definition recorded video; according to the timestamp, a split-screen node and a corresponding auxiliary picture are inserted into the high-definition recorded video, a video recording result is obtained, and the auxiliary picture comprises auxiliary content corresponding to the key content; the recorded video obtained through the method is high in definition and highlighted in key content, and is convenient for trained personnel to learn.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to an online training video recording method and system. Background Art

[0002] With the continuous development and popularization of Internet technology, online training, as a new learning method, has gradually been recognized and accepted by the majority of users. Usually, the training content is recorded during the training process, so the clarity of the training video recording is directly related to the later review. At present, screen recording software is usually used to record the online training process. Although this method can record the training content completely, it cannot improve the clarity of the recorded video and cannot highlight the key content of the training process. Summary of the invention

[0003] The present invention provides an online training video recording method and system, which solves the problem that the existing video recording method cannot improve the clarity of the video recording picture and cannot highlight the key content of the training process.

[0004] In order to solve the above technical problems, the first aspect of the present invention provides an online training video recording method, comprising:

[0005] Acquire a recorded video image of a target subject, and add a time stamp to key content in the recorded video image;

[0006] Identify the target subject in each video frame of the recorded video image to obtain a first video picture;

[0007] Extracting features of the first video image through a preset neural network model, and performing denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image;

[0008] Combining the high-definition recorded video images in chronological order to obtain a high-definition recorded video;

[0009] A split-screen node and its corresponding sub-picture are inserted into the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-picture includes auxiliary content corresponding to the key content.

[0010] As one of the preferred schemes, the preset neural network model includes a deep residual network layer, a first convolutional neural network layer, a recurrent neural network layer, a second convolutional neural network layer, a third convolutional neural network layer and a fourth convolutional neural network layer.

[0011] As one of the preferred solutions, the features of the first video picture are extracted by a preset neural network model, and the recorded video image is denoised based on the features to obtain a high-definition recorded video image, including:

[0012] Extracting image features of the first video picture through the deep residual network layer, and extracting spatial features of the first video picture through the first convolutional neural network layer;

[0013] Processing the spatial features of the first video picture in the current video frame and the spatial features of the first video picture in the previous video frame through the recurrent neural network layer to obtain the temporal features of the first video picture in the current video frame;

[0014] Extracting state features of the first video picture through the second convolutional neural network layer, and calculating a contribution value of the first video picture according to the state features and the image features;

[0015] The temporal features and the spatial features of the corresponding video frame are fused to obtain spatiotemporal features, and based on the spatiotemporal features, the first video picture is convolved through the third convolutional neural network layer to obtain a convolution matrix;

[0016] The convolution matrix is ​​subjected to residual processing through the fourth convolutional neural network layer to obtain the first video picture after denoising, and the first video picture after denoising is combined according to the contribution value to obtain a high-definition recorded video image picture.

[0017] As one preferred solution, the timestamp includes a start time and an end time; wherein,

[0018] Inserting a split-screen node and its corresponding sub-picture into the high-definition recorded video according to the timestamp to obtain a video recording result includes:

[0019] Obtaining a secondary picture and its corresponding secondary video frame, and setting the time length of the secondary picture by using the timestamp;

[0020] Based on the timestamp, extract a video frame from the high-definition recorded video and use it as a main video frame in the main picture;

[0021] On the frame corresponding to the timestamp, a new video frame is created and the main video frame and the auxiliary video frame are added to obtain a new fused video frame;

[0022] A split screen node is set at the start time, and the new fused video frame is inserted according to the split screen node according to the timestamp to replace the original video frame to obtain a video recording result.

[0023] As one of the preferred solutions, creating a new video frame on the frame corresponding to the timestamp and adding the main video frame and the auxiliary video frame to obtain a new fused video frame includes:

[0024] On the frame corresponding to the timestamp, create a new video frame, and divide the new video frame into regions to obtain a first region and a second region;

[0025] The main video frame is added to the first area, and the auxiliary video frame is added to the second area to obtain a new fused video frame.

[0026] As one of the preferred solutions, after inserting a split screen node and its corresponding sub-picture into the high-definition recorded video according to the timestamp and obtaining the video recording result, the method further includes:

[0027] The video recording result is condensed to obtain a condensed video.

[0028] As one of the preferred solutions, the video recording result is condensed to obtain a condensed video, including:

[0029] Identify each high-definition recorded video frame in the video recording result to obtain a second video picture corresponding to each high-definition recorded video frame;

[0030] Selecting target video pictures that meet preset conditions from the second video pictures to obtain a target picture list;

[0031] The video recording result is cropped according to the target screen list to obtain a plurality of video clips and combine them to obtain a condensed video.

[0032] As one of the preferred solutions, target video pictures meeting preset conditions are selected from the second video pictures to obtain a target picture list, including:

[0033] Calculating the feature similarity of the second video picture, traversing each high-definition recorded video frame, and comparing the feature similarity of the second video picture located in the current high-definition recorded video frame with the feature similarity of the second video picture located in an adjacent high-definition recorded video frame;

[0034] The two second video pictures whose comparison results meet the preset comparison conditions and whose appearance duration reaches the preset duration threshold are taken as the same target video picture, and the target video pictures whose pixels are not less than the preset pixel threshold are combined to obtain a target picture list.

[0035] As one of the preferred solutions, target video pictures meeting preset conditions are selected from the second video pictures to obtain a target picture list, including:

[0036] Extracting color data of the second video picture, and removing background color data from the color data to obtain content color data;

[0037] If the content color data belongs to a pre-constructed color database, and the appearance duration of the second video picture corresponding to the content color data reaches a preset duration threshold, then the video picture is used as the target video picture;

[0038] The target video pictures are combined to obtain a target video picture list.

[0039] A second aspect of the present invention provides an online training video recording system, comprising:

[0040] A video image processing module, used to obtain a recorded video image of a target subject and add a time stamp to key content in the recorded video image;

[0041] A video picture recognition module, used for recognizing the target subject in each video frame in the recorded video image to obtain a first video picture;

[0042] A video image denoising module, configured to extract features of the first video image through a preset neural network model, and perform denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image;

[0043] A first video generation module, used for combining the high-definition recorded video images in chronological order to obtain a high-definition recorded video;

[0044] The second video generation module is used to insert a split-screen node and its corresponding sub-screen into the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-screen includes auxiliary content corresponding to the key content.

[0045] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0046] (1) The present application captures the recorded video image of the target subject and adds timestamps to important content for subsequent processing; uses a preset neural network model to extract features of the target subject in each frame of the video, and performs denoising based on these features to improve the clarity of the video; reassembles the processed high-definition images to generate a high-definition recorded video; according to the previously added timestamps, inserts split-screen nodes and their sub-images at corresponding positions in the high-definition video. These sub-images contain auxiliary information of the key content, thereby enhancing the information content and viewing experience of the video, and improving the quality of the recorded video, thereby generating a high-definition and information-rich video recording result;

[0047] (2) This application adds timestamps to key content, which not only facilitates post-editing and quick positioning, but also provides learners with intuitive reference points for reviewing and sharing specific content; through the feature extraction and denoising processing technology of the neural network model, the noise and interference in the video are effectively removed, and the clarity of the video is significantly improved, so that learners can obtain a more delicate and realistic visual experience; the design of the split-screen node and its sub-screen makes the video content presentation more flexible and diverse, which not only increases the layering of the video, but also greatly enriches the content of the video, and can be customized according to the needs and interests of the learners, thereby improving the fun and interactivity of watching; it can effectively convey knowledge points and details in the fields of education and training, helping learners to better understand and master the content they have learned, thereby improving the educational effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0049] Figure 1 is a flow chart of an online training video recording method provided by a certain embodiment of the present invention;

[0050] Figure 2 is a flowchart of step S3 provided by an embodiment of the present invention;

[0051] Figure 3 is a flowchart of step S5 provided by an embodiment of the present invention;

[0052] Figure 4 is a flowchart of step S53 provided by an embodiment of the present invention;

[0053] Figure 5 is a flowchart of step S6 provided in one embodiment of the present invention;

[0054] Figure 6 is a flowchart of step S62 provided by an embodiment of the present invention;

[0055] Figure 7 is a flowchart of step S62 provided in another embodiment of the present invention;

[0056] Figure 8 It is a structural diagram of an online training video recording system provided by a certain embodiment of the present invention. DETAILED DESCRIPTION

[0057] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0059] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two components. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are for illustrative purposes only, and do not indicate or imply that the system or component referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0060] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0061] In one embodiment, if Figure 1 As shown, the first aspect of the present invention provides an online training video recording method, comprising:

[0062] S1. Obtain a recorded video image of a target subject, and add a timestamp to key content in the recorded video image; wherein the timestamp includes a start time and an end time; the target subject is a training file commonly used by trainers for auxiliary purposes during online training, such as PPT, documents, pictures, etc.; the recorded video image is usually recorded using a tool with a recording function such as a camera; the key content is the content that learners need to focus on learning and mastering during the training process. Usually, the key content will be explained or demonstrated by trainers during the training process, so when recording the key content, a timestamp is added to mark it, including the start time and the end time, for subsequent processing.

[0063] S2, identifying the target subject in each video frame in the recorded video image to obtain a first video picture;

[0064] Specifically, since the target subject can be text or a picture, but during training presentations, in order to facilitate learning personnel to watch, the training files are formed through a certain arrangement, and the recorded video image cannot be filled with the target subject, so the target subject in each video frame in the recorded video image needs to be identified and processed to obtain the first video picture, that is, the image picture containing the key content.

[0065] The present application first processes the recorded video image to obtain the corresponding grayscale image and pixel image respectively; then uses the particle swarm algorithm to search the grayscale image with the first preset parameters (the parameters include resolution, weight, learning factor, and speed) and calculates the optimal grayscale threshold (that is, the optimal parameter obtained in the search process), and based on the optimal grayscale threshold obtained in the previous search, uses the cooperative particle swarm algorithm to search the grayscale image with the second preset parameters and update the optimal grayscale threshold, repeats the update step until the preset search times are reached, obtains the final optimal grayscale threshold and uses it to segment the original image to obtain a number of image blocks; then Determine the boundary range of the pixel map and set a central seed point at the center of the pixel map, and randomly select multiple initial seed points in the pixel map, and merge the adjacent pixels whose pixel value difference with the central seed point is less than the pixel threshold and the pixels whose color distance (such as Euclidean distance, Manhattan distance, etc.) between pixels is less than the color threshold as the growth criterion for growth, and stop when there are no new pixels to obtain several segmented areas; finally, compare several image blocks with several segmented areas, use the overlapping part as the initial video picture, and use the boundary of the segmented area to optimize the boundary of the initial video picture, so as to obtain the first video picture.

[0066] This application uses a multi-particle algorithm and a region growing algorithm to identify and segment images, combines global search and local optimization, integrates different modal information from the image through the multi-particle algorithm, and uses this information through the growing algorithm to construct a more accurate similarity criterion, thereby more accurately dividing the image area, improving the segmentation accuracy and robustness, and can also improve the efficiency and convergence speed of the algorithm, and enhance the adaptability of the algorithm.

[0067] S3, extracting features of the first video image through a preset neural network model, and performing denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image;

[0068] In one embodiment, the preset neural network model includes a deep residual network layer, a first convolutional neural network layer, a recurrent neural network layer, a second convolutional neural network layer, a third convolutional neural network layer and a fourth convolutional neural network layer; specifically, the preset neural network model can be trained by original and denoised initial video image samples, the original video image samples are input into the preset neural network model, the denoised target video image samples are output, the error loss between the target video image samples and the denoised initial video image samples is calculated, and the preset neural network model is trained by the error loss to obtain the trained preset neural network model, and it is used for denoising in the scheme described in the present application to improve the clarity of the video image.

[0069] In one embodiment, step S3 is as follows Figure 2 As shown, including:

[0070] S31, extracting image features of the first video picture through the deep residual network layer, and extracting spatial features of the first video picture through the first convolutional neural network layer;

[0071] S32, processing the spatial features of the first video picture in the current video frame and the spatial features of the first video picture in the previous video frame through the recurrent neural network layer to obtain the temporal features of the first video picture in the current video frame;

[0072] S33, extracting state features of the first video picture through the second convolutional neural network layer, and calculating a contribution value of the first video picture according to the state features and the image features;

[0073] S34, fusing the temporal features and the spatial features of the corresponding video frame to obtain spatiotemporal features, and based on the spatiotemporal features, performing convolution processing on the first video picture through the third convolutional neural network layer to obtain a convolution matrix;

[0074] S35. Perform residual processing on the convolution matrix through the fourth convolutional neural network layer to obtain the first video picture after denoising, and combine the first video picture after denoising according to the contribution value to obtain a high-definition recorded video image picture.

[0075] Specifically, the present application uses a deep residual network layer to extract image features of video pictures corresponding to video frames, which can quickly extract important information from frame images; the spatial features of the first video picture are extracted through the first convolutional neural network layer, and the spatial features are used to represent the position of noise elements in the video picture, and the characteristics of the recurrent neural network layer are used to summarize and extract the time features of the video picture corresponding to the current video frame based on the spatial features of the video pictures corresponding to the current and previous video frames, and so on, the time features of the video pictures corresponding to the entire video frame sequence are obtained according to the time features of the video pictures corresponding to the current video frame; and then by fusing the time features and spatial features of the corresponding video images in the video frames, the time and space correlations between adjacent frame images are established, which is more helpful to identify and extract the corresponding target subject and remove noise; at the same time, considering that the proportions of different video pictures are different, the proportions of various video components in the same video picture are also different, and the present application calculates the contribution value of the corresponding video picture through the image features and hidden state features corresponding to the video pictures in the video frames, and combines the denoised video pictures in chronological order according to the proportions corresponding to the contribution values ​​to obtain high-definition recorded video image pictures.

[0076] This application removes noise through a preset neural network model after training, so that the quality of video images is significantly improved. Viewers can see the details in the image more clearly, thereby improving the overall viewing experience. The noise reduction process helps to accurately capture and display important information in the video image, reduce the interference and confusion of noise on image details, so as to better summarize and highlight the key content of the training process.

[0077] S4. Combining the high-definition recorded video images in chronological order to obtain a high-definition recorded video; specifically, the high-definition recorded video can be used as video material, and through video editing software such as Adobe Premiere Pro, FinalCut Pro, Wondershare Filmora, etc., the high-definition recorded video can be generated by adding the video material to the timeline, adjusting the order of the video material, editing and adjusting the video, setting export parameters, etc.

[0078] S5, inserting a split-screen node and its corresponding sub-screen into the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-screen includes auxiliary content corresponding to the key content;

[0079] In one embodiment, step S5 is as follows: Figure 3 As shown, including:

[0080] S51, obtaining a secondary picture and its corresponding secondary video frame, and setting the time length of the secondary picture by the timestamp; specifically, the secondary picture contains the auxiliary file content corresponding to the training file to further supplement the training file, which may be annotations, small window videos or other auxiliary content, etc. The start time in the timestamp is used as the start time of the secondary picture, and the end time in the timestamp is used as the end time of the secondary picture, so that the secondary picture can be played synchronously with the main picture.

[0081] S52, extracting a video frame from the high-definition recorded video based on the timestamp and using it as a main video frame in the main picture;

[0082] S53, creating a new video frame on the frame corresponding to the timestamp and adding the main video frame and the auxiliary video frame to obtain a new fused video frame;

[0083] In one embodiment, step S53 is as follows: Figure 4 As shown, including:

[0084] S531, creating a new video frame on the frame corresponding to the timestamp, and dividing the new video frame into regions to obtain a first region and a second region;

[0085] S532: Add the main video frame to the first area, and add the auxiliary video frame to the second area to obtain a new fused video frame.

[0086] Specifically, the present application merges the main and sub-video frames by creating a new fused video frame, and adds the corresponding main and sub-video frames to the corresponding areas according to the timestamp, and displays the contents of two or more video sources in the same frame, which can enrich the information content of the video and enable learners to obtain more visual information at the same time, thereby enhancing the expressiveness and attractiveness of the video; when dividing the area, a unique video layout can be designed as needed to achieve personalized visual effects; through area division and content fusion to optimize the presentation of video content, important content is more prominent, and secondary information is reasonably supplemented, thereby improving the overall quality and viewing experience of the video, so that learners can effectively learn the key content of the training process.

[0087] S54, setting a split screen node at the start time, and inserting the new fused video frame according to the split screen node according to the timestamp to replace the original video frame, to obtain a video recording result;

[0088] Specifically, the present application sets a split-screen node at a specified time point to mark the starting position of the split-screen operation to be performed, and determines the precise position of one or more video frames according to the split-screen node, and replaces the original video frames with new fused video frames at these positions to generate a video recording result containing new content. By inserting new fused video frames, the present application can introduce additional information or perspectives into high-definition recorded videos, thereby greatly improving the richness and diversity of video content; it can quickly add or modify content in high-definition recorded videos without reprocessing the entire video file, thereby improving the efficiency of video production.

[0089] In one embodiment, after step S5, the method further includes:

[0090] S6, condensing the video recording result to obtain a condensed video;

[0091] In one embodiment, step S6 is as follows: Figure 5 As shown, including:

[0092] S61, identifying each high-definition recorded video frame in the video recording result, and obtaining a second video picture corresponding to each high-definition recorded video frame;

[0093] S62, selecting target video pictures that meet preset conditions from the second video pictures to obtain a target picture list;

[0094] S63, cropping the video recording result according to the target picture list to obtain a plurality of video clips and combining them to obtain a condensed video.

[0095] Specifically, the recording of the online training process is to facilitate the re-learning of later learners, but the entire training process does not necessarily require key learning content. Learners view the playback video to obtain a lot of useful information. Therefore, after obtaining the video recording results, this application also condenses it to extract the key content, so as to improve learning efficiency. Among them, the recognition method of the second video screen can refer to the recognition process of the first video screen, which will not be elaborated here; and the preset conditions include the duration of the appearance reaching the duration threshold. Under normal circumstances, the key content often requires the training personnel to focus on explaining, so it will take more time in this part. The duration of the key content in the previous training videos is counted, and the shortest duration threshold can be evaluated. The video screen that exceeds this value is the target video screen to form a target video screen list. The video segments with the target video screen can be cropped and combined in chronological order to obtain a condensed video.

[0096] By selecting target video images that meet preset conditions, the present application can remove redundant or unnecessary parts in the video, so that the final condensed video content is more compact and refined, and directly focuses on the information that learners need to learn, so as to improve viewing efficiency; and the condensed video obtained by cropping has a file size greatly reduced compared to the original video recording result, which can not only save storage space and reduce storage costs, but also reduce the bandwidth required during video transmission and improve transmission efficiency.

[0097] In one embodiment, step S62 is as follows: Figure 6 As shown, including:

[0098] S6211, calculating the feature similarity of the second video picture, traversing each high-definition recorded video frame, and comparing the feature similarity of the second video picture located in the current high-definition recorded video frame with the feature similarity of the second video picture located in the adjacent high-definition recorded video frame;

[0099] S6212. The two second video pictures whose comparison results meet the preset comparison condition and whose appearance duration reaches the preset duration threshold are taken as the same target video picture, and the target video pictures whose pixels are not less than the preset pixel threshold are combined to obtain a target picture list.

[0100] Specifically, the feature similarity of video pictures can be calculated by SSIM based on structural similarity. By calculating the brightness, contrast and structural information of the video pictures, the similarity between video pictures can be more comprehensively evaluated; the preset comparison condition is the similarity between video pictures. It should be noted that the preset class and threshold class values ​​in the text are set according to actual conditions and actual needs, and are not specifically limited here. The present application determines the target video picture by the feature similarity of the video picture, and eliminates the video picture with insufficient duration and too small pixel value, which can effectively avoid using useless video pictures as target video pictures for video concentration, thereby improving the efficiency of video concentration, so as to provide better and useful video image data, which is convenient for the later learning of learners.

[0101] In another embodiment, step S62 is as follows: Figure 7 As shown, including:

[0102] S6221, extracting color data of the second video picture, and removing background color data from the color data to obtain content color data;

[0103] S6222: If the content color data belongs to a pre-built color database, and the appearance duration of the second video picture corresponding to the content color data reaches a preset duration threshold, then the video picture is used as the target video picture;

[0104] S6223: Combine the target video frames to obtain a target video frame list.

[0105] Specifically, during online training, trainers usually mark or interpret the displayed training documents. After removing the color of the video screen corresponding to the training file, the remaining color data are all color data corresponding to the text added when the training file is directly demonstrated later. Since the text added later is used to highlight key points and make distinctions, it is very different from the color of the training file itself. The two cannot be converted to each other, and the more important the content, the more text is added, which makes the accuracy of the target video screen obtained extremely high. Among them, the pre-constructed color database includes the annotation colors commonly used in the training process, and the target video screens are usually combined in chronological order. By setting the duration threshold of the second video screen corresponding to the content color data, the present application can further screen out the important screens that appear continuously in the video, which helps to remove short and unimportant screens and retain only the key learning content that is valuable to the learners.

[0106] In the embodiment of the present application, based on the problem that the existing video recording method cannot improve the clarity of the video recording screen and cannot highlight the key content of the training process, an online training video recording method is designed, which realizes obtaining a recorded video image of a target subject and adding a timestamp to the key content in the recorded video image; identifying the target subject in each video frame in the recorded video image to obtain a first video screen; extracting the features of the first video screen through a preset neural network model, and denoising the recorded video image based on the features to obtain a high-definition recorded video image screen; combining the high-definition recorded video image screens in chronological order to obtain a high-definition recorded video; inserting a split-screen node and its corresponding sub-screen in the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-screen includes a technical solution of auxiliary content corresponding to the key content; enhancing the information content and viewing experience of the video, improving the quality of the recorded video, and generating a high-definition and information-rich video recording result.

[0107] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0108] In another embodiment, if Figure 8 As shown, the second aspect of the present invention provides an online training video recording system, comprising:

[0109] The video image processing module 10 is used to obtain a recorded video image of a target subject and add a time stamp to key contents in the recorded video image;

[0110] A video frame recognition module 20 is used to recognize the target subject in each video frame in the recorded video image to obtain a first video frame;

[0111] A video image denoising module 30, configured to extract features of the first video image through a preset neural network model, and perform denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image;

[0112] A first video generation module 40, configured to combine the high-definition recorded video images in chronological order to obtain a high-definition recorded video;

[0113] The second video generation module 50 is used to insert a split-screen node and its corresponding sub-picture in the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-picture includes auxiliary content corresponding to the key content.

[0114] It should be noted that each module in the above-mentioned online training video recording system can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For the specific definition of an online training video recording system, please refer to the definition of an online training video recording method above. The two have the same functions and effects, which will not be repeated here.

[0115] In summary, the present invention relates to the field of communication technology, and discloses an online training video recording method and system, wherein the method includes obtaining a recorded video image of a target subject, and adding a timestamp to key content in the recorded video image; identifying the target subject in each video frame in the recorded video image to obtain a first video picture; extracting features of the first video picture through a preset neural network model, and denoising the recorded video image based on the features to obtain a high-definition recorded video image picture; combining the high-definition recorded video image pictures in chronological order to obtain a high-definition recorded video; inserting a split-screen node and its corresponding sub-picture in the high-definition recorded video according to the timestamp to obtain a video recording result, and the sub-picture includes auxiliary content corresponding to the key content; the problem that the existing video recording method cannot improve the clarity of the video recording picture, nor can it highlight the key content of the training process, is facilitated for the later learning of the trainees.

[0116] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0117] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. An online training video recording method, characterized in that: include: Acquire a recorded video image of a target subject, and add a time stamp to key content in the recorded video image; Identify the target subject in each video frame of the recorded video image to obtain a first video picture; Extracting features of the first video image through a preset neural network model, and performing denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image; Combining the high-definition recorded video images in chronological order to obtain a high-definition recorded video; A split-screen node and its corresponding sub-picture are inserted into the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-picture includes auxiliary content corresponding to the key content.

2. The online training video recording method according to claim 1, characterized in that: The preset neural network model includes a deep residual network layer, a first convolutional neural network layer, a recurrent neural network layer, a second convolutional neural network layer, a third convolutional neural network layer and a fourth convolutional neural network layer.

3. The online training video recording method according to claim 2, characterized in that: The extracting the features of the first video picture by using a preset neural network model, and performing denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image picture, includes: Extracting image features of the first video picture through the deep residual network layer, and extracting spatial features of the first video picture through the first convolutional neural network layer; Processing the spatial features of the first video picture in the current video frame and the spatial features of the first video picture in the previous video frame through the recurrent neural network layer to obtain the temporal features of the first video picture in the current video frame; Extracting state features of the first video picture through the second convolutional neural network layer, and calculating a contribution value of the first video picture according to the state features and the image features; The temporal features and the spatial features of the corresponding video frame are fused to obtain spatiotemporal features, and based on the spatiotemporal features, the first video picture is convolved through the third convolutional neural network layer to obtain a convolution matrix; The convolution matrix is ​​subjected to residual processing through the fourth convolutional neural network layer to obtain the first video picture after denoising, and the first video picture after denoising is combined according to the contribution value to obtain a high-definition recorded video image picture.

4. The online training video recording method according to claim 1, characterized in that: The timestamp includes a start time and an end time; The inserting of the split-screen node and its corresponding sub-screen into the high-definition recorded video according to the timestamp to obtain the video recording result includes: Obtaining a secondary picture and its corresponding secondary video frame, and setting the time length of the secondary picture by using the timestamp; Based on the timestamp, extract a video frame from the high-definition recorded video and use it as a main video frame in the main picture; On the frame corresponding to the timestamp, a new video frame is created and the main video frame and the auxiliary video frame are added to obtain a new fused video frame; A split screen node is set at the start time, and the new fused video frame is inserted according to the split screen node according to the timestamp to replace the original video frame to obtain a video recording result.

5. The online training video recording method according to claim 4, characterized in that: The creating a new video frame and adding the main video frame and the auxiliary video frame on the frame corresponding to the timestamp to obtain a new fused video frame includes: On the frame corresponding to the timestamp, create a new video frame, and divide the new video frame into regions to obtain a first region and a second region; The main video frame is added to the first area, and the auxiliary video frame is added to the second area to obtain a new fused video frame.

6. The online training video recording method according to claim 1, characterized in that: After inserting the split screen node and its corresponding sub-picture into the high-definition recorded video according to the timestamp to obtain the video recording result, the method further includes: The video recording result is condensed to obtain a condensed video.

7. The online training video recording method according to claim 6, characterized in that: The condensing process of the video recording result to obtain a condensed video includes: Identify each high-definition recorded video frame in the video recording result to obtain a second video picture corresponding to each high-definition recorded video frame; Selecting target video pictures that meet preset conditions from the second video pictures to obtain a target picture list; The video recording result is cropped according to the target screen list to obtain a plurality of video clips and combine them to obtain a condensed video.

8. The online training video recording method according to claim 7, characterized in that: The step of selecting target video pictures that meet a preset condition from the second video pictures to obtain a target picture list includes: Calculating the feature similarity of the second video picture, traversing each high-definition recorded video frame, and comparing the feature similarity of the second video picture located in the current high-definition recorded video frame with the feature similarity of the second video picture located in an adjacent high-definition recorded video frame; The two second video pictures whose comparison results meet the preset comparison conditions and whose appearance duration reaches the preset duration threshold are taken as the same target video picture, and the target video pictures whose pixels are not less than the preset pixel threshold are combined to obtain a target picture list.

9. The online training video recording method according to claim 1, characterized in that: The step of selecting target video pictures that meet a preset condition from the second video pictures to obtain a target picture list includes: Extracting color data of the second video picture, and removing background color data from the color data to obtain content color data; If the content color data belongs to a pre-constructed color database, and the appearance duration of the second video picture corresponding to the content color data reaches a preset duration threshold, then the video picture is used as the target video picture; The target video pictures are combined to obtain a target video picture list.

10. An online training video recording system, characterized in that: include: A video image processing module, used to obtain a recorded video image of a target subject and add a time stamp to key content in the recorded video image; A video picture recognition module, used for recognizing the target subject in each video frame in the recorded video image to obtain a first video picture; A video image denoising module, configured to extract features of the first video image through a preset neural network model, and perform denoising processing on the recorded video image based on the features to obtain a high-definition recorded video image; A first video generation module, used for combining the high-definition recorded video images in chronological order to obtain a high-definition recorded video; The second video generation module is used to insert a split-screen node and its corresponding sub-picture in the high-definition recorded video according to the timestamp to obtain a video recording result, wherein the sub-picture includes auxiliary content corresponding to the key content.

Images