A video processing method, storage medium and related apparatus

Abstract

The application relates to the technical field of Internet of Things devices and discloses a video processing method, a storage medium and related devices, which comprises the following steps: receiving video data uploaded by a device end and marking each video data with a corresponding event mark; inputting each piece of video data and a preset prompt word into an AI large model to obtain a text description of each piece of video data; when a video summary request is received, performing deduplication on video data in a past preset time window according to the similarity between the text descriptions, so that the similarity between the text descriptions of any two pieces of video data in the deduplicated video data set is all lower than a preset similarity threshold; selecting a preset number of video data from the video data set as to-be-edited video data according to a preset summary rule; editing and splicing the to-be-edited video data into a summary video in chronological order, and pushing the summary video to a user terminal. The method can improve the information amount of the summary video and the acquisition efficiency of important monitoring information.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) device technology, specifically to a video processing method, storage medium, and related apparatus. Background Technology

[0002] As people's living standards improve, home appliances are becoming increasingly intelligent, and various smart home monitoring devices are becoming more and more widespread. Some existing monitoring devices compress video data of key events detected the previous day into a short video at fixed times each day and push it to the user's terminal device so that the user can quickly review the key events of the previous day. However, to avoid missing crucial information, existing monitoring devices usually set relatively sensitive thresholds for detecting abnormal events. This results in the detection of many useless information events, and a large number of repeated events within the same time period. Consequently, the compressed video of the key events pushed out at the end is still quite long and contains relatively little information. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the present invention aims to provide a video processing method, storage medium and related apparatus that can avoid the inclusion of too much duplicate information in the devices pushed to users, further increase the amount of information carried by the thumbnail videos pushed to users, and improve the user experience.

[0004] To solve the above problems, the technical solution adopted by the present invention is as follows: A video processing method, comprising the following steps: Receive video data uploaded by the receiving device; Each video data point is labeled with a corresponding event tag based on its content and recording time. Each video segment and preset prompts are input into the AI ​​model to obtain a text description of each video segment. When a video summary request is received, the similarity between all text descriptions within the past preset time window is calculated, and the video data within the past preset time window is deduplicated based on the similarity, so that the similarity between the text descriptions of any two video data in the deduplicated video dataset is lower than the preset similarity threshold. Based on the preset summary rules and the event tags of each video data segment in the video dataset, a preset number of video data segments are selected from the video dataset as video data to be edited. The video data to be edited is edited and spliced ​​together in chronological order to form a summary video, which is then pushed to the user's terminal.

[0005] The above video processing method, where event tagging includes event type tagging, and the step of selecting a preset number of video data as video data to be edited from the video dataset based on preset summarization rules and the event tagging of each video data segment in the video dataset, includes: Based on the user-defined set of key event types and event type tags, the video data in the video dataset is divided into key video sets and non-key video sets; Extract a preset number of key events from the key video set as the video data to be edited, and extract a preset number of non-key events from the non-key video set as the video data to be edited.

[0006] The above-described video processing method further includes time stamping for event marking. The steps of extracting a preset number of video data representing key events from a key video set as video data to be edited, and extracting a preset number of video data representing non-key events from a non-key video set as video data to be edited, include: The video data in the key video set is divided into multiple time periods based on the preset first time interval and time marker; Iterate through each time period in the key video set. If there is video data in the current time period, select the highest priority video data from the video data in the current time period as the video data to be edited, and reduce the priority of that event type by one level. If there is no video data in the current time period, continue to iterate through the next time period. Repeat the iteration through each time period in the key video set until the number of video data extracted from the key video set reaches the preset number of key events. Based on the preset second time interval and time marker, the video data in the non-key video set is divided into multiple time periods; Iterate through each time period in the non-key video set. If there is video data in the current time period, randomly select one video data from the video data in the current time period as the video data to be edited. If there is no video data in the current time period, continue to iterate through the next time period. Repeat the iteration through each time period in the non-key video set until the number of video data extracted from the non-key video set reaches the preset number of non-key events.

[0007] The above video processing method further includes, before the step of receiving video data uploaded by the receiving device: The device collects ambient audio and video information. The device determines whether there are any abnormal events in the collected audio and video information. If there are abnormal events, the audio and video information within a preset time period before and after the abnormal event is detected is converted into video data and uploaded to the server.

[0008] A storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the above-described video processing method.

[0009] An electronic device includes a memory and a processor, the processor and the memory being communicatively connected via a data bus, the processor being able to implement the above-described video processing method by calling and executing a computer program stored in the memory.

[0010] A video processing service system includes a video acquisition device and a server. The video acquisition device is communicatively connected to the server. The server includes an acquisition module, a tagging module, a content summarization module, a similarity deduplication module, a filtering module, an editing module, and a push module. The acquisition module receives video data from the device. The tagging module assigns a corresponding event tag to each video data segment based on its content and recording time. The content summarization module inputs each video data segment and preset prompts into an AI model to obtain a text description of each video data segment. The similarity deduplication module calculates the text description within a preset time window when a video summarization request is received. The similarity between all the text descriptions is used to deduplicate the video data within the preset time window, ensuring that the similarity between the text descriptions of any two segments of video data in the deduplicated video dataset is below a preset similarity threshold. The filtering module selects a preset number of video data as video data to be edited from the video dataset according to preset summary rules and event markers for each segment of video data in the video dataset. The editing module edits and splices the video data to be edited into a summary video in chronological order. The push module pushes the summary video to the user terminal at fixed time nodes or based on user requests.

[0011] The aforementioned video processing service system, wherein the filtering module is further configured to: divide the video data in the video dataset into a key video set and a non-key video set according to the key event type set and event type marker set set by the user; extract a preset number of video data with key events from the key video set as the video data to be edited, and extract a preset number of video data with non-key events from the non-key video set as the video data to be edited.

[0012] The aforementioned video processing service system, wherein the filtering module is further configured to: divide the video data in the key video set into multiple time periods according to a preset first time interval and a time marker; traverse each time period of the key video set; if there is video data in the current time period, select the highest priority video data from the video data in the current time period as the video data to be edited according to the priority of each event type in the key event type set, and lower the priority of that event type by one level; if there is no video data in the current time period, continue traversing the next time period; repeat traversing each time period of the key video set until the video data in the key video set is selected from the video data in the current time period. The number of video data extracted from the video set reaches the preset number of key events; the video data in the non-key video set is divided into multiple time periods according to the preset second time interval and time marker; each time period in the non-key video set is traversed, and if there is video data in the current time period, a video data is randomly selected from the video data in the current time period as the video data to be edited; if there is no video data in the current time period, the traversal continues to the next time period; the traversal of each time period in the non-key video set is repeated until the number of video data extracted from the non-key video set reaches the preset number of non-key events.

[0013] The video processing service system described above includes a video acquisition device comprising an acquisition module and an anomaly detection module. The acquisition module controls the hardware acquisition module of the video acquisition device to acquire audio-visual information surrounding the video acquisition device. The anomaly detection module determines whether there are any abnormal events in the acquired audio-visual information, and when an abnormal event is detected, converts the audio-visual information within a preset time period before and after the detection of the abnormal event into video data and uploads it to the server.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: The video processing method, storage medium and related device of the present invention utilize AI large model to summarize and extract text descriptions of video data uploaded by the device, and deduplicate the video data uploaded by the device according to the similarity between the text descriptions, so that the similarity of the deduplicated video data has a certain difference, thereby avoiding a lot of repetitive content in the summary video after editing and splicing these video data, increasing the information content of the summary video, so that users can more efficiently browse the important event information monitored by the monitoring equipment in the past day, and improve the user experience.

[0015] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0016] Figure 1This is a flowchart of the video processing method according to the first embodiment of the present invention.

[0017] Figure 2 This is a flowchart of a video processing method according to a second embodiment of the present invention.

[0018] Figure 3 This is a flowchart illustrating the video data selection process for key event types in an embodiment of the present invention.

[0019] Figure 4 This is a schematic block diagram of an electronic device according to an embodiment of the present invention.

[0020] Figure 5 This is a schematic diagram of the video processing service system according to an embodiment of the present invention. Detailed Implementation

[0021] The embodiments of the present invention are described in detail below, with reference to... Figure 1 The present invention provides a video processing method, comprising the following steps: Receive video data uploaded by the receiving device; Each video data point is labeled with a corresponding event tag based on its content and recording time. Each video segment and preset prompts are input into the AI ​​model to obtain a text description of each video segment. When a video summary request is received, the similarity between all text descriptions within the past preset time window is calculated, and the video data within the past preset time window is deduplicated based on the similarity, so that the similarity between the text descriptions of any two video data in the deduplicated video dataset is lower than the preset similarity threshold. Based on the preset summary rules and the event tags of each video data segment in the video dataset, a preset number of video data segments are selected from the video dataset as video data to be edited. The video data to be edited is edited and spliced ​​together in chronological order to form a summary video, which is then pushed to the user's terminal.

[0022] This method utilizes a large AI model to analyze video data uploaded from each device. It generates textual descriptions for each video segment based on the video data content. When generating the summary video, it deduplicates the uploaded video data based on the similarity of the textual descriptions, thereby reducing the total length of the spliced ​​summary video or preventing it from containing a large amount of repetitive information. This increases the information content of the summary video, allowing users to quickly grasp the important events monitored by the device the previous day, improving the user experience. This method, which uses the reasoning capabilities of a large AI model to analyze the textual descriptions of the video data for deduplication, requires less computation than directly comparing the similarity of audio and video information. It also improves deduplication accuracy, avoiding the filtering of video data containing different event information as duplicate content simply because of high image similarity, and preventing the filtering of important information during deduplication.

[0023] Understandably, large-scale AI models should ideally employ multimodal models trained jointly by integrating various data types such as text, images, videos, and audio, such as Qwen-VL-Max or GPT-4V. By inputting each video segment along with prompts like "Please generate a text description summary based on the content of this video data, outputting in JSON format" into the large-scale AI model, the model obtains the text description corresponding to each video segment. After establishing a mapping relationship between the text descriptions and the video data, this information is stored in the server's database.

[0024] Understandably, the server can obtain the similarity between any two video segments by converting the text description of each video segment into corresponding text vectors using a pre-trained encoder and calculating the cosine similarity between the text vectors corresponding to any two video segments. If the similarity between the text vectors of any two video segments is higher than a preset similarity threshold, then one video is retained from the two segments according to a preset filtering rule. By repeatedly sampling, calculating similarity, and filtering the video data uploaded from the device, the process continues until the similarity between the text descriptions of any two videos in the final filtered video dataset is lower than the preset similarity threshold. Understandably, in some embodiments, a large AI model can also be used to calculate the similarity between the text descriptions of two video segments.

[0025] In some embodiments, to further improve the deduplication accuracy of video data collected by the device and avoid the misdeduplication of video data with higher information value, which would prevent users from quickly obtaining important information from the previous day or several hours by summarizing the videos, the server can also analyze the danger weight of each video data segment by analyzing the number of security risk events contained in the video data while outputting the text description of each video data segment through AI large model analysis. During the filtering process, if the similarity threshold of the text descriptions corresponding to any two video data segments is higher than the preset similarity threshold, the video data with the higher danger weight is selected and retained first, so as to avoid important security-related information being filtered out.

[0026] Reference Figure 2In this embodiment, the device continuously collects audio-visual information from its surroundings and inputs this information into a pre-trained lightweight model on the device to monitor for abnormal sounds, changes in the visuals, or the appearance of faces or pets. Upon detecting an abnormal event, the device converts the audio-visual information for a period before and after the event (e.g., 4 seconds) into an 8-second video and uploads it to the server. The server receives the video data from the device and inputs it along with preset prompts into the AI ​​model to obtain a text description for each video segment. The server then establishes a correlation between each video segment and its corresponding text description and stores it in its database. Simultaneously, the server analyzes the event type in the video data based on the text description (e.g., a child crying, broken objects, or the appearance of a shadowy figure) and, combined with the date and time of video data collection, tags each video segment with an event tag containing both the event type and a time stamp. When the preset summary time is reached, such as 0:00 every day, the server automatically sends a video summary request, or receives a video summary request actively sent by the user through the terminal device. If the total number of video data collected the previous day is less than the preset maximum number of videos to select, such as 15, then there is no need to deduplicate the video data based on the text description of each video segment. Instead, all the video data is directly edited and spliced ​​into a 2-minute summary video according to the time stamp order, and pushed to the user's terminal device, such as a mobile app or mini-program, at the preset push time, such as 8:00 every morning. If the number of video data from the previous day is greater than 15, then the similarity of the text descriptions of any two video segments is calculated pairwise. If the similarity is higher than 70%, one segment is saved in the video dataset, and the other segment is excluded from the video dataset, until the similarity of the text descriptions of any consecutive video segments in the video dataset is less than or equal to 70%. If the number of videos in the video dataset after deduplication is less than or equal to 15, the videos in the video dataset will be directly edited and spliced ​​into a summary video. Alternatively, some videos will be recalled from the videos excluded during deduplication to make up the number of videos in the video dataset to 15 before editing and splicing into a summary video. If the number of videos in the video dataset is greater than 15, the videos in the video dataset will be divided into a key video set and a non-key video set according to the event type label of each video segment, based on the user's preset key event type set. Twelve videos will be extracted from the key video set and three videos will be extracted from the non-key video set as video data to be edited and added to the video dataset to be edited and spliced ​​into a summary video.

[0027] Reference Figure 3In some embodiments, to ensure the summary video content covers each time period of the previous day as evenly as possible, and to cover each event type set by the user in the key event type set, the server divides the video data from the key video set of the previous day's 24 hours into 12 time periods based on the timestamp of each video segment, according to a preset first time interval, such as 2 hours. The priority of each event type in the event type set is initialized to 12, and the dataset of videos to be edited is cleared. Then, each time period is iterated. If video data exists in the current time period, the video data with the highest priority corresponding to the event type is extracted and added to the dataset of videos to be edited. The priority of the selected event type is decremented by 1, and then the process continues to the next time period. If, after iterating through all time periods, the number of video data selected from the key video set is less than 12, several more videos are randomly selected from the unselected videos in the key video set to make up the number of selected videos to 12; or the process continues to iterate from the first time period until the number of selected video data reaches 12. Similarly, for video data in the non-priority video set, the video data from the previous day's 24 hours is divided into three time periods based on the timestamp of each video segment, according to a second time interval, such as 8 hours. One video from each time period is then randomly selected and added to the dataset to be edited. It's understood that if multiple videos of the same event type exist within the same time period, one video of that type is randomly selected and added to the dataset to be edited.

[0028] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing a computer program, which, when executed by a processor, implements the video processing method described above.

[0029] In some possible implementations, various aspects of the video processing method provided by the present invention can also be implemented as a program product comprising program code that, when the program product is run on a device, causes the control device to perform the steps of the video processing method according to the various exemplary embodiments of the present application described above.

[0030] By designing and programming the processor, the code corresponding to the video processing method described in the foregoing embodiments can be embedded into the chip, enabling the chip to execute the steps of the video processing method shown in the embodiments of the present invention during operation. How to design and program the processor is a technique well-known to those skilled in the art and will not be elaborated upon here.

[0031] Based on the same inventive concept, referring to Figure 4The present invention also provides an electronic device, including a processor, a memory and a data bus. The processor is communicatively connected to the memory via the data bus, and the processor can implement the above-mentioned video processing method by calling and executing a computer program in the memory.

[0032] In one possible design, the processor may include one or more processing units. The processor and memory may be implemented on the same chip or on separate chips. The processor may be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the video processing method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0033] Memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited to this. The memory in the embodiments of this application can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0034] Based on the same inventive concept, referring to Figure 5Embodiments of the present invention also provide a video processing service system, including a video acquisition device and a server. The video acquisition device can be a network-enabled monitoring device such as an IP network camera, and the video acquisition device communicates with the server via the Internet. Both the video acquisition device and the server can include a processor and a memory. The processor communicates with the memory via a data bus, and the processor implements various software modules by calling and executing computer programs in the memory to implement the above-described video processing method. Specifically, the server includes an acquisition module, a tagging module, a content summarization module, a similarity deduplication module, a filtering module, an editing module, and a push module. The acquisition module receives video data from the device; the tagging module tags each video data with a corresponding event tag based on its content and shooting time; the content summarization module inputs each video data segment and preset prompts into the AI ​​model to obtain a text description for each video data segment; the similarity deduplication module calculates the similarity between all text descriptions within a preset time window when a video summary request is received, and deduplicates the video data within the preset time window based on the similarity, ensuring that the similarity between the text descriptions of any two video data segments in the deduplicated video dataset is below a preset similarity threshold; the filtering module selects a preset number of video data segments as video data to be edited based on preset summary rules and the event tags of each video data segment in the video dataset; the editing module edits and splices the video data to be edited into a summary video in chronological order; and the push module pushes the summary video to the user terminal at fixed time nodes or based on user requests.

[0035] Reference Figure 5 In some embodiments, the video acquisition device includes an acquisition module and an anomaly detection module. The acquisition module controls the device's hardware, such as a microphone and image sensor, to acquire audio-visual information around the device. The anomaly detection module determines whether there are any abnormal times in the acquired audio-visual information, and when an abnormal time is detected, converts the audio-visual information within a preset time period before and after the detected abnormal time into video data, and uploads it to the server via the communication module.

[0036] Understandably, in some embodiments, the filtering module includes a rule engine and a filtering process control module. The rule engine is used to set filtering rules for the video data to be edited, and outputs decision results based on input data such as the user-defined set of key event types, the event markers of the video data uploaded by the input device segment, and the number of video data. The filtering process control module is used to input the input data of each filtering step into the rule engine according to the preset filtering process, obtain the decision results of the rule engine, and perform processing such as selection on the video data according to the decision results. By implementing the video data selection process in the video processing method of this embodiment, the final video data to be edited can contain as much information as possible, including each key event type set by the user and video data from each time period of 24 hours a day.

[0037] It should be noted that in the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. If "first" or "second" is mentioned, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0038] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0039] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0040] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0041] In the description of this invention, unless otherwise explicitly defined, terms such as "set up," "install," and "connect" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0042] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A video processing method, characterized in that, Includes the following steps: Receive video data uploaded by the receiving device; Each video data point is labeled with a corresponding event tag based on its content and recording time. Each video segment and preset prompts are input into the AI ​​model to obtain a text description of each video segment. When a video summary request is received, the similarity between all text descriptions within the past preset time window is calculated, and the video data within the past preset time window is deduplicated based on the similarity, so that the similarity between the text descriptions of any two video data in the deduplicated video dataset is lower than the preset similarity threshold. Based on the preset summary rules and the event tags of each video data segment in the video dataset, a preset number of video data segments are selected from the video dataset as video data to be edited. The video data to be edited is edited and spliced ​​together in chronological order to form a summary video, which is then pushed to the user's terminal.

2. The video processing method according to claim 1, characterized in that, Event tagging includes event type tagging. The step of selecting a preset number of video data as video data to be edited from the video dataset based on preset summarization rules and the event tagging of each video data segment in the video dataset includes: Based on the user-defined set of key event types and event type tags, the video data in the video dataset is divided into key video sets and non-key video sets; Extract a preset number of key events from the key video set as the video data to be edited, and extract a preset number of non-key events from the non-key video set as the video data to be edited.

3. The video processing method according to claim 2, characterized in that, Event tagging also includes time tagging. The steps of extracting a preset number of key events from the key video set as video data to be edited, and extracting a preset number of non-key events from the non-key video set as video data to be edited, include: The video data in the key video set is divided into multiple time periods based on the preset first time interval and time marker; Iterate through each time period in the key video set. If there is video data in the current time period, select the highest priority video data from the video data in the current time period as the video data to be edited, and reduce the priority of that event type by one level. If there is no video data in the current time period, continue to iterate through the next time period. Repeat the iteration through each time period in the key video set until the number of video data extracted from the key video set reaches the preset number of key events. Based on the preset second time interval and time marker, the video data in the non-key video set is divided into multiple time periods; Iterate through each time period in the non-key video set. If there is video data in the current time period, randomly select one video data from the video data in the current time period as the video data to be edited. If there is no video data in the current time period, continue to iterate through the next time period. Repeat each time period in the non-key video set until the number of video data extracted from the non-key video set reaches the preset number of non-key events.

4. The video processing method according to claim 1, characterized in that, Before the step of receiving video data uploaded by the receiving device, the following also applies: The device collects ambient audio and video information. The device determines whether there are any abnormal events in the collected audio and video information. If there are abnormal events, the audio and video information within a preset time period before and after the abnormal event is detected is converted into video data and uploaded to the server.

5. A storage medium storing a computer program, characterized in that, When the computer program is invoked and executed by the processor, it implements the video processing method according to any one of claims 1 to 4.

6. An electronic device, characterized in that, The device includes a memory and a processor, the processor and the memory being communicatively connected via a data bus, the processor being able to implement the video processing method according to any one of claims 1 to 4 by calling and executing a computer program in the memory.

7. A video processing service system, characterized in that, The system includes a video capture device and a server. The video capture device is communicatively connected to the server. The server includes an acquisition module, a tagging module, a content summarization module, a similarity deduplication module, a filtering module, an editing module, and a push module. The acquisition module receives video data from the device. The tagging module assigns a corresponding event tag to each video data based on its content and shooting time. The content summarization module inputs each video data segment and preset prompts into an AI model to obtain a text description for each video data segment. The similarity deduplication module, upon receiving a video summary request, calculates the similarity between all text descriptions within a preset time window and deduplicates the video data within the preset time window based on the similarity, ensuring that the similarity between the text descriptions of any two video data segments in the deduplicated video dataset is below a preset similarity threshold. The filtering module is used to select a preset number of video data as video data to be edited from the video dataset according to preset summarization rules and event markers of each video data segment in the video dataset. The editing module is used to edit and splice the video data to be edited into a summary video in chronological order; The push module is used to push the summary video to the user terminal at fixed time points or based on user requests.

8. The video processing service system according to claim 7, characterized in that, The filtering module is further configured to: divide the video data in the video dataset into a key video set and a non-key video set according to the key event type set and event type marker set set by the user; extract a preset number of video data with key events from the key video set as the video data to be edited, and extract a preset number of video data with non-key events from the non-key video set as the video data to be edited.

9. The video processing service system according to claim 8, characterized in that, The filtering module is further configured to: divide the video data in the key video set into multiple time periods according to a preset first time interval and a time marker; traverse each time period in the key video set; if there is video data in the current time period, select the highest priority video data from the video data in the current time period as the video data to be edited according to the priority of each event type in the key event type set, and reduce the priority of that event type by one level; if there is no video data in the current time period, continue traversing the next time period; repeat the traversal of each time period in the key video set until the video data is extracted from the key video set. The number of video data reaches the preset number of key events; the video data in the non-key video set is divided into multiple time periods according to the preset second time interval and time marker; each time period in the non-key video set is traversed, and if there is video data in the current time period, one video data is randomly selected from the video data in the current time period as the video data to be edited; if there is no video data in the current time period, the next time period is traversed; each time period in the non-key video set is traversed repeatedly until the number of video data extracted from the non-key video set reaches the preset number of non-key events.

10. The video processing service system according to claim 7, characterized in that, The video acquisition device includes an acquisition module and an anomaly detection module. The acquisition module is used to control the hardware acquisition module of the video acquisition device to acquire audio-visual information around the video acquisition device. The anomaly detection module is used to determine whether there is an abnormal event in the acquired audio-visual information, and when the abnormal event is detected, convert the audio-visual information within a preset time period before and after the abnormal event into video data and upload it to the server.

Images