Video processing method and device, electronic equipment, storage medium and product

Abstract

The invention provides a video processing method and apparatus, an electronic device, a storage medium and a product. The method comprises the steps of obtaining audio data and a video frame set from a to-be-processed video; performing sensitive word detection on an audio text corresponding to the audio data, determining a target sensitive word in the audio text and time information corresponding to the target sensitive word, and performing desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain target audio data; determining a sensitive content area of each video frame in the video frame set, and desensitizing the sensitive content area to obtain a target video frame set; and combining the target audio data with the target video data synthesized by the target video frame set to obtain a desensitized video file. According to the scheme, the sensitive words in the audio data and the sensitive content areas in the video frames are automatically desensitized, so that the labor cost for desensitizing the video is reduced, and the desensitizing efficiency of the video is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a video processing method, device, electronic device, storage medium and product. Background Art

[0002] As a carrier of information transmission, video is inseparable from personal life and organizational development. Due to the random nature of video content and the changes in the environment, the original video content may not meet the current standards, or there is a risk that the video recipient may maliciously exploit the private content after the video is disseminated. Therefore, sensitive content in the video needs to be desensitized.

[0003] The current mainstream technical solution relies on post-production manual operations. That is, professional post-production personnel edit the video content in order to remove sensitive content. However, manual operations result in high labor costs and low efficiency in desensitizing videos. Summary of the invention

[0004] Based on the above needs, the present application proposes a video processing method, device, electronic device, storage medium and product, which can reduce the labor cost of desensitizing the video and improve the efficiency of desensitizing the video.

[0005] To achieve the above objectives, this application proposes the following technical solutions:

[0006] According to a first aspect of an embodiment of the present application, a video processing method is provided, including:

[0007] Obtain audio data and a set of video frames from the video to be processed;

[0008] Performing sensitive word detection on the audio text corresponding to the audio data, determining the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, and performing desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data;

[0009] Determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set;

[0010] The target audio data is combined with the target video data synthesized by the target video frame set to obtain a video file after desensitizing the video to be processed.

[0011] Optionally, performing sensitive word detection on an audio text corresponding to the audio data to determine a target sensitive word in the audio text and time information of the target sensitive word in the audio data includes:

[0012] Performing a matching analysis on the audio text corresponding to the audio data and the sensitive words in the pre-built sensitive word library, and taking the words in the audio text that match the sensitive words as the target sensitive words;

[0013] The time information of the target sensitive word in the audio data is obtained as the time information corresponding to the target sensitive word.

[0014] Optionally, performing sensitive word detection on an audio text corresponding to the audio data to determine a target sensitive word in the audio text and time information of the target sensitive word in the audio data includes:

[0015] Performing semantic analysis on each sentence in the audio text corresponding to the audio data, determining semantically incoherent words in each sentence, and outputting the sentence after marking the semantically incoherent words, so that the user can confirm whether the semantically incoherent words are sensitive words;

[0016] If the user's feedback information indicates that the semantically incoherent word is a sensitive word, the semantically incoherent word is used as a target sensitive word, and time information of the target sensitive word in the audio data is obtained as the time information corresponding to the target sensitive word.

[0017] Optionally, desensitizing the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data includes:

[0018] The audio segment in the time information corresponding to the target sensitive word in the audio data is replaced with a preset shielded audio to obtain the target audio data.

[0019] Optionally, determining the sensitive content area of ​​each video frame in the video frame set includes:

[0020] According to the similarity between the video frames, the video frames in the video frame set are divided into at least one sub-video frame set, and the similarity between any two video frames in the sub-video frame set reaches a preset similarity threshold;

[0021] Performing image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame; wherein the representative video frame in the sub-video frame set is any video frame in the sub-video frame set;

[0022] The coordinate information of the sensitive content area in the representative video frame is used as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set to which the representative video frame belongs.

[0023] Optionally, performing image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame includes:

[0024] Performing picture analysis on a representative video frame in the sub-video frame set to determine description information of a target object and coordinate information of the target object contained in the representative video frame;

[0025] The coordinate information of the sensitive content area in the representative video frame is determined according to the description information of the target object contained in the representative video frame and the coordinate information of the target object.

[0026] Optionally, determining the coordinate information of the sensitive content area in the representative video frame according to the description information of the target object contained in the representative video frame and the coordinate information of the target object includes:

[0027] Determining whether the representative video frame contains negative content according to the description information of the target object contained in the representative video frame;

[0028] If the representative video frame contains negative content, the coordinate information of each pixel point on the outer edge of the representative video frame is used as the coordinate information of the sensitive content area in the representative video frame.

[0029] Optionally, determining the coordinate information of the sensitive content area in the representative video frame according to the description information of the target object contained in the representative video frame and the coordinate information of the target object further includes:

[0030] If the representative video frame does not contain negative content, the description information of the target object contained in the representative video frame is matched and analyzed with the sensitive description information in a pre-built sensitive image description library, and the coordinate information of the target object corresponding to the description information in the representative video frame that matches the sensitive description information is used as the coordinate information of the sensitive content area in the representative video frame.

[0031] Optionally, desensitizing the sensitive content area to obtain a target video frame set includes:

[0032] Performing differential privacy processing on all pixels in the sensitive content area of ​​the video frames in the video frame set to obtain a target video frame corresponding to the video frame;

[0033] The target video frames corresponding to the respective video frames in the video frame set are sequentially combined into a target video frame set.

[0034] According to a second aspect of an embodiment of the present application, a video processing device is provided, including:

[0035] An acquisition module, used for acquiring audio data and a video frame set from the video to be processed;

[0036] A sensitive word processing module is used to perform sensitive word detection on the audio text corresponding to the audio data, determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, and perform desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data;

[0037] A sensitive area processing module, used to determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set;

[0038] The audio and video merging module is used to merge the target audio data with the target video data synthesized by the target video frame set to obtain a video file after desensitizing the video to be processed.

[0039] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including: a memory and a processor;

[0040] The memory is connected to the processor and is used to store programs;

[0041] The processor is used to implement the above-mentioned video processing method by running the program in the memory.

[0042] According to a fourth aspect of an embodiment of the present application, a storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned video processing method is implemented.

[0043] According to a fifth aspect of an embodiment of the present application, a computer program product is provided, comprising computer program instructions, which, when executed by a processor, enable the processor to implement the above-mentioned video processing method.

[0044] The video processing method proposed in the present application obtains audio data and a video frame set from a video to be processed; performs sensitive word detection on the audio text corresponding to the audio data, determines the target sensitive words in the audio text and the time information of the target sensitive words in the audio data, and desensitizes the audio data according to the time information corresponding to the target sensitive words to obtain the target audio data; determines the sensitive content area of ​​each video frame in the video frame set, and desensitizes the sensitive content area to obtain the target video frame set; merges the target audio data and the target video frame set to synthesize the target video data to obtain a video file after the video to be processed is desensitized. The technical solution of the present application can automatically perform desensitization on the sensitive words in the audio data of the video to be processed and the sensitive content areas in each video frame in the video to be processed. Compared with manual desensitization of the video, the labor cost of desensitizing the video is reduced and the efficiency of desensitizing the video is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0046] Figure 1 A schematic diagram of a video processing method provided in an embodiment of the present application;

[0047] Figure 2 A schematic diagram of a processing flow for sensitive word detection provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of another processing flow of sensitive word detection provided in an embodiment of the present application;

[0049] Figure 4 A schematic diagram of a processing flow for sensitive content area detection provided in an embodiment of the present application;

[0050] Figure 5 A schematic diagram of the structure of a video processing device provided in an embodiment of the present application;

[0051] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The technical solution of the embodiment of the present application is applicable to the application scenario of video processing, and is specifically used in the application scenario of desensitizing the video and audio in the video. The technical solution of the embodiment of the present application can automatically desensitize the sensitive words in the audio data of the video to be processed and the sensitive content areas in each video frame in the video to be processed. Compared with manual desensitization of the video, the labor cost of desensitizing the video is reduced and the efficiency of desensitizing the video is improved.

[0053] As a channel for information transmission, the Internet has now been deeply integrated into different areas of social development. As a carrier of information transmission, video is inseparable from personal life and organizational development. Since video content is random and environmental changes may cause the original video content to not conform to current regulations, or there is a risk that the video recipient will maliciously exploit the privacy content after the video is disseminated, it is necessary to desensitize the video.

[0054] The current mainstream technical solution relies on post-production manual operations. That is, when the video producer knows that some content in the video does not meet the current dissemination standards, post-production professional production personnel will perform post-production editing and other processing on the video content to remove sensitive words or mosaics. However, manual desensitization of videos has high labor costs and low processing efficiency. In addition, when the video contains a lot of sensitive content or the workload of post-production professional production personnel is large, sensitive content may leak out, thereby affecting the accuracy of the video desensitization processing.

[0055] Based on this, the present application proposes a video processing method, which can automatically desensitize sensitive words in the audio data of the video to be processed and sensitive content areas in each video frame in the video to be processed, thereby solving the problem of high labor cost, low efficiency and accuracy in desensitizing video in the prior art.

[0056] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0057] Exemplary Methods

[0058] See also Figure 1 As shown, the embodiment of the present application proposes a video processing method. The method includes:

[0059] S101. Acquire audio data and a video frame set from a video to be processed.

[0060] In this embodiment, after obtaining a video that needs to remove sensitive content, the video is used as a video to be processed. The video to be processed includes audio stream information and video image information. In this embodiment, it is first necessary to extract the audio stream information from the video to be processed as audio data, and to extract video frames from the video to be processed to obtain a video frame set consisting of all video frames in the video to be processed.

[0061] In this embodiment, it is preferred to use ffmpeg (open source audio and video data processing library) to extract audio stream information from the video to be processed. It is also possible to use ffmpeg (open source audio and video data processing library) to extract all video frames from the video to be processed, and organize all video frames into a video frame set in chronological order.

[0062] S102: Perform sensitive word detection on the audio text corresponding to the audio data, determine the target sensitive words in the audio text and the time information of the target sensitive words in the audio data, and desensitize the audio data according to the time information corresponding to the target sensitive words to obtain the target audio data.

[0063] After extracting the audio data in the video to be processed through the above steps, it is necessary to perform desensitization processing on the sensitive content in the audio data. Specifically, it is necessary to first perform speech recognition on the audio data to identify the audio text corresponding to the audio data, and then perform sensitive word detection on the audio text corresponding to the audio data, and detect the sensitive words in the audio text corresponding to the audio data as the target sensitive words. In addition, the present embodiment can use an existing speech recognition system to perform speech recognition on the audio data. During the speech recognition process of the audio data, the speech recognition system can not only identify the audio text corresponding to the audio data, but also identify the time information corresponding to each character in the audio data, wherein the time information includes the start and end time of the character in the audio data. Therefore, after detecting the target sensitive words in the audio text corresponding to the audio data, the present embodiment needs to determine the time information of the target sensitive words in the audio data according to the time information corresponding to each character in the target sensitive words in the audio data, that is, the start and end time of the target sensitive words in the audio data.

[0064] After detecting the target sensitive words in the audio text corresponding to the audio data and the time information corresponding to the target sensitive words, this embodiment needs to desensitize the target sensitive words in the audio data according to the time information of the target sensitive words in the audio data, so as to obtain the target audio data, that is, according to the start and end time of the target sensitive words in the audio data, the audio segments corresponding to the start and end time of the target sensitive words in the target audio data are desensitized.

[0065] Specifically, sensitive word detection is performed on the audio text corresponding to the audio data, and a sensitive word library can be pre-constructed, which contains several sensitive words. Each word in the audio text corresponding to the audio data is matched and analyzed with the sensitive words in the sensitive word library, and the words in the audio text corresponding to the audio data that match any sensitive word in the sensitive word library are used as target sensitive words. Among them, if the similarity of two words reaches a preset similarity threshold, it can be judged that the two words match, or semantic analysis is performed on the two words respectively. If the two words have the same semantics, it means that the two words match. In addition, sensitive word detection is performed on the audio text corresponding to the audio data, and semantic analysis can also be performed on each sentence in the audio text corresponding to the audio data to determine whether there are semantically incoherent words in the sentence. Semantically incoherent words may be replaced by other words by the speaker in order to avoid saying sensitive words, but replacing sensitive words with other words may cause the sentence to be semantically incoherent and incoherent. Therefore, in this embodiment, semantically incoherent words in the sentence can be used as target sensitive words in the audio text corresponding to the audio data.

[0066] Further, in this embodiment, after determining the target sensitive words in the audio text and the time information of the target sensitive words in the audio data, the specific steps of desensitizing the audio data according to the time information corresponding to the target sensitive words are as follows:

[0067] The audio segment in the time information corresponding to the target sensitive word in the audio data is replaced with the preset shielded audio to obtain the target audio data.

[0068] The present embodiment pre-sets shielded audio, which can be user-defined audio. Among them, the time information corresponding to the target sensitive word is the start and end time corresponding to the target sensitive word. According to the start and end time corresponding to the target sensitive word, the audio segment involved in the target sensitive word in the audio data can be determined, and then the audio segment involved in the target sensitive word in the audio data is replaced with the preset shielded audio, thereby realizing the desensitization of the audio data. In addition, the present embodiment can also directly mute the audio segment involved in the target sensitive word in the audio data, and can also realize the desensitization of the audio data. After replacing the audio of all target sensitive words in the audio data or silencing the audio, the final audio data is used as the target audio data after the audio data is desensitized.

[0069] S103: Determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set.

[0070] After extracting a video frame set containing all video frames from the video to be processed, this embodiment needs to perform sensitive content detection on each video frame in the video frame set, so as to detect the sensitive content area in each video frame, and then desensitize the sensitive content area in each video frame, and then sort the desensitized video frames in chronological order to form a target video frame set.

[0071] Specifically, this embodiment determines the sensitive content area of ​​each video frame in the video frame set, and can pre-build a sensitive content description library, which stores a number of sensitive content description information, wherein the sensitive content description information may include the category of sensitive content (for example, animals, flowers, etc.), and may also include the description content of sensitive content (for example, the description content of actions), etc. This embodiment first needs to perform image analysis and understanding on the video frame to detect the target objects contained in the video frame, the specific category to which each target object belongs, the description content of each target object, and the coordinate information of each target object in the video frame, etc. This embodiment uses the specific category and description content of the target object in the video frame as the description information of the target object. This embodiment matches and analyzes the description information of all target objects contained in the video frame with the sensitive content description information contained in the sensitive content description library, and uses the coordinate information of the target object whose description information in the video frame matches the sensitive content description information contained in the sensitive content description library as the coordinate information of the sensitive content area in the video frame. In addition, this embodiment can also make a tendency judgment on the description information of the target object contained in the video frame to determine whether the target object contained in the video frame is negative content, and then directly use the entire picture of the video frame as the sensitive content area. Among them, this embodiment can use a pre-trained neural network model to perform image understanding on the feature information corresponding to the video frame for image analysis and understanding of the video frame. This embodiment preferably uses a large language model (LLM).

[0072] In this embodiment, the above-mentioned sensitive content area detection can be performed on each video frame in the video frame set, or the video frame set can be divided first, and the video frames whose similarity reaches a preset similarity threshold (i.e., the similarity is relatively large) are divided into a set, thereby obtaining a plurality of sub-video frame sets divided from the video frame set, and any two video frames in the sub-video frame set are video frames whose similarity reaches a preset similarity threshold (i.e., the similarity is relatively large). Then, any one video frame is directly selected from the sub-video frame set to perform the above-mentioned sensitive content area detection, and the coordinate information of the detected sensitive content area is used as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set. In this way, the coordinate information of the sensitive content area of ​​all video frames in each sub-video frame set can be determined, and this method can reduce the number of times the sensitive content area is detected and improve the detection efficiency.

[0073] Specifically, after determining the sensitive content area of ​​each video frame in the video frame set, this embodiment performs desensitization processing on the sensitive content area to obtain a target video frame set, which specifically includes the following steps:

[0074] First, all pixels in the sensitive content area of ​​the video frames in the video frame set are differentially privacy processed to obtain a target video frame corresponding to the video frame.

[0075] After obtaining the coordinate information of the sensitive content area of ​​the video frame in the video frame set, this embodiment performs differential privacy processing on all pixels within the coordinate information range to achieve desensitization of the sensitive content area, thereby obtaining a desensitized video frame, and using the video frame obtained after the desensitization of the video frame as the target video frame corresponding to the video frame. Among them, differential privacy processing for all pixels within the coordinate information range can be performed by mosaicking the image area within the coordinate information of the sensitive content area. For example, the average color value of all pixels within the coordinate information of the sensitive content area can be calculated, and then the color values ​​of each pixel within the coordinate information of the sensitive content area are replaced with the average color value, thereby achieving differential privacy processing. The image area within the coordinate information of the sensitive content area can also be divided into several small areas, and then the color average value of all pixels in the small area is calculated for each small area, and the color value of each pixel in the small area is replaced with the color average value corresponding to the small area. Dividing the sensitive content area into multiple small areas for differential privacy processing can improve the visual effect of the target video frame after differential privacy processing, so that the sensitive content area after differential privacy processing and the area other than the sensitive content area in the video frame are better connected.

[0076] Second, the target video frames corresponding to the respective video frames in the video frame set are sequentially combined into a target video frame set.

[0077] This embodiment performs differential privacy processing on the sensitive content area of ​​each video frame in the video frame set through the above steps. For the video frame, all sensitive content areas in the video frame can be desensitized to obtain the target video frame corresponding to the video frame. After obtaining the target video frame corresponding to each video frame in the video frame set, the target video frames corresponding to each video frame are sorted according to the time sequence of each video frame in the video to be processed, and all the sorted target video frames are combined into a target video frame set.

[0078] S104: Merge the target audio data and the target video data synthesized by the target video frame set to obtain a video file after desensitizing the video to be processed.

[0079] Through the above steps, after the audio data and the video frame set in the video to be processed are desensitized, the target audio data corresponding to the audio data and the target video frame set corresponding to the video frame set are obtained. Then, all the target video frames contained in the target video frame set are merged in order to form target video stream information, which is the video stream information after the video stream information in the video to be processed is desensitized. Finally, the target audio data after desensitization and the target video stream information after desensitization are merged to form a complete video file with audio, that is, to obtain a video file after the video to be processed is desensitized. Among them, the target audio data and the target video stream information are merged preferably using ffmpeg to suppress the target audio data into the target video stream information to obtain a desensitized video file.

[0080] In this embodiment, the execution order of step S102 and step S103 is not limited. Step S102 can be executed first and then step S103, or step S103 can be executed first and then step S102, or step S102 and step S103 can be executed simultaneously.

[0081] Through the above introduction, it can be seen that the video processing method proposed in the embodiment of the present application obtains audio data and a video frame set from the video to be processed; performs sensitive word detection on the audio text corresponding to the audio data, determines the target sensitive words in the audio text and the time information of the target sensitive words in the audio data, and desensitizes the audio data according to the time information corresponding to the target sensitive words to obtain the target audio data; determines the sensitive content area of ​​each video frame in the video frame set, and desensitizes the sensitive content area to obtain the target video frame set; merges the target audio data and the target video frame set to synthesize the target video data to obtain the video file after the video to be processed is desensitized. The technical solution of this embodiment can automatically perform desensitization on the sensitive words in the audio data of the video to be processed and the sensitive content areas in each video frame in the video to be processed. Compared with manual desensitization of the video, the labor cost of desensitizing the video is reduced and the efficiency of desensitizing the video is improved.

[0082] As an optional implementation, see Figure 2 As shown, in another embodiment of the present application, in step S102 in the above embodiment, sensitive word detection is performed on the audio text corresponding to the audio data to determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, which specifically includes the following steps:

[0083] S201: Perform matching analysis on the audio text corresponding to the audio data and the sensitive words in the pre-built sensitive word library, and use the words in the audio text that match the sensitive words as target sensitive words.

[0084] According to actual needs, this embodiment constructs a sensitive word library, which contains several sensitive words. This embodiment can also classify the sensitive word library to obtain multiple sensitive word libraries of different categories, such as economic, medical, political, etc. Each sensitive word library can pre-store several sensitive words of this category. In order to improve the detection accuracy of the target sensitive words contained in the audio text corresponding to the audio data, this embodiment can also expand the sensitive word library of each category. According to the category of the sensitive word library, the words that are similar or related to the sensitive words in the sensitive word library are stored in the sensitive word library, so as to achieve the effect of enriching the content of the sensitive word library. In this embodiment, a neural network model can be used to generate words that are similar or related to sensitive words, preferably using a large language model (LLM) to achieve it.

[0085] This embodiment requires dividing the audio text corresponding to the audio data according to the sentence dimension to obtain a number of sentences contained in the audio text corresponding to the audio data, and then matching and analyzing the sentences in the audio text with each sensitive word in the sensitive word library to detect whether there are words matching each sensitive word in the sentences of the audio text. If so, the words matching the sensitive words in the sentences of the audio text are determined as the target sensitive words in the audio text. The target sensitive words in each sentence are determined according to the above method for each sentence in the audio text, thereby obtaining all the target sensitive words contained in the audio text corresponding to the audio data.

[0086] S202: Obtain time information of the target sensitive word in the audio data as time information corresponding to the target sensitive word.

[0087] When the speech recognition system is used to perform speech recognition on audio data, this embodiment can not only recognize the audio text corresponding to the audio data, but also recognize the time information of each character in the audio text in the audio data, wherein the time information of the character in the audio data includes the start time and end time of the character in the audio data.

[0088] After determining the target sensitive words included in the audio text, this embodiment can determine the time information of each target sensitive word included in the audio text in the audio data according to the time information of each character in the audio text in the audio data, and use the time information of the target sensitive word in the audio data as the time information corresponding to the target sensitive word. The time information corresponding to the target sensitive word includes the start time and end time of the target sensitive word in the audio data.

[0089] As an optional implementation, see Figure 2 As shown, in another embodiment of the present application, in step S102 in the above embodiment, sensitive word detection is performed on the audio text corresponding to the audio data to determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, specifically including the following steps:

[0090] S301. Perform semantic analysis on each sentence in the audio text corresponding to the audio data, determine semantically incoherent words in each sentence, and output the sentence after marking the semantically incoherent words, so that the user can confirm whether the semantically incoherent words are sensitive words.

[0091] This embodiment needs to divide the audio text corresponding to the audio data according to the sentence dimension to obtain a number of sentences contained in the audio text corresponding to the audio data, and then perform semantic analysis on each sentence in the audio text corresponding to the audio data, and analyze whether there are semantically incoherent words in each sentence in combination with the context. If there are semantically incoherent words in the sentence, the semantically incoherent words in the sentence are marked, and then the sentence is output to the user for viewing. The user views the semantically incoherent words marked in the sentence and determines whether the semantically incoherent words are sensitive words, and feeds back the confirmation result as feedback information, wherein, if the user views the semantically incoherent words marked in the sentence and confirms that the marked semantically incoherent words are sensitive words, the feedback information indicates that the semantically incoherent words are sensitive words, and if the user views the semantically incoherent words marked in the sentence and confirms that the marked semantically incoherent words are not sensitive words, the feedback information indicates that the semantically incoherent words are not sensitive words.

[0092] In this embodiment, the specific steps of performing semantic analysis on each sentence in the audio text corresponding to the audio data and determining the semantically incoherent words in each sentence include:

[0093] First, each sentence in the audio text corresponding to the audio data is preprocessed separately, including steps such as removing irrelevant characters and segmenting sentences, so as to form separate words for better understanding and analysis later. For example, for the sentence "Today the weather is very good, I went to the park.", after preprocessing, the words "today", "weather", "very good", "I", "went", and "park" are obtained.

[0094] Secondly, this embodiment can use a neural network model to perform semantic analysis on sentences, including syntactic analysis and semantic understanding of words in context, preferably using a large language model (LLM). Through syntactic analysis, the subject, predicate, object and other components in the sentence, as well as the grammatical relationship between them, are identified, which helps the large language model understand the sentence structure. By combining the lexical semantic understanding with the context, the large language model refers to the context of the sentence, performs in-depth analysis on the semantics of the words, and evaluates the semantic relationship between the words.

[0095] Secondly, the neural network model is used to judge the coherence of the sentence, including semantic consistency and logical coherence checks, preferably using a large language model (LLM). Semantic consistency checks, such as whether the subject and predicate, object and predicate actions match. Logical coherence checks, whether the words in the sentence are arranged in a reasonable logical order, whether there are logical connections such as cause and effect, transition, etc.

[0096] Finally, after the above processing, the neural network model can determine the coherence of the sentence and obtain a coherence score for each word. If the score is lower than the preset coherence threshold, the word is considered to be semantically incoherent.

[0097] S302: If the user's feedback information indicates that the semantically incoherent word is a sensitive word, the semantically incoherent word is used as a target sensitive word, and time information of the target sensitive word in the audio data is obtained as the time information corresponding to the target sensitive word.

[0098] If the user's feedback information indicates that the semantically incoherent words are sensitive words, the semantically incoherent words in the sentence are used as the target sensitive words in the sentence. In this way, the target sensitive words in each sentence in the audio text corresponding to the audio data can be determined. Then, the time information of the target sensitive words in the audio data is obtained as the time information corresponding to the target sensitive words. The method of obtaining the time information of the target sensitive words in the audio data is the same as the method of obtaining the time information of the target sensitive words in the audio data in the above embodiment, and this embodiment will not be described in detail.

[0099] If the user's feedback information indicates that the semantically incoherent word is not a sensitive word, the semantically incoherent word is not used as a target sensitive word.

[0100] As a specific embodiment, sensitive word detection for the audio text corresponding to the audio data can be performed by Figure 2 The embodiment described in the above can detect sensitive words in audio text, and can also be used Figure 3 The embodiment described in the above can detect sensitive words in audio text, and can also be used Figure 2 The embodiments described in Figure 3Sensitive word detection is performed by combining the embodiments described in Figure 2 The embodiment described in the first round of sensitive word detection is performed on the audio text, and then the Figure 3 The embodiment described in performs two rounds of sensitive word detection on the audio text.

[0101] As an optional implementation, see Figure 4 As shown, in another embodiment of the present application, it is disclosed that in step S103 in the above embodiment, determining the sensitive content area of ​​each video frame in the video frame set specifically includes the following steps:

[0102] S401 . Divide video frames in a video frame set into at least one sub-video frame set according to similarities between video frames.

[0103] This embodiment needs to divide the video frames in the video frame set into at least one sub-video frame set according to the similarity between the video frames in the video frame set, wherein the similarity between any two video frames in the sub-video frame set reaches a preset similarity threshold. If the similarity between two video frames reaches the preset similarity threshold, it means that the two video frames are similar video frames. If the similarity between the two video frames does not reach the preset similarity threshold, it means that the two video frames are dissimilar video frames.

[0104] In this embodiment, the similarity between two video frames can be calculated by using the cosine similarity calculation method, or the Average Hash algorithm. Among them, the Average Hash algorithm is preferably used. The specific steps of using the Average Hash algorithm to calculate the similarity between two video frames are as follows:

[0105] First, the two video frames are scaled to 32×32 resolution respectively, which reduces the number of pixels in the video frames and improves the calculation efficiency of the similarity.

[0106] Second, the scaled video frame is grayed to obtain a grayed video frame.

[0107] Third, for the grayscale processed video frames, the grayscale average values ​​of all pixels are calculated according to the grayscale values ​​of all pixels in the video frames, so as to obtain the grayscale average values ​​corresponding to each grayscale processed video frame.

[0108] Fourth, for the grayscale processed video frame, the grayscale value of each pixel in the video frame is compared with the grayscale average value corresponding to the video frame. If the grayscale value of the similarity in the video frame is greater than or equal to the grayscale average value corresponding to the video frame, it is recorded as 1. If the grayscale value of the similarity in the video frame is less than the grayscale average value corresponding to the video frame, it is recorded as 0, thereby obtaining the 32×32-bit hash value corresponding to the video frame. In this way, the hash values ​​corresponding to two video frames can be determined.

[0109] Fifth, the Hamming distance between the two image frames is calculated using the hash values ​​corresponding to the two video frames.

[0110] Sixth, the Hamming distance between the two video frames is converted into a percentage similarity. If the similarity reaches a preset similarity threshold (e.g., 90%), the two video frames are determined to be similar. If the similarity does not reach the preset similarity threshold (e.g., 90%), the two video frames are determined to be dissimilar.

[0111] Specifically, this embodiment divides the video frames in the video frame set into at least one sub-video frame set according to the similarity between the video frames, first starting from the first video frame in the video frame set, and calculating the similarity with each subsequent video frame in sequence until the calculated similarity is lower than the preset similarity threshold, then pausing, and then dividing the first video frame and all the video frames whose calculated similarity with the first video frame is not lower than the preset similarity threshold into the same sub-video frame set, and then continuing to divide the sub-video frame set in the above manner starting from the first video frame in the remaining video frames of the video frame set, until all the video frames in the video frame set are divided and then stopping. The similarity between any two video frames in the sub-video frame set divided in this way reaches the preset similarity threshold, that is, all the video frames in the sub-video frame set are similar video frames.

[0112] S402: Perform image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame.

[0113] After dividing the video frame set into at least one sub-video frame set through the above steps, it is first necessary to select a representative video frame corresponding to the sub-video frame set from each sub-video frame set. In this embodiment, any video frame in the sub-video frame set can be selected as the representative video frame corresponding to the sub-video frame set. Then, image analysis is performed on the representative video frame in the sub-video frame set to determine the coordinate information of the sensitive content area in the representative video frame. The specific steps are as follows:

[0114] First, image analysis is performed on a representative video frame in the sub-video frame set to determine description information of a target object and coordinate information of the target object contained in the representative video frame.

[0115] This embodiment can use a neural network model to perform image analysis on representative video frames in a set of sub-video frames, preferably a large language model (LLM). Before performing image analysis on representative video frames, first, the representative video frames need to be preprocessed, for example, the video frame size is changed to meet the model input requirements, noise data in the video frame is removed to improve the quality of the video frame, etc. Secondly, the features in the preprocessed representative video frame (such as color features, texture features, shape features, etc. in the video frame) are extracted to obtain feature information of the representative video frame. These feature information can capture local information and global information in the video frame, and provide assistance for subsequent image analysis. Thirdly, based on the preprocessed representative video frames and the feature information of the representative video frames, a large language model (LLM) is used to perform image understanding analysis on the representative video frames, so as to determine the target objects contained in the representative video frames, and locate the target objects contained in the representative video frames to obtain the coordinate information of each target object. In addition, a large language model (LLM) is also needed to analyze the specific category (such as people, cats, plants, etc.) of each target object and the description content of each target object, and use the specific category and description content of the target object as the description information of the target object.

[0116] Second, coordinate information representing a sensitive content area in the video frame is determined based on the description information representing the target object contained in the video frame and the coordinate information of the target object.

[0117] Specifically, this embodiment determines the sensitive content area of ​​each video frame in the video frame set, and can pre-build a sensitive content description library, which stores a number of sensitive content description information, wherein the sensitive content description information may include the category of sensitive content (for example, animals, flowers, etc.), and may also include the description content of sensitive content (for example, the description content of actions), etc. This embodiment first needs to perform image analysis and understanding on the video frame to detect the target objects contained in the video frame, the specific category to which each target object belongs, the description content of each target object, and the coordinate information of each target object in the video frame, etc. This embodiment uses the specific category and description content of the target object in the video frame as the description information of the target object. This embodiment matches and analyzes the description information of all target objects contained in the video frame with the sensitive content description information contained in the sensitive content description library, and uses the coordinate information of the target object whose description information in the video frame matches the sensitive content description information contained in the sensitive content description library as the coordinate information of the sensitive content area in the video frame. In addition, this embodiment can also make a tendency judgment on the description information of the target object contained in the video frame to determine whether the target object contained in the video frame is negative content, and then directly use the entire picture of the video frame as the sensitive content area. Among them, this embodiment can use a pre-trained neural network model to perform image understanding on the feature information corresponding to the video frame for image analysis and understanding of the video frame. This embodiment preferably uses a large language model (LLM).

[0118] First, according to the description information of the target object contained in the representative video frame, it is determined whether the representative video frame contains negative content.

[0119] This embodiment performs a tendency judgment on the representative video frame based on the description information of all target objects contained in the representative video frame to determine whether the representative video frame contains negative content. Specifically, this embodiment can use the reasoning and judgment ability of a large language model to analyze and judge the description information of all target objects contained in the representative video frame to determine whether the representative video frame contains negative content.

[0120] Secondly, if the representative video frame contains negative content, the coordinate information of each pixel point on the outer edge of the representative video frame is used as the coordinate information of the sensitive content area in the representative video frame.

[0121] If it is determined that the representative video frame contains negative content, then the entire representative video frame is directly judged to be sensitive content. At this time, the coordinate information of each pixel point on the outer edge of the representative video frame is used as the coordinate information of the sensitive content area in the representative video frame.

[0122] Finally, if the representative video frame does not contain negative content, the description information of the target object contained in the representative video frame is matched and analyzed with the sensitive description information in the pre-built sensitive image description library, and the coordinate information of the target object corresponding to the description information in the representative video frame that matches the sensitive description information is used as the coordinate information of the sensitive content area in the representative video frame.

[0123] If it is determined that the representative video frame contains negative content, the reasoning and judgment ability of the large language model is used again to determine whether the description information of the target object contained in the representative video frame matches the sensitive description information in the pre-built sensitive image description library. The target objects whose description information in the representative video frame matches any sensitive description information in the pre-built sensitive image description library are all regarded as sensitive content, and the area where the target object belonging to the sensitive content is located is regarded as the sensitive content area, and the coordinate information of the target object of the sensitive content is used as the coordinate information of the sensitive content area.

[0124] In addition, when the present embodiment determines that the description information of the target object contained in the representative video frame matches the sensitive description information in the sensitive image description library, the target object is first marked in the representative video frame, and then the marked representative video frame is output to the user for viewing, so that the user can determine whether the target object marked in the representative video frame is sensitive content. If the user's feedback information indicates that the target object marked in the representative video frame is sensitive content, the target object marked in the representative video frame is treated as sensitive content. By treating the target object as sensitive content after the user confirms, misjudgment can be avoided and the accuracy of determining the sensitive content area can be improved.

[0125] S403: Use the coordinate information of the sensitive content area in the representative video frame as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set to which the representative video frame belongs.

[0126] In this embodiment, since all video frames included in the sub-video frame set are similar video frames, the sensitive content areas of all video frames included in the sub-video frame set are also similar, and the coordinate information of the sensitive content area of ​​any video frame in the sub-video frame set can be used as the coordinate information of the sensitive content areas of all video frames in the sub-video frame set. Therefore, in this embodiment, the coordinate information of the sensitive content area in the representative video frame of the determined sub-video frame set is used as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set to which the representative video frame belongs, so that the coordinate information of the sensitive content areas of all video frames in the video frame set can be determined.

[0127] Exemplary Devices

[0128] Accordingly, the present application also provides a video processing device, see Figure 5 As shown, the device comprises:

[0129] An acquisition module 100 is used to acquire audio data and a video frame set from a video to be processed;

[0130] The sensitive word processing module 110 is used to perform sensitive word detection on the audio text corresponding to the audio data, determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, and perform desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data;

[0131] The sensitive area processing module 120 is used to determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set;

[0132] The audio and video merging module 130 is used to merge the target audio data with the target video frame set to synthesize the target video data to obtain a video file after desensitizing the video to be processed.

[0133] From the above introduction, it can be seen that the video processing device proposed in the embodiment of the present application can automatically desensitize sensitive words in the audio data of the video to be processed and sensitive content areas in each video frame in the video to be processed. Compared with manual desensitization of the video, it reduces the labor cost of desensitization of the video and improves the efficiency of desensitization of the video.

[0134] As an optional implementation, another embodiment of the present application discloses that the sensitive word processing module 110 includes: a sensitive word matching unit and a time determination unit.

[0135] A sensitive word matching unit is used to match and analyze the audio text corresponding to the audio data with the sensitive words in the pre-built sensitive word library, and use the words in the audio text that match the sensitive words as target sensitive words;

[0136] The time determination unit is used to obtain the time information of the target sensitive word in the audio data as the time information corresponding to the target sensitive word.

[0137] As an optional implementation, another embodiment of the present application discloses that the sensitive word processing module 110 further includes: a semantic analysis unit.

[0138] A semantic analysis unit, configured to perform semantic analysis on each sentence in the audio text corresponding to the audio data, determine semantically incoherent words in each sentence, and output the sentence after marking the semantically incoherent words, so that the user can confirm whether the semantically incoherent words are sensitive words;

[0139] The time determination unit is also used to, if the user's feedback information indicates that the semantically incoherent word is a sensitive word, use the semantically incoherent word as a target sensitive word, and obtain the time information of the target sensitive word in the audio data as the time information corresponding to the target sensitive word.

[0140] As an optional implementation, it is disclosed in another embodiment of the present application that the sensitive word processing module 110 also includes: an audio desensitization processing unit.

[0141] The audio desensitization processing unit is used to replace the audio segment in the time information corresponding to the target sensitive word in the audio data with the preset shielded audio to obtain the target audio data.

[0142] As an optional implementation, another embodiment of the present application discloses that the sensitive area processing module 120 includes: a dividing unit, a first sensitive area determining unit and a second sensitive area determining unit.

[0143] A division unit, configured to divide the video frames in the video frame set into at least one sub-video frame set according to the similarity between the video frames, wherein the similarity between any two video frames in the sub-video frame set reaches a preset similarity threshold;

[0144] A first sensitive area determination unit is used to perform image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame; wherein the representative video frame in the sub-video frame set is any video frame in the sub-video frame set;

[0145] The second sensitive area determination unit is used to use the coordinate information representing the sensitive content area in the video frame as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set to which the representative video frame belongs.

[0146] As an optional implementation, another embodiment of the present application discloses that the first sensitive area determination unit includes: a picture analysis unit and a coordinate determination unit.

[0147] A picture analysis unit, used to perform picture analysis on a representative video frame in the sub-video frame set, and determine description information of a target object and coordinate information of the target object contained in the representative video frame;

[0148] The coordinate determination unit is used to determine the coordinate information of the sensitive content area in the video frame according to the description information of the target object contained in the video frame and the coordinate information of the target object.

[0149] As an optional implementation, another embodiment of the present application discloses that the coordinate determination unit is specifically used for:

[0150] Determining whether the representative video frame contains negative content according to the description information of the target object contained in the representative video frame;

[0151] If the representative video frame contains negative content, the coordinate information of each pixel point on the outer edge of the representative video frame is used as the coordinate information of the sensitive content area in the representative video frame.

[0152] As an optional implementation, another embodiment of the present application discloses that the coordinate determination unit is further used for:

[0153] If the representative video frame does not contain negative content, the description information of the target object contained in the representative video frame is matched and analyzed with the sensitive description information in the pre-built sensitive image description library, and the coordinate information of the target object corresponding to the description information in the representative video frame that matches the sensitive description information is used as the coordinate information of the sensitive content area in the representative video frame.

[0154] As an optional implementation, another embodiment of the present application discloses that the sensitive region processing module 120 includes: a region desensitization processing unit. The region desensitization processing unit is specifically used to:

[0155] Perform differential privacy processing on all pixels in the sensitive content area of ​​the video frame in the video frame set to obtain a target video frame corresponding to the video frame;

[0156] The target video frames corresponding to the respective video frames in the video frame set are sequentially combined into a target video frame set.

[0157] The video processing device provided in this embodiment belongs to the same application concept as the video processing method provided in the above embodiments of this application, can execute the video processing method provided in any of the above embodiments of this application, and has the corresponding functional modules and beneficial effects of executing the video processing method. For technical details not fully described in this embodiment, please refer to the specific processing content of the video processing method provided in the above embodiments of this application, which will not be repeated here.

[0158] Exemplary Electronic Devices

[0159] Another embodiment of the present application also provides an electronic device, see Figure 6 As shown, the device includes:

[0160] Memory 200 and processor 210;

[0161] The memory 200 is connected to the processor 210 and is used to store programs;

[0162] The processor 210 is used to implement the video processing method disclosed in any of the above embodiments by running the program stored in the memory 200.

[0163] Specifically, the electronic device may further include: a bus, a communication interface 220 , an input device 230 and an output device 240 .

[0164] The processor 210, the memory 200, the communication interface 220, the input device 230 and the output device 240 are connected to each other via a bus.

[0165] A bus may include a pathway that transfers information between components of a computer system.

[0166] The processor 210 may be a general-purpose processor, such as a general-purpose central processing unit (CPU), a microprocessor, etc., or an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the scheme of the present invention. It may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0167] The processor 210 may include a main processor, and may also include a baseband chip, a modem, and the like.

[0168] The memory 200 stores a program for executing the technical solution of the present invention, and may also store an operating system and other key services. Specifically, the program may include a program code, and the program code includes a computer operation instruction. More specifically, the memory 200 may include a read-only memory (ROM), other types of static storage devices that can store static information and instructions, a random access memory (RAM), other types of dynamic storage devices that can store information and instructions, a disk storage, a flash, and the like.

[0169] The input device 230 may include a device for receiving data and information input by a user, such as a keyboard, a mouse, a camera, a scanner, a light pen, a voice input device, a touch screen, a pedometer, or a gravity sensor.

[0170] Output device 240 may include devices that allow information to be output to a user, such as a display screen, printer, speaker, etc.

[0171] The communication interface 220 may include any transceiver or the like to communicate with other devices or communication networks, such as Ethernet, Radio Access Network (RAN), Wireless Local Area Network (WLAN), etc.

[0172] The processor 210 executes the program stored in the memory 200 and calls other devices, and can be used to implement each step of any video processing method provided in the above embodiments of the present application.

[0173] Exemplary computer program products and storage media

[0174] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the video processing method according to various embodiments of the present application described in the above-mentioned "Exemplary Method" section of this specification.

[0175] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0176] In addition, an embodiment of the present application may also be a storage medium on which a computer program is stored, and the computer program is executed by a processor to execute the steps of the video processing method according to various embodiments of the present application described in the above "Exemplary Method" section of this specification.

[0177] For the aforementioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0178] It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the device 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.

[0179] The steps in the methods of each embodiment of the present application can be adjusted in order, combined and deleted according to actual needs, and the technical features recorded in each embodiment can be replaced or combined.

[0180] The modules and sub-modules in the devices and terminals of the various embodiments of the present application can be combined, divided and deleted according to actual needs.

[0181] In the several embodiments provided in the present application, it should be understood that the disclosed terminals, devices and methods can be implemented in other ways. For example, the terminal embodiments described above are only schematic, for example, the division of modules or submodules is only a logical function division, and there may be other division methods in actual implementation, for example, multiple submodules or modules can be combined or integrated into another module, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0182] The modules or submodules described as separate components may or may not be physically separated, and the components of the modules or submodules may or may not be physical modules or submodules, that is, they may be located in one place, or they may be distributed on multiple network modules or submodules. Some or all of the modules or submodules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] In addition, each functional module or submodule in each embodiment of the present application may be integrated into one processing module, or each module or submodule may exist physically separately, or two or more modules or submodules may be integrated into one module. The above-mentioned integrated modules or submodules may be implemented in the form of hardware or in the form of software functional modules or submodules.

[0184] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0185] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly by hardware, software units executed by a processor, or a combination of the two. The software units may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0186] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0187] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A video processing method, characterized in that: include: Obtain audio data and a set of video frames from the video to be processed; Performing sensitive word detection on the audio text corresponding to the audio data, determining the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, and performing desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data; Determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set; The target audio data is combined with the target video data synthesized by the target video frame set to obtain a video file after desensitizing the video to be processed.

2. The method according to claim 1, characterized in that Performing sensitive word detection on the audio text corresponding to the audio data to determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data includes: Performing a matching analysis on the audio text corresponding to the audio data and the sensitive words in the pre-built sensitive word library, and taking the words in the audio text that match the sensitive words as the target sensitive words; The time information of the target sensitive word in the audio data is obtained as the time information corresponding to the target sensitive word.

3. The method according to claim 1, characterized in that Performing sensitive word detection on the audio text corresponding to the audio data to determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data includes: Performing semantic analysis on each sentence in the audio text corresponding to the audio data, determining semantically incoherent words in each sentence, and outputting the sentence after marking the semantically incoherent words, so that the user can confirm whether the semantically incoherent words are sensitive words; If the user's feedback information indicates that the semantically incoherent word is a sensitive word, the semantically incoherent word is used as a target sensitive word, and time information of the target sensitive word in the audio data is obtained as the time information corresponding to the target sensitive word.

4. The method according to claim 1, characterized in that: Desensitizing the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data includes: The audio segment in the time information corresponding to the target sensitive word in the audio data is replaced with a preset shielded audio to obtain the target audio data.

5. The method according to claim 1, characterized in that Determining the sensitive content area of ​​each video frame in the video frame set includes: According to the similarity between the video frames, the video frames in the video frame set are divided into at least one sub-video frame set, and the similarity between any two video frames in the sub-video frame set reaches a preset similarity threshold; Performing image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame; wherein the representative video frame in the sub-video frame set is any video frame in the sub-video frame set; The coordinate information of the sensitive content area in the representative video frame is used as the coordinate information of the sensitive content area of ​​each video frame in the sub-video frame set to which the representative video frame belongs.

6. The method according to claim 5, characterized in that Performing image analysis on a representative video frame in the sub-video frame set to determine coordinate information of a sensitive content area in the representative video frame includes: Performing picture analysis on a representative video frame in the sub-video frame set to determine description information of a target object and coordinate information of the target object contained in the representative video frame; The coordinate information of the sensitive content area in the representative video frame is determined according to the description information of the target object contained in the representative video frame and the coordinate information of the target object.

7. The method according to claim 6, characterized in that Determining coordinate information of a sensitive content area in the representative video frame according to description information of the target object contained in the representative video frame and coordinate information of the target object includes: Determining whether the representative video frame contains negative content according to the description information of the target object contained in the representative video frame; If the representative video frame contains negative content, the coordinate information of each pixel point on the outer edge of the representative video frame is used as the coordinate information of the sensitive content area in the representative video frame.

8. The method according to claim 7, characterized in that Determining the coordinate information of the sensitive content area in the representative video frame according to the description information of the target object contained in the representative video frame and the coordinate information of the target object, further comprising: If the representative video frame does not contain negative content, the description information of the target object contained in the representative video frame is matched and analyzed with the sensitive description information in a pre-built sensitive image description library, and the coordinate information of the target object corresponding to the description information in the representative video frame that matches the sensitive description information is used as the coordinate information of the sensitive content area in the representative video frame.

9. The method according to claim 1, characterized in that: Desensitizing the sensitive content area to obtain a target video frame set, including: Performing differential privacy processing on all pixels in the sensitive content area of ​​the video frames in the video frame set to obtain a target video frame corresponding to the video frame; The target video frames corresponding to the respective video frames in the video frame set are sequentially combined into a target video frame set.

10. A video processing device, characterized in that: include: An acquisition module, used for acquiring audio data and a video frame set from the video to be processed; A sensitive word processing module is used to perform sensitive word detection on the audio text corresponding to the audio data, determine the target sensitive word in the audio text and the time information of the target sensitive word in the audio data, and perform desensitization processing on the audio data according to the time information corresponding to the target sensitive word to obtain the target audio data; A sensitive area processing module, used to determine the sensitive content area of ​​each video frame in the video frame set, and perform desensitization processing on the sensitive content area to obtain a target video frame set; The audio and video merging module is used to merge the target audio data with the target video data synthesized by the target video frame set to obtain a video file after desensitizing the video to be processed.

11. An electronic device, characterized in that: include: Memory and processor; The memory is connected to the processor and is used to store programs; The processor is used to implement the video processing method according to any one of claims 1 to 9 by running the program in the memory.

12. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the processor, the video processing method according to any one of claims 1 to 9 is implemented.

13. A computer program product, characterized in that The method comprises computer program instructions, which, when executed by a processor, enable the processor to implement the video processing method according to any one of claims 1 to 9.

Images