Recording and broadcasting detection method and device, equipment, storage medium and product

Abstract

The embodiment of the invention provides a recording and broadcasting detection method and device, equipment, a storage medium and a product. According to the technical scheme provided by the embodiment of the invention, the sampling frame vectors of the plurality of live broadcast sampling frames of the to-be-detected live broadcast room are determined, the frame vector matching is performed on the sampling frame vectors in the recording and broadcasting vector library to obtain the first matching result, and the first recording and broadcasting detection result is determined according to the first matching result. When the first recorded broadcast detection result is that the to-be-detected live broadcast room is a non-recorded broadcast live broadcast room, frame vector matching is carried out on the sampling frame vector in a live broadcast vector library to obtain a second matching result, and a second recorded broadcast detection result is determined according to the second matching result, whether the to-be-detected live broadcast room is the recorded broadcast live broadcast room can be determined according to the second recorded broadcast detection result, the recorded broadcast detection accuracy is effectively ensured while the recorded broadcast detection efficiency is improved, the dependence of recorded broadcast detection on live broadcast audio and voice-to-text conversion capability can be effectively reduced, and the recorded broadcast detection effect is improved.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of computer technology, and in particular to a recording and broadcasting detection method, device, equipment, storage medium and product. Background Art

[0002] With the development of communication technology and audio and video technology, the application of live video is becoming more and more extensive, and the regulatory requirements for live video are becoming more and more stringent.

[0003] In some live video broadcasts, recorded videos are used for live broadcasts. Compared with live video broadcasts, recorded content lacks real-time interaction, and the audience cannot communicate with the anchor in real time, which loses the interactivity and sense of participation unique to live broadcasts. In addition, recorded content is easily tampered with, and there is a risk of spreading false information, which damages the audience's trust. Recorded broadcasts may also involve copyright issues. Unauthorized broadcasts will infringe on the legitimate rights and interests of the copyright owner. Therefore, it is necessary to detect and handle recorded broadcasts.

[0004] Existing recording and broadcast detection is generally performed through live broadcast text matching, that is, when the anchor starts live broadcasting, the live broadcast text is obtained through the speech-to-text service, and the live broadcast text is matched in the text library, and the recording and broadcast detection result is obtained based on the matching result. However, this recording and broadcast detection method is highly dependent on the live broadcast audio and speech-to-text capabilities. For scenarios where audio cannot be effectively converted into text, recording and broadcast detection often cannot be performed correctly, and the recording and broadcast detection effect is poor. Summary of the invention

[0005] The embodiments of the present application provide a recording and broadcasting detection method, device, equipment, storage medium and product to solve the technical problem that the recording and broadcasting detection method based on live text in the related technology is highly dependent on the live audio and speech-to-text capabilities, and the recording and broadcasting detection effect is poor, which can effectively improve the recording and broadcasting detection effect.

[0006] In a first aspect, an embodiment of the present application provides a recording and broadcasting detection method, including:

[0007] Acquire multiple live broadcast sampling frames of the live broadcast room to be detected, and determine sampling frame vectors of the multiple live broadcast sampling frames;

[0008] Performing frame vector matching on the sampled frame vector in a recording and broadcasting vector library to obtain a first matching result, and determining a first recording and broadcasting detection result according to the first matching result, wherein the recording and broadcasting vector library records a plurality of recording and broadcasting frame vectors of a plurality of recording and broadcasting live broadcasting rooms;

[0009] When the first recording and broadcast detection result is that the live broadcast room to be detected is a non-recorded and broadcast live broadcast room, frame vector matching is performed on the sampling frame vector in the live broadcast vector library to obtain a second matching result, and a second recording and broadcast detection result is determined according to the second matching result. The live broadcast vector library records multiple video frame vectors of multiple video live broadcast rooms, and the second recording and broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded and broadcast live broadcast room.

[0010] In a second aspect, an embodiment of the present application provides a recording and broadcasting detection device, including a vector acquisition module, a first matching module, and a second matching module, wherein:

[0011] The vector acquisition module is configured to acquire multiple live sampling frames of the live broadcast room to be detected, and determine the sampling frame vectors of the multiple live sampling frames;

[0012] The first matching module is configured to perform frame vector matching on the sampled frame vector in a recording and broadcasting vector library to obtain a first matching result, and determine a first recording and broadcasting detection result according to the first matching result, wherein the recording and broadcasting vector library records a plurality of recording and broadcasting frame vectors of a plurality of recording and broadcasting live broadcasting rooms;

[0013] The second matching module is configured to perform frame vector matching on the sampling frame vector in the live broadcast vector library to obtain a second matching result when the first recording and broadcast detection result indicates that the live broadcast room to be detected is a non-recorded and broadcast room, and determine a second recording and broadcast detection result based on the second matching result. The live broadcast vector library records multiple video frame vectors of multiple video live broadcast rooms, and the second recording and broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded and broadcast room.

[0014] In a third aspect, an embodiment of the present application provides a recording and broadcasting detection device, including: a memory and one or more processors;

[0015] The memory is used to store one or more programs;

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the recording and broadcasting detection method as described in the first aspect.

[0017] In a fourth aspect, an embodiment of the present application provides a non-volatile storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the recording and broadcasting detection method as described in the first aspect when executed by a computer processor.

[0018] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor of the device reads and executes the computer program from the computer-readable storage medium, so that the device performs the recording and broadcast detection method described in the first aspect.

[0019] The embodiment of the present application determines the sampling frame vectors of multiple live sampling frames of the live broadcast room to be detected, performs frame vector matching on the sampling frame vectors in the recording and broadcasting vector library to obtain a first matching result, and determines a first recording and broadcasting detection result based on the first matching result. When the first recording and broadcasting detection result is that the live broadcast room to be detected is not a recording and broadcasting live broadcast room, frame vector matching is performed on the sampling frame vectors in the live broadcast vector library to obtain a second matching result, and determines a second recording and broadcasting detection result based on the second matching result. Whether the live broadcast room to be detected is a recording and broadcasting live broadcast room can be determined based on the second recording and broadcasting detection result. First, a fast and accurate recording and broadcasting detection is performed in the recording and broadcasting vector library using the recording and broadcasting frame vector corresponding to the recording and broadcasting live broadcast room. When the recording and broadcasting live broadcast room is not detected in the recording and broadcasting vector library, a more comprehensive recording and broadcasting detection is performed in the live broadcast vector library. While improving the efficiency of recording and broadcasting detection, the accuracy of recording and broadcasting detection is effectively guaranteed, and the dependence of recording and broadcasting detection on live audio and speech-to-text capabilities can be effectively reduced, thereby improving the effect of recording and broadcasting detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of a recording and broadcasting detection method provided in an embodiment of the present application;

[0021] Figure 2 is a flow chart of another recording and broadcasting detection method provided in an embodiment of the present application;

[0022] Figure 3 It is a schematic diagram of a process for determining a first similarity matching score and a first temporal matching score provided in an embodiment of the present application;

[0023] Figure 4 It is a schematic diagram of a process for determining a second similarity matching score and a second timing matching score provided in an embodiment of the present application;

[0024] Figure 5 It is a structural schematic diagram of a recording and broadcasting detection device provided in an embodiment of the present application;

[0025] Figure 6 It is a structural diagram of a recording and broadcasting detection device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical scheme and advantages of the present application clearer, the specific embodiments of the present application are further described in detail below in conjunction with the accompanying drawings. It is understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for the convenience of description, only the part related to the present application but not all the contents are shown in the accompanying drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow chart describes each operation (or step) as a sequential process, many of the operations can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. The above process can be terminated when its operation is completed, but it can also have additional steps not included in the accompanying drawings. The above process can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0027] The recording and broadcast detection method provided in this application can be applied to the recording and broadcast detection of the live broadcast room in the live broadcast platform. It aims to first perform fast and accurate recording and broadcast detection using the recording and broadcast frame vector corresponding to the recording and broadcast live broadcast room in the recording and broadcast vector library, and then perform a more comprehensive recording and broadcast detection in the live broadcast vector library when the recording and broadcast room is not detected in the recording and broadcast vector library.

[0028] In the existing recording and broadcasting detection scheme, recording and broadcasting detection is usually performed through the live text matching of the live broadcast room. For example, when the anchor starts the live broadcast, the audio stream of the live broadcast room is pulled, and the live text corresponding to the audio stream is obtained through the speech-to-text service (ASR service), and then the live text is sampled and recalled and matched in the text library, and the recording and broadcasting detection result is obtained according to the recall matching result. In this detection mode, the recording and broadcasting detection is strongly dependent on the live audio and speech-to-text capabilities, and the recording and broadcasting detection failure will be caused for some scenes where the screen is hung up and recorded and broadcasted, such as in scenes such as PPT, background cloth loop playback, pure background music, etc., the live text cannot be effectively obtained in such scenes, resulting in the failure of recording and broadcasting detection. At the same time, the speech-to-text service needs to adapt to different languages ​​of the anchor, and requires a higher speech-to-text configuration cost. In addition, in the matching process, the similarity between samples is considered more, and the consideration of timing attributes is lacking. Based on this, a recording and broadcasting detection method of an embodiment of the present application is provided to solve the technical problem that the existing recording and broadcasting detection method strongly relies on the live audio and speech-to-text capabilities, and the recording and broadcasting detection effect is poor.

[0029] Figure 1 A flowchart of a recording and broadcast detection method provided in an embodiment of the present application is given. The recording and broadcast detection method provided in an embodiment of the present application can be executed by a recording and broadcast detection device, which can be implemented in hardware and / or software and integrated in a recording and broadcast detection device.

[0030] The following description is made by taking the recording and broadcasting detection device executing the recording and broadcasting detection method as an example. Figure 1 , the recording and broadcasting detection method comprises:

[0031] S110: Acquire multiple live broadcast sampling frames of the live broadcast room to be detected, and determine sampling frame vectors of the multiple live broadcast sampling frames.

[0032] Exemplarily, the live video stream of the live broadcast room to be detected is sampled to obtain multiple live sampling frames, and sampling frame vectors of the multiple live sampling frames are determined. For example, feature extraction is performed on the live sampling frames, and the extracted features are converted into vector form to obtain sampling frame vectors corresponding to the live sampling frames.

[0033] Optionally, feature extraction is performed on the live sampling frames based on key point detection, feature descriptors, deep learning (using a deep learning model for feature extraction, which can automatically learn multi-level feature representations from edges, textures to complex objects), etc.

[0034] S120: Perform frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain a first matching result, and determine a first recording and broadcasting detection result according to the first matching result. The recording and broadcasting vector library records multiple recording and broadcasting frame vectors of multiple recording and broadcasting live broadcasting rooms.

[0035] Exemplarily, after obtaining the sampling frame vectors of multiple live sampling frames, the sampling frame vectors are matched by frame vector in a pre-configured recording and broadcasting vector library to obtain a first matching result. Among them, the recording and broadcasting vector library provided by this solution records multiple recording and broadcasting frame vectors of multiple recording and broadcasting live rooms (one recording and broadcasting live room may correspond to one or more recording and broadcasting frame vectors). For example, when a live room is determined to be a recording and broadcasting live room, the frame vector corresponding to the recording and broadcasting live room may be saved as a recording and broadcasting frame vector in the recording and broadcasting vector library.

[0036] Optionally, frame vector matching of the sampling frame vector in the recording and broadcasting vector library may be performed by calculating similarity information between the sampling frame vector and each recording and broadcasting frame vector in the recording and broadcasting vector library, and determining a first matching result of the sampling frame vector based on the similarity information. The similarity information (including the first similarity information and the second similarity information) provided by the present scheme may be obtained by the vector distance between the sampling frame vector and the recording and broadcasting frame vector (for example, Euclidean distance, Manhattan distance, Hamming distance, cosine distance, etc., the larger the vector distance, the lower the similarity between the sampling frame vector and the recording and broadcasting frame vector) or the vector similarity (for example, cosine similarity, Jaccard similarity, etc., the larger the vector similarity, the higher the similarity between the sampling frame vector and the recording and broadcasting frame vector).

[0037] After determining the first matching results of multiple sampling frame vectors relative to each recorded broadcast frame vector in the recorded broadcast vector library, the first recorded broadcast detection result of the live broadcast room to be detected in the recorded broadcast vector library can be determined based on the multiple first matching results. The first recorded broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded broadcast room.

[0038] For example, it is determined whether there is a first matching result whose similarity information reaches a set similarity threshold, or it is determined whether there are a first set number of sampling frame vectors whose similarity information reaches a set similarity threshold with the recorded broadcast frame vector, or it can be determined whether there are a second set number of sampling frame vectors whose similarity information reaches a set similarity threshold with the recorded broadcast frame vector in the same recorded broadcast live room. If so, it is determined that the first recorded broadcast detection result is that the live broadcast room to be detected is a recorded broadcast live room, otherwise it is determined that the first recorded broadcast detection result is that the live broadcast room to be detected is a non-recorded broadcast live room. Among them, the similarity information between the sampling frame vector and the recorded broadcast frame vector reaches the set similarity threshold, which can be that the vector distance is less than the set vector distance threshold, or the vector similarity is greater than or equal to the set vector similarity threshold.

[0039] It needs to be explained that when the first recording and broadcasting detection result indicates that the live broadcast room to be detected is a recording and broadcasting live broadcast room, it can be directly determined that the detected live broadcast room is a recording and broadcasting live broadcast room, and there is no need to perform recording and broadcasting detection in the live broadcast vector library. The recording and broadcasting detection process for the current live broadcast room to be detected can be ended, and the recording and broadcasting detection efficiency can be improved while accurately detecting the recording and broadcasting live broadcast room. When the first recording and broadcasting detection result shows that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room, it does not mean that the live broadcast room to be detected is definitely a non-recording and broadcasting live broadcast room, but that the live broadcast room to be detected is not detected as a recording and broadcasting live broadcast room in the recording and broadcasting vector library, and further more comprehensive detection is required in the live broadcast vector library.

[0040] In a possible embodiment, after determining the first recording and broadcast detection result according to the first matching result, the recording and broadcast detection method provided by the present scheme may also save the sampling frame vector to the recording and broadcast vector library when the first recording and broadcast detection result is that the live broadcast room to be detected is a recording and broadcast live broadcast room, for frame vector matching of the subsequently determined sampling frame vector.

[0041] Exemplarily, when the first recording and broadcast detection result is that the live broadcast room to be detected is a recording and broadcast live broadcast room, the live broadcast room to be detected and the corresponding one or more sampling frame vectors can be saved in the recording and broadcast vector library. The live broadcast room to be detected and the corresponding one or more sampling frame vectors saved in the live broadcast vector library can be used as a newly added recording and broadcast live broadcast room and the corresponding recording and broadcast frame vectors to perform frame vector matching on the sampling frame vectors determined in the subsequent recording and broadcast detection process, thereby improving the detection accuracy of the recording and broadcast live broadcast room.

[0042] In a possible embodiment, the recording and broadcasting detection method provided by the present solution may further perform downsampling processing on multiple sampling frame vectors before performing frame vector matching on the sampling frame vectors in the recording and broadcasting vector library to obtain the first matching result. Accordingly, the recording and broadcasting detection method provided by the present solution may perform frame vector matching on the sampling frame vectors in the recording and broadcasting vector library to obtain the first matching result, which may be to perform frame vector matching on the sampling frame vectors after the downsampling processing in the recording and broadcasting vector library to obtain the first matching result.

[0043] Exemplarily, before performing frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain the first matching result, multiple sampled frame vectors are downsampled, and some of the sampled frame vectors are matched with the frame vector in the recording and broadcasting vector library to obtain the first matching result. Since the recording and broadcasting vector library records the recording and broadcasting frame vectors that are determined to correspond to the recording and broadcasting live broadcast room, the recording and broadcasting live broadcast room can be detected more accurately, and the sampled frame vectors can be downsampled, which effectively reduces the amount of data processing and improves the efficiency of recording and broadcasting detection while ensuring the accuracy of recording and broadcasting detection.

[0044] S130: When the first recording and broadcasting detection result shows that the live broadcasting room to be detected is a non-recording and broadcasting live broadcasting room, frame vector matching is performed on the sampled frame vector in the live broadcasting vector library to obtain a second matching result, and a second recording and broadcasting detection result is determined according to the second matching result.

[0045] Exemplarily, when the first recording and broadcasting detection result is that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room, the sampled frame vector is matched with the frame vector in the pre-configured live broadcast vector library to obtain the second matching result. Among them, the live broadcast vector library provided by this solution records multiple video frame vectors (a video live broadcast room is a live broadcast room that has ended live broadcast or is in the process of live broadcast, including a live broadcast room determined to be a recording and broadcasting live broadcast room and a non-recording and broadcasting live broadcast room) (one video live broadcast room can correspond to one or more video frame vectors). For example, after sampling the frame vector of a live broadcast room, the frame vector is saved as a video frame vector in the live broadcast vector library.

[0046] Optionally, frame vector matching is performed on the sampled frame vector in the live broadcast vector library, which may be to calculate similarity information between the sampled frame vector and each video frame vector in the live broadcast vector library, and determine a second matching result of the sampled frame vector based on the similarity information.

[0047] After determining the second matching results of multiple sampling frame vectors relative to each video frame vector in the live broadcast vector library, the second recording and broadcast detection result of the live broadcast room to be detected in the live broadcast vector library can be determined based on the multiple second matching results. The second recording and broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded and broadcast live room.

[0048] For example, it is determined whether there is a second matching result whose similarity information reaches a set similarity threshold, or it is determined whether there are a third set number of sampling frame vectors whose similarity information reaches a set similarity threshold with the video frame vector, or it can be determined whether there are a fourth set number of sampling frame vectors whose similarity information reaches a set similarity threshold with the video frame vector in the same recorded live broadcast room. If so, it is determined that the second recorded broadcast detection result is that the live broadcast room to be detected is a recorded broadcast room, otherwise it is determined that the second recorded broadcast detection result is that the live broadcast room to be detected is a non-recorded live broadcast room. Among them, the similarity information between the sampling frame vector and the video frame vector reaches the set similarity threshold, which can be that the vector distance is less than the set vector distance threshold, or the vector similarity is greater than or equal to the set vector similarity threshold.

[0049] In a possible embodiment, the recording and broadcast detection method provided by the present scheme can also save the sampling frame vector to the live broadcast vector library after determining the first recording and broadcast detection result according to the first matching result, or after determining the second recording and broadcast detection result according to the second matching result, for frame vector matching of the subsequently determined sampling frame vector.

[0050] Exemplarily, no matter what the first recording and broadcast detection results and the second recording and broadcast detection results are for the recording and broadcast detection results of the live broadcast room to be detected, the multiple live broadcast sampling frames of the live broadcast room to be detected are all sampling frames that have been broadcast live. The live broadcast room to be detected and the corresponding sampling frame vectors can be saved to the live broadcast vector library, which can be used as newly added video live rooms and corresponding video frame vectors to perform frame vector matching on the sampling frame vectors determined in the subsequent recording and broadcast detection process, thereby improving the detection accuracy of the recorded and broadcast live rooms.

[0051] In one embodiment, a validity period can be set for the video live room and the corresponding video frame vector recorded in the live vector library. When the storage time of the video live room and the corresponding video frame vector exceeds the validity period, the video live room and the corresponding video frame vector are deleted to reduce the unlimited expansion of the data volume in the live vector library and ensure the efficiency of recording and broadcasting detection.

[0052] In the above, by determining the sampling frame vectors of multiple live sampling frames of the live broadcast room to be detected, frame vector matching is performed on the sampling frame vectors in the recording and broadcasting vector library to obtain a first matching result, and the first recording and broadcasting detection result is determined according to the first matching result. When the first recording and broadcasting detection result is that the live broadcast room to be detected is not a recording and broadcasting live broadcast room, frame vector matching is performed on the sampling frame vectors in the live broadcast vector library to obtain a second matching result, and the second recording and broadcasting detection result is determined according to the second matching result. It can be determined whether the live broadcast room to be detected is a recording and broadcasting live broadcast room according to the second recording and broadcasting detection result. First, a fast and accurate recording and broadcasting detection is performed in the recording and broadcasting vector library using the recording and broadcasting frame vector corresponding to the recording and broadcasting live broadcast room. When the recording and broadcasting live broadcast room is not detected in the recording and broadcasting vector library, a more comprehensive recording and broadcasting detection is performed in the live broadcast vector library. While improving the efficiency of recording and broadcasting detection, the accuracy of recording and broadcasting detection is effectively guaranteed, and the dependence of recording and broadcasting detection on live audio and speech-to-text capabilities can be effectively reduced, thereby improving the effect of recording and broadcasting detection.

[0053] Based on the above embodiments, Figure 2 A flowchart of another recording and broadcasting detection method provided in an embodiment of the present application is given, and the recording and broadcasting detection method is a specific embodiment of the above recording and broadcasting detection method. Figure 2 , the recording and broadcasting detection method comprises:

[0054] S210: Acquire multiple live broadcast sampling frames of the live broadcast room to be detected, and determine sampling frame vectors of the multiple live broadcast sampling frames.

[0055] S220: Perform frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain a first matching result, wherein the recording and broadcasting vector library records a plurality of recording and broadcasting frame vectors of a plurality of recording and broadcasting live studios.

[0056] S230: Determine a first number of similar recorded and broadcasted frame vectors corresponding to each sampling frame vector according to the first matching result.

[0057] S240: Determine a first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector, based on a first preset similarity threshold and first similarity information between each sampling frame vector and a similar recorded broadcast frame vector.

[0058] S250: Determine a first recording and broadcasting detection result according to the first similarity matching score and the first time sequence matching score.

[0059] Exemplarily, after performing frame vector matching on multiple sampling frame vectors in the recording and broadcasting vector library to obtain multiple first matching results, for each sampling frame vector, a first number of similar recording and broadcasting frame vectors are determined from multiple recording and broadcasting frame vectors in the recording and broadcasting vector library in the order of the similarity reflected by the first matching result from large to small. For example, when the first matching result is represented by vector distance, assuming that the first number is p, after performing frame vector matching on m sampling frame vectors (for example, downsampling n sampling frame vectors to obtain m sampling frame vectors) in the recording and broadcasting vector library, for one of the sampling frame vectors, the recording and broadcasting frame vector with the corresponding vector distance p in front can be used as the similar recording and broadcasting frame vector of the sampling frame vector, and finally m*p similar recording and broadcasting frame vectors can be determined. Optionally, the first number of similar recording and broadcasting frame vectors can be quickly obtained through a vector database (for example, chroma, Milvus, Faiss and other vector databases).

[0060] In one embodiment, after determining the first number of similar recorded broadcast frame vectors corresponding to each sampling frame vector, the first preset similarity threshold is compared with the first similarity information of each sampling frame vector and the similar recorded broadcast frame vector, and the first similarity matching score of each sampling frame vector corresponding to the plurality of similar recorded broadcast frame vectors is determined according to the comparison result, and the first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector is determined according to the comparison result. The first similarity information may be the vector distance or vector similarity between the sampling frame vector and the similar recorded broadcast frame vector.

[0061] Among them, the first similarity matching score can be used to indicate the degree of similarity between multiple sampling frame vectors and a first number of similar recorded broadcast frame vectors. The more sampling frame vectors whose degree of similarity reflected by the first similarity information reaches the degree of similarity reflected by the first preset similarity threshold, the greater the first similarity matching score. Among them, the first timing matching score can be used to indicate the degree of similarity between multiple sampling frame vectors and similar recorded broadcast frame vectors in multiple similar recorded broadcast frame vectors in the same recorded broadcast live room. The more recorded broadcast live rooms there are with similar recorded broadcast frame vectors whose degree of similarity reflected by the first similarity information reaches the degree of similarity reflected by the first preset similarity threshold, the greater the first timing matching score.

[0062] In one embodiment, after determining the first similarity matching score and the first timing matching score of each sampling frame vector relative to the corresponding multiple similar recording and broadcasting frame vectors, the first recording and broadcasting detection result is determined according to the first similarity matching score and the first timing matching score. For example, the sum of the first similarity matching score and the first timing matching score is used as the first recording and broadcasting detection result, or the weighted sum of the first similarity matching score and the first timing matching score is used as the first recording and broadcasting detection result, or the ratio of the weighted sum of the first similarity matching score and the first timing matching score to the number of sampling frame vectors is used as the first recording and broadcasting detection result.

[0063] In one embodiment, after obtaining the first recording and broadcasting detection result, the first recording and broadcasting detection result and the first preset result threshold value may be compared. When the first recording and broadcasting detection result is greater than or equal to the first preset result threshold value, it may be determined that the live broadcast room to be detected is a recording and broadcasting live broadcast room, and when the first recording and broadcasting detection result is less than the first preset result threshold value, it may be determined that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room.

[0064] This solution determines a first number of similar recorded broadcast frame vectors corresponding to each sampling frame vector according to the first matching result, determines a first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector according to a first preset similarity threshold and the first similarity information between each sampling frame vector and the similar recorded broadcast frame vector. The first recorded broadcast detection result of the live broadcast room to be detected can be accurately determined according to the first similarity matching score and the first timing matching score, thereby effectively improving the recorded broadcast detection efficiency.

[0065] In one possible embodiment, Figure 3 As shown in a flowchart of determining a first similarity matching score and a first timing matching score, the recording and broadcasting detection method provided by the present solution determines a first similarity matching score of a sampling frame vector relative to a similar recording and broadcasting frame vector, and a first timing matching score of a live broadcast room to be detected relative to a recording and broadcasting live broadcast room corresponding to the similar recording and broadcasting frame vector, including:

[0066] S241: Determine a first similarity score of each sampling frame vector relative to a similar recorded frame.

[0067] S242: Determine a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector according to the first similarity score.

[0068] S243: Determine a first timing matching score of the live broadcast room to be detected relative to the live broadcast room corresponding to the similar recorded broadcast frame vector according to the first preset number threshold, the first similarity score of each sampling frame vector, and the recorded broadcast room corresponding to each similar recorded broadcast frame.

[0069] Exemplarily, after determining the first number of similar recorded and broadcast frame vectors corresponding to each sampling frame vector, determine the first similarity score of each sampling frame vector relative to the similar recorded and broadcast frame. The first similarity score can be used to indicate whether the degree of similarity reflected by the first similarity information reaches the degree of similarity reflected by the first preset similarity threshold. For example, when the first similarity information is represented by vector distance, the corresponding first preset similarity threshold is the distance threshold. For a sampling frame vector, the first similarity information corresponding to the first number of similar recorded and broadcast frame vectors is compared with the first preset similarity threshold respectively. When the first similarity information is less than or equal to the first preset similarity threshold, the first similarity score is a first preset value (e.g., 1). When the first similarity information is greater than the first preset similarity threshold, the first similarity score is a second preset value (e.g., 0).

[0070] In one embodiment, after determining the first similarity scores of each sampling frame vector relative to the corresponding similar recorded and broadcasted frames, the first similarity matching scores of the sampling frame vector relative to the similar recorded and broadcasted frame vector are determined according to the first similarity scores. For example, the sum of the first similarity scores of each sampling frame vector is used as the first similarity matching score; or, alternatively, a bitwise OR operation is performed on the first similarity scores of each sampling frame vector respectively (for example, when a first preset value exists among the multiple first similarity scores of a sampling frame vector, the bitwise OR operation result is the first preset value, and when the multiple first similarity scores of a sampling frame vector are all second preset values, the bitwise OR operation result is the second preset value), and the sum of the bitwise OR operation results corresponding to the various first similarity scores is used as the first similarity matching score.

[0071] In one embodiment, the recorded broadcast live rooms corresponding to the first number of similar recorded broadcast frame vectors corresponding to each sampling frame vector are determined. For a sampling frame vector and a corresponding similar recorded broadcast frame vector, when the first similarity score is a first preset value, the number of times the recorded broadcast live room corresponding to the similar recorded broadcast frame vector appears (and the corresponding first similarity score is a first preset value) is counted. When the number of occurrences reaches the first preset number threshold, the first timing information under the combination of the sampling frame vector and the similar recorded broadcast frame vector is determined to be the first preset value. When the number of occurrences does not reach the first preset number threshold, the first timing information under the combination of the sampling frame vector and the similar recorded broadcast frame vector is determined to be the second preset value. For a sampling frame vector and a corresponding similar recorded broadcast frame vector, when the first similarity score is a second preset value, the first timing information under the combination of the sampling frame vector and the similar recorded broadcast frame vector is determined to be the second preset value. The first timing matching score of the live room to be detected relative to the recorded broadcast live room corresponding to the similar recorded broadcast frame vector can be determined based on the first timing information under the combination of each sampling frame vector and the similar recorded broadcast frame vector. For example, the sum of the first timing information of each sampling frame vector is used as the first timing matching score; or, the first timing information of each sampling frame vector can be subjected to a bitwise OR operation (for example, when a first preset value exists in multiple first timing information of a sampling frame vector, the bitwise OR operation result is the first preset value; when multiple first timing information of a sampling frame vector are all second preset values, the bitwise OR operation result is the second preset value), and the sum of the bitwise OR operation results corresponding to each first timing information is used as the first timing matching score.

[0072] This solution accurately determines the first similarity matching score based on the first similarity score of each sampling frame vector relative to the similar recorded broadcast frame, and accurately determines the first timing matching score based on the first preset number threshold, the first similarity score of each sampling frame vector and the recorded broadcast room corresponding to each similar recorded broadcast frame. The first recorded broadcast detection result of the live broadcast room to be detected can be accurately determined based on the first similarity matching score and the first timing matching score, thereby effectively improving the recorded broadcast detection efficiency.

[0073] In a possible embodiment, the first similarity score provided by the present solution is recorded in a first vector similarity matrix, and different elements in the first vector similarity matrix record first similarity scores of different sampling frame vectors relative to different similar recorded frames.

[0074] Exemplarily, after determining the first matching result of each sampling frame vector relative to the first number of similar recorded and broadcast frames, a first matching result matrix is ​​formed according to the first matching result. For example, the number of sampling frame vectors after downsampling is m, the first number is p, the first matching result matrix is ​​an m*p matrix, and each column element in the first matching result matrix is ​​a first matching result of a sampling frame vector relative to the first number of similar recorded and broadcast frames.

[0075] In one embodiment, a first similarity score of the sampling frame vector corresponding to each element position relative to a similar recorded frame is determined based on a comparison result of the first matching result corresponding to each element position of the first matching result matrix with a first preset similarity threshold.

[0076] For example, when the first matching result is represented by vector distance, the first preset similarity threshold is the vector distance threshold, when the first matching result is less than or equal to the first preset similarity threshold, the first similarity score of the position corresponding to the first matching result is the first preset value, and when the first matching result is greater than the first preset similarity threshold, the first similarity score of the position corresponding to the first matching result is the second preset value. When the first matching result is represented by vector similarity, the first preset similarity threshold is the vector similarity threshold, when the first matching result is greater than or equal to the first preset similarity threshold, the first similarity score of the position corresponding to the first matching result is the first preset value, and when the first matching result is less than the first preset similarity threshold, the first similarity score of the position corresponding to the first matching result is the second preset value. After determining the first similarity scores of the positions corresponding to the first matching results, a first vector similarity matrix can be formed based on these first similarity scores. This scheme records the first similarity scores of different sampling frame vectors relative to similar recorded and broadcast frames through the first vector similarity matrix, effectively improving the processing efficiency of the first similarity matching score and the first temporal matching score, and effectively improving the recording and broadcast detection efficiency.

[0077] In a possible embodiment, the recording and broadcasting detection method provided by the present scheme determines the first similarity matching score of the sampling frame vector relative to the similar recording and broadcasting frame vector based on the first similarity score, which can be: performing bitwise OR addition processing on the first similarity scores corresponding to multiple sampling frame vectors in the first vector similarity matrix and summing them up to obtain the first similarity matching score of the sampling frame vector relative to the similar recording and broadcasting frame vector.

[0078] Exemplarily, the first similarity scores corresponding to each sampling frame vector in the first vector similarity matrix are subjected to bitwise OR addition processing and summed (each column is subjected to bitwise OR addition processing, and the bitwise OR addition processing results of each column are added together) to obtain the first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector. This solution obtains the first similarity matching score by performing bitwise OR addition processing on the first vector similarity matrix and summing the scores, thereby effectively improving the efficiency of determining the first similarity matching score and effectively improving the efficiency of recorded broadcast detection.

[0079] In a possible embodiment, the recording and broadcasting detection method provided by the present solution determines the first timing matching score of the live broadcast room to be detected relative to the recorded and broadcasting live broadcast room corresponding to the similar recorded and broadcasting frame vector according to the first preset number threshold, the first similarity score of each sampling frame vector, and the recorded and broadcasting live broadcast room corresponding to each similar recorded and broadcasting frame, which can be:

[0080] S2431: Determine the recording and broadcasting room matrix according to the recording and broadcasting rooms corresponding to each similar recording and broadcasting frame.

[0081] S2432: Determine a first timing matching matrix based on the first number threshold and the first similarity scores of the recorded and broadcast live rooms at each position in the recorded and broadcast live room matrix in the corresponding position of the first vector similarity matrix, wherein different elements in the first timing matching matrix record first timing information of whether the number of times the first similarity score corresponding to the recorded and broadcast live room at the corresponding position is a first preset value reaches the first number threshold.

[0082] S2433: Perform bitwise OR addition processing on the first timing information corresponding to multiple sampling frame vectors in the first timing matching matrix and sum them up to obtain a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector.

[0083] Exemplarily, the live broadcast rooms corresponding to the first number of similar live broadcast frame vectors corresponding to each sampling frame vector are determined, and a live broadcast room matrix is ​​formed based on these live broadcast rooms. Wherein, each column of the live broadcast room matrix is ​​a live broadcast room corresponding to the first number of similar live broadcast frame vectors corresponding to a sampling frame vector.

[0084] In one embodiment, for the recorded live broadcast room corresponding to each element position in the recorded live broadcast room matrix, it is determined whether the first similarity score corresponding to the recorded live broadcast room is a first preset value. When the first similarity score is a second preset value, the first timing information corresponding to the element position is determined to be a second preset value. When the first similarity score corresponding to the recorded live broadcast room is a first preset value, it is determined that the number of recorded live broadcast rooms that are consistent with the recorded live broadcast room (including the recorded live broadcast room) and have a corresponding first similarity score of the first preset value in each element position in the recorded live broadcast room matrix reaches a first numerical threshold (at this time, the number of times the first similarity score of the recorded live broadcast room is the first preset value reaches the first numerical threshold), and the first timing information corresponding to the element position is determined to be the first preset value, and when the number of recorded live broadcast rooms with a corresponding first similarity score of the first preset value does not reach the first numerical threshold (at this time, the number of times the first similarity score of the recorded live broadcast room is the first preset value does not reach the first numerical threshold), the first timing information corresponding to the element position is determined to be the second preset value. A first timing matching matrix can be formed based on the first timing information corresponding to each element position in the recorded live broadcast room matrix.

[0085] In one embodiment, after obtaining the first timing matching matrix, the first timing information corresponding to the multiple sampling frame vectors in the first timing matching matrix is ​​subjected to bitwise OR addition processing and summed (each column is subjected to bitwise OR addition processing, and the bitwise OR addition processing results of each column are added together) to obtain the first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector. This scheme determines the recorded broadcast room matrix according to the recorded broadcast rooms corresponding to each similar recorded broadcast frame, determines the first timing matching matrix according to the first number threshold and the recorded broadcast room matrix, and performs bitwise OR addition processing on the first timing information corresponding to the multiple sampling frame vectors in the first timing matching matrix and sums them to obtain the first timing matching score, thereby effectively improving the efficiency of determining the first timing matching score and effectively improving the efficiency of recorded broadcast detection.

[0086] In a possible embodiment, the recording and broadcast detection method provided by the present scheme determines a first recording and broadcast detection result based on a first similarity matching score and a first timing matching score. The first similarity matching score and the first timing matching score may be weightedly summed to obtain a first weighted summation result, and a first ratio of the first weighted summation result to the number of sampling frame vectors (the number of sampling frame vectors after downsampling) is determined, and the first ratio is determined as the first recording and broadcast detection result. The first recording and broadcast detection result is accurately determined based on the first similarity matching score and the first timing matching score, thereby ensuring the accuracy of the recording and broadcast detection.

[0087] For example, when the first matching result is represented by vector distance, assuming that the sampling frame vector is m=5, the first number is p=6, and the first preset similarity threshold is t=10, the first matching result matrix (vector distance matrix) is:

[0088]

[0089] Correspondingly, the first vector similarity matrix is:

[0090]

[0091] The first similarity match score can be obtained by performing bitwise OR addition on the first vector similarity matrix and summing it up:

[0092] sum(B)=1+1+1+1+1=5

[0093] Furthermore, the recording and live broadcasting rooms corresponding to the first number of similar recording and broadcasting frame vectors corresponding to each sampling frame vector are used to form a recording and live broadcasting room matrix:

[0094]

[0095] Assuming that the first numerical threshold is the ratio of the number of sampling frame vectors (the number of sampling frame vectors after downsampling) to the set coefficient (for example, 2), for example, the first numerical threshold ε=m / 2, it can be seen that the number of times the first similarity score of the recorded live broadcast room ee1 is the first preset value (1) is 4 times, reaching the first numerical threshold (4>m / 2=2.5), the first timing information corresponding to the recorded live broadcast room whose first similarity score is the first preset value is the first preset value, and the first timing information corresponding to the remaining recorded live broadcast rooms is the second preset value (0), and the first timing matching matrix is ​​obtained:

[0096]

[0097] The first timing matching score can be obtained by performing bitwise OR addition on the first timing matching matrix and summing the results:

[0098] sum(D)=1+1+0+1+1=4

[0099] Among them, the video frame timing is also taken into account by matching the first timing matching matrix, which can effectively improve the accuracy and reliability of recorded broadcast detection. In addition, on the basis that the corresponding sampled video frame and the live broadcast room video frame are sufficiently similar, the more similar the sequence is to a certain live broadcast room, the greater the possibility that the live broadcast room to be detected is a recorded broadcast room. On the basis of calculating the similarity between samples, the timing properties of the samples are considered at the same time, which can effectively improve the accuracy and reliability of recorded broadcast detection.

[0100] Correspondingly, the first matching result can be expressed as:

[0101]

[0102] Wherein, α is the first preset weight, β is the second preset weight, and 1≥α≥0, 1≥β≥0, α+β=1. Assuming α=0.4, β=0.6, the first matching result is:

[0103]

[0104] The first recording and broadcast detection result can be compared with the first preset result threshold. When the first recording and broadcast detection result is greater than or equal to the first preset result threshold, it can be determined that the live broadcast room to be detected is a recording and broadcast room. When the first recording and broadcast detection result is less than the first preset result threshold, it can be determined that the live broadcast room to be detected is a non-recording and broadcast room.

[0105] S260: When the first recording and broadcasting detection result is that the live broadcasting room to be detected is a non-recording and broadcasting live broadcasting room, frame vector matching is performed on the sampled frame vector in the live broadcasting vector library to obtain a second matching result.

[0106] S270: Determine a second number of similar video frame vectors corresponding to each sampling frame vector according to the second matching result.

[0107] S280: Determine a second similarity matching score of the sampling frame vector relative to the similar video frame vector and a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector based on a second preset similarity threshold and second similarity information between each sampling frame vector and the similar video frame vector.

[0108] S290: Determine a second recording and broadcasting detection result according to the second similarity matching score and the second time sequence matching score.

[0109] When the first recording and broadcasting detection result is that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room, the sampled frame vector is further matched with the frame vector in the live broadcast vector library to obtain a second matching result. Exemplarily, for each sampling frame vector, a second number of similar video frame vectors are determined from multiple video frame vectors in the live broadcast vector library in the order of the similarity reflected by the second matching result from large to small. For example, when the second matching result is represented by the vector distance, assuming that the second number is k (the second number is greater than the first number), after the frame vector matching of n sampling frame vectors is performed in the live broadcast vector library, for one of the sampling frame vectors, the video frame vector with the corresponding vector distance in front of k can be used as the similar video frame vector of the sampling frame vector, and finally n*k similar video frame vectors can be determined. Optionally, the second number of similar video frame vectors can be quickly acquired through the vector database.

[0110] In one embodiment, after determining the second number of similar video frame vectors corresponding to each sampling frame vector, the second preset similarity threshold is compared with the second similarity information of each sampling frame vector and the similar video frame vector, and the second similarity matching score of each sampling frame vector corresponding to the plurality of similar video frame vectors is determined according to the comparison result, and the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector is determined according to the comparison result. The second similarity information can be the vector distance or vector similarity between the sampling frame vector and the similar video frame vector.

[0111] The second similarity matching score can be used to indicate the degree of similarity between multiple sampling frame vectors and a second number of similar video frame vectors. The more sampling frame vectors whose degree of similarity reflected by the second similarity information reaches the degree of similarity reflected by the second preset similarity threshold, the greater the second similarity matching score. The second timing matching score can be used to indicate the degree of similarity between multiple sampling frame vectors and similar video frame vectors in the same video live broadcast room. The more video live broadcast rooms there are similar video frame vectors whose degree of similarity reflected by the second similarity information reaches the degree of similarity reflected by the second preset similarity threshold, the greater the second timing matching score.

[0112] In one embodiment, after determining the second similarity matching score and the second timing matching score of each sampling frame vector relative to the corresponding multiple similar video frame vectors, the second recording and broadcasting detection result is determined according to the second similarity matching score and the second timing matching score. For example, the sum of the second similarity matching score and the second timing matching score is used as the second recording and broadcasting detection result, or the weighted sum of the second similarity matching score and the second timing matching score is used as the second recording and broadcasting detection result, or the ratio of the weighted sum of the second similarity matching score and the second timing matching score to the number of sampling frame vectors is used as the second recording and broadcasting detection result.

[0113] In one embodiment, after obtaining the second recording and broadcasting detection result, the second recording and broadcasting detection result and the first preset result threshold value may be compared. When the second recording and broadcasting detection result is greater than or equal to the first preset result threshold value, it may be determined that the live broadcast room to be detected is a recording and broadcasting live broadcast room, and when the second recording and broadcasting detection result is less than the first preset result threshold value, it may be determined that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room.

[0114] This solution determines a second number of similar video frame vectors corresponding to each sampling frame vector according to the second matching result, determines a second similarity matching score of the sampling frame vector relative to the similar video frame vector, and a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector according to a second preset similarity threshold and second similarity information between each sampling frame vector and the similar video frame vector. The second recording and broadcast detection result of the live broadcast room to be detected can be accurately determined according to the second similarity matching score and the second timing matching score, thereby effectively improving the recording and broadcast detection efficiency.

[0115] In one possible embodiment, Figure 4 As shown in a flowchart of determining a second similarity matching score and a second timing matching score, the recording and broadcasting detection method provided by the present solution determines the second similarity matching score of the sampling frame vector relative to the similar video frame vector, and the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector, including:

[0116] S281: Determine a second similarity score of each sampling frame vector relative to a similar live frame.

[0117] S282: Determine a second similarity matching score of the sample frame vector relative to the similar video frame vector according to the second similarity score.

[0118] S283: Determine the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector based on the second preset number threshold, the second similarity score of each sampling frame vector and the video live broadcast room corresponding to each similar live broadcast frame.

[0119] Exemplarily, after determining the second number of similar video frame vectors corresponding to each sampling frame vector, determine the second similarity score of each sampling frame vector relative to the similar live frame. The second similarity score can be used to indicate whether the degree of similarity reflected by the second similarity information reaches the degree of similarity reflected by the second preset similarity threshold. For example, when the second similarity information is represented by vector distance, the corresponding second preset similarity threshold is the distance threshold. For a sampling frame vector, the second similarity information corresponding to the second number of similar video frame vectors is compared with the second preset similarity threshold respectively. When the second similarity information is less than or equal to the second preset similarity threshold, the second similarity score is the first preset value (e.g., 1). When the second similarity information is greater than the second preset similarity threshold, the second similarity score is the second preset value (e.g., 0).

[0120] In one embodiment, after determining the second similarity scores of each sampling frame vector relative to the corresponding similar live frames, the second similarity matching scores of the sampling frame vector relative to the similar video frame vector are determined according to the second similarity scores. For example, the sum of the second similarity scores of each sampling frame vector is used as the second similarity matching score; or a bitwise OR operation is performed on the second similarity scores of each sampling frame vector respectively (for example, when a first preset value exists among multiple second similarity scores of a sampling frame vector, the bitwise OR operation result is the first preset value, and when multiple second similarity scores of a sampling frame vector are all second preset values, the bitwise OR operation result is the second preset value), and the sum of the bitwise OR operation results corresponding to each second similarity score is used as the second similarity matching score.

[0121] In one embodiment, the video live broadcast rooms corresponding to the second number of similar video frame vectors corresponding to each sampling frame vector are determined. For a sampling frame vector and a corresponding similar video frame vector, when the second similarity score is a first preset value, the number of times the video live broadcast room corresponding to the similar video frame vector appears (and the corresponding second similarity score is a first preset value) is counted. When the number of occurrences reaches a second preset number threshold, the second timing information under the combination of the sampling frame vector and the similar video frame vector is determined to be a first preset value. When the number of occurrences does not reach the second preset number threshold, the second timing information under the combination of the sampling frame vector and the similar video frame vector is determined to be a second preset value. For a sampling frame vector and a corresponding similar video frame vector, when the second similarity score is a second preset value, the second timing information under the combination of the sampling frame vector and the similar video frame vector is determined to be a second preset value. The second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector can be determined based on the second timing information under the combination of each sampling frame vector and the similar video frame vector. For example, the sum of the second timing information of each sampling frame vector is used as the second timing matching score; or, a bitwise OR operation may be performed on the second timing information of each sampling frame vector respectively (for example, when a first preset value exists in multiple second timing information of a sampling frame vector, the bitwise OR operation result is the first preset value; when multiple second timing information of a sampling frame vector are all second preset values, the bitwise OR operation result is the second preset value), and the sum of the bitwise OR operation results corresponding to each second timing information is used as the second timing matching score.

[0122] This solution accurately determines the second similarity matching score based on the second similarity score of each sampling frame vector relative to the similar live frame, and accurately determines the second timing matching score based on the second preset number threshold, the second similarity score of each sampling frame vector and the video live room corresponding to each similar live frame. The second recording and broadcast detection result of the live room to be detected can be accurately determined based on the second similarity matching score and the second timing matching score, thereby effectively improving the recording and broadcast detection efficiency.

[0123] In a possible embodiment, the second similarity score provided by the present solution is recorded in a second vector similarity matrix, and different elements in the second vector similarity matrix record second similarity scores of different sampling frame vectors relative to different similar live frames.

[0124] Exemplarily, after determining the second matching results of each sampling frame vector relative to the second number of similar live frames, a second matching result matrix is ​​formed according to the second matching results. For example, the number of sampling frame vectors is n, the second number is k, the second matching result matrix is ​​an n*k matrix, and each column element in the second matching result matrix is ​​a second matching result of a sampling frame vector relative to the second number of similar live frames.

[0125] In one embodiment, based on the comparison result of the second matching result corresponding to each element position of the second matching result matrix and the second preset similarity threshold, the second similarity score of the sampling frame vector corresponding to each element position relative to the similar live frame is determined.

[0126] For example, when the second matching result is represented by vector distance, the second preset similarity threshold is the vector distance threshold, when the second matching result is less than or equal to the second preset similarity threshold, the second similarity score of the position corresponding to the second matching result is the first preset value, and when the second matching result is greater than the second preset similarity threshold, the second similarity score of the position corresponding to the second matching result is the second preset value. When the second matching result is represented by vector similarity, the second preset similarity threshold is the vector similarity threshold, when the second matching result is greater than or equal to the second preset similarity threshold, the second similarity score of the position corresponding to the second matching result is the first preset value, and when the second matching result is less than the second preset similarity threshold, the second similarity score of the position corresponding to the second matching result is the second preset value. After determining the second similarity scores of the positions corresponding to each second matching result, a second vector similarity matrix can be formed based on these second similarity scores. This scheme records the second similarity scores of different sampling frame vectors relative to similar live frames through the second vector similarity matrix, effectively improving the processing efficiency of the second similarity matching score and the second timing matching score, and effectively improving the recording and broadcasting detection efficiency.

[0127] In a possible embodiment, the recording and broadcasting detection method provided by the present scheme determines the second similarity matching score of the sampling frame vector relative to the similar video frame vector based on the second similarity score, which can be: performing bitwise OR addition processing on the second similarity scores corresponding to multiple sampling frame vectors in the second vector similarity matrix and summing them up to obtain the second similarity matching score of the sampling frame vector relative to the similar video frame vector.

[0128] Exemplarily, the second similarity scores corresponding to each sampling frame vector in the second vector similarity matrix are subjected to bitwise OR addition processing and summed (each column is subjected to bitwise OR addition processing, and the bitwise OR addition processing results of each column are added together) to obtain the second similarity matching score of the sampling frame vector relative to the similar video frame vector. This solution effectively improves the efficiency of determining the second similarity matching score and effectively improves the efficiency of recording and broadcasting detection by performing bitwise OR addition processing on the second vector similarity matrix and summing up the second similarity matching score.

[0129] In a possible embodiment, the recording and broadcasting detection method provided by the present solution determines the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector according to the second preset number threshold, the second similarity score of each sampling frame vector, and the video live broadcast room corresponding to each similar live broadcast frame, which can be:

[0130] S2831: Determine the video live room matrix according to the video live rooms corresponding to each similar live frame.

[0131] S2832: Determine the second timing matching matrix based on the second numerical threshold and the second similarity scores of the video live rooms at each position in the video live room matrix in the corresponding position of the second vector similarity matrix, and different elements in the second timing matching matrix record the second timing information of whether the number of times the second similarity score corresponding to the video live room at the corresponding position is the first preset value reaches the second numerical threshold.

[0132] S2833: Perform bitwise OR addition processing on the second timing information corresponding to multiple sampling frame vectors in the second timing matching matrix and sum them up to obtain a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector.

[0133] Exemplarily, the video live broadcast rooms corresponding to the second number of similar video frame vectors corresponding to each sampling frame vector are determined, and a video live broadcast room matrix is ​​formed based on these video live broadcast rooms. Wherein, each column of the video live broadcast room matrix is ​​a video live broadcast room corresponding to the second number of similar video frame vectors corresponding to a sampling frame vector.

[0134] In one embodiment, for the video live room corresponding to each element position in the video live room matrix, it is determined whether the second similarity score corresponding to the video live room is a first preset value, and when the second similarity score is a second preset value, the second timing information corresponding to the element position is determined as the second preset value. When the second similarity score corresponding to the video live room is the first preset value, it is determined that in each element position in the video live room matrix, the number of video live rooms that are consistent with the video live room (including the video live room) and have a corresponding second similarity score of the first preset value reaches a second numerical threshold (at this time, the number of times the second similarity score of the video live room is the first preset value reaches the second numerical threshold), and the second timing information corresponding to the element position is determined as the first preset value, and when the number of video live rooms with a corresponding second similarity score of the first preset value does not reach the second numerical threshold (at this time, the number of times the second similarity score of the video live room is the first preset value does not reach the second numerical threshold), the second timing information corresponding to the element position is determined as the second preset value. A second timing matching matrix can be formed based on the second timing information corresponding to each element position in the video live room matrix.

[0135] In one embodiment, after obtaining the second timing matching matrix, the second timing information corresponding to the multiple sampling frame vectors in the second timing matching matrix is ​​subjected to bitwise OR addition processing and summed (bitwise OR addition processing is performed on each column, and the bitwise OR addition processing results of each column are added together) to obtain the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector. This scheme determines the video broadcast room matrix according to the video broadcast rooms corresponding to each similar live broadcast frame, determines the second timing matching matrix according to the second numerical threshold and the video broadcast room matrix, and performs bitwise OR addition processing on the second timing information corresponding to the multiple sampling frame vectors in the second timing matching matrix and sums them to obtain the second timing matching score, thereby effectively improving the efficiency of determining the second timing matching score and effectively improving the efficiency of recorded broadcast detection.

[0136] In a possible embodiment, the recording and broadcast detection method provided by the present solution determines the second recording and broadcast detection result according to the second similarity matching score and the second timing matching score. The second similarity matching score and the second timing matching score can be weightedly summed to obtain a second weighted summation result, and a second ratio of the second weighted summation result to the number of sampling frame vectors is determined, and the second ratio is determined as the second recording and broadcast detection result. The second recording and broadcast detection result is accurately determined according to the second similarity matching score and the second timing matching score, thereby ensuring the accuracy of the recording and broadcast detection.

[0137] The second vector similarity matrix F and the second similarity matching score sum(F) can be determined with reference to the first vector similarity matrix and the method for determining the first similarity matching score. The second timing matching matrix H and the second timing matching score sum(H) can be determined with reference to the first timing matching matrix and the method for determining the first timing matching score, wherein the video frame timing is also taken into account by matching the second timing matching matrix, which can effectively improve the accuracy and reliability of recorded broadcast detection, and the corresponding sampled video frame and the live broadcast room video frame are sufficiently similar, and the more similar the sequence is to a certain live broadcast room, the greater the possibility that the live broadcast room to be detected is a video live broadcast room, and on the basis of calculating the similarity between samples, the timing attributes of the samples are considered at the same time, which can effectively improve the accuracy and reliability of recorded broadcast detection.

[0138] Correspondingly, the second matching result can be expressed as:

[0139]

[0140] Wherein, ω is the third preset weight, η is the fourth preset weight, and 1≥ω≥0, 1≥η≥0, ω+η=1. The second recording and broadcasting detection result may be compared with the second preset result threshold, and when the second recording and broadcasting detection result is greater than or equal to the second preset result threshold, it may be determined that the live broadcast room to be detected is a recording and broadcasting live broadcast room, and when the second recording and broadcasting detection result is less than the second preset result threshold, it may be determined that the live broadcast room to be detected is a non-recording and broadcasting live broadcast room.

[0141] In the above, by determining the sampling frame vectors of multiple live sampling frames of the live broadcast room to be detected, frame vector matching is performed on the sampling frame vectors in the recording and broadcasting vector library to obtain a first matching result, and the first recording and broadcasting detection result is determined according to the first matching result. When the first recording and broadcasting detection result is that the live broadcast room to be detected is not a recording and broadcasting live broadcast room, frame vector matching is performed on the sampling frame vectors in the live broadcast vector library to obtain a second matching result, and the second recording and broadcasting detection result is determined according to the second matching result. It can be determined whether the live broadcast room to be detected is a recording and broadcasting live broadcast room according to the second recording and broadcasting detection result. First, a fast and accurate recording and broadcasting detection is performed in the recording and broadcasting vector library using the recording and broadcasting frame vector corresponding to the recording and broadcasting live broadcast room. When the recording and broadcasting live broadcast room is not detected in the recording and broadcasting vector library, a more comprehensive recording and broadcasting detection is performed in the live broadcast vector library. While improving the efficiency of recording and broadcasting detection, the accuracy of recording and broadcasting detection is effectively guaranteed, and the dependence of recording and broadcasting detection on live audio and speech-to-text capabilities can be effectively reduced, thereby improving the effect of recording and broadcasting detection. At the same time, by determining the first number of similar recorded broadcast frame vectors corresponding to each sampling frame vector according to the first matching result, determining the first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector according to the first preset similarity threshold and the first similarity information between each sampling frame vector and the similar recorded broadcast frame vector, and the first timing matching score of the live broadcast room to be detected relative to the recorded broadcast live broadcast room corresponding to the similar recorded broadcast frame vector, the first recorded broadcast detection result of the live broadcast room to be detected can be accurately determined according to the first similarity matching score and the first timing matching score, effectively improving the recorded broadcast detection efficiency. And by determining the second number of similar video frame vectors corresponding to each sampling frame vector according to the second matching result, determining the second similarity matching score of the sampling frame vector relative to the similar video frame vector according to the second preset similarity threshold and the second similarity information between each sampling frame vector and the similar video frame vector, and the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector, the second recorded broadcast detection result of the live broadcast room to be detected can be accurately determined according to the second similarity matching score and the second timing matching score, effectively improving the recorded broadcast detection efficiency.

[0142] Figure 5 Schematic diagram of a recording and broadcasting detection device provided in an embodiment of the present application. Figure 5 The recording and broadcasting detection device includes a vector acquisition module 51, a first matching module 52 and a second matching module 53.

[0143] Among them, the vector acquisition module 51 is configured to acquire multiple live sampling frames of the live broadcast room to be detected, and determine the sampling frame vectors of the multiple live sampling frames; the first matching module 52 is configured to perform frame vector matching on the sampling frame vector in the recorded broadcast vector library to obtain a first matching result, and determine the first recorded broadcast detection result according to the first matching result, and the recorded broadcast vector library records multiple recorded broadcast frame vectors of multiple recorded broadcast rooms; the second matching module 53 is configured to perform frame vector matching on the sampling frame vector in the live broadcast vector library to obtain a second matching result when the first recorded broadcast detection result is that the live broadcast room to be detected is a non-recorded broadcast room, and determine the second recorded broadcast detection result according to the second matching result, and the live broadcast vector library records multiple video frame vectors of multiple video broadcast rooms.

[0144] In the above, by determining the sampling frame vectors of multiple live sampling frames of the live broadcast room to be detected, frame vector matching is performed on the sampling frame vectors in the recording and broadcasting vector library to obtain a first matching result, and the first recording and broadcasting detection result is determined according to the first matching result. When the first recording and broadcasting detection result is that the live broadcast room to be detected is not a recording and broadcasting live broadcast room, frame vector matching is performed on the sampling frame vectors in the live broadcast vector library to obtain a second matching result, and the second recording and broadcasting detection result is determined according to the second matching result. It can be determined whether the live broadcast room to be detected is a recording and broadcasting live broadcast room according to the second recording and broadcasting detection result. First, a fast and accurate recording and broadcasting detection is performed in the recording and broadcasting vector library using the recording and broadcasting frame vector corresponding to the recording and broadcasting live broadcast room. When the recording and broadcasting live broadcast room is not detected in the recording and broadcasting vector library, a more comprehensive recording and broadcasting detection is performed in the live broadcast vector library. While improving the efficiency of recording and broadcasting detection, the accuracy of recording and broadcasting detection is effectively guaranteed, and the dependence of recording and broadcasting detection on live audio and speech-to-text capabilities can be effectively reduced, thereby improving the effect of recording and broadcasting detection.

[0145] In a possible embodiment, the first matching module 52 determines the first recording and broadcasting detection result according to the first matching result, and is configured as follows:

[0146] Determine a first number of similar recorded and broadcast frame vectors corresponding to each sampling frame vector according to the first matching result;

[0147] Determine, according to a first preset similarity threshold and first similarity information between each sampling frame vector and a similar recorded broadcast frame vector, a first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector;

[0148] A first recording and broadcasting detection result is determined according to the first similarity matching score and the first time sequence matching score.

[0149] In a possible embodiment, the first matching module 52 determines a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded and broadcasted live broadcast room corresponding to the similar recorded and broadcasted frame vector, and is configured as follows:

[0150] Determining a first similarity score of each sampled frame vector relative to a similar recorded frame;

[0151] Determine a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector according to the first similarity score;

[0152] According to the first preset number threshold, the first similarity score of each sampling frame vector and the recorded live broadcast room corresponding to each similar recorded broadcast frame, a first timing matching score of the live broadcast room to be detected relative to the recorded live broadcast room corresponding to the similar recorded broadcast frame vector is determined.

[0153] In a possible embodiment, the first similarity score is recorded in a first vector similarity matrix, and different elements in the first vector similarity matrix record first similarity scores of different sampling frame vectors relative to similar recorded frames.

[0154] In a possible embodiment, the first matching module 52 determines a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector according to the first similarity score, and is configured as follows:

[0155] The first similarity scores corresponding to the plurality of sampling frame vectors in the first vector similarity matrix are subjected to bitwise OR addition processing and summed to obtain a first similarity matching score of the sampling frame vector relative to the similar recorded and broadcasted frame vector.

[0156] In a possible embodiment, the first matching module 52 determines the first timing matching score of the live broadcast room to be detected relative to the live broadcast room corresponding to the similar recorded broadcast frame vector according to the first preset number threshold, the first similarity score of each sampling frame vector, and the recorded broadcast room corresponding to each similar recorded broadcast frame, and is configured as follows:

[0157] Determine the recording and broadcasting room matrix according to the recording and broadcasting rooms corresponding to each similar recording and broadcasting frame;

[0158] Determine a first time sequence matching matrix according to the first number threshold and the first similarity scores of the recorded and broadcasted live rooms at each position in the recorded and broadcasted live room matrix in the corresponding position of the first vector similarity matrix, wherein different elements in the first time sequence matching matrix record first time sequence information of whether the number of times that the first similarity score corresponding to the recorded and broadcasted live room at the corresponding position is the first preset value reaches the first number threshold;

[0159] The first timing information corresponding to multiple sampling frame vectors in the first timing matching matrix is ​​subjected to bitwise OR addition processing and summed to obtain a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector.

[0160] In a possible embodiment, the first matching module 52 determines the first recording and broadcasting detection result according to the first similarity matching score and the first time sequence matching score, and is configured as follows:

[0161] A first weighted sum result is obtained by weighted summing the first similarity matching score and the first timing matching score, a first ratio of the first weighted sum result to the number of sampling frame vectors is determined, and the first ratio is determined as a first recording and broadcasting detection result.

[0162] In a possible embodiment, before the first matching module 52 performs frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain the first matching result, it is further configured as follows:

[0163] Down-sampling the plurality of sampling frame vectors;

[0164] Accordingly, in a possible embodiment, the first matching module 52 performs frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain a first matching result, and is configured as follows:

[0165] A frame vector matching is performed on the down-sampled sampling frame vector in the recording and broadcasting vector library to obtain a first matching result.

[0166] In a possible embodiment, the second matching module 53 determines the second recording and broadcasting detection result according to the second matching result, and is configured as follows:

[0167] Determine a second number of similar video frame vectors corresponding to each sampling frame vector according to the second matching result;

[0168] According to the second preset similarity threshold and the second similarity information between each sampling frame vector and the similar video frame vector, determine the second similarity matching score of the sampling frame vector relative to the similar video frame vector, and the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector;

[0169] A second recording and broadcasting detection result is determined according to the second similarity matching score and the second time sequence matching score.

[0170] In a possible embodiment, the second matching module 53 determines a second similarity matching score of the sampled frame vector relative to the similar video frame vector, and a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector, and is configured as follows:

[0171] Determining a second similarity score for each sampled frame vector relative to a similar live frame;

[0172] Determining a second similarity matching score of the sample frame vector relative to the similar video frame vector according to the second similarity score;

[0173] According to the second preset number threshold, the second similarity score of each sampling frame vector and the video live room corresponding to each similar live frame, the second timing matching score of the live room to be detected relative to the video live room corresponding to the similar video frame vector is determined.

[0174] In a possible embodiment, the second similarity score is recorded in a second vector similarity matrix, and different elements in the second vector similarity matrix record second similarity scores of different sampling frame vectors relative to similar live frames.

[0175] In a possible embodiment, the second matching module 53 determines a second similarity matching score of the sample frame vector relative to the similar video frame vector according to the second similarity score, and is configured as follows:

[0176] The second similarity scores corresponding to the multiple sampling frame vectors in the second vector similarity matrix are subjected to bitwise OR addition processing and summed to obtain a second similar matching score of the sampling frame vector relative to the similar video frame vector.

[0177] In a possible embodiment, the second matching module 53 determines the second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector according to the second preset number threshold, the second similarity score of each sampling frame vector, and the video live broadcast room corresponding to each similar live broadcast frame, and is configured as follows:

[0178] Determine a video live room matrix according to the video live rooms corresponding to each similar live frame;

[0179] Determine a second timing matching matrix according to the second numerical threshold and the second similarity scores of the video live rooms at each position in the video live room matrix in the corresponding position of the second vector similarity matrix, wherein different elements in the second timing matching matrix record second timing information of whether the number of times that the second similarity score corresponding to the video live room at the corresponding position is the first preset value reaches the second numerical threshold;

[0180] The second timing information corresponding to multiple sampling frame vectors in the second timing matching matrix is ​​subjected to bitwise OR addition processing and summed to obtain a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector.

[0181] In a possible embodiment, the second matching module 53 determines the second recording and broadcasting detection result according to the second similarity matching score and the second timing matching score, and is configured as follows:

[0182] A second weighted sum result is obtained by weighted summing the second similarity matching score and the second timing matching score, a second ratio of the second weighted sum result to the number of sampling frame vectors is determined, and the second ratio is determined as the second recording and broadcasting detection result.

[0183] In a possible embodiment, the recording and broadcasting detection device further includes a first updating module, and the first updating module is configured to save the sampling frame vector to a live broadcast vector library for performing frame vector matching on the subsequently determined sampling frame vector.

[0184] In a possible embodiment, the recording and broadcasting detection device also includes a second update module, and the second update module is configured to: when the first recording and broadcasting detection result is that the live broadcast room to be detected is a recording and broadcasting live broadcast room, save the sampling frame vector to the recording and broadcasting vector library for frame vector matching of the subsequently determined sampling frame vector.

[0185] It is worth noting that in the embodiment of the above-mentioned recording and broadcasting detection device, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the protection scope of the embodiments of the present application.

[0186] The embodiment of the present application further provides a recording and broadcasting detection device, which can be integrated with the recording and broadcasting detection device provided in the embodiment of the present application. Figure 6 Schematic diagram of a recording and broadcasting detection device provided in an embodiment of the present application. Figure 6 The recording and broadcasting detection device includes: an input device 63, an output device 64, a memory 62 and one or more processors 61; the memory 62 is used to store one or more programs; when one or more programs are executed by one or more processors 61, one or more processors 61 implement the recording and broadcasting detection method provided in the above embodiment. The recording and broadcasting detection device, equipment and computer provided above can be used to execute the recording and broadcasting detection method provided in any of the above embodiments, and have corresponding functions and beneficial effects.

[0187] The embodiments of the present application also provide a non-volatile storage medium storing computer executable instructions, which are used to execute the recording and broadcast detection method provided in the above embodiments when executed by a computer processor. Of course, the non-volatile storage medium storing computer executable instructions provided in the embodiments of the present application, whose computer executable instructions are not limited to the recording and broadcast detection method provided above, can also execute the related operations in the recording and broadcast detection method provided in any embodiment of the present application. The recording and broadcast detection device, equipment and storage medium provided in the above embodiments can execute the recording and broadcast detection method provided in any embodiment of the present application. For technical details not described in detail in the above embodiments, please refer to the recording and broadcast detection method provided in any embodiment of the present application.

[0188] On the basis of the above embodiments, the embodiments of the present application also provide a computer program product. The technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer program product is stored in a storage medium, including a number of instructions for enabling a computer device, a mobile terminal or a processor therein to execute all or part of the steps of the recording and broadcasting detection method provided in each embodiment of the present application.

Claims

1. A recording and broadcasting detection method, characterized in that: include: Acquire multiple live broadcast sampling frames of the live broadcast room to be detected, and determine sampling frame vectors of the multiple live broadcast sampling frames; Performing frame vector matching on the sampled frame vector in a recording and broadcasting vector library to obtain a first matching result, and determining a first recording and broadcasting detection result according to the first matching result, wherein the recording and broadcasting vector library records a plurality of recording and broadcasting frame vectors of a plurality of recording and broadcasting live broadcasting rooms; When the first recording and broadcast detection result is that the live broadcast room to be detected is a non-recorded and broadcast live broadcast room, frame vector matching is performed on the sampling frame vector in the live broadcast vector library to obtain a second matching result, and a second recording and broadcast detection result is determined according to the second matching result. The live broadcast vector library records multiple video frame vectors of multiple video live broadcast rooms, and the second recording and broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded and broadcast live broadcast room.

2. The recording and broadcasting detection method according to claim 1, characterized in that: The determining a first recording and broadcasting detection result according to the first matching result includes: Determine a first number of similar recorded and broadcasted frame vectors corresponding to each of the sampled frame vectors according to the first matching result; According to a first preset similarity threshold and first similarity information between each of the sampled frame vectors and the similar recorded and broadcast frame vectors, determine a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcast frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded and broadcast live broadcast room corresponding to the similar recorded and broadcast frame vector; A first recording and broadcasting detection result is determined according to the first similarity matching score and the first time sequence matching score.

3. The recording and broadcasting detection method according to claim 2, characterized in that: The determining of a first similarity matching score of the sampling frame vector relative to the similar recorded broadcast frame vector, and a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector, comprises: Determining a first similarity score of each of the sampled frame vectors relative to the similar recorded and broadcasted frame; Determining a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector according to the first similarity score; According to the first preset number threshold, the first similarity score of each of the sampling frame vectors, and the recorded live broadcast room corresponding to each of the similar recorded broadcast frames, a first timing matching score of the live broadcast room to be detected relative to the recorded live broadcast room corresponding to the similar recorded broadcast frame vector is determined.

4. The recording and broadcasting detection method according to claim 3, characterized in that: The first similarity score is recorded in a first vector similarity matrix, and different elements in the first vector similarity matrix record different first similarity scores of the sampling frame vectors relative to the similar recorded and broadcasted frames.

5. The recording and broadcasting detection method according to claim 4, characterized in that: The determining, according to the first similarity score, a first similarity matching score of the sampled frame vector relative to the similar recorded and broadcasted frame vector comprises: The first similarity scores corresponding to the plurality of the sampling frame vectors in the first vector similarity matrix are subjected to bitwise OR addition processing and summed to obtain a first similarity matching score of the sampling frame vector relative to the similar recorded and broadcasted frame vector.

6. The recording and broadcasting detection method according to claim 4, characterized in that: The determining, according to the first preset number threshold, the first similarity score of each of the sampling frame vectors, and the recorded and broadcast live room corresponding to each of the similar recorded and broadcast frames, a first timing matching score of the live room to be detected relative to the recorded and broadcast live room corresponding to the similar recorded and broadcast frame vectors, comprises: Determine a recording and live broadcast room matrix according to the recording and live broadcast rooms corresponding to each of the similar recording and live broadcast frames; Determine a first timing matching matrix according to the first number threshold and the first similarity scores of the recording and live broadcast rooms at each position in the recording and live broadcast room matrix in the corresponding position of the first vector similarity matrix, wherein different elements in the first timing matching matrix record first timing information of whether the number of times that the first similarity score corresponding to the recording and live broadcast room at the corresponding position is a first preset value reaches the first number threshold; The first timing information corresponding to multiple sampling frame vectors in the first timing matching matrix is ​​subjected to bitwise OR addition processing and summed to obtain a first timing matching score of the live broadcast room to be detected relative to the recorded broadcast room corresponding to the similar recorded broadcast frame vector.

7. The recording and broadcasting detection method according to claim 2, characterized in that: The determining the first recording and broadcasting detection result according to the first similarity matching score and the first time sequence matching score includes: A first weighted sum result is obtained by weighted summing the first similarity matching score and the first timing matching score, a first ratio of the first weighted sum result to the number of sampling frame vectors is determined, and the first ratio is determined as a first recording and broadcasting detection result.

8. The recording and broadcasting detection method according to claim 1, characterized in that: Before performing frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain a first matching result, the method further includes: Performing down-sampling processing on a plurality of the sampling frame vectors; Correspondingly, performing frame vector matching on the sampled frame vector in the recording and broadcasting vector library to obtain a first matching result includes: A frame vector matching is performed on the sampling frame vector after the downsampling process in the recording and broadcasting vector library to obtain a first matching result.

9. The recording and broadcasting detection method according to claim 1, characterized in that: The determining a second recording and broadcasting detection result according to the second matching result includes: Determine a second number of similar video frame vectors corresponding to each of the sampling frame vectors according to the second matching result; According to a second preset similarity threshold and second similarity information between each of the sampled frame vectors and the similar video frame vector, determine a second similarity matching score of the sampled frame vector relative to the similar video frame vector, and a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector; A second recording and broadcasting detection result is determined according to the second similarity matching score and the second timing matching score.

10. The recording and broadcasting detection method according to claim 9, characterized in that: The determining of a second similarity matching score of the sampling frame vector relative to the similar video frame vector, and a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector, comprises: Determining a second similarity score of each of the sampled frame vectors relative to the similar live frame; Determining a second similarity matching score of the sample frame vector relative to the similar video frame vector according to the second similarity score; According to the second preset number threshold, the second similarity score of each of the sampling frame vectors and the video live room corresponding to each of the similar live frames, the second timing matching score of the live room to be detected relative to the video live room corresponding to the similar video frame vector is determined.

11. The recording and broadcasting detection method according to claim 10, characterized in that: The second similarity score is recorded in a second vector similarity matrix, and different elements in the second vector similarity matrix record second similarity scores of different sampling frame vectors relative to the similar live broadcast frame.

12. The recording and broadcasting detection method according to claim 11, characterized in that: The determining, according to the second similarity score, a second similarity matching score of the sample frame vector relative to the similar video frame vector comprises: The second similarity scores corresponding to the plurality of the sampling frame vectors in the second vector similarity matrix are subjected to bitwise OR addition processing and summed to obtain a second similarity matching score of the sampling frame vector relative to the similar video frame vector.

13. The recording and broadcasting detection method according to claim 11, characterized in that: The determining, according to the second preset number threshold, the second similarity score of each of the sampling frame vectors, and the video live room corresponding to each of the similar live frames, a second timing matching score of the live room to be detected relative to the video live room corresponding to the similar video frame vector comprises: Determine a video live room matrix according to the video live rooms corresponding to each of the similar live frames; Determine a second timing matching matrix according to the second number threshold and the second similarity scores of the video live rooms at each position in the video live room matrix in the corresponding position of the second vector similarity matrix, wherein different elements in the second timing matching matrix record second timing information of whether the number of times that the second similarity score corresponding to the video live room at the corresponding position is a first preset value reaches the second number threshold; The second timing information corresponding to multiple sampling frame vectors in the second timing matching matrix is ​​subjected to bitwise OR addition processing and summed up to obtain a second timing matching score of the live broadcast room to be detected relative to the video live broadcast room corresponding to the similar video frame vector.

14. The recording and broadcasting detection method according to claim 9, characterized in that: The determining the second recording and broadcasting detection result according to the second similarity matching score and the second timing matching score includes: A second weighted sum result is obtained by weighted summing the second similarity matching score and the second timing matching score, a second ratio of the second weighted sum result to the number of sampling frame vectors is determined, and the second ratio is determined as a second recording and broadcasting detection result.

15. The recording and broadcasting detection method according to claim 1, characterized in that: After determining a first recording and broadcasting detection result according to the first matching result, or after determining a second recording and broadcasting detection result according to the second matching result, the method further includes: The sampling frame vector is saved in the live broadcast vector library for frame vector matching of the subsequently determined sampling frame vector.

16. The recording and broadcasting detection method according to claim 1, characterized in that: After determining the first recording and broadcasting detection result according to the first matching result, the method further includes: When the first recording and broadcasting detection result is that the live broadcast room to be detected is a recording and broadcasting live broadcast room, the sampling frame vector is saved to the recording and broadcasting vector library for frame vector matching of the subsequently determined sampling frame vector.

17. A recording and broadcasting detection device, characterized in that: It includes a vector acquisition module, a first matching module and a second matching module, wherein: The vector acquisition module is configured to acquire multiple live sampling frames of the live broadcast room to be detected, and determine the sampling frame vectors of the multiple live sampling frames; The first matching module is configured to perform frame vector matching on the sampled frame vector in a recording and broadcasting vector library to obtain a first matching result, and determine a first recording and broadcasting detection result according to the first matching result, wherein the recording and broadcasting vector library records a plurality of recording and broadcasting frame vectors of a plurality of recording and broadcasting live broadcasting rooms; The second matching module is configured to perform frame vector matching on the sampling frame vector in the live broadcast vector library to obtain a second matching result when the first recording and broadcast detection result indicates that the live broadcast room to be detected is a non-recorded and broadcast room, and determine a second recording and broadcast detection result based on the second matching result. The live broadcast vector library records multiple video frame vectors of multiple video live broadcast rooms, and the second recording and broadcast detection result is used to indicate whether the live broadcast room to be detected is a recorded and broadcast room.

18. A recording and broadcasting detection device, characterized in that: include: memory and one or more processors; The memory is used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the recording and broadcasting detection method according to any one of claims 1 to 16.

19. A non-volatile storage medium storing computer executable instructions, characterized in that: The computer executable instructions are used to execute the recording and broadcasting detection method as described in any one of claims 1 to 16 when executed by a computer processor.

20. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the recording and broadcasting detection method according to any one of claims 1 to 16 is implemented.

Images