Similar video detection method and device

Abstract

The invention discloses a similar video detection method and device, and relates to the field of video similarity analysis, and the method comprises a basic data acquisition step, a key frame extraction step and a video similarity analysis step. According to the method, the initial position of comparison video similarity analysis is scientifically and effectively determined, the video type corresponding to the target video is acquired, information extraction is performed on the video graph corresponding to the target video, and a text information set, a feature graph set and a non-feature information set are obtained, so that comprehensive data acquisition of the target video and the comparison video is realized; the accuracy and effectiveness of information acquisition are improved, the accuracy and effectiveness of video similarity analysis are further improved, and the development and progress of a similar video detection technology are promoted.

Description

Technical Field

[0001] The present application relates to the field of video similarity analysis, and in particular to a similar video detection method and device. Background Art

[0002] Similar video detection has important practical applications in copyright protection, content review, advertising monitoring, video recommendation, video retrieval, and public safety. With the continuous advancement of technology and the continuous expansion of application scenarios, similar video detection will play an even more important role in the future. However, existing video similarity analysis technology still has the following shortcomings:

[0003] When performing similar video recognition on a video, existing technologies are usually unable to effectively and accurately find the initial comparison points corresponding to the target video and the comparison video, which increases the error of video similarity analysis and reduces the accuracy and effectiveness of video similarity analysis.

[0004] When existing technologies identify similar videos, they lack data division according to video types, which reduces the accuracy of data acquisition and indirectly reduces the accuracy and effectiveness of video similarity analysis, which is not conducive to the development and progress of similar video detection technology. Summary of the Invention

[0005] The purpose of the present invention is to provide a similar video detection method and device to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a similar video detection method, comprising:

[0007] Basic data acquisition step: used to acquire the video, and perform data analysis based on the basic information corresponding to the video obtained by video extraction, to obtain each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video;

[0008] Key frame extraction step: for performing data analysis based on the basic information corresponding to the video and each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, and intelligently extracting the key frame information corresponding to the target task;

[0009] Video similarity analysis step: used to perform data analysis based on the key frame information corresponding to the target task, and obtain the video similarity between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected.

[0010] In a preferred embodiment of this solution, the basic data acquisition step is specifically performed as follows:

[0011] Obtain the target task video that needs to be tested for video similarity corresponding to the detection task;

[0012] Obtain each video to be compared for the video similarity detection corresponding to the detection task;

[0013] Extract data from the target task video to obtain basic information corresponding to the target task video, including the file size and playback time.

[0014] Extract data from each video to be compared and tested to obtain basic information corresponding to each video to be compared and tested, where the basic information includes file size and playback time;

[0015] By calculating the formula Calculate the initial matching coefficient Vs of the target task video corresponding to each video to be compared i , where td i 、fs i They represent the file size and playback duration of each video to be compared and tested, td′ and fs′ represent the file size and playback duration of the target task video, and i represents the number of each video to be compared and tested;

[0016] The preliminary conformity coefficients of the target task video and the detection videos to be compared are compared and analyzed with the preset preliminary conformity coefficient threshold. If the preliminary conformity coefficient of the detection video to be compared is less than or equal to the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the first sequence detection video to be compared. If the preliminary conformity coefficient of the detection video to be compared is greater than the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the second sequence detection video to be compared. The first sequence detection videos to be compared and the second sequence detection videos to be compared corresponding to the target task video are statistically obtained.

[0017] In the preferred embodiment of this solution, the key frame extraction step is specifically performed as follows:

[0018] Establishing a data extraction relationship between the key frame extraction step and the database, extracting each key frame extraction set stored in the database, wherein the key frame extraction set includes the video playback duration, the detection sequence corresponding to the video, the preliminary matching coefficient corresponding to the video, and the extraction control set corresponding to the video, wherein the extraction control set includes the number of key frame extractions and the time nodes of each key frame;

[0019] The key frame refers to an important frame in the video that can represent its content;

[0020] The detection sequence includes a first sequence and a second sequence;

[0021] Extracting the target task video, the video images of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, performing image comparison analysis on the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video, and obtaining image similarity between the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video;

[0022] If the image similarity with the first sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the first sequence, the first frame of the first sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the first sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the first sequence, the video images of the remaining frames corresponding to the first sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the first frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the first sequence. The frame is recorded as the initial frame of the first sequence of video to be compared and detected, and the video duration from the initial frame to the first frame of the first sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the first sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed corresponding to each first sequence of video to be compared and detected and the dissimilar video duration corresponding to each first sequence of video to be compared and detected of the target task video are statistically obtained.

[0023] If the image similarity with the second sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the second sequence, the second frame of the second sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the second sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the second sequence, the video images of the remaining frames corresponding to the second sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the second frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the second sequence. The frame is recorded as the initial frame of the second sequence of video to be compared and detected, and the video duration from the initial frame to the second frame of the second sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the second sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed of each second sequence of video to be compared and detected corresponding to the target task video and the dissimilar video duration corresponding to each second sequence of video to be compared and detected are statistically obtained.

[0024] Screening is performed based on the playback duration and preliminary matching coefficient corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested, to obtain an extraction control set corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested;

[0025] Through the extraction control sets corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, key frames are extracted for the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, and key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained, and video images of the key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained. Key frames are extracted for the target task video according to the time nodes of the key frames corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, and video images of the key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained;

[0026] The video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed, and the video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed of the target task video are recorded as the key frame information corresponding to the target task.

[0027] In the preferred embodiment of this solution, the specific implementation of the video similarity analysis step is as follows:

[0028] Obtain a set of feature graphics corresponding to the target task video of the detection task, and filter to obtain a set of feature graphics corresponding to the target task video and each detection video to be compared;

[0029] Feature information is extracted from the video images of each key frame according to the feature graphic set corresponding to the target task video, and feature information sets of the target task video and the video images of each key frame corresponding to each sub-video to be analyzed are obtained, wherein the feature information set includes a text information set, a feature graphic set, and a non-feature information set. The text information set refers to the text content in the video image, the feature image set includes the type of feature graphic corresponding to the video image and the number of each type of feature graphic, and the non-feature information set refers to the content information of the non-feature video image area excluding the text information set and the feature graphic set, and the non-feature information set includes the corresponding color histogram in the non-feature video image area;

[0030] The text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the target task video and the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is matched with the text content, and the text feature conformity coefficient F of the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is obtained. i p ;

[0031] By calculating the formula Calculate the characteristic graph matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ' p ;

[0032] By calculating the formula Calculate the non-feature matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ″ p ;

[0033] By calculating the formula Calculate the video similarity coefficient Q between the target task video and each video to be compared and tested i , where α represents the number of each type of feature graphics, β represents the number of types of feature graphics, u represents the number of each type of color, n represents the number of color types, p represents the number of each key frame, and q represents the number of key frames. It is expressed as the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed corresponding to each target task video. It is represented by the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed in each video to be compared and detected. It is expressed as the proportion of each color in each key frame video image of the target task video corresponding to each to-be-compared detection video corresponding to the to-be-analyzed sub-video. It is expressed as the proportion of each color corresponding to each key frame video image in the sub-video to be analyzed in each video to be compared and detected. It is represented by the length of dissimilar videos corresponding to the videos to be compared and detected;

[0034] The video similarity coefficients between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected and the corresponding numbers of each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected are screened to create a task log, and the task log is uploaded to the task management center.

[0035] To achieve the above object, the present invention further provides the following technical solution: a similar video detection device, comprising the following steps:

[0036] The video is acquired, and basic information corresponding to the video is obtained according to the video extraction to perform data analysis, so as to obtain each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos corresponding to the target task video;

[0037] Perform data analysis based on the basic information corresponding to the video and the first sequence of detection videos to be compared and the second sequence of detection videos to be compared corresponding to the target task video, and intelligently extract the key frame information corresponding to the target task;

[0038] Data analysis is performed based on the key frame information corresponding to the target task to obtain the video similarity between the target task video and each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The present invention analyzes and compares the video image of the first frame corresponding to the target video with the comparison video, scientifically and effectively determines the initial position of the comparison video similarity analysis, and effectively improves the accuracy and effectiveness of the video similarity analysis;

[0041] The present invention obtains the video type corresponding to the target video and extracts information from the video graphics corresponding to the target video to obtain a text information set, a feature graphic set and a non-feature information set, thereby achieving comprehensive data collection of the target video and the comparison video, improving the accuracy and effectiveness of information collection, further improving the accuracy and effectiveness of video similarity analysis, and being conducive to promoting the development and progress of similar video detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0043] Figure 1 This is a schematic diagram of the connection steps of an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] See also Figure 1 , the present invention provides a similar video detection method, which includes a basic data acquisition step, a key frame extraction step and a video similarity analysis step;

[0046] The basic data acquisition step is connected to the key frame extraction step, and the key frame extraction step is connected to the video similarity analysis step;

[0047] Basic data acquisition step: used to acquire the video, and perform data analysis based on the basic information corresponding to the video obtained by video extraction, to obtain each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video;

[0048] Furthermore, the specific execution method of the basic data acquisition step is as follows:

[0049] Obtain the target task video that needs to be tested for video similarity corresponding to the detection task;

[0050] Obtain each video to be compared for the video similarity detection corresponding to the detection task;

[0051] Extract data from the target task video to obtain basic information corresponding to the target task video, including the file size and playback time.

[0052] Extract data from each video to be compared and tested to obtain basic information corresponding to each video to be compared and tested, where the basic information includes file size and playback time;

[0053] By calculating the formula Calculate the initial matching coefficient Vs of the target task video corresponding to each video to be compared i , where td i 、fs i They represent the file size and playback duration of each video to be compared and tested, td′ and fs′ represent the file size and playback duration of the target task video, and i represents the number of each video to be compared and tested;

[0054] The preliminary conformity coefficients of the target task video and the detection videos to be compared are compared and analyzed with the preset preliminary conformity coefficient threshold. If the preliminary conformity coefficient of the detection video to be compared is less than or equal to the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the first sequence detection video to be compared. If the preliminary conformity coefficient of the detection video to be compared is greater than the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the second sequence detection video to be compared. The first sequence detection videos to be compared and the second sequence detection videos to be compared corresponding to the target task video are statistically obtained.

[0055] Key frame extraction step: for performing data analysis based on the basic information corresponding to the video and each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, and intelligently extracting the key frame information corresponding to the target task;

[0056] Furthermore, the specific execution method of the key frame extraction step is as follows:

[0057] Establishing a data extraction relationship between the key frame extraction step and the database, extracting each key frame extraction set stored in the database, wherein the key frame extraction set includes the video playback duration, the detection sequence corresponding to the video, the preliminary matching coefficient corresponding to the video, and the extraction control set corresponding to the video, wherein the extraction control set includes the number of key frame extractions and the time nodes of each key frame;

[0058] The key frame refers to an important frame in the video that can represent its content;

[0059] The detection sequence includes a first sequence and a second sequence;

[0060] Extracting the target task video, the video images of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, performing image comparison analysis on the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video, and obtaining image similarity between the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video;

[0061] If the image similarity with the first sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the first sequence, the first frame of the first sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the first sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the first sequence, the video images of the remaining frames corresponding to the first sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the first frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the first sequence. The frame is recorded as the initial frame of the first sequence of video to be compared and detected, and the video duration from the initial frame to the first frame of the first sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the first sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed corresponding to each first sequence of video to be compared and detected and the dissimilar video duration corresponding to each first sequence of video to be compared and detected of the target task video are statistically obtained.

[0062] If the image similarity with the second sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the second sequence, the second frame of the second sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the second sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the second sequence, the video images of the remaining frames corresponding to the second sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the second frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the second sequence. The frame is recorded as the initial frame of the second sequence of video to be compared and detected, and the video duration from the initial frame to the second frame of the second sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the second sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed of each second sequence of video to be compared and detected corresponding to the target task video and the dissimilar video duration corresponding to each second sequence of video to be compared and detected are statistically obtained.

[0063] Screening is performed based on the playback duration and preliminary matching coefficient corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested, to obtain an extraction control set corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested;

[0064] Through the extraction control set pairs corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared, key frames are extracted from the sub-videos to be analyzed corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared, and key frames of each first sequence of detection videos to be compared and each second sequence of detection videos to be compared are obtained. Video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed are obtained, and key frames of the target task video are extracted according to the time nodes of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared, and video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared are obtained;

[0065] The video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed, and the video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed of the target task video are recorded as the key frame information corresponding to the target task.

[0066] Video similarity analysis step: used to perform data analysis based on the key frame information corresponding to the target task, and obtain the video similarity between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected.

[0067] Furthermore, the specific execution method of the video similarity analysis step is as follows:

[0068] Obtain a set of feature graphics corresponding to the target task video of the detection task, and filter to obtain a set of feature graphics corresponding to the target task video and each detection video to be compared;

[0069] Feature information is extracted from the video images of each key frame according to the feature graphic set corresponding to the target task video, and feature information sets of the target task video and the video images of each key frame corresponding to each sub-video to be analyzed are obtained, wherein the feature information set includes a text information set, a feature graphic set, and a non-feature information set. The text information set refers to the text content in the video image, the feature image set includes the type of feature graphic corresponding to the video image and the number of each type of feature graphic, and the non-feature information set refers to the content information of the non-feature video image area excluding the text information set and the feature graphic set, and the non-feature information set includes the corresponding color histogram in the non-feature video image area;

[0070] The text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the target task video and the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is matched with the text content, and the text feature conformity coefficient F of the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is obtained. i p ;

[0071] It should be noted that: each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos still use the numbers of the to-be-compared detection videos.

[0072] By calculating the formula Calculate the characteristic graph matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ' p ;

[0073] By calculating the formula Calculate the non-feature matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ″ p ;

[0074] By calculating the formula Calculate the video similarity coefficient Q between the target task video and each video to be compared and tested i , where α represents the number of each type of feature graphics, β represents the number of types of feature graphics, u represents the number of each type of color, n represents the number of color types, p represents the number of each key frame, and q represents the number of key frames. It is expressed as the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed corresponding to each target task video. It is represented by the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed in each video to be compared and detected. It is expressed as the proportion of each color in each key frame video image of the target task video corresponding to each to-be-compared detection video corresponding to the to-be-analyzed sub-video. It is expressed as the proportion of each color corresponding to each key frame video image in the sub-video to be analyzed in each video to be compared and detected. It is represented by the length of dissimilar videos corresponding to the videos to be compared and detected;

[0075] The video similarity coefficients between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected and the corresponding numbers of each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected are screened to create a task log, and the task log is uploaded to the task management center.

[0076] To achieve the above object, the present invention further provides the following technical solution: a similar video detection device, comprising the following steps:

[0077] The video is acquired, and basic information corresponding to the video is obtained according to the video extraction to perform data analysis, so as to obtain each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos corresponding to the target task video;

[0078] Perform data analysis based on the basic information corresponding to the video and the first sequence of detection videos to be compared and the second sequence of detection videos to be compared corresponding to the target task video, and intelligently extract the key frame information corresponding to the target task;

[0079] Data analysis is performed based on the key frame information corresponding to the target task to obtain the video similarity between the target task video and each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos.

[0080] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A similar video detection method, characterized by: include: Basic data acquisition step: used to acquire the video, and perform data analysis based on the basic information corresponding to the video obtained by video extraction, to obtain each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video; Key frame extraction step: for performing data analysis based on the basic information corresponding to the video and each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, and intelligently extracting the key frame information corresponding to the target task; Video similarity analysis step: used to perform data analysis based on the key frame information corresponding to the target task, and obtain the video similarity between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected.

2. The similar video detection method according to claim 1, wherein: The specific implementation method of the basic data acquisition step is as follows: Obtain the target task video that needs to be tested for video similarity corresponding to the detection task; Obtain each video to be compared for the video similarity detection corresponding to the detection task; Extract data from the target task video to obtain basic information corresponding to the target task video, including the file size and playback time. Extract data from each video to be compared and tested to obtain basic information corresponding to each video to be compared and tested, where the basic information includes file size and playback time; By calculating the formula Calculate the initial matching coefficient Vs of the target task video corresponding to each video to be compared i , where td i 、fs i They represent the file size and playback duration of each video to be compared and tested, td′ and fs′ represent the file size and playback duration of the target task video, and i represents the number of each video to be compared and tested; The preliminary conformity coefficients of the target task video and the detection videos to be compared are compared and analyzed with the preset preliminary conformity coefficient threshold. If the preliminary conformity coefficient of the detection video to be compared is less than or equal to the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the first sequence detection video to be compared. If the preliminary conformity coefficient of the detection video to be compared is greater than the preset preliminary conformity coefficient threshold, the detection video to be compared is recorded as the second sequence detection video to be compared. The first sequence detection videos to be compared and the second sequence detection videos to be compared corresponding to the target task video are statistically obtained.

3. The similar video detection method according to claim 2, wherein: The specific implementation of the key frame extraction step is as follows: Establishing a data extraction relationship between the key frame extraction step and the database, extracting each key frame extraction set stored in the database, wherein the key frame extraction set includes the video playback duration, the detection sequence corresponding to the video, the preliminary matching coefficient corresponding to the video, and the extraction control set corresponding to the video, wherein the extraction control set includes the number of key frame extractions and the time nodes of each key frame; The key frame refers to an important frame in the video that can represent its content; The detection sequence includes a first sequence and a second sequence; Extracting the target task video, the video images of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video corresponding to the target task video, performing image comparison analysis on the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video, and obtaining image similarity between the video image of the first frame of the target task video and the video images of the first frame of each first sequence of to-be-compared detection video and each second sequence of to-be-compared detection video; If the image similarity with the first sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the first sequence, the first frame of the first sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the first sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the first sequence, the video images of the remaining frames corresponding to the first sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the first frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the first sequence. The frame is recorded as the initial frame of the first sequence of video to be compared and detected, and the video duration from the initial frame to the first frame of the first sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the first sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed corresponding to each first sequence of video to be compared and detected and the dissimilar video duration corresponding to each first sequence of video to be compared and detected of the target task video are statistically obtained. If the image similarity with the second sequence of video to be compared and detected is greater than the image similarity threshold corresponding to the second sequence, the second frame of the second sequence of video to be compared and detected is recorded as the initial frame. If the image similarity with the second sequence of video to be compared and detected is less than or equal to the image similarity threshold corresponding to the second sequence, the video images of the remaining frames corresponding to the second sequence of video to be compared and detected are sequentially extracted and image comparison and analysis are performed with the video image of the second frame corresponding to the target task video until the image similarity of the video image corresponding to a certain frame is greater than the image similarity threshold corresponding to the second sequence. The frame is recorded as the initial frame of the second sequence of video to be compared and detected, and the video duration from the initial frame to the second frame of the second sequence of video to be compared and detected is calculated and recorded as the dissimilar video duration corresponding to the second sequence of video to be compared and detected. The video after the initial frame is recorded as the sub-video to be analyzed, and the sub-video to be analyzed of each second sequence of video to be compared and detected corresponding to the target task video and the dissimilar video duration corresponding to each second sequence of video to be compared and detected are statistically obtained. Screening is performed based on the playback duration and preliminary matching coefficient corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested, to obtain an extraction control set corresponding to each first sequence of videos to be compared and tested and each second sequence of videos to be compared and tested; Through the extraction control sets corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, key frames are extracted for the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, and key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained, and video images of the key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained. Key frames are extracted for the target task video according to the time nodes of the key frames corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared, and video images of the key frames of the sub-videos to be analyzed corresponding to the first sequence detection videos to be compared and the second sequence detection videos to be compared are obtained; The video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed, and the video images of each key frame corresponding to each first sequence of detection videos to be compared and each second sequence of detection videos to be compared and each sub-video to be analyzed of the target task video are recorded as the key frame information corresponding to the target task.

4. The similar video detection method according to claim 3, wherein: The specific execution method of the video similarity analysis step is as follows: Obtain a set of feature graphics corresponding to the target task video of the detection task, and filter to obtain a set of feature graphics corresponding to the target task video and each detection video to be compared; Feature information is extracted from the video images of each key frame according to the feature graphic set corresponding to the target task video, and feature information sets of the target task video and the video images of each key frame corresponding to each sub-video to be analyzed are obtained, wherein the feature information set includes a text information set, a feature graphic set, and a non-feature information set. The text information set refers to the text content in the video image, the feature image set includes the type of feature graphic corresponding to the video image and the number of each type of feature graphic, and the non-feature information set refers to the content information of the non-feature video image area excluding the text information set and the feature graphic set, and the non-feature information set includes the corresponding color histogram in the non-feature video image area; The text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the target task video and the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is matched with the text content, and the text feature conformity coefficient F of the text content of each key frame video image in each first sequence of detection videos to be compared and each second sequence of detection videos to be compared corresponding to the sub-video to be analyzed is obtained. i p ; By calculating the formula Calculate the characteristic graph matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ' p ; By calculating the formula Calculate the non-feature matching coefficient F of each key frame video image in the target task video and each to-be-compared detection video corresponding to the to-be-analyzed sub-video i ″ p ; By calculating the formula Calculate the video similarity coefficient Q between the target task video and each video to be compared and tested i , where α represents the number of each type of feature graphics, β represents the number of types of feature graphics, u represents the number of each type of color, n represents the number of color types, p represents the number of each key frame, and q represents the number of key frames. It is expressed as the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed corresponding to each target task video. It is represented by the number of various feature graphics corresponding to each key frame video image in each sub-video to be analyzed in each video to be compared and detected. It is expressed as the proportion of each color in each key frame video image of the target task video corresponding to each to-be-compared detection video corresponding to the to-be-analyzed sub-video. It is expressed as the proportion of each color corresponding to each key frame video image in the sub-video to be analyzed in each video to be compared and detected. It is represented by the length of dissimilar videos corresponding to the videos to be compared and detected; The video similarity coefficients between the target task video and each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected and the corresponding numbers of each first sequence of videos to be compared and detected and each second sequence of videos to be compared and detected are screened to create a task log, and the task log is uploaded to the task management center.

5. A similar video detection device, applied to a similar video detection method according to any one of claims 1 to 4, characterized in that: include: The video is acquired, and basic information corresponding to the video is obtained according to the video extraction to perform data analysis, so as to obtain each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos corresponding to the target task video; Perform data analysis based on the basic information corresponding to the video and the first sequence of detection videos to be compared and the second sequence of detection videos to be compared corresponding to the target task video, and intelligently extract the key frame information corresponding to the target task; Data analysis is performed based on the key frame information corresponding to the target task to obtain the video similarity between the target task video and each first sequence of to-be-compared detection videos and each second sequence of to-be-compared detection videos.

Images