WO2021175040A1

WO2021175040A1 - Video processing method and related device

Info

Publication number: WO2021175040A1
Application number: PCT/CN2021/073333
Authority: WO
Inventors: 尹康
Original assignee: Oppo广东移动通信有限公司
Priority date: 2020-03-02
Filing date: 2021-01-22
Publication date: 2021-09-10
Also published as: CN111274446A

Abstract

The present application provides a video processing method and a related device, comprising: first, extracting N pieces of video feature data of N videos comprised in a video data set, N being a positive integer; next, obtaining matching degree data of every two pieces of video feature data among the N pieces of video feature data; then, dividing the N videos into M video clustering clusters on the basis of the matching degree data, M being a positive integer less than or equal to N; and finally, performing deduplication processing on the M video clustering clusters one by one on the basis of a preset deduplication rule to obtain a deduplicated video data set, the deduplicated video data set comprising M videos. Repeated videos in the video data set can be accurately clustered by means of an efficient feature extraction algorithm, and then the clustered repeated videos are subjected to deduplication, thereby greatly improving the accuracy of video deduplication.

Description

Video processing method and related device

Technical field

This application relates to the technical field of data deduplication, in particular to a video processing method and related devices.

Background technique

With the development of technology, deep learning theory has become the mainstream solution in basic image processing fields such as image classification and target detection, and has also received extensive attention in the field of video processing. When building a video processing-related model, a large amount of training data is needed for training, and the scale and quality of the training data set directly affect the speed and accuracy of model building. However, the video data collection process will inevitably introduce a large amount of duplicate data. Therefore, in order to improve the performance of the video processing model based on the deep learning theory, it is necessary to perform deduplication operations on the data set in advance.

The current common video deduplication algorithm is based on key point matching to remove duplication, but the process of using key points to extract image features is too cumbersome, and clustering algorithms such as k-means used in feature matching require manual pre-setting of the number of categories Such parameters cannot guarantee the accuracy of the final de-duplication processing.

Summary of the invention

Based on the above problems, this application proposes a video processing method and related devices, which can accurately cluster the repeated videos in the video data set through an efficient feature extraction algorithm, and then de-duplicate the clustered repeated videos, which greatly improves The accuracy of video deduplication.

The first aspect of the embodiments of the present application provides a video processing method, including:

Extract N video feature data of N videos included in the video data set, where N is a positive integer;

Acquiring the matching degree data of every two pieces of video feature data among the N pieces of video feature data;

Dividing the N videos into M video clusters based on the matching degree data, where M is a positive integer less than or equal to N;

Perform deduplication processing on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set, and the deduplicated video data set includes M videos.

A second aspect of the embodiments of the present application provides a video processing device. The device includes a processing unit and a communication unit, wherein:

The processing unit is configured to extract N video feature data of N videos included in the video data set, where N is a positive integer; obtain the matching degree data of every two video feature data between the N video feature data; based on The matching degree data divides the N videos into M video clusters, where M is a positive integer less than or equal to N; perform deduplication processing on the M video clusters one by one based on a preset deduplication rule To obtain a deduplicated video data set, where the deduplicated video data set includes M videos.

The third aspect of the embodiments of the present application provides an electronic device, including an application processor, a communication interface, and a memory. The application processor, the communication interface, and the memory are connected to each other. The memory is used to store a computer program. The computer program includes program instructions, and the application processor is configured to invoke the program instructions to execute all or part of the steps of the method described in the first aspect of the embodiments of the present application.

The fourth aspect of the embodiments of the present application provides a computer storage medium, the computer storage medium stores a computer program, the computer program includes program instructions, and when executed by a processor, the program instructions cause the processor to execute such as All or part of the steps of the method described in the first aspect of the embodiments of the present application.

The fifth aspect of the embodiments of the present application provides a computer program product, wherein the above-mentioned computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the above-mentioned computer program is operable to cause a computer to execute the computer program as in the embodiment of the present application. Part or all of the steps described in any method of the first aspect. The computer program product may be a software installation package.

By implementing the above application examples, the following beneficial effects can be obtained:

In the above video processing method and related device, firstly, N video feature data of N videos included in the video data set are extracted, where N is a positive integer; then, the value of every two video feature data between the N video feature data is obtained. Matching degree data; then, based on the matching degree data, the N videos are divided into M video clusters, where M is a positive integer less than or equal to N; finally, one pair of the M is paired based on a preset deduplication rule Deduplication processing is performed on two video clusters to obtain a deduplicated video data set, and the deduplicated video data set includes M videos. The repeated videos in the video data set can be accurately clustered through an efficient feature extraction algorithm, and then the clustered repeated videos can be deduplicated, which greatly improves the accuracy of the video deduplication.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the following will briefly introduce the drawings used in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. Ordinary technicians can obtain other drawings based on these drawings without creative work.

FIG. 1 is a system architecture diagram of a video processing method provided by an embodiment of this application;

2 is a schematic flowchart of a video processing method provided by an embodiment of this application;

FIG. 3 is a schematic structural diagram of an electronic device provided by an embodiment of the application;

FIG. 4 is a block diagram of functional units of a video processing device provided by an embodiment of the application.

Detailed ways

In order to enable those skilled in the art to better understand the solutions of the application, the technical solutions in the embodiments of the application will be clearly and completely described below in conjunction with the drawings in the embodiments of the application. Obviously, the described embodiments are only It is a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific sequence. In addition, the terms "including" and "having" and any variations of them are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes unlisted steps or units, or optionally also includes Other steps or units inherent to these processes, methods, products or equipment.

The reference to "embodiments" herein means that a specific feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art clearly and implicitly understand that the embodiments described herein can be combined with other embodiments.

The electronic devices involved in the embodiments of the application may be electronic devices with communication capabilities. The electronic devices may include various handheld devices with wireless communication functions, vehicle-mounted devices, wearable devices, computing devices, or other devices connected to wireless modems. Processing equipment, as well as various forms of user equipment (User Equipment, UE), mobile station (Mobile Station, MS), terminal equipment (terminal device), and so on.

The following describes the embodiments of the application in detail.

Figure 1 is a system architecture diagram of a video processing method provided by an embodiment of the application, including a video acquisition module 110, a matching module 120, a classification module 130, and a deduplication processing module 140. The video acquisition module 110, the matching module 120, The classification module 130 and the deduplication processing module 140 are connected to each other. The video acquisition module 110 may acquire a video data set composed of videos to be processed, and send the video data set to the matching module 120, and the matching module 120 The video to be processed in the received video data set may be matched, and the matching result may be sent to the classification module 130. The classification module 130 may classify the to-be-processed video according to the matching result to obtain multiple videos. Clustering, each video cluster is one video or multiple repeated videos, and finally the deduplication processing module 140 performs deduplication processing on each video cluster to obtain a deduplicated video data set, and complete the video Steps to de-duplication.

It should be noted that the training data of the neural network model may have a large amount of duplicate data. Using all training data to train the model is not efficient and will cause the accuracy of the neural model to decrease. Therefore, a large amount of training data is deduplicated and automatically processed. It is very important to select training data with better training effects. The system architecture in the embodiment of the present application can be applied to a scenario of screening training data of a neural network model related to video processing.

Through the above system architecture, the repeated videos in the video data set can be accurately clustered through an efficient feature extraction algorithm, and then the clustered repeated videos can be deduplicated, which greatly improves the accuracy of video deduplication.

The following describes in detail a video processing method in an embodiment of the present application with reference to FIG. 2. FIG. 2 is a schematic flowchart of a video processing method provided by an embodiment of the present application, which specifically includes the following steps:

Step 201: Extract N video feature data of N videos included in the video data set.

Among them, the above-mentioned video data set is a set of N to-be-processed videos, and N is any positive integer, and each video can be processed to extract the video feature data corresponding to each video. For ease of understanding, the following is a set of any video The video feature data extraction step is explained.

First, read the video frame by frame to obtain the image data of each frame of the video, and then extract the feature vector of each frame of image data through the Perceptual Hash Algorithm (PHA). The feature vector extraction steps of one frame of image data are described in detail. The above single frame of image data is a color image, including three color channels of red (Red), green (Green), and blue (Blue). The single-frame image is converted into a single-channel gray-scale image, and the size of the converted gray-scale image is normalized to a size of 32×32 pixels through a bilinear interpolation algorithm to improve the extraction efficiency, and then a discrete cosine transform (Discrete cosine transform) is performed. Cosine Transform, DCT), obtain a 32×32 coefficient matrix, and then select 64 coefficients in the 8×8 area at the upper left corner of each coefficient matrix to perform ordinary quantization to obtain a binary image. The above-mentioned ordinary quantization processing rules are as follows:

Finally, the 64-dimensional vector obtained by the above-mentioned binary image flattening is used as the feature vector of the frame of image data. Similarly, each frame of image data is processed to obtain the feature vector corresponding to each frame of image.

As described above, after the feature vector corresponding to each frame of image of a video is obtained, the video feature data corresponding to the video can be obtained based on the aforementioned feature vector.

Optionally, the above-mentioned video feature data may be a feature sequence, and the above-mentioned one feature sequence may be understood as a collection of all feature vectors corresponding to each frame of image data of a video, and the feature vector of each frame of the video can be graded. To obtain the above feature sequence. Specifically, you can initialize an empty list first, then obtain the timestamp corresponding to each frame of image data of the video data, and then calculate the 64-dimensional vector corresponding to each frame of image data according to the order of the timestamp Arrange them in sequence and add the above empty list to obtain a feature sequence corresponding to the video. Repeat the above steps until N feature sequences corresponding to the above N videos are obtained. The above different videos correspond to different feature sequences. It should be noted that, When generating a feature sequence, the above feature vectors can be down-sampled based on different application scenarios, that is, a feature vector cascade is extracted every 2 frames, 4 frames, etc. to obtain the feature sequence. The number of video frames is different and its corresponding features The length of the sequence may also be different.

Optionally, the above-mentioned video feature data may be a video feature vector, and the above-mentioned video feature vector is a multi-dimensional vector composed of image feature vectors of each frame of image. In order to improve the extraction efficiency, the size of the converted gray image is normalized to a size of 32×32 pixels through the bilinear interpolation algorithm, and then the discrete cosine transform (DCT) is performed to obtain 32 ×32 coefficient matrix, and then select 64 coefficients in the 8×8 area at the upper left corner of each coefficient matrix to perform special quantization to obtain a special binary image. The above-mentioned special quantization processing rules are as follows:

After that, the 64-dimensional special vector obtained by flattening all the special binary images after the above special quantization is superimposed, and finally the superimposed 64-dimensional special vector is subjected to the above-mentioned ordinary quantization to generate the above-mentioned video feature vector, and the above-mentioned video feature vector can reflect the corresponding The content information of the video.

By extracting N video feature data of N videos included in the video data set, two types of video feature data can be extracted in two ways, which can cope with multiple video processing scenarios, and greatly improves the flexibility of subsequent video processing.

Step 202: Obtain the matching degree data of every two pieces of video feature data among the N pieces of video feature data.

Wherein, the above-mentioned matching degree data represents the similarity between every two video characteristic data in the N pieces of video feature data, and the above-mentioned matching degree data may be equivalent to the similarity between every two videos in the N pieces of video.

Optionally, when the above-mentioned video feature data is a feature sequence, since the length of the feature sequence may be different, conventional methods such as measuring the Euclidean distance of the feature sequence cannot be used to calculate the similarity between every two feature sequences. The function calculates the length of the longest common subsequence between every two feature sequences in N feature sequences. For example, the video data set is represented by V={v1,v2,...,vN}, that is, the above video data set V There are N videos v, and any two different feature sequences in the N feature sequences are represented by Fi and Fj, where Fi={fi1,fi2,...,fin}, Fj={fj1,fj2,...,fjm} , That is, any feature sequence Fi is composed of n feature vectors f corresponding to n frames of image data, and another feature sequence Fj is composed of m feature vectors f corresponding to m frames of image data. The above m and n may be the same or different , The matching function match(Fi,Fj) can determine the longest common subsequence F*=(f*1,f*2,...,f*k) between any two feature sequences, where F*∈Fi, F *∈Fj and F* is the longest common subsequence among all common subsequences, k is the length of the above-mentioned longest common subsequence, and calculate the length of the longest common subsequence of every two feature sequences in turn until N The feature sequence is calculated, combined with the combination formula

get

The length of the longest common subsequence.

Optionally, when the above-mentioned video feature data is a video feature vector, the Manhattan distance (Manhattan Distance) is calculated, and the Manhattan distance between every two video feature vectors is used as the above-mentioned matching degree data. The specific Manhattan distance calculation step can be The existing algorithm is used, and will not be repeated here.

Step 203: Divide the N videos into M video clusters based on the matching degree data.

Where M is a positive integer less than or equal to N, each video cluster can include at least one video, that is, multiple videos with repeated content will be classified into the same video cluster, and there is no video with repeated content. A single video can form a video cluster by itself.

Optionally, when the matching degree data is the length of the longest common subsequence, a preset length threshold can be set. If the longest common subsequence of any two videos is greater than the preset length threshold, it means the longest common subsequence The corresponding two videos are repeated video sets, and each of the N videos needs to be matched with other videos pairwise to obtain the length of the corresponding longest common subsequence.

It can be described _{in pseudo code in combination with the expressions of F i} and F _j above, as shown below:

Among them, the above-mentioned output video cluster cluster set C includes M video clusters, and the above-mentioned N-dimensional flag vector is used to determine whether the video has been added to a certain video cluster, if a certain video cluster has been added Cluster, it is not necessary to judge again whether the longest common subsequence between it and other videos is greater than the above-mentioned preset length threshold.

Specifically, taking the first video as an example, it can be determined whether the length of the longest common subsequence between the first video and the second video is greater than the aforementioned preset length threshold. The length of the longest common subsequence is greater than the above preset length threshold, it means that the first video and the second video are repeated video sets, and the first video and the second video need to be divided into the first video cluster. If the length of the longest common subsequence between the first video and the second video is less than or equal to the aforementioned preset length threshold, the first video and the second video are different videos and do not belong to the same video cluster; Then determine in turn whether the length of the longest common subsequence between the first video and the third video, and the fourth video to the Nth video is greater than the preset length threshold. The length of the longest common subsequence is greater than the above preset length threshold, it means that the first video and the third video are also repeated video sets, and the first video, the second video, and the third video are all repeated videos If the length of the longest common subsequence between the first video and the third video is less than or equal to the aforementioned preset length threshold, the third video needs to be divided into the first video cluster. It does not belong to the same video cluster as the third video; by analogy, the videos that are duplicate video sets with the first video among the N videos are screened out, and are divided into the first video cluster; the first After the video cluster is determined, you can continue to determine the second video cluster, that is, determine whether the length of the longest common subsequence between the second video and the third video to the Nth video is greater than the aforementioned preset length threshold As described above, the second video among the N videos is screened out as the videos of the repeated video set, and the second video is divided into the second video clusters, until the above N videos are divided into M video clusters.

Optionally, when the matching degree data is the Manhattan distance between video feature vectors, a hierarchical clustering algorithm (Hierarchical Density-Based Spatial Clustering of Applications with Noise, HDBSCAN) to divide N videos into M video clusters. It should be noted that, compared with the method of using matching functions to divide video clusters, HDBSCAN is used to divide video clusters. The clustering speed can be improved, but the method using the matching function is more accurate, and the method of dividing video clusters can be flexibly switched based on different application requirements.

It can be seen that dividing the N videos into M video clusters based on the matching degree data can quickly and accurately divide the repeated videos into the corresponding video clusters, which greatly improves the video deduplication processing. accuracy.

Step 204: Perform deduplication processing on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set.

Wherein, the video data set after deduplication includes M videos, that is, only one video is reserved for each video cluster. The preset deduplication rule may include at least one deduplication index data, and the deduplication index data may include videos. Any one or any combination of video-related data such as duration index, video editing frequency index, video image quality index, video format index, video quality index, etc., select different deduplication index data based on different application scenarios, the above video duration index It can be the longest video duration or the shortest video duration, etc. The above-mentioned video editing frequency index can be the editing limit such as the least video editing frequency or the most video editing frequency, the above-mentioned video quality index can be the clearest video quality or the video quality The blurriest picture quality is limited, the above-mentioned video format indicator may be the MP4 format or the AVI format, and the above-mentioned video quality indicator may be the highest video quality or the lowest video quality.

For example, if the deduplication index data is a video duration index and the video duration index at this time is the limit of the longest video duration, the preset deduplication rule is to retain the video with the longest video duration in each video cluster, and delete Go to other videos to get the video data set after deduplication. Similarly, when the de-duplication index data is any one or any combination of video-related data such as video editing times index, video image quality index, video format index, video quality index, etc., it is also based on the corresponding preset de-duplication rule to Deduplication processing is performed on each video cluster to obtain a corresponding deduplication video data set, which will not be repeated here.

Among them, the video data set can have a mapping relationship with the deduplication index data in the preset deduplication rules. The deduplication index data can be changed manually, or the deduplication index data that best fits the video data set can be automatically selected according to the video data set. , There is no specific limitation here.

It can be seen that based on the preset deduplication rules, the M video clusters are deduplicated one by one to obtain the deduplicated video data set, which can be flexibly adapted to different application scenarios to perform the most appropriate deduplication on the video data set. Processing greatly improves the accuracy and versatility of video deduplication.

The following describes an electronic device 300 in an embodiment of the application with reference to FIG. 3. FIG. 3 is a schematic structural diagram of an electronic device 300 provided in an embodiment of the application, including an application processor 301, a communication interface 302, and a memory 303. The application processor 301, the communication interface 302, and the memory 303 are connected to each other through a bus 304, which can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) Bus and so on. The bus 304 can be divided into an address bus, a data bus, a control bus, and so on. For ease of presentation, only a thick line is used in FIG. 3 to represent it, but it does not mean that there is only one bus or one type of bus. Wherein, the memory 303 is used to store a computer program, the computer program includes program instructions, and the application processor 301 is configured to call the program instructions to perform the following steps:

In a possible embodiment, the video feature data includes a feature sequence; in terms of extracting the N video feature data of the N videos in the video data set, the instructions in the program are specifically used to perform the following operations:

Acquiring each frame of image data of each video;

Extracting the feature vector of each frame of image data through a perceptual hash algorithm;

The feature vector of each frame of image data is concatenated to obtain a feature sequence corresponding to each video, and the feature sequence is used to represent the content feature of the video.

In a possible embodiment, in terms of acquiring the matching degree data of every two pieces of video feature data among the N pieces of video feature data, the instructions in the program are specifically used to perform the following operations:

Obtain the longest common subsequence between every two feature sequences in N feature sequences based on the matching function;

The length of each longest common subsequence is determined as the matching degree data of every two pieces of video feature data among the N pieces of video feature data.

In a possible embodiment, in terms of dividing the N videos into M video clusters based on the matching degree data, the instructions in the program are specifically used to perform the following operations:

Determining two videos corresponding to the longest common subsequence whose length of the longest common subsequence is greater than a preset length threshold as a repeated video set;

Acquiring coincident video information between each repeated video set, where the coincident video information is used to indicate whether the same video exists between each repeated video set;

According to the coincident video information, the N videos included in all repeated video sets are divided into the M video clusters.

In a possible embodiment, in terms of performing deduplication processing on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set, the instructions in the program are specifically used To do the following:

Acquiring a preset deduplication rule corresponding to the video data set, where the preset deduplication rule includes deduplication index data;

Screen out the videos that meet the deduplication index data in each video cluster for retention;

The M videos retained in the M video clusters are used as the deduplicated video data set.

In a possible embodiment, where the feature vector includes a 64-dimensional vector; in terms of extracting the feature vector of each frame of image data through a perceptual hash algorithm, the instructions in the program are specifically used to perform the following operations:

Converting the image data of each frame into 32×32 pixel grayscale image data;

Processing the gray image data by discrete cosine transform to obtain a 32×32 coefficient matrix;

The 64 coefficients in the 8×8 area at the upper left position of each coefficient matrix are selected for quantization, and the 64-dimensional vector of each frame of image data is obtained.

In a possible embodiment, in terms of cascading the feature vectors of each frame of image data to obtain the feature sequence corresponding to each video, the instructions in the program are specifically used to perform the following operations:

Acquiring the time stamp of each frame of image data;

The 64-dimensional vectors are arranged in sequence according to the sequence of the time stamps to generate the feature sequence corresponding to each video.

The foregoing mainly introduces the solution of the embodiment of the present application from the perspective of the execution process on the method side. It can be understood that, in order to implement the above-mentioned functions, an electronic device includes hardware structures and/or software modules corresponding to each function. Those skilled in the art should easily realize that in combination with the units and algorithm steps of the examples described in the embodiments provided herein, this application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed by hardware or computer software-driven hardware depends on the specific application and design constraint conditions of the technical solution. Professionals and technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered beyond the scope of this application.

The embodiment of the present application may divide the electronic device into functional units according to the foregoing method examples. For example, each functional unit may be divided corresponding to each function, or two or more functions may be integrated into one processing unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, and is only a logical function division, and there may be other division methods in actual implementation.

FIG. 4 is a block diagram of functional units of a video processing device 400 provided by an embodiment of the present application. The video processing device 400 is applied to an electronic device, and includes a processing unit 401, a communication unit 402, and a storage unit 403. The processing unit 401 is configured to perform any step in the above-mentioned method embodiment, and is performing such as During data transmission such as sending, the communication unit 402 can be optionally invoked to complete the corresponding operation. The detailed description will be given below.

The processing unit 401 is configured to extract N video feature data of N videos included in a video data set, where N is a positive integer;

In a possible embodiment, the video feature data includes a feature sequence; in terms of extracting N video feature data of N videos in the video data set, the processing unit 401 is specifically configured to:

Acquiring each frame of image data of each video;

In a possible embodiment, the processing unit 401 is specifically configured to:

In a possible embodiment, in terms of dividing the N videos into M video clusters based on the matching degree data, the processing unit 401 is specifically configured to:

In a possible embodiment, the processing unit 401 is specifically configured to perform deduplication processing on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set. :

In a possible embodiment, where the feature vector includes a 64-dimensional vector; the processing unit 401 is specifically configured to:

Converting the image data of each frame into 32×32 pixel grayscale image data;

In a possible embodiment, in the aspect of cascading the feature vectors of each frame of image data to obtain the feature sequence corresponding to each video, the processing unit 401 is specifically configured to:

Acquiring the time stamp of each frame of image data;

An embodiment of the present application also provides a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute part or all of the steps of any method as recorded in the above method embodiment , The above-mentioned computer includes electronic equipment.

The embodiments of the present application also provide a computer program product. The above-mentioned computer program product includes a non-transitory computer-readable storage medium storing a computer program. Part or all of the steps of the method. The computer program product may be a software installation package, and the above-mentioned computer includes electronic equipment.

It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that this application is not limited by the described sequence of actions. Because according to this application, some steps can be performed in other order or at the same time. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

In the above-mentioned embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in an embodiment, reference may be made to related descriptions of other embodiments.

In the several embodiments provided in this application, it should be understood that the disclosed device may be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of the above-mentioned units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components can be combined or integrated. To another system, or some features can be ignored, or not implemented. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical or other forms.

The units described above as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

In addition, the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

If the above integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable memory. Based on this understanding, the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory. A number of instructions are included to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the foregoing methods of the various embodiments of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.

Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by a program instructing relevant hardware. The program can be stored in a computer-readable memory, and the memory can include: a flash disk , Read-only memory (English: Read-Only Memory, abbreviation: ROM), random access device (English: Random Access Memory, abbreviation: RAM), magnetic disk or optical disc, etc.

The embodiments of the application are described in detail above, and specific examples are used in this article to illustrate the principles and implementation of the application. The descriptions of the above embodiments are only used to help understand the methods and core ideas of the application; at the same time, for Those of ordinary skill in the art, based on the idea of the application, will have changes in the specific implementation and the scope of application. In summary, the content of this specification should not be construed as a limitation to the application.

Claims

A video processing method, characterized in that the method includes:

Extract N video feature data of N videos included in the video data set, where N is a positive integer;

Acquiring the matching degree data of every two pieces of video feature data among the N pieces of video feature data;

Dividing the N videos into M video clusters based on the matching degree data, where M is a positive integer less than or equal to N;

Perform deduplication processing on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set, and the deduplicated video data set includes M videos.
The method according to claim 1, wherein the video feature data includes a feature sequence; and the extraction of N video feature data of N videos in a video data set includes:

Acquiring each frame of image data of each video;

Extracting the feature vector of each frame of image data through a perceptual hash algorithm;

The feature vector of each frame of image data is concatenated to obtain a feature sequence corresponding to each video, and the feature sequence is used to represent the content feature of the video.
The method according to claim 2, wherein the obtaining the matching degree data of every two pieces of video feature data among the N pieces of video feature data comprises:

Obtain the longest common subsequence between every two feature sequences in N feature sequences based on the matching function;

The length of each longest common subsequence is determined as the matching degree data of every two pieces of video feature data among the N pieces of video feature data.
The method according to claim 3, wherein the dividing the N videos into M video clusters based on the matching degree data comprises:

Determining two videos corresponding to the longest common subsequence whose length of the longest common subsequence is greater than a preset length threshold as a repeated video set;

Acquiring coincident video information between each repeated video set, where the coincident video information is used to indicate whether the same video exists between each repeated video set;

According to the coincident video information, the N videos included in all repeated video sets are divided into the M video clusters.
The method according to claim 1, wherein the video feature data includes a video feature vector; and the extracting N video feature data of N videos in a video data set includes:

Acquiring each frame of image data of each video;

Extracting the image feature vector of each frame of image data through a perceptual hash algorithm;

The image feature vector of each frame of image data is superimposed to form the video feature vector.
The method according to claim 5, wherein the obtaining the matching degree data of every two pieces of video feature data among the N pieces of video feature data comprises:

Acquiring Manhattan distance data between every two video feature vectors in the N video feature vectors;

Each Manhattan distance data is determined as the matching degree data of every two pieces of video feature data among the N pieces of video feature data.
The method according to any one of claims 1 to 6, wherein the deduplication processing is performed on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video data set, include:

Acquiring a preset deduplication rule corresponding to the video data set, where the preset deduplication rule includes deduplication index data;

Screen out the videos that meet the deduplication index data in each video cluster for retention;

The M videos retained in the M video clusters are used as the deduplicated video data set.
The method according to claim 2, wherein the feature vector includes a 64-dimensional vector; and the extraction of the feature vector of each frame of image data through a perceptual hash algorithm includes:

Converting the image data of each frame into 32×32 pixel grayscale image data;

Processing the gray image data by discrete cosine transform to obtain a 32×32 coefficient matrix;

The 64 coefficients in the 8×8 area at the upper left position of each coefficient matrix are selected for quantization, and the 64-dimensional vector of each frame of image data is obtained.
The method according to claim 6, wherein the cascading the feature vectors of each frame of image data to obtain the feature sequence corresponding to each video comprises:

Acquiring the time stamp of each frame of image data;

The 64-dimensional vectors are arranged in sequence according to the sequence of the time stamps to generate the feature sequence corresponding to each video.
A video processing device, characterized in that the device includes a processing unit and a communication unit, wherein:

The processing unit is configured to extract N video feature data of N videos included in the video data set, where N is a positive integer; obtain the matching degree data of every two video feature data between the N video feature data; based on The matching degree data divides the N videos into M video clusters, where M is a positive integer less than or equal to N; perform deduplication processing on the M video clusters one by one based on a preset deduplication rule To obtain a deduplicated video data set, where the deduplicated video data set includes M videos.
The video processing device according to claim 10, wherein the video feature data includes a feature sequence; in terms of extracting N video feature data of N videos in the video data set, the processing unit is configured to:

Acquiring each frame of image data of each video;

Extracting the feature vector of each frame of image data through a perceptual hash algorithm;

The feature vector of each frame of image data is concatenated to obtain a feature sequence corresponding to each video, and the feature sequence is used to represent the content feature of the video.
The video processing device according to claim 11, wherein, in terms of obtaining the matching degree data of every two pieces of video feature data among the N pieces of video feature data, the processing unit is configured to:

Obtain the longest common subsequence between every two feature sequences in N feature sequences based on the matching function;

The length of each longest common subsequence is determined as the matching degree data of every two pieces of video feature data among the N pieces of video feature data.
The video processing device according to claim 12, wherein in the aspect of dividing the N videos into M video clusters based on the matching degree data, the processing unit is configured to:

Determining two videos corresponding to the longest common subsequence whose length of the longest common subsequence is greater than a preset length threshold as a repeated video set;

Acquiring coincident video information between each repeated video set, where the coincident video information is used to indicate whether the same video exists between each repeated video set;

According to the coincident video information, the N videos included in all repeated video sets are divided into the M video clusters.
The video processing device according to claim 10, wherein the video feature data includes a video feature vector; in terms of extracting N video feature data of N videos in the video data set, the processing unit is configured to :

Acquiring each frame of image data of each video;

Extracting the image feature vector of each frame of image data through a perceptual hash algorithm;

The image feature vector of each frame of image data is superimposed to form the video feature vector.
The video processing device according to claim 14, wherein, in the aspect of obtaining the matching degree data of every two pieces of video feature data among the N pieces of video feature data, the processing unit is configured to:

Acquiring Manhattan distance data between every two video feature vectors in the N video feature vectors;

Each Manhattan distance data is determined as the matching degree data of every two pieces of video feature data among the N pieces of video feature data.
The video processing device according to any one of claims 10 to 15, wherein the deduplication processing is performed on the M video clusters one by one based on a preset deduplication rule to obtain a deduplicated video In terms of data sets, the processing unit is used to:

Acquiring a preset deduplication rule corresponding to the video data set, where the preset deduplication rule includes deduplication index data;

Screen out the videos that meet the deduplication index data in each video cluster for retention;

The M videos retained in the M video clusters are used as the deduplicated video data set.
The video processing device according to claim 11, wherein the feature vector comprises a 64-dimensional vector; in the aspect of extracting the feature vector of each frame of image data through the perceptual hash algorithm, the processing unit is configured to :

Converting the image data of each frame into 32×32 pixel grayscale image data;

Processing the gray image data by discrete cosine transform to obtain a 32×32 coefficient matrix;

The 64 coefficients in the 8×8 area at the upper left position of each coefficient matrix are selected for quantization, and the 64-dimensional vector of each frame of image data is obtained.
The video processing device according to claim 15, wherein in the aspect of concatenating the feature vectors of each frame of image data to obtain the feature sequence corresponding to each video, the processing unit is configured to:

Acquiring the time stamp of each frame of image data;

The 64-dimensional vectors are arranged in sequence according to the sequence of the time stamps to generate the feature sequence corresponding to each video.
An electronic device, characterized by comprising an application processor, a communication interface, and a memory, the application processor, the communication interface, and the memory are connected to each other, wherein the memory is used to store a computer program, and the computer program includes program instructions The application processor is configured to call the program instructions to execute the method according to any one of claims 1-9.
A computer storage medium, wherein the computer storage medium stores a computer program, the computer program includes program instructions, and when executed by a processor, the program instructions cause the processor to execute as claimed in claims 1-9. Any of the methods.