CN111368143A - Video similarity retrieval method and device, electronic equipment and storage medium - Google Patents

Abstract

The application relates to a video similarity retrieval method, a video similarity retrieval device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring a video clip to be retrieved of a video to be retrieved; analyzing the video clip to be retrieved to obtain a first feature vector; and comparing the first feature vector with a video feature library to obtain a retrieval result of the video to be retrieved. The technical scheme carries out retrieval by extracting the characteristic vector of the video segment to be retrieved to replace the traditional image characteristic vector. On the basis of ensuring the feasibility, the stability and the robustness of the characteristics are enhanced, and meanwhile, the retrieval precision is also improved.

Description

A video similarity retrieval method, device, electronic device and storage medium

technical field

The present application relates to the field of video retrieval, and in particular, to a method, apparatus, electronic device and storage medium for video similarity retrieval.

Background technique

Video retrieval is an important means to deal with copyright detection, infringement query, and video duplication check. At present, video retrieval methods are based on images in videos. The keyframes are compared with the video keyframes in the library to measure their similarity. This approach has the following drawbacks:

1. A single frame of image cannot replace the content of the video clip to be retrieved. There is a big limitation in using the comparison result between frames as the mapping relationship between video clips. This limitation will reduce the robustness of the algorithm. Slight displacement and distortion can cause inaccurate results.

2. Retrieval is performed by extracting key frames, so that the retrieval accuracy and speed are antagonistic. If you want to maintain a high retrieval accuracy, you need to reduce the extraction interval of key frames. Selecting multiple frames for comparison will definitely cost more time. Conversely, to speed up the retrieval speed, it is necessary to increase the interval for extracting key frames and reduce the number of comparisons, which may lead to a higher false detection rate.

3. Affected by the image feature extraction method and the feature comparison method, it is necessary to adjust the features according to the situation and increase the complexity of the hyperparameters.

SUMMARY OF THE INVENTION

In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides a video similarity retrieval method, apparatus, electronic device and storage medium.

In a first aspect, an embodiment of the present application provides a video similarity retrieval method, including:

Obtain the to-be-retrieved video clip of the to-be-retrieved video;

Analyzing the to-be-retrieved video segment to obtain a first feature vector;

The first feature vector is compared with the video feature library to obtain the retrieval result of the video to be retrieved.

Optionally, the first feature vector obtained by parsing the to-be-retrieved video clip includes:

Inputting the video segment to be retrieved into a feature extraction model, the feature extraction model includes: a plurality of 3D convolutional layers;

The plurality of 3D convolution layers sequentially convolve the to-be-retrieved video segments to obtain the first feature vector.

Optionally, the video feature library includes: a second feature vector, where the second feature vector is obtained by inputting video clips in the video library into the feature extraction model.

Optionally, comparing the first feature vector with a video feature library to obtain a retrieval result of the video to be retrieved, including:

Obtain a third eigenvector matching the first eigenvector from the second eigenvector;

Obtain the video set corresponding to the third feature vector;

establishing a mapping relationship between the video to be retrieved and all videos in the video set;

The mapping relationship is used as the retrieval result of the video to be retrieved.

Optionally, obtaining a third eigenvector matching the first eigenvector from the second eigenvector includes:

calculating the similarity between the first feature vector and the second feature vector;

The second feature vector whose similarity satisfies a preset condition is used as the third feature vector.

Optionally, the method further includes:

obtain at least two adjacent first feature vectors;

When the videos in the video set include at least two adjacent third feature vectors that match the at least two adjacent first feature vectors and have the same timing information, then confirm that the to-be-retrieved video matches the The videos in the video set are identical or partially identical.

Optionally, the method further includes:

Determine the repetition rate of the video to be retrieved according to the retrieval result;

When the repetition rate is greater than or equal to a preset threshold, a corresponding processing operation is performed on the video to be retrieved.

In a second aspect, an embodiment of the present application provides a video similarity retrieval device, including:

A determination module, used to determine the to-be-retrieved video segment of the to-be-retrieved video;

an extraction module for extracting at least one first feature vector based on the video to be retrieved;

A retrieval module, configured to compare the first feature vector with a preset video library to obtain a retrieval result of the video to be retrieved.

In a third aspect, the present application provides an electronic device, including: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

the memory for storing computer programs;

The processor implements the above method steps when executing the computer program.

In a fourth aspect, the present application provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the above method steps.

Compared with the prior art, the above technical solutions provided by the embodiments of the present application have the following advantages: the retrieval is performed by extracting the feature vector of the video segment to be retrieved instead of the traditional image feature vector. On the basis of ensuring the feasibility, the stability and robustness of the features are enhanced, and the retrieval accuracy is also improved.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and together with the description serve to explain the principles of the invention.

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. In other words, on the premise of no creative labor, other drawings can also be obtained from these drawings.

1 is a flowchart of a video similarity retrieval method provided by an embodiment of the present application;

2 is a working schematic diagram of a feature extraction model provided by an embodiment of the present application;

3 is a schematic diagram of a mapping relationship between a video to be retrieved and a video set provided by an embodiment of the present application;

4 is a flowchart of a video similarity retrieval method provided by another embodiment of the present application;

5 is a flowchart of a video similarity retrieval method provided by another embodiment of the present application;

6 is a block diagram of a video similarity retrieval apparatus provided by an embodiment of the present application;

FIG. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

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

The present application provides a video similarity retrieval method, device, electronic device, and storage medium. The method provided by the embodiment of the present invention can be applied to any required electronic device, for example, it can be an electronic device such as a server and a terminal, which is not described here. For specific limitations, for the convenience of description, it is hereinafter referred to as an electronic device.

The following first introduces a video similarity retrieval method provided by an embodiment of the present invention.

FIG. 1 is a flowchart of a video similarity retrieval method provided by an embodiment of the present application. As shown in Figure 1, the method includes the following steps:

Step S11, obtaining to-be-retrieved video clips of the to-be-retrieved video;

Step S12, analyzing the video segment to be retrieved to obtain a first feature vector;

Step S13, comparing the first feature vector with the video feature library to obtain a retrieval result of the video to be retrieved.

The video similarity retrieval method provided in this embodiment replaces the traditional image feature retrieval method by extracting the feature vector of the video segment to be retrieved, thereby enhancing the stability of the feature on the basis of ensuring the feasibility , robustness, and also improve the retrieval accuracy.

In this embodiment, the to-be-retrieved video is acquired, and then the to-be-retrieved video is divided according to a continuous preset number of frames to obtain to-be-retrieved video segments of the to-be-retrieved video.

Among them, the preset number of frames involved in this embodiment is 16 frames, the images of 16 consecutive frames are used as the video clips to be retrieved, and the video clips of 1 second less than one second are used as the retrieval unit to perform video retrieval, which significantly reduces the retrieval granularity. At the same time, the recognition accuracy is improved, and retrieval results can be given for different segments of the same video.

In this embodiment, the first feature vector is obtained by parsing the video clip to be retrieved, which is specifically implemented in the following manner: inputting the video clip to be retrieved into a feature extraction model, and the feature extraction model includes: multiple 3D convolution layers; The convolution layer sequentially convolves the video segments to be retrieved to obtain the first feature vector.

Fig. 2 is a working schematic diagram of a feature extraction model provided by an embodiment of the present application. As shown in Fig. 2, images of 16 consecutive frames are input into a pre-trained feature extraction model, the feature extraction model performs convolution calculation, and outputs a 256-dimensional array , taking the 256-dimensional data as the first feature vector.

The parameters of the 16 consecutive frames of images in this embodiment include: a length of 112 pixels, a width of 112 pixels, and a 3-channel image. The feature extraction model used in this embodiment includes 6 layers of convolution layers. The function of the convolution layer is to perform feature extraction on the input data, and it contains multiple convolution kernels. Each element constituting the convolution kernel corresponds to a weight coefficient and A bias vector, similar to the neuron of a feedforward neural network. The parameters of the convolutional layer include convolution kernel size, stride and padding, which together determine the size of the output feature map of the convolutional layer. The size of the convolution kernel can be specified as an arbitrary value smaller than the size of the input image. The larger the convolution kernel, the more complex the input features that can be extracted.

The convolution layer used in this embodiment is a 3D convolution layer. By using 3D convolution to compress data, it can ensure fast feature extraction, high compression efficiency, save operation time and reduce storage cost. The parameters of each layer are as follows:

Conv1 convolution kernel: 1×3×3×3×64, stride[1,1,2,2,1], padding 0;

Conv2 convolution kernel: 1×5×5×64×128, stride[1,1,5,5,1], padding 0;

Conv3 convolution kernel: 3×1×1×128×256, stride[1,2,1,1,1], padding 0;

Conv4 convolution kernel: 3×1×1×256×512, stride[1,2,1,1,1], padding 0;

Conv5 convolution kernel: 2×2×2×512×1536, stride[1,2,2,2,1], padding 0;

Conv6 convolution kernel: 2×7×7×1536×256, stride[1,1,1,1,1], padding 0.

The feature extraction model is trained by the following methods: obtaining training samples, wherein the training samples may be 16 consecutive frames of images, and using the training samples to train a preset convolutional neural network model to obtain a feature extraction model.

In this embodiment, the first feature vector is compared with the video feature library to obtain the retrieval result of the video to be retrieved, which is specifically implemented in the following manner:

First, obtain the video feature library and the video library. It can be understood that the video library is a collection of complete videos. The video feature library stores a second feature vector, and the second feature vector is obtained in the following way: according to 16 frames as a video The segment segments the videos in the video library to obtain multiple video segments, and input the obtained video segments into a pre-trained feature extraction model to obtain a second feature vector.

It should be noted that, in this embodiment, the 3D convolution layer in the feature extraction model is used to convolve the video clips of 16 consecutive frames. On the one hand, the speed of video feature extraction is ensured, the convolution efficiency is high, and the time cost is saved. . On the other hand, the present embodiment uses video clips of 16 consecutive frames as a retrieval unit, which significantly reduces the retrieval granularity and improves the accuracy of retrieval results compared with the prior art.

Then, obtain a third feature vector matching the first feature vector from the second feature vector, obtain a video set corresponding to the third feature vector, establish a mapping relationship between the video to be retrieved and all videos in the video set, and use the mapping relationship as The retrieval result of the video to be retrieved.

Optionally, obtaining a third eigenvector matching the first eigenvector from the second eigenvector is by calculating the similarity between the first eigenvector and the second eigenvector. The second feature vector whose similarity satisfies the preset condition is used as the third feature vector. As an example, the similarity between the first feature vector and all the second feature vectors in the video feature library is calculated, and the second feature vector with the similarity greater than or equal to 0 is used as the third feature vector. The similarity referred to in this embodiment is the cosine similarity. Assuming the existing 256-dimensional first eigenvector A and second eigenvector B, the cosine similarity calculation formula is as follows:

In the formula, ||A|| is the modulus of vector A, and ||B|| is the modulus of vector B.

After obtaining the third feature vector, determine the video clip corresponding to each third feature vector, query the video to which each video clip belongs, and obtain a video set according to the video to which each video clip belongs, that is, the video set corresponding to the third feature vector .

Establishing the mapping relationship between the video to be retrieved and all the videos in the video set may be to count the video clips corresponding to the video clips to be retrieved, sort the video clips according to the similarity, and eliminate videos with a similarity less than a preset similarity. Thus, the mapping relationship between each to-be-retrieved video segment in the to-be-retrieved video and all the videos in the video set is obtained.

For example: the video M to be retrieved includes: the video segment to be retrieved M1, the video segment to be retrieved M2, the video segment to be retrieved M3 and the video segment to be retrieved M4, the video set includes: video Q, video P, video O, video G, video K and video R.

The mapping relationship is: the video segment M1 to be retrieved corresponds to the video segment Q2 (similarity 95%) and the video segment P2 (similarity 90%).

The video segment M2 to be retrieved corresponds to the video segment Q3 (similarity 99%) and video segment P3 (similarity 96%).

The video segment O1 (similarity 98%) and the video segment G1 (similarity 94%) matching the to-be-retrieved video segment M3.

The video segment K2 (similarity 97%) and the video segment R3 (similarity 90%) matching the to-be-retrieved video segment M4.

Thus, the mapping relationship between the to-be-retrieved video M and all the videos in the video set is obtained.

FIG. 3 is a flowchart of a video similarity retrieval method provided by another embodiment of the present application. As shown in Figure 3, the method further includes the following steps:

Step S21, obtaining at least two adjacent first feature vectors;

Step S22, when the video in the video set includes at least two adjacent third feature vectors that match with at least two adjacent first feature vectors and have the same timing information, then confirm that the video to be retrieved is the same as the video set in the video set. The videos are identical or partially identical.

In this embodiment, a video corresponding to a certain part of the video to be retrieved can be obtained by verifying the adjacent first feature vector and the third feature vector in the video library, thereby improving the validity and accuracy of the retrieval result.

FIG. 4 is a schematic diagram of a mapping relationship between a video to be retrieved and a video set provided by an embodiment of the present application. As shown in FIG. 4 , the video to be retrieved includes: a first part Z ₁ , a second part Z ₂ and a third part Z ₃ .

It is assumed that the first part of the video to be retrieved includes: the video segment to be retrieved Z ₁₁ , the video segment to be retrieved Z ₁₂ , the video segment to be retrieved Z ₁₃ and the video segment to be retrieved Z ₁₄ . After retrieval, the video segment Z ₁₁ to be retrieved corresponds to the segment A _j1 of the video A in the video set, the video segment Z ₁₂ to be retrieved corresponds to the segment A _j2 of the video A in the video set, and the video segment Z ₁₃ to be retrieved corresponds to the segment A j2 of the video A in the video set. The segment A _j3 of the video A corresponds to the segment A j4 of the video A to be retrieved, and the to-be-retrieved video segment Z ₁₄ corresponds to the segment A _j4 of the video A in the video set.

At this time, the timing information is used to judge. If the timing information of the four segments in the first part of the video to be retrieved is the same as that of the four segments of the video A in the video set, it is considered that the first part of the video to be retrieved is the same as the video A part in the video set. same.

At the same time, the to-be-retrieved video segment Z ₁₁ corresponds to the segment B _j1 of the video B in the video set, the to-be-retrieved video segment Z ₁₂ corresponds to the segment B _j2 of the video B in the video set, and the to-be-retrieved video segment Z ₁₃ corresponds to the video B in the video set The segment B _j3 of the to-be-retrieved video segment Z ₁₄ corresponds to the segment B _j4 of the video B in the video set.

At this time, the timing information is used to judge. If the timing information of the four segments in the first part of the video to be retrieved is the same as the timing information of the four segments in the video B in the video set, it is considered that the first part of the video to be retrieved is the same as the video B in the video set. same.

The second part of the video to be checked includes: the video segment to be retrieved Z ₂₁ , the video segment to be retrieved Z ₂₂ , the video segment to be retrieved _{Z 23} and the video segment to be retrieved Z ₂₄ . The segment C j1 of the video C corresponds to the segment C _j1 of the video C to be retrieved, the segment Z ₂₂ of the video to be retrieved corresponds to the segment C _j2 of the video C in the video set, and the segment Z ₂₃ of the video to be retrieved corresponds to the segment C _j3 of the video C in the video set.

At this time, the timing information is used to judge. If the timing information of the four segments in the first part of the video to be retrieved is the same as that of the four segments in the video C in the video set, it is considered that the first part of the video to be retrieved is the same as the video C in the video set. same.

The second part of the video to be checked includes: a video segment Z ₃₁ to be retrieved and a video segment Z ₃₂ to be retrieved. After retrieval, it is obtained that the video segment Z ₃₁ to be retrieved corresponds to the segment D j1 of the video D in the video set, and the video segment Z ₃₂ to be retrieved corresponds to the segment D _j1 of the video D in the video set. Corresponds to segment D _j2 of video D in the video set.

At this time, the timing information is used to judge. If the timing information of the four segments in the first part of the video to be retrieved is the same as that of the four segments in the video D in the video set, it is considered that the first part of the video to be retrieved is the same as the video D in the video set. same.

FIG. 5 is a flowchart of a video similarity retrieval method provided by another embodiment of the present application. As shown in Figure 5, the method further includes the following steps:

Step 31, determine the repetition rate of the video to be retrieved according to the retrieval result;

Step 32, when the repetition rate is greater than or equal to the preset threshold, perform corresponding processing operations on the video to be retrieved.

In this embodiment, the repetition rate of the video to be retrieved is determined according to the retrieval result, and whether the video to be retrieved is a redundant video is determined by the repetition rate. As an example: the video to be retrieved includes: a first part, a second part and a third part, where the first part and the second part correspond to a certain video in the video set, and then calculate the video duration of the first part and the second part, according to the video The repetition rate is calculated based on the duration and the total duration of the video to be retrieved. For example, if the video duration of the first part and the second part is 4.5 minutes, and the total duration of the video to be retrieved is 15 minutes, the repetition rate is 30%.

When the repetition rate is greater than or equal to the preset threshold, a corresponding processing operation is performed on the video to be retrieved. As an example: when the repetition rate is greater than the preset threshold of 90%, it is determined that the video to be retrieved is a redundant video, and it can be eliminated, so as to prevent a large number of redundant videos from occupying the resources of the video library.

The video similarity retrieval method disclosed in this embodiment can also be applied to video management. For example, there is a 24-hour live stream, which includes multiple programs of interest to users and other common videos. Reminders to watch or perform other actions. Considering that the program of interest has a relatively fixed title, the title can be extracted and stored as a sample in the video library, and then retrieved in the video library using the live stream as the video to be checked. When the similarity of the inner samples is greater than the threshold, it can be considered that the current segment is the title of the program, and the user is prompted to start subsequent operations.

In addition, the video similarity retrieval method disclosed in this embodiment can also be used to find and recommend similar videos. For example, after a user watches a certain video, he hopes to see content related to this video. At this time, the video watched by the user can be The videos to be retrieved are compared in the entire video library and the retrieved videos are sorted in descending order of similarity, and finally the sorting result is used as the video content recommended to the user.

FIG. 6 is a block diagram of a video similarity retrieval apparatus provided by an embodiment of the present application, and the apparatus may be implemented as part or all of an electronic device through software, hardware, or a combination of the two. As shown in Figure 6, the device includes:

A determination module 41, configured to determine the to-be-retrieved video segment of the to-be-retrieved video;

Extraction module 42, for extracting at least one first feature vector based on the video to be retrieved;

The retrieval module 43 is configured to compare the first feature vector with a preset video library to obtain a retrieval result of the video to be retrieved.

In this embodiment, the extraction module 42 is specifically configured to obtain a pre-trained feature extraction model, and input the video segment to be retrieved into the feature extraction model to obtain a first feature vector.

In this embodiment, the video feature library includes: a second feature vector, where the second feature vector is obtained by inputting the video clips in the video library into the feature extraction model.

In this embodiment, the retrieval module 43 includes:

The first obtaining submodule is used to obtain the third feature vector matching the first feature vector from the second feature vector;

The second acquisition sub-module is used to acquire the video set corresponding to the third feature vector;

The storage submodule is used to establish a mapping relationship between the video to be retrieved and all the videos in the video set, and use the mapping relationship as the retrieval result of the video to be retrieved.

Wherein, the first obtaining sub-module is specifically configured to: calculate the similarity between the first feature vector and the second feature vector, and use the second feature vector whose similarity meets a preset condition as the third feature vector.

The video similarity retrieval apparatus provided in this embodiment further includes: an acquisition module, configured to acquire at least two adjacent first feature vectors, when the videos in the video set include at least two adjacent first feature vectors that match , and at least two adjacent third feature vectors with the same timing information, it is confirmed that the video to be retrieved is the same or partially the same as the video in the video set.

The video similarity retrieval apparatus provided in this embodiment further includes: a processing module configured to determine the repetition rate of the video to be retrieved according to the retrieval result, and perform a corresponding processing operation on the video to be retrieved when the repetition rate is greater than or equal to a preset threshold.

This embodiment of the present application further provides an electronic device. As shown in FIG. 7 , the electronic device may include: a processor 1501, a communication interface 1502, a memory 1503, and a communication bus 1504, wherein the processor 1501, the communication interface 1502, and the memory 1503 pass through The communication bus 1504 performs communication with each other.

The memory 1503 is used to store computer programs;

The processor 1501 is configured to implement the steps of the foregoing embodiments when executing the computer program stored in the memory 1503 .

The communication bus mentioned in the above electronic device may be a peripheral component interconnect standard (Peripheral Component Interconnect, PCI) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus or the like. The communication bus can be divided into an address bus, a data bus, a control bus, and the like. For ease of presentation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

The communication interface is used for communication between the above electronic device and other devices.

The memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory, NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

The above-mentioned processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; it may also be a digital signal processor (Digital Signal Processing, DSP), an application-specific integrated circuit ( Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

The present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the steps of the above-mentioned embodiments.

It should be noted that, for the above embodiments of the apparatus, electronic device and computer-readable storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and for related details, please refer to some descriptions of the method embodiments.

It is further to be noted that in this document, relational terms such as "first" and "second" etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that Any such actual relationship or sequence exists between these entities or operations. Moreover, the terms "comprising", "comprising" or any other variation thereof are intended to encompass a non-exclusive inclusion such that a process, method, article or device that includes a list of elements includes not only those elements, but also includes not explicitly listed or other elements inherent to such a process, method, article or apparatus. Without further limitation, an element qualified by the phrase "comprising a..." does not preclude the presence of additional identical elements in a process, method, article or apparatus that includes the element.

The above descriptions are only specific embodiments of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims (10)

1. a video similarity retrieval method, is characterized in that, comprises: Obtain the to-be-retrieved video clip of the to-be-retrieved video; Analyzing the to-be-retrieved video segment to obtain a first feature vector; The first feature vector is compared with the video feature library to obtain the retrieval result of the video to be retrieved. 2. The method according to claim 1, wherein the obtaining the first feature vector by parsing the to-be-retrieved video segment comprises: Inputting the video segment to be retrieved into a feature extraction model, the feature extraction model includes: a plurality of 3D convolutional layers; The plurality of 3D convolution layers sequentially convolve the to-be-retrieved video segments to obtain the first feature vector. 3. The method according to claim 2, wherein the video feature library comprises: a second feature vector, wherein the second feature vector is obtained by inputting video clips in the video library into the feature extraction model of. 4. The method according to claim 3, wherein the comparing the first feature vector with a video feature library to obtain a retrieval result of the video to be retrieved comprises: Obtain a third eigenvector matching the first eigenvector from the second eigenvector; Obtain the video set corresponding to the third feature vector; establishing a mapping relationship between the video to be retrieved and all videos in the video set; The mapping relationship is used as the retrieval result of the video to be retrieved. 5. The method according to claim 4, wherein the obtaining a third eigenvector matching the first eigenvector from the second eigenvector comprises: calculating the similarity between the first feature vector and the second feature vector; The second feature vector whose similarity satisfies a preset condition is used as the third feature vector. 6. The method according to claim 4, wherein the method further comprises: obtain at least two adjacent first feature vectors; When the videos in the video set include at least two adjacent third feature vectors that match the at least two adjacent first feature vectors and have the same timing information, then confirm that the to-be-retrieved video matches the The videos in the video set are identical or partially identical. 7. The method of claim 1, wherein the method further comprises: Determine the repetition rate of the video to be retrieved according to the retrieval result; When the repetition rate is greater than or equal to a preset threshold, a corresponding processing operation is performed on the video to be retrieved. 8. A video similarity retrieval device, characterized in that, comprising: A determination module, used to determine the to-be-retrieved video segment of the to-be-retrieved video; an extraction module for extracting at least one first feature vector based on the video to be retrieved; A retrieval module, configured to compare the first feature vector with a preset video library to obtain a retrieval result of the video to be retrieved. 9. An electronic device, comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; the memory for storing computer programs; The processor, when executing the computer program, implements the method steps of any one of claims 1-7. 10. A computer-readable storage medium on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the method steps of any one of claims 1-7 are implemented.

Images