Intelligent management method and system for high-definition video resources

Abstract

The invention discloses a high-definition video resource intelligent management method and system, and relates to the technical field of video management, and the method comprises the steps: obtaining video resources and user behavior data, dividing the video resources into video frames, extracting the multi-dimensional features of the video frames through an image processing algorithm, dividing the video frames into sub-band channels, and calculating local distortion; a Minkowski convergence mode is applied, and local distortion of all channels is synthesized into frame perception complexity score expansion video frame multi-dimensional features; and obtaining the complexity level and the membership degree corresponding to each video frame through FCM clustering complexity grading, calculating the complexity of a single frame, and calculating the resolution of each video frame according to the complexity of the single frame to form a video resource. According to the invention, accurate analysis of the video frame is realized, the resolution of the video frame is adjusted by calculating the complexity of the single frame to form the video resource, dynamic adjustment of different image quality requirements is realized, and the watching experience of a user is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of video management, and in particular to a method and system for intelligent management of high-quality video resources. Background Art

[0002] With the rapid development of information technology, the storage, transmission and processing technologies of video resources have also been greatly improved. Especially in the fields of video streaming, video surveillance and high-definition video playback, how to effectively manage video resources and improve user experience has become a hot issue in current technical research and application. Traditional video processing methods mostly rely on simple compression and encoding technologies, using standard video encoding formats (such as H.264, HEVC, etc.) for video compression. These methods can effectively reduce the size of video files and ensure the smoothness of video transmission under limited bandwidth. In recent years, with the continuous advancement of deep learning and big data technologies, the research on video content analysis and management methods has gradually developed towards intelligent and personalized directions. However, existing technologies still face some challenges in processing video resources. Existing methods rely too much on traditional visual perception models and lack in-depth analysis of the multi-dimensional features of video frames, resulting in insufficient precision in the management and optimization of high-quality video resources. In addition, there are certain limitations in accurately measuring the complexity of video frames and refining user behavior analysis. Summary of the Invention

[0003] In view of the above existing problems, the present invention is proposed.

[0004] Therefore, the present invention provides a method and system for intelligent management of high-definition video resources to solve the problem that the existing methods are overly dependent on traditional visual perception models and lack in-depth analysis of the multi-dimensional features of video frames, resulting in insufficiently refined management and optimization of high-definition video resources, and certain limitations in the accurate measurement of video frame complexity and the refinement of user behavior analysis.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a method for intelligent management of high-quality video resources, which comprises:

[0007] Obtain video resources and user behavior data, divide the video resources into video frames, use image processing algorithms to extract multidimensional features of the video frames, divide the video frames into sub-band channels and calculate local distortion, apply Minkowski aggregation, combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frames;

[0008] Obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. Calculate the resolution of each video frame based on the complexity of a single frame to form a video resource.

[0009] Use a multi-layer perceptron to classify the multi-dimensional features of video frames and comprehensively analyze video categories. Analyze user preferences based on user behavior data and push videos to users based on video categories.

[0010] The user behavior data includes browsing history and rating data.

[0011] As a preferred solution of the high-quality video resource intelligent management method of the present invention, wherein: the multi-dimensional features of the video frame are extracted using an image processing algorithm, the video frame is divided into sub-band channels and local distortion is calculated, and the Minkowski convergence method is applied to synthesize the local distortion of all channels into a frame perception complexity score to expand the multi-dimensional features of the video frame, including:

[0012] Use Sobel operator and Canny to extract the number of edge pixels in the video frame as the edge intensity E t ;

[0013] Convert the video frame into a grayscale image and construct the gray-level co-occurrence matrix as the texture feature T t ;

[0014] The optical flow method is used to calculate the pixel difference between two adjacent video frames as the motion change M t ;

[0015] The edge strength E t , texture feature T t And the motion change M t Synthetic feature vector x t ;

[0016] For each video frame, Wavelet transform is used to divide the video frame into sub-band channels C t，l , using contrast sensitivity function filtering to weight each channel to simulate the sensitivity of the human eye to video frame images

[0017] Taking the adjacent video frames as reference, calculate the error signal e of each sub-band channel t，l ;

[0018] Adjusting perceptual sensitivity B based on local contrast energy t，l , simulating saturation of optic nerve response;

[0019] Based on adjusting the perceptual sensitivity B t，l The error signal e of each sub-band channel t，l Normalized to "perceptible difference" units as local distortion;

[0020] Apply Minkowski aggregation to combine the local distortions of all channels into a frame perception complexity score A t ;

[0021] Synchronously use the structural similarity index to evaluate the structural consistency Q between video frames t ;

[0022] Score the frame perceptual complexity A t and structural consistency Q t Expand and add the multi-dimensional features of the video frame, and express the expanded multi-dimensional features of the video frame as x′ t .

[0023] As a preferred solution of the intelligent management method of high-quality video resources of the present invention, wherein: the complexity level and membership of each video frame are obtained by FCM clustering complexity classification and the complexity of a single frame is calculated, and the resolution of each video frame is calculated according to the complexity of the single frame to form a video resource index based on the expanded multi-dimensional feature x′ of the video frame. t Define the fuzzy clustering objective function J m Build an FCM model;

[0024] Based on the FCM model, the membership degree and cluster center are iteratively updated and optimized until the maximum membership degree after iteration converges and then the iteration is stopped;

[0025] Output the membership of the video frame, and take the complexity level with the highest membership as the complexity level of the video frame.

[0026] And perform weighted summation to output the video frame single frame complexity P t ;

[0027] Calculate the resolution Z of each video frame based on the complexity of the single frame of the video frame t ;

[0028] The video frames are adjusted according to the calculated video frame resolution, and the adjusted video frames are recomposed into the video assets.

[0029] As a preferred solution of the intelligent management method of high-quality video resources of the present invention, wherein: the multi-layer perceptron is used to classify the multi-dimensional features of the video frame and the comprehensive analysis of the video category refers to the multi-dimensional feature x′ of each video frame in the video resource. t Perform average pooling to obtain the overall feature vector F of the video resource;

[0030] A multilayer perceptron is selected as the classifier, which includes an input layer, a hidden layer, and an output layer. The input layer inputs the overall feature vector F, and the output is the video category. The multilayer perceptron is trained using the cross entropy loss function, and the overall feature vector F is input into the multilayer perceptron to obtain the video category.

[0031] As a preferred embodiment of the high-quality video resource intelligent management method of the present invention, the method of analyzing user preferences based on user behavior data and pushing videos to users based on video categories comprises extracting the types of videos viewed, viewing duration, and viewing frequency from the user's browsing history, and obtaining the priority of each video category of the user by taking a weighted sum of the viewing duration and viewing frequency of each video category;

[0032] For each video category, the user's rating data for all videos in that video category is counted, and the average score is calculated as the user's video evaluation;

[0033] The user's video category priority and user video evaluation are weighted and summed to obtain the user's comprehensive score for the video category. The video categories are sorted in descending order according to the comprehensive score to form a list of recommended video categories, and videos are pushed to users according to the video categories in the list.

[0034] As a preferred solution of the high-definition video resource intelligent management method described in the present invention, the obtaining of video resources and user behavior data, and dividing the video resources into video frames refers to collecting user behavior data by embedding data acquisition software in a video player, synchronously collecting video resources, dividing the video resources into frames to form video frames, and preprocessing the collected user behavior data.

[0035] As a preferred solution of the high-definition video resource intelligent management method described in the present invention, category tags are assigned to videos after they are classified, video resources are classified and stored according to the category tags, and push records are generated and stored synchronously after each video push.

[0036] In a second aspect, the present invention provides a high-quality video resource intelligent management system, comprising:

[0037] Data acquisition module, used to obtain video resources and user behavior data and perform preprocessing;

[0038] The feature extraction module is used to extract multidimensional features of video frames using image processing algorithms, divide the video frames into sub-band channels and calculate local distortion. Then, the Minkowski aggregation method is applied to combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frame.

[0039] The video adjustment module is used to obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. The resolution of each video frame is calculated based on the complexity of the single frame to form a video resource;

[0040] The video push module is used to classify the multi-dimensional features of video frames through a multi-layer perceptron, comprehensively analyze video categories, analyze user preferences based on user behavior data, and push videos to users based on video categories.

[0041] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the high-definition video resource intelligent management method as described in the first aspect of the present invention is implemented.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the high-definition video resource intelligent management method as described in the first aspect of the present invention is implemented.

[0043] The beneficial effects of the present invention are as follows: the present invention expands the multidimensional features of video frames by collecting video resources, extracting multidimensional features of video frames, and calculating frame perception complexity scores, thereby realizing accurate analysis of video frames, adjusting the video frame resolution by calculating the complexity of a single frame to form video resources, realizing dynamic adjustment of different picture quality requirements, improving user viewing experience, and combining user behavior data analysis to more accurately push personalized video recommendations to users. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a flowchart of the high-definition video resource intelligent management method in Example 1.

[0046] Figure 2 This is a structural diagram of the high-definition video resource intelligent management system in Example 1. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0050] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for intelligent management of high-quality video resources, comprising the following steps:

[0051] S1. Obtain video resources and user behavior data, divide the video resources into video frames, use image processing algorithms to extract multidimensional features of the video frames, divide the video frames into sub-band channels and calculate local distortion, apply Minkowski aggregation, combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frames;

[0052] Specifically, obtaining video resources and user behavior data, and dividing video resources into video frames means collecting user behavior data by embedding data acquisition software in a video player, synchronously collecting video resources, dividing the video resources into frames to form video frames, and preprocessing the collected user behavior data.

[0053] Furthermore, we use image processing algorithms to extract multidimensional features of video frames, divide the video frames into sub-band channels and calculate local distortion. Then, we apply Minkowski aggregation to synthesize the local distortion of all channels into a frame perception complexity score. The multidimensional features of video frames are expanded to include:

[0054] Use Sobel operator and Canny to extract the number of edge pixels in the video frame as the edge intensity E t ;

[0055] By applying the Sobel operator and the Canny edge detection algorithm, edge information in video frames can be effectively extracted and the structural features of the image can be enhanced, which is crucial for video quality evaluation. The efficient extraction of edge strength allows the edge information in the video to be fully considered, improving the ability to recognize video details.

[0056] Convert the video frame into a grayscale image and construct the gray-level co-occurrence matrix as the texture feature T t ;

[0057] The gray-level co-occurrence matrix is used to capture the texture features of an image. It can accurately reflect the changes and details of the texture in the image, and helps to improve the analysis accuracy of the video frame.

[0058] The optical flow method is used to calculate the pixel difference between two adjacent video frames as the motion change M t ;

[0059] The optical flow method is used to calculate the motion changes of adjacent video frames, so that the dynamic changes of video content can be evaluated, providing supplementary information of motion features for the processing of subsequent video frames.

[0060] The edge strength E t , texture feature T t And the motion change M t Synthetic feature vector x t ;

[0061] For each video frame, Wavelet transform is used to divide the video frame into sub-band channels C t，l , using contrast sensitivity function filtering to weight each channel to simulate the sensitivity of the human eye to video frame images

[0062]

[0063] Among them H l is the CSF (contrast sensitivity function) filter weight;

[0064] The Wavelet transform's frequency domain division of video frames can meticulously capture all levels of video content, enabling more accurate weighting of different features in subsequent analysis. This improves the comprehensiveness of video frame content and the sophistication of processing. By weighting different frequency bands, it can visually enhance the contrast perception of video frames, simulating the visual sensitivity of the human eye, and further improving the quality of video frames and the viewing experience.

[0065] Taking the adjacent video frames as reference, calculate the error signal e of each sub-band channel t，l :

[0066]

[0067] in and is the sensitivity of the current video frame and the previous video frame of the l-th sub-band channel;

[0068] Adjusting perceptual sensitivity B based on local contrast energy t，l , simulating saturation of the optic nerve response:

[0069] B t，l =B0+α*Lt，l ;

[0070] Where B0 is the initial perception sensitivity, L t，l is the local contrast energy, which is obtained by the sum of the squares of the brightness gradient in the local area, and α is the adjustment coefficient;

[0071] Based on adjusting the perceptual sensitivity B t，l The error signal e of each sub-band channel t，l Normalized to "perceptible difference" units As a local distortion:

[0072]

[0073] Apply Minkowski aggregation to combine the local distortions of all channels into a frame perception complexity score A t :

[0074]

[0075] Where L is the total number of sub-band channels, β is the nonlinear sensitivity coefficient;

[0076] By calculating local distortion and adjusting perceptual sensitivity, we can accurately simulate the human eye's perceptual differences in different video frame areas, ensuring that each processed video frame can achieve reasonable quality optimization under different viewing conditions. The Minkowski convergence method provides a nonlinear convergence method that makes the distortion synthesis of each sub-band channel more sensitive and adaptable, and can produce more accurate complexity scores under different visual conditions. In this way, the generated frame perceptual complexity score can not only reflect the visual complexity of the video frame, but also provide efficient support for subsequent resource management, recommendation systems and quality adjustment.

[0077] Synchronously use the structural similarity index to evaluate the structural consistency Q between video frames t :

[0078]

[0079] where σ xy is the covariance, μ x and μ y is the mean of the current video frame and the previous video frame, σ x and σ y is the standard deviation between the current video frame and the previous video frame;

[0080] Score the frame perceptual complexity A t and structural consistency Q t Expand and add the multi-dimensional features of the video frame, and express the expanded multi-dimensional features of the video frame as x′ t .

[0081] Combined with local distortion, the structural similarity index further enhances the ability to preserve the structure between video frames, making the video content more coherent and the visual effect more natural. Especially during video playback, it can effectively avoid the discomfort caused by sudden changes in the inter-frame structure and improve the user experience.

[0082] S2. Obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. Calculate the resolution of each video frame based on the complexity of the single frame to form a video resource.

[0083] Specifically, the complexity level and membership of each video frame are obtained through FCM clustering complexity classification and the complexity of a single frame is calculated. The resolution of each video frame is calculated based on the complexity of a single frame to form a video resource index based on the expanded multi-dimensional feature x′ of the video frame. t Define the fuzzy clustering objective function J m Building the FCM model:

[0084]

[0085] Where N is the total number of video frames, Q is the complexity classification, u tj is the membership of the video frame t to the jth class of complexity, obtained by the fuzzy membership formula, m is the fuzzy coefficient, q j is the jth cluster center, initially obtained by setting;

[0086] Based on the FCM model, the membership degree and cluster center are iteratively updated and optimized until the maximum membership degree after iteration converges and then the iteration is stopped;

[0087] The formula for iteratively updating the membership degree and cluster center is as follows:

[0088]

[0089] where q o is the oth cluster center;

[0090] The total number of complexity classifications is determined based on fuzzy partition coefficient (PC) and fuzzy entropy (PE);

[0091] Output the membership of the video frame, take the complexity level with the highest membership as the complexity level of the video frame, and perform weighted summation to output the single frame complexity P of the video frame t :

[0092]

[0093] where w j is the weight of the j-th category complexity;

[0094] Calculate the resolution Z of each video frame based on the complexity of the single frame of the video frame t :

[0095]

[0096] where Z max is the maximum resolution, P max is the maximum single-frame complexity of the video frame;

[0097] The video frames are adjusted according to the calculated video frame resolution, and the adjusted video frames are recomposed into the video assets.

[0098] Through the FCM clustering method, video frames can be divided into multiple complexity levels. This fuzzy classification method enables the membership of each video frame between complexity levels to be accurately evaluated. Compared with the traditional hard clustering method (that is, each video frame belongs to only one category), FCM provides a more fine-grained classification, effectively avoiding the risk of over-simplification of video frame complexity. The use of fuzzy membership allows video frames to dynamically adapt to different complexity categories, which is crucial for subsequent resolution adjustment and video quality management. By calculating the membership of each video frame in different complexity categories, it is possible to more flexibly assign appropriate processing strategies to each video frame. By iteratively updating the cluster center, the FCM algorithm can effectively avoid the deviation that may be caused by the initial cluster center, making the complexity grading of video frames more accurate. Each iterative optimization will improve the accuracy of clustering, ensuring that the final classification result can reflect the true complexity of the video frame. The iterative update process stops when the maximum membership converges, thereby ensuring the stability of the clustering results, avoiding overfitting or underfitting problems in the clustering process, and ensuring the efficiency and rationality of complexity grading. The weighted summation calculation of single-frame complexity can more accurately evaluate the overall complexity of each video frame and take into account the different effects of different complexity levels on video quality. In this way, multi-dimensional weights can be introduced into complexity evaluation to better reflect the real perceived complexity of the video frame, providing accurate single-frame complexity data for subsequent video resource management, so that the resolution adjustment of the video frame can be personalized and optimized according to the actual complexity requirements. By adjusting the resolution of frames with higher complexity, the video quality of high-complexity areas can be effectively improved, thereby improving the user's viewing experience. By adjusting the resolution of each video frame according to its complexity, the present invention realizes dynamic management of video resources. Frames with higher complexity may require higher resolution to maintain image quality, while frames with lower complexity can use lower resolution, thereby effectively reducing bandwidth occupancy while ensuring video quality. This method can greatly improve the utilization efficiency of video resources and adapt to different viewing environments.

[0099] S3. Classify the multi-dimensional features of video frames through a multi-layer perceptron, comprehensively analyze video categories, analyze user preferences based on user behavior data, and push videos to users based on video categories;

[0100] Specifically, the multi-dimensional features of the video frames are classified by a multi-layer perceptron, and the video category is comprehensively analyzed to convert the multi-dimensional features x′ of each video frame in the video resource into t Perform average pooling to obtain the overall feature vector F of the video resource;

[0101] A multilayer perceptron is selected as the classifier, which includes an input layer, a hidden layer, and an output layer. The input layer inputs the overall feature vector F, and the output is the video category, including sports videos, movie videos, news videos, etc. The cross entropy loss function is used to train the multilayer perceptron, and the overall feature vector F is input into the multilayer perceptron to obtain the video category.

[0102] By performing average pooling on the multi-dimensional features of video frames, the invention can effectively reduce the data dimension while retaining the key information of the video resources. This pooling process enhances the feature expression of video resources and improves the accuracy and efficiency of subsequent classification by reducing redundant information. Merging single-frame features into an overall feature vector helps to describe the video content from a global perspective, avoiding the errors that may be caused by the local features of a single frame, making the classification results more accurate and reliable. The multi-layer perceptron can automatically extract and learn high-level features in video frames through training, which means that the model can efficiently classify without manual feature extraction. This automatic learning capability reduces the need for human intervention and improves the classification efficiency. Accuracy. Multilayer perceptrons can process complex input data through multiple layers of nonlinear activation functions. Therefore, they are particularly suitable for processing complex patterns in video data and can fully explore the subtle differences between different categories in video content. For example, sports videos, movie videos, and news videos have significant differences in image, sound, and motion features. MLP can adaptively extract key patterns from these high-dimensional information and output precise video categories, which helps to quickly identify video content and provide accurate classification basis for subsequent applications such as video recommendation, playback, or advertising. For example, sports videos may require special push strategies or advertising, while news videos can be recommended based on current hot topics.

[0103] Furthermore, analyzing user preferences based on user behavior data and pushing videos to users based on video categories means extracting the types of videos viewed, viewing time, and viewing frequency from the user's browsing history, and obtaining the priority of each video category of the user by taking a weighted sum of the viewing time and viewing frequency of each video category;

[0104] For each video category, the user's rating data for all videos in that video category is counted, and the average score is calculated as the user's video evaluation;

[0105] The user's video category priority and user video evaluation are weighted and summed to obtain the user's comprehensive score for the video category. The video categories are sorted in descending order according to the comprehensive score to form a list of recommended video categories, and videos are pushed to users according to the video categories in the list.

[0106] By analyzing the user's browsing history and viewing time on different video categories, the system can obtain detailed information about the user's interests. For example, if a user watches sports videos for a long time and at a high frequency, it means that he or she has a strong interest in sports content. The system can make more accurate recommendations based on this information. Through these behavioral data, the system can not only identify the user's basic interests, but also generate personalized content push strategies for users based on multi-dimensional information such as browsing frequency and viewing time, thereby avoiding overly single and mechanical recommendation methods and improving the personalization and intelligence of recommendations. Through weighted summation, it can be based on the importance of video content in different categories. And the actual preferences of users, flexibly adjust the priority of each video category. For example, some users may occasionally browse certain video categories, but when they watch and repeatedly browse certain specific categories of content for a long time, the priority of the category will be significantly improved, which helps to more accurately capture changes in user preferences. The comprehensive score combines the user's interest depth (through priority) and quality perception (through rating) in each video category, which can more accurately reflect the user's comprehensive interests. Through descending sorting, the system can dynamically adjust the recommendation strategy according to the user's comprehensive preferences, making the recommendation more in line with user needs and avoiding recommendations biased towards one aspect (such as based only on viewing time or ratings).

[0107] The user behavior data includes browsing history and rating data.

[0108] After the video is classified, a category label is assigned to the video, and the video resources are classified and stored according to the category label. After each video push, a push record is generated and stored synchronously.

[0109] This embodiment also provides a high-quality video resource intelligent management system, including:

[0110] Data acquisition module, used to obtain video resources and user behavior data and perform preprocessing;

[0111] The feature extraction module is used to extract multidimensional features of video frames using image processing algorithms, divide the video frames into sub-band channels and calculate local distortion. Then, the Minkowski aggregation method is applied to combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frame.

[0112] The video adjustment module is used to obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. The resolution of each video frame is calculated based on the complexity of the single frame to form a video resource.

[0113] The video push module is used to classify the multi-dimensional features of video frames through a multi-layer perceptron, comprehensively analyze video categories, analyze user preferences based on user behavior data, and push videos to users based on video categories.

[0114] This embodiment also provides a computer device suitable for the intelligent management method of high-definition video resources, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the intelligent management method of high-definition video resources proposed in the above embodiment.

[0115] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0116] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for intelligent management of high-quality video resources as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0117] In summary, the present invention expands the multidimensional features of video frames by collecting video resources, extracting multidimensional features of video frames, and calculating frame perception complexity scores, thereby realizing accurate analysis of video frames. By calculating the complexity of single frames, the video frame resolution is adjusted to form video resources, dynamic adjustment of different picture quality requirements is achieved, the user viewing experience is improved, and personalized video recommendations are pushed to users more accurately in combination with user behavior data analysis.

[0118] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A high-quality video resource intelligent management method, characterized by: include, Obtain video resources and user behavior data, divide the video resources into video frames, use image processing algorithms to extract multidimensional features of the video frames, divide the video frames into sub-band channels and calculate local distortion, apply Minkowski aggregation, combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frames; Obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. Calculate the resolution of each video frame based on the complexity of a single frame to form a video resource. Use a multi-layer perceptron to classify the multi-dimensional features of video frames and comprehensively analyze video categories. Analyze user preferences based on user behavior data and push videos to users based on video categories. The user behavior data includes browsing history and rating data.

2. The intelligent management method for high-quality video resources according to claim 1, wherein: The method uses an image processing algorithm to extract multidimensional features of video frames, divides the video frames into sub-band channels and calculates local distortion, and applies the Minkowski convergence method to synthesize the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of video frames. Use Sobel operator and Canny to extract the number of edge pixels in the video frame as the edge intensity E t ; Convert the video frame into a grayscale image and construct the gray-level co-occurrence matrix as the texture feature T t ; The optical flow method is used to calculate the pixel difference between two adjacent video frames as the motion change M t ; The edge strength E t , texture feature T t And the motion change M t Synthetic feature vector x t ; For each video frame, Wavelet transform is used to divide the video frame into sub-band channels C t，l , using contrast sensitivity function filtering to weight each channel to simulate the sensitivity of the human eye to video frame images Taking the adjacent video frames as reference, calculate the error signal e of each sub-band channel t，l ; Adjusting perceptual sensitivity B based on local contrast energy t，l , simulating saturation of optic nerve response; Based on adjusting the perceptual sensitivity B t，l The error signal e of each sub-band channel t，l Normalized to "perceptible difference" units as local distortion; Apply Minkowski aggregation to combine the local distortions of all channels into a frame perception complexity score A t ; Synchronously use the structural similarity index to evaluate the structural consistency Q between video frames t ; Score the frame perceptual complexity A t and structural consistency Q t Expand and add the multi-dimensional features of the video frame, and express the expanded multi-dimensional features of the video frame as x′ t .

3. The intelligent management method for high-quality video resources according to claim 2, wherein: The complexity level and membership of each video frame are obtained by FCM clustering complexity classification and the complexity of a single frame is calculated. The resolution of each video frame is calculated according to the complexity of a single frame to form a video resource based on the expanded multi-dimensional feature x′ of the video frame. t Define the fuzzy clustering objective function J m Build an FCM model; Based on the FCM model, the membership degree and cluster center are iteratively updated and optimized until the maximum membership degree after iteration converges and then the iteration is stopped; Output the membership of the video frame, and take the complexity level with the highest membership as the complexity level of the video frame. And perform weighted summation to output the video frame single frame complexity P t ; Calculate the resolution Z of each video frame based on the complexity of the single frame of the video frame t ; The video frames are adjusted according to the calculated video frame resolution, and the adjusted video frames are recomposed into the video assets.

4. The intelligent management method for high-quality video resources according to claim 3, wherein: The multi-dimensional features of the video frames are classified by the multi-layer perceptron and the video categories are analyzed comprehensively, which means that the multi-dimensional features x′ of each video frame in the video resource are classified by the multi-layer perceptron. t Perform average pooling to obtain the overall feature vector F of the video resource; A multilayer perceptron is selected as the classifier, which includes an input layer, a hidden layer, and an output layer. The input layer inputs the overall feature vector F, and the output is the video category. The multilayer perceptron is trained using the cross entropy loss function, and the overall feature vector F is input into the multilayer perceptron to obtain the video category.

5. The intelligent management method for high-quality video resources according to claim 4, wherein: Analyzing user preferences based on user behavior data and pushing videos to users based on video categories refers to extracting the types of videos viewed, viewing time, and viewing frequency from the user's browsing history, and obtaining the priority of each video category of the user by taking a weighted sum of the user's viewing time and viewing frequency of each video category; For each video category, the user's rating data for all videos in that video category is counted, and the average score is calculated as the user's video evaluation; The user's video category priority and user video evaluation are weighted and summed to obtain the user's comprehensive score for the video category. The video categories are sorted in descending order according to the comprehensive score to form a list of recommended video categories, and videos are pushed to users according to the video categories in the list.

6. The intelligent management method for high-quality video resources according to claim 5, wherein: The obtaining of video resources and user behavior data and dividing the video resources into video frames refers to collecting user behavior data by embedding data acquisition software in a video player, synchronously collecting video resources, dividing the video resources into frames to form video frames, and preprocessing the collected user behavior data.

7. The intelligent management method for high-quality video resources according to claim 4 or 5, characterized in that: After the video is classified, a category label is assigned to the video, and the video resources are classified and stored according to the category label. After each video push, a push record is generated and stored synchronously.

8. A high-definition video resource intelligent management system, based on the high-definition video resource intelligent management method according to any one of claims 1 to 7, characterized in that: include, Data acquisition module, used to obtain video resources and user behavior data and perform preprocessing; The feature extraction module is used to extract multidimensional features of video frames using image processing algorithms, divide the video frames into sub-band channels and calculate local distortion. Then, the Minkowski aggregation method is applied to combine the local distortion of all channels into a frame perception complexity score to expand the multidimensional features of the video frame. The video adjustment module is used to obtain the complexity level and membership of each video frame through FCM clustering complexity classification and calculate the complexity of a single frame. The resolution of each video frame is calculated based on the complexity of the single frame to form a video resource. The video push module is used to classify the multi-dimensional features of video frames through a multi-layer perceptron, comprehensively analyze video categories, analyze user preferences based on user behavior data, and push videos to users based on video categories.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the high-definition video resource intelligent management method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the high-definition video resource intelligent management method according to any one of claims 1 to 7 are implemented.

Images