Video acceleration method based on multi-scene dynamic perception intelligent switching

Abstract

The invention discloses a video acceleration method based on multi-scene dynamic perception intelligent switching, and belongs to the technical field of video playing. Comprising the following steps: S1, acquiring scene data in real time, and playing a video in a first acceleration mode according to the scene data; s2, in a set statistical period, under the condition that the video request frequency received by the terminal equipment exceeds an access frequency threshold value, judging that the terminal equipment has a repeated access behavior and calculating a video acceleration score, and under the condition that the video acceleration score is greater than or equal to an acceleration starting threshold value, executing the step S2; and loading the video frame of the terminal equipment into the cache, activating the second acceleration mode, and playing the video through the second acceleration mode. The method has the beneficial effects that by intelligently switching the acceleration mode of the video, the transmission speed and the playing stability of the video can be kept in different scenes, and the phenomena of playing lagging, black screen and the like are effectively avoided.

Description

Technical Field

[0001] The present invention relates to the field of video playback technology, and in particular to a video acceleration method. Background Art

[0002] With the widespread adoption and application of digital multimedia content, video playback has become deeply integrated into people's daily lives, becoming an important form of entertainment and learning, and an effective means for officials to monitor road conditions and ensure public safety. However, the smoothness and viewing quality of video playback are often limited by various factors, notably fluctuations in network bandwidth, differences in terminal device performance, and network congestion.

[0003] Existing technologies struggle to maintain consistent performance across diverse scenarios, resulting in a varying user experience. In practice, users often encounter issues such as video freezes, delays, and image quality degradation. Faced with complex network environments and diverse device conditions, limited processing capabilities for complex scenarios prevent them from effectively addressing challenges such as network fluctuations and device performance differences. Consequently, video playback quality can be severely impacted when the network signal is unstable or device performance is low. Summary of the Invention

[0004] The purpose of the present invention is to provide a video acceleration method based on multi-scene dynamic perception and intelligent switching to solve the above technical problems;

[0005] A video acceleration method based on multi-scene dynamic perception and intelligent switching, comprising:

[0006] Step S1, acquiring scene data in real time, and playing a video in a first acceleration mode according to the scene data;

[0007] Step S2: If the number of video requests received by the terminal device exceeds the access number threshold within the set statistical period, it is determined that the terminal device has repeated access behavior and the video acceleration score is calculated. If the video acceleration score is greater than or equal to the acceleration start threshold, the video frame of the terminal device is loaded into the cache and the second acceleration mode is activated, and the video is played through the second acceleration mode.

[0008] Preferably, step S1 includes,

[0009] Step S11, obtaining hardware information of the video playback platform when starting streaming through monitoring;

[0010] Step S12, selecting a suitable acceleration strategy to play the video according to the monitoring results;

[0011] Step S13, determine whether the current acceleration strategy meets the minimum requirements. If so, optimize the playback efficiency by compressing the video frames, reducing the resolution and reducing the noise. If not, extract the key frames of the video to obtain key information, combine the user model to filter out the key pictures, and output the video stream.

[0012] Preferably, the hardware information in step S11 includes network bandwidth, CPU usage, and memory usage.

[0013] Preferably, the calculation formula for the motion amount of the video frame in step S13 is:

[0014]

[0015] Where M(K) represents the motion amount of the (K)th frame;

[0016] L x (i, j, k) represents the component of the optical flow in the x direction at the pixel point (i, j) of the (K)th frame;

[0017] L y (i, j, k) represents the component of the optical flow in the y direction at the pixel point (i, j) of the (K)th frame;

[0018] The minimum value of M(K) represents the key frame.

[0019] Preferably, the method further includes step S14 of collecting feedback information of video playback and optimizing and adjusting the user model according to the feedback information.

[0020] Preferably, the step of using the second acceleration mode to accelerate the video in step S2 is:

[0021] Step S21, collecting video playback behavior data;

[0022] Step S22, analyzing the behavior data and counting the viewing frequency and duration of each terminal device, and identifying the terminal devices that are frequently called;

[0023] Step S23, identifying a device combination associated with the frequently called terminal device based on an association analysis technology, obtaining a first device combination, and calculating a confidence level of the first device combination;

[0024] Step S24: According to the confidence level, the first device combination that meets the association requirement is added to the resource pool of the second acceleration mode, and the video is played through the second acceleration mode.

[0025] Preferably, the behavior data in step S21 includes device information, viewing time and video content.

[0026] Preferably, the confidence calculation formula in step S23 is,

[0027] Confidence(X→Y)=Support(X∪Y) / Support(X);

[0028] Wherein, Confidence(X→Y) represents the confidence level;

[0029] Support(X∪Y) indicates the frequency of item set X and item set Y appearing simultaneously;

[0030] Support(X) represents the frequency of occurrence of item set X.

[0031] Preferably, the calculation formula for the video acceleration score in step S2 is:

[0032] S=α×A+β×B+γ×C+δ×D;

[0033] Wherein, S represents the video acceleration score;

[0034] A represents the frequency of visits;

[0035] B represents the average length of stay;

[0036] C represents the user preference matching degree;

[0037] D represents the management strategy score;

[0038] α represents the weight of access frequency;

[0039] β represents the weight of the average residence time;

[0040] γ represents the weight of user preference matching;

[0041] δ represents the weight of the management strategy score.

[0042] Preferably, the first acceleration mode in step S1 is a dynamic acceleration mode, and the second acceleration mode in step S2 is a static acceleration mode.

[0043] The beneficial effect of the present invention is that by intelligently switching the acceleration mode of the video, the transmission speed and playback stability of the video can be maintained in different scenarios, effectively avoiding playback freezes, black screen and other phenomena. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a step diagram of the video acceleration method based on multi-scene dynamic perception and intelligent switching of the present invention;

[0045] Figure 2 is a schematic diagram of step S1 of the present invention;

[0046] Figure 3 This is a diagram of the steps of performing video acceleration in the second acceleration mode of the present invention;

[0047] Figure 4 This is a flow chart of the video acceleration method based on multi-scene dynamic perception and intelligent switching of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0050] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0051] A video acceleration method based on multi-scene dynamic perception and intelligent switching, such as Figure 1 、 Figure 4 Shown, including,

[0052] Step S1, acquiring scene data in real time, and playing a video in a first acceleration mode according to the scene data;

[0053] Step S2: If the number of video requests received by the terminal device exceeds the access number threshold within the set statistical period, it is determined that the terminal device has repeated access behavior and the video acceleration score is calculated. If the video acceleration score is greater than or equal to the acceleration start threshold, the video frame of the terminal device is loaded into the cache and the second acceleration mode is activated to play the video through the second acceleration mode.

[0054] Specifically, the present invention provides a video acceleration method based on multi-scene dynamic perception and intelligent switching. By intelligently switching the acceleration mode of the video, it can maintain the transmission speed and playback stability of the video in different scenarios, and effectively avoid playback freezes, black screens and other phenomena.

[0055] In a preferred embodiment, referring to Figure 2 , step S1 includes,

[0056] Step S11, obtaining hardware information of the video playback platform when starting streaming through monitoring;

[0057] Step S12, selecting an appropriate acceleration strategy to play the video based on the monitoring results;

[0058] Step S13, determine whether the current acceleration strategy meets the minimum requirements. If so, optimize the playback efficiency by compressing the video frames, reducing the resolution and reducing the noise. If not, extract the key frames of the video to obtain key information, combine the user model to filter out the key pictures, and output the video stream.

[0059] Specifically, in the first acceleration mode, the system monitors the environment and device status in real time to select the appropriate acceleration strategy. First, the system checks whether the network environment and playback device meet minimum requirements, such as network bandwidth and device CPU usage. If so, the system uses algorithms to intelligently process the video, such as compressing video frames, reducing resolution, and reducing noise, to achieve optimal playback quality and resource efficiency.

[0060] If the requirements are not met, the system will use a visual saliency detection algorithm to extract key frames, ensuring that users can quickly access key information during video playback. Furthermore, the system will filter out videos of interest to users based on the user model, further refining key information to provide users with more personalized video services. The user model is refined based on user actions.

[0061] After all processing and screening is complete, the system outputs an optimized video stream for users to watch. The entire process uses real-time monitoring, intelligent processing, and personalized screening to ensure users can enjoy a smooth and clear video playback experience in any environment.

[0062] Different hardware conditions and network environments have different requirements for video playback. When network bandwidth is sufficient and device performance is good, choose a high-quality acceleration strategy, such as maintaining high resolution and high frame rate playback. When network bandwidth is limited or device performance is insufficient, choose strategies such as reducing video quality and resolution to reduce data transmission and device processing burden, while maintaining normal video playback.

[0063] The first acceleration mode improves video transmission speed and playback stability based on the user's real-time scenario and network environment, effectively avoiding playback freezes and black screens, and achieving a smooth viewing experience. It ensures video playback quality in different network environments, effectively reducing image quality degradation, and providing users with a high-quality visual experience.

[0064] In a preferred embodiment, the hardware information in step S11 includes network bandwidth, central processing unit (CPU) usage, and memory usage.

[0065] Specifically, network bandwidth, CPU usage, and memory usage directly affect video playback performance. Obtaining this information can help you understand the current usage of system resources.

[0066] For example, if the network bandwidth is low, corresponding strategies need to be adopted to avoid video freezes; if the CPU or memory usage is too high, the complexity of video processing needs to be reduced to ensure smooth video playback.

[0067] In a preferred embodiment, the calculation formula for the motion amount of the video frame in step S13 is:

[0068]

[0069] Where M(K) represents the motion amount of the (K)th frame;

[0070] L x (i, j, k) represents the component of the optical flow in the x direction at the pixel point (i, j) of the (K)th frame;

[0071] L y (i, j, k) represents the component of the optical flow in the y direction at the pixel point (i, j) of the (K)th frame;

[0072] The minimum value of M(K) indicates a key frame.

[0073] Specifically, subsequent actions are determined by determining whether the acceleration strategy meets minimum requirements. If so, the video is optimized to reduce data volume while maintaining a good viewing experience. If not, key frames are extracted based on the motion of the video frames, as frames with minimal motion often contain critical information, ensuring that users see important content.

[0074] Under different playback conditions, users can be guaranteed to obtain key video information. When acceleration strategies are feasible, playback efficiency is optimized by compressing video frames, reducing resolution, and reducing noise, thereby conserving system resources. When these strategies are not feasible, key frames are extracted to display key scenes, avoiding the loss of important information due to video playback issues.

[0075] In a preferred embodiment, the method further includes step S14 of collecting feedback information of video playback and optimizing and adjusting the user model based on the feedback information.

[0076] Specifically, data such as user viewing history, search history, feedback information, etc. are collected, features such as viewing time, viewing frequency, and search keywords are extracted, and machine learning algorithms (such as collaborative filtering, deep learning, etc.) are used to train and build a large user model.

[0077] The model is continuously updated and optimized based on user feedback and behavioral data to make it more accurately reflect user interests and needs, improve the accuracy and effectiveness of recommendations, realize personalized video services, improve the efficiency of surveillance video processing and retrieval, and enhance user experience and security effects.

[0078] By collecting this information, such as lag, viewing preferences, etc., the user model is optimized so that the system can better understand user preferences.

[0079] For example, if a user frequently watches a certain type of video and gives good feedback, the system can subsequently recommend this type of video to the user first and optimize the playback strategy for related videos.

[0080] In a preferred embodiment, referring to Figure 3 The step of using the second acceleration mode to accelerate the video in step S2 is:

[0081] Step S21, collecting video playback behavior data;

[0082] Step S22: Analyze the behavior data and count the viewing frequency and duration of each terminal device to identify the terminal devices that are frequently called;

[0083] Step S23: identifying a device combination associated with the frequently called terminal device based on an association analysis technology, obtaining a first device combination, and calculating a confidence level of the first device combination;

[0084] Step S24: adding the first device combination that meets the association requirement to the resource pool of the second acceleration mode according to the confidence level, and playing the video in the second acceleration mode.

[0085] Specifically, in the second acceleration mode, user behavior data is collected and correlated to identify frequently used videos and their associated devices. Video frames for these videos are then pre-cached. The system then directly reads the cached video frames from the local computer for playback. This mode is suitable for scenarios where video resources have been pre-downloaded locally or when users are focusing on a particular device.

[0086] In a preferred embodiment, the behavior data in step S21 includes device information, viewing time, and video content.

[0087] Specifically, behavioral data such as device information, viewing time, and video content can reflect user behavior patterns.

[0088] For example, analyzing device information can reveal a user's frequently used playback devices. Viewing duration and video content can reveal a user's preferences for different videos. This data provides a basis for subsequent, precise optimization of video playback. This helps the system better understand user viewing habits and needs, providing strong support for static acceleration mode.

[0089] In a preferred embodiment, the confidence calculation formula in step S23 is:

[0090] Confidence(X→Y)=Support(X∪Y) / Support(X);

[0091] Among them, Confidence(X→Y) represents confidence;

[0092] Support(X∪Y) indicates the frequency of item set X and item set Y appearing simultaneously;

[0093] Support(X) represents the frequency of occurrence of item set X.

[0094] Specifically, association analysis technology can uncover potential connections between data. By calculating confidence levels, the reliability of associations between device combinations can be measured. A higher confidence level indicates a greater likelihood that a device combination will occur simultaneously. Optimizing the video resources of these combinations can better meet user needs.

[0095] In a preferred embodiment, the calculation formula for the video acceleration score in step S2 is:

[0096] S=α×A+β×B+γ×C+δ×D;

[0097] Among them, S represents the video acceleration score;

[0098] A represents the frequency of visits;

[0099] B represents the average length of stay;

[0100] C represents the user preference matching degree;

[0101] D represents the management strategy score;

[0102] α represents the weight of access frequency;

[0103] β represents the weight of the average residence time;

[0104] γ represents the weight of user preference matching;

[0105] δ represents the weight of the management strategy score.

[0106] Specifically, when playing a video, the system will optimize the playback effect according to the set acceleration mode. The default playback is in the first acceleration mode. The activation conditions of the second acceleration mode are as follows:

[0107] User Conduct:

[0108] 1. The number of video requests from a user to a device exceeds the access threshold N within a preset statistical period (e.g., 7 days);

[0109] 2. The user has repeated access behavior on multiple device combinations.

[0110] S=α×A+β×B+γ×C+δ×D;

[0111] Among them, S represents the video acceleration score;

[0112] A represents the frequency of visits;

[0113] B represents the average length of stay;

[0114] C represents the user preference matching degree;

[0115] D represents the management strategy score;

[0116] α represents the weight of access frequency;

[0117] β represents the weight of the average residence time;

[0118] γ represents the weight of user preference matching;

[0119] δ represents the weight of the management strategy score.

[0120] When S ≥ T (acceleration start threshold) and resource conditions are met, the system automatically loads the video frames of the target device into the cache and activates the static acceleration mode.

[0121] The video resources of the target device have been downloaded to the local or local cache node;

[0122] The video file has completed frame parsing and has fast decoding and positioning index structure;

[0123] At this time, the device will be bound to specific tags (such as "high-frequency access", "key monitoring", etc.); the system default policy enables the static acceleration function.

[0124] In a preferred embodiment, the first acceleration mode in step S1 is a dynamic acceleration mode, and the second acceleration mode in step S2 is a static acceleration mode.

[0125] Specifically, by intelligently switching between two acceleration modes, the video transmission speed and playback stability can be maintained in different scenarios, effectively avoiding playback freezes, black screens and other phenomena, and comprehensively improving the user's video viewing experience.

[0126] Dynamic acceleration mode flexibly adjusts video playback strategies based on real-time hardware information and network conditions to adapt to changing playback conditions. Static acceleration mode, based on user behavior analysis and data mining, pre-caches video frames likely to be viewed, quickly delivering video content when users access it. These two modes work together: dynamic mode responds to real-time changes, while static mode optimizes the playback of frequently used videos, jointly ensuring video playback quality.

[0127] Taking a certain video playback platform as an example, through video surveillance, it can monitor single-route vehicles, flood and typhoon prevention, sudden fires, etc. in real time.

[0128] In order to ensure the quality of video playback, the multi-scene dynamic perception and intelligent switching video acceleration method of this application is adopted to achieve timely response to emergencies, and the video is played in dynamic acceleration mode by default.

[0129] The steps for implementing static acceleration include:

[0130] Step 1: Data collection:

[0131] Use the big model to collect the behavioral data of platform users when watching videos in advance through log records, API calls, etc., including device information, viewing time, video content, etc., and store it in the database.

[0132] Step 2: Filter frequently called devices:

[0133] By analyzing the collected data and counting the viewing frequency and duration of each device, we can identify frequently used devices. These devices all have certain characteristics or behavioral patterns, which are worth further exploring in relation to other devices.

[0134] Step 3: Correlation Analysis:

[0135] Use association analysis technology to identify device combinations associated with frequently called devices, discover association rules between item sets from large-scale data sets, and find device combinations with high co-occurrence frequency.

[0136] Step 4: Calculate the confidence level:

[0137] For the identified association rule X→Y, calculate its confidence.

[0138] The confidence level is calculated using the following formula,

[0139]

[0140] Among them, Support(X∪Y) represents the frequency of item sets X and Y appearing simultaneously;

[0141] Support(X) represents the frequency of occurrence of item set X.

[0142] The higher the confidence level, the higher the correlation between Y and X.

[0143] Step 5: Update the static acceleration resource pool:

[0144] Based on the calculated confidence level, highly correlated device combinations are added to the static acceleration resource pool. These device combinations are considered to have a certain degree of correlation and can serve as the basis for static acceleration strategies to improve the efficiency and quality of video transmission.

[0145] Step 6: Load and play static video:

[0146] When the platform loads the matching video resources and plays them when the user watches the video, since the selected video resources have been optimized, they can achieve faster loading speed and lower lag rate with the support of static acceleration, improving the user's viewing experience.

[0147] The dynamic acceleration steps include,

[0148] Step 1: Environment and equipment inspection:

[0149] Use Prometheus monitoring to obtain real-time hardware information when the platform opens a stream, such as network bandwidth, device CPU usage, memory usage, etc.

[0150] Step 2: Determine whether the environmental information meets the set threshold:

[0151] Based on the monitoring results, the platform will select an appropriate acceleration strategy. If the network environment is good and the device performance is sufficient, the system will select a high-quality acceleration strategy to maintain video clarity and smoothness. If the network environment is poor or the device performance is limited, the system will select an acceleration strategy that reduces video quality or resolution to reduce network bandwidth usage and device load.

[0152] If the current strategy cannot meet the minimum requirements, the system will use a visual saliency detection algorithm to extract key frames in the video, such as vehicle outlines and riverbanks, so that users can quickly obtain important information.

[0153] First, a shot-based method is used to segment the video into multiple shots. In each shot, a motion analysis-based method is used to extract key frames.

[0154] Specifically, the optical flow changes between video frames are calculated. Optical flow is a method that describes the direction and speed of movement of pixels in an image in the time domain. The formula for calculating the amount of motion of a video frame using the optical flow method is as follows:

[0155]

[0156] Where M(K) represents the motion amount of the (K)th frame;

[0157] L x (i, j, k) represents the component of the optical flow in the x direction at the pixel point (i, j) of the (K)th frame;

[0158] L y (i, j, k) represents the component of the optical flow in the y direction at the pixel point (i, j) of the (K)th frame;

[0159] In each shot, the frame with the smallest M(K) is selected as the key frame. This is because, in most cases, the frame with the smallest amount of motion often contains the key information of the shot. This key information is displayed while other parts of the video are blurred to ensure normal video playback. This prevents the missed viewing of sudden events such as fires and car accidents due to video playback failures.

[0160] The platform also collects data such as user viewing history, search history, and feedback. Features are extracted from this data, such as viewing duration, viewing frequency, and search keywords. Machine learning algorithms (such as collaborative filtering and deep learning) are used to train these extracted features and build a large user model.

[0161] Based on the predictions from the user model, relevant videos or events are recommended to users. The user model is continuously updated and optimized based on user feedback and behavioral data. This allows the model to more accurately reflect user interests and needs, improving the accuracy and effectiveness of recommendations.

[0162] By combining keyframe extraction and user large model algorithms, the platform can provide users with a more efficient and intelligent monitoring experience.

[0163] For example, the system can automatically extract key frames from a video and generate a video summary so that users can quickly browse and retrieve video content;

[0164] At the same time, the recommendation function based on the user's big model can recommend relevant surveillance videos or events to users based on their interests and needs, improving the processing and retrieval efficiency of surveillance videos, enhancing user experience and security effectiveness.

[0165] This application dynamically senses scene information and intelligently switches acceleration modes, ensuring smooth video playback without the user noticing. This video playback acceleration method combines static and dynamic acceleration to pre-acquire and cache key frames of the video. Based on the user's real-time scenario and needs, intelligent strategies and algorithms are used to optimize the video transmission process, thereby improving video transmission speed and playback stability.

[0166] Among them, static acceleration counts user behavior data, associates and analyzes the devices that users are interested in, configures video resources in the background in advance, obtains and caches video frames, and enables users to first display the cached key frames when playing videos. It will not cause problems such as black screen, freeze, or spinning due to network or signaling delays.

[0167] Dynamic acceleration includes the following means:

[0168] 1. Scene perception: Through real-time monitoring and analysis of the current environment's network status, device performance and other parameters, the current scene is determined, and the normal playback of the video is ensured by adjusting the video resolution and compressing the video.

[0169] 2. Data transmission optimization: Based on the selected acceleration strategy, the audio-visual media data is optimized to improve transmission efficiency and reduce latency.

[0170] 3. Intelligent distribution and selection algorithm: Based on the user's network environment and device conditions, an intelligent video distribution and selection algorithm is used to select the best transmission path and server node to improve the video transmission speed and stability.

[0171] 4. Video Keyframes: When rigid requirements cannot be met, we extract video keyframes using a visual saliency detection algorithm and a large-scale model to analyze user preferences. These frames are then output as key scenes, the scenes that users are most interested in, to ensure a consistent user experience. We also reduce the keyframe encoding format to reduce transmission efficiency and improve keyframe display speed. Retrieving keyframes consumes server performance, so we prioritize only devices with high user frequency. We prioritize displaying low-frame-rate images. If a stream fails to arrive within a certain period of time, we progressively retrieve higher-frame-rate images.

[0172] Static acceleration technology enables fast video loading by pre-acquiring and caching video frames, allowing users to quickly start watching movies without waiting. Dynamic acceleration technology uses intelligent strategies and algorithms to optimize the video transmission process based on the user's real-time scenario and network environment, improving the video transmission speed and playback stability, effectively avoiding playback freezes, black screens, and other phenomena, and achieving a smooth viewing experience. Through intelligent video transmission optimization and dynamic adjustment strategies, the present invention ensures video playback quality in different network environments, effectively reduces the problem of image quality degradation, and enables users to obtain a high-quality visual experience.

[0173] The above description is only a preferred embodiment of the present invention and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A video acceleration method based on multi-scene dynamic perception and intelligent switching, characterized in that: include, Step S1, acquiring scene data in real time, and playing a video in a first acceleration mode according to the scene data; Step S2: If the number of video requests received by the terminal device exceeds the access number threshold within the set statistical period, it is determined that the terminal device has repeated access behavior and the video acceleration score is calculated. If the video acceleration score is greater than or equal to the acceleration start threshold, the video frame of the terminal device is loaded into the cache and the second acceleration mode is activated, and the video is played through the second acceleration mode.

2. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 1 is characterized in that: Step S1 includes, Step S11, obtaining hardware information of the video playback platform when starting streaming through monitoring; Step S12, selecting a suitable acceleration strategy to play the video according to the monitoring results; Step S13, determine whether the current acceleration strategy meets the minimum requirements. If so, optimize the playback efficiency by compressing the video frames, reducing the resolution and reducing the noise. If not, extract the key frames of the video to obtain key information, combine the user model to filter out the key pictures, and output the video stream.

3. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 2 is characterized in that: The hardware information in step S11 includes network bandwidth, CPU usage, and memory usage.

4. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 2 is characterized in that: The calculation formula for the motion amount of the video frame in step S13 is: Where M(K) represents the motion amount of the (K)th frame; L x (i, j, k) represents the component of the optical flow in the x direction at the pixel point (i, j) of the (K)th frame; L y (i, j, k) represents the component of the optical flow in the y direction at the pixel point (i, j) of the (K)th frame; The minimum value of M(K) represents the key frame.

5. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 2 is characterized in that: The method further includes step S14 of collecting feedback information of video playback and optimizing and adjusting the user model according to the feedback information.

6. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 1 is characterized in that: The step of using the second acceleration mode to accelerate the video in step S2 is: Step S21, collecting video playback behavior data; Step S22, analyzing the behavior data and counting the viewing frequency and duration of each terminal device, and identifying the terminal devices that are frequently called; Step S23: identifying a device combination associated with the frequently called terminal device based on an association analysis technology, obtaining a first device combination, and calculating a confidence level of the first device combination; Step S24: According to the confidence level, the first device combination that meets the association requirement is added to the resource pool of the second acceleration mode, and the video is played through the second acceleration mode.

7. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 6 is characterized in that: The behavior data in step S21 includes device information, viewing time, and video content.

8. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 6 is characterized in that: The confidence calculation formula in step S23 is: Confidence(X→Y)=Support(X∪Y) / Support(X); Wherein, Confidece(X→Y) represents the confidence level; Support(X∪Y) indicates the frequency of item set X and item set Y appearing simultaneously; Support(X) represents the frequency of occurrence of item set X.

9. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 1, characterized in that: The calculation formula for the video acceleration score in step S2 is: S=α×A+β×B+γ×C+δ×D; Wherein, S represents the video acceleration score; A represents the frequency of visits; B represents the average length of stay; C represents the user preference matching degree; D represents the management strategy score; α represents the weight of access frequency; β represents the weight of the average residence time; γ represents the weight of user preference matching; δ represents the weight of the management strategy score.

10. The video acceleration method based on multi-scene dynamic perception and intelligent switching according to claim 1, characterized in that: In step S1, the first acceleration mode is a dynamic acceleration mode, and in step S2, the second acceleration mode is a static acceleration mode.

Images