Weak network environment video compression method based on deep learning

Abstract

The invention discloses a weak network environment video compression method based on deep learning, and relates to the technical field of video compression, and the method comprises the following steps: 1, carrying out the feature extraction of video content through a deep learning model; 2, dynamically adjusting video compression parameters according to the extracted features and real-time network conditions; step 3, compressing the video by using a self-adaptive coding strategy; according to the video compression method based on deep learning, the video compression efficiency and adaptability in a weak network environment are remarkably improved, the video content is subjected to feature extraction by using the deep learning model, key information in the video can be intelligently identified, such as a motion vector, texture complexity and scene change, and the video compression efficiency and adaptability are improved. Therefore, a more accurate compression strategy is realized, the compression ratio of the video data is optimized, and the smoothness and the stability of video playing are ensured to be kept under variable network conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of video compression, and in particular to a video compression method in a weak network environment based on deep learning. Background Art

[0002] In the current network environment, video streaming services are becoming more and more popular. However, users often watch videos under bandwidth-limited network conditions, such as mobile networks or crowded public Wi-Fi. These so-called "weak network environments" pose a challenge to video transmission because they are usually accompanied by high latency, low bandwidth and unstable connections, resulting in video buffering, freezes and quality degradation.

[0003] Traditional video compression technology often cannot effectively balance compression rate and video quality under these conditions, resulting in poor user experience. Therefore, it is particularly important to develop a technology that can maintain video transmission efficiency and playback quality in a weak network environment.

[0004] Existing video compression technologies mainly rely on fixed encoding parameters and static compression strategies, which cannot adapt to dynamic changes in network conditions. For example, although standards such as H.264 / AVC and H.265 / HEVC have made significant progress in video compression efficiency, they do not take into account changes in real-time network conditions and therefore cannot provide optimal compression effects when the network fluctuates.

[0005] In addition, these traditional methods usually require complex hardware support, which limits their application on resource-constrained devices. To overcome these limitations, a video compression method that can adapt to changing network conditions is needed, which can adjust compression parameters in real time to optimize video quality and transmission efficiency.

[0006] In order to solve the above defects, a technical solution is now provided. Summary of the invention

[0007] The purpose of the present invention is to solve the problems of low video transmission efficiency and poor playback quality in a weak network environment, and to propose a video compression method for a weak network environment based on deep learning.

[0008] The purpose of the present invention can be achieved through the following technical solutions: The video compression method in a weak network environment based on deep learning includes the following steps: Step 1: Use deep learning models to extract features from video content; Step 2: Dynamically adjust video compression parameters based on the extracted features and real-time network conditions; Step 3: Compress the video using an adaptive coding strategy; Step 4: Transmit the compressed video data to the client, and perform real-time decoding and playback on the client; Step 5: Collect playback quality information fed back by the client to further optimize the compression model and parameters.

[0009] Furthermore, the specific process of step one is as follows: Decompose the video into frames, normalize and resize the frames to adapt to the deep learning model, and choose convolutional neural network as the model architecture for video feature extraction; Design a CNN model, including convolutional layers, activation layers, pooling layers, and fully connected layers. Use convolutional layers to extract primary features, and use deep layers of the network to extract advanced features. Build and annotate a dataset containing various video types for model training, select the corresponding loss function and optimization algorithm, and train the model using the annotated dataset; Select and fuse features from the forward propagation model output for video compression, evaluate the feature extraction effect using accuracy and recall metrics, and optimize the model based on the evaluation results.

[0010] Furthermore, the specific operation steps of step 2 are as follows: Monitor bandwidth, latency, and packet loss rate, evaluate network status by sending test packets, and analyze motion features, texture complexity, and scene changes in the video to determine compression strategies; Build a deep learning model to predict compression parameters, including bit rate, quantization parameter, and frame rate, based on network and video features; map network conditions and video features to compression parameters, including bit rate, quantization parameter, and frame rate, through a mapping mechanism; Select an adaptive algorithm to dynamically adjust compression parameters, where the adaptive algorithm includes rules-based, reinforcement learning, or online learning, which adjusts compression parameters in real time and makes fine adjustments based on playback quality feedback; Test compression strategies in a simulated weak network environment and evaluate their impact on video quality and transmission efficiency.

[0011] Furthermore, the mapping mechanism in step 2 includes: Quantify and normalize network conditions including bandwidth, delay and packet loss rate and video features including motion vectors, texture complexity and scene changes; According to the impact of video features on compression effect, weights are assigned to determine their importance in decision making, and corresponding models are selected to achieve mapping. Models include decision trees, neural networks, or support vector machines. The models are trained using labeled data sets to learn the mapping relationship between network conditions and video features to compression parameters. Design algorithms to optimize compression parameters, including bit rate and frame rate adjustment, and integrate the algorithms into the mapping model to ensure the output of optimal compression parameters; Preset specific rules for specific compression parameters under specific conditions, and integrate the preset rules into the mapping model to handle complex decision-making scenarios; Monitor the performance of the mapping mechanism, identify areas for improvement, and adjust model parameters and rules based on monitoring and feedback to optimize compression results.

[0012] Furthermore, the specific operation steps of step three are as follows: Set baseline encoding parameters, based on preliminary analysis of video content and preset network conditions, dynamically adjust bit rate, frame rate, key frame interval and quantization parameters according to video characteristics and real-time network conditions; Split the video into small segments or frames and apply appropriate encoding parameters to each segment; For static video segments or video segments with a change rate less than a preset change rate, a fixed encoding mode is selected to reduce the amount of calculation; for dynamic video segments or video segments with a change rate greater than a preset change rate, a dynamic encoding mode is selected to optimize compression efficiency; Implement bitrate control algorithms, optimize client buffer management, reduce playback freezes, apply intra-frame coding to reduce inter-frame dependencies, and use inter-frame coding technology to improve compression efficiency; Test different encoding parameter combinations to determine the optimal settings.

[0013] Furthermore, the specific operation steps of testing different encoding parameter combinations in step 3 to determine the optimal setting are as follows: The evaluation parameters are obtained through testing and comprehensively analyzed, including: Time required: Measure the time required for the encoding process and the time required for the client to decode the video. Calculate the average encoding time and decoding time through multiple measurements. After normalization, use the encoding time average as the radius to build a bottom circle model and the decoding time average as the height to build a cone model. Calculate the surface area of ​​the cone model and record it as the time evaluation value. This time evaluation value is used as the standard for measuring the time required. Startup delay: measures the delay before the video starts playing; Video quality: Use objective quality evaluation indicators, including peak signal-to-noise ratio and structural similarity, and after normalization, obtain the quality judgment value through weighted summation, and use this quality judgment value as the standard for measuring video quality; Playback smoothness: Tests the smoothness of video playback and quantifies it in percentage, recorded as the smoothness ratio; The obtained time evaluation value, delay time, quality judgment value and fluency ratio are normalized and then input into the following formula: To get the grade judgment value JPZ, where T, Y, Z and L are the time evaluation value, delay time, quality judgment value and smoothness ratio respectively. are preset weight coefficients of the time evaluation value, delay time, quality judgment value and smoothness ratio respectively; and the obtained level judgment value JPZ is used as the priority standard for measuring different coding parameter combinations; The level judgment values ​​of different coding parameter combinations are sorted according to size, and the coding parameter combination with the highest level judgment value is selected as the final choice.

[0014] Furthermore, the specific operation steps of step 4 are as follows: Encapsulate compressed video into FLV, MP4 or streaming formats and embed metadata. Select HLS, DASH or WebRTC protocols and support adaptive bitrate streaming technology. Split the video into small segments to optimize transmission, control the transmission rate to reduce network congestion, set buffers on the client, and implement buffering strategies to cope with network fluctuations; Select a decoder and use hardware acceleration for decoding to ensure audio and video synchronization and optimize playback performance; Monitor network conditions in real time, dynamically adjust video quality, detect transmission or decoding errors, implement retransmission or forward error correction, provide playback control functions, and collect user feedback to optimize the playback experience.

[0015] Furthermore, the specific operation steps of step 5 are as follows: Determine key indicators for evaluating playback quality, including video clarity, playback smoothness, buffering time, and startup delay, and obtain users' subjective satisfaction through questionnaires or direct feedback; Integrate monitoring tools in client applications to collect playback quality data in real time, set data collection points, and ensure data is collected during playback; Through the automatic data reporting mechanism, data is automatically sent and users are provided with the option of manual feedback, allowing users to proactively report when they encounter problems; Implement real-time data analysis to identify and respond to playback quality issues, and periodically perform batch data analysis to identify long-term trends and potential issues; Integrate client feedback into the optimization process of video compression and transmission, adjust the parameters of the deep learning model based on the feedback, and optimize the video compression strategy; Use machine learning algorithms to analyze data, predict and optimize compression parameters, and implement a rule engine to automatically adjust compression and transmission strategies based on analysis results; Design A / B tests to compare the playback quality under different compression parameter settings, evaluate the test results, and select the optimal compression parameter configuration; Continuously monitor playback quality, make adjustments and optimizations based on new data, and regularly update client applications and compression models to integrate the latest optimization results.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention significantly improves the video compression efficiency and adaptability in a weak network environment by adopting a deep learning-based video compression method. It uses a deep learning model to extract features from video content and can intelligently identify key information in the video, such as motion vectors, texture complexity, and scene changes, thereby achieving a more accurate compression strategy. It not only optimizes the compression rate of video data, but also ensures the smoothness and stability of video playback under variable network conditions. The present invention effectively optimizes the utilization of network resources by dynamically adjusting compression parameters to adapt to real-time network conditions. The deep learning model predicts and adjusts parameters such as bit rate and frame rate to ensure that the transmission efficiency of the video stream matches the network bandwidth, reducing buffering and jamming. The adaptive encoding strategy further ensures that high-quality playback experience can be provided in different network environments, thereby enhancing user satisfaction. The present invention has a mechanism for collecting playback quality feedback from the client, and the feedback is used to further optimize the compression model and parameters. Through real-time data analysis and user feedback, the system can continuously learn and adjust to adapt to changing viewing habits and network conditions. The intelligent processing method not only improves the flexibility and accuracy of video compression technology, but also ensures the continuous improvement of video service quality in the long term. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings; Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] It should be understood that the terms "include" and "comprising" used in the specification and claims of the present disclosure indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or collections thereof.

[0020] It should also be understood that the terms used in this disclosure are only for the purpose of describing specific embodiments and are not intended to limit the disclosure. As used in this disclosure and claims, the singular forms of "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise. It should also be further understood that the term "and / or" used in this disclosure and claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0021] like Figure 1 As shown, the video compression method in a weak network environment based on deep learning includes the following steps: Step 1: Use deep learning models to extract features from video content; Decompose the video into individual frames, which will be used as input to the deep learning model. Normalize the pixel values ​​of the video frames so that they usually range between 0 and 1 to facilitate model processing. Adjust the video frames to the fixed size required for model input. Select a deep learning model architecture suitable for video feature extraction, such as convolutional neural network (CNN). Build a model that includes multiple convolutional layers, activation layers (such as ReLU), pooling layers (such as max pooling), and fully connected layers. Use convolutional layers to extract primary features in the video, such as edges and textures, and use deep network structures to extract high-level features in the video, such as scenes and objects. Build a dataset containing multiple types of videos for training the model, and annotate the dataset to provide the labels required for training the model; select a suitable loss function, such as cross entropy loss or mean square error, based on the goal of feature extraction, and select an optimization algorithm, such as Adam or SGD, for training the model. Train the model using the training dataset and adjust the model parameters through the back-propagation algorithm; input the preprocessed video frames into the trained model, perform forward propagation, and obtain outputs from specific layers of the model (usually the middle layers), which represent the characteristics of the video content; Select the most useful features from the features extracted by the model for subsequent compression steps, and fuse multiple features to obtain a more comprehensive video representation; use indicators such as accuracy and recall rate to evaluate the effect of feature extraction, and tune the model based on the evaluation results to improve the accuracy and efficiency of feature extraction.

[0022] Step 2: Dynamically adjust video compression parameters based on the extracted features and real-time network conditions; First, monitor the available bandwidth of the current network, evaluate the network's download and upload speeds by sending test packets, measure the network's latency, including the round-trip time (RTT) and the transmission time of the packet, and monitor the packet loss in the network, which may affect the stability of the video stream; analyze the motion features in the video, such as the size and direction of the motion vector, to determine the dynamic degree of the scene, evaluate the texture complexity in the video frame, which may affect the compression efficiency and the required bit rate, and detect scene changes in the video, such as switching and transitions, which require different compression strategies; build a deep learning model that can predict the optimal compression parameters based on network conditions and video features, and design a mapping mechanism to map network conditions and video features to compression parameters, including bit rate, quantization parameter and frame rate; the mapping mechanism includes: Standardize network conditions and video features. Specifically, convert the real-time monitored network bandwidth, delay, and packet loss rate data into a unified metric for easy processing. Quantify and standardize the extracted video features (such as motion vectors, texture complexity, and scene changes). Evaluate the impact of different video features on the compression effect, determine their importance, and assign weights to each feature. These weights reflect the influence of the feature in the compression parameter decision. Select a suitable model to achieve the mapping, such as a decision tree, neural network, or support vector machine. Use a labeled data set (including network conditions, video features, and optimal compression parameters) to train the model. Design algorithms to optimize compression parameters, such as bit rate control algorithms, frame rate adjustment strategies, etc., and integrate the algorithms into the mapping model to ensure that the optimal compression parameters can be output based on the input characteristics and conditions; for complex decision-making scenarios, preset specific rules for specific compression parameters under specific conditions, and integrate the preset rules into the mapping model; monitor the performance of the mapping mechanism, identify possible improvement points, and adjust the model parameters and mapping rules based on the monitoring results and feedback information to improve the compression effect.

[0023] Select adaptive algorithms, such as rule-based algorithms, reinforcement learning, or online learning algorithms, to dynamically adjust compression parameters, and through optimization mechanisms, ensure video quality while minimizing the required network resources; dynamically adjust compression parameters based on real-time monitored network conditions and extracted video features, and use feedback mechanisms to further fine-tune compression parameters based on real-time feedback on video playback quality; test the effectiveness of compression parameter adjustment strategies in simulated weak network environments, and evaluate the impact of adjusted compression parameters on video quality and transmission efficiency.

[0024] Step 3: Compress the video using an adaptive coding strategy; Design an adaptive encoding strategy to dynamically adjust encoding settings based on video content characteristics and network conditions, and determine adjustable parameters during the encoding process, including bit rate, frame rate, key frame interval, and quantization parameters; set a set of baseline parameters for the encoding process, which can be set based on initial analysis of the video and general network conditions, and dynamically adjust the baseline parameters based on the compression parameters adjusted in step 2; The video content is divided into smaller segments or frames, and different encoding parameters are applied to each part. The video features extracted in step 1 are applied to each video segment to guide the adjustment of the encoding parameters. For a static video segment or a video segment with a change rate less than a preset change rate, a fixed encoding mode is selected to reduce the amount of calculation. For a dynamic video segment or a video segment with a change rate greater than a preset change rate, a dynamic encoding mode is selected to optimize the compression efficiency. Implement bitrate control algorithms to ensure that the bitrate of the video stream matches the network conditions, manage the client buffer to reduce playback lag and delay; perform intra-frame encoding on video frames to reduce the dependency on previous and next frames, and use inter-frame coding techniques such as motion estimation and compensation to improve compression efficiency; Different coding parameter combinations are tested to find the optimal coding parameter combination, and evaluation parameters are obtained through testing. The evaluation parameters are then comprehensively analyzed, where the evaluation parameters include: Required time: Measure the time required for the encoding process and the time required for the client to decode the video. Calculate the mean encoding time and decoding time through multiple measurements. After normalization, use the mean encoding time as the radius to build a bottom circle model, and use the mean decoding time as the height to build a cone model. Calculate the surface area of ​​the cone model and record it as the time evaluation value, and use this time evaluation value as the standard for measuring the required time; Startup delay: Measure the delay time before the video starts playing; Video quality: Use objective quality evaluation indicators, including peak signal-to-noise ratio (PSNR) and structural similarity (SSIM). After normalization, obtain the quality judgment value by weighted summation, and use this quality judgment value as the standard for measuring video quality; Playback fluency: Test the fluency of video playback and quantify it as a percentage, which is recorded as the fluency ratio; The obtained time evaluation value, delay time, quality judgment value and fluency ratio are normalized and then input into the following formula: To get the grade judgment value JPZ, where T, Y, Z and L are the time evaluation value, delay time, quality judgment value and smoothness ratio respectively. The preset weight coefficients are respectively the time evaluation value, delay time, quality judgment value and smoothness ratio; and the obtained level judgment value JPZ is used as the priority standard for measuring different coding parameter combinations; the level judgment values ​​of different coding parameter combinations are sorted according to size, and the coding parameter combination with the highest level judgment value is selected as the final choice; Evaluate the impact of different encoding parameters on video quality and file size; evaluate the quality of the encoded video in real time to ensure that it meets the predetermined quality standards, and based on the quality evaluation results, provide feedback to adjust the encoding parameters to optimize the encoding of subsequent video segments; encapsulate the encoded video data into a format suitable for network transmission, and implement error recovery mechanisms such as forward error correction (FEC) to improve transmission robustness in weak network environments.

[0025] Step 4: Transmit the compressed video data to the client, and perform real-time decoding and playback on the client; Convert the compressed video data into a format suitable for network transmission, such as encapsulating the encoded data into FLV, MP4 or other streaming formats, and embedding necessary metadata in the video data, such as encoding information, timestamps, key frame positions, etc., to facilitate client processing; select a suitable network transmission protocol, such as HTTP Live Streaming (HLS), DynamicAdaptive Streaming over HTTP (DASH) or WebRTC to support the transmission of video streams, and implement adaptive bitrate streaming technology to dynamically adjust the video quality according to the client's network conditions; Split video data into small blocks or fragments to facilitate parallel transmission and reduce buffering time, control data transmission rate, ensure stable transmission of video data, and reduce network congestion and packet loss; set up a buffer on the client to store temporary video data to cope with network fluctuations and delays, and implement intelligent buffering strategies such as preloading, buffer size adjustment, and data refresh to optimize the playback experience; select a suitable decoder based on the encoding format of the video data, and use the client's hardware acceleration capabilities (such as GPU) for video decoding to improve decoding efficiency and reduce energy consumption; Synchronize video playback with audio to ensure synchronous playback of audio and video, optimize playback performance, such as frame rate control, image rendering optimization, etc., to provide a smooth viewing experience; monitor the client's network conditions in real time, such as bandwidth, latency, and packet loss rate, and dynamically adjust the video quality according to the monitoring results, such as switching to a lower or higher bit rate stream; detect errors that may occur during transmission or decoding, and implement error recovery mechanisms, such as retransmitting lost data packets and using forward error correction (FEC); provide user control functions such as play, pause, fast forward, and rewind, allowing users to feedback playback quality, such as freezes, blurs, etc., in order to further optimize services.

[0026] Step 5: Collect playback quality information fed back by the client to further optimize the compression model and parameters.

[0027] Determine key indicators for evaluating playback quality, such as video clarity, playback smoothness, buffering time, startup delay, etc., and obtain users' subjective satisfaction through questionnaires or direct feedback; integrate monitoring tools in client applications to collect playback quality data in real time, set data collection points, and ensure data collection in key playback links (such as buffering, freezes, and quality switching); Through the automatic data reporting mechanism, the client automatically sends data when encountering specific events (such as buffering completion, playback freezes), and provides users with the option of manual feedback, allowing users to actively report when encountering problems; implement real-time data analysis to quickly identify and respond to playback quality issues, and conduct batch data analysis regularly to identify long-term trends and potential problems; integrate client feedback into the optimization process of video compression and transmission, adjust the parameters of the deep learning model based on the feedback, and optimize the video compression strategy; Use machine learning algorithms to analyze data, predict and optimize compression parameters, implement a rule engine, and automatically adjust compression and transmission strategies based on analysis results; design A / B tests, compare playback quality under different compression parameter settings, evaluate test results, and select the optimal compression parameter configuration; continuously monitor playback quality, and continuously adjust and optimize based on new data, and regularly update client applications and compression models to integrate the latest optimization results.

[0028] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A video compression method for a weak network environment based on deep learning, characterized in that: The following steps are involved: Step 1: Use deep learning models to extract features from video content; Step 2: Dynamically adjust video compression parameters based on the extracted features and real-time network conditions; Step 3: Compress the video using an adaptive coding strategy; Step 4: Transmit the compressed video data to the client, and perform real-time decoding and playback on the client; Step 5: Collect playback quality information fed back by the client to further optimize the compression model and parameters.

2. The method for video compression in a weak network environment based on deep learning according to claim 1, characterized in that: The specific process of step one is as follows: Decompose the video into frames, normalize and resize the frames to adapt to the deep learning model, and choose convolutional neural network as the model architecture for video feature extraction; Design a CNN model, including convolutional layers, activation layers, pooling layers, and fully connected layers. Use convolutional layers to extract primary features, and use deep layers of the network to extract advanced features. Build and annotate a dataset containing various video types for model training, select the corresponding loss function and optimization algorithm, and train the model using the annotated dataset; Select and fuse features from the forward propagation model output for video compression, evaluate the feature extraction effect using accuracy and recall metrics, and optimize the model based on the evaluation results.

3. The method for video compression in a weak network environment based on deep learning according to claim 1, characterized in that: The specific operation steps of step 2 are as follows: Monitor bandwidth, latency, and packet loss rate, evaluate network status by sending test packets, and analyze motion features, texture complexity, and scene changes in the video to determine compression strategies; Build a deep learning model to predict compression parameters, including bit rate, quantization parameter, and frame rate, based on network and video features; map network conditions and video features to compression parameters, including bit rate, quantization parameter, and frame rate, through a mapping mechanism; Select an adaptive algorithm to dynamically adjust compression parameters, where the adaptive algorithm includes rules-based, reinforcement learning, or online learning, which adjusts compression parameters in real time and makes fine adjustments based on playback quality feedback; Test compression strategies in a simulated weak network environment and evaluate their impact on video quality and transmission efficiency.

4. The method for video compression in a weak network environment based on deep learning according to claim 3, characterized in that: The mapping mechanism in step 2 includes: Quantify and normalize network conditions including bandwidth, delay and packet loss rate and video features including motion vectors, texture complexity and scene changes; According to the impact of video features on compression effect, weights are assigned to determine their importance in decision making, and corresponding models are selected to achieve mapping. Models include decision trees, neural networks, or support vector machines. The models are trained using labeled data sets to learn the mapping relationship between network conditions and video features to compression parameters. Design algorithms to optimize compression parameters, including bit rate and frame rate adjustment, and integrate the algorithms into the mapping model to ensure the output of optimal compression parameters; Preset specific rules for specific compression parameters under specific conditions, and integrate the preset rules into the mapping model to handle complex decision-making scenarios; Monitor the performance of the mapping mechanism, identify areas for improvement, and adjust model parameters and rules based on monitoring and feedback to optimize compression results.

5. The method for video compression in a weak network environment based on deep learning according to claim 1, characterized in that: The specific operation steps of step three are as follows: Set baseline encoding parameters, based on preliminary analysis of video content and preset network conditions, dynamically adjust bit rate, frame rate, key frame interval and quantization parameters according to video characteristics and real-time network conditions; Split the video into small segments or frames and apply appropriate encoding parameters to each segment; For static video segments or video segments with a change rate less than a preset change rate, a fixed encoding mode is selected to reduce the amount of calculation; for dynamic video segments or video segments with a change rate greater than a preset change rate, a dynamic encoding mode is selected to optimize compression efficiency; Implement bitrate control algorithms, optimize client buffer management, reduce playback freezes, apply intra-frame coding to reduce inter-frame dependencies, and use inter-frame coding technology to improve compression efficiency; Test different encoding parameter combinations to determine the optimal settings.

6. The method for video compression in a weak network environment based on deep learning according to claim 5, characterized in that: The specific steps for testing different encoding parameter combinations and determining the optimal settings in step 3 are as follows: The evaluation parameters are obtained through testing and comprehensively analyzed, including: Time required: Measure the time required for the encoding process and the time required for the client to decode the video. Calculate the average encoding time and decoding time through multiple measurements. After normalization, use the encoding time average as the radius to build a bottom circle model and the decoding time average as the height to build a cone model. Calculate the surface area of ​​the cone model and record it as the time evaluation value. This time evaluation value is used as the standard for measuring the time required. Startup delay: measures the delay before the video starts playing; Video quality: Use objective quality evaluation indicators, including peak signal-to-noise ratio and structural similarity, and after normalization, obtain the quality judgment value through weighted summation, and use this quality judgment value as the standard for measuring video quality; Playback smoothness: Tests the smoothness of video playback and quantifies it in percentage, recorded as the smoothness ratio; The obtained time evaluation value, delay time, quality judgment value and fluency ratio are normalized and then input into the following formula: To get the grade judgment value JPZ, where T, Y, Z and L are the time evaluation value, delay time, quality judgment value and smoothness ratio respectively. are preset weight coefficients of the time evaluation value, delay time, quality judgment value and smoothness ratio respectively; and the obtained level judgment value JPZ is used as the priority standard for measuring different coding parameter combinations; The level judgment values ​​of different coding parameter combinations are sorted according to size, and the coding parameter combination with the highest level judgment value is selected as the final choice.

7. The method for video compression in a weak network environment based on deep learning according to claim 1, characterized in that: The specific operation steps of step 4 are as follows: Encapsulate compressed video into FLV, MP4 or streaming formats and embed metadata. Select HLS, DASH or WebRTC protocols and support adaptive bitrate streaming technology. Split the video into small segments to optimize transmission, control the transmission rate to reduce network congestion, set buffers on the client, and implement buffering strategies to cope with network fluctuations; Select a decoder and use hardware acceleration for decoding to ensure audio and video synchronization and optimize playback performance; Monitor network conditions in real time, dynamically adjust video quality, detect transmission or decoding errors, implement retransmission or forward error correction, provide playback control functions, and collect user feedback to optimize the playback experience.

8. The method for video compression in a weak network environment based on deep learning according to claim 1, characterized in that: The specific operation steps of step 5 are as follows: Determine key indicators for evaluating playback quality, including video clarity, playback smoothness, buffering time, and startup delay, and obtain users' subjective satisfaction through questionnaires or direct feedback; Integrate monitoring tools in client applications to collect playback quality data in real time, set data collection points, and ensure data is collected during playback; Through the automatic data reporting mechanism, data is automatically sent and users are provided with the option of manual feedback, allowing users to proactively report when they encounter problems; Implement real-time data analysis to identify and respond to playback quality issues, and periodically perform batch data analysis to identify long-term trends and potential issues; Integrate client feedback into the optimization process of video compression and transmission, adjust the parameters of the deep learning model based on the feedback, and optimize the video compression strategy; Use machine learning algorithms to analyze data, predict and optimize compression parameters, and implement a rule engine to automatically adjust compression and transmission strategies based on analysis results; Design A / B tests to compare the playback quality under different compression parameter settings, evaluate the test results, and select the optimal compression parameter configuration; Continuously monitor playback quality, make adjustments and optimizations based on new data, and regularly update client applications and compression models to integrate the latest optimization results.

Images