AI reasoning adaptive video push-pull stream speed adjusting method

Abstract

The invention discloses an AI reasoning adaptive video push-pull stream speed adjusting method, and relates to the technical field of video data management. Comprising the steps that 1, key parameters of video stream transmission are collected regularly, state data of video streams are monitored in real time, the state data comprise the current frame rate, the video cache queue length and the reasoning queue state, and data analysis is carried out according to the key parameters and the state data to predict possible network bottlenecks; 3, dynamically calculating the optimal speed of the push-and-pull flow based on the length of a video cache queue, the state of the reasoning queue and the change of the display time of a single picture, and calculating the optimal speed of the push-and-pull flow according to the current system load, the network condition, the user experience and the predicted network bottleneck. And step 4, obtaining adjusted feedback information, setting a threshold value, and dynamically adjusting the push-pull stream speed so as to maintain synchronization between the video stream and the reasoning task and prevent overload of the reasoning queue.

Description

Technical Field

[0001] The invention discloses an AI inference adaptive video push-pull stream speed adjustment method, which relates to the technical field of video data management. Background Art

[0002] In video processing and transmission technology, with the widespread application of AI reasoning technology, high-concurrency processing in the push-pull process often brings about problems of reasoning delay and result loss. Especially when the AI ​​reasoning module in the system processes the video stream, due to the uncertainty of the reasoning time, the reasoning queue may be overloaded, resulting in the loss of some reasoning results, which in turn affects the overall performance and accuracy of the system. In addition, the existing push-pull method cannot dynamically adjust the push-pull speed according to the changes in the reasoning queue status and the video playback cache, which easily leads to video quality degradation or reasoning result loss in high-concurrency scenarios. Summary of the invention

[0003] In response to the problems of the prior art, the present invention provides an AI reasoning adaptive video push-pull stream speed adjustment method, which can be widely used in scenarios with high requirements on video stream transmission quality, such as coal mine monitoring, industrial quality inspection, and drone inspection, and effectively solves the performance bottleneck problem of video transmission and AI reasoning in a large concurrent environment.

[0004] The specific scheme proposed by the present invention is:

[0005] The present invention provides an AI inference adaptive video push-pull stream speed adjustment method, comprising:

[0006] Step 1: Regularly collect key parameters of video stream transmission and monitor the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length, and inference queue status. Perform data analysis based on key parameters and status data to predict possible network bottlenecks.

[0007] Step 2: Monitor the performance indicators of the AI ​​reasoning module, collect feedback information during the reasoning process in real time, and analyze the reasoning delay and result loss.

[0008] Step 3: Based on the changes in the video cache queue length, inference queue status, and single image display time, dynamically calculate the optimal speed of the push and pull streams. According to the current system load, network conditions, user experience, and predicted network bottlenecks, adjust the push and pull stream speeds based on the optimal speed.

[0009] Step 4: Obtain the adjusted feedback information, set the threshold to dynamically adjust the push and pull stream speed to maintain synchronization between the video stream and the inference task, and prevent the inference queue from being overloaded.

[0010] Furthermore, in step 1 of the AI ​​inference adaptive video push-pull stream speed adjustment method, a PID control algorithm or an adaptive algorithm is selected to analyze and predict the current system load and network conditions.

[0011] Furthermore, the integrated multi-GPU architecture of the AI ​​reasoning adaptive video push-pull stream speed adjustment method uses the parallel computing power of the GPU to accelerate the AI ​​reasoning process in step 2, and uses the parallel computing power of the GPU to calculate the optimal speed of the push-pull stream in step 3, which is used to adjust and accelerate the video push stream.

[0012] Furthermore, the AI ​​reasoning adaptive video push-pull stream speed adjustment method also includes step 5: dynamically adjusting bandwidth: automatically adjusting the bandwidth usage of the video stream according to network conditions, using an adaptive bit rate ABR method, and selecting a higher quality video stream for push when the network conditions are good; in the event of network fluctuations or bandwidth limitations, automatically reducing the video quality to maintain the stability of the reasoning task.

[0013] The present invention also provides an AI inference adaptive video push-pull stream speed adjustment device, comprising a video stream monitoring module, an AI inference capability module, an adaptive adjustment algorithm module and a feedback control module.

[0014] The video stream monitoring module regularly collects key parameters of video stream transmission and monitors the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length and inference queue status. It performs data analysis based on key parameters and status data to predict possible network bottlenecks.

[0015] The AI ​​reasoning capability module monitors the performance indicators of the AI ​​reasoning module, collects feedback information during the reasoning process in real time, and analyzes reasoning delays and result loss.

[0016] The adaptive adjustment algorithm module dynamically calculates the optimal speed of push and pull streams based on the changes in the video cache queue length, inference queue status, and single image display time. It adjusts the push and pull stream speed based on the optimal speed according to the current system load, network conditions, user experience, and predicted network bottlenecks.

[0017] The feedback control module obtains the adjusted feedback information, sets the threshold and dynamically adjusts the push and pull stream speed to maintain the synchronization between the video stream and the inference task, and prevents the inference queue from being overloaded.

[0018] Furthermore, the feedback control module of the AI-inference adaptive video push-pull stream speed adjustment device selects a PID control algorithm or an adaptive algorithm to analyze and predict the current system load and network conditions.

[0019] Furthermore, the AI ​​reasoning adaptive video push-pull stream speed adjustment device is based on an integrated multi-GPU architecture. The AI ​​reasoning capability module uses the parallel computing capability of the GPU to accelerate the AI ​​reasoning process, and the adaptive adjustment algorithm module uses the parallel computing capability of the GPU to calculate the optimal speed of the push-pull stream, which is used to adjust and accelerate the video push stream.

[0020] Furthermore, the AI ​​inference adaptive video push-pull stream speed adjustment device also includes a bandwidth adjustment module, which dynamically adjusts the bandwidth: automatically adjusts the bandwidth usage of the video stream according to network conditions, and adopts an adaptive bit rate ABR method. When the network conditions are good, a higher quality video stream is selected for pushing; in the case of network fluctuations or bandwidth limitations, the video quality is automatically reduced to maintain the stability of the reasoning task.

[0021] The benefits of the present invention are:

[0022] (1) Improving system stability: By real-time monitoring of video stream status, inference queues, and network conditions, and dynamically adjusting the push and pull stream speeds, we can effectively avoid the loss of inference results due to network fluctuations or inference capacity bottlenecks, thereby improving the overall stability and reliability of the system.

[0023] (2) Optimize video quality:

[0024] This technology can automatically adjust the quality of the video stream according to changes in network bandwidth, ensuring high-quality video transmission when bandwidth is sufficient and maintaining a smooth playback experience when bandwidth is insufficient, thereby improving user experience.

[0025] (3) Enhanced AI reasoning capabilities: By combining GPU accelerated reasoning with video processing, the present invention significantly improves the speed and efficiency of AI reasoning, and can complete reasoning tasks in a timely manner under large-scale concurrency, reduce reasoning delays, and ensure the real-time nature of the results.

[0026] (4) Reduce resource waste: The combination of dynamic bandwidth adjustment and adaptive adjustment algorithm effectively avoids the waste of resources in an unstable network environment, ensures the rational allocation of resources in different usage scenarios, and improves the operating efficiency of the system.

[0027] (5) Adapt to a variety of application scenarios: It can be widely used in coal mine monitoring, industrial quality inspection, drone inspection and other scenarios with high requirements for real-time and reliability, meeting the needs of various complex application environments and having good market prospects.

[0028] (6) Real-time feedback and optimization: Through the feedback control mechanism, the system can continuously optimize the push and pull flow speed and reasoning task scheduling, so that it can self-adjust during operation to adapt to changing loads and user needs, thereby enhancing the flexibility of the system.

[0029] (7) Improve user experience: In high-concurrency situations, users can enjoy smooth video playback and timely AI inference results, reducing waiting time and information delays, thereby improving overall user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 It is a schematic diagram of the application framework of the method of the present invention.

[0031] Figure 2 It is a schematic diagram of the application framework of the device of the present invention. DETAILED DESCRIPTION

[0032] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0033] Example 1

[0034] The present invention provides an AI inference adaptive video push-pull stream speed adjustment method, comprising:

[0035] Step 1: Regularly collect key parameters of video stream transmission and monitor the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length, and inference queue status. Perform data analysis based on key parameters and status data to predict possible network bottlenecks. The API interface can be used to obtain network bandwidth and latency information to ensure a comprehensive understanding of the video stream transmission status so that the push and pull stream speed can be adjusted in a timely manner.

[0036] Step 2: Monitor the performance indicators of the AI ​​reasoning module, collect feedback information in the reasoning process in real time, and analyze the reasoning delay and result loss. The multi-GPU architecture can be integrated to use the parallel computing power of the GPU to accelerate the AI ​​reasoning process in step 2, integrate the GPU accelerated computing power, use the parallel processing advantages of the GPU, significantly improve the reasoning efficiency, and reduce the completion time of the reasoning task.

[0037] Step 3: Based on the changes in the video cache queue length, inference queue status, and single image display time, dynamically calculate the optimal speed of push and pull streams, and adjust the push and pull stream speed based on the current system load, network conditions, user experience, and predicted network bottlenecks. The parallel computing power of the GPU is used to calculate the optimal speed of the push and pull streams, and the PID control algorithm or adaptive algorithm is selected to analyze and predict the current system load and network conditions for adjusting and accelerating video push streaming.

[0038] Step 4: Obtain the adjusted feedback information, set the threshold to dynamically adjust the push and pull stream speed to maintain synchronization between the video stream and the inference task, and prevent the inference queue from being overloaded.

[0039] Step 5: Dynamically adjust bandwidth: Automatically adjust the bandwidth usage of the video stream according to network conditions, using the adaptive bit rate ABR method. When the network condition is good, select a higher quality video stream to push; when the network fluctuates or the bandwidth is limited, automatically reduce the video quality to maintain the stability of the reasoning task. This technology ensures that the playback quality of the video stream is not affected when the network environment changes, while effectively reducing data packet loss and delay.

[0040] Example 2

[0041] The present invention also provides an AI inference adaptive video push-pull stream speed adjustment device, comprising a video stream monitoring module, an AI inference capability module, an adaptive adjustment algorithm module and a feedback control module.

[0042] The video stream monitoring module regularly collects key parameters of video stream transmission and monitors the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length and inference queue status. It performs data analysis based on key parameters and status data to predict possible network bottlenecks.

[0043] The AI ​​reasoning capability module monitors the performance indicators of the AI ​​reasoning module, collects feedback information during the reasoning process in real time, and analyzes reasoning delays and result loss.

[0044] The adaptive adjustment algorithm module dynamically calculates the optimal speed of push and pull streams based on the changes in the video cache queue length, inference queue status, and single image display time. It adjusts the push and pull stream speed based on the optimal speed according to the current system load, network conditions, user experience, and predicted network bottlenecks.

[0045] The feedback control module obtains the adjusted feedback information, sets the threshold and dynamically adjusts the push and pull stream speed to maintain the synchronization between the video stream and the inference task, and prevents the inference queue from being overloaded.

[0046] As the information interaction and execution process between the modules in the above-mentioned device are based on the same concept as the embodiment of the method of the present invention, the specific contents can be found in the description of the embodiment of the method of the present invention and will not be repeated here.

[0047] Likewise, the benefits of the device of the present invention are:

[0048] (1) Improving system stability: By real-time monitoring of video stream status, inference queues, and network conditions, and dynamically adjusting the push and pull stream speeds, we can effectively avoid the loss of inference results due to network fluctuations or inference capacity bottlenecks, thereby improving the overall stability and reliability of the system.

[0049] (2) Optimize video quality:

[0050] This technology can automatically adjust the quality of the video stream according to changes in network bandwidth, ensuring high-quality video transmission when bandwidth is sufficient and maintaining a smooth playback experience when bandwidth is insufficient, thereby improving user experience.

[0051] (3) Enhanced AI reasoning capabilities: By combining GPU accelerated reasoning with video processing, the present invention significantly improves the speed and efficiency of AI reasoning, and can complete reasoning tasks in a timely manner under large-scale concurrency, reduce reasoning delays, and ensure the real-time nature of the results.

[0052] (4) Reduce resource waste: The combination of dynamic bandwidth adjustment and adaptive adjustment algorithm effectively avoids the waste of resources in an unstable network environment, ensures the rational allocation of resources in different usage scenarios, and improves the operating efficiency of the system.

[0053] (5) Adapt to a variety of application scenarios: It can be widely used in coal mine monitoring, industrial quality inspection, drone inspection and other scenarios with high requirements for real-time and reliability, meeting the needs of various complex application environments and having good market prospects.

[0054] (6) Real-time feedback and optimization: Through the feedback control mechanism, the system can continuously optimize the push and pull flow speed and reasoning task scheduling, so that it can self-adjust during operation to adapt to changing loads and user needs, thereby enhancing the flexibility of the system.

[0055] (7) Improve user experience: In high-concurrency situations, users can enjoy smooth video playback and timely AI inference results, reducing waiting time and information delays, thereby improving overall user satisfaction.

[0056] It should be noted that not all steps and modules in the above-mentioned processes and device structures are necessary, and some steps or modules can be ignored according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or some components in multiple independent devices may be implemented together.

[0057] The above-described embodiments are only preferred embodiments for fully illustrating the present invention, and the protection scope of the present invention is not limited thereto. Equivalent substitutions or changes made by those skilled in the art based on the present invention are within the protection scope of the present invention. The protection scope of the present invention shall be subject to the claims.

Claims

1. An AI inference adaptive video push-pull stream speed adjustment method, characterized by include: Step 1: Regularly collect key parameters of video stream transmission and monitor the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length, and inference queue status. Perform data analysis based on key parameters and status data to predict possible network bottlenecks. Step 2: Monitor the performance indicators of the AI ​​reasoning module, collect feedback information during the reasoning process in real time, and analyze the reasoning delay and result loss. Step 3: Based on the changes in the video cache queue length, inference queue status, and single image display time, dynamically calculate the optimal speed of the push and pull streams. According to the current system load, network conditions, user experience, and predicted network bottlenecks, adjust the push and pull stream speeds based on the optimal speed. Step 4: Obtain the adjusted feedback information, set the threshold to dynamically adjust the push and pull stream speed to maintain synchronization between the video stream and the inference task, and prevent the inference queue from being overloaded.

2. According to claim 1, the AI ​​inference adaptive video push-pull stream speed adjustment method is characterized by: In step 4, a PID control algorithm or an adaptive algorithm is selected to analyze and predict the current system load and network conditions.

3. According to claim 1, the AI ​​reasoning adaptive video push-pull stream speed adjustment method is characterized by: The multi-GPU architecture is integrated. In step 2, the parallel computing capability of the GPU is used to accelerate the AI ​​reasoning process. In step 3, the parallel computing capability of the GPU is used to calculate the optimal speed of the push and pull streams, which is used to adjust and accelerate the video push stream.

4. According to claim 1, the AI ​​reasoning adaptive video push-pull stream speed adjustment method is characterized by: It also includes step 5: Dynamically adjust bandwidth: Automatically adjust the bandwidth usage of the video stream according to network conditions, using the adaptive bit rate ABR method. When the network conditions are good, a higher quality video stream is selected for streaming; in the case of network fluctuations or bandwidth limitations, the video quality is automatically reduced to maintain the stability of the reasoning task.

5. An AI inference adaptive video push-pull stream speed adjustment device, characterized in that It includes video stream monitoring module, AI reasoning capability module, adaptive adjustment algorithm module and feedback control module. The video stream monitoring module regularly collects key parameters of video stream transmission and monitors the status data of the video stream in real time. The status data includes the current frame rate, video cache queue length and inference queue status. It performs data analysis based on key parameters and status data to predict possible network bottlenecks. The AI ​​reasoning capability module monitors the performance indicators of the AI ​​reasoning module, collects feedback information during the reasoning process in real time, and analyzes reasoning delays and result loss. The adaptive adjustment algorithm module dynamically calculates the optimal speed of push and pull streams based on the changes in the video cache queue length, inference queue status, and single image display time. It adjusts the push and pull stream speed based on the optimal speed according to the current system load, network conditions, user experience, and predicted network bottlenecks. The feedback control module obtains the adjusted feedback information, sets the threshold and dynamically adjusts the push and pull stream speed to maintain the synchronization between the video stream and the inference task, and prevents the inference queue from being overloaded.

6. The AI ​​inference adaptive video push-pull stream speed adjustment device according to claim 5 is characterized in that The feedback control module selects PID control algorithm or adaptive algorithm to analyze and predict the current system load and network conditions.

7. The AI ​​inference adaptive video push-pull stream speed adjustment device according to claim 5 is characterized in that The device is based on an integrated multi-GPU architecture. The AI ​​reasoning capability module uses the parallel computing capability of the GPU to accelerate the AI ​​reasoning process. The adaptive adjustment algorithm module uses the parallel computing capability of the GPU to calculate the optimal speed of the push and pull streams, which is used to adjust and accelerate the video push stream.

8. The AI ​​inference adaptive video push-pull stream speed adjustment device according to claim 5 is characterized in that It also includes a bandwidth adjustment module, which dynamically adjusts the bandwidth: it automatically adjusts the bandwidth usage of the video stream according to network conditions, using the adaptive bit rate ABR method. When the network conditions are good, it selects a higher-quality video stream for streaming; in the case of network fluctuations or bandwidth limitations, it automatically reduces the video quality to maintain the stability of the reasoning task.

Images