Intelligent network control methods and related equipment for low-latency video backhaul

Abstract

This invention relates to the field of multimedia communication and network control, specifically to an intelligent network control method and related equipment for low-latency video backhaul. The intelligent network control method includes the following steps: real-time acquisition of key network indicators of the network link, including bandwidth, latency, jitter, and packet loss rate, to provide real-time network environment data support for transmission strategy adjustments. This invention effectively solves the technical challenge of low-latency video backhaul in complex network environments by constructing an intelligent control mechanism that combines network state awareness and video content feature recognition. Specifically, by real-time acquisition of key network link indicators and the introduction of a lightweight CNN model to analyze video content activity, a dual perception capability of network environment and content features is formed, providing data support for dynamically adjusting encoding parameters and achieving a precise balance between video quality and network adaptability.

Description

Technical Field

[0001] This invention relates to the field of multimedia communication and network control, specifically to an intelligent network control method and related equipment for low-latency video backhaul. Background Technology

[0002] With the widespread adoption of high-definition video capture equipment and the increasing demand for remote control of unmanned vehicles, drones, and robots, video transmission in complex mobile network environments faces multiple challenges. Traditional video backhaul solutions rely on fixed bandwidth allocation strategies, making it difficult to dynamically adapt to network congestion, bandwidth fluctuations, and resource contention issues arising from concurrent transmission of multiple video streams. This results in excessively high end-to-end latency and insufficient video smoothness, severely impacting the real-time performance of remote control. Existing technologies, such as mainstream transmission protocols like RTMP and RTSP, lack deep awareness of network conditions and cannot dynamically adjust encoding parameters and transmission priorities based on video content characteristics, exhibiting a technological gap in the collaborative optimization of source coding and channel transmission. Furthermore, facing different network failure modes such as random packet loss and continuous packet loss, traditional packet loss mitigation mechanisms struggle to achieve an intelligent balance between bandwidth overhead and recovery capability, further exacerbating video transmission stability issues. Therefore, the technical problem proposed in this invention is: how to construct an end-to-end optimization mechanism that integrates network state awareness, video content feature recognition, and intelligent resource scheduling in complex and dynamic network environments to achieve low-latency, high-reliability transmission of multiple video streams, while also considering encoding efficiency and network adaptability. Summary of the Invention

[0003] This disclosure proposes an intelligent network control method and related equipment for low-latency video backhaul, aiming to overcome at least one of the defects existing in the prior art.

[0004] To achieve the above objectives, the technical solution disclosed in this invention is as follows:

[0005] According to one aspect of this disclosure, a smart network control method for low-latency video backhaul is provided, comprising the following steps:

[0006] Real-time acquisition of key network indicators, including bandwidth, latency, jitter, and packet loss rate, provides real-time network environment data support for adjusting transmission strategies;

[0007] By extracting motion vectors between video frames using a lightweight CNN model, the activity level of video content can be determined, and the frequency and magnitude of content changes between adjacent frames can be evaluated, providing content feature basis for encoding strategy optimization.

[0008] Based on the aforementioned key network indicators and video content activity, the video encoding parameters, including resolution, bitrate, and frame rate, are dynamically adjusted to balance video quality and network adaptability.

[0009] Based on the priority and resource usage of multiple video streams, network bandwidth resources are allocated in a coordinated manner, and the transmission parameters of each video stream are controlled to ensure the efficiency of multi-video transmission.

[0010] Based on the real-time network environment, the system intelligently identifies packet loss types, dynamically adjusts FEC redundancy parameters, and adds redundant data to network transmission to combat packet loss problems, thereby balancing packet loss recovery capability and bandwidth overhead.

[0011] Furthermore, the steps for collecting the key network indicators include:

[0012] The uplink available bandwidth is estimated by sending probe packets and analyzing response time, and the bandwidth change trend is tracked in real time to identify network congestion and the peak bandwidth per unit time is statistically analyzed.

[0013] Measure the round-trip time and one-way transmission delay of data packets, and analyze the stability of delay jitter to determine the size of the decoding buffer;

[0014] By statistically analyzing the instantaneous packet loss rate within a short time window and the average packet loss rate over a long time period, patterns of continuous or random packet loss can be identified, providing a basis for adjusting the FEC redundancy parameters.

[0015] Furthermore, the step of extracting the motion vectors between video frames includes:

[0016] Reduce the computational complexity and resource consumption of lightweight CNN models for real-time video analysis scenarios;

[0017] Based on the pixel differences and motion trajectories of adjacent video frames, motion vector features representing content changes are extracted, the dynamic level of video content is quantified, and a video content activity evaluation result is generated to drive the dynamic adjustment of the encoding parameters.

[0018] Furthermore, the steps for dynamically adjusting video encoding parameters include:

[0019] When sufficient network bandwidth is detected, high bitrate and high resolution encoding are used to improve video detail quality; when network bandwidth is limited, the bitrate and resolution are automatically reduced to ensure transmission continuity.

[0020] For high dynamic range video content, increase frame rate and resolution to preserve motion details; for static or low dynamic range content, reduce frame rate and bit rate to reduce data volume.

[0021] When network congestion or bandwidth drops sharply, a fast bitrate adjustment mechanism is triggered to prioritize the transmission of key I-frames to ensure continuous display at the video decoding end.

[0022] Furthermore, the steps for coordinating and allocating network bandwidth resources include:

[0023] Establish a video stream priority evaluation model to assign priorities based on the business type and content importance of the video stream;

[0024] Based on real-time network bandwidth and the priority of each video stream, scheduling instructions are sent to the network transport layer through a standard interface to dynamically adjust the transmission bandwidth quota and transmission priority of each video stream, thereby avoiding resource contention among multiple video streams.

[0025] Furthermore, the steps for dynamically adjusting the FEC redundancy parameters include:

[0026] When an increase in network packet loss rate is detected and a continuous packet loss pattern is observed, the FEC redundancy is increased to improve packet loss recovery capability. When the network packet loss rate is low and the packet loss is random, the FEC redundancy is reduced to reduce bandwidth overhead.

[0027] Based on the latency and jitter characteristics of the network link, the block size and redundancy ratio of FEC encoding are dynamically adjusted to balance packet loss resistance and encoding processing latency, so as to ensure real-time video transmission.

[0028] Furthermore, the network control method achieves low-latency transmission through joint optimization of the source and channel: on the source side, an adaptive coding control module reduces redundant data, and on the channel side, an FEC anti-packet loss module and an intelligent scheduling decision module improve transmission reliability and bandwidth utilization, forming end-to-end dynamic optimization by combining feedback strategies.

[0029] According to another aspect of this disclosure, a low-latency video backhaul intelligent network control system is provided for implementing the low-latency video backhaul intelligent network control method described above, comprising:

[0030] The network status detection module is used to collect key network indicators in real time to provide data support for adjusting transmission strategies. The key network indicators include network link bandwidth, latency, jitter, and packet loss rate.

[0031] The video content analysis module is used to extract inter-frame motion vectors from video using a lightweight CNN model, determine the activity level of video content, and output content feature evaluation results.

[0032] An adaptive coding control module is used to receive the evaluation results of the network key indicators and content features, dynamically adjust the video coding parameters, and achieve a balance between video quality and network adaptability.

[0033] The intelligent scheduling decision module is used to allocate network bandwidth resources and control the transmission parameters of each video stream based on the priority and resource usage of multiple video streams.

[0034] The feedback adjustment mechanism module is used to periodically collect module operation data, evaluate system performance, and update control strategies.

[0035] The FEC packet loss prevention module is used to intelligently identify packet loss types based on real-time network conditions, dynamically adjust FEC redundancy parameters, and add redundant data to network transmission to combat packet loss problems.

[0036] According to another aspect of this disclosure, a video conferencing device is provided, comprising:

[0037] Encoder devices, deployed at the video acquisition end, are used to perform image acquisition and preprocessing;

[0038] Decoder devices, deployed at the video receiver, are used to perform network data decoding and image display;

[0039] The intelligent network control system for low-latency video backhaul described above is used to connect the encoder device and the decoder device to achieve low-latency backhaul control of the video stream.

[0040] According to another aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent network control method for low-latency video backhaul as described above.

[0041] The beneficial effects of this invention are:

[0042] This invention effectively solves the technical challenge of low-latency video backhaul in complex network environments by constructing an intelligent control mechanism that combines network state awareness with video content feature recognition. Specifically, by collecting key indicators of the network link in real time and introducing a lightweight CNN model to analyze video content activity, a dual perception capability of network environment and content features is formed, providing data support for dynamically adjusting encoding parameters and achieving a precise balance between video quality and network adaptability.

[0043] Furthermore, the intelligent scheduling decision module allocates bandwidth resources for multiple video streams based on a priority evaluation model, avoiding performance degradation caused by resource contention and improving the overall efficiency of multi-channel transmission. The FEC (Fulfilled Error Correction) module dynamically adjusts redundancy strategies according to real-time packet loss types, optimizing bandwidth utilization while ensuring transmission reliability. The feedback adjustment mechanism forms a closed-loop control through periodic strategy updates, continuously adapting to changes in the network environment. The combined optimization of adaptive coding control on the source side and intelligent scheduling and packet loss correction technologies on the channel side significantly reduces end-to-end latency and improves video smoothness, maintaining stable transmission even under network fluctuation scenarios.

[0044] Furthermore, this invention, through a multi-module collaborative intelligent control system, breaks through the limitations of traditional fixed strategies, providing a comprehensive solution for video backhaul in remote control scenarios that combines real-time performance, reliability, and resource efficiency, effectively meeting the stringent requirements of remote control of unmanned equipment for low-latency video transmission.

[0045] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description

[0046] Figure 1 This is a flowchart of a low-latency video backhaul control method in one embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram of a low-latency video backhaul control method in one embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram of the spatiotemporal evolution of key network link indicators in one embodiment of the present invention;

[0049] Figure 4 This is a schematic diagram of the inter-frame motion vector field in one embodiment of the present invention;

[0050] Figure 5 This is a schematic diagram of an adaptive decision surface for encoding parameters in one embodiment of the present invention;

[0051] Figure 6 This is a schematic diagram illustrating the performance analysis of the FEC redundancy strategy in one embodiment of the present invention;

[0052] Figure 7 This is a schematic diagram of intelligent bandwidth scheduling for multiple video streams in one embodiment of the present invention;

[0053] Figure 8 This is an architecture diagram of a low-latency video backhaul control system according to an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0055] The present invention provides the following preferred embodiments:

[0056] Example 1: To address the problem of insufficient adaptation between dynamic changes in the network environment and the characteristics of video content in low-latency video backhaul, this example provides an intelligent network control method for low-latency video backhaul. Through coordinated perception and dynamic adjustment of network status and video content, it optimizes transmission efficiency and quality. Figure 1 and Figure 2 As shown, the control method flow is as follows:

[0057] S100: Real-time acquisition of key network indicators, including bandwidth, latency, jitter, and packet loss rate, to provide real-time network environment data support for transmission strategy adjustments.

[0058] S200: Extracts motion vectors between video frames using a lightweight CNN model, determines the activity level of video content, and evaluates the frequency and magnitude of content changes between adjacent frames, providing content feature basis for encoding strategy optimization.

[0059] S300: Based on key network indicators and video content activity, it dynamically adjusts video encoding parameters, including resolution, bitrate, and frame rate, to balance video quality and network adaptability.

[0060] S400: Based on the priority and resource usage of multiple video streams, it coordinates the allocation of network bandwidth resources and controls the transmission parameters of each video stream to ensure the efficiency of multi-video transmission.

[0061] S500: Intelligently identifies packet loss types based on the real-time network environment, dynamically adjusts FEC redundancy parameters, and adds redundant data in network transmission to combat packet loss problems, thereby balancing packet loss recovery capability and bandwidth overhead.

[0062] Furthermore, in the network status awareness stage, multi-dimensional indicators such as network link bandwidth, latency, jitter, and packet loss rate are continuously collected. For example... Figure 3 As shown, a time-series sliding window is used to record bandwidth fluctuation characteristics and identify network congestion events in specific time periods, such as t=28-35 seconds. Real-time quantified spatiotemporal evolution data provides support for policy decisions. It is important to understand that the delay heatmap also reflects the stability characteristics of the network path, which plays a fundamental role in subsequent transmission parameter control.

[0063] Furthermore, a lightweight CNN model is used to extract inter-frame motion vectors from the video. For example... Figure 4 As shown, the pixel displacement vector field is visualized based on a streamline diagram, where the red motion trajectory marks the motion characteristics of fast-moving objects. The spatial distribution uniformity of the motion vectors, such as the contrast between sparse and dense regions and their instantaneous intensity, are represented by color mapping, which together constitute the quantitative basis for the activity of video content, providing a foundation for analyzing the dynamic characteristics of content for encoding optimization.

[0064] Furthermore, in the dynamic adjustment of encoding parameters, the network status collected previously and the content activity are integrated for decision-making. For example... Figure 5As shown in the three-dimensional decision surface, a non-linear mapping relationship to the target bitrate (Z-axis) is established using available bandwidth (X-axis) and content activity (Y-axis) as input variables. During implementation, the combination of encoding parameters is dynamically selected along the black optimization path. For example, when bandwidth is limited (Bw=2Mbps), the resolution is reduced instead of the frame rate to maintain the recognizability of key motion information.

[0065] Furthermore, for multi-stream video transmission, bandwidth resource scheduling is implemented based on preset service priorities. For example... Figure 7 The layered filling diagram shown implements guaranteed allocation for high-priority streams and elastic compression for medium- and low-priority streams under total bandwidth constraints. A congestion event at a specific time, such as t=28 seconds, triggers a dynamic resource reallocation mechanism, temporarily limiting the bandwidth of low-priority streams to ensure the continuity of critical video stream transmission.

[0066] Furthermore, the packet loss control strategy distinguishes between random packet loss and continuous packet loss based on real-time network monitoring data. For example... Figure 6 The twin-graph analysis employs a linearly increasing FEC redundancy (5-30%) for random packet loss scenarios, while an exponential redundancy strategy is used for sudden, continuous packet loss, requiring 25% redundancy when the packet loss length is greater than 8. This strategy establishes a packet loss type identification module in the control system kernel to dynamically generate redundancy parameters that match the current network impairment characteristics.

[0067] This embodiment establishes a control link between network status, content characteristics, and transmission strategy, forming a multi-dimensional collaborative dynamic adjustment mechanism to improve network resource utilization while ensuring basic video quality.

[0068] Example 2: To address the issue of insufficient network status monitoring accuracy leading to transmission parameter mismatch during video backhaul, this example further optimizes the acquisition process of key network indicators and constructs an accurate network environment model through multi-dimensional indicator collaborative analysis. The following details this with reference to the accompanying drawings:

[0069] Specifically, during the bandwidth monitoring phase, uplink available bandwidth is estimated by periodically sending probe packets and analyzing the response time. For example... Figure 3 As shown, a time-series sliding window (X-axis represents time) records bandwidth fluctuations, with the bandwidth curve and congestion event annotations capturing real-time trends. After identifying bandwidth drop characteristics within a specific time period (t=28-35 seconds), the peak bandwidth within each unit time window is statistically analyzed to form a dynamic view of bandwidth capacity. It's important to understand that this process indirectly verifies the estimation accuracy by calculating RTT×BDP (bandwidth-delay product), providing a spatiotemporal evolution basis for congestion identification.

[0070] Furthermore, in the delay measurement phase, the round-trip time and one-way transmission delay of data packets are collected simultaneously. For example... Figure 3As shown, the color depth mapping jitter variance σ of each data point in the delay heatmap (Y-axis represents delay values) is... j 2 The spatial distribution of low-jitter and high-jitter regions reflects network path stability. By analyzing the statistical characteristics of this heatmap, such as the change in jitter standard deviation over time, the adjustment threshold for the decoder buffer size is determined. It is understood that the buffer configuration needs to match the latency jitter stability curve to avoid playback jitter.

[0071] Furthermore, for packet loss analysis, a dual-time-scale statistical model is set up: the instantaneous packet loss rate η is calculated using a short-term window (second-level). loss Long-term window (minute-level) analysis of average packet loss rate. For example... Figure 3 The bottom subplot, a packet loss rate histogram, separates random packet loss from continuous packet loss, identifying pattern characteristics based on burst packet loss events such as continuous packet loss pulses at t=68-73 seconds. This pattern classification result is directly output to the subsequent FEC parameter controller, providing network impairment type input for redundancy strategies.

[0072] The advantage of this embodiment is that, through, as shown in... Figure 3 The multi-dimensional index spatiotemporal evolution collaborative acquisition mechanism shown enables efficient modeling of network state. The coupled analysis of bandwidth, latency, and packet loss patterns provides a joint input space for transmission decisions, avoiding policy adaptation bias caused by distortion of a single index.

[0073] Example 3: To address the issue of increased latency due to excessive computational resource consumption in real-time video analysis, this example further optimizes the extraction process of motion vectors between video frames. It achieves accurate quantification of dynamic features of video content through a lightweight neural network model, thereby adapting to intelligent adjustment of encoding parameters.

[0074] To address the issues of computational complexity and resource consumption, this embodiment employs a lightweight CNN model. Model compression techniques, such as pruning and quantization, are used to reduce the parameter size, making it suitable for resource-constrained embedded devices. Figure 4 As shown, the motion vector field of the video frame is obtained by extracting features from the pixel differences between adjacent frames. The streamline map is generated based on the u and v components of the x and y grid points, using color mapping to calculate motion intensity. It's important to understand that this model utilizes the pixel gradient information of the input frame to construct the motion trajectory map, as shown below. Figure 4 The red curve in the image captures the motion pattern of objects by analyzing the continuity of the trajectory, thereby reducing the computational overhead of real-time inference.

[0075] Furthermore, the quantization process of motion vector features includes deriving video content activity assessment results from the vector field. For example... Figure 4As shown, the comparison between the trajectory of a rapidly moving object and the motion intensity of other areas generates a normalized activity index, Av. This index is calculated using the overall distribution characteristics of the statistical vector field, such as the average velocity magnitude and the rate of change of direction. Understandably, the activity assessment combines the inter-frame pixel variation amplitude and motion vector consistency; for example, when an object's trajectory at time t=10 shows rapid motion, Av increases significantly, providing input for subsequent coding control.

[0076] Furthermore, the activity assessment results directly drive the dynamic adjustment of encoding parameters. For example... Figure 5 As shown, content activity Av, as one of the key input parameters (Y-axis), works in conjunction with network bandwidth Bw in the adaptive decision surface to generate an optimized value for the target bitrate Rc. In implementation, the spatiotemporal evolution data of Av is output in real time through a lightweight CNN, forming part of the encoder's input space. Understandably, this mechanism avoids repetitive analysis of the original frames, achieving efficient content awareness by utilizing motion vector features.

[0077] The advantage of this embodiment is that it allows for: Figure 4 and Figure 5 The linkage mechanism between motion vector extraction and encoding decisions forms a closed-loop control. The lightweight CNN model reduces resource consumption while ensuring the real-time reliability of the activity index, avoiding lag in encoding parameter adjustments compared to changes in video content.

[0078] Example 4: To address the dual impact of network fluctuations and the dynamic nature of video content on encoding parameters, this example further optimizes the dynamic encoding adjustment mechanism. Figure 5 The decision surface shown achieves dual adaptation of network state and content features.

[0079] When sufficient network bandwidth resources are detected, i.e. Figure 5 In regions where the Bw value is higher than 10 Mbps, this embodiment improves quantization accuracy and sampling density to enhance video detail reproduction capabilities. For example... Figure 5 As shown in the right half of the decision surface, under conditions of high Bw and high Av, such as Bw=15Mbps and Av>0.6, the coding bitrate Rc increases non-linearly to over 2000kbps. It's important to understand that this strategy improves image quality by increasing motion compensation accuracy and reducing macroblock segmentation granularity, while avoiding overloading the encoder's computational resources.

[0080] Furthermore, the dynamic range of the video content is determined through... Figure 4 The motion vector field is quantized to drive parameter adjustment. For example... Figure 4 The region of high motion corresponding to the trajectory of a fast-moving object is represented by orange-red in color mapping (value range > 1.2). In this case, the system... Figure 5The gradient change along the vertical dimension (Av axis) of the decision surface actively increases the frame sampling rate to 60fps and employs full-resolution encoding. This is understandable when the vector field exhibits a low-intensity blue distribution, such as... Figure 4 Top right corner If the value is less than 0.3, the frame rate reduction mechanism will be triggered and time-domain downsampling will be enabled.

[0081] In network congestion scenarios, such as Figure 3 In the event of a sudden bandwidth drop between 28 and 35 seconds, this embodiment initiates a fast bitrate adjustment mechanism. This mechanism comprises a three-layer response: firstly, based on... Figure 3 Real-time Bw monitoring values ​​cause Rc to drop sharply to the lowest region of the surface. Figure 5 The bottom left corner Rc < 500kbps; secondly, non-keyframe encoding is interrupted, and only I-frames carrying complete information are transmitted; finally, as shown... Figure 7 As shown, the bandwidth allocation module allocates a high-priority channel to the I-frame stream, indicated by the red "high-priority" bandwidth area. It's important to understand that this process utilizes... Figure 3 The RTT×BDP product value is used to predict congestion risk, and a bandwidth reduction plan is initiated 300ms before the bandwidth drops below the threshold (e.g., <4Mbps).

[0082] The advantage of this embodiment is that: Figure 5 The decision surface constructs a collaborative control mechanism between Bw and Av, avoiding image quality fluctuations caused by single-factor decision-making. For example... Figure 7 As shown in multi-stream scheduling, the Rc stability of high-priority video streams such as surveillance alarm footage is maintained during dynamic resource allocation, while content activity assessment is integrated into the vertical dimension of the surface. Figure 5 The Av axis achieves a quantitative balance between motion detail and bandwidth consumption.

[0083] Example 5: To address the transmission quality degradation caused by resource contention during concurrent transmission of multiple video streams, this example further optimizes the global scheduling mechanism for network bandwidth resources. By establishing a service-driven priority evaluation model and a dynamic quota adjustment system, efficient and coordinated allocation of network resources is achieved. Detailed implementation methods are combined with... Figure 4 Motion vector analysis and Figure 7 The intelligent scheduling architecture unfolds the complete workflow.

[0084] First, a video stream priority evaluation model is constructed, which achieves priority quantification and classification through two-dimensional feature extraction. The business type dimension uses deep packet inspection technology to identify the application scenario characteristics of the video stream, marking key business flows such as video conferencing and video backhaul control as high-priority channels (red). The content importance dimension is based on… Figure 4 The inter-frame motion field analysis technique shown calculates the magnitude of motion vectors in a specific region within consecutive video frames. The urgency of the event is determined by the duration of the event. It should be understood that when the vector amplitude in a specific area continuously exceeds a set threshold, the system automatically escalates the video stream to the emergency transmission queue.

[0085] Furthermore, the scheduling engine issues dynamic resource allocation instructions to the underlying transport stack through a standardized southbound interface. For example... Figure 7 In the hierarchical quota model shown, under the constraint of total available bandwidth, the system constructs a three-layer bandwidth allocation channel: red, orange, and green. High-priority flows occupy the top layer of the transmission channel (red area) to ensure the continuous transmission of critical service video frames; medium-priority flows (orange area) and low-priority flows (green area) dynamically shrink proportionally. It can be understood that this scheduling algorithm monitors network congestion characteristics in real time, and when it detects... Figure 7 When a competition signal occurs in the resource contention zone, such as a scheduling event at 28s, the system prioritizes compressing the quota of low-priority flows while maintaining stable bandwidth usage for high-priority flows.

[0086] The advantage of this embodiment is that it allows for: Figure 4 Content Analysis and Figure 7 The organic coordination of scheduling models establishes a precise mapping path from business semantics to network resources. A bandwidth isolation mechanism based on flow priority is implemented at the transport layer, transforming resource conflicts during network congestion into a controlled hierarchical allocation process, effectively avoiding the risk of transmission interruptions for high-importance video streams. The dynamic quota model, by sensing real-time network state changes, ensures the scientific allocation of bandwidth resources among streams of different priorities, providing a stable quality assurance foundation for multi-channel video transmission.

[0087] Example 6: To address the video transmission quality degradation caused by network channel instability, this example further refines the control strategy of the forward error correction mechanism. Through refined identification of network packet loss characteristics and coordinated adjustment of FEC parameters, a balance between transmission reliability and bandwidth efficiency is achieved. Detailed implementation methods are combined... Figure 3 Network link monitoring data and Figure 6 The effectiveness of redundancy strategies is analyzed to construct a closed-loop optimization system.

[0088] Furthermore, based on Figure 3 The network link multi-dimensional monitoring system shown includes bandwidth mutation events, latency change trends, and jitter heatmaps, capturing the packet loss rate magnitude and its temporal distribution characteristics in real time. When the packet loss pattern recognition module detects, for example... Figure 3 A network congestion event marked at 28 seconds—characterized by a sudden drop in bandwidth and a dense red area on the latency heatmap—is identified by the system as a risk of continuous packet loss. This is automatically triggered at this time. Figure 6The block redundancy enhancement strategy shown in the right figure increases the FEC redundancy ratio to the 15%-25% range. The key technical point is the use of a sliding window to analyze the characteristics of consecutive packet loss lengths, such as... Figure 6 As shown on the horizontal axis, when the packet loss length exceeds 3 data packets, the redundancy allocation coefficient is increased linearly according to the packet loss sequence length. At this time, the FEC coding block size is synchronously expanded to the range of 32-64 packets to ensure the recovery capability of long sequence packet loss.

[0089] Furthermore, during the stable phase of the network channel, such as Figure 3 In the 0-20 second range, if the jitter heatmap shows a blue-green distribution and the packet loss rate is less than 2%, the system automatically identifies it as a random packet loss mode. Activation is performed under this state. Figure 6 The performance curve in the left figure corresponds to a refined control strategy: compressing FEC redundancy to the 5%-10% range while reducing the code block size to 8-16 packets. It can be understood that this operation... Figure 6 The overlapping characteristics of multiple curves in the low packet loss rate region are realized - when the random packet loss rate is less than 5%, 5%-10% redundancy can meet more than 95% of the recovery requirements, avoiding bandwidth waste caused by high redundancy.

[0090] Furthermore, the system is based on Figure 3 The color intensity changes of the bottom latency jitter heatmap are used to calculate the latency stability index of the transmission path in real time. When a jitter intensification event is detected in the yellow to red area, such as... Figure 3 Marking the 85-second mark, an automatic block size compression strategy is implemented: the upper limit of FEC encoded blocks is constrained to within 16 packets. It is important to emphasize that this strategy addresses the processing latency issue caused by large-block FEC encoding. By limiting the size of the encoded blocks, it reduces data encapsulation time and ensures that the encoding and decoding process is completed under a 100ms latency constraint.

[0091] Through this embodiment: Figure 3 Network state perception and Figure 6 The performance model forms a coordinated control system, establishing a closed-loop control logic of packet loss feature identification → parameter decision → performance verification. In random packet loss scenarios, it reduces redundancy overhead to save bandwidth resources; under sudden packet loss conditions, it enhances error correction capabilities to ensure transmission continuity; and it balances the conflicting requirements of packet loss resistance and processing latency through a block-scale dynamic adaptation mechanism. This hierarchical control model ensures that the FEC redundancy parameters are always optimally matched to the network channel characteristics, providing technical assurance for end-to-end reliable transmission of video streams.

[0092] Example 7: To address the challenge of balancing video coding redundancy and network transmission efficiency, this example refines the source and channel collaborative control mechanism. Figure 5 Adaptive coding control module Figure 6 FEC anti-packet loss module and Figure 7 The intelligent scheduling and decision-making module constitutes a two-tiered optimization system. The specific implementation method is as follows:

[0093] based on Figure 4 Dynamic analysis of the motion vector field between video frames; when large-scale directional motion is detected, triggering... Figure 5 The high-motion compression strategy for the decision surface shown. In high-motion scenarios with content activity > 0.6, the corresponding... Figure 4 In the bright yellow area on the right, the system automatically lowers Rs along the gradient descent direction of the curved surface, while simultaneously switching the GOP structure to IBBP mode. Understandably, this reduces redundant data by lowering the intra-frame refresh rate, while motion compensation coding preserves key motion information, resulting in a reduction of the source output bit rate by approximately 30%.

[0094] Furthermore, at the network transport layer, Figure 3 The spatiotemporal evolution data shown are Figure 6 The effectiveness model forms the basis for joint decision-making:

[0095] when Figure 3 When the packet loss rate curve detects consecutive packet loss events, activate Figure 6 The block redundancy strategy selects a redundancy range of 25%-30% based on the packet loss length, and the coding block is synchronously enlarged to 48 packets.

[0096] Figure 7 The bandwidth scheduling module receives the characteristics of the multi-stream output by the encoder in real time. When a high-priority stream triggers a low-resolution mode, the bandwidth resources it releases are dynamically reallocated to medium and low-priority streams. Figure 7 As shown, at t=28 seconds, the bandwidth demand of the high-priority flow drops by 1.2Mbps, and this resource is immediately allocated to the medium-priority flow in the green area.

[0097] Furthermore, the received QoE evaluation data triggers two-parameter correction via the feedback channel:

[0098] When the motion blur index at the decoder is greater than 0.4, forced boosting is applied. Figure 5 The Rs parameter of the decision surface should be lowered when the packet retransmission rate is >8%. Figure 6 The FEC block size is increased to 24 packets. This feedback mechanism allows the channel control parameters of the source to continuously approach the optimal operating point, forming... Figure 5 curved surfaces and Figure 6 Dynamic coupling of curves.

[0099] In this embodiment, redundancy control in the source coding stage provides flexibility for channel transmission, while FEC and bandwidth scheduling strategies on the channel side conversely ensure source quality. This bidirectional collaborative mechanism avoids the problem of coding parameters being disconnected from network conditions in traditional schemes, enabling the video stream to automatically trigger a graded degradation strategy when bandwidth is limited, and to achieve the optimal level of damage resistance protection when the network is congested. Especially in Figure 7 In the scheduling strategy, priority routing and resource reallocation form a dynamic balance system to ensure that critical video streams still meet end-to-end latency constraints when the link fluctuates.

[0100] Example 8: To address the control problem of dynamically matching multi-dimensional parameters with video content features in complex network environments, this example constructs an intelligent network control system architecture with a hierarchical collaborative mechanism, such as... Figure 8 As shown, precise control of low-latency video backhaul is achieved through data interaction and strategy linkage between modules.

[0101] Furthermore, the network status detection module constructs a spatiotemporal network parameter monitoring system through distributed sampling nodes. The bandwidth detection unit uses a time window sliding algorithm to estimate available bandwidth in real time. By periodically sending probe packet sequences and calculating the statistical characteristics of the sequence response intervals, it forms a data structure such as... Figure 3 The upper part shows the bandwidth variation curve. During network congestion events, such as the red-filled area between 28-35 seconds, the algorithm automatically switches to a high-density sampling mode to capture sudden bandwidth changes. The latency monitoring unit generates bidirectional probe packet pairs based on the UDP protocol and constructs a latency jitter heatmap by calculating the probability density distribution of the round-trip time (RTT). This heatmap, with time as the vertical axis and jitter amplitude as the horizontal axis, clearly presents the temporal characteristics of latency fluctuations. The packet loss statistics unit maintains the packet sequence number sequence within a sliding window, identifies packet loss types such as random or continuous packet loss in real time, and outputs a feature vector containing parameters such as packet loss rate, packet loss continuity, and burst duration, providing basic data support for subsequent FEC strategy adjustments. It's important to understand that each monitoring unit shares the underlying transmission channel through a time-division multiplexing mechanism, reducing system overhead while ensuring data acquisition accuracy.

[0102] Furthermore, considering the computing power constraints of edge computing nodes, this embodiment optimizes the video content analysis module, such as... Figure 4 As shown. The lightweight CNN processing unit uses depthwise separable convolutions instead of traditional convolutional layers, compressing the number of model parameters to less than 30% of the original ResNet18. Simultaneously, a 15×15 normalized grid is introduced in the input layer, keeping the processing time per frame below 20ms. The motion vector extraction unit calculates pixel-level motion vectors based on the difference operation between adjacent frames using the Farneback optical flow algorithm, generating a vector field containing direction and velocity information, such as... Figure 4As shown in the diagram. Based on this, the system divides the video frames into blocks. When the motion vector density within a single block exceeds a preset threshold, the region is determined to be high-activity content. Finally, a global content feature evaluation result is generated by weighted averaging of regional activity. This hierarchical processing mechanism preserves the detailed features of dynamic video changes while avoiding the performance bottleneck caused by intensive full-frame computation.

[0103] Furthermore, such as Figure 5 As shown, the adaptive coding control module constructs a three-dimensional decision space based on network bandwidth and content activity. The parameter adjustment algorithm unit receives the available bandwidth value (X-axis) output by the network state detection module and the content activity (Y-axis) output by the video content analysis module, and calculates the optimal coding parameters (Z-axis represents the target bitrate) through a preset nonlinear mapping model. This model uses a power function fitting method to represent the bitrate adjustment strategy as a joint function of network resources and content characteristics. For example, when available bandwidth increases and content activity is high, the system automatically increases the coding bitrate to retain more detailed information; while when bandwidth is limited or content tends to be static, it maintains smooth transmission by reducing the bitrate and adjusting resolution parameters. The generation process of the decision surface integrates historical transmission data training and real-time parameter calibration to ensure that the coding strategy can dynamically adapt to the dual changes of network and content, forming an optimal balance between video quality and transmission latency.

[0104] Furthermore, for scenarios involving concurrent transmission of multiple video streams, the intelligent scheduling decision module constructs a priority-based bandwidth allocation model, such as... Figure 7 As shown. The system presets a Quality of Service (QoS) level for each video stream. High-priority streams, such as video backhaul for autonomous vehicle control, enjoy resource reservation rights during bandwidth allocation, while low-priority streams, such as non-real-time recording video, adopt an elastic allocation strategy. The scheduling algorithm monitors the total available bandwidth in real time. Figure 7 The fluctuation curves show the resource occupancy of each stream. An optimal allocation scheme is generated using dynamic programming: when bandwidth is ample, resources are allocated according to the peak demand of each stream; when bandwidth bottlenecks occur, low-priority streams are compressed in bitrate or their transmission intervals are adjusted based on priority coefficients to ensure the transmission stability of critical video streams. It should be noted that this module introduces a prediction mechanism, predicting bandwidth demand trends for future periods based on historical scheduling data, and adjusting the transmission parameters of each stream in advance to reduce decision-making latency, forming a proactive resource management strategy.

[0105] Furthermore, the FEC packet loss mitigation module achieves precise control by establishing a mapping relationship between packet loss types and redundancy parameters. The redundancy strategy decision unit analyzes the packet loss pattern characteristics output by the network state detection module, such as... Figure 6As shown, when an increase in random packet loss rate is detected, a linearly increasing strategy is used to adjust redundancy to ensure effective recovery with low redundancy overhead. In the face of continuous packet loss events, the FEC block size is dynamically increased based on the duration of packet loss bursts, improving the recovery success rate by expanding the protection scope of redundant data. In practice, the system maintains a pre-set redundancy strategy library containing optimal redundancy and block size combinations for different packet loss scenarios. The decision unit quickly retrieves the corresponding strategy using a pattern matching algorithm, avoiding increased latency caused by online computation. This scenario-based differentiated processing mechanism enables the FEC scheme to maintain high efficiency when network packet loss characteristics change, reducing the additional bandwidth consumption of redundant data.

[0106] Furthermore, the feedback adjustment mechanism module periodically collects operational data from each module, such as the frequency of encoding parameter adjustments, scheduling decision response time, and FEC recovery success rate, and analyzes the effectiveness of the current control strategy through a preset performance evaluation index system. When a key indicator is detected to deviate from a preset threshold, the system triggers a strategy update process: for example, if the bitrate adjustment of the adaptive encoding module lags behind network changes, the parameter update cycle of the decision model is automatically optimized; if the intelligent scheduling module experiences resource allocation conflicts, the priority weights of each flow are recalibrated. This closed-loop adjustment mechanism enables the system to continuously adapt to changes in the network environment and business needs during long-term operation, achieving incremental optimization of the control strategy through a data-driven approach.

[0107] Through the modular design and collaborative control architecture of this embodiment, the system achieves intelligent processing throughout the entire process, from network status awareness and content feature extraction to transmission strategy generation. Each module has an independent functional implementation unit, and they also form a strategy linkage through standardized data interfaces, ensuring a dynamic balance between video quality, transmission latency, and resource utilization under complex network conditions, providing a systematic solution for low-latency video backhaul scenarios.

[0108] Example 9: To address the latency control and quality assurance issues in multi-device collaborative transmission during video conferencing scenarios, this example provides a video conferencing device integrating intelligent network control functions. Through a deep coupling design of encoder devices, decoder devices, and an intelligent control system, end-to-end low-latency video return control is achieved. The following details the device's hardware architecture, module collaboration mechanism, and data processing flow.

[0109] The video conferencing equipment in this embodiment adopts an end-to-end layered architecture. Encoder devices are deployed at the video acquisition end, such as conference room camera terminals, while decoder devices are deployed at the video receiving end, such as participant terminal displays. Both achieve bidirectional data interaction through an intelligent network control system. It is important to understand that the intelligent network control system can be integrated into the device as an independent hardware module, or it can be deployed on edge computing nodes via software containerization to form a distributed control architecture. This deployment method balances the real-time requirements of local device processing with the coordination requirements of global network strategies, reducing the computing load on the edge while ensuring the dynamic adaptability of the control strategy.

[0110] Furthermore, the core functional units of the encoder device include an image acquisition module and a preprocessing module. The image acquisition module uses a progressive scan CMOS sensor, supporting dynamic switching from 1080P to 4K resolution. Combined with automatic white balance and exposure control algorithms, it ensures the color reproduction and brightness uniformity of the original video stream. The preprocessing module integrates lightweight video content analysis functions, based on a lightweight CNN model such as... Figure 4 The motion vector extraction mechanism shown performs differential operations on consecutive frames to generate an inter-frame motion vector field. Specifically, this module divides video frames into 16×16 pixel macroblocks. By calculating the motion vector density and direction distribution of each macroblock, it outputs real-time content activity assessment results, such as the proportion of high dynamic range regions and the average speed of moving objects. These parameters are directly input to the adaptive coding control module of the intelligent network control system, providing content feature basis for dynamically adjusting coding parameters such as target bitrate, GOP structure, and entropy coding mode. It should be noted that the preprocessing module adopts a pipelined processing architecture, parallelizing the three processing stages of image noise reduction, resolution scaling, and motion feature extraction, keeping the single-frame processing latency within 15ms to meet the timing requirements of low-latency transmission.

[0111] Furthermore, the decoder device undertakes the dual functions of network data decoding and image display. In the decoding stage, it supports adaptive switching between multiple encoding formats such as H.264, H.265, and AV1, automatically configuring the decoding engine parameters by parsing the parameter set information in the bitstream. To address potential packet loss or bit error issues in network transmission, the decoder integrates a forward error correction (FEC) module, forming a closed loop with the intelligent control system's FEC anti-packet loss module: when continuous packet loss is detected causing decoding failure, the decoder sends packet loss location and type information, such as random or sudden packet loss, to the control system via a feedback channel, triggering dynamic adjustment of the FEC redundancy parameters. Figure 6The redundancy strategy performance analysis mechanism is shown below. In the image display stage, the decoded video stream first passes through the frame rate synchronization module, which dynamically adjusts the display frame rate through a buffer queue to eliminate frame rate fluctuations caused by network transmission. It then enters the resolution adaptation module, which scales the image according to the physical resolution of the receiving end's display screen, such as 1920×1080 or 3840×2160, using a bicubic interpolation algorithm to improve the image quality after scaling. Furthermore, the decoder device is equipped with a real-time monitoring interface to continuously collect performance indicators such as decoding latency, buffer occupancy rate, and image stuttering rate, and feeds this data back to the feedback adjustment mechanism module of the intelligent control system, providing backend measured data for optimizing the transmission strategy.

[0112] Furthermore, the intelligent network control system constructs a multi-level data interaction channel between the encoder and decoder. The network status detection module collects key indicators such as bandwidth, latency, jitter, and packet loss rate of the transmission link in real time, such as... Figure 3 The spatiotemporal evolution monitoring mechanism shown provides network-layer foundational data for encoding parameter adjustment and FEC strategy configuration. Specifically, when a decrease in available bandwidth is detected, the control system synchronously triggers the encoder's rate control module, based on... Figure 5 The adaptive decision surface model shown dynamically reduces the target bitrate under the constraint of content activity assessment results, while adjusting quantization parameters to maintain relative stability of visual quality. If a sudden packet loss event is detected, the FEC anti-packet loss module immediately switches to a high redundancy strategy, increasing the size and number of redundant data blocks, such as... Figure 6 The continuous packet loss scenario handling mechanism shown improves the packet loss resistance of the bitstream in harsh network environments.

[0113] Furthermore, in scenarios with concurrent multi-video streams, such as multi-party video conferences, the intelligent scheduling and decision-making module plays a crucial coordinating role. This module assigns priority tags to each video stream, such as setting the speaker's video to the highest priority and the participants' videos to a standard priority, based on... Figure 7 The intelligent bandwidth scheduling model shown calculates the bandwidth allocation scheme for each stream in real time. When the total available bandwidth is insufficient, the system prioritizes the transmission quality of high-priority streams, releasing bandwidth resources by reducing the encoding resolution of low-priority streams or increasing the I-frame interval. When the network is stable, a load balancing strategy is used to optimize the overall resource utilization. This hierarchical scheduling mechanism, together with encoding control and packet loss prevention strategies, ensures that video streams of different importance can meet latency and quality requirements when sharing network resources.

[0114] Furthermore, to enhance the environmental adaptability of the equipment, this embodiment introduces several optimized designs at the hardware level: the encoder integrates a temperature sensor and a fan speed control module, ensuring stability during long-term high-load operation through dynamic adjustment of heat dissipation strategies; the decoder is equipped with a dual-port redundant design, supporting automatic switching between wired and wireless networks, seamlessly switching to the backup link in the event of a single link failure. At the software level, the intelligent control system adopts a microservice architecture, encapsulating functions such as network detection, content analysis, and encoding control into independent service units, and achieving low-latency communication between modules through message queues, facilitating subsequent function expansion and algorithm upgrades. In addition, the equipment supports a remote management interface, allowing maintenance personnel to monitor the equipment's operating status and configure policy parameters in real time through a secure channel, forming a centralized equipment management system.

[0115] Through the device architecture design in this embodiment, the video conferencing system achieves intelligent control across the entire chain, from video acquisition and encoding transmission to decoding and display. The encoder and decoder form a closed-loop strategy through an intelligent network control system, enabling front-end acquisition and processing, and back-end display adaptation to respond in real-time to changes in network status and content characteristics. This effectively reduces end-to-end transmission latency in complex network environments while ensuring the clarity and smoothness of video content. This device-level deep integration solution provides an engineered implementation path for low-latency applications such as remote video conferencing and real-time collaboration. Through inter-module collaborative optimization and deep coupling of hardware and software, it ensures that the system maintains stable and reliable operating performance in different deployment environments.

[0116] Example 10: To address the portability and efficient execution of low-latency video backhaul methods across different computing platforms, this example provides a specific implementation method using a computer-readable storage medium. By using structured programming, the method steps are transformed into an executable code system, ensuring optimized adaptation between software logic and hardware resources.

[0117] Specifically, computer-readable storage media include, but are not limited to, non-volatile storage carriers such as solid-state storage devices (SSDs), disk arrays, and cloud storage units. The computer program stored on it adopts a modular architecture, corresponding to the steps of implementing an intelligent network control method for low-latency video backhaul. The main program is divided into six functional modules: network status monitoring, video content analysis, encoding strategy control, intelligent scheduling decision-making, packet loss mitigation, and feedback adjustment. Each module achieves data interaction and logical collaboration through standardized interfaces. It is important to understand that the storage media is compatible with mainstream processor architectures such as x86 and ARM, and supports Docker containerized deployment. By encapsulating runtime environment dependencies through container images, seamless migration of the program across different hardware platforms such as edge computing nodes and server clusters is ensured.

[0118] Furthermore, the network status monitoring module employs asynchronous I / O technology, using multi-threaded concurrent acquisition of network link parameters such as bandwidth, latency, jitter, and packet loss rate. The data acquisition frequency can be dynamically adjusted via configuration files (e.g., a default of 50ms / time). The video content analysis module integrates the inference code of a lightweight CNN model. The model file is stored in PROTOBUF format on the medium, loaded into memory at runtime, and uses GPU-accelerated libraries (such as CUDA) to quickly extract inter-frame motion vectors. The content activity features output by this module are transmitted to the encoding strategy control module in real time via a shared memory mechanism. The encoding strategy control module contains the code implementation of a nonlinear mapping algorithm, constructing a three-dimensional decision logic based on network status parameters and content features. It dynamically adjusts encoding parameters, such as target bitrate and GOP structure, through conditional statements. These parameters are transmitted to the encoder device via a socket interface.

[0119] Furthermore, the intelligent scheduling decision module employs a priority queue data structure to manage multiple video streams, and uses dynamic programming algorithm code to achieve real-time allocation of bandwidth resources. The algorithm logic includes a pre-defined service quality level mapping table and supports reading configuration files for user-defined priority strategies. The packet loss mitigation module contains index code for the FEC redundancy strategy library, matches pre-defined strategies based on real-time packet loss characteristics, and uses bitwise operations to dynamically generate and verify redundant data. The feedback adjustment module periodically collects the operation logs of each module, evaluates the effectiveness of the strategies through statistical analysis code, and triggers an automatic update process for the configuration file when abnormal performance indicators are detected, achieving closed-loop optimization of the control strategy.

[0120] Furthermore, an error handling mechanism is embedded in the program code, with catch functions set up to handle exceptions such as network connection interruptions and model loading failures, and exception stack information is recorded through a logging system. The storage medium supports incremental updates; when an algorithm module needs to be upgraded, only the corresponding code file needs to be replaced without affecting the overall program operation, improving system maintainability. It should be noted that a memory paging mechanism is used to manage data caching during program execution, storing frequently accessed network state data and encoding parameters in a high-speed cache area, reducing processor memory access latency and ensuring the real-time execution requirements of method steps.

[0121] Through the computer-readable storage medium of this embodiment, the intelligent network control method for low-latency video backhaul can be transformed into an independently deployable software entity. Its modular design and hardware-independent architecture provide a standardized solution for system integration in different application scenarios. The efficient execution of the program in the storage medium depends on the optimized design of the data interaction interface and the completeness of the exception handling mechanism, ensuring that the logical integrity of the method steps and the dynamic adaptability of the control strategy can be maintained even in complex computing environments, forming an intelligent video transmission solution that combines software and hardware.

[0122] Although the present invention has been specifically described above with reference to preferred embodiments, it should be understood that the present invention is not limited to the embodiments described above. Various modifications and variations can be made by those skilled in the art without departing from the spirit of the present invention, and such modifications and variations should fall within the scope defined by the appended claims and their equivalents.

Claims

1. An intelligent network control method for low-latency video backhaul, characterized by, The method comprises the following steps: Real-time collection of network key indicators of network links, including bandwidth, delay, jitter and packet loss rate, provides real-time network environment data support for transmission strategy adjustment; Extracting the motion vector between video frames through a lightweight CNN model to determine the video content activity and assess the frequency and amplitude of content changes between adjacent frames, providing content characteristics for encoding strategy optimization; Comprehensive network key indicators and video content activity, dynamically adjust the video encoding parameters, including resolution, code rate, frame rate, to balance video quality and network adaptability; According to the priority and resource occupation of multiple video streams, allocate network bandwidth resources, control the transmission parameters of each video stream to ensure the efficiency of multi-channel video transmission; According to the real-time network environment, intelligently identify the packet loss type, dynamically adjust the FEC redundancy parameters, add redundant data in network transmission to resist packet loss problems, and balance the packet loss recovery ability and bandwidth overhead; The step of dynamically adjusting the video encoding parameters comprises: When the network bandwidth is sufficient, high code rate and high resolution encoding are used to improve the video detail quality, and when the network bandwidth is limited, the code rate and resolution are automatically reduced to ensure transmission continuity; For high dynamic video content, increase the frame rate and resolution to retain motion details, and for static or low dynamic content, reduce the frame rate and code rate to reduce data volume; When the network is congested or the bandwidth drops sharply, trigger the fast code rate adjustment mechanism, and preferentially transmit key I frames to ensure continuous display at the video decoding end; The step of allocating network bandwidth resources comprises: Establish a video stream priority evaluation model to allocate priority according to the service type and content importance of the video stream; Based on real-time network bandwidth and the priority of each video stream, issue scheduling instructions to the network transmission layer through a standard interface to dynamically adjust the transmission bandwidth quota and transmission priority of each video stream, avoiding resource contention among multiple video streams; The step of dynamically adjusting the FEC redundancy parameters comprises: When detecting that the network packet loss rate is rising and showing a continuous packet loss mode, increase the FEC redundancy to improve the packet loss recovery ability, and when the network packet loss rate is low and random, reduce the FEC redundancy to reduce the bandwidth overhead; According to the delay and jitter characteristics of the network link, dynamically adjust the block size and redundancy ratio of the FEC encoding, balance the anti-packet loss performance and encoding processing delay, and ensure real-time video transmission. 2.The intelligent network control method of low-latency video backhauling of claim 1, wherein, The collection step of the network key indicators comprises: Estimate the available uplink bandwidth by sending probe packets and analyzing response time, track the bandwidth change trend in real time to identify network congestion, and count the peak bandwidth per unit time; Measure the round-trip delay and one-way transmission delay of data packets, analyze the stability of delay jitter to determine the decoding end buffer size; Statistical instantaneous packet loss rate in a short time window and average packet loss rate in a long time period to identify the mode of continuous packet loss or random packet loss, and provide a basis for adjusting the FEC redundancy parameters. 3.The intelligent network control method of low-latency video backhauling of claim 1, wherein, The extraction step of the motion vector between video frames comprises: For real-time video analysis scenarios, reduce the computational complexity and resource consumption of the lightweight CNN model; Based on the pixel difference and motion trajectory of adjacent video frames, a motion vector feature representing content change is extracted, the dynamic degree of video content is quantified, and a video content activity evaluation result is generated to drive dynamic adjustment of the encoding parameters. 4.The intelligent network control method of low-latency video backhauling of claim 1, wherein, The network control method optimizes the source and channel jointly to transmit with low latency: at the source side, an adaptive coding control module is used to reduce redundant data, at the channel side, an FEC anti-packet loss module and an intelligent scheduling decision module are used to improve transmission reliability and bandwidth utilization, and a feedback strategy is combined to form end-to-end dynamic optimization.

5. An intelligent network control system for low-latency video backhauling, configured to implement the intelligent network control method for low-latency video backhauling according to any one of claims 1 to 4, characterized in that The network control method comprises: a network state detection module for real-time collection of network key indicators to provide data support for transmission strategy adjustment, the network key indicators including bandwidth, delay, jitter and packet loss rate of network links; a video content analysis module for extracting motion vectors between video frames through a lightweight CNN model, judging the activity level of video content, and outputting content feature evaluation results; an adaptive coding control module for receiving the network key indicators and content feature evaluation results, dynamically adjusting video encoding parameters, and achieving a balance between video quality and network adaptability; an intelligent scheduling decision module for allocating network bandwidth resources according to the priority and resource occupation of multiple video streams, and controlling the transmission parameters of each video stream; a feedback adjustment mechanism module for periodically collecting module running data, evaluating system performance, and updating control strategies; an FEC anti-packet loss module for intelligently identifying packet loss types according to real-time network conditions, dynamically adjusting FEC redundancy parameters, and adding redundant data in network transmission to combat packet loss problems.

6. A video conferencing device, characterized by The network control method comprises: an encoder device deployed at the video acquisition end for image acquisition and preprocessing; a decoder device deployed at the video receiving end for network data decoding and image display; The intelligent network control system for low-latency video backhaul according to claim 5 is used to connect the encoder device and the decoder device to realize low-latency backhaul control of video streams.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by a processor to realize the steps of the intelligent network control method for low-latency video backhaul according to any one of claims 1-4.

Images