Adaptive method and device for monitoring video transmission parameters

Abstract

The invention relates to the technical field of video transmission, in particular to an adaptive method and device for monitoring video transmission parameters, and the method comprises the steps: calculating the filling acceleration representing the transmission congestion trend, and the frequency gravity center representing the image complexity; taking a quantization parameter of a to-be-coded frame as a particle, and constructing a fitness function; initializing a particle swarm algorithm and carrying out iteration, calculating the vector divergence and the divergence change rate of the particle swarm in real time in the iteration process, terminating when the divergence change rate is smaller than a preset stable threshold value or the number of iterations reaches the maximum number, and outputting a globally optimal solution and the final vector divergence; and determining video coding parameters of different areas in the to-be-coded frame according to the globally optimal solution, the final vector divergence and the frequency gravity center, and executing video coding transmission. According to the technical scheme, under the condition that the bandwidth is limited, the definition of the region of interest is guaranteed, and video jamming is prevented.

Description

Technical Field

[0001] This application relates to the field of video transmission technology, and in particular to an adaptive method and apparatus for monitoring video transmission parameters. Background Technology

[0002] Communication command vehicles have become core equipment in scenarios such as forest fire prevention and geological disaster rescue. These vehicles are typically equipped with high-definition pan-tilt cameras and video encoding and transmission systems, responsible for transmitting high-definition video from the scene back to the command center in real time. Because the environment at the scene of an emergency is often complex, command vehicles may be in a state of rapid movement or parked in areas with severe signal obstruction, causing the bandwidth, packet loss rate, and jitter of the transmission link to exhibit time-varying and non-linear fluctuations. Therefore, how to ensure smooth video transmission while improving video quality under extremely unstable channel conditions is a major challenge currently facing video transmission technology. Summary of the Invention

[0003] To address the technical challenge of improving video quality while ensuring smooth video transmission, this application provides an adaptive method and apparatus for monitoring video transmission parameters. This method and apparatus can guarantee the clarity of the region of interest and prevent video stuttering under conditions of extremely unstable channel conditions, thus ensuring smooth video transmission.

[0004] In a first aspect, this application provides an adaptive method for monitoring video transmission parameters. The adaptive method includes: acquiring a state sequence of an encoding buffer and frequency domain coefficients in the spectrogram corresponding to at least one frame to be encoded; calculating a filling acceleration representing a transmission congestion trend based on the state sequence of the encoding buffer, and calculating a frequency centroid representing image complexity based on the frequency domain coefficients; using the quantization parameters of the frame to be encoded as particles and constructing a fitness function, wherein the fitness function is the product of video quality plus a penalty weight and a transmission cost, the video quality being negatively correlated with the quantization parameters, the transmission cost being positively correlated with the transmission bitrate corresponding to the quantization parameters, and the penalty weight being dynamically adjusted based on the filling acceleration; initializing a particle swarm optimization algorithm and iterating thereafter, calculating the vector divergence and the rate of change of divergence of the particle swarm in real time during the iteration process, terminating when the rate of change of divergence is less than a preset stability threshold, or when the number of iterations reaches the maximum number, and outputting the global optimal solution and the final vector divergence; determining video encoding parameters for different regions in the frame to be encoded based on the global optimal solution, the final vector divergence, and the frequency centroid, and performing video encoding transmission using the video encoding parameters.

[0005] By using the quantization parameters of the frames to be encoded as particles and constructing a fitness function based on video quality, transmission cost, and a penalty weight dynamically adjusted by padding acceleration, the vector divergence of the particle swarm is monitored in real time during the iteration process. The convergence state of the algorithm is determined based on the final vector divergence, and then the video coding parameters for different regions are determined by combining the frequency centroid. This achieves a dynamic balance between video quality and transmission bandwidth, ensuring that the output can quickly balance the image quality of key areas and the overall smoothness when the network fluctuates.

[0006] Preferably, the step of calculating the filling acceleration characterizing the transmission congestion trend based on the state sequence of the coding buffer includes: obtaining the water level value of the coding buffer at each sampling time to obtain a state sequence; and calculating the second derivative at any sampling time in the state sequence, wherein the second derivative corresponds to the filling acceleration at the sampling time.

[0007] Accurately capturing the rapid upward trend of the nonlinearity of the coding buffer level can identify signs of worsening transmission congestion before the buffer overflows, thus providing a predictive quantitative indicator for the dynamic adjustment of subsequent penalty weights and improving the response speed of congestion control.

[0008] Preferably, the step of calculating the frequency centroid representing the complexity of the image based on the frequency domain coefficients includes: using the distance from the coordinate point corresponding to the frequency domain coefficient to the origin of the spectrum graph as a weight to sum the frequency domain coefficients to obtain the high-frequency energy; calculating the sum of squares of all frequency domain coefficients to obtain the total spectrum energy; and dividing the sum of the high-frequency energy by the total spectrum energy to obtain the normalized frequency centroid.

[0009] The frequency domain quantitatively characterizes the richness of image texture and coding complexity; a higher frequency centroid indicates more image details and greater compression difficulty, which provides data support for subsequent judgment of the degree of image quality sacrifice in the background area.

[0010] Preferably, the penalty weight is dynamically adjusted according to the filling acceleration, which includes: setting a base penalty weight and an acceleration threshold; in response to the filling acceleration not being greater than the acceleration threshold, the penalty weight is the base penalty weight, otherwise, the penalty weight is positively correlated with the filling acceleration.

[0011] When the link is in good condition, the basic weight is maintained to allow for high image quality, while when a congestion trend is detected, the penalty weight is increased exponentially, forcing the particle swarm algorithm to converge rapidly toward a lower bit rate, thereby proactively reducing the transmission load before congestion occurs and avoiding data backlog.

[0012] Preferably, the transmission cost is calculated as follows: based on the current quantization parameters of the particle, a preset code rate prediction model is used to calculate the predicted code rate; the predicted code rate is compared with the current channel's detection bandwidth; if the predicted code rate is less than or equal to the detection bandwidth, the transmission cost is set to zero; if the predicted code rate is greater than the detection bandwidth, the difference between the predicted code rate and the detection bandwidth is calculated, and the transmission cost is positively correlated with the difference.

[0013] This ensures that the transmission cost in the fitness function only plays a constraining role when bandwidth resources are insufficient, and ensures that video quality is prioritized when bandwidth is sufficient, thereby maximizing bandwidth utilization.

[0014] Preferably, the video quality is calculated as follows: based on the quantization parameters of the particles, a preset video quality model is used to calculate the video quality, wherein the video quality model is a decreasing function of the quantization parameters.

[0015] Preferably, the vector divergence is calculated by: calculating the variance of the optimal position of each particle in the particle swarm, and normalizing the variance to obtain the vector divergence.

[0016] Vector divergence measures the degree of disagreement among individual experiences within a particle swarm, and can intuitively reflect the convergence performance of the particle swarm algorithm, providing a reliable basis for determining whether to switch to a regional differentiation coding strategy.

[0017] Preferably, determining the video coding parameters for different regions in each frame to be encoded based on the global optimal solution, the final vector divergence, and the frequency centroid includes: in response to the final vector divergence being less than a divergence threshold, using the quantization parameters corresponding to the global optimal solution as the video coding parameters for the frame to be encoded; in response to the final vector divergence being not less than the divergence threshold, dividing the frame to be encoded into a region of interest and a background region; using the quantization parameters corresponding to the global optimal solution as the video coding parameters for the region of interest; calculating the quantization parameter suppression increment using the final vector divergence and the frequency centroid; and adding the quantization parameters corresponding to the global optimal solution to the quantization parameter suppression increment as the video coding parameters for the background region.

[0018] In the absence of consensus among algorithms, a regional differential coding strategy is adopted. Priority is given to ensuring that the region of interest uses the globally optimal parameters, while additional compression suppression is applied to the background region. This ensures the clarity of the region of interest and effectively prevents video stuttering by suppressing the bitrate of the background region when bandwidth is limited or the link is unstable.

[0019] Preferably, the step of calculating the quantization parameter suppression increment using the final vector divergence and frequency centroid includes: calculating the product of the final vector divergence and the frequency centroid; multiplying the normalized product by a preset maximum quantization step size and then rounding down to obtain the quantization parameter suppression increment.

[0020] In a second aspect, this application also provides an adaptive device for monitoring video transmission parameters, including a processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement an adaptive method for monitoring video transmission parameters according to the first aspect of this application.

[0021] The technical solution of this application has the following beneficial technical effects: By acquiring the state sequence of the encoding buffer and the frequency domain coefficients of the frame to be encoded, the filling acceleration, which characterizes the transmission congestion trend, and the frequency centroid, which characterizes the image complexity, are calculated. A multi-dimensional optimization algorithm is then used, where the penalty weight of the fitness function is dynamically adjusted based on the filling acceleration, and the vector divergence of the particle swarm is monitored during iteration. In complex scenarios where the particle swarm algorithm struggles to converge, differentiated video encoding parameter configurations are applied to the region of interest (ROI) and the background region based on the global optimal solution and the frequency centroid. This achieves adaptive adjustment of transmission parameters, ensuring the clarity of the ROI under bandwidth-constrained or unstable link conditions while effectively preventing video stuttering by suppressing the background bitrate, thus improving the real-time performance and reliability of surveillance video transmission. Attached Figure Description

[0022] Figure 1 This is a flowchart of an adaptive method for monitoring video transmission parameters according to an embodiment of this application.

[0023] Figure 2 This is a schematic diagram illustrating the differential encoding of the region of interest and the background region in the frame to be encoded according to an embodiment of this application.

[0024] Figure 3 This is a structural block diagram of an adaptive device for monitoring video transmission parameters according to an embodiment of this application. Detailed Implementation

[0025] According to a first aspect of this application, this application provides an adaptive method for monitoring video transmission parameters. Figure 1 This is a flowchart of an adaptive method for monitoring video transmission parameters according to an embodiment of this application. Figure 1 As shown, the adaptive method for monitoring video transmission parameters includes steps S101 to S105, which are described in detail below.

[0026] S101, obtain the state sequence of the encoding buffer and the frequency domain coefficients in the spectrum diagram corresponding to at least one frame to be encoded.

[0027] In one embodiment, after the video transmission system is started, a control period is set at a group of images or a fixed time slice (e.g., 200ms). At the beginning of each period, real-time data is acquired from the system's underlying hardware interface. The acquired real-time data includes at least one frame to be encoded, and the video encoding parameters of the frame to be encoded are subsequently adaptively optimized.

[0028] The encoding buffer refers to the data queue (FIFO) connecting the video encoder and the wireless transmission module. The amount of data in the encoding buffer directly reflects the current load status of the communication link, and the state sequence refers to the buffer level value collected at each continuous sampling time.

[0029] Furthermore, the frequency domain coefficients of the frame to be encoded refer to the frequency domain coefficients at each coordinate point in the spectrum obtained after performing a discrete cosine transform on the frame to be encoded. These frequency domain coefficients reflect the energy distribution of the frame to be encoded in the frequency domain. It should be noted that each frame to be encoded has its own set of frequency domain coefficients.

[0030] S102, calculate the filling acceleration representing the transmission congestion trend based on the coding buffer state sequence, and calculate the frequency centroid representing the image complexity based on the frequency domain coefficients.

[0031] In one embodiment, the rapid movement of the command vehicle may cause the signal to attenuate rapidly, especially in complex areas such as forests and mountains. The water level change of the encoding buffer is often non-linear. When the water level of the encoding buffer rises rapidly, it indicates that the current communication link is showing a tendency to be congested.

[0032] Specifically, the step of calculating the filling acceleration characterizing the transmission congestion trend based on the state sequence of the coding buffer includes: obtaining the water level value of the coding buffer at each sampling time to obtain a state sequence; and calculating the second derivative at any sampling time in the state sequence, wherein the second derivative corresponds to the filling acceleration at the sampling time.

[0033] Get the current sampling time Encoding buffer water level value Simultaneously acquire sampling time and sampling time water level and Sampling time Filling acceleration Satisfying the relation: ; in, The sampling interval is, for example, set to 50ms. When A positive and large value indicates that the water level in the encoding buffer is rising rapidly. It should be noted that, to avoid the influence of noise on the filling acceleration, a moving average filter is applied to the encoding buffer state sequence before calculating the filling acceleration to remove noise from the state sequence.

[0034] Furthermore, for complex scenarios such as forest fires or population evacuations, the frames to be encoded contain a large amount of high-frequency information and are extremely sensitive to quantization parameters.

[0035] Specifically, for the frequency domain coefficients of the frame to be encoded ,in The frequency coordinates represent the frequency. The calculation of the frequency centroid representing the image complexity based on the frequency domain coefficients includes: using the distance from the coordinate point corresponding to the frequency domain coefficient to the origin of the spectrum graph as a weight to sum the frequency domain coefficients to obtain the high-frequency energy; calculating the sum of squares of all frequency domain coefficients to obtain the total spectral energy; and dividing the sum of the high-frequency energy by the total spectral energy to obtain the normalized frequency centroid.

[0036] The origin of the spectrum graph is... ,but corresponding coordinates The distance to the origin of the spectrum graph is the sum of the squares of the coordinate points. ; normalized frequency centroid The calculation formula is: ; in, To prevent positive numbers with a denominator of zero, the value can be 1; the frame to be encoded The larger the value, the more concentrated the energy of the frame to be encoded is in the high-frequency region, and the richer the details and information contained in the frame to be encoded. This provides a basis for subsequent judgment on whether to sacrifice background image quality.

[0037] S103, the quantization parameters of the frame to be encoded are used as particles, and a fitness function is constructed. The fitness function is the product of video quality, penalty weight, and transmission cost. The video quality is negatively correlated with the quantization parameters, the transmission cost is positively correlated with the transmission bitrate corresponding to the quantization parameters, and the penalty weight is dynamically adjusted according to the fill acceleration.

[0038] In one embodiment, the quantization parameters of each frame to be encoded are used as particles. The position of each particle in the particle swarm represents a set of candidate quantization parameters. If the number of frames to be encoded is 1, then a particle contains only one quantization parameter. If the number of frames to be encoded is 10, then a particle contains only 10 quantization parameters.

[0039] Construct a fitness function that aims to maximize video quality while minimizing transmission cost. The fitness function consists of video quality, penalty weights, and transmission cost, and is equal to the video quality plus the product of the penalty weights and transmission cost.

[0040] It should be noted that when a particle includes quantization parameters of multiple frames to be encoded, for any frame to be encoded, the product of video quality plus penalty weight and transmission cost can characterize the quality of transmitting the frame to be encoded using the quantization parameters of that frame. The sum of the quality of all frames to be encoded is used as the fitness function of the particle.

[0041] The quantization parameter (QP) is a key parameter in video coding, used to control the degree of detail loss during compression. A larger QP value results in stronger compression but lower image quality; a smaller QP value results in weaker compression but higher image quality. Its main function is to balance video quality and bitrate. Therefore, the video quality model is a decreasing function of the QP, and the video quality model is as follows: ;in, For the frames to be encoded in the particles Quantization parameters, and These are the first constant coefficient and the second constant coefficient, respectively. The value is 100. The value is 20.

[0042] The transmission cost is positively correlated with the transmission code rate corresponding to the quantization parameters. The transmission cost is calculated as follows: based on the current quantization parameters of the particle, the predicted code rate is calculated using a preset code rate prediction model; the predicted code rate is compared with the current channel's detection bandwidth; if the predicted code rate is less than or equal to the detection bandwidth, the transmission cost is set to zero; if the predicted code rate is greater than the detection bandwidth, the difference between the predicted code rate and the detection bandwidth is calculated, and the transmission cost is positively correlated with the difference.

[0043] Specifically, the probe bandwidth is the bandwidth available for data transmission on the current communication link, which can be calculated by statistically analyzing the total size of data packets successfully sent and acknowledged per unit time. For example, if 5Mb of data was successfully sent in the past 500ms, then the probe bandwidth is 10Mbps. The preset bit rate prediction model adopts a quadratic RQ model or... The model is used in this application without limitation.

[0044] When the prediction bit rate is greater than the probe bandwidth, it indicates that the current quantization parameters of the particles are causing backlog in the transmit buffer, increasing transmission cost. Satisfying the relation: ;in, For the frames to be encoded in the particles The predicted bitrate corresponding to the quantization parameters. To detect bandwidth, This is the cost coefficient, and its value is 1.

[0045] Furthermore, the penalty weight in the fitness function is dynamically adjusted based on the filling acceleration. Specifically, a base penalty weight and an acceleration threshold are set. In response to the filling acceleration not being greater than the acceleration threshold, the penalty weight is the base penalty weight; otherwise, the penalty weight is positively correlated with the filling acceleration.

[0046] In response to the filling acceleration being greater than the acceleration threshold, a penalty weight is applied. The formula is as follows: ; in, Based on the penalty weight, For acceleration threshold, For filling acceleration, Let be the penalty coefficient, when When the acceleration is greater than the acceleration threshold, Rapid growth forces the particle swarm to abandon its pursuit of high image quality and proactively reduce the bitrate before congestion occurs on the current communication link. The base penalty weight is 1, and the penalty coefficient is 5.

[0047] S104: Initialize the particle swarm algorithm and perform iterations. During the iteration process, calculate the vector divergence and the rate of change of divergence of the particle swarm in real time. Terminate when the rate of change of divergence is less than the preset stability threshold or when the number of iterations reaches the maximum number. Output the global optimal solution and the final vector divergence.

[0048] In one embodiment, the particle swarm size is set to 50, meaning there are 50 particles in the swarm. The particle swarm algorithm is used to update the position of each particle in the swarm. However, in complex scenarios where both transmission bandwidth and video quality must be maintained, the particle swarm algorithm may be in a state of not reaching a consensus, meaning that the search directions of each particle are relatively divergent. In this case, the particle swarm algorithm cannot converge and is forced to output intermediate results, resulting in the inability to output the global optimal solution.

[0049] During the iteration process, the vector divergence of the particle swarm is calculated in real time. The vector divergence is calculated as follows: the variance of the individual optimal position of each particle in the particle swarm is calculated, and the variance is normalized to obtain the vector divergence. The larger the variance of the individual optimal position of each particle, the more dispersed the distribution of the individual optimal positions. The vector divergence can quantify the divergence between the individual experiences of particles. The greater the divergence, the greater the divergence, indicating that the algorithm has not yet converged.

[0050] Furthermore, obtain the current iteration round. The absolute value of the difference between the vector divergence of the current iteration and the vector divergence of the previous iteration is used as the rate of change of divergence. rate of change of divergence Satisfying the relation: ; in, For the current iteration round vector divergence, This represents the vector divergence of the previous round.

[0051] The preset stability threshold is set to 0.01. When the rate of change of divergence is less than the preset stability threshold, it indicates that the opinions within the particle swarm have tended to be unified. At this point, continuing the iteration would only waste computing power and would not bring significant performance improvement, so the iteration is terminated. Alternatively, when the maximum number of iterations is reached, the iteration is forcibly terminated. After terminating the iteration, the global optimal solution and the final vector divergence are output.

[0052] S105: Determine the video coding parameters for different regions in the frame to be encoded based on the global optimal solution, the final vector divergence, and the frequency centroid, and then use the video coding parameters to perform video encoding and transmission.

[0053] In one embodiment, after the iteration ends, in response to the final vector divergence being less than the divergence threshold, the quantization parameters corresponding to the global optimal solution are used as the video coding parameters of the frame to be encoded; in response to the final vector divergence being not less than the divergence threshold, the frame to be encoded is divided into a region of interest and a background region; the quantization parameters corresponding to the global optimal solution are used as the video coding parameters of the region of interest; the quantization parameter suppression increment is calculated using the final vector divergence and the frequency centroid; the quantization parameters corresponding to the global optimal solution are added to the quantization parameter suppression increment to obtain the video coding parameters for the background region.

[0054] The divergence threshold is 0.05. Understandably, if the final vector divergence is not less than the divergence threshold, it indicates that the individual optimal solutions of each particle within the particle swarm cannot converge, meaning the bandwidth of the current communication link cannot transmit high-definition textures. In this case, the frame to be encoded is divided into a region of interest and a background region, and a spatially non-uniform coding strategy is used to calculate the quantization parameters of the region of interest and the background region respectively.

[0055] In this method, a lightweight target detection algorithm built into the vehicle terminal can be used to divide the frame to be encoded into a region of interest (ROI) and a background region. The target detection algorithm can be a YOLO series model, which identifies specific disaster features such as people, vehicles, flames, and smoke in the frame to be encoded. The rectangular area containing these identified features is marked as the ROI, and the remaining areas in the frame to be encoded are marked as the background region. In other optional embodiments, a difference map between the frame to be encoded and the previous adjacent video frame can be calculated to extract the moving foreground region, which is the ROI, and the remaining areas in the frame to be encoded are marked as the background region.

[0056] The step of calculating the quantization parameter suppression increment using the final vector divergence and frequency centroid includes: calculating the product of the final vector divergence and the frequency centroid; multiplying the normalized product by a preset maximum quantization step size and then rounding down to obtain the quantization parameter suppression increment.

[0057] Quantization parameters suppress increment Satisfying the relation: ; in, This is the preset maximum quantization step size coefficient, which can be 15. For adjustment coefficients, This is a floor function. The product of the final vector divergence and the frequency centroid. The larger the value, the more complex the image texture, indicating that the particle swarm optimization algorithm is in a state of not having reached a consensus, and the more complex the image texture, the more difficult compression becomes. To increase the required QP value, the quantization parameters corresponding to the global optimal solution are... The video encoding parameters for the region of interest are set as follows, while the encoding parameters for the background region are set to... By sacrificing the image quality of the background area, and given the limited bandwidth and difficulty in compression, this method not only preserves the clarity of the area of ​​interest but also prevents video stuttering by significantly compressing the background.

[0058] Please see Figure 2 This is a schematic diagram of differential encoding of the region of interest and the background region in a frame to be encoded according to an embodiment of this application. For any frame to be encoded, when the fill acceleration is 440, the frequency centroid is 2.81, and the vector divergence of the particle swarm algorithm is 0.1515, since the vector divergence is greater than the divergence threshold, differential encoding is performed on the frame to be encoded by region. The video encoding parameters of the region of interest are 28.11, and the video encoding parameters of the background region are 42.11, thus obtaining the video encoding parameters of the frame to be encoded.

[0059] After obtaining the video encoding parameters of all frames to be encoded, the video encoding parameters are used to encode the video, and the encoded video stream is input into the transmission system to be sent to the command center.

[0060] According to a second aspect of this application, this application also provides an adaptive device for monitoring video transmission parameters. Figure 3 This is a structural block diagram of an adaptive device for monitoring video transmission parameters according to an embodiment of this application. Figure 3 As shown, the device 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an adaptive method for monitoring video transmission parameters according to the first aspect of this application. The device also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0061] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. An adaptive method for monitoring video transmission parameters, characterized in that, The adaptive method includes: Obtain the state sequence of the encoding buffer and the frequency domain coefficients in the spectrogram of at least one frame to be encoded; The filling acceleration, which characterizes the transmission congestion trend, is calculated based on the state sequence of the encoded buffer, and the frequency centroid, which characterizes the image complexity, is calculated based on the frequency domain coefficients. The quantization parameters of the frame to be encoded are used as particles, and a fitness function is constructed. The fitness function is the product of video quality, penalty weight, and transmission cost. The video quality is negatively correlated with the quantization parameters, the transmission cost is positively correlated with the transmission bitrate corresponding to the quantization parameters, and the penalty weight is dynamically adjusted according to the fill acceleration. Initialize the particle swarm optimization algorithm and iterate it. During the iteration process, calculate the vector divergence and the rate of change of divergence of the particle swarm in real time. Terminate when the rate of change of divergence is less than the preset stability threshold or when the number of iterations reaches the maximum number. Output the global optimal solution and the final vector divergence. Based on the global optimal solution, the final vector divergence, and the frequency centroid, the video coding parameters for different regions in the frame to be encoded are determined, and the video coding parameters are used to perform video coding and transmission.

2. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The calculation of the filling acceleration characterizing the transmission congestion trend based on the encoded buffer state sequence includes: Obtain the water level value of the encoding buffer at each sampling time to obtain the state sequence; In the state sequence, the second derivative is calculated at any sampling time, and the second derivative corresponds to the filling acceleration at that sampling time.

3. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The calculation of the frequency centroid representing image complexity based on frequency domain coefficients includes: The high-frequency energy is obtained by using the distance from the coordinate point corresponding to the frequency domain coefficient to the origin of the spectrum graph as a weight to sum the frequency domain coefficients; the total spectral energy is obtained by calculating the sum of squares of all frequency domain coefficients; the normalized frequency centroid is obtained by dividing the sum of the high-frequency energy by the total spectral energy.

4. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The penalty weight is dynamically adjusted according to the filling acceleration, including setting a basic penalty weight and an acceleration threshold. In response to the filling acceleration not being greater than the acceleration threshold, the penalty weight is the base penalty weight; otherwise, the penalty weight is positively correlated with the filling acceleration.

5. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The transmission cost is calculated as follows: Based on the current quantization parameters of the particles, the predicted bit rate is calculated using a preset bit rate prediction model. The predicted code rate is compared with the current channel's probe bandwidth; If the predicted bit rate is less than or equal to the probe bandwidth, the transmission cost is set to zero. If the predicted bit rate is greater than the probe bandwidth, the difference between the predicted bit rate and the probe bandwidth is calculated, and the transmission cost is positively correlated with the difference.

6. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The video quality is calculated as follows: based on the particle quantization parameters, a preset video quality model is used to calculate the video quality, where the video quality model is a decreasing function of the quantization parameters.

7. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The vector divergence is calculated as follows: the variance of the optimal position of each particle in the particle swarm is calculated, and the variance is normalized to obtain the vector divergence.

8. The adaptive method for monitoring video transmission parameters according to claim 1, characterized in that, The process of determining video coding parameters for different regions in each frame to be encoded based on the global optimal solution, final vector divergence, and frequency centroid includes: In response to the final vector divergence being less than the divergence threshold, the quantization parameters corresponding to the global optimal solution are used as the video coding parameters of the frame to be encoded; in response to the final vector divergence being not less than the divergence threshold, the frame to be encoded is divided into a region of interest and a background region; the quantization parameters corresponding to the global optimal solution are used as the video coding parameters of the region of interest; the quantization parameter suppression increment is calculated using the final vector divergence and the frequency centroid; the quantization parameters corresponding to the global optimal solution are added to the quantization parameter suppression increment to obtain the video coding parameters for the background region.

9. The adaptive method for monitoring video transmission parameters according to claim 8, characterized in that, The step of using the final vector divergence and frequency centroid to calculate quantization parameters to suppress increments includes: Calculate the product of the final vector divergence and the frequency centroid; The normalized product is multiplied by the preset maximum quantization step size and then rounded down to obtain the quantization parameter suppression increment.

10. An adaptive device for monitoring video transmission parameters, characterized in that, It includes a processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement an adaptive method for monitoring video transmission parameters according to any one of claims 1 to 9.

Images