A dynamic video remote monitoring management method

Abstract

The application discloses a kind of dynamic video remote monitoring management methods, it is related to image processing technical field, including: establishing monitoring edge device history transmission database, the historical resource utilization state of each monitoring edge device is analyzed, the resource demand of each edge device is predicted;Determine the maximum carrying capacity of the resource of each edge device, the resource load state of each edge device is calculated;Based on the resource demand of each edge device, the resource load state of each edge device and the video stream of historical transmission data, initialize the video transmission code rate of each edge device;Real-time transmission data of each video sensor is obtained, according to the resource demand and resource load state of each edge device, calculate video stream processing cost index;Based on video transmission code rate and video stream processing cost index, update the video transmission code rate of each edge device;The application has the advantages that: optimize video transmission quality, improve system monitoring overall performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a dynamic video remote monitoring management method. BACKGROUND

[0002] Dynamic video remote monitoring management refers to transmitting real-time video monitoring pictures to a platform through network technology for centralized management, analysis and storage, thereby realizing safety monitoring of different regions and places.

[0003] Since the video sensor edge device is difficult to update and iterate hardware and software resources in a large area after successful deployment, the old video sensor edge device is limited by the body resources when facing scenes with large dynamic traffic pictures, resulting in transmission delay, poor real-time performance of the video, and is not conducive to remote monitoring management. SUMMARY

[0004] To solve the above technical problems, a dynamic video remote monitoring management method is provided, which solves the above problems that the video sensor edge device is difficult to update and iterate hardware and software resources in a large area after successful deployment, the old video sensor edge device is limited by the body resources when facing scenes with large dynamic traffic pictures, resulting in transmission delay, poor real-time performance of the video, and is not conducive to remote monitoring management.

[0005] To achieve the above purposes, the technical scheme adopted by the present application is:

[0006] A dynamic video remote monitoring management method, comprising:

[0007] Based on the background API interface of a plurality of video sensors of the monitoring target area, the historical background API interface transmission data of each video sensor is obtained, and a monitoring edge device historical transmission database is established;

[0008] According to the monitoring edge device historical transmission database, the historical resource utilization state of each monitoring edge device is analyzed according to time sequence prediction, and the resource demand of each edge device is predicted;

[0009] The construction parameters of each edge device are obtained, the maximum resource carrying capacity of each edge device is determined, and the resource load state of each edge device under the resource demand of each edge device is calculated;

[0010] Based on the resource demand of each edge device, the resource load state of each edge device and the historical transmission data of the video stream, the video transmission code rate of each edge device is initialized;

[0011] Acquiring real-time transmission data of each video sensor, calculating a video stream processing cost index of the real-time transmission data according to resource requirements of each edge device and a resource load state of each edge device;

[0012] Based on the video transmission code rate of each edge device and the video stream processing cost index of the real-time transmission data, the video transmission code rate of each edge device is updated;

[0013] The video transmission code rate of each edge device is updated, and the video transmission code rate of each edge device is updated.

[0014]

[0015] In the formula, The video transmission code rate of the kth edge device in the tth unit time is updated. k The video transmission code rate of the kth edge device in the tth unit time is updated. t The video transmission code rate of the kth edge device in the tth unit time is updated. The video transmission code rate of the kth edge device in the tth unit time is updated. The video stream processing cost index of the real-time transmission data of the kth edge device in the tth unit time is updated.

[0016] Preferably, according to the historical transmission database of the monitoring edge device, the historical resource utilization state of each monitoring edge device is analyzed according to time sequence prediction, and the resource requirements of each edge device are predicted, which specifically includes:

[0017] Based on ARIMA time sequence prediction, an edge device resource utilization state analysis model is established.

[0018] Based on the resource occupation value of the historical transmission data, an observation window is established according to unit time, the resource occupation value of the historical transmission data is taken as an observation object, and through a sliding window, historical transmission resource occupation time sequence characteristic data of the monitoring edge device is obtained; The resource occupation includes: resource occupation CPU occupation rate, memory occupation rate, network bandwidth occupation rate;

[0019] The historical transmission resource occupation time sequence characteristic data of the monitoring edge device is preprocessed, and the autocorrelation function and the partial autocorrelation function are used to determine the autoregressive order and the moving average order of the edge device resource utilization state analysis model;

[0020] Based on the historical transmission resource occupation time sequence characteristic data of the monitoring edge device, the edge device resource utilization state analysis model is substituted, the autoregressive order and the moving average order are taken as influence factors, and the resource requirements of each edge device are generated;

[0021] The edge device resource utilization state analysis model is specifically as follows:

[0022]

[0023] wherein, is the resource requirement of the edge device, k is the resource requirement of the edge device, t is the resource requirement of the edge device, is the resource requirement of the edge device, k is the resource requirement of the edge device, is the resource requirement of the edge device, is a constant term, is an autoregressive coefficient, i is an autoregressive coefficient, is a moving average coefficient, j is a moving average coefficient, is an error term, is an error term, is a white noise error term, is the total number of autoregressive terms, is the total number of moving average terms.

[0024] Preferably, initializing the video transmission bit rate of each edge device based on the resource requirement of each edge device, the resource load state of each edge device and the historical transmission data of the video stream includes:

[0025] According to the video frame rate, the video stream of the historical transmission data is divided to obtain the monitoring video frame image data of each edge device;

[0026] The monitoring video frame image data of each edge device is subjected to grayscale conversion, and the absolute difference matrix of pixels between images is generated by comparing the grayscale value difference of adjacent frame images pixel by pixel per unit time;

[0027] Based on the grayscale threshold segmentation algorithm, the dynamic pixels in the absolute difference matrix of pixels between images are segmented with a preset grayscale threshold to determine the total number of dynamic region pixels per unit time of the edge device;

[0028] Based on the total number of motion region pixels in the absolute difference matrix of pixels between images, the total number of static region pixels per unit time of the edge device;

[0029] Based on the total number of dynamic region pixels per unit time of the edge device, the total number of static region pixels per unit time of the edge device and the resource requirement of each edge device, the video transmission bit rate of each edge device is initialized;

[0030] The initializing the video transmission bit rate of each edge device specifically includes:

[0031]

[0032] wherein, a video transmission code rate of the kth edge device in the tth unit time, a total number of dynamic region pixels of the kth edge device in the tth unit time, a total number of static region pixels of the kth edge device in the tth unit time, a resource influence weight, a content complexity weight;

[0033] Preferably, real-time transmission data of each video sensor is acquired, and a video stream processing cost index of the real-time transmission data is calculated according to resource requirements of each edge device and resource load states of each edge device, and the video stream processing cost index specifically comprises:

[0034] Based on video stream feature parameters of the real-time transmission data of each video sensor, resource requirements of each edge device in a unit time are determined;

[0035] Based on resource load state indexes of each edge device in a unit time and resource requirements of each edge device in a next unit time, the video stream processing cost index of the real-time transmission data is calculated and normalized;

[0036] The calculation of the video stream processing cost index of the real-time transmission data specifically comprises:

[0037]

[0038] In the formula, a video stream processing cost index of real-time transmission data of the kth edge device in the tth unit time, a resource occupation rate weight coefficient, a maximum bearing capacity of the kth edge device, a resource requirement of the kth edge device in the tth unit time, a resource load state index of the kth edge device in the tth unit time, a resource cost coefficient, cpu is a computing resource, bw is a bandwidth resource, and mem is a memory resource.

[0039] Compared with the prior art, the application has the beneficial effects that:

[0040] The application provides a dynamic video remote monitoring management scheme, which dynamically adjusts a video transmission code rate by establishing a historical transmission database of monitoring edge devices, combining time sequence prediction to analyze resource requirements and load states of each device, and calculating a video stream processing cost according to bearing capacity and load states of device resources after real-time acquisition of video sensor data, and updating the video transmission code rate. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 A dynamic video remote monitoring management method flow chart is shown in the figure;

[0042] Figure 2 A method flow chart for predicting resource requirements of each edge device is shown in the figure;

[0043] Figure 3 A method flow chart for initializing video transmission code rates of each edge device is shown in the figure;

[0044] Figure 4 A method flow chart for calculating video stream processing cost indicators of real-time transmission data is shown in the figure. DETAILED DESCRIPTION

[0045] The following description is provided to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only examples, and other obvious modifications can be made by those skilled in the art.

[0046] Referring to Figure 1 A dynamic video remote monitoring management method is shown in the figure, which comprises:

[0047] Based on the background API interface of a plurality of video sensors of the monitoring target area, the historical background API interface transmission data of each video sensor is obtained, and a monitoring edge device historical transmission database is established;

[0048] According to the monitoring edge device historical transmission database, the historical resource utilization state of each monitoring edge device is analyzed according to time sequence prediction, and the resource requirements of each edge device are predicted;

[0049] The construction parameters of each edge device are obtained, the maximum resource carrying capacity of each edge device is determined, and the resource load state of each edge device under the resource requirements of each edge device is calculated;

[0050] Based on the resource requirements of each edge device, the resource load state of each edge device and the historical transmission data of the video stream, the video transmission code rate of each edge device is initialized;

[0051] The real-time transmission data of each video sensor is obtained, and the video stream processing cost indicators of real-time transmission data are calculated according to the resource requirements of each edge device and the resource load state of each edge device;

[0052] Based on the initialization of the video transmission code rate of each edge device and the video stream processing cost indicators of real-time transmission data, the video transmission code rate of each edge device is updated;

[0053] The video transmission code rate of each edge device is updated, which is specifically:

[0054]

[0055] In the formula, To update the number k The first edge device t Video transmission bitrate per unit time Let be the video transmission bitrate of the k-th edge device in the t-th unit of time. This represents the video stream processing cost metric for real-time data transmission from the k-th edge device in the t-th unit of time.

[0056] This solution establishes a historical transmission database for monitoring edge devices and combines it with time-series predictive analysis of the resource requirements and load status of each device to dynamically adjust the video transmission bitrate. After acquiring video sensor data in real time, it calculates the video stream processing cost based on the device's resource capacity and load status, and updates the video transmission bitrate accordingly. The advantages of this invention are: optimized video transmission quality and improved overall system monitoring performance.

[0057] Reference Figure 2 As shown, based on the historical transmission database of the monitoring edge devices, the historical resource utilization status of each monitoring edge device is analyzed according to time series prediction, and the resource requirements of each edge device are predicted, specifically including:

[0058] Based on ARIMA time series forecasting, an analysis model for the resource utilization status of edge devices is established.

[0059] Based on the resource usage values ​​of historical transmission data, an observation window is established according to unit time. The resource usage values ​​of historical transmission data are used as the observation objects. By using a sliding window, the time-series characteristic data of historical transmission resource usage of the monitoring edge device is obtained. The resource usage includes: CPU usage rate, memory usage rate, and network bandwidth usage rate.

[0060] A stationary preprocessing method is used to perform time-series characteristic data of historical transmission resource occupancy of monitoring edge devices. The autoregression order and moving average order of the edge device resource utilization status analysis model are determined by using autocorrelation function and partial autocorrelation function.

[0061] Based on the historical transmission resource occupancy time-series characteristic data of monitoring edge devices, the resource utilization status analysis model of edge devices is substituted into the autoregression order and the moving average order as influencing factors to generate the resource requirements of each edge device.

[0062] Specifically, the edge device resource utilization status analysis model is as follows:

[0063]

[0064] In the formula, For the firstk resource requirement of the edge device, t resource requirement of the edge device, resource requirement of the edge device, k resource requirement of the edge device, resource occupation time series feature data of the edge device, constant term, autoregressive coefficient of the edge device, i autoregressive coefficient of the edge device, moving average coefficient of the edge device, j moving average coefficient of the edge device, moving average coefficient of the edge device, error term of the edge device, white noise error term of the edge device, total number of autoregressive, total number of moving average.

[0065] In use, in combination with the contents in the above steps:

[0066] As further content, based on the ARIMA time series prediction, the resource utilization state analysis model of the edge device is established, the resource occupation value (CPU, memory and network bandwidth occupation rate) of the historical transmission data is used to construct an observation window and obtain time series feature data, after the autoregressive and moving average order is determined through stationary preprocessing, the model is substituted to generate the resource requirement of each edge device, and the device resource utilization law is realized. The reference variable is provided for subsequent resource allocation, so as to improve the resource utilization rate of the edge device.

[0067] Further, the following implementation examples are proposed for the above steps:

[0068] An edge device A of an intelligent park entrance is responsible for processing video streams from three high-definition cameras.

[0069] 1. Establishing an observation window and obtaining time series feature data:

[0070] From the historical transmission database of the monitoring edge device, the historical data of device A in the past 30 days is extracted.

[0071] Taking every 5 minutes as a unit of time, an observation window is established. In each window, we record the average resource occupation value in this time period, including: CPU occupation rate (%), memory occupation rate (%), network bandwidth occupation rate (Mbps).

[0072] Through the sliding window, we get the historical transmission resource occupation time series feature data of device A for 30 days with 5-minute interval sampling.

[0073] 2. Stationary preprocessing and model order determination:

[0074] Firstly, we perform stationarity test on the time series data and find that it has a clear daily cycle pattern (high during the day and low at night), which is a non-stationary series. Therefore, we perform first-order difference and periodic difference (period 288, i.e. 24 hours / 5 minutes) stationarity preprocessing on it to obtain a stationary time series.

[0075] Then, we calculate the autocorrelation function and partial autocorrelation function of the preprocessed sequence. By observing the tailing and trailing characteristics of the ACF and PACF graphs, we determine the ARIMA model for predicting CPU occupancy, with autoregressive order p=2 and moving average order q=1.

[0076] 3、Model training and resource demand generation:

[0077] We put the preprocessed time series data into the ARIMA(2,1,1) model for training and establish the edge device resource utilization state analysis model.

[0078] After the model training is completed, we predict the CPU occupancy for the next 4 time units (i.e. the next 20 minutes). With the autoregressive order (p=2) and moving average order (q=1) as the core influencing factors inside the model, the model generates the predicted CPU occupancy of device A at each 5-minute time point in the next 20 minutes, for example: [68%, 72%, 75%, 74%]. This predicted time series is the specific resource demand of the edge device in the future period of time.

[0079] Referring to Figure 3 , the video transmission bit rate of each edge device is initialized based on the resource demand of each edge device, the resource load state of each edge device, and the historical transmission data video stream, which specifically includes:

[0080] According to the video frame rate, the historical transmission data video stream is divided to obtain the monitoring video frame image data of each edge device;

[0081] The monitoring video frame image data of each edge device is converted to grayscale, and the absolute difference matrix between adjacent frames is generated by comparing the grayscale value difference between adjacent frames pixel by pixel per unit time;

[0082] Based on the grayscale threshold segmentation algorithm, the dynamic pixels in the absolute difference matrix between adjacent frames are segmented by a preset grayscale threshold to determine the total number of dynamic region pixels per unit time of the edge device;

[0083] Based on the total number of motion region pixels in the absolute difference matrix between adjacent frames, the total number of static region pixels per unit time of the edge device is determined;

[0084] based on the total number of pixels in the dynamic area of the edge device per unit time, the total number of pixels in the static area of the edge device per unit time, and the resource requirements of each edge device, initialize the video transmission code rate of each edge device;

[0085] The initialization of the video transmission code rate of each edge device specifically includes:

[0086]

[0087] wherein, is the video transmission code rate of the kth edge device per unit time t, is the total number of pixels in the dynamic area of the kth edge device per unit time t, is the total number of pixels in the static area of the kth edge device per unit time t, is the resource impact weight, is the content complexity weight.

[0088] It should be noted that, is the resource impact weight, is the content complexity weight obtained according to the Q-learning algorithm.

[0089] It can be understood that, The resource impact weight reflects the reliability of the network transmission capability, and its value is dynamically adjusted according to historical bandwidth fluctuation rate, packet loss rate and other network indicators through reinforcement learning. For example, when the network stability decreases, it will be automatically reduced to weaken the impact of unreliable bandwidth prediction.

[0090] The content complexity weight represents the weight of the motion complexity of the video content, and is used to quantify the priority of the demand for code rate by the dynamic area. Its adjustment is based on the real-time feedback of the video analysis task (such as target tracking accuracy, motion blur degree). When the scene motion is intense, it will be automatically increased to improve the encoding quality of the key area.

[0091] In use, the contents in the above steps are combined:

[0092] As a further content, by inter-frame difference and gray threshold segmentation technology, the video content is quantitatively analyzed into dynamic and static pixel regions, so as to accurately evaluate the scene complexity; then, the complexity is combined with the predicted edge device resource demand to serve as a decision basis for initializing the most matched video transmission code rate for different devices and scenes. The refinement and intelligentization of code rate allocation are realized, which can significantly optimize the utilization efficiency of network bandwidth and storage resources while ensuring the quality of key dynamic pictures; by matching the code rate with the actual load capacity of the device, the risk of resource overload is avoided from the system entrance, thereby greatly improving the stability and resource utilization of the entire monitoring system;

[0093] Further, the following implementation examples are proposed for the above steps:

[0094] The video transmission code rate of the edge device A at the entrance of the smart park is initialized, and it is predicted from the S2 step that its CPU resource demand will be high in the future period of time.

[0095] 1. Divide the video frame image data:

[0096] From the historical transmission database of the device A, a representative video stream in the recent period of time is intercepted.

[0097] According to the video frame rate (such as 25 frames per second), the video stream is divided to obtain a series of continuous monitoring video frame image data.

[0098] 2. Generate the inter-pixel absolute difference matrix:

[0099] For the obtained frame image data, first, the gray scale conversion is performed to convert the color image into a gray scale image to simplify the calculation.

[0100] According to the unit time (for example, every adjacent two frames as a group), the gray scale values of adjacent frame images are compared pixel by pixel, and the absolute value of the difference is calculated.

[0101] The calculation process generates an inter-pixel absolute difference matrix with the same size as the image, and the value at each position in the matrix represents the change degree of the corresponding pixel point between two frames.

[0102] 3. Determine the total number of dynamic region pixels:

[0103] Based on the gray scale threshold segmentation algorithm, a preset gray scale threshold (for example, the gray scale difference is 15) is set.

[0104] The threshold is used to segment the inter-pixel absolute difference matrix, and the pixels with a value greater than 15 in the matrix are determined as dynamic pixels. The collection of all these dynamic pixels constitutes the motion region.

[0105] The number of these dynamic pixels is counted, and the total number of dynamic area pixels per unit time of the edge device is determined to be 150,000 pixels. At the same time, we subtract the number of dynamic pixels from the total number of pixels to obtain the total number of static area pixels, which is 1,850,000 pixels (assuming the total resolution is 2,000,000 pixels).

[0106] 4. Initialize the video transmission code rate:

[0107] Based on the calculated total number of dynamic area pixels (150,000), the total number of static area pixels (1,850,000), and the resource requirements of each edge device obtained from step S2 (here, high CPU requirement), we make a comprehensive decision.

[0108] The dynamic pixel ratio is calculated (15 / 200 = 7.5%), which belongs to a medium dynamic scene. However, considering the high resource requirement of the device, in order to avoid placing too heavy a coding burden on the already strained CPU, we decide not to use the highest code rate.

[0109] Therefore, the system initializes the video transmission code rate of device A to a medium-high value, such as 2048 Kbps. This value can better describe the 7.5% motion area, and will not cause excessive pressure on the device during periods of high resource demand.

[0110] Referring to Figure 4 , real-time transmission data of each video sensor is obtained, and based on the resource requirements of each edge device and the resource load state of each edge device, a video stream processing cost index of real-time transmission data is calculated, which specifically includes:

[0111] Based on the video stream feature parameters of the real-time transmission data of each video sensor, the resource requirements of each edge device per unit time are determined;

[0112] Based on the resource load state index of each edge device per unit time and the resource requirements of each edge device per unit time, the video stream processing cost index of real-time transmission data is calculated and normalized;

[0113] Wherein, the calculation of the video stream processing cost index of real-time transmission data is specifically:

[0114]

[0115] In the formula, is the video stream processing cost index of real-time transmission data of the kth edge device at the tth unit time, is the resource occupancy rate weight coefficient, is the maximum resource carrying capacity of the kth edge device, is the resource requirement of the kth edge device at the tth unit time, Resource load state index of the kth edge device in the tth unit time, Resource cost coefficient, cpu is a computing resource, bw is a bandwidth resource, and mem is a memory resource.

[0116] It should be noted that the resource cost is a multi-dimensional coefficient, including but not limited to: when calculating the cost coefficient of the CPU, the usage rate and usage duration of the CPU are obtained for calculation. Assuming that the usage cost of the CPU is per core per hour, the average number of cores of the CPU during video stream processing is , and the processing duration is t hours, then the CPU cost is Total cost = × × t, and the resource cost coefficients of other resources can be implemented according to the above scheme.

[0117] The above shows and describes the basic principles, main features and advantages of the present application. It should be understood by those skilled in the art that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A dynamic video remote monitoring management method, characterized in that, The method comprises the following steps: Based on the background API interface of the video sensor monitoring target area, the historical data of each video sensor background API interface is obtained, and a monitoring edge device historical transmission database is established; According to the monitoring edge device historical transmission database, the historical resource utilization state of each monitoring edge device is analyzed according to time sequence prediction, and the resource demand of each edge device is predicted; Obtain the construction parameters of each edge device, determine the maximum resource carrying capacity of each edge device, and calculate the resource load state of each edge device under the resource demand of each edge device; Based on the resource demand of each edge device, the resource load state of each edge device and the video stream of historical transmission data, the video transmission code rate of each edge device is initialized; According to the resource demand of each edge device and the resource load state of each edge device, the video stream processing cost index of real-time transmission data is calculated; Based on the initialization of the video transmission code rate of each edge device and the video stream processing cost index of real-time transmission data, the video transmission code rate of each edge device is updated; The video transmission code rate of each edge device is updated, which comprises: ； In the formula, is the video transmission code rate of the kth edge device in the tth unit time, k is the video transmission code rate of the kth edge device in the tth unit time, t is the video transmission code rate of the kth edge device in the tth unit time, is the video transmission code rate of the kth edge device in the tth unit time, is the video stream processing cost index of the real-time transmission data of the kth edge device in the tth unit time.

2. The dynamic video remote monitoring management method according to claim 1, wherein, According to the monitoring edge device historical transmission database, the historical resource utilization state of each monitoring edge device is analyzed according to time sequence prediction, and the resource demand of each edge device is predicted, which comprises: Based on ARIMA time sequence prediction, an edge device resource utilization state analysis model is established; Based on the resource occupation value of historical transmission data, an observation window is established according to unit time, the resource occupation value of historical transmission data is taken as the observation object, and the monitoring edge device historical transmission resource occupation time series characteristic data is obtained through sliding window; The resource occupation includes: CPU occupation rate, memory occupation rate and network bandwidth occupation rate; The monitoring edge device historical transmission resource occupation time series characteristic data is preprocessed, the autocorrelation function and the partial autocorrelation function are used to determine the autoregressive order and the moving average order of the edge device resource utilization state analysis model; Based on the monitoring edge device historical transmission resource occupation time series characteristic data, the edge device resource utilization state analysis model is substituted, the autoregressive order and the moving average order are taken as the influence factor, and the resource demand of each edge device is generated.

3. The dynamic video remote monitoring management method according to claim 2, wherein, The edge device resource utilization state analysis model comprises: ； In the formula, is the resource requirement of the nth edge device in the mth unit time, k is the resource requirement of the nth edge device in the mth unit time, t is the resource requirement of the nth edge device in the mth unit time, is the resource requirement of the nth edge device in the mth unit time, k is the resource requirement of the nth edge device in the mth unit time, is the resource requirement of the nth edge device in the mth unit time, is the constant term, is the constant term, i is the nth autoregressive coefficient, is the nth autoregressive coefficient, j is the nth moving average coefficient, is the nth moving average coefficient, is the error term in the mth unit time, is the white noise error term in the unit time, is the total number of autoregressive terms, is the total number of moving average terms.

4. The dynamic video remote monitoring management method according to claim 3, characterized in that, According to the resource demand of each edge device, the resource load state of each edge device and the video stream of historical transmission data, the video transmission code rate of each edge device is initialized, which comprises: According to the video frame rate, the video stream of historical transmission data is divided to obtain the monitoring video frame image data of each edge device; The gray scale of each edge device monitoring video frame image data is converted, and the gray scale value difference of adjacent frame images is compared pixel by pixel according to unit time to generate an image pixel absolute difference matrix; Based on the gray threshold segmentation algorithm, the dynamic pixels in the image pixel absolute difference matrix are segmented with a preset gray threshold to determine the total number of dynamic area pixels per unit time of the edge device; Based on the total number of motion region pixels in the inter-image pixel absolute difference value matrix, the total number of static region pixels per unit time of the edge device; Based on the total number of dynamic region pixels per unit time of the edge device, the total number of static region pixels per unit time of the edge device, and the resource requirements of each edge device, the video transmission code rate of each edge device is initialized.

5. The dynamic video remote monitoring management method according to claim 4, characterized in that, The initialization of the video transmission code rate of each edge device specifically includes: ； In the formula, is the video transmission code rate of the kth edge device in the tth unit time, is the total number of dynamic region pixels of the kth edge device in the tth unit time, is the total number of static region pixels of the kth edge device in the tth unit time, is the resource influence weight, is the content complexity weight.

6. The dynamic video remote monitoring management method according to claim 5, wherein, Obtaining real-time transmission data of each video sensor, and calculating a video stream processing cost index of the real-time transmission data according to the resource requirements of each edge device and the resource load state of each edge device specifically includes: Based on the video stream feature parameters of the real-time transmission data of each video sensor, the resource requirements of each edge device per unit time are determined; Based on the resource load state index of each edge device per unit time and the resource requirements of each edge device per unit time, the video stream processing cost index of the real-time transmission data is calculated and normalized.

7. The dynamic video remote monitoring management method according to claim 6, wherein, The calculation of the video stream processing cost index of the real-time transmission data specifically includes: ； In the formula, is a video stream processing cost index of real-time transmission data of the kth edge device in the tth unit time, is a resource occupation rate weight coefficient, is a maximum resource carrying capacity of the kth edge device, is a resource demand of the kth edge device in the tth unit time, is a resource load state index of the kth edge device in the tth unit time, is a resource cost coefficient, cpu is a computing resource, bw is a bandwidth resource, and mem is a memory resource.

Images