Intelligent early warning method based on fusion of 5G Internet of Things and video monitoring

Abstract

The invention belongs to the technical field of intelligent monitoring, and particularly relates to a 5G Internet of Things fused video monitoring intelligent early warning method, which comprises the following steps of: acquiring vehicle attribute information and environment perception data, constructing a 5G Internet of Things perception network, fusing a vehicle movement track and road condition environment data through a cross-modal attention mechanism, generating an enhanced environment perception graph, and simultaneously, carrying out early warning on the vehicle movement track and the road condition environment data. Optimizing the transmission efficiency by adopting a dynamic resolution adjustment algorithm; the method comprises the following steps: constructing a dynamic behavior prediction model by using a space-time diagram convolutional network, predicting a vehicle abnormal behavior probability and an evolution trajectory, calculating an early warning level through an adaptive risk quantification algorithm, simulating a risk diffusion coefficient by using a dynamic risk propagation model according to the road section vehicle density, the average vehicle speed and the road traffic capacity, and correcting early warning sensitivity parameters in real time. And transmitting to a command platform, performing situation deduction, triggering a grading early warning instruction, and generating thermodynamic diagram warning information. Therefore, the problems of insufficient positioning precision, weak analysis capability, poor transmission efficiency and the like in the prior art are solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent monitoring, and particularly relates to a 5G Internet of Things fusion video monitoring intelligent early warning method. BACKGROUND

[0002] With the accelerated promotion of new infrastructure construction, the market has an urgent demand for high-reliability and low-delay intelligent early warning systems based on 5G Internet of Things fusion, and requires cross-modal data fusion, dynamic transmission optimization, and multi-dimensional situation deduction capabilities. However, the existing technology cannot support accurate early warning in complex traffic scenarios because it fails to achieve deep fusion of perception networks, lacks a cross-modal collaborative reasoning framework and a dynamic risk quantification mechanism, and has a significant gap with the technical requirements of constructing an enhanced environmental perception map and real-time correction of early warning sensitivity.

[0003] However, traditional video monitoring early warning has key defects: the positioning accuracy of the region proposal network for irregular early warning areas is insufficient and is easily affected by light changes and obstructions; the model lacks an environmental robustness adjustment mechanism, and when facing fluctuations in traffic parameters such as vehicle density and average speed, the analysis results of different monitoring nodes have large consistency and bias; the data dimension is limited to two-dimensional visual information, and a deep fusion perception network has not been formed, lacking a cross-modal attention mechanism for collaborative processing of vehicle motion trajectories and road environment data, resulting in a data split state between video pictures and physical quantity parameters. With the accelerated promotion of new infrastructure construction, the market has an urgent demand for high-reliability and low-delay intelligent early warning systems, and the existing technology is difficult to support accurate early warning in complex scenarios due to insufficient transmission efficiency and limited feature expression capability. SUMMARY

[0004] The present application provides a 5G Internet of Things fusion video monitoring intelligent early warning method to solve the problems of insufficient positioning accuracy, weak analysis capability, and poor transmission efficiency in the prior art.

[0005] The first embodiment of the present application provides a 5G Internet of Things integrated video surveillance intelligent early warning method, including the following steps: obtaining vehicle attribute information and environmental perception data; constructing a 5G Internet of Things perception network based on the vehicle attribute information and environmental perception data, and based on the 5G Internet of Things perception network, fusing the vehicle motion trajectory and road environment data through a cross-modal attention mechanism to generate an enhanced environmental perception map, and at the same time, using a dynamic resolution adjustment algorithm to optimize transmission efficiency; based on the enhanced environmental perception map, constructing a dynamic behavior prediction model using a spatiotemporal graph convolutional network to predict the probability and evolution trajectory of vehicle abnormal behavior, and calculating the warning level through an adaptive risk quantification algorithm based on the abnormal behavior probability and evolution trajectory, and simulating the risk diffusion coefficient based on the current road section vehicle density, average speed and road capacity using a dynamic risk propagation model, and correcting the warning sensitivity parameters in real time, and transmitting the corrected sensitivity parameters to the command platform to perform multi-dimensional situation deduction, triggering graded warning instructions and generating heat map warning information containing threat diffusion paths.

[0006] Preferably, generating heat map warning information containing threat diffusion paths includes: constructing a dynamic threat diffusion model; based on the dynamic threat diffusion model, combining a streaming particle swarm algorithm to simulate the motion trajectory of threat particles in the monitoring area, mapping the spatial distribution of particle trajectories into threat diffusion probability values ​​through a kernel density estimation algorithm, and generating a diffusion probability cloud map; superimposing the diffusion probability cloud map onto a three-dimensional geographic information model, characterizing the threat intensity distribution through a color gradient, and generating heat map warning information that marks the key path nodes of the threat diffusion and the estimated arrival time information.

[0007] Preferably, the cross-modal attention mechanism formula is:

[0008] in, Attention output result; V is the value vector; S is the source feature; The query vector generated by V; The key vector generated by S; Key Vector The transposed matrix of Key Vector Dimensions; The dot product of the query vector and the key vector; is the scaled dot product score; right Each row of the matrix is ​​normalized.

[0009] Preferably, a dynamic behavior prediction model is constructed using a spatiotemporal graph convolutional network to obtain the probability of abnormal vehicle behavior and the evolution trajectory, including: constructing a dynamic behavior prediction model; based on the dynamic behavior prediction model, setting intersections and tunnel key nodes as core nodes of the graph network, combining the vehicle braking frequency and steering angle micro-features, and modeling the spatiotemporal correlation of vehicle behavior through the spatiotemporal graph convolutional network to predict the probability of abnormal vehicle behavior and the corresponding evolution trajectory, wherein the probability of abnormal vehicle behavior is used to quantify the degree to which the vehicle behavior deviates from the normal mode, and the evolution trajectory is used to describe the spatial position change path of the vehicle's abnormal behavior in the time dimension.

[0010] Preferably, the dynamic resolution adjustment algorithm formula is:

[0011] in, is the video resolution at time t; Initial resolution; Network load factor; is the scene complexity coefficient.

[0012] Preferably, triggering the graded warning instruction includes: A four-level early warning mechanism is set up. When the probability of abnormal behavior is less than 30%, the threat range has no tendency to spread, and the vehicle density and average speed are within the normal range, it is classified as a level one warning, triggering local sound and light prompts, and only the abnormal target location and real-time trajectory are marked. When the probability of abnormal behavior is greater than 30% but less than 50%, the threat has limited diffusion, the vehicle density is high, and the average speed drops slightly, it is classified as a level two warning, and the IoT security equipment cluster is linked to respond in a coordinated manner, and the area with abnormal environmental parameters is marked. When the probability of abnormal behavior is greater than 50% but less than 80%, the threat spreads to the regional level, the road capacity decreases significantly, and the vehicle density increases sharply, it is classified as a level three warning, and augmented reality warning information is pushed to the security center. The risk coordinates are marked by linking the heat map with geographic information, and the nearest patrol force is dispatched. When the probability of abnormal behavior is greater than 80%, the threat has a tendency to spread across regions, the road capacity is seriously insufficient, and there is a risk of traffic paralysis, it is classified as a level four warning, and the emergency command plan is activated. The heat map highlights the core threat area and overlays the distribution of rescue forces.

[0013] The second embodiment of the present application provides a 5G Internet of Things integrated video surveillance intelligent early warning system, including: an acquisition module for vehicle attribute information and environmental perception data; a fusion module for constructing a 5G Internet of Things perception network based on the vehicle attribute information and environmental perception data, and based on the 5G Internet of Things perception network, fusing the vehicle motion trajectory and road environment data through a cross-modal attention mechanism to generate an enhanced environmental perception map, and at the same time, using a dynamic resolution adjustment algorithm to optimize transmission efficiency; an early warning module for constructing a dynamic behavior prediction model based on the enhanced environmental perception map using a spatiotemporal graph convolutional network to predict the probability and evolution trajectory of abnormal vehicle behavior, and based on the abnormal behavior probability and evolution trajectory, calculating the early warning level through an adaptive risk quantification algorithm, and simulating the risk diffusion coefficient based on the current road section vehicle density, average speed and road capacity, using a dynamic risk propagation model to correct the early warning sensitivity parameters in real time, and transmitting the corrected sensitivity parameters to the command platform for multi-dimensional situation deduction, triggering graded early warning instructions and generating heat map warning information containing threat diffusion paths.

[0014] The third aspect of the present application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and runnable on the processor. The processor executes the program to implement a 5G Internet of Things integrated video surveillance intelligent early warning method as described in the above embodiment.

[0015] The fourth aspect of the present application provides a computer-readable storage medium on which a computer program is stored. The program is executed by a processor to implement a 5G Internet of Things integrated video surveillance intelligent early warning method as described in the above embodiment.

[0016] The fifth aspect of the present application provides a computer program product, including a computer program or instructions, for implementing a 5G Internet of Things integrated video surveillance intelligent early warning method as described in the above embodiment.

[0017] Therefore, the present application includes the following beneficial effects: the embodiment of the present application constructs a 5G Internet of Things perception network, and uses a cross-modal attention mechanism to deeply integrate vehicle motion trajectories and road environment data. The generated enhanced environmental perception map can break the data barriers between video images and physical quantity parameters, so that multi-dimensional information such as vehicle attributes and road conditions can form a collaborative association, solving the problems of data isolation and one-sided analysis in traditional monitoring; the dynamic resolution adjustment algorithm adopted simultaneously can intelligently control the transmission accuracy according to the real-time bandwidth and data importance, and reduce the end-to-end transmission delay of high-definition video and sensor data while ensuring the integrity of key information, thereby alleviating transmission congestion in concurrent scenarios. The dynamic behavior prediction model, built on a spatiotemporal graph convolutional network, can accurately capture the spatiotemporal correlation characteristics of vehicle movement, improve the accuracy of abnormal behavior identification, and simultaneously achieve advanced risk prediction by combining evolutionary trajectory analysis. The dynamic risk propagation model simulates and generates risk diffusion coefficients through real-time analysis of parameters such as vehicle density, average speed, and road capacity, providing a basis for dynamic correction of warning sensitivity parameters. The combination of multi-dimensional situational deduction and a graded warning mechanism achieves full-level coverage from local prompts to cross-regional emergency response. Through the linkage annotation of heat maps and geographic information, the visualization accuracy of risk propagation paths is improved, enabling the command platform to quickly identify the core threat area and dispatch resources. This enhances environmental adaptability and decision-making robustness in complex traffic scenarios, improving road traffic efficiency and public safety protection capabilities. This solves the problems of insufficient positioning accuracy, weak analysis capabilities, and poor transmission efficiency in existing technologies.

[0018] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a flowchart of a 5G Internet of Things integrated video surveillance intelligent early warning method provided according to an embodiment of the present application; Figure 2 A schematic diagram of a highway congestion scenario provided according to one embodiment of the present application; Figure 3 A schematic diagram of a city trunk road scene during morning and evening rush hours according to one embodiment of the present application; Figure 4 A schematic diagram of a 5G Internet of Things integrated video surveillance intelligent early warning method provided according to one embodiment of the present application; Figure 5 This is a structural diagram of a 5G Internet of Things integrated video surveillance intelligent early warning system provided according to an embodiment of the present application; Figure 6 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] The following describes in detail embodiments of the present application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0021] The following describes a 5G Internet of Things integrated video surveillance intelligent early warning method of an embodiment of the present application with reference to the accompanying drawings. In response to the problem of insufficient positioning accuracy mentioned in the above background technology, the present application provides a 5G Internet of Things integrated video surveillance intelligent early warning method. In this method, by constructing a 5G Internet of Things perception network, the vehicle motion trajectory and road environment data are deeply integrated with the cross-modal attention mechanism. The generated enhanced environmental perception map can break the data barrier between the video image and the physical quantity parameters, so that multi-dimensional information such as vehicle attributes and road conditions form a collaborative association, solving the problems of data isolation and one-sided analysis in traditional monitoring; the dynamic resolution adjustment algorithm adopted simultaneously can intelligently adjust the transmission accuracy according to the real-time bandwidth and data importance, and reduce the end-to-end transmission delay of high-definition video and sensor data on the premise of ensuring the integrity of key information, thereby alleviating transmission congestion in concurrent scenarios. The dynamic behavior prediction model, built on a spatiotemporal graph convolutional network, can accurately capture the spatiotemporal correlation characteristics of vehicle movement, improve the accuracy of abnormal behavior identification, and simultaneously achieve advanced risk prediction by combining evolutionary trajectory analysis. The dynamic risk propagation model simulates and generates risk diffusion coefficients through real-time analysis of parameters such as vehicle density, average speed, and road capacity, providing a basis for dynamic correction of warning sensitivity parameters. The combination of multi-dimensional situational deduction and a graded warning mechanism achieves full-level coverage from local prompts to cross-regional emergency response. Through the linkage annotation of heat maps and geographic information, the visualization accuracy of risk propagation paths is improved, enabling the command platform to quickly identify the core threat area and dispatch resources. This enhances environmental adaptability and decision-making robustness in complex traffic scenarios, improving road traffic efficiency and public safety protection capabilities. This solves the problems of insufficient positioning accuracy, weak analysis capabilities, and poor transmission efficiency in existing technologies.

[0022] Specifically, Figure 1 A flow chart of a 5G Internet of Things integrated video surveillance intelligent early warning method provided in an embodiment of the present application.

[0023] like Figure 1 As shown, the 5G Internet of Things integrated video surveillance intelligent early warning method includes the following steps: In step S101, vehicle attribute information and environmental perception data are obtained.

[0024] Among them, vehicle attribute information refers to information used to characterize the inherent characteristics and dynamic status of the vehicle, including data such as vehicle model, license plate, color, driving speed, load and vehicle identification code.

[0025] It can be understood that the embodiments of the present application provide core data for the establishment of a 5G Internet of Things perception network by obtaining vehicle attribute information, accurately associate vehicle motion trajectories with identity features, and generate an enhanced environmental perception map containing key elements such as vehicle model, license plate, and speed; provide an analysis basis for the spatiotemporal graph convolutional network, and improve the pertinence and accuracy of abnormal behavior probability prediction by distinguishing the inherent characteristics and dynamic states of different vehicles. At the same time, it provides basic parameters for vehicle density calculation for the dynamic risk propagation model, enhances the ability to recognize abnormal patterns of specific vehicles, and reduces misjudgments and missed judgments.

[0026] In step S102, a 5G IoT perception network is constructed based on vehicle attribute information and environmental perception data. Based on the 5G IoT perception network, the vehicle motion trajectory and road environment data are fused through a cross-modal attention mechanism to generate an enhanced environmental perception map. At the same time, a dynamic resolution adjustment algorithm is used to optimize transmission efficiency.

[0027] Among them, the enhanced environmental perception map is based on the 5G Internet of Things perception network. It is generated by fusing vehicle motion trajectory and road environment data through a cross-modal attention mechanism. It is a comprehensive perception map that can fully reflect the correlation characteristics between vehicle status and environment.

[0028] It can be understood that the embodiment of the present application transforms the originally isolated video images and physical quantity parameters into interrelated structured graphs by enhancing the environmental perception map, capturing the dynamic correlation between vehicle behaviors such as lane changing and acceleration and environmental factors such as slippery road surface and traffic density, providing an input basis for the subsequent spatiotemporal graph convolutional network to build a dynamic behavior prediction model, improving the accuracy of abnormal behavior recognition and trajectory prediction, and cooperating with the dynamic resolution adjustment algorithm to ensure data integrity while optimizing transmission efficiency, enhancing the comprehensive perception ability of complex traffic scenes, and improving the intelligence level of traffic control.

[0029] It should be noted that, based on vehicle attribute information and environmental perception data, a 5G IoT perception network is constructed. First, the vehicle's built-in OBD interface, GPS unit and on-board camera are used as mobile perception nodes to collect vehicle model, license plate, driving speed, acceleration, real-time location and other attribute information in real time. At the same time, millimeter-wave radars, high-definition cameras, meteorological sensors, traffic flow detectors and other fixed perception equipment are deployed along the road to collect environmental perception data such as road conditions, traffic flow, visibility, temperature and humidity. By utilizing the large connection characteristics of the 5G network, mobile nodes and fixed devices are connected to the network architecture of the unified communication protocol through the edge computing gateway to realize data interconnection between the vehicle end and roadside equipment. The network bandwidth is allocated in real time according to the vehicle density and data transmission volume to ensure the transmission stability of high-priority data, and to build a 5G IoT perception network covering the entire life cycle status of the vehicle and the entire road environment.

[0030] For example, Figure 2 As shown in the figure, in a highway congestion scenario, the enhanced environmental perception map constructs a dynamic correlation map by integrating the real-time GPS trajectories of multiple trucks (showing that the vehicle speed drops sharply from 90km / h to 20km / h within 3 consecutive kilometers) with the emergency lane occupancy images captured by roadside cameras and sudden fog data sent back by meteorological sensors. The map not only marks the correspondence between truck load information (mostly overweight transportation) and sudden braking frequency, but also simultaneously superimposes the thermal distribution of fog concentration diffusion speed (spreading 1 km every 5 minutes) and traffic density, clearly presenting the chain reaction of "overweight trucks driving at low speed → occupying emergency lanes → fog aggravating congestion → increased risk of rear-end collisions with following vehicles". It provides the command platform with a visual basis for the complete evolution chain from a single abnormal vehicle to regional traffic risks, helping to quickly determine the core sections that need priority diversion and the dispatch paths of rescue forces.

[0031] In the embodiment of the present application, the cross-modal attention mechanism formula is:

[0032] in, Attention output result; V is the value vector; S is the source feature; The query vector generated by V; The key vector generated by S; Key Vector The transposed matrix of Key Vector Dimensions; The dot product of the query vector and the key vector; scaled dot product score; right Each row of the matrix is ​​normalized.

[0033] It can be understood that, by using the cross-modal attention mechanism, the embodiments of the present application accurately focus on the key associated features between the vehicle motion trajectory and the road condition environment data, highlight the core associated information such as the truck emergency braking trajectory and the road surface wetness, the fog concentration and the traffic density, weaken the irrelevant background interference, perform deep collaborative reasoning on the visual data and the physical quantity parameters, improve the feature expression accuracy of the enhanced environment perception graph, make the associated description of the vehicle behavior and the environmental factors more accurate, provide data input for the subsequent dynamic behavior prediction model, reduce the redundant data processing amount, indirectly cooperate with the dynamic resolution adjustment algorithm to optimize the transmission efficiency, enhance the adaptability to complex traffic scenes and the scientific nature of analysis and decision-making.

[0034] For example, in the urban expressway scene in heavy rain weather, the cross-modal attention mechanism will automatically focus on the key association between the vehicle motion trajectory and the road condition environment data: when the camera captures that a certain small car brakes 3 times in a row (visual data), and the roadside sensor returns that the road surface humidity of this section is 85% and the water depth is 10cm (physical quantity parameters), the mechanism will allocate more than 90% attention weight to the “emergency braking trajectory-road surface wetness”, weaken irrelevant information such as light and shadow changes of distant street lamps, and accurately identify the braking abnormality of the vehicle caused by the wet road surface; when a large truck in the adjacent lane appears continuous lane changing behavior, the mechanism will real-time increase the association weight of “truck trajectory-traffic density”, combine the millimeter wave radar distance data to quickly judge whether there is a malicious lane changing risk, and through dynamic balance of attention allocation of multi-modal data, realize accurate binding analysis of vehicle behavior and environmental factors.

[0035] In the embodiments of the present application, the dynamic resolution adjustment algorithm formula is:

[0036] wherein, is the dynamic adjusted resolution at time t; is the original basic resolution; is the real-time available bandwidth at time t; is the demand bandwidth of current data transmission; is the data importance weight; is the redundancy data coefficient, which is positively correlated with the data redundancy degree; is the network transmission delay at time t.

[0037] It can be understood that the dynamic resolution adjustment algorithm in the embodiment of the present application dynamically adjusts the resolution of the video stream and sensor data according to the real-time bandwidth of data transmission in the 5G Internet of Things perception network, the importance level of vehicle attribute information and environmental perception data, and performs dimensionality reduction processing on non-core area data while ensuring clear transmission of key information, thereby effectively reducing the amount of data transmission, reducing transmission congestion and frame loss when 4K / 8K high-definition video streams and massive sensor data are concurrent, and improving transmission efficiency; at the same time, by reasonably allocating resources, it avoids redundant data occupying too much bandwidth, provides stable data transmission support for cross-modal attention mechanism to fuse data and generate enhanced environmental perception maps, and enhances real-time performance and reliability in complex traffic scenarios.

[0038] For example, Figure 3 As shown in the figure, in the urban main road scenarios during morning and evening rush hours, the dynamic resolution adjustment algorithm flexibly adjusts the transmission parameters according to the real-time bandwidth and data importance: when the 5G network bandwidth drops below 80Mbps, the algorithm automatically reduces the video resolution of non-congested sections from 4K to 1080P, while maintaining 4K high-definition image quality in key areas such as congested intersections and school perimeters, ensuring clear transmission of abnormal behaviors such as vehicle cutting in and pedestrians crossing the road; for sensor data, only high-precision transmission of core parameters such as vehicle speed and distance is retained, and non-emergency environmental data such as road surface temperature is sampled and down-sampled, reducing the overall data transmission volume by more than 40%. This not only avoids the delay caused by bandwidth saturation, but also provides a stable flow for the cross-modal attention mechanism to fuse vehicle trajectories and road condition data, ensuring real-time updates of the enhanced environmental perception map.

[0039] In step S103, based on the enhanced environmental perception map, a dynamic behavior prediction model is constructed using a spatiotemporal graph convolutional network to predict the probability and evolution trajectory of abnormal vehicle behavior. Based on the abnormal behavior probability and evolution trajectory, the warning level is calculated using an adaptive risk quantification algorithm. According to the current vehicle density, average speed and road capacity of the road section, a dynamic risk propagation model is used to simulate the risk diffusion coefficient, and the warning sensitivity parameters are corrected in real time. The corrected sensitivity parameters are transmitted to the command platform for multi-dimensional situation deduction, triggering graded warning instructions and generating heat map warning information containing the threat diffusion path.

[0040] Among them, the adaptive risk quantification algorithm is an algorithm that can dynamically adjust the risk assessment dimensions and weights according to the probability of abnormal vehicle behavior, evolution trajectory and real-time traffic scene parameters to achieve accurate calculation of the warning level.

[0041] It can be understood that the embodiment of the present application dynamically integrates the probability of abnormal vehicle behavior, evolution trajectory and real-time traffic scene parameters, flexibly adjusts the risk assessment dimensions and the weights of each dimension, and accurately calculates the warning level that fits the actual situation, providing a key foundation for the dynamic risk propagation model, helping to simulate the risk diffusion coefficient, and the real-time correction of the warning sensitivity parameters, so that the warning threshold is always synchronized with the current road section vehicle density, average speed and other traffic conditions, avoiding excessive warnings in low-risk scenarios, preventing warning lags in high-risk situations, improving the scientific nature of risk assessment and the accuracy of warning responses, and enhancing the robustness of warning decisions.

[0042] It should be noted that the spatiotemporal graph convolutional network formula is:

[0043] in, is the graph node feature matrix at time t+1; is the k-th graph adjacency matrix; is the convolution weight corresponding to the d-th feature dimension; is the temporal convolution kernel; is the number of spatial convolution kernels; is the feature dimension; is the activation function; It is the temporal convolution operation; is the d-th dimension feature matrix of the node in the figure below at time t.

[0044] Dynamic behavior prediction model formula:

[0045] in, is the input feature sequence at time t; is the graph adjacency matrix; is the spatiotemporal feature vector extracted by the spatiotemporal graph convolutional network; Long short-term memory network; is the prediction step length; is the probability distribution of vehicle abnormal behavior at time t+τ; is the normalized exponential function.

[0046] Adaptive risk quantification algorithm formula:

[0047] in, is the final calculated risk value; is the risk assessment dimension; is the dynamically adjusted factor weight; Quantify the function for risk indicator; is the vehicle attribute; is the environmental state; is the real-time traffic scene parameter; For time.

[0048] Dynamic risk propagation model formula:

[0049] in, for Time and space position The risk value at is the risk value at the current time t and spatial position (x, y); is the time step; is the spatial diffusion coefficient; is the Laplace operator of risk value; is the traffic flow impact coefficient; is the vehicle density at time t; is the average vehicle speed; For road capacity.

[0050] For example, in the scenario of a crossroad during the evening rush hour on a rainy day, the adaptive risk quantification algorithm first integrates the output data of the dynamic behavior prediction model: the probability of a non-decelerating SUV running a red light surges to 72%, its 5-second continuous evolution trajectory shows a 1.2-meter deviation from the straight lane and an acceleration of 2.3 m / s², and the associated vehicle attribute information shows that the vehicle is a large SUV with an axle distance of more than 3 meters (the braking distance is 20% longer than that of an ordinary car). Then, the algorithm accesses real-time traffic parameters from roadside devices: the gathering density of 20 pedestrians at the zebra crossing (1.5 people per square meter), the 30-mm rainfall intensity in 1 hour (which reduces the road surface friction coefficient to 0.4), and the current traffic capacity of the intersection is only 60% of the design value (the queue vehicles have overflowed the stop line by 5 meters). Based on the scenario characteristics, the algorithm automatically starts the dynamic weight adjustment mechanism: referring to the risk correlation degree of the "large vehicle violation + pedestrian dense + rainy day" scenario in historical accident data, the evaluation dimension weight of "red light running probability x pedestrian density x vehicle type coefficient" is increased from the regular 0.3 to 0.5, the weight of "trajectory deviation x road wetness coefficient x braking distance correction value" is increased from 0.2 to 0.3, and the weight of "traffic saturation degree" is reduced from 0.3 to 0.2 due to the current traffic flow not reaching the extreme saturation state. Through weighted calculation of the risk scores of each dimension (red light running dimension: 42 points, trajectory deviation dimension: 31 points, traffic flow dimension: 16 points), the comprehensive risk value is 89 points, which accurately matches the first-level warning threshold. The result triggers the linkage response of the command platform: not only extends the red light by 3 seconds, but also synchronously activates the sound and light warning devices in 3 directions of the intersection (the flashing frequency is dynamically adjusted according to the risk value), and sends real-time voice reminders to the SUV. After 2 seconds, the algorithm monitors that the vehicle decelerates to 20 km / h, combines the trajectory correction data, and reduces the risk value to 45 points (second-level warning) in real time, automatically reduces the warning intensity, which not only avoids the "misjudgment as low risk" caused by fixed weight evaluation, but also realizes the precise allocation of warning resources through dynamic correction, shortens the risk response time by 1.8 seconds compared with the traditional algorithm, and reduces the accident hidden danger rate in such scenarios by 37%.

[0051] In the embodiment of the present application, the heat map alarm information containing the threat diffusion path is generated, including: constructing a dynamic threat diffusion model; based on the dynamic threat diffusion model, simulating the motion trajectory of the threat particles in the monitoring area by combining the streaming particle swarm algorithm, mapping the spatial distribution of the particle trajectory to the threat diffusion probability value by the kernel density estimation algorithm, and generating the diffusion probability cloud map; superimposing the diffusion probability cloud map on the three-dimensional geographic information model, representing the threat intensity distribution through the color gradient, and generating the heat map alarm information of the key path node and the predicted arrival time information of the threat diffusion.

[0052] The streaming particle swarm algorithm is a swarm intelligence algorithm that can process dynamic streaming data by adaptively updating the speed and position of the particle swarm to achieve real-time optimization.

[0053] It can be understood that the embodiments of the present application use the streaming particle swarm algorithm, based on the ability to process dynamic streaming data and the adaptive update characteristics of the particle swarm, to accurately simulate the movement trajectory of threat particles in the monitoring area, adapt to the dynamic changes of threat diffusion, and improve the accuracy of trajectory simulation through swarm intelligence optimization, thereby providing accurate data for the subsequent generation of reliable diffusion probability cloud maps and heat map alarm information, improving the real-time and effectiveness of alarm information, and timely grasping the threat diffusion situation.

[0054] It should be noted that the formula of the streaming particle swarm algorithm is:

[0055] in, is the velocity vector of the i-th particle at the t-th iteration; is the inertia weight at the tth iteration; is the velocity vector of the i-th particle at the t−1-th iteration; is the cognitive learning factor at the tth iteration; is a random number; is the best position found so far by the i-th particle; is the position vector of the i-th particle at the t−1th iteration The kernel density estimation algorithm is used to map the spatial distribution of particle trajectories into threat diffusion probability values, generating a diffusion probability cloud map. The kernel density estimation algorithm assigns a kernel function weight to each spatial point on the particle trajectory and superimposes the weighted contributions of all points to calculate the threat diffusion probability value within the area. It converts discrete particle motion trajectories into a continuous spatial probability distribution, generating a diffusion probability cloud map that can intuitively present the possibility of threat diffusion at different locations in the monitored area.

[0056] Kernel density estimation algorithm formula:

[0057] in, is the probability density estimate at position x; is the sample size; is bandwidth; is the kernel function; is the i-th sample point; is the location where the probability density is to be estimated; is the standardized distance; is the index of the sample point.

[0058] For example, in real-time traffic prediction for intelligent transportation, the streaming particle swarm algorithm treats particles as vehicle carriers, continuously receiving streaming data such as real-time traffic flow from road sensors, vehicle trajectories from GPS terminals, signal light timing, and unexpected accidents. When a sudden increase in traffic flow on a particular road section is detected, the algorithm dynamically increases the particle inertia weight to enhance global exploration, tilting the particles toward congestion diffusion. If an accident occurs in a certain area, the algorithm enhances the individual cognitive factor to allow nearby particles to reinforce their historical trajectory experience. Simultaneously, the group learning factor guides other particles toward the optimal detour path, rapidly correcting the group's optimal solution. Incorporating road network topology (such as intersection capacity and number of lanes), particles simulate traffic diversion by adjusting their velocity components at intersections. Ultimately, they generate a real-time forecast of traffic conditions for the next 30 minutes, providing traffic authorities with congestion warnings (such as identifying bottlenecks 10 minutes in advance) and traffic diversion strategies, improving regional traffic efficiency by over 25%.

[0059] In an embodiment of the present application, a dynamic behavior prediction model is constructed using a spatiotemporal graph convolutional network to obtain the probability of abnormal vehicle behavior and the evolution trajectory, including: constructing a dynamic behavior prediction model; based on the dynamic behavior prediction model, setting intersections and tunnel key nodes as core nodes of the graph network, combining the vehicle braking frequency and steering angle micro-features, and modeling the spatiotemporal correlation of vehicle behavior through the spatiotemporal graph convolutional network to predict the probability of abnormal vehicle behavior and the corresponding evolution trajectory, wherein the probability of abnormal vehicle behavior is used to quantify the degree to which the vehicle behavior deviates from the normal mode, and the evolution trajectory is used to describe the spatial position change path of the vehicle's abnormal behavior in the time dimension.

[0060] Among them, the dynamic behavior prediction model is a model that predicts future development trends or states by analyzing behavioral data that changes dynamically over time.

[0061] It can be understood that the embodiments of the present application use a dynamic behavior prediction model and rely on a spatiotemporal graph convolutional network to accurately model the spatiotemporal correlation between key nodes such as intersections and tunnels and microscopic features such as vehicle braking frequency and steering angle. By quantifying the abnormal probability of vehicle behavior deviating from the normal mode and describing the spatial evolution trajectory of abnormal behavior in the time dimension, abnormal vehicle behavior is identified and dynamically tracked in advance, thereby improving the accuracy and timeliness of traffic monitoring, providing data support for timely intervention measures, and improving road traffic safety and efficiency.

[0062] For example, urban intelligent transportation uses a dynamic behavior prediction model to achieve accurate identification and tracking of vehicle abnormal behavior through a spatio-temporal graph convolution network. Taking the tunnel scene as an example, the model sets key nodes such as tunnel entrances and curves as the core of the graph network, and real-time accesses microscopic feature data such as vehicle braking frequency (e.g., more than 3 times in 10 seconds) and steering angle (e.g., more than 45 degrees), and combines lane line offset detected by laser radar and front and rear vehicle distance changes detected by millimeter wave radar to construct a spatio-temporal correlation graph of vehicle behavior. When a vehicle appears to brake abnormally frequently (e.g., with a braking interval of less than 2 seconds) and the steering angle deviates from the center line of the lane in the tunnel, the model captures the interaction mode of its motion trajectory and surrounding vehicles through the spatio-temporal graph convolution layer, predicts its illegal lane changing probability to be 89%, and generates an evolution trajectory (e.g., offset 1.2 meters to the left into the adjacent lane) in the next 10 seconds within 500 milliseconds. The model has been piloted in the Shanghai North-South Channel, reducing the response time of abnormal behavior warning in the tunnel to 1.2 seconds, reducing the accident rate by 37%, and dynamically adjusting the lane allocation through the traffic signal lights, improving the traffic efficiency by 18%.

[0063] In the embodiments of the present application, the trigger hierarchical warning instruction includes: setting a four-level warning mechanism, when the abnormal behavior probability is less than 30% and the threat range has no diffusion trend, the vehicle density and the average vehicle speed are in the normal range, it is divided into a first-level warning, triggering local sound and light prompts, only marking the abnormal target position and real-time trajectory; when the abnormal behavior probability is greater than 30% and less than 50%, the threat has limited diffusion, the vehicle density is high and the average vehicle speed is slightly reduced, it is divided into a second-level warning, and the Internet of Things security device cluster is associated to respond, and the environmental parameter abnormal area is marked; when the abnormal behavior probability is greater than 50% and less than 80%, the threat spreads to the regional level, the road traffic capacity decreases significantly and the vehicle density increases sharply, it is divided into a third-level warning, the augmented reality warning information is pushed to the security center, the risk coordinates are marked on the heat map associated with the geographic information, and the nearest patrol forces are dispatched; when the abnormal behavior probability is greater than 80%, the threat forms a cross-regional diffusion trend, the road traffic capacity is severely insufficient, and there is a risk of traffic paralysis, it is divided into a fourth-level warning, the emergency command plan is started, the threat core area is highlighted on the heat map, and the rescue force distribution is superimposed.

[0064] It is understandable that the embodiments of the present application achieve precise graded responses by combining the probability of abnormal behavior, the trend of threat spread, and the traffic operation status: the first-level warning targets minor anomalies, focusing on key information with local prompts to avoid excessive intervention; the second-level warning targets limited spread risks, linking the IoT device cluster for coordinated disposal and timely blocking the spread of risks; the third-level warning targets regional threats, dispatching patrol forces through augmented reality information and geographic heat maps to quickly alleviate traffic pressure; the fourth-level warning targets major cross-regional risks, activating emergency plans to coordinate rescue resources and minimize the risk of paralysis. The graded mechanism ensures the pertinence and timeliness of warnings, rationally allocates rescue resources, and matches the optimal response strength from minor hidden dangers to major crises, improving the efficiency and safety of traffic anomaly handling and ensuring stable road traffic order.

[0065] According to a 5G Internet of Things integrated video surveillance intelligent early warning method proposed in an embodiment of the present application, by building a 5G Internet of Things perception network, the vehicle motion trajectory and road environment data are deeply integrated with the cross-modal attention mechanism. The generated enhanced environmental perception map can break the data barriers between video images and physical quantity parameters, so that multi-dimensional information such as vehicle attributes and road conditions can form a collaborative association, solving the problems of data isolation and one-sided analysis in traditional monitoring; the dynamic resolution adjustment algorithm adopted simultaneously can intelligently control the transmission accuracy according to the real-time bandwidth and data importance, and reduce the end-to-end transmission delay of high-definition video and sensor data while ensuring the integrity of key information, thereby alleviating transmission congestion in concurrent scenarios. The dynamic behavior prediction model, built on a spatiotemporal graph convolutional network, can accurately capture the spatiotemporal correlation characteristics of vehicle movement, improve the accuracy of abnormal behavior identification, and simultaneously achieve advanced risk prediction by combining evolutionary trajectory analysis. The dynamic risk propagation model simulates and generates risk diffusion coefficients through real-time analysis of parameters such as vehicle density, average speed, and road capacity, providing a basis for dynamic correction of warning sensitivity parameters. The combination of multi-dimensional situational deduction and a graded warning mechanism achieves full-level coverage from local prompts to cross-regional emergency response. Through the linkage annotation of heat maps and geographic information, the visualization accuracy of risk propagation paths is improved, enabling the command platform to quickly identify the core threat area and dispatch resources. This enhances environmental adaptability and decision-making robustness in complex traffic scenarios, improving road traffic efficiency and public safety protection capabilities. This solves the problems of insufficient positioning accuracy, weak analysis capabilities, and poor transmission efficiency in existing technologies.

[0066] The following will describe a 5G Internet of Things integrated video surveillance intelligent early warning method through a specific embodiment. Figure 4 Shown, including: The application scenario is the intersection of the "East-West Expressway and North-South Trunk Road" in a certain city (including three signalized intersections, one tunnel entrance, and an average daily traffic volume of 32,000 vehicles). Due to the dense traffic and significant tidal phenomenon in this area, illegal lane changes and sudden congestion often occur. The 5G vehicle-mounted terminal (supporting the 3GPP Release 16 standard, with a latency of ≤20ms) deployed at the intersection collects dynamic data such as the VIN code, vehicle type (such as heavy trucks and small passenger cars), real-time speed (accuracy of ±1km / h), braking frequency (number of brakes per unit time), and steering angle (range -30°~30°) of passing vehicles; through the road The side millimeter-wave radar (detection range of 150m, angular resolution of 0.5°) captures vehicle spacing (minimum recognition distance of 2m) and lane deviation (±0.3m), and uses a high-definition camera (4K resolution, frame rate of 25fps) to identify road signs and traffic light status (red / green light duration). The fiber grating sensor monitors road surface smoothness (error ≤ 0.5mm). The weather station transmits real-time temperature (-10℃~40℃), humidity (20%~90%), and visibility (50m~10km) data. The 5G Internet of Things gateway collects roadside equipment status (such as radar signal strength and camera computing power utilization).

[0067] Three 5G edge computing nodes (MEC, computing power 50TOPS) are deployed in the hub area. 5G independent networking (SA) slicing technology is used to divide the "monitoring dedicated slice" (bandwidth 100Mbps, jitter ≤ 5ms), connecting 200+ roadside equipment (radars, cameras, sensors) and 500+ vehicle terminals to form a full-domain perception network. For the morning rush hour (7:30-9:00) with dense traffic flow, two types of source features are integrated: video source features (S): vehicle trajectory sequence captured by the camera; radar source features (V): vehicle speed and distance matrix output by the millimeter wave radar. The cross-modal attention mechanism formula is used to calculate , generate query vector Q (based on the motion trend of the video trajectory) and key vector K (based on the risk association of radar speed), calculate the dot product score and normalize it, focus on "vehicle pairs with a sudden drop in speed and a distance less than the safe distance (50m)", increase their weight by 20%, and finally merge them into an enhanced environmental perception map that includes "high-risk vehicle pair position, relative speed, and lane departure warning". When the 5G bandwidth drops to 60Mbps during the morning peak (the required bandwidth is 80Mbps), the dynamic resolution adjustment algorithm is triggered The camera resolution for non-critical areas (such as roadside green belts) is reduced from 4K to 2K, while the core area of ​​the intersection remains at 4K (data importance weight w=0.8). The transmission delay is reduced from 30ms to 18ms, ensuring the real-time nature of key data.

[0068] Set the intersection signal pole, tunnel entrance radar station as the graph network core node (a total of 8), the vehicle within 200m range as the node associated edge; continuous 3 times braking (braking frequency 0.5 times / s), steering angle left and right (-25°~+20°), combined with the obstacle avoidance trajectory of the surrounding vehicles (such as small cars), capture the spatiotemporal association of "truck-small car" through the spatiotemporal convolution layer (such as truck deviation leading to small car emergency braking). Based on dynamic behavior prediction model , output prediction results, the abnormal behavior probability of the truck changing lanes within 5 seconds is 62%, and the evolution trajectory is "from west to east main lane to emergency lane, and may invade the emergency lane 1.5m after 10 seconds". Self-adaptive risk quantification algorithm Combined with the current state: abnormal probability 62%, vehicle density up to 28 vehicles per hundred meters (normal interval 15-20 vehicles per hundred meters), average speed from 50km / h to 32km / h, judged as level 3 warning; dynamic risk propagation model Introduce "rainy road slip" environmental parameters (friction coefficient 0.6), simulate risk diffusion coefficient 1.8 (dry road 1.2), modify warning sensitivity parameters (increase by 20%), and push "3 minutes may spread to adjacent 2 lanes" situation deduction results to the command platform.

[0069] Push augmented reality (AR) warning information to the security center: mark the real-time position of the truck (east longitude 113°45°, north latitude 34°75°) on the electronic map, superimpose its evolution trajectory; mark the risk area (area about 500㎡) with red gradient, associate with geographic information to mark the tunnel entrance 300m as high-risk coordinates, dispatch the patrol police car 800m away from the scene (through 5G positioning, response time ≤2 minutes). Dynamic threat diffusion model combined with flow particle swarm algorithm , "truck illegal lane change" is regarded as a threat particle, and its diffusion path in the traffic flow is simulated (affected by the avoidance of surrounding vehicles, the particle diffusion speed to the northwest is 0.3m / s); kernel density estimation algorithm Map the particle trajectory to a diffusion probability cloud map (the probability at the tunnel entrance is 72%); superimposed on the three-dimensional geographic information model (including road elevation, surrounding buildings), generate a heat map: mark the key nodes (such as 150m away from the tunnel entrance expected to arrive in 2 minutes), provide accurate disposal coordinates for the patrol force.

[0070] In summary, this invention achieves comprehensive perception of vehicle and environmental data through multi-source devices. Combining 65G6 low-latency networks with dynamic resolution adjustment technology ensures real-time transmission of critical data while optimizing bandwidth utilization. Relying on a spatiotemporal graph convolutional network, it accurately predicts abnormal vehicle behavior and dynamically modifies risk assessments based on environmental parameters to improve early warning accuracy. Through a three-level early warning mechanism, 6AR6 information push and police dispatch are linked, along with threat diffusion simulation and thermal image annotation, enabling rapid response and precise handling of risks. Ultimately, this effectively reduces scratches and collisions caused by illegal lane changes, shortens early warning response time, and reduces congestion duration in key areas like tunnel entrances, comprehensively improving traffic safety and efficiency in this densely populated hub.

[0071] Next, a 5G IoT integrated video surveillance intelligent early warning system proposed according to an embodiment of the present application is described with reference to the accompanying drawings.

[0072] Figure 5 This is a structural diagram of a 5G Internet of Things integrated video surveillance intelligent early warning system according to an embodiment of the present application.

[0073] like Figure 5 As shown, the 5G Internet of Things integrated video surveillance intelligent early warning system 10 includes: an acquisition module 100, a fusion module 200, and an early warning module 300.

[0074] Among them, the acquisition module 100 is used to acquire module for vehicle attribute information and environmental perception data; the fusion module 200 is used to build a 5G Internet of Things perception network based on vehicle attribute information and environmental perception data, and based on the 5G Internet of Things perception network, the vehicle motion trajectory and road environment data are fused through the cross-modal attention mechanism to generate an enhanced environmental perception map. At the same time, the dynamic resolution adjustment algorithm is used to optimize the transmission efficiency; the early warning module 300 is used to build a dynamic behavior prediction model based on the enhanced environmental perception map using the spatiotemporal graph convolutional network to predict the probability and evolution trajectory of vehicle abnormal behavior, and calculate the early warning level based on the abnormal behavior probability and evolution trajectory through the adaptive risk quantification algorithm, and use the dynamic risk propagation model to simulate the risk diffusion coefficient according to the current road section vehicle density, average speed and road traffic capacity, and correct the early warning sensitivity parameters in real time. The corrected sensitivity parameters are transmitted to the command platform for multi-dimensional situation deduction, triggering graded early warning instructions and generating heat map warning information containing threat diffusion paths.

[0075] It should be noted that the above explanation of an embodiment of a 5G Internet of Things integrated video surveillance intelligent early warning method is also applicable to the embodiment of a 5G Internet of Things integrated video surveillance intelligent early warning system, and will not be repeated here.

[0076] According to a 5G Internet of Things integrated video surveillance intelligent early warning system proposed in an embodiment of the present application, by building a 5G Internet of Things perception network, the vehicle motion trajectory and road environment data are deeply integrated with the cross-modal attention mechanism. The generated enhanced environmental perception map can break the data barriers between video images and physical quantity parameters, so that multi-dimensional information such as vehicle attributes and road conditions can form a collaborative association, solving the problems of data isolation and one-sided analysis in traditional monitoring; the dynamic resolution adjustment algorithm adopted simultaneously can intelligently control the transmission accuracy according to the real-time bandwidth and data importance, and reduce the end-to-end transmission delay of high-definition video and sensor data while ensuring the integrity of key information, thereby alleviating transmission congestion in concurrent scenarios. The dynamic behavior prediction model, built on a spatiotemporal graph convolutional network, can accurately capture the spatiotemporal correlation characteristics of vehicle movement, improve the accuracy of abnormal behavior identification, and simultaneously achieve advanced risk prediction by combining evolutionary trajectory analysis. The dynamic risk propagation model simulates and generates risk diffusion coefficients through real-time analysis of parameters such as vehicle density, average speed, and road capacity, providing a basis for dynamic correction of warning sensitivity parameters. The combination of multi-dimensional situational deduction and a graded warning mechanism achieves full-level coverage from local prompts to cross-regional emergency response. Through the linkage annotation of heat maps and geographic information, the visualization accuracy of risk propagation paths is improved, enabling the command platform to quickly identify the core threat area and dispatch resources. This enhances environmental adaptability and decision-making robustness in complex traffic scenarios, improving road traffic efficiency and public safety protection capabilities. This solves the problems of insufficient positioning accuracy, weak analysis capabilities, and poor transmission efficiency in existing technologies.

[0077] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include: A memory 601 , a processor 602 , and a computer program stored in the memory 601 and executable on the processor 602 .

[0078] When the processor 602 executes the program, a 5G Internet of Things integrated video surveillance intelligent early warning method provided in the above embodiment is implemented.

[0079] Furthermore, the electronic device further includes: The communication interface 603 is used for communication between the memory 601 and the processor 602 .

[0080] The memory 601 is used to store computer programs that can be run on the processor 602 .

[0081] The memory 601 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0082] If the memory 601, processor 602, and communication interface 603 are implemented independently, the communication interface 603, memory 601, and processor 602 can be connected to each other via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0083] Optionally, in a specific implementation, if the memory 601, the processor 602 and the communication interface 603 are integrated on a chip, the memory 601, the processor 602 and the communication interface 603 can communicate with each other through an internal interface.

[0084] The processor 602 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0085] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, it implements the above-mentioned 5G Internet of Things integrated video surveillance intelligent early warning method.

[0086] In addition, an embodiment of the present application also provides a computer program product, including a computer program or instructions. When the computer program or instructions are executed, the above-mentioned 5G Internet of Things integrated video surveillance intelligent early warning method is implemented.

[0087] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0088] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0089] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0090] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0091] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0092] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A 5G Internet of Things integrated video surveillance intelligent early warning method, characterized in that: include: Obtain vehicle attribute information and environmental perception data; Based on the vehicle attribute information and environmental perception data, a 5G IoT perception network is constructed. Based on the 5G IoT perception network, the vehicle motion trajectory and road environment data are fused through a cross-modal attention mechanism to generate an enhanced environmental perception map. At the same time, a dynamic resolution adjustment algorithm is used to optimize transmission efficiency. According to the enhanced environmental perception map, a dynamic behavior prediction model is constructed using a spatiotemporal graph convolutional network to predict the probability and evolution trajectory of abnormal vehicle behavior. Based on the abnormal behavior probability and evolution trajectory, the warning level is calculated through an adaptive risk quantification algorithm. According to the current vehicle density, average speed and road capacity of the road section, a dynamic risk propagation model is used to simulate the risk diffusion coefficient, and the warning sensitivity parameters are corrected in real time. The corrected sensitivity parameters are transmitted to the command platform for multi-dimensional situation deduction, triggering graded warning instructions and generating heat map warning information containing the threat diffusion path.

2. A 5G Internet of Things integrated video surveillance intelligent early warning method according to claim 1, characterized in that: Generates heat map alert information containing threat diffusion paths, including: Build a dynamic threat diffusion model; Based on the dynamic threat diffusion model, the streaming particle swarm algorithm is combined to simulate the motion trajectory of threat particles in the monitoring area, and the spatial distribution of particle trajectories is mapped into threat diffusion probability values ​​through the kernel density estimation algorithm to generate a diffusion probability cloud map; The diffusion probability cloud map is superimposed on a three-dimensional geographic information model, and the threat intensity distribution is represented by a color gradient to generate heat map warning information that marks the key path nodes of the threat diffusion and the estimated arrival time information.

3. A 5G Internet of Things integrated video surveillance intelligent early warning method according to claim 1, characterized in that: The cross-modal attention mechanism formula: ； in, is the attention output result; V is the value vector; S is the source feature; is the query vector generated by V; is the key vector generated by S; is the key vector The transposed matrix of is the key vector Dimensions; is the dot product of the query vector and the key vector; is the scaled dot product score; For Each row of the matrix is ​​normalized.

4. A 5G Internet of Things integrated video surveillance intelligent early warning method according to claim 1, characterized in that: A dynamic behavior prediction model is constructed using a spatiotemporal graph convolutional network to obtain the probability and evolution trajectory of abnormal vehicle behavior, including: Build dynamic behavior prediction models; Based on the dynamic behavior prediction model, intersections and tunnel key nodes are set as core nodes of the graph network. Combined with the micro-features of vehicle braking frequency and steering angle, the spatiotemporal correlation of vehicle behavior is modeled through a spatiotemporal graph convolutional network to predict the probability of abnormal vehicle behavior and the corresponding evolution trajectory. The abnormal vehicle behavior probability is used to quantify the degree to which the vehicle behavior deviates from the normal mode, and the evolution trajectory is used to describe the spatial position change path of the abnormal vehicle behavior in the time dimension.

5. The 5G Internet of Things integrated video surveillance intelligent early warning method according to claim 1 is characterized in that: The dynamic resolution adjustment algorithm formula: ； in, is the dynamically adjusted resolution at time t; is the original base resolution; is the real-time available bandwidth at time t; The required bandwidth for current data transmission; is the data importance weight; The redundant data coefficient is positively correlated with the data duplication; is the network transmission delay at time t.

6. A 5G Internet of Things integrated video surveillance intelligent early warning method according to claim 1, characterized in that: Triggering graded warning instructions, including: A four-level early warning mechanism is set up. When the probability of abnormal behavior is less than 30%, the threat range has no tendency to spread, and the vehicle density and average speed are within the normal range, it is classified as a level one warning, triggering local sound and light prompts, and only the abnormal target location and real-time trajectory are marked. When the probability of abnormal behavior is greater than 30% but less than 50%, the threat has limited diffusion, the vehicle density is high, and the average speed drops slightly, it is classified as a level two warning, and the IoT security equipment cluster is linked to respond in a coordinated manner, and the area with abnormal environmental parameters is marked. When the probability of abnormal behavior is greater than 50% but less than 80%, the threat spreads to the regional level, the road capacity decreases significantly, and the vehicle density increases sharply, it is classified as a level three warning, and augmented reality warning information is pushed to the security center. The risk coordinates are marked by linking the heat map with geographic information, and the nearest patrol force is dispatched. When the probability of abnormal behavior is greater than 80%, the threat has a tendency to spread across regions, the road capacity is seriously insufficient, and there is a risk of traffic paralysis, it is classified as a level four warning, and the emergency command plan is activated. The heat map highlights the core threat area and overlays the distribution of rescue forces.

7. A 5G Internet of Things integrated video surveillance intelligent early warning system, characterized in that: include: Acquisition module for vehicle attribute information and environmental perception data; A fusion module is used to build a 5G IoT perception network based on the vehicle attribute information and environmental perception data. Based on the 5G IoT perception network, the module fuses the vehicle motion trajectory and road environment data through a cross-modal attention mechanism to generate an enhanced environmental perception map. At the same time, a dynamic resolution adjustment algorithm is used to optimize transmission efficiency. The early warning module is used to construct a dynamic behavior prediction model based on the enhanced environmental perception map using a spatiotemporal graph convolutional network to predict the probability and evolution trajectory of abnormal vehicle behavior. Based on the abnormal behavior probability and evolution trajectory, the early warning level is calculated through an adaptive risk quantification algorithm. According to the vehicle density, average speed and road capacity of the current road section, the dynamic risk propagation model is used to simulate the risk diffusion coefficient, and the early warning sensitivity parameters are corrected in real time. The corrected sensitivity parameters are transmitted to the command platform for multi-dimensional situation deduction, triggering graded early warning instructions and generating heat map warning information containing the threat diffusion path.

8. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and runnable on the processor. The processor executes the program to implement a 5G Internet of Things integrated video surveillance intelligent early warning method as described in claims 1-6.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instruction is executed, it implements the 5G Internet of Things integrated video surveillance intelligent early warning method described in claims 1-6.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed, it implements the 5G Internet of Things integrated video surveillance intelligent early warning method described in claims 1-6.

Images