A low-power traffic monitoring method, system and device based on edge computing

Abstract

The present application provides a low-power traffic monitoring method, system and device based on edge computing, belonging to the technical field of traffic control systems. The method obtains road traffic data from multi-source traffic monitoring sensors; based on the road traffic data and a preset device communication distance, determines a corresponding preset range and one or more corresponding edge computing device groups. According to each edge computing device group and its corresponding road traffic data, switches the real-time operation detection models of the edge computing devices in the edge computing device group, so that the switched real-time operation detection models process the corresponding road traffic data; based on the data processing results of the real-time operation detection models, determines whether there are traffic monitoring alarm events at the monitoring positions corresponding to the edge computing devices, and generates monitoring alarm information and sends it to the user terminal when there are traffic monitoring alarm events. Through the above solution, a low-power traffic monitoring with a dynamic balance among energy consumption, efficiency and accuracy is achieved.

Description

Technical Field

[0001] This application relates to the technical field of traffic control systems, and in particular, to a low-power traffic monitoring method, system, and device based on edge computing. Background Art

[0002] With the rapid development of intelligent transportation systems, edge computing devices are increasingly widely used in traffic monitoring. Existing technologies usually deploy edge computing devices along roadsides and use deep learning models such as Convolutional Neural Networks (CNNs) to analyze traffic data in real time to detect abnormal events.

[0003] However, traditional edge computing devices run a single and high-power CNN model (typical power consumption > 15 watts) for a long time and cannot dynamically adjust computing resources according to traffic flow, exacerbating the contradiction between energy consumption and computing power. For example, during low-traffic periods at night, complex models are still running continuously, resulting in a large amount of ineffective energy consumption (accounting for more than 40% of the daily total energy consumption). At the same time, when an accident occurs on a certain section of the road, multiple edge devices have the phenomenon of repeatedly triggering high-precision models, which not only causes computational redundancy but also leads to waste of computing resources.

[0004] Based on this, there is an urgent need for a traffic monitoring technical solution that can achieve dynamic balance among energy consumption, efficiency, and accuracy. Summary of the Invention

[0005] Embodiments of this application provide a low-power traffic monitoring method, system, and device based on edge computing, which are used to solve the technical problems of traffic monitoring that current roadside edge computing devices run a single model, have high power consumption, and have low-efficiency computational redundancy phenomena, and it is difficult to achieve dynamic balance among energy consumption, efficiency, and accuracy.

[0006] On the one hand, embodiments of this application provide a low-power traffic monitoring method based on edge computing, and the method includes:

[0007] Obtain road traffic data from multi-source traffic monitoring sensors;

[0008] Based on the road traffic data and a preset device communication distance, determine a corresponding preset range and one or more edge computing device groups corresponding thereto; wherein, each of the edge computing device groups executes different model switching rules; the model switching rules are at least used for an edge computing device to dynamically switch a real-time running detection model to one of a plurality of different pre-deployed traffic monitoring detection models;

[0009] According to each of the edge computing device groups and the corresponding road traffic data, switch the real-time operation detection models of the edge computing devices in the edge computing device groups, so that the switched real-time operation detection models process the corresponding road traffic data;

[0010] Based on the data processing results of the real-time operation detection models, determine whether there are traffic monitoring warning events at the monitoring positions corresponding to the edge computing devices, so as to generate monitoring warning information and send it to the user terminal when there are such traffic monitoring warning events.

[0011] In an implementation manner of the present application, the multi-source traffic monitoring sensors at least include one or more of the following: lidar sensors, microphone arrays, infrared thermal imagers, and geomagnetic sensors.

[0012] In an implementation manner of the present application, based on the road traffic data and the preset device communication distance, determine the corresponding preset range and one or more edge computing device groups corresponding thereto, specifically including:

[0013] Based on the traffic flow density value, average vehicle speed, and preset congestion conditions in the road traffic data, determine the duration of continuous road congestion; wherein, the preset congestion conditions are used to determine whether the traffic flow density value and the average vehicle speed are both within the corresponding preset congestion determination intervals within a preset time window; the preset congestion determination intervals include a preset congestion density sub-interval and a preset average vehicle speed sub-interval;

[0014] Based on the traffic flow density value, the preset maximum road carrying density value, and the duration of continuous road congestion, calculate the traffic flow saturation coefficient;

[0015] According to the preset device communication distance and the traffic flow saturation coefficient, determine the spatial neighborhood radius corresponding to the edge computing device, and take the location of the edge computing device as the center and the spatial neighborhood radius as the radius to determine the preset range;

[0016] Compare the current traffic flow density values corresponding to the edge computing devices within the preset range with the preset traffic flow density threshold respectively to obtain respective first comparison results, and compare the data processing amount per unit time corresponding to the edge computing devices with the preset data amount threshold respectively to obtain respective second comparison results;

[0017] Generate each of the edge computing device groups according to the respective first comparison results and the respective second comparison results.

[0018] In an implementation manner of the present application, generating each of the edge computing device groups according to the respective first comparison results and the respective second comparison results specifically includes:

[0019] When the current traffic flow density value is greater than or equal to the preset traffic flow density threshold and the amount of data processed per unit time is greater than or equal to the preset data amount threshold, add the corresponding edge computing device to the first edge computing device group;

[0020] When the current traffic flow density value is greater than or equal to the preset traffic flow density threshold or the amount of data processed per unit time is greater than or equal to the preset data amount threshold, add the corresponding edge computing device to the second edge computing device group;

[0021] When the current traffic flow density value is less than the preset traffic flow density threshold and the amount of data processed per unit time is less than the preset data amount threshold, add the corresponding edge computing device to the third edge computing device group.

[0022] In one implementation manner of the present application, according to each of the edge computing device groups and the corresponding road traffic data, switch the real-time operation detection models of the edge computing devices in the edge computing device group, specifically including:

[0023] According to the current traffic flow density value, the preset maximum road carrying density value, the preset minimum road carrying density value, the standard deviation of vehicle speed, and the average vehicle speed respectively corresponding to each first edge computing device in the first edge computing device group, calculate the normalized traffic flow density index corresponding to each first edge computing device;

[0024] According to the traffic flow density index weights respectively corresponding to each first edge computing device, calculate the weighted average value corresponding to each normalized traffic flow density index; wherein, the traffic flow density index weight is positively correlated with the monitored area area corresponding to the first edge computing device;

[0025] When the weighted average value is greater than the first preset threshold, switch the first traffic monitoring detection model to the real-time operation detection model of each first edge computing device, otherwise, switch the second traffic monitoring detection model to the real-time operation detection model;

[0026] When the real-time operation detection model of the first edge computing device group is switched, synchronously switch the corresponding real-time operation detection model of the second edge computing device group;

[0027] Switch each third edge computing device in the third edge computing device group to the sleep mode to disable the corresponding real-time operation detection model.

[0028] In one implementation manner of the present application, the method further includes:

[0029] Before switching the real-time operation detection model of the first edge computing device group, calculate the switching decision weights corresponding to each of the first edge computing devices respectively according to the current traffic flow density values and the corresponding device interval distances of the first edge computing devices, and compare the sum value of the switching decision weights with a second preset threshold;

[0030] When the sum value is greater than the second preset threshold, determine that the first edge computing device group meets the switching condition.

[0031] In an implementation manner of the present application, the multiple different traffic monitoring detection models at least include a spiking neural network (SNN) model and a convolutional neural network (CNN) model;

[0032] The method further includes:

[0033] When the real-time operation detection model of the edge computing device is an SNN model, determine whether there is the traffic monitoring warning event in the edge computing device; the traffic monitoring warning event at least includes one or more of the following: sudden braking of a vehicle, speeding, traffic accident, abnormal parking, sudden congestion;

[0034] If so, switch the real-time operation detection model of the edge computing device to a CNN model, and continuously run the CNN model within a preset time period.

[0035] In an implementation manner of the present application, a microchannel cooling structure and an integrated Peltier effect temperature control module are embedded in the chip of the edge computing device; the method further includes:

[0036] When the chip temperature of the edge computing device is greater than a first temperature threshold, generate a first temperature control signal to increase the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the first temperature control signal;

[0037] When the chip temperature of the edge computing device is less than a second temperature threshold, generate a second temperature control signal to reduce the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the second temperature control signal.

[0038] On the other hand, an embodiment of the present application further provides a low-power traffic monitoring system based on edge computing, and the system includes:

[0039] An acquisition module, configured to acquire road traffic data from multi-source traffic monitoring sensors;

[0040] A first determination module, configured to determine a corresponding preset range and one or more corresponding edge computing device groups based on the road traffic data and a preset device communication distance; wherein, each of the edge computing device groups executes different model switching rules; the model switching rules are at least used for an edge computing device to dynamically switch a real-time running detection model to one of a plurality of pre-deployed different traffic monitoring detection models.

[0041] A switching module, configured to switch the real-time running detection models of each edge computing device in the edge computing device group according to each of the edge computing device groups and the corresponding road traffic data, so that the switched real-time running detection model processes the corresponding road traffic data.

[0042] A second determination module, configured to determine whether there is a traffic monitoring alarm event at the monitoring position corresponding to each edge computing device based on the data processing result of the real-time running detection model, and generate a monitoring alarm message and send it to a user terminal when there is the traffic monitoring alarm event.

[0043] On the other hand, an embodiment of the present application further provides a low-power traffic monitoring device based on edge computing, and the device includes:

[0044] At least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a low-power traffic monitoring method based on edge computing as described above.

[0045] Compared with the prior art, the present application has the following remarkable effects:

[0046] Through the above solution, the present application completes a dynamic model switching strategy according to dynamic parameters such as traffic flow density and vehicle speed, and can effectively achieve the balance between high precision and low power consumption; at the same time, based on the preset device communication distance and road traffic data, an edge computing device group is constructed to realize the optimization of task allocation and resource scheduling, and improve the device utilization rate and data processing efficiency; the present application can also quickly identify traffic monitoring alarm events, and ensure the event response time and alarm accuracy rate. Furthermore, through the above dynamic model switching, device collaborative optimization and real-time alarm generation, the problems of high energy consumption, low efficiency and insufficient flexibility in traditional traffic monitoring methods are solved, and high-precision, low-power and high-efficiency traffic monitoring is realized, providing reliable technical support for the intelligent transportation system. Description of the Drawings

[0047] The accompanying drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0048] Figure 1 It is a schematic flowchart of a low-power traffic monitoring method based on edge computing in an embodiment of the present application;

[0049] Figure 2 It is a schematic structural diagram of a low-power traffic monitoring system based on edge computing in an embodiment of the present application;

[0050] Figure 3 It is a schematic structural diagram of a low-power traffic monitoring device based on edge computing in an embodiment of the present application. Detailed implementation manners

[0051] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0052] The embodiments of the present application provide a low-power traffic monitoring method, system, and device based on edge computing, which are used to solve the technical problems of the current road edge computing device having a single operation model, high power consumption, and the existence of inefficient computing redundancy, and it is difficult to achieve a dynamic balance among energy consumption, efficiency, and accuracy in traffic monitoring.

[0053] The following will describe each embodiment of the present application in detail with reference to the drawings.

[0054] The embodiments of the present application provide a low-power traffic monitoring method based on edge computing. As Figure 1 shown, the method may include steps S101-S104:

[0055] S101, the edge computing device cluster obtains road traffic data from multi-source traffic monitoring sensors.

[0056] It should be noted that the edge computing device cluster, as the execution entity of the low-power traffic monitoring method based on edge computing, is a distributed computing network composed of multiple edge computing devices set on the road.

[0057] The multi-source traffic monitoring sensors include, but are not limited to, lidar sensors, microphone arrays, infrared thermal imagers, and geomagnetic sensors. Among them, lidar sensors can be used to generate three-dimensional contour features of vehicles; microphone arrays can collect voiceprint features of road sounds for identifying vehicle types and abnormal sounds; infrared thermal imagers can detect the heat source distribution of vehicles; and geomagnetic sensors can obtain vehicle passing frequencies, vehicle passing speeds, etc.

[0058] After each edge computing device in the edge computing device cluster receives the monitoring data from its corresponding multi-source traffic monitoring sensor through its sensor interface, it takes each monitoring data as road traffic data.

[0059] S102. The edge computing device cluster determines a corresponding preset range and one or more corresponding edge computing device groups based on the road traffic data and the preset device communication distance.

[0060] Among them, each edge computing device group executes different model switching rules. The model switching rules are at least used for the edge computing device to dynamically switch the real-time running detection model to one of multiple pre-deployed different traffic monitoring detection models.

[0061] In the embodiment of the present application, the above determination of the corresponding preset range and one or more corresponding edge computing device groups based on the road traffic data and the preset device communication distance specifically includes:

[0062] Based on the traffic flow density value, average vehicle speed, and preset congestion condition in the road traffic data, determine the duration of continuous road congestion. Among them, the preset congestion condition is used to determine whether the traffic flow density value and the average vehicle speed are both within the corresponding preset congestion determination interval within a preset time window. The preset congestion determination interval includes a preset congestion density sub-interval and a preset average vehicle speed sub-interval. Calculate the traffic flow saturation coefficient based on the traffic flow density value, the preset maximum road carrying density value, and the duration of continuous road congestion. Determine the spatial neighborhood radius corresponding to the corresponding edge computing device according to the preset device communication distance and the traffic flow saturation coefficient, and take the location of the edge computing device as the center and the spatial neighborhood radius as the radius to determine the preset range. Compare the current traffic flow density value corresponding to each edge computing device within the preset range with the preset traffic flow density threshold to obtain each first comparison result, and compare the data processing amount per unit time corresponding to each edge computing device with the preset data amount threshold to obtain each second comparison result. Generate each edge computing device group according to each first comparison result and each second comparison result.

[0063] In other words, for each edge computing device in the edge computing device cluster, data sub-items such as traffic flow density values and average vehicle speeds in the road traffic data are extracted respectively, and the continuous road congestion duration in the monitoring area of each edge computing device is calculated respectively. Among them, the traffic flow density value can be obtained by dividing the number of vehicles detected by the device by the area of the monitoring area of the edge computing device.

[0064] Among them, the calculation of the continuous road congestion duration is performed using a preset congestion condition. Specifically, it can be: through a preset time window with a preset time window length such as 1 minute, it is judged whether the traffic flow density value continuously stays within a preset congestion density sub-interval and whether the average vehicle speed continuously stays within a preset average vehicle speed sub-interval within this preset time window. If both conditions are met, this preset time window meets the preset congestion condition. At this time, the continuous road congestion duration is accumulated as a preset time window length, that is, 1 minute. Subsequently, the number of time windows that continuously meet the preset congestion condition is accumulated. If the next preset time window does not meet the preset congestion condition, the continuous road congestion duration is obtained according to the product value of the number of time windows that meet the preset congestion condition and the time window length before.

[0065] For example, if the time window length of the preset time window is 1 minute, the preset congestion density sub-interval is ( ), and the preset average vehicle speed sub-interval is ( ), where represents 80% of the preset maximum road carrying density value , and represents 50% of the preset road indicated speed . If within the preset time window, the continuous traffic flow density value and average vehicle speed meet: the traffic flow density value and the average vehicle speed , it is determined that the road corresponding to the monitoring area of the corresponding edge computing device is in a congested state, and the congestion duration is accumulated by one time window length, that is, 1 minute; if the next preset time window still meets the above preset congestion condition, on the basis of the above 1 minute, another time window length is accumulated. At this time, the continuous road congestion duration is 2 minutes. The judgment process of this preset congestion condition continues until the next preset time window does not meet the above preset congestion condition, and the previously accumulated congestion duration is used as the continuous road congestion duration. In addition, if the edge computing device cluster calculates that a certain edge computing device still meets the preset congestion condition until the current time, the currently accumulated congestion duration is used as the continuous road congestion duration.

[0066] After obtaining the continuous road congestion duration, the present application also calculates the traffic flow saturation coefficient through the traffic flow density value, the preset maximum road carrying density value, and the continuous road congestion duration. This traffic flow saturation coefficient is used to characterize the traffic flow saturation degree of the current road and dynamically evaluate the traffic state. The formula is as follows:

[0067]

[0068] Among them, represents the traffic flow saturation coefficient of the th edge computing device, represents the traffic flow density value of the th edge computing device, represents the preset maximum road carrying density value of the th edge computing device, represents the duration of continuous road congestion corresponding to the th edge computing device. It is used to evaluate the immediate congestion degree of the road, and is used to reflect the influence degree of the duration of continuous road congestion on the traffic state. The duration and the influence degree are in a positive correlation. Among them, the preset maximum road carrying density value and the preset minimum road carrying density value can be preset by the user, and the present application does not make specific limitations thereon.

[0069] Furthermore, the present application will also calculate the spatial neighborhood radius according to the preset device communication distance and the calculated traffic flow saturation coefficient above, so as to perform clustering processing on the edge computing devices. The spatial neighborhood radius is as follows: , where represents the preset device communication distance of the th edge computing device, which can be stored and shared in the edge computing device cluster; represents the spatial neighborhood radius of the th edge computing device. Through this spatial neighborhood radius and the location of the edge computing device, a preset range is generated.

[0070] Then, the present application compares the current traffic flow density value of each edge computing device within the preset range with the preset traffic flow density threshold, and compares the data volume processed per unit time with the preset data volume threshold. Among them, the preset traffic flow density threshold and the preset data volume threshold can be set by the user according to the actual usage scenario, and the present application does not make specific limitations thereon. The data volume processed per unit time can be obtained by the ratio of the sum of the sizes of each data packet processed by the edge computing device within a unit time to the unit time. The data packet size can be megabytes or gigabytes.

[0071] In an embodiment of the present application, generating each edge computing device group according to each first comparison result and each second comparison result specifically includes:

[0072] When the current traffic density value is greater than or equal to the preset traffic density threshold and the amount of data processed per unit time is greater than or equal to the preset data amount threshold, the corresponding edge computing device is added to the first edge computing device group.

[0073] When the current traffic density value is greater than or equal to the preset traffic density threshold or the amount of data processed per unit time is greater than or equal to the preset data amount threshold, the corresponding edge computing device is added to the second edge computing device group.

[0074] When the current traffic density value is less than the preset traffic density threshold and the amount of data processed per unit time is less than the preset data amount threshold, the corresponding edge computing device is added to the third edge computing device group.

[0075] That is to say, this application clusters and groups edge computing devices according to the traffic density value and the amount of data processed per unit time, divides them into three different edge computing device groups, and executes different model switching rules. By grouping edge computing devices, the edge computing devices can be uniformly scheduled to complete model switching. On the one hand, it avoids the consumption of computing resources by controlling each edge computing device separately, increasing system energy consumption and communication overhead; on the other hand, it enables edge computing devices to form a unified monitoring network during traffic monitoring, enhance the linkage between devices, and provide equipment utilization while ensuring the accuracy of event detection.

[0076] S103, the edge computing device cluster switches the real-time operation detection model of each edge computing device in the edge computing device group according to each edge computing device group and its corresponding road traffic data, so that the switched real-time operation detection model processes the corresponding road traffic data.

[0077] In the embodiment of the present application, the real-time operation detection model of each edge computing device in the edge computing device group is switched according to each edge computing device group and its corresponding road traffic data, specifically including:

[0078] According to the current traffic flow density values, preset maximum road carrying density values, preset minimum road carrying density values, vehicle speed standard deviations, and average vehicle speeds corresponding to the first edge computing devices in the first group of edge computing devices, calculate the normalized traffic flow density indices corresponding to the first edge computing devices respectively. According to the traffic flow density index weights corresponding to the first edge computing devices respectively, calculate the weighted average values corresponding to the normalized traffic flow density indices. Among them, the traffic flow density index weight is positively correlated with the monitored area of the first edge computing device. When the weighted average value is greater than the first preset threshold, switch the first traffic monitoring detection model to the real-time operation detection model for each first edge computing device; otherwise, switch the second traffic monitoring detection model to the real-time operation detection model. When the first group of edge computing devices switches the real-time operation detection model, synchronously switch the corresponding real-time operation detection model of the second group of edge computing devices. Switch each third edge computing device in the third group of edge computing devices to the sleep mode to disable the corresponding real-time operation detection model.

[0079] In other words, after obtaining the above different groups of edge computing devices, for the first group of edge computing devices, the edge computing device cluster calculates the normalized traffic flow density index through the following formula:

[0080]

[0081] Among them, represents the normalized traffic flow density index of the th edge computing device, represents the preset minimum road carrying density value of the th edge computing device, represents the vehicle speed standard deviation of the th edge computing device, represents the average vehicle speed monitored by the th edge computing device.

[0082] Subsequently, calculate the weighted average value of each normalized traffic flow density index corresponding to the first group of edge computing devices :

[0083]

[0084] Among them, is the traffic flow density index weight of the th edge computing device, , is the monitored area of the th edge computing device, is the total number of edge computing devices in the first group of edge computing devices. Among them, , .

[0085] Subsequently, the edge computing device cluster compares the above weighted average value with a first preset threshold, which is set according to the actual usage scenario and is not specifically limited herein. If the weighted average value is greater than the first preset threshold, the real-time operation detection model of each first edge computing device will be switched to the first traffic monitoring detection model at this time; otherwise, it will be switched to the second traffic monitoring detection model. Among them, the multiple different traffic monitoring detection models at least include a Spiking Neural Network (SNN) model and a Convolutional Neural Network (CNN) model. The first traffic monitoring detection model is the CNN model, and the second traffic monitoring detection model is the SNN model. The CNN model has a higher operating power consumption than the SNN model, but the CNN model can perform real-time and high-precision data processing and analyze more complex events. By flexibly switching between the CNN model and the SNN model, it can not only ensure effective traffic monitoring but also reduce device energy consumption.

[0086] For the second group of edge computing devices, it follows the first group of edge computing devices to switch the model. If the first group of edge computing devices has completed the switching of the real-time operation detection model, the second group of edge computing devices will synchronously switch to the same real-time operation detection model as the first group of edge computing devices, keeping the real-time operation detection model consistent with the first group of edge computing devices.

[0087] For the third group of edge computing devices, each third edge computing device will be put into the sleep mode, that is, in low-power standby, without executing the SNN model or the CNN model.

[0088] Through the above solution, different model switching rules can be used to manage different groups of edge computing devices, achieving more energy-efficient and effective traffic monitoring.

[0089] In addition, when the real-time operation detection model of each of the above edge computing devices is the SNN model, it is determined whether there is a traffic monitoring alarm event for the edge computing device. The traffic monitoring alarm event at least includes one or more of the following: sudden braking of a vehicle, speeding, traffic accident, abnormal parking, and sudden congestion. If so, the real-time operation detection model of the edge computing device will be switched to the CNN model and the CNN model will continue to run within a preset period.

[0090] That is to say, if a traffic monitoring alarm event occurs in the monitoring area monitored by the edge computing device, in order to better achieve monitoring, the edge computing device will maintain the CNN model within a preset period to accurately monitor traffic anomalies. The preset period can be set by the user according to the actual usage scenario and is not specifically limited in this application.

[0091] In another embodiment of the present application, for the above-mentioned third edge computing device group, when it maintains the sleep mode, it is possible to determine whether to execute the SNN model according to the change rate of the real-time vehicle density value of the third edge computing device. For example, if the change rate of the real-time vehicle density value within 30 minutes is 0.5, which is greater than a preset change rate threshold of 0.3, at this time, the real-time operation detection model of the third edge computing device will be switched to the SNN model to detect traffic monitoring warning events.

[0092] In yet another embodiment of the present application, to avoid frequent model switching of the first edge computing device group and balance reasonable power consumption and monitoring accuracy, the following embodiments are also included:

[0093] Before switching the real-time operation detection model of the first edge computing device group, according to the current traffic flow density value and the corresponding device interval distance of each first edge computing device, calculate the switching decision weight corresponding to each first edge computing device respectively, and compare the sum value of each switching decision weight with a second preset threshold. When the sum value is greater than the second preset threshold, it is determined that the first edge computing device group meets the switching condition.

[0094] In other words, before performing model switching, the edge computing device cluster will calculate the switching decision weight for each first edge computing device, and the formula is as follows:

[0095]

[0096] Among them, is the switching decision weight of the th first edge computing device, is the node set composed of each first edge computing device in the first edge computing device group, represents the traffic flow density value of the th first edge computing device, is the th first edge computing device and the th first edge computing device. The actual physical distance between them. The actual physical distance can be preset by the user or calculated as the Euclidean distance according to the position coordinates of the edge computing device. The present application does not make specific limitations on this. is the distance attenuation factor, which reflects the degree of influence attenuation of nodes affected by the actual physical distance.

[0097] After obtaining the switching decision weights of the above-mentioned edge computing devices, add up the switching decision weights of each device to calculate the sum value, and compare the size of the sum value with a preset second preset threshold. If the sum value is greater than the second preset threshold, it indicates that the switching condition is met, and the model switching of the first edge computing device group can be performed as described above; otherwise, no model switching is performed. Through the above solution, regional model switching of the first edge computing device group can be achieved, and high-precision and low-power optimization of traffic monitoring can be better realized.

[0098] S104. Based on the data processing results of the real-time operation detection model, the edge computing device cluster determines whether there is a traffic monitoring warning event at the monitoring location corresponding to each edge computing device, and generates a monitoring warning message and sends it to the user terminal in the case of a traffic monitoring warning event.

[0099] That is to say, the edge computing device can perform traffic monitoring on the monitoring area through the real-time operation detection model, and when a traffic monitoring warning event is detected, generate a notification of the existence of the traffic monitoring warning event and send it to the edge computing device cluster. The edge computing device cluster further generates a monitoring warning message and sends it to the user terminal. The user terminal can be understood as devices such as the mobile phones and computers of the traffic management department or traffic police, and this application does not make specific limitations on this.

[0100] In an embodiment of the present application, a microchannel cooling structure and an integrated Peltier effect temperature control module are also embedded in the chip of the above-mentioned edge computing device.

[0101] Specifically, a microchannel cooling structure is embedded in the chip packaging layer of the edge computing device. The microchannel includes: an inlet channel connected to the coolant supply system for introducing coolant; branch channels distributed in the chip heat source area for uniform heat dissipation; and an outlet channel connected to the coolant recovery system for discharging coolant. A Peltier effect temperature control module is integrated in the heat source area of the chip packaging layer. The module may include thermocouple pairs formed by alternating P-type semiconductors and N-type semiconductors; a cold end in contact with the chip heat source area for absorbing heat; and a hot end in contact with the microchannel cooling structure for releasing heat.

[0102] In addition, the edge computing device may further include a coolant circulation system, including: a coolant pump for driving the coolant to circulate in the microchannel; a radiator for cooling the coolant flowing through the outlet channel; and a temperature sensor for real-time monitoring of the coolant temperature and feedback control of the coolant flow rate.

[0103] In an embodiment of the present application, when the chip temperature of the edge computing device is greater than the first temperature threshold, a first temperature control signal is generated to increase the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the first temperature control signal. When the chip temperature of the edge computing device is less than the second temperature threshold, a second temperature control signal is generated to reduce the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the second temperature control signal.

[0104] Through the collaborative work of the above microchannel cooling structure and the Peltier effect temperature control module, efficient heat dissipation of the chip can be carried out; also based on the active temperature control algorithm of the above real-time temperature, precise temperature control of the chip can be achieved. When paired with solar power supply, the computing power of the chip can be more effectively maintained, realizing low-power traffic monitoring.

[0105] Through the above solution of the present application, a dynamic model switching strategy is completed according to dynamic parameters such as traffic flow density and vehicle speed, which can effectively achieve the balance between high precision and low power consumption; at the same time, based on the preset device communication distance and road traffic data, an edge computing device group is constructed to optimize task allocation and resource scheduling, increasing the device utilization rate and data processing efficiency; the present application can also quickly identify traffic monitoring alarm events, ensuring the event response time and alarm accuracy rate. Furthermore, through the above dynamic model switching, device collaborative optimization and real-time alarm generation, the problems of high energy consumption, low efficiency and lack of flexibility in traditional traffic monitoring methods are solved, realizing high-precision, low-power and high-efficiency traffic monitoring, and providing reliable technical support for the intelligent transportation system.

[0106] Figure 2 It is a schematic structural diagram of a low-power traffic monitoring system based on edge computing provided by an embodiment of the present application, as Figure 2 shown, the low-power traffic monitoring system 200 based on edge computing includes:

[0107] An acquisition module 201 is configured to acquire road traffic data from multi-source traffic monitoring sensors. A first determination module 202 is configured to determine a corresponding preset range and one or more edge computing device groups corresponding thereto based on the road traffic data and a preset device communication distance. Each of the edge computing device groups executes different model switching rules. The model switching rule is at least used for an edge computing device to dynamically switch a real-time running detection model to one of a plurality of pre-deployed different traffic monitoring detection models. A switching module 203 is configured to switch the real-time running detection models of the edge computing devices in the edge computing device group according to each edge computing device group and the corresponding road traffic data, so that the switched real-time running detection model processes the corresponding road traffic data. A second determination module 204 is configured to determine whether there is a traffic monitoring warning event at the monitoring position corresponding to each edge computing device based on the data processing result of the real-time running detection model, so as to generate a monitoring warning message and send it to a user terminal when there is a traffic monitoring warning event.

[0108] Figure 3 FIG. is a schematic structural diagram of a low-power traffic monitoring device based on edge computing provided by an embodiment of the present application, as Figure 3 shown, the device includes:

[0109] At least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to:

[0110] Acquire road traffic data from multi-source traffic monitoring sensors. Determine a corresponding preset range and one or more edge computing device groups corresponding thereto based on the road traffic data and a preset device communication distance. Each of the edge computing device groups executes different model switching rules. The model switching rule is at least used for an edge computing device to dynamically switch a real-time running detection model to one of a plurality of pre-deployed different traffic monitoring detection models. Switch the real-time running detection models of the edge computing devices in the edge computing device group according to each edge computing device group and the corresponding road traffic data, so that the switched real-time running detection model processes the corresponding road traffic data. Determine whether there is a traffic monitoring warning event at the monitoring position corresponding to each edge computing device based on the data processing result of the real-time running detection model, so as to generate a monitoring warning message and send it to a user terminal when there is a traffic monitoring warning event.

[0111] Each embodiment in this application is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system and device embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiments.

[0112] The systems and devices provided in the embodiments of this application correspond one by one to the methods. Therefore, the systems and devices also have beneficial technical effects similar to those of their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and devices will not be elaborated here.

[0113] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the said element.

[0114] The above description is only for the embodiments of this application and is not intended to limit this application. For those skilled in the art, various changes and modifications can be made to this application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the scope of the claims of this application.

Claims

1. A low-power traffic monitoring method based on edge computing, characterized in that: The method comprises: Obtain road traffic data from multi-source traffic monitoring sensors; Based on the road traffic data and the preset device communication distance, determine the corresponding preset range and its corresponding one or more edge computing device groups; wherein each edge computing device group executes different model switching rules respectively; the model switching rules are at least used for the edge computing device to dynamically switch the real-time running detection model to one of the multiple different traffic monitoring detection models deployed in advance; According to each edge computing device group and the corresponding road traffic data, switching the real-time operation detection model of each edge computing device in the edge computing device group, so that the switched real-time operation detection model processes the corresponding road traffic data; Based on the data processing result of the real-time running detection model, determine whether there is a traffic monitoring alarm event at the monitoring location corresponding to each edge computing device, so as to generate monitoring alarm information and send it to the user terminal when the traffic monitoring alarm event exists; Wherein, based on the road traffic data and the preset device communication distance, determining the corresponding preset range and its corresponding one or more edge computing device groups specifically includes: Based on the traffic density value and average vehicle speed in the road traffic data and the preset congestion condition, the duration of continuous road congestion is determined; wherein the preset congestion condition is used to determine whether the traffic density value and the average vehicle speed are both within a corresponding preset congestion determination interval within a preset time window; the preset congestion determination interval includes a preset congestion density subinterval and a preset average vehicle speed subinterval; Calculating a traffic flow saturation coefficient based on the traffic density value, a preset maximum road load density value, and the duration of continuous congestion on the road; Determine the spatial neighborhood radius corresponding to the edge computing device according to the preset device communication distance and the traffic flow saturation coefficient, and determine the preset range with the location of the edge computing device as the center and the spatial neighborhood radius as the radius; Compare the current traffic density value corresponding to each edge computing device within the preset range with the preset traffic density threshold to obtain each first comparison result, and compare the data volume processed per unit time corresponding to each edge computing device with the preset data volume threshold to obtain each second comparison result; Each of the edge computing device groups is generated according to each of the first comparison results and each of the second comparison results.

2. A low-power traffic monitoring method based on edge computing according to claim 1, characterized in that: The multi-source traffic monitoring sensor includes at least one or more of the following: a lidar sensor, a microphone array, an infrared thermal imager, and a geomagnetic sensor.

3. A low-power traffic monitoring method based on edge computing according to claim 1, characterized in that: Generating each of the edge computing device groups according to each of the first comparison results and each of the second comparison results specifically includes: When the current traffic density value is greater than or equal to the preset traffic density threshold and the amount of data processed per unit time is greater than or equal to the preset data amount threshold, adding the corresponding edge computing device to the first edge computing device group; When the current traffic density value is greater than or equal to the preset traffic density threshold or the amount of data processed per unit time is greater than or equal to the preset data amount threshold, adding the corresponding edge computing device to the second edge computing device group; When the current traffic density value is less than the preset traffic density threshold and the amount of data processed per unit time is less than the preset data amount threshold, the corresponding edge computing device is added to the third edge computing device group.

4. A low-power traffic monitoring method based on edge computing according to claim 3, characterized in that: According to each edge computing device group and the corresponding road traffic data, switching the real-time operation detection model of each edge computing device in the edge computing device group specifically includes: Calculate the normalized traffic density index corresponding to each first edge computing device in the first edge computing device group according to the current traffic density value, the preset road maximum load density value, the preset road minimum load density value, the vehicle speed standard deviation and the average vehicle speed respectively corresponding to each first edge computing device in the first edge computing device group; According to the vehicle flow density index weights corresponding to the first edge computing devices, a weighted average value corresponding to each normalized vehicle flow density index is calculated; wherein the vehicle flow density index weight is positively correlated with the monitoring area corresponding to the first edge computing device; When the weighted average value is greater than a first preset threshold, switching the first traffic monitoring detection model to the real-time operation detection model of each of the first edge computing devices; otherwise, switching the second traffic monitoring detection model to the real-time operation detection model; When the first edge computing device group switches the real-time operation detection model, synchronously switching the corresponding real-time operation detection model of the second edge computing device group; Switch each third edge computing device of the third edge computing device group to a sleep mode to disable the corresponding real-time running detection model.

5. A low-power traffic monitoring method based on edge computing according to claim 4, characterized in that: The method further comprises: Before switching the real-time operation detection model of the first edge computing device group, respectively calculating the switching decision weights corresponding to each of the first edge computing devices according to the current traffic density value and the corresponding device interval distance of each of the first edge computing devices, and comparing the sum of the switching decision weights with a second preset threshold; When the sum is greater than the second preset threshold, it is determined that the first edge computing device group meets the switching condition.

6. A low-power traffic monitoring method based on edge computing according to any one of claims 1 to 5, characterized in that: The multiple different traffic monitoring detection models include at least a spike neural network SNN model and a convolutional neural network CNN model; The method further comprises: When the real-time operation detection model of the edge computing device is an SNN model, determining whether the edge computing device has the traffic monitoring alarm event; the traffic monitoring alarm event includes at least one or more of the following: vehicle emergency braking, speeding, traffic accidents, abnormal parking, and sudden congestion; If so, switch the real-time running detection model of the edge computing device to a CNN model, and continue to run the CNN model within a preset period of time.

7. A low-power traffic monitoring method based on edge computing according to claim 1, characterized in that: The chip of the edge computing device is embedded with a microchannel cooling structure and an integrated Peltier effect temperature control module; the method also includes: When the chip temperature of the edge computing device is greater than a first temperature threshold, a first temperature control signal is generated to increase the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the first temperature control signal; When the chip temperature of the edge computing device is less than a second temperature threshold, a second temperature control signal is generated to reduce the coolant flow rate of the microchannel cooling structure and the driving voltage of the Peltier effect temperature control module through the second temperature control signal.

8. A low-power traffic monitoring system based on edge computing, characterized in that: The system comprises: An acquisition module, used to acquire road traffic data from multi-source traffic monitoring sensors; A first determination module is used to determine a corresponding preset range and one or more edge computing device groups corresponding thereto based on the road traffic data and the preset device communication distance; wherein each of the edge computing device groups respectively executes a different model switching rule; and the model switching rule is at least used for the edge computing device to dynamically switch the real-time running detection model to one of a plurality of different traffic monitoring detection models deployed in advance; A switching module, configured to switch the real-time operation detection model of each edge computing device in the edge computing device group according to each edge computing device group and the corresponding road traffic data, so that the switched real-time operation detection model processes the corresponding road traffic data; A second determination module is used to determine whether there is a traffic monitoring alarm event at the monitoring location corresponding to each edge computing device based on the data processing result of the real-time operation detection model, so as to generate monitoring alarm information and send it to the user terminal when the traffic monitoring alarm event exists; Wherein, the first determining module is specifically used for: Based on the traffic density value and average vehicle speed in the road traffic data and the preset congestion condition, the duration of continuous road congestion is determined; wherein the preset congestion condition is used to determine whether the traffic density value and the average vehicle speed are both within a corresponding preset congestion determination interval within a preset time window; the preset congestion determination interval includes a preset congestion density subinterval and a preset average vehicle speed subinterval; Calculating a traffic flow saturation coefficient based on the traffic density value, a preset maximum road load density value, and the duration of continuous congestion on the road; Determine the spatial neighborhood radius corresponding to the edge computing device according to the preset device communication distance and the traffic flow saturation coefficient, and determine the preset range with the location of the edge computing device as the center and the spatial neighborhood radius as the radius; Compare the current traffic density value corresponding to each edge computing device within the preset range with the preset traffic density threshold to obtain each first comparison result, and compare the data volume processed per unit time corresponding to each edge computing device with the preset data volume threshold to obtain each second comparison result; Each of the edge computing device groups is generated according to each of the first comparison results and each of the second comparison results.

9. A low-power traffic monitoring device based on edge computing, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a low-power traffic monitoring method based on edge computing as described in any one of claims 1 to 7 above.

Images