Municipal road monitoring method and system based on edge computing technology

Abstract

The invention discloses a municipal road monitoring method and system based on an edge computing technology, and relates to the technical field of signal processing. The method comprises the steps that road edge nodes collect data through optical, vibration and infrared sensors, and the pavement crack index and the real-time traffic flow are calculated in parallel; fusing environment temperature and humidity data, weighting to generate a road health initial judgment value, and triggering a dynamic threshold alarm; exchanging initial judgment values at adjacent nodes through a low-power-consumption Mesh network, jointly calculating the regional-level traffic flow density and road surface damage association degree, and generating an encrypted regional road state feature vector; the central management platform aggregates the regional feature vectors, calculates a road comprehensive health score and an early warning weight coefficient in combination with a historical baseline, generates a self-adaptive dredging strategy and issues the self-adaptive dredging strategy to an edge node; according to node load and energy consumption data, the sampling frequency of the sensor is dynamically adjusted, and the crack recognition model is updated. According to the invention, the road state monitoring real-time performance and the early warning accuracy are obviously improved, and the data transmission energy consumption is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation infrastructure, and particularly to a municipal road monitoring method and system based on edge computing technology. Background Art

[0002] Existing municipal road monitoring systems mainly rely on a centralized cloud computing architecture, where sensors transmit raw data to a remote server for centralized analysis and processing.

[0003] This mode has significant technical bottlenecks: First, the large-scale data backhaul leads to a sharp increase in network bandwidth pressure, and the data processing delay of the central platform often reaches the minute level, making it difficult to meet the real-time early warning requirements of road surface conditions; Second, road monitoring nodes usually operate independently, lacking a collaborative analysis mechanism between adjacent regions, and unable to capture the spatial correlation laws of traffic flow density and road surface damage, resulting in a passive response only when local risks evolve into regional congestion; Third, the sensor sampling frequency and computing resources of edge nodes adopt fixed configurations, which are prone to data loss due to overload during peak hours, while causing energy waste during off-peak hours.

[0004] Although there have been attempts to apply edge computing to road monitoring, there are still bottlenecks such as the lack of regional collaborative computing and the blank of dynamic resource scheduling mechanism, and it is urgent to build a closed-loop optimization system with three-level linkage of "edge-region-center". Summary of the Invention

[0005] The present invention proposes a municipal road monitoring method based on edge computing technology, including: Edge computing nodes deployed on municipal roads collect multi-source monitoring data through optical, vibration, and infrared sensors, and calculate the road surface crack index and real-time traffic volume in parallel; Based on the road surface crack index and real-time traffic volume, and integrating environmental temperature and humidity data, edge computing nodes on municipal roads generate a preliminary judgment value of road health by weighting and trigger a dynamic threshold alarm.

[0006] The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculating the regional traffic flow density value and the correlation degree value of road surface damage, and generating an encrypted regional road state feature vector.

[0007] The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with historical baselines, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes.

[0008] Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency and incrementally updates the crack recognition model parameters.

[0009] A municipal road monitoring method based on edge computing technology as described above, wherein edge computing nodes deployed on municipal roads collect multi-source monitoring data through optical, vibration, and infrared sensors, and perform parallel calculations on the road surface crack index and real-time traffic flow, including the following sub-steps: Use an optical sensor to capture road surface image data, identify crack features through a convolutional neural network, and output the road surface crack index; Based on the rut track data and heat map distribution synchronously collected by vibration sensors and infrared sensors, calculate the real-time traffic flow in combination with the time window statistical model.

[0010] A municipal road monitoring method based on edge computing technology as described above, wherein, based on the road surface crack index and real-time traffic flow, fusing environmental temperature and humidity data, the edge computing nodes of the municipal road generate a preliminary judgment value of road health by weighting and trigger a dynamic threshold alarm, including the following sub-steps: Input the road surface crack index, real-time traffic flow, and environmental temperature and humidity data into a weighted fusion model, and calculate the preliminary judgment value of road health according to the preset weight coefficients; Dynamically adjust the alarm trigger threshold according to the mutation amplitude of the real-time traffic flow and the growth rate of the road surface crack index.

[0011] A municipal road monitoring method based on edge computing technology as described above, wherein the preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculate the regional traffic flow density value and the pavement damage correlation value, and generate an encrypted regional road state feature vector, including the following sub-steps: Adjacent edge computing nodes exchange the preliminary judgment value of road health through a low-power Mesh network, and calculate the regional traffic flow density value using a spatio-temporal correlation algorithm; Generate a pavement damage correlation value based on the spatial distribution characteristics of the multi-node road surface crack index; Perform encryption encoding on the traffic flow density value and the pavement damage correlation value to generate a regional road state feature vector.

[0012] A municipal road monitoring method based on edge computing technology as described above, wherein the central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and the early warning weight coefficient in combination with the historical baseline, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes, including the following sub-steps: Perform dimensionality reduction processing on the aggregated regional road state feature vectors to extract key road state parameters; Compare the key road state parameters with the historical baseline data, and calculate the comprehensive road health score and the early warning weight coefficient; Generate an adaptive guidance strategy including detour route planning and speed limit instructions according to the early warning weight coefficient.

[0013] A municipal road monitoring method based on edge computing technology as described above, wherein, based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack identification model parameters, including the following sub-steps: Calculate the resource allocation coefficient according to the periodic characteristics of the load peak and energy consumption curve; Scale the image acquisition frequency of the optical sensor proportionally based on the resource allocation coefficient, and reduce the sampling resolution of the vibration sensor; Incrementally update the convolutional neural network weight parameters according to the inference delay of the crack identification model at the edge node.

[0014] A municipal road monitoring method based on edge computing technology as described above, wherein, according to the inference delay of the crack identification model at the edge node, incrementally update the convolutional neural network weight parameters, including the following sub-steps: Construct an incremental training set from the crack false alarm samples reported by the edge computing node and the new crack feature data; Fine-tune the convolutional layer parameters of the crack identification model using transfer learning algorithm, and retain the fully connected layer structure of the original model; Distribute the updated model parameters to the corresponding edge computing nodes through an encrypted channel.

[0015] The present invention also proposes a municipal road monitoring system based on edge computing technology, including: Information acquisition and preprocessing module: The edge computing nodes deployed on the municipal road collect multi-source monitoring data through optical, vibration, and infrared sensors, and calculate the pavement crack index and real-time traffic flow in parallel; Dynamic health assessment module: Based on the pavement crack index and real-time traffic flow, fusing environmental temperature and humidity data, the edge computing nodes of the municipal road generate an initial judgment value of road health and trigger a dynamic threshold alarm; Regional collaborative processing module: The initial judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculate the regional traffic flow density value and the pavement damage correlation value, and generate an encrypted regional road state feature vector; Central decision-making module: The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with the historical baseline, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes; Resource scheduling module: Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack identification model parameters.

[0016] The present invention also proposes a computer storage medium, including: at least one memory and at least one processor; A memory for storing one or more program instructions; A processor for running one or more program instructions to execute a municipal road monitoring method based on edge computing technology described in any one of the above.

[0017] The beneficial effects achieved by the present invention are as follows: (1) By deploying edge computing nodes on municipal roads, using multi-source sensors such as optical, vibration, and infrared to collect monitoring data, and calculating the pavement crack index and real-time traffic flow in parallel, the comprehensiveness of road condition data collection and the real-time nature of processing are effectively improved, providing diverse and timely data support for road health assessment.

[0018] (2) Based on the pavement crack index, real-time traffic flow, and by fusing environmental temperature and humidity data, the edge computing node weights to generate a preliminary judgment value of road health and trigger a dynamic threshold alarm, realizing the accurate assessment of road health status and adaptive alarm. It can flexibly adjust the alarm strategy according to environmental changes, improving the accuracy and adaptability of the alarm.

[0019] (3) The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculating the regional traffic flow density value and the pavement damage correlation value, generating an encrypted regional road condition feature vector, enhancing the collaborative processing ability of edge computing nodes within the region, realizing the fusion analysis and secure transmission of regional road condition data, and providing more comprehensive regional feature information for global road management.

[0020] (4) The central management platform aggregates all regional feature vectors, calculates the comprehensive health score and early warning weight coefficient of the road in combination with historical baselines, generates an adaptive guidance strategy and issues it to the corresponding edge computing nodes, realizing the comprehensive assessment of the global road health status and global decision-making. It can generate targeted guidance strategies according to the actual road conditions, improving the intelligence and refinement level of municipal road management; (5) Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency and incrementally updates the crack recognition model parameters, realizing the optimal allocation of resources in the monitoring system and the continuous evolution of the model. While reducing the system energy consumption, it continuously improves the accuracy of pavement crack recognition, enabling the entire monitoring system to better adapt to the complex environment and long-term operation requirements of municipal roads. Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other accompanying drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a flowchart of a municipal road monitoring method based on edge computing technology provided by an embodiment of the present application.

[0023] Figure 2 It is a schematic diagram of a municipal road monitoring system based on edge computing technology provided by an embodiment of the present application. Specific embodiments

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0025] Embodiment 1

[0026] As Figure 1 shown, a municipal road monitoring method based on edge computing technology in an embodiment of the present application includes: Step S1: The edge computing nodes deployed on the municipal road collect multi-source monitoring data through optical, vibration, and infrared sensors, and perform parallel computing on the road surface crack index and real-time traffic flow, specifically including the following sub-steps: Step S11: Use an optical sensor to capture road surface image data, identify crack features through a convolutional neural network, and output the road surface crack index; The optical sensor captures the road surface RGB and depth image data streams at a fixed sampling frequency, and performs end-to-end crack feature extraction through a pre-trained lightweight convolutional neural network. The network first performs geometric distortion correction and illumination adaptive compensation on the input image, and then uses a depthwise separable convolutional layer for multi-scale feature fusion to generate a pixel-level crack probability heat map; based on the connected component analysis of the heat map, the total crack length and maximum width are quantified, and combined with the normalization processing of the road surface area, finally, the road surface crack index within the 0-1 standardization interval is output.

[0027] Step S12: Based on the rut track data and heat map distribution synchronously collected by the vibration sensor and the infrared sensor, calculate the real-time traffic flow in combination with the time window statistical model; The vibration sensor array synchronously collects the road surface micro-strain waveforms caused by the vehicle axle load with millisecond-level accuracy. Meanwhile, the infrared thermal imaging unit generates the temperature distribution matrix of the vehicle tire contact surface. After performing time-frequency transformation on the vibration signal, the timestamp sequence of the wheel passing event is identified through the peak detection algorithm; the infrared thermal map locates the rut track boundary through spatial gradient operation and performs spatio-temporal alignment with the vibration event timestamp. Within the sliding time window, based on the prior model of vehicle length, continuous detection events are clustered and de-duplicated, and the vehicle presence is verified by combining the temperature change rate in the thermal map region. Finally, the calibrated real-time traffic flow value is output, and the traffic confidence evaluation results of each lane are synchronously appended.

[0028] Step S2: Based on the road surface crack index and real-time traffic flow, fuse the ambient temperature and humidity data, and the edge computing node of the municipal road generates a preliminary judgment value of road health by weighting and triggers a dynamic threshold alarm, which specifically includes the following sub-steps: Step S21: Based on the road surface crack index and real-time traffic flow, fuse the ambient temperature and humidity data, and the edge computing node of the municipal road generates a preliminary judgment value of road health by weighting and triggers a dynamic threshold alarm; Specifically, the edge computing node receives the data streams of the road surface crack index, real-time traffic flow, and ambient temperature and humidity sensors in real time. Through the pre-constructed road material degradation model, the temperature and humidity data are mapped to the elastic modulus attenuation coefficient of asphalt concrete, which is expressed by the following formula:

[0029] Among them, represents the asphalt elastic modulus attenuation coefficient; represents the real-time temperature, with the unit of °C; represents the reference temperature, which is set to 20 °C and serves as the standard temperature for comparison; represents the critical temperature of material failure, which is 60 °C. When this temperature is reached, the performance of the asphalt material may undergo failure changes; represents the real-time humidity, the actual humidity of the environment where the asphalt is located; represents the saturation humidity threshold, which is 90%RH and is the critical value of the influence of asphalt humidity. When it is exceeded, the influence law of humidity on the material may change; represents the material experiment calibration parameter, which is obtained by fitting through the triaxial compression test and is used to quantify the specific coefficients of the influence of temperature and humidity on the attenuation of asphalt elastic modulus; Meanwhile, based on the principle of equivalent axle load conversion, the traffic flow is converted into the action frequency of the standard axle load, which is expressed by the following formula:

[0030] Among them, represents the equivalent axle load frequency; represents the total number of axle load categories; represents the index of the axle load category; Indicates the actual axle weight of the th axle type; Indicates the standard axle weight; Indicates the damage index; Indicates the vehicle speed of the vehicle with the axle weight of the Indicates the reference vehicle speed; Indicates the coefficient related to the vehicle speed impact; Indicates the humidity impact factor; The multi-layer perceptron fusion network is used to perform feature-level weighted fusion on the crack index, axle load frequency and elastic modulus attenuation coefficient. Among them, the crack index weight increases exponentially with the increase of temperature, and the axle load weight has a negative correlation with the humidity data. Finally, the initial judgment value of road health on a scale of 0-100 is output. When this value is lower than the current dynamic alarm threshold, a level 3 alarm event is immediately triggered and the data packet is encapsulated and uploaded to the cloud supervision platform.

[0031] Step S22: Dynamically adjust the alarm trigger threshold according to the mutation amplitude of the real-time traffic flow and the growth rate of the road surface crack index; Establish a traffic flow mutation monitoring window in hours, and calculate the traffic flow fluctuation intensity through the sliding standard deviation algorithm; at the same time, use the exponential smoothing method to track the change trajectory of the crack index, and extract its maximum slope within 24 hours as the growth rate characterization quantity. Based on the historical road damage case library, train the threshold adjustment model. When the traffic flow fluctuation intensity exceeds the critical interval, the alarm threshold is lowered by a preset ratio to enhance the sensitivity; if the crack growth rate continues to be in the high-risk interval for two hours, the rainstorm mode threshold strategy is automatically enabled, and preventive threshold calibration is implemented in combination with the short-term precipitation forecast data of the real-time meteorological bureau. All threshold adjustment operations record the version label and generate an adjustment confidence evaluation report locally at the edge node.

[0032] Step S3: The initial judgment value of road health is exchanged between adjacent nodes through the low-power Mesh network, and the regional traffic flow density value and the road surface damage correlation value are jointly calculated to generate an encrypted regional road state feature vector, which specifically includes the following sub-steps: Step S31: Adjacent edge computing nodes exchange the initial judgment value of road health through the low-power Mesh network, and calculate the regional traffic flow density value by using the spatio-temporal correlation algorithm; Specifically, adjacent edge computing nodes utilize the multi-hop communication mechanism of a low-power Mesh network to interact the initially judged road health values collected in real time. Each node calculates based on the traffic flow data of itself and adjacent nodes in the time series, combined with the geographical location information, using a spatio-temporal correlation algorithm. This algorithm constructs a spatio-temporal data matrix, analyzes the traffic flow change patterns between different time points and spatial positions, introduces a time decay factor and a spatial distance weight coefficient, quantifies the contribution of data from different nodes to the calculation of the regional traffic flow density, and finally obtains the regional traffic flow density value, reflecting the overall traffic congestion situation within the region.

[0033] Step S32: Generate a pavement damage correlation value based on the spatial distribution characteristics of the multi-node pavement crack index; First, each edge computing node collects pavement image data at its location, uses image recognition technology to extract pavement crack information, and calculates the pavement crack index. Then, the pavement crack indices and their geographical coordinate information of multiple nodes are integrated to construct a spatial distribution model. By analyzing the clustering characteristics, dispersion degree, and change trend of the crack indices in space in the model, a spatial autocorrelation analysis method is used to calculate the correlation coefficients between the crack indices of each node. These correlation coefficients are weighted and summed to finally generate a pavement damage correlation value that can characterize the degree of pavement damage correlation within the region, thereby measuring the mutual influence relationship of pavement damage at different locations. The damage correlation is expressed by the following formula:

[0034] Where, represents the damage correlation; represents the total number of spatial units in the study area; and represent the index for traversing spatial units; represents the element of the spatial weight matrix, reflecting the strength of the spatial relationship between spatial units and ; , respectively represent the attribute values corresponding to spatial units and ; represents the average value of the attribute values of all spatial units ; represents the variance of the attribute value , reflecting the dispersion degree of the attribute value among spatial units; represents the square of the mean of the attribute value .

[0035] Step S33: Encrypt and encode the traffic flow density value and the pavement damage correlation value to generate a regional road state feature vector; Adopt a hybrid encryption method that combines symmetric encryption and asymmetric encryption to process the correlation values between traffic flow density and road surface damage. First, use the symmetric encryption algorithm to quickly encrypt the data with a high-strength encryption key, and then use the asymmetric encryption algorithm to encrypt and transmit the symmetric encryption key. Encode the encrypted data according to specific coding rules and convert it into a binary sequence. According to the preset requirements for the dimension of the feature vector, perform grouping, splicing, and padding operations on the encoded binary sequence, and finally generate an encrypted regional road status feature vector containing regional traffic flow and road surface damage information to ensure the security and integrity of the data during transmission and storage.

[0036] Step S4: The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and the early warning weight coefficient in combination with the historical baseline, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes, which specifically includes the following sub-steps: Step S41: Perform dimensionality reduction processing on the aggregated regional road status feature vectors to extract key road status parameters; Step S42: Compare the key road status parameters with the historical baseline data to calculate the comprehensive road health score and the early warning weight coefficient; The key road status parameters extracted by the platform are compared with the dynamic historical baseline data stored in the database in real time. The historical baseline data is constructed based on the moving average and standard deviation of long-term monitoring data under the same date type and the same time period, and is updated regularly. The comprehensive road health score is calculated using a weighted comprehensive evaluation model. The input of the model is the deviation degree of each key parameter from its historical baseline, and the weight of each parameter is determined according to expert experience or historical data analysis of its impact on the road network traffic efficiency. The scoring result is mapped to a continuous value from 0 to 100, and the higher the score, the better the overall health status of the road network. The early warning weight coefficient is comprehensively calculated based on the degree of deviation of the comprehensive health score from the preset health threshold, the abnormal fluctuation range of the key parameters, and the importance of this area in the overall road network topology structure.

[0037] Step S43: Generate an adaptive guidance strategy containing detour path planning and speed limit instructions according to the early warning weight coefficient.

[0038] Based on the obtained early warning weight coefficient and the specific abnormal conditions of the key road status parameters, the platform generates a targeted adaptive guidance strategy. The core of the strategy includes two parts: detour path planning and dynamic speed limit instructions. The detour path planning dynamically adjusts the path cost function using the real-time road condition topology map and the early warning weight coefficient. Let the road network topology map be , where represents the set of nodes in the road network; the real-time cost of the road section , where Represents a road section, which is an element in

[0039] The implementation cost of the road section is expressed by the following formula:

[0040] Among them, represents the basic travel time of the road section ; represents the warning area to which the road section belongs, and is used to identify which warning-related area range the current road section is in; represents the warning weight coefficient related to the warning area to which the road section belongs, reflecting the intensity of the enhancement of the warning area on the road section cost; represents the location information of the road section ; represents the central position coordinates of the warning area , and the attenuation of the Gaussian term is determined by calculating the distance between the road section location and the center of the warning area; represents the variance of the Gaussian function, which controls the attenuation rate of the Gaussian term.

[0041] Parameters related to the detour path generation formula:

[0042] Among them, represents the optimal detour path; represents the independent variable that finds the minimum of the following expression; represents the limited range, is the path, and is the set of all possible paths; represents calculating the sum of the "cost " of all road sections included in the path ; "; represents the regularization coefficient; represents the length of the path .

[0043] For high-warning-weight areas, significantly increase the travel "cost" of the roads inside or related to them, guide the path planning algorithm to preferentially bypass this area or select less affected alternative paths, and the generated detour recommended paths will cover the main traffic flow source directions. The dynamic speed limit instruction calculates the upper limit of the safe recommended speed for this area or adjacent related road sections according to the regional health score decline rate, real-time traffic flow and density, combined with the historical accident data model.

[0044] This speed limit value is a dynamic threshold, designed to smooth traffic flow, prevent congestion from worsening or accidents from occurring, and is adjusted as the evaluation results are updated. The finally generated adaptive traffic guidance policy package will be accurately sent to the edge computing nodes responsible for the corresponding road sections for execution according to the area division and warning weights.

[0045] Step S5: Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack identification model parameters; Specifically, the central platform receives the load peak and energy consumption curve data uploaded by each edge computing node, analyzes the characteristics of these data, calculates the resource allocation coefficient, and then optimizes and adjusts the sampling strategy of the sensor and the crack identification model, including the following sub-steps: Step S51: Calculate the resource allocation coefficient according to the periodic characteristics of the load peak and energy consumption curve; First, preprocess the load peak and energy consumption curve reported by the edge computing node, decompose the different frequency components of the curve using wavelet transform, and extract the periodic characteristic parameters of the curve. By analyzing the change trend of the load peak and the fluctuation amplitude of the energy consumption curve within adjacent periods, a periodic characteristic vector is constructed.

[0046] Based on this characteristic vector, use the support vector regression algorithm to establish a mapping relationship model between the load peak, energy consumption, and resource allocation, and calculate the resource allocation coefficient of each edge computing node:

[0047] Among them,

[0048]

[0049] Among them, represents the resource allocation coefficient; represents the number of a certain sample, element, or vector participating in the calculation; represents the index; represents the original coefficient parameter; represents a certain correction value, reference value, or target value; represents the bias term; represents the kernel function; is the specific form definition of the kernel function; represents two vectors input into the kernel function, corresponding to and in the formula; represents a parameter in the kernel function, used to control the influence degree of the square term of the vector difference, and adjust the sensitivity of the kernel function to the vector difference; represents the vector and The square of the Euclidean distance measures the distance between two vectors in space; denotes the vector and inner product; denotes the exponent of the inner product, used to adjust the influence degree of the inner product term on the result of the kernel function; denotes an eigenvector, composed of multiple sub-features; denotes a function or operator related to a certain left, with the input being related variables; denotes the norm of the result of this function; denotes a function or operator related to a certain right, with the input being related variables; denotes the norm of the result of this function; denotes a constant used to normalize the norm results related to and ; denotes the adjusted peak length, represents the peak, denotes the adjustment; denotes a certain variance related to ; This coefficient is used to subsequently adjust the sampling frequency of the sensor and the priority of model parameter update.

[0050] Step S52. Scale the image acquisition frequency of the optical sensor proportionally based on the resource allocation coefficient, and reduce the sampling resolution of the vibration sensor; According to the calculated resource allocation coefficient, formulate a sensor sampling strategy adjustment plan. For the optical sensor, dynamically adjust its image acquisition frequency according to the proportional relationship of the resource allocation coefficient, appropriately reduce the acquisition frequency when resources are scarce, and at the same time use an image enhancement algorithm to compensate for the information loss that may be caused by the frequency reduction. For the vibration sensor, reduce the sampling resolution by reducing the sampling bits and increasing the sampling interval, and effectively reduce the data volume and processing load while ensuring that key vibration characteristics can be detected, so as to achieve the optimal allocation of computing resources.

[0051] Step S53. Incrementally update the weight parameters of the convolutional neural network according to the inference delay of the crack recognition model at the edge node; Step S531. Construct an incremental training set from the crack false alarm samples reported by the edge computing node and the new crack feature data; Analyze the crack detection results reported by each edge computing node, and screen out the samples misjudged as cracks and the newly discovered real crack feature data. Label and extract features from these data, divide them into a training set and a validation set according to a preset ratio, and construct an incremental training set for model update to ensure that the training set contains sufficient positive and negative samples to improve the generalization ability of the model.

[0052] Step S532: Use the transfer learning algorithm to fine-tune the convolutional layer parameters of the crack recognition model, and retain the fully connected layer structure of the original model; Using transfer learning technology, based on the original crack recognition model, freeze the fully connected layer parameters and only fine-tune the convolutional layer parameters. By performing iterative training on the incremental training set, the model can learn new crack feature patterns while retaining the recognition ability for the original features. During the training process, adopt an adaptive learning rate adjustment strategy and an early stopping mechanism to prevent the model from overfitting and improve the efficiency and stability of model update.

[0053] Step S533: Distribute the updated model parameters to the corresponding edge computing nodes through an encrypted channel.

[0054] Embodiment 2

[0055] As Figure 2 shown, Embodiment 2 of the present application provides a municipal road monitoring system based on edge computing technology, including: Information acquisition and preprocessing module 21: The edge computing nodes deployed on the municipal road collect multi-source monitoring data through optical, vibration, and infrared sensors, and perform parallel calculations on the road surface crack index and real-time traffic flow; the information acquisition and preprocessing module includes the following sub-modules: Road surface crack recognition sub-module 211: Use an optical sensor to capture road surface image data, identify crack features through a convolutional neural network, and output the road surface crack index; Real-time traffic flow calculation sub-module 212: Based on the rut track data and heat map distribution synchronously collected by the vibration sensor and the infrared sensor, calculate the real-time traffic flow in combination with the time window statistical model; combine the vibration signal detected by the vibration sensor when the vehicle passes and the vehicle heat signal captured by the infrared sensor, and statistically analyze the results of processing these two types of signals according to time periods to obtain the current number of vehicles passing on the road.

[0056] Dynamic health assessment module 22: Based on the road surface crack index and real-time traffic flow, fuse the environmental temperature and humidity data, and the edge computing nodes of the municipal road generate a preliminary judgment value of road health by weighting and trigger a dynamic threshold alarm; include the following sub-modules: Core evaluation sub-module 221: Based on the road surface crack index and real-time traffic flow, integrating environmental temperature and humidity data, the edge computing node of the municipal road generates a preliminary judgment value of road health with weights and triggers a dynamic threshold alarm; on the edge node, the received road surface crack index, real-time traffic flow, and environmental temperature and humidity data are comprehensively calculated according to the preset weight ratio to generate a preliminary road health score. At the same time, based on the currently calculated health score and the preset dynamic baseline, it is immediately judged whether the alarm condition is reached and the corresponding alarm is triggered.

[0057] Threshold management sub-module 222: Dynamically adjusts the alarm trigger threshold according to the mutation amplitude of the real-time traffic flow and the growth rate of the road surface crack index; Regional collaborative processing module 23: The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, and the regional-level traffic flow density value and the correlation value of road surface damage are jointly calculated to generate an encrypted regional road state feature vector; it includes the following sub-modules: Regional traffic flow density collaborative calculation sub-module 231: Adjacent edge computing nodes exchange the preliminary judgment value of road health through a low-power Mesh network, and calculate the regional-level traffic flow density value using the spatio-temporal correlation algorithm; Road surface damage correlation analysis sub-module 232: Generates the road surface damage correlation value based on the spatial distribution characteristics of the road surface crack index of multiple nodes; Regional state feature encryption generation sub-module 233: Encodes and encrypts the traffic flow density value and the road surface damage correlation value to generate a regional road state feature vector.

[0058] Central decision-making module 24: The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with the historical baseline, and generates an adaptive diversion strategy to be sent to the corresponding edge computing node; it includes the following sub-modules: Key feature extraction sub-module 241: Performs dimensionality reduction processing on the aggregated regional road state feature vector to extract key road state parameters; Comprehensive score and early warning calculation sub-module 242: Compares the key road state parameters with the historical baseline data to calculate the comprehensive road health score and early warning weight coefficient.

[0059] Adaptive diversion strategy generation sub-module 243: Generates an adaptive diversion strategy including detour route planning and speed limit instructions according to the early warning weight coefficient; Resource scheduling module 25: Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency and incrementally updates the crack identification model parameters; it includes the following sub-modules: Resource Allocation Coefficient Calculation Sub-module 251: Calculate the resource allocation coefficient according to the periodic characteristics of the load peak and the energy consumption curve; the central platform analyzes the historical load peak data and the energy consumption curve reported by each edge node, identifies the recurring patterns therein, and calculates a distribution coefficient value for guiding the dynamic adjustment of resources accordingly.

[0060] Sensor Sampling Dynamic Adjustment Sub-module 252: Scale the image acquisition frequency of the optical sensor proportionally based on the resource allocation coefficient, and reduce the sampling resolution of the vibration sensor; Model Parameter Incremental Update Sub-module 253: Incrementally update the convolutional neural network weight parameters according to the inference delay of the crack identification model at the edge node; Corresponding to the above embodiments, an embodiment of the present invention provides a computer storage medium, including: at least one memory and at least one processor; The memory is used to store one or more program instructions; The processor is used to run one or more program instructions to execute a municipal road monitoring method based on edge computing technology.

[0061] Corresponding to the above embodiments, an embodiment of the present invention provides a computer-readable storage medium. The computer storage medium contains one or more program instructions, and the one or more program instructions are used to be executed by the processor to execute a municipal road monitoring method based on edge computing technology.

[0062] The disclosed embodiment of the present invention provides a computer-readable storage medium. Computer program instructions are stored in the computer-readable storage medium. When the computer program instructions run on a computer, the computer is caused to execute the above-mentioned municipal road monitoring method based on edge computing technology.

[0063] In the embodiment of the present invention, the processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0064] The various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor, or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed and completed by a hardware decoding processor, or can be executed and completed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register. The processor reads the information in the storage medium and combines its hardware to complete the steps of the above method.

[0065] The storage medium can be a memory, for example, it can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0066] Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory.

[0067] The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM).

[0068] The storage medium described in the embodiments of the present invention is intended to include but not be limited to these and any other suitable types of memories.

[0069] Those skilled in the art should be able to realize that in one or more of the above examples, the functions described in the present invention can be implemented by a combination of hardware and software. When applying software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. The computer-readable medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transfer of a computer program from one place to another. The storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0070] The specific embodiments described above have further elaborated on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, and improvements made on the basis of the technical solutions of the present invention should be included in the protection scope of the present invention.

Claims

1. A municipal road monitoring method based on edge computing technology, characterized in that, Including: Edge computing nodes deployed on municipal roads collect multi-source monitoring data through optical, vibration, and infrared sensors, and concurrently calculate the road surface crack index and real-time traffic flow; Based on the road surface crack index and real-time traffic flow, integrating environmental temperature and humidity data, the edge computing nodes of the municipal road generate a preliminary judgment value of road health by weighting and trigger a dynamic threshold alarm; The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculating the regional traffic flow density value and the correlation degree value of road surface damage, and generating an encrypted regional road state feature vector; The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with historical baselines, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes; Based on the load peak and energy consumption curve reported by the edge computing nodes, the central platform calculates the resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack recognition model parameters.

2. The municipal road monitoring method based on edge computing technology according to claim 1, wherein, Edge computing nodes deployed on municipal roads collect multi-source monitoring data through optical, vibration, and infrared sensors, and concurrently calculate the road surface crack index and real-time traffic flow, including the following sub-steps: Use an optical sensor to capture road surface image data, identify crack features through a convolutional neural network, and output the road surface crack index; Based on the rut track data and heat map distribution synchronously collected by vibration sensors and infrared sensors, calculate the real-time traffic flow in combination with the time window statistical model.

3. The method for monitoring municipal roads based on edge computing technology according to claim 1, wherein, Based on the road surface crack index and real-time traffic flow, integrating environmental temperature and humidity data, the edge computing nodes of the municipal road generate a preliminary judgment value of road health by weighting and trigger a dynamic threshold alarm, including the following sub-steps: Input the road surface crack index, real-time traffic flow, and environmental temperature and humidity data into a weighted fusion model, and calculate the preliminary judgment value of road health according to the preset weight coefficient; Dynamically adjust the alarm trigger threshold according to the mutation amplitude of the real-time traffic flow and the growth rate of the road surface crack index.

4. A municipal road monitoring method based on edge computing technology according to claim 1, characterized in that, The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, jointly calculating the regional traffic flow density value and the correlation degree value of road surface damage, and generating an encrypted regional road state feature vector, including the following sub-steps: Adjacent edge computing nodes exchange the preliminary judgment value of road health through a low-power Mesh network, and calculate the regional traffic flow density value using a spatio-temporal correlation algorithm; Generate the correlation degree value of road surface damage based on the spatial distribution characteristics of the multi-node road surface crack index; Encrypt and encode the traffic flow density value and the correlation degree value of road surface damage to generate a regional road state feature vector.

5. A method for monitoring municipal roads based on edge computing technology according to claim 1, characterized in that, The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with historical baselines, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes, including the following sub-steps: Perform dimensionality reduction processing on the aggregated regional road state feature vectors to extract key road state parameters; Compare the key road state parameters with historical baseline data, and calculate the comprehensive road health score and early warning weight coefficient; Generate an adaptive guidance strategy including detour path planning and speed limit instructions according to the early warning weight coefficient.

6. The municipal road monitoring method based on edge computing technology according to claim 1, characterized in that Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the computing resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack identification model parameters, including the following sub-steps: Calculate the computing resource allocation coefficient according to the periodic characteristics of the load peak and energy consumption curve; Scale the image acquisition frequency of the optical sensor proportionally based on the resource allocation coefficient, and reduce the sampling resolution of the vibration sensor; Incrementally update the convolutional neural network weight parameters according to the inference delay of the crack identification model at the edge node.

7. The method for monitoring municipal roads based on edge computing technology according to claim 6, characterized in that, Incrementally update the convolutional neural network weight parameters according to the inference delay of the crack identification model at the edge node, including the following sub-steps: Construct an incremental training set from the crack false alarm samples reported by the edge computing node and the new crack feature data; Fine-tune the convolutional layer parameters of the crack identification model using transfer learning algorithms, while retaining the fully connected layer structure of the original model; Distribute the updated model parameters to the corresponding edge computing nodes through an encrypted channel.

8. A municipal road monitoring system based on edge computing technology, characterized in that, Including: Information acquisition and preprocessing module: The edge computing nodes deployed on the municipal road collect multi-source monitoring data through optical, vibration, and infrared sensors, and perform parallel calculations on the road surface crack index and real-time traffic flow; Dynamic health assessment module: Based on the road surface crack index and real-time traffic flow, and integrating environmental temperature and humidity data, the edge computing nodes of the municipal road generate a preliminary judgment value of road health and trigger a dynamic threshold alarm; Regional collaborative processing module: The preliminary judgment value of road health is exchanged between adjacent nodes through a low-power Mesh network, and the regional traffic flow density value and the correlation value between road surface damage are jointly calculated to generate an encrypted regional road state feature vector; Central decision-making module: The central management platform aggregates all regional feature vectors, calculates the comprehensive road health score and early warning weight coefficient in combination with the historical baseline, and generates an adaptive guidance strategy to be sent to the corresponding edge computing nodes; Resource scheduling module: Based on the load peak and energy consumption curve reported by the edge computing node, the central platform calculates the computing resource allocation coefficient, dynamically adjusts the sensor sampling frequency, and incrementally updates the crack identification model parameters.

9. A computer storage medium, characterized in that, Including: At least one memory and at least one processor; The memory is used to store one or more program instructions; The processor is used to run one or more program instructions to execute a method for monitoring a municipal road based on edge computing technology as described in any one of claims 1-7.

Images