Municipal road abnormity monitoring method and system based on cloud platform

Abstract

The invention provides a municipal road abnormity monitoring method and system based on a cloud platform. The method comprises the following steps: acquiring vibration signals and deformation data of a road in real time through a self-powered piezoelectric and optical fiber fusion sensing device, and transmitting original data to a cloud processing platform; the cloud platform automatically completes data denoising, feature extraction and time sequence reconstruction to form effective road abnormal feature data; a deep learning anomaly recognition module fusing a Transform model and a time sequence convolutional network analyzes and calculates the feature data to realize accurate recognition and spatial positioning of road anomaly types; further based on the real-time traffic flow data, a road risk level is calculated and determined through an intelligent risk assessment module, and early warning information and response suggestions are pushed to a municipal management terminal or a user in time; and finally, traffic signals and identifiers in the high-risk area are automatically optimized through an edge cooperation control module, vehicles are guided to avoid the risk area, and the safety and the intelligent level of municipal road management are improved.

Description

Technical Field

[0001] The present invention relates to the field of smart city and cloud computing technology, and in particular to a method and system for monitoring municipal road anomalies based on a cloud platform. Background Art

[0002] With the acceleration of urbanization and the increasing complexity of urban transportation networks, the safety and efficiency of municipal roads have become crucial issues in urban management. Over the long term, municipal roads often suffer structural damage such as cracks, potholes, and collapses due to factors such as vehicle traffic, climate change, ground settlement, and structural aging. Furthermore, sudden traffic accidents or extreme weather conditions can seriously threaten the capacity and safety of municipal roads, further exacerbating traffic congestion and safety risks.

[0003] Existing municipal road monitoring systems typically rely on cameras, traffic flow sensors, or manual inspections to detect road anomalies and traffic congestion. However, these methods have the following shortcomings: First, camera monitoring is limited by weather, lighting, and occlusion, resulting in a limited monitoring range and difficulty identifying underground structural damage or small cracks. Second, traffic flow sensors often only monitor surface vehicle flow and speed, failing to capture the structural health of the road's interior, resulting in insufficient early warning capabilities for road anomalies. Third, manual inspections require significant manpower and time, resulting in blind spots and delays, making them unable to meet the real-time monitoring needs of complex traffic environments.

[0004] Against this backdrop, intelligent monitoring systems based on piezoelectric and fiber optic sensors are becoming a key development direction for road anomaly monitoring. Piezoelectric sensors can capture dynamic changes in the road surface in real time through vibration signals generated by vehicle movement; fiber optic sensors can accurately detect road structural deformation and crack growth through the interference and reflection properties of light. Furthermore, with the rapid development of cloud computing and artificial intelligence technologies, the application of deep learning models for anomaly signal recognition and feature extraction has also provided new solutions for road monitoring systems. By combining real-time monitoring, intelligent identification, and dynamic traffic guidance, efficient monitoring and rapid response to municipal road anomalies can be achieved.

[0005] Building on existing technologies, this application proposes a cloud-based municipal road anomaly monitoring method and system. By integrating piezoelectric and fiber-optic fusion sensing technology, cloud-based data processing, and deep learning models, the system can capture road vibration and deformation data in real time, automatically identify anomaly types, accurately locate anomaly locations, and generate optimized traffic guidance plans based on current traffic conditions. Through an edge-based collaborative control module, the system can dynamically adjust traffic signals and road guidance signs in high-risk areas, improving road safety and traffic efficiency. Summary of the Invention

[0006] The present invention provides a method for monitoring municipal road anomalies based on a cloud platform, which includes:

[0007] The self-powered piezoelectric and fiber optic fusion sensor monitors road vibration signals and deformation data, and transmits the monitoring data to a cloud processing platform;

[0008] The cloud processing platform performs denoising, abnormal feature extraction, and time series reconstruction on the received monitoring data to obtain road abnormal feature data;

[0009] A deep learning anomaly recognition module that integrates the Transformer model and a temporal convolutional network is used to analyze road anomaly feature data, identify the type of road anomaly, and determine the location of the anomaly.

[0010] The intelligent risk assessment module calculates and generates a road risk level based on the identified anomaly type, anomaly location, and real-time traffic flow data, and pushes the risk level and corresponding recommended measures to the municipal management terminal or end user;

[0011] The edge collaborative control module automatically optimizes and controls traffic lights and digital signs in areas with high risk levels to guide vehicles to avoid dangerous road sections in real time.

[0012] In the above method, the self-powered piezoelectric and optical fiber fusion sensing device includes a piezoelectric element and an optical fiber sensing unit, the piezoelectric element is used to collect road vibration signals and convert vibration energy into electrical energy, and the optical fiber sensing unit is used to monitor the strain or deformation data of the road surface.

[0013] In the above method, the cloud processing platform uses a multi-scale filtering algorithm to remove environmental noise when denoising the monitoring data, extracts characteristic parameters of road anomalies through frequency domain and time domain analysis, and reconstructs the characteristic parameters in time series to generate an abnormal feature sequence reflecting the road status.

[0014] In the above method, the anomaly recognition module adopts a deep learning model that integrates Transformer and temporal convolutional network, including using Transformer to extract long-term dependent features of road anomaly feature data, and using temporal convolutional network to extract local temporal patterns, thereby improving the recognition accuracy of road anomaly types and locations.

[0015] In the above method, the abnormality recognition module can identify road abnormality types including potholes, cracks and settlements, and determine the location of the abnormality based on the layout position data of the sensor device.

[0016] In the above method, the intelligent risk assessment module calculates the road risk level based on factors such as the severity of the anomaly type, the real-time traffic flow and vehicle speed of the road section where the anomaly is located, and divides the road risk level into multiple levels, each level corresponding to a preset handling measure.

[0017] In the above method, the edge collaboration control module guides passing vehicles to slow down or choose detour routes by adjusting the traffic light timing in high-risk areas and dynamically updating road digital signs, thereby avoiding dangerous sections of the road.

[0018] The present invention also provides a municipal road anomaly monitoring system based on a cloud platform, which includes:

[0019] A self-powered piezoelectric and fiber optic fusion sensor device is used to monitor road vibration signals and deformation data, and upload the monitoring data to a cloud processing platform;

[0020] The cloud processing platform is used to perform denoising, abnormal feature extraction, and time series reconstruction on the received monitoring data to obtain road abnormal feature data;

[0021] The anomaly recognition module uses a deep learning model that integrates Transformer and temporal convolutional networks to analyze road anomaly feature data, identify the type of road anomaly, and determine the location of the anomaly.

[0022] An intelligent risk assessment module generates a road risk rating based on anomaly type, location, and real-time traffic flow data, and pushes the risk rating and recommended measures to municipal management terminals or end users;

[0023] The edge collaborative control module is used to automatically optimize the control of traffic lights and digital signs in areas with high risk levels, so as to guide vehicles to avoid dangerous sections of road in real time.

[0024] The beneficial effects achieved by the present invention are as follows: the present invention obtains road vibration and deformation data in real time through an autonomously powered piezoelectric and fiber optic fusion sensing device, utilizes cloud-based automated processing and a deep learning model that integrates Transformer and temporal convolutional networks to accurately identify road anomalies and determine their locations, assesses risk levels in combination with real-time traffic flow information, automatically pushes risk warnings and response measures, and utilizes edge collaborative control to optimize traffic guidance. This can significantly improve the real-time, accuracy, and intelligence level of municipal road anomaly monitoring, thereby enhancing traffic safety and management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0026] Figure 1 This is a flow chart of a method for monitoring abnormal municipal roads based on a cloud platform provided in Example 1 of the present application. DETAILED DESCRIPTION

[0027] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] Example 1

[0029] like Figure 1 As shown, the first embodiment of the present application provides a method for monitoring municipal road anomalies based on a cloud platform, comprising the following steps:

[0030] Step S10: monitoring road vibration signals and deformation data through a self-powered piezoelectric and optical fiber fusion sensing device, and transmitting the monitoring data to a cloud processing platform;

[0031] Specifically, the sensing device is installed in key monitoring areas of municipal roads. By combining piezoelectric and fiber optic technologies, it can fully sense the dynamic vibration and deformation of the road surface. The device includes a piezoelectric sensing module, a fiber optic sensing module, an autonomous power supply module, and a data transmission module. Its working process is as follows:

[0032] The piezoelectric sensing module, based on piezoelectric materials, uses stress changes caused by vehicles passing by and road load variations to collect vibration signals in real time. The module records characteristic parameters such as vibration amplitude, frequency, and duration, and stores this data in digital form in a local cache.

[0033] The fiber optic sensing module uses distributed fiber optic sensing technology to monitor the deformation state of the road surface, including information such as displacement, stress distribution and crack development. By monitoring the wavelength or scattering changes of the optical signal, accurate deformation data can be extracted.

[0034] The sensor device utilizes an autonomous power supply, utilizing the electrical energy generated by piezoelectric materials when subjected to pressure or vibration, along with the photoelectric conversion equipment included with the fiber optic sensing module, to convert mechanical and optical energy into electrical energy. The power supply module integrates an energy management unit that stores the harvested energy in a high-efficiency battery, providing long-term, stable power for the entire device, adapting to the complex environments and long-term deployment requirements of municipal roads.

[0035] The data transmission module has multiple built-in communication interfaces, allowing users to select the appropriate transmission protocol based on their needs, such as low-power wide area networks (LPWAN), cellular networks (such as NB-IoT and 5G), or wired networks. After the sensor performs preliminary processing on the collected vibration signals and deformation data, the data is packaged and encrypted and uploaded to the cloud processing platform via the transmission module.

[0036] Step S20: The cloud processing platform performs denoising, abnormal feature extraction, and time series reconstruction on the received monitoring data to obtain road abnormal feature data;

[0037] In this step, the cloud-based processing platform first receives road vibration signals and deformation data from the sensors. Due to the complex environment of municipal roads, the collected signals may contain a large amount of background noise and environmental interference (such as vehicle noise, construction noise, and meteorological noise). Therefore, the platform needs to denoise this data to extract valid features related to road anomalies.

[0038] The cloud-based processing platform uses a joint adaptive denoising method based on Fourier transform and wavelet transform. During the denoising process, the platform first performs a Fourier transform on the input time domain signal, converting it into the frequency domain to distinguish valid signals from noise signals.

[0039] In the frequency domain, the platform dynamically adjusts the filter bandwidth based on the noise distribution characteristics, thereby suppressing noise while maximizing the preservation of effective signals. The platform then further uses wavelet decomposition to split the signal into different frequency components and reconstructs the denoised signal using dynamic weighting and regularization methods. The core calculation formula is: Among them, X denoised (t) represents the signal after denoising; F and F -1 represent Fourier transform and inverse transform respectively; H(f) is an adaptive filter that dynamically adjusts the frequency range according to the noise level; x(t) represents the original time domain signal; x i (t) is the i-th component after wavelet decomposition; W i (f) is the adaptive weighting function for the wavelet component, which indicates the importance of different frequency components; N is the number of components after wavelet decomposition; is a regularization factor that is automatically adjusted based on signal power and noise level. This effectively suppresses interference in complex noise environments and maximizes the retention of abnormal characteristic signals.

[0040] After denoising, the cloud platform extracts features from the filtered and reconstructed signal. The platform uses a feature extraction method based on a dynamic feature combination model, combining features from the time, frequency, and spatial domains into a complete feature vector to form an anomaly feature dataset.

[0041] Time-domain features describe the signal's changing characteristics over time. First, the platform extracts the signal's overall energy level over time. This energy level reflects the intensity and stability of vibrations and is an important indicator for identifying sudden anomalies. The energy level is expressed using the following formula: Where E(t) represents the average signal energy in the time window T; x denoised (t) represents the denoised vibration signal; T represents the time window size. This item is mainly used to identify sudden vibrations and persistent anomalies.

[0042] Frequency domain features are used to describe the energy distribution and amplitude changes of a signal at different frequencies. The platform converts the signal from the time domain to the frequency domain through Fourier transform and extracts the center frequency. The center frequency represents the dominant frequency component of the signal in the frequency domain. The higher the center frequency, the more high-frequency components the signal contains, which may be due to anomalies caused by short-term impact or high-frequency vibration. For example, a vehicle running over quickly may cause high-frequency vibration; structural damage may manifest as enhanced low-frequency components. The platform divides the spectrum into multiple frequency bands and extracts the energy proportion in each frequency band. The energy proportion within the frequency band is expressed as: Among them, P k represents the energy ratio of the kth frequency band; f k represents the center frequency of the kth frequency band; f min and f max The numerator represents the total energy within a specific frequency band, and the denominator represents the total energy of the entire spectrum. The platform dynamically adjusts the number and width of frequency bands to automatically adapt to different types of abnormal characteristics. For example, crack expansion may manifest as an enhancement of specific low-frequency components, while a high-speed vehicle impact may manifest as an enhancement of specific high-frequency components. The platform also determines the type of abnormality by extracting morphological features of the spectrum (such as symmetry and sharpness). For example, enhanced symmetry may indicate periodic vibration, while increased sharpness may indicate sudden vibration or impact.

[0043] Spatial domain features are used to describe the propagation and interaction of signals between different sensors. The platform constructs a spatial correlation matrix to analyze the coordinated change patterns between different sensors. The spatial matrix formula is: Where R represents the correlation coefficient between sensors; n, m represent the number of sensors; W ij represents the weighting factor between sensors i and j; x i , x j Represents the output signals of different sensors; Represents the average value of the sensor output signal. Using the spatial correlation matrix, the platform can identify abnormal propagation patterns between different sensors. If a specific element in the correlation matrix increases abnormally, it indicates that the abnormality is propagating spatially. If the overall value of the correlation matrix increases, it indicates possible global structural damage or wide-area vibration anomalies. If the correlation matrix changes dramatically at a specific time step, it indicates the possibility of localized sudden abnormalities.

[0044] The platform dynamically adjusts weighting parameters in the time, frequency, and spatial domains to combine these features into a complete feature vector. This comprehensive and dynamic feature combination enables the platform to fully and accurately describe road anomalies, providing comprehensive data support for subsequent deep learning models.

[0045] Step S30: Analyze the road anomaly feature data using a deep learning anomaly recognition module that integrates a Transformer model and a temporal convolutional network to identify the road anomaly type and determine the anomaly location.

[0046] After completing denoising and feature extraction, the cloud-based processing platform generates a comprehensive dataset of road anomaly features. This data captures the interactions and dynamic changes between signals across different time periods, frequency ranges, and spatial locations. To conduct in-depth analysis of this data, the platform employs a deep learning approach that combines the Transformer model with a temporal convolutional network (TCN). This approach captures both long-term and short-term dynamic changes, enabling accurate identification of anomaly types and pinpointing their locations.

[0047] During the analysis process, the cloud platform first reorganizes the feature data according to time sequence and spatial position to form a complete multidimensional time series tensor. The structure of the time series tensor includes: time dimension, space dimension and feature dimension. The time dimension represents the changing trend of the feature at different time steps, the space dimension represents the interaction between different sensors, and the feature dimension represents the state of different types of features (such as time domain, frequency domain and spatial domain features) at each time step and each position. The platform uses a sliding window method to group the feature data. Each window represents a complete time series, and there is partial overlap between windows to maintain the continuity and integrity of the time series. The platform will dynamically adjust the window length and overlap based on the periodicity and signal strength of the historical data to ensure that the model can capture short-term dynamic changes while taking into account long-term trends.

[0048] After completing data reorganization, the platform first inputs the feature data into the Transformer model. The Transformer model is a deep learning architecture based on self-attention and multi-head attention mechanisms, and excels at modeling temporal dependencies and dynamic correlations over long time spans. In the self-attention mechanism, the model performs a weighted combination of features across time steps to generate an attention matrix that represents the similarities and dependencies between them. If the features of a time step align with the trends of the preceding and following time steps, the model assigns a higher attention weight, indicating a strong correlation between the current time step and the others. If the features of a time step do not align with the trends of the preceding and following time steps, the model lowers the attention weight, indicating that the current time step may represent a sudden or abnormal signal. For example, when a crack in a road expands, the model may detect similar frequency increases or amplitude increases in multiple adjacent time steps, thereby increasing the attention weights between these time steps and determining that the signal may be an anomaly caused by structural damage. If the instantaneous energy or high-frequency component of a certain time step suddenly increases but lacks correlation with other time steps, the model will reduce its attention weight and judge that it may be instantaneous noise or unstructured vibration.

[0049] To enhance the Transformer model's ability to perceive the order of time series, the platform introduces a positional encoding mechanism. Since the Transformer itself lacks the ability to perceive time order, the platform injects the order information of time steps into the model by mapping the time step index into a combination of sine and cosine functions. Through positional encoding, the platform ensures that the model maintains the integrity and order perception of time series when modeling periodic and trending signals. Furthermore, the platform uses a multi-head attention mechanism to simultaneously model different time steps, different frequency components, and different sensors, generating multiple independent attention maps. By merging these attention maps, the platform can simultaneously capture the dynamic characteristics of vibration signals at different time scales and spatial dimensions, thereby enhancing the model's modeling capabilities in complex environments.

[0050] After completing the global time feature modeling, the platform inputs the output feature tensor of the Transformer into the Temporal Convolutional Network (TCN) to further extract local time dynamic features. TCN uses causal convolution and dilated convolution to model time series. Causal convolution is a special convolution method that ensures that the output of the current time step depends only on the data of the past time step, prevents future information from "leaking" into the current state, and maintains the integrity of the time series. For example, when the vibration caused by a vehicle running over propagates to different sensors, the model will use causal convolution to determine that the vibration is caused by the impact of the previous time step, rather than the result of a future event. Through causal convolution, the platform can establish dynamic causal relationships between different time steps, thereby more accurately identifying the temporal dependency of signals.

[0051] Dilated convolution expands the model's receptive field by introducing dilation between convolution kernels, thereby capturing long-term trends without increasing the number of parameters. The platform dynamically adjusts the dilation rate to enable the model to switch freely between short-term dynamics and long-term trends. For example, when road cracks expand or structural damage occurs, the signal may gradually increase over a longer time window; when a vehicle impacts or the ground shakes, the signal may show strong high-frequency changes within a short time window. By combining causal convolution and dilated convolution, the platform can accurately extract local temporal dynamic features while taking into account both short-term sudden changes and long-term cumulative changes.

[0052] After extracting global and local temporal features, the platform uses a fully connected classifier (FullyConnected Layer) to classify anomaly types. The classifier generates a probability distribution for different anomaly types based on the feature vectors output by the Transformer and TCN. Anomaly types include road crack expansion, traffic impact, structural damage, ground subsidence, and sensor failure. The platform uses the cross entropy loss function to train the classifier to ensure the confidence and accuracy of the classification output. For example, if an abnormal event occurs at multiple sensor locations, the model will distribute the output probability across different locations based on the global attention matrix and local convolutional responses, thereby identifying the primary location of the anomaly.

[0053] After completing anomaly classification, the platform determines the location of the anomaly based on the Transformer's attention matrix and the TCN's convolutional response. Based on the weight distribution of different sensors and time steps in the attention matrix, the platform determines the geographic location or specific sensor number where the anomaly may have occurred. If a sensor has the highest attention weight at a certain time step, the platform determines that location may be the source of the anomaly; if a time step has the strongest response on the convolution kernel, the platform determines that the anomaly may have occurred within the time window covered by that time step. Furthermore, the platform combines the spatial correlation matrix between sensors to determine the propagation of the anomaly between different locations and determine whether the anomaly is a localized or widespread event.

[0054] Ultimately, the platform outputs complete anomaly identification results, including anomaly type (such as cracks, collapses, and impacts), anomaly location (based on sensor location or geographic coordinates), anomaly time (specific time step location), and confidence level (based on classifier output probability). Based on these outputs, the platform automatically adjusts the Transformer and TCN parameter configurations to enhance the model's stability and adaptability in complex environments.

[0055] Step S40: The intelligent risk assessment module calculates and generates a road risk level based on the identified anomaly type, anomaly location, and real-time traffic flow data, and pushes the risk level and corresponding recommended measures to the municipal management terminal or end user;

[0056] After identifying the anomaly type and locating its location, the cloud-based processing platform inputs the identified anomaly characteristics into the intelligent risk assessment module. This module dynamically calculates the road's risk level based on the anomaly type, location, and real-time traffic flow data. It then generates corresponding management measures or warning information and pushes it to municipal management terminals or end users. This module enables the platform to quickly and accurately assess road safety status after an anomaly occurs, providing effective decision-making support for municipal management departments.

[0057] During the risk assessment process, the platform first formats and normalizes the input anomaly feature data. This data includes anomaly type, location, temporal characteristics, frequency characteristics, and spatial characteristics. The platform then combines real-time traffic data (such as traffic volume, speed, and lane occupancy) with historical road anomaly records to generate a complete feature vector. This feature vector comprehensively reflects the overall characteristics of the road anomaly and the dynamic changes in the current traffic environment.

[0058] After completing the feature data construction, the platform uses a risk assessment method based on dynamic adjustment and joint modeling to calculate road risk. The risk level is comprehensively modeled using a nonlinear formula to capture the complex relationship between different features. The final road risk level is expressed as follows: Among them, R represents the road risk level; T represents the anomaly type characteristics, such as cracks, collapses or impacts; F represents the frequency characteristics, high-frequency anomalies correspond to vehicle impacts or short-term vibrations, while low-frequency anomalies represent structural damage or foundation settlement; S represents the spatial propagation characteristics, reflecting the expansion pattern of the anomaly between different sensors; L represents the location characteristics, indicating the geographical location or specific location of the road segment where the anomaly occurs; V represents the traffic flow characteristics, such as traffic volume, speed and road load level; P represents the degree of matching of historical anomaly patterns in the current state. If the current anomaly characteristics are similar to the historical data pattern, the model will strengthen the response strength to the anomaly; the regularization parameter τ prevents the model output from being abnormal or unstable in extreme cases; α, β, γ, δ, ∈ are dynamic adjustment coefficients used to dynamically correct the risk assessment according to different traffic conditions and anomaly characteristics; Indicates the importance of historical patterns; ω and θ are balance factors that control the balance between current features and historical data.

[0059] The risk level generated by the platform is a discrete integer between 1 and 5, which is divided into the following categories:

[0060] Level 1 (low risk): indicates good road conditions and the anomaly may be a short-term disturbance or a sensor false alarm;

[0061] Level 2 (medium-low risk): indicates that there are minor abnormalities on the road, which may be fluctuations caused by vehicle impact or short-term vibration;

[0062] Level 3 (medium risk): indicates moderate damage to the road, possibly caused by crack expansion or periodic impact;

[0063] Level 4 (high risk): indicates that the road has serious damage, which may be structural damage or long-term abnormalities caused by settlement;

[0064] Level 5 (extremely high risk): Indicates that there are serious safety hazards on the road, which may be collapse or large-scale damage.

[0065] After generating the risk level, the platform will generate corresponding management measures based on the different risk levels and push them to the municipal management terminal or end user. For medium and low-level risks (levels 1-3), the platform will trigger an alert mechanism and notify relevant management departments to monitor or track. In high-risk (level 4) or extremely high-risk (level 5) situations, the platform will automatically trigger an emergency response mechanism, for example:

[0066] Generate detours in traffic signal systems to avoid vehicles entering high-risk areas.

[0067] Set up dynamic warning signs on the road to remind drivers to pay attention to road conditions.

[0068] Send emergency dispatch instructions to the municipal management department and arrange engineering personnel to conduct on-site inspection and repairs.

[0069] Output generated by the platform includes:

[0070] Anomaly Type: Indicates the specific anomaly type identified, such as crack, collapse, or impact.

[0071] Anomaly location: Indicates the sensor number or geographic coordinates where the anomaly occurred.

[0072] Risk Level: A generated discrete integer (1–5) that represents the safety status of the current road.

[0073] Response measures: Specific measures generated based on risk levels, such as traffic guidance or emergency dispatch.

[0074] Confidence: Indicates the reliability of the evaluation results, based on the dynamic confidence level of the model output.

[0075] The platform dynamically adjusts model parameters, combining current anomaly characteristics with historical data, to achieve rapid adaptation and precise response to different scenarios. After generating a risk level, the platform continuously updates the model based on subsequent road and traffic conditions, ensuring accurate and timely risk assessments.

[0076] Step S50: The edge collaboration control module automatically optimizes and controls traffic lights and digital signs in areas with high risk levels to guide vehicles to avoid dangerous sections in real time.

[0077] After completing the risk assessment and generating management measures, the cloud processing platform sends the generated risk level and response measures to the edge collaborative control module. The edge collaborative control module's primary task is to automatically optimize the configuration of traffic lights and digital road signs based on the generated risk level and real-time traffic conditions to guide vehicles away from dangerous areas and reduce the likelihood of accidents or congestion caused by road anomalies.

[0078] When the platform detects a high or extremely high risk area (i.e., a risk level of 4 or 5) in a specific area, the edge collaborative control module automatically initiates an optimization mechanism. First, the module determines the traffic conditions of the affected road section and its surroundings based on the location and traffic conditions of the anomaly. Through its integrated traffic monitoring system, the platform obtains information such as current road traffic flow, vehicle speed, and lane occupancy, allowing for real-time assessment of the overall operational status of the road network.

[0079] After acquiring traffic data, the module will automatically generate a dynamic traffic guidance plan based on the current traffic load. Specifically, the module will guide traffic in the following ways:

[0080] Dynamically adjust the timing strategy of traffic lights.

[0081] If a road section presents a high risk, the module may extend the green light time of adjacent roads, increase the vehicle passing rate, and guide vehicles to detour to avoid high-risk areas.

[0082] Traffic lights on the road section where the abnormality occurs may be set to red, temporarily blocking vehicles from entering to prevent traffic accidents or vehicle delays.

[0083] If the traffic load in a certain area is close to saturation, the module may coordinate the adjustment of traffic light timing among multiple intersections to optimize the overall traffic flow and alleviate traffic pressure.

[0084] Control dynamic digital signs on roads.

[0085] On electronic display screens near high-risk road sections, the module will display real-time road status information and detour prompts.

[0086] If the abnormal condition of a road section requires a long time to repair, the module will set up continuous detour guidance signs at multiple locations to guide vehicles to choose alternative routes.

[0087] In the event of an accident or extreme weather conditions, the module will prompt the driver to slow down or change lanes through the display screen.

[0088] Adjust the path planning of the intelligent navigation system.

[0089] The platform will synchronize the location information and risk level of abnormal road sections to the intelligent navigation system.

[0090] The module will automatically re-plan routes for vehicles approaching the area based on real-time traffic flow data and road conditions to prevent vehicles from entering high-risk areas.

[0091] In the navigation system, the platform may dynamically adjust the detour path based on historical data and current status to ensure the optimality and safety of the navigation path.

[0092] During actual control, the edge collaborative control module employs an adaptive dynamic adjustment strategy, continuously optimizing traffic signals and guidance plans based on real-time changes in road conditions. For example, if a road section experiences congestion during peak hours, the module prioritizes directing traffic to alternative routes with greater capacity to ensure overall traffic balance. If a high-risk road section returns to normal within a short period of time, the module automatically lifts the temporary traffic restrictions and restores the original signal timing and traffic signs.

[0093] Furthermore, the edge collaborative control module, through its integrated traffic monitoring system, can monitor optimization results in real time. The module dynamically adjusts traffic light timing parameters and guidance strategies based on data such as vehicle traffic efficiency, road flow, and accident rates. For example, if vehicle traffic efficiency improves on a particular road section after optimization measures are implemented, the module will store the optimization strategy as a historical template for subsequent reference in similar situations. If, after optimization measures are implemented, a road section still experiences vehicle delays or traffic accidents, the module will automatically adjust traffic light timing or guidance paths to further optimize traffic flow.

[0094] While implementing traffic control, the platform also continuously provides feedback on optimization results to municipal management terminals and traffic management departments. The platform generates a comprehensive optimization results report, which primarily includes: changes in risk levels for affected road sections; comparisons of traffic efficiency before and after signal adjustments; adjustments to navigation system guidance paths; changes in vehicle detour rates and travel times; and changes in traffic accidents or vehicle delays.

[0095] If the optimization results are satisfactory, the platform automatically stores the current optimization plan in a cloud database, creating a template for future use in similar situations. If the optimization results fall short of expectations, the platform automatically adjusts the optimization strategy based on real-time monitoring data, ensuring stable and efficient traffic guidance under varying traffic conditions and abnormalities.

[0096] Example 2

[0097] The second embodiment of the present application provides a municipal road abnormality monitoring system based on a cloud platform, including:

[0098] A self-powered piezoelectric and fiber optic fusion sensing device 21 is used to monitor road vibration signals and deformation data, and upload the monitoring data to a cloud processing platform;

[0099] The self-powered piezoelectric and fiber optic fusion sensing device is installed at key locations on municipal roads, such as lane joints, curves, bridges, and tunnels. The device integrates piezoelectric and fiber optic sensors, enabling highly sensitive monitoring of road surface vibration signals and structural deformation data through the fusion of these two sensing mechanisms.

[0100] Through the mechanical stress caused by vehicle driving or environmental vibration, the piezoelectric sensor generates a charge signal, thereby extracting dynamic vibration data caused by vehicle impact, road crack expansion, and ground collapse.

[0101] Through the interference and reflection characteristics of light, fiber optic sensors can monitor road structure deformation and displacement with high precision at the sub-millimeter level, and extract structural abnormality signals such as road crack expansion and bridge subsidence.

[0102] The sensor device is integrated with an energy recovery module, which can generate electricity through photovoltaic or piezoelectric effects to continuously power the sensor, reduce external power supply requirements, and ensure long-term stable operation of the sensor device.

[0103] After collecting vibration and deformation data, the sensor uses a built-in data preprocessing chip to perform preliminary filtering and normalization on the raw signals to reduce background noise. The preprocessed monitoring data is uploaded to a cloud processing platform 22 in real time via a wireless communication network (such as 5G or LoRa).

[0104] The cloud processing platform 22 is used to perform denoising, abnormal feature extraction and time series reconstruction on the received monitoring data to obtain road abnormal feature data;

[0105] The cloud processing platform 22 receives the vibration signals and deformation data from the sensor device 21 and extracts effective road anomaly features through a series of signal processing algorithms, mainly including the following:

[0106] Denoising: The platform uses a joint denoising method based on Fourier transform and wavelet decomposition. It first removes environmental noise and interference signals through frequency domain filters, then extracts effective signal components at different scales through wavelet decomposition, and finally generates the denoised original signal.

[0107] Abnormal feature extraction: Based on a dynamic feature combination model, the platform extracts the time domain features (such as instantaneous energy), frequency domain features (such as center frequency and frequency band energy distribution), and spatial features (such as correlation between sensors) of the vibration signal to generate a complete abnormal feature vector.

[0108] Time series reconstruction: The platform reconstructs the time series of feature vectors through singular value decomposition (SVD) or adaptive filters, eliminating noise components, retaining the main characteristic patterns, and ensuring the integrity and consistency of the generated feature vectors.

[0109] The cloud platform dynamically adjusts the filter parameters (such as bandwidth, center frequency, etc.) and wavelet decomposition scale to adapt to different types of road anomalies and background noise, ensuring that accurate road anomaly features can still be extracted in complex environments.

[0110] The anomaly recognition module 23 uses a deep learning model that integrates Transformer and temporal convolutional network to analyze road anomaly feature data, identify road anomaly types and determine anomaly locations. The anomaly recognition module 23 integrates a deep learning model that integrates Transformer and temporal convolutional network (TCN) to intelligently analyze anomaly feature vectors generated by the cloud processing platform and accurately identify anomaly types and locations.

[0111] Intelligent risk assessment module 24, used to generate road risk levels based on anomaly type, anomaly location and real-time traffic flow data, and push the risk level and recommended measures to municipal management terminals or end users;

[0112] The intelligent risk assessment module 24 generates a comprehensive road risk level based on the identified anomaly type, location, and traffic conditions. A nonlinear mapping model is used to generate a comprehensive risk level distribution, with risk levels ranging from 1 to 5, representing low risk to extremely high risk. Dynamic management measures are then generated based on the risk level, such as adjusting traffic light timing, redirecting vehicles, or issuing emergency alerts.

[0113] The intelligent risk assessment module generates the following output results in real time: road risk level; corresponding traffic guidance plan and management measures; confidence level of the output results.

[0114] The edge collaboration control module 25 is used to automatically optimize and control traffic lights and digital signs in areas with high risk levels, so as to guide vehicles to avoid dangerous sections of road in real time.

[0115] The edge collaborative control module 25 integrates with the traffic signal system and intelligent navigation system to dynamically guide vehicles to avoid high-risk areas, ensuring smooth and safe traffic. It mainly includes:

[0116] Traffic light optimization control: Automatically adjust traffic light timing according to current traffic conditions and abnormal road locations, extend safe passage time or set traffic restrictions.

[0117] Dynamic Guidance System: Detour routes and traffic warning information are published on electronic display screens, navigation systems and intelligent traffic signs on the road.

[0118] Real-time adaptation mechanism: Dynamically optimizes guidance paths and signal light configurations based on traffic flow data and vehicle feedback to ensure efficient traffic guidance capabilities when traffic conditions change.

[0119] When the risk level of a road section rises to Level 4 or 5, the module will automatically trigger the following measures: prevent vehicles from entering high-risk areas through the traffic light system; generate detour plans in the navigation system to prevent vehicles from passing through dangerous areas; dispatch the traffic management system to send an emergency report to the municipal management department and activate the emergency response mechanism.

[0120] Corresponding to the above embodiment, an embodiment of the present invention provides a computer storage medium, comprising: at least one memory and at least one processor;

[0121] The memory is used to store one or more program instructions;

[0122] The processor is used to run one or more program instructions to execute a municipal road abnormality monitoring method based on a cloud platform.

[0123] Corresponding to the above embodiment, an embodiment of the present invention provides a computer-readable storage medium, which contains one or more program instructions, and the one or more program instructions are used by a processor to execute a municipal road abnormality monitoring method based on a cloud platform.

[0124] The embodiments disclosed in the present invention provide a computer-readable storage medium, in which computer program instructions are stored. When the computer program instructions are executed on a computer, the computer executes the above-mentioned cloud platform-based municipal road abnormality monitoring method.

[0125] In the embodiments of the present invention, the processor may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0126] The methods, steps, and logic diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads the information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned methods.

[0127] The storage medium may be a memory and may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0128] 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.

[0129] Volatile memory may be random access memory (RAM), which is used as an external cache memory. By way of example and 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 (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM).

[0130] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0131] Those skilled in the art will appreciate that in one or more of the above examples, the functions described herein can be implemented using a combination of hardware and software. When software is used, 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. Computer-readable media include computer storage media and communication media, wherein communication media includes any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0132] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring abnormalities of municipal roads based on a cloud platform, characterized in that: The following steps are involved: The self-powered piezoelectric and fiber optic fusion sensor monitors road vibration signals and deformation data, and transmits the monitoring data to a cloud processing platform; The cloud processing platform performs denoising, abnormal feature extraction, and time series reconstruction on the received monitoring data to obtain road abnormal feature data; A deep learning anomaly recognition module that integrates the Transformer model and a temporal convolutional network is used to analyze road anomaly feature data, identify the type of road anomaly, and determine the location of the anomaly. The intelligent risk assessment module calculates and generates a road risk level based on the identified anomaly type, anomaly location, and real-time traffic flow data, and pushes the risk level and corresponding recommended measures to the municipal management terminal or end user; The edge collaborative control module automatically optimizes and controls traffic lights and digital signs in areas with high risk levels to guide vehicles to avoid dangerous road sections in real time.

2. The method according to claim 1, wherein The self-powered piezoelectric and fiber optic fusion sensing device includes a piezoelectric element and a fiber optic sensing unit. The piezoelectric element is used to collect road vibration signals and convert vibration energy into electrical energy, and the fiber optic sensing unit is used to monitor the strain or deformation data of the road surface.

3. The method according to claim 1, wherein The cloud-based processing platform uses a multi-scale filtering algorithm to remove environmental noise when denoising monitoring data, extracts characteristic parameters of road anomalies through frequency and time domain analysis, and reconstructs the characteristic parameters in time series to generate an abnormal feature sequence reflecting the road status.

4. The method according to claim 1, wherein The anomaly recognition module adopts a deep learning model that integrates Transformer and temporal convolutional network. It uses Transformer to extract long-term dependent features of road anomaly feature data, and uses temporal convolutional network to extract local temporal patterns, thereby improving the recognition accuracy of road anomaly types and locations.

5. The method according to claim 1, wherein The anomaly recognition module can identify road anomaly types including potholes, cracks and subsidence, and determine the location of the anomaly based on the layout position data of the sensor device.

6. The method according to claim 1, wherein The intelligent risk assessment module calculates the road risk level based on the severity of the anomaly type, the real-time traffic flow of the road section where the anomaly is located, and the vehicle speed factors, and divides the road risk level into multiple levels, each of which corresponds to a preset handling measure.

7. The method according to claim 1, wherein The edge collaborative control module guides passing vehicles to slow down or choose detour routes to avoid dangerous sections of the road by adjusting the timing of traffic lights in high-risk areas and dynamically updating road digital signs.

8. A municipal road abnormality monitoring system based on a cloud platform, characterized in that: include: A self-powered piezoelectric and fiber optic fusion sensor device is used to monitor road vibration signals and deformation data, and upload the monitoring data to a cloud processing platform; The cloud processing platform is used to perform denoising, abnormal feature extraction, and time series reconstruction on the received monitoring data to obtain road abnormal feature data; The anomaly recognition module uses a deep learning model that integrates Transformer and temporal convolutional networks to analyze road anomaly feature data, identify the type of road anomaly, and determine the location of the anomaly. An intelligent risk assessment module generates a road risk rating based on anomaly type, location, and real-time traffic flow data, and pushes the risk rating and recommended measures to municipal management terminals or end users; The edge collaborative control module is used to automatically optimize the control of traffic lights and digital signs in areas with high risk levels, so as to guide vehicles to avoid dangerous sections of road in real time.

Images