Urban lifeline engineering safety risk identification method and system based on intelligent lamp post

Abstract

The invention discloses an urban lifeline engineering safety risk identification method and system based on an intelligent lamp post, and relates to the technical field of urban safety monitoring and intelligent identification. The method comprises the following steps: receiving multi-source data of urban lifeline engineering, wherein the multi-source data comprises image data or monitoring data of fire, city appearance violation and bridge safety risk; and performing data preprocessing on the multi-source data, deploying a lightweight model to perform feature extraction and real-time identification on image data or monitoring data of three types of risks, triggering hierarchical alarm, outputting a risk identification result and alarm information, and uploading the risk identification result and the alarm information to a cloud end. According to the invention, monitoring and identification functions of three types of core risks of fire, city appearance violation and bridge safety are integrated on a unified platform, an information island formed by dispersive deployment of traditional monitoring equipment is thoroughly broken, comprehensive acquisition and efficient fusion of multi-source data are realized, and a full-process technical support is provided for urban safety management.

Description

Technical Field

[0001] This invention relates to the field of urban safety monitoring and intelligent identification technology, specifically to a method and system for identifying safety risks in urban lifeline projects based on smart light poles. Background Technology

[0002] With the accelerated pace of urbanization, the scale of urban lifeline projects (including bridges, gas pipelines, water supply and drainage networks, power and communication facilities, etc.) continues to expand, and their safe operation is directly related to urban public safety and people's livelihood. At the same time, the standardized and orderly management of the urban environment has increasingly become an important part of modern urban governance. Currently, the following technical means are mainly used for safety monitoring of urban lifeline projects and supervision of violations of urban appearance regulations: In terms of safety monitoring of urban lifeline projects, traditional solutions often adopt a model of independently deploying monitoring equipment at each risk point, such as installing structural health monitoring systems at key parts of bridges, deploying leak detection sensors along gas pipelines, and setting up water level monitoring devices at flood-prone areas. Various monitoring devices transmit data to the back-end monitoring center via wired or wireless means, where the back-end system stores the data and performs simple threshold judgment and alarm. In recent years, some cities have begun to explore the use of smart light poles to mount monitoring equipment, initially realizing the attempt at integrated equipment deployment; However, existing technical solutions involve the scattered deployment of monitoring equipment, a lack of unified planning, and the independent operation of various sensors, which creates information silos. This makes it difficult to integrate and share data, resulting in low efficiency in cross-departmental collaboration and limited risk identification methods. Therefore, this invention proposes a method and system for identifying safety risks in urban lifeline engineering based on smart light poles. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for identifying safety risks in urban lifeline projects based on smart light poles. Through a three-level architecture of "equipment deployment - edge processing - cloud application", it realizes multi-source data collection, real-time risk identification, urban appearance violation detection and visualization simulation, thereby improving the intelligence and efficiency of urban lifeline project safety management and urban appearance governance.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying safety risks in urban lifeline engineering based on smart light poles, comprising the following steps: Receive multi-source data from urban lifeline projects, including image data or monitoring data of three types of risks: fire, urban appearance violations, and bridge safety. Data preprocessing is performed on multi-source data, and a lightweight model is deployed to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger hierarchical alarms, output risk identification results and alarm information, and upload them to the cloud. In the three types of risk identification, the CNN fire detection model is used to identify the preprocessed fire image data, the ellipse fitting method for subpixel circle center identification and the IMU error compensation model are used to identify bridge safety data, and the YOLOv11 model is used to identify urban appearance violation image data. A 3D simulation platform is built in the cloud and integrated to form a complete safety risk identification system. The risk identification results and alarm information are input into the safety risk identification system, and the system outputs a visualized situation, emergency plan and system-level management functions.

[0005] Furthermore, the multi-source data is collected through chain cameras or surveillance cameras deployed on smart light poles, in conjunction with an IMU inertial navigation unit and a GPS synchronization module, as detailed below: (21) Risk-adaptive equipment selection: Different monitoring equipment is selected according to the monitoring needs of different risk types. For fire scenarios, a monitoring camera with a resolution of 1280×720 is used; for urban appearance violation scenarios, a monitoring camera with a resolution of 2560×1504 is used; and for bridge safety, a chain camera with a high frame rate of 60Hz and a measurement accuracy of 0.1mm is used, along with an IMU inertial navigation unit and a GPS synchronization module. (22) Layered point deployment: Based on the equipment collection coverage, monitoring area characteristics and risk distribution, determine the deployment location and interval of various monitoring devices on smart light poles, prioritize points with fewer monitoring blind spots and coverage of core risk areas, and add demonstration points or densify the deployment for key structures or road sections prone to accidents in combination with actual scenarios. (23) Data acquisition and transmission link construction: The edge computing unit receives data through the local area network, completes temporary caching and format standardization processing, and establishes data association tags containing device identification, acquisition time and location; the cloud receives data that needs to be stored for a long time or analyzed globally through the high-speed communication module.

[0006] Furthermore, in step (23), fire monitoring collects images at 25 frames per second and converts them to JPEG format, bridge safety collects structural images at 30Hz and tilt angle data at 10Hz and converts them to JSON format, and urban appearance violations collect video data at 30 frames per second.

[0007] Furthermore, multi-source data preprocessing is performed, and a lightweight model is deployed to extract features from image data or monitoring data of the three types of risks, identify them in real time, and trigger tiered alarms. The risk identification results and alarm information are output and uploaded to the cloud, as detailed below: (41) For fire risk identification scenarios: (41.1) The effective frames of the image are extracted using the OpenCV module, and environmental noise is removed by Gaussian filtering. The formula is as follows: in, Image pixel coordinates, The image was converted to JPEG format and a data index containing device ID, acquisition time, and location was created using a Gaussian kernel standard deviation. (41.2) Based on the fire risk identification requirements, deploy a CNN fire detection model trained and optimized with 10,000 fire images and load the pre-trained weight file; perform parameter pruning and precision quantization processing on the model by converting 32-bit floating-point to 16-bit integer to accurately identify flame and smoke features. (41.3) By analyzing the video stream in parallel through multi-threading, when the dual conditions of "preset time interval + appearance of flame / smoke features in the image" are met, a fire alarm is triggered and the flame / smoke area is selected in the image; the alarm information is synchronously transmitted to the cloud. (41.4) Containerization technology is used to deploy the CNN fire detection model, and the parameters of the fire detection model are updated regularly to adapt to the changes in flame / smoke characteristics of fires caused by different seasons and different combustibles; (42) For bridge safety monitoring scenarios: (42.1) For the bridge structure image data acquired by the chain camera, the effective frames are extracted by the frame extraction algorithm and the environmental interference is removed by Gaussian filtering noise reduction; for the tilt data acquired by the IMU inertial navigation unit, the jump points are removed by sliding window analysis, and the invalid data is filtered by referring to the anomaly identification logic in the multi-source fusion cleaning method of bridge service data; the processed image data and tilt data are converted into a standard format compatible with the edge-end algorithm, and a data index containing device ID, acquisition time and bridge monitoring point location is established to facilitate subsequent model calls; (42.2) Based on the requirements for bridge safety risk identification, an ellipse fitting method sub-pixel circle center identification model and an IMU error compensation model are deployed. The core formulas are as follows: in, The sub-pixel center coordinates of the circular marker at different times. The IMU pole tilt angle after error compensation; Indicates the bridge settlement value; The actual displacement of the fixed point is calculated using the lamp post tilt angle output by the IMU. The compensation formula is as follows: Calculation of the vertical height of the chain camera unit from the ground: in, Vertical height Mounting height for camera unit The horizontal tilt angle of the light pole is the angle of inclination towards the bridge. The angle of inclination of the light pole along the bridge direction; Calculation of transverse displacement of the chain camera bridge: in, The transverse bridge displacement is represented by an IMU error compensation model, which corrects the error caused by pole vibration on the tilt angle data. Parameter pruning and accuracy quantization are performed on both models. (42.3) Calculate the bridge settlement value through the model. When the settlement value exceeds the preset safety threshold, trigger an alarm and record the bridge location and settlement data at the time of the alarm. (42.4) The ellipse fitting model is lightweighted to ensure that the bridge deformation measurement accuracy is <0.1mm. The model is deployed using containerization technology, and the communication link status of the edge computing device is monitored in real time. When an anomaly occurs, a maintenance reminder is triggered. (43) Targeting violations of urban appearance regulations (43.1) The OpenCV module is used to extract the effective frames of the video, Gaussian filtering is used to remove environmental noise, the image is converted to JPEG format and a data index containing device ID, acquisition time and location is established; (43.2) Based on the requirements for identifying violations of urban appearance regulations, deploy a lightweight YOLOv11 object detection model trained and optimized by the urban appearance violation dataset, and load the pre-trained weight file; perform parameter pruning and precision quantization on the model to identify non-motorized vehicles and features of illegal parking, specifically including vehicles exceeding the parking line and parking on the sidewalk; (43.3) By analyzing the video stream frame by frame in parallel through multi-threading, when the characteristics of non-motorized vehicles being parked randomly are detected, an alarm is triggered and the illegal area is selected in the image; (43.4) Perform parameter pruning and precision quantization on the YOLOv11 model, deploy the model using containerization technology, monitor the operating status of edge computing devices in real time, and trigger maintenance reminders when abnormalities occur.

[0008] Furthermore, after training and optimization using fire images, the CNN fire detection model employs a fire risk confidence formula to quantify the reliability of the identification results and ensure alarm accuracy. The formula is as follows: in, A true positive result indicates that the actual fire was correctly identified. A false positive indicates that a non-fire incident was misdiagnosed. A false negative indicates that the fire has not been identified; an alarm is triggered when the confidence level is ≥85%.

[0009] Furthermore, a 3D simulation and deduction platform is built in the cloud, integrating it into a complete safety risk identification system. The risk identification results and alarm information are input into the safety risk identification system, which outputs visualized situational awareness, emergency plans, and system-level management functions, as detailed below: (61) Multi-source data fusion processing: After receiving data from the edge, the integrity and validity are verified, the consistency of associated risk data is verified and contradictory data is removed, and a three-dimensional index of "device-time-space" is used for storage. It has historical backtracking function and graded data quality; bridge data and water level data of flood-prone road sections are cross-verified, and data consistency is verified through the deviation rate formula: Data A is the bridge deck settlement-related water level data collected by the bridge monitoring unit, and data B is the water level data collected by the water level sensor in the flood-prone section during the same period. When the deviation rate is ≤5%, the data is judged to be consistent; when the deviation rate is >5%, it is marked as contradictory data and removed. At the same time, the collection time and device ID of the contradictory data are recorded for subsequent tracing and troubleshooting of equipment failure or environmental interference issues. (62) Construction of a three-dimensional simulation and deduction platform: In the modeling stage, Blender is used to build a three-dimensional model of the infrastructure; in the rendering stage, PBR material textures are used to simulate lighting / weather; in the functional development stage, dynamic simulation of risk scenarios is realized, and wind force / temperature / vehicle load parameter configuration is supported. (63) Decision support application development: Visual dashboards present safety status and risk information, contingency plan generation function provides rescue route / maintenance suggestions, and data export function supports exporting reports by filtering according to conditions; (64) Construction of a safety risk identification system: integrating smart light pole monitoring unit, edge computing unit and cloud platform.

[0010] According to a second aspect of the present invention, the present invention provides a safety risk identification system for urban lifeline engineering projects based on smart light poles, used to implement the safety risk identification method for urban lifeline engineering projects based on smart light poles described in the first aspect, comprising: The data receiving module is used to receive multi-source data from urban lifeline projects. The multi-source data includes image data or monitoring data of three types of risks: fire, urban appearance violations, and bridge safety. Data preprocessing is performed on multi-source data, and a lightweight model is deployed to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger hierarchical alarms, output risk identification results and alarm information, and upload them to the cloud. In the three types of risk identification, a CNN fire detection model is used to identify preprocessed fire image data, an ellipse fitting method for subpixel circle center identification and an IMU error compensation model are used to identify bridge safety data, a YOLOv11 model is used to identify urban appearance violation image data, and ... is used to identify road flooding image data; The platform construction and output module is used to build a 3D simulation and deduction platform in the cloud, integrate it to form a complete safety risk identification system, input the risk identification results and alarm information into the safety risk identification system, and output visualized situation, emergency plan and system-level management functions.

[0011] According to a third aspect of the present invention, the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein when the processor loads and executes the computer program, it employs the urban lifeline engineering safety risk identification method based on smart light poles described in the first aspect.

[0012] This invention has at least the following beneficial effects: 1. This invention fully utilizes the widely distributed hardware mounting capabilities of smart light poles and the advantages of edge computing pre-processing to integrate the monitoring and identification functions of four core risks—fire, road flooding, urban appearance violations, and bridge safety—into a unified platform. This completely breaks down the information silos formed by the decentralized deployment of traditional monitoring equipment and achieves comprehensive collection and efficient fusion of multi-source data.

[0013] 2. In this invention, a lightweight identification model with parameter pruning and integer quantization is deployed at the edge. For fire risks, it can achieve flame and smoke feature identification in seconds, with an average confidence level of over 92%. For bridge safety, it adopts ellipse fitting sub-pixel circle center identification combined with IMU vibration error compensation technology, which significantly reduces settlement calculation error. Through edge computing unit preprocessing, it avoids network congestion and delay caused by massive raw data back transmission. The risk identification response time is shortened from the traditional minutes to the seconds, winning valuable time for emergency response.

[0014] 3. The cloud-based 3D simulation platform of this invention can dynamically simulate complex risk scenarios such as fire spread and bridge settlement evolution. It can also be extended to the field of intelligent supervision of urban appearance violations, supporting configuration of multiple parameters such as wind force, temperature, and vehicle load, enabling managers to intuitively predict the development trend of risks.

[0015] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the method described in this invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0018] Example 1: Please see Figure 1 This invention provides a technical solution: a method for identifying safety risks in urban lifeline engineering based on smart light poles, comprising the following steps: S1. Deployment of smart light pole equipment and multi-source data acquisition Focusing on the four core risks of urban lifeline engineering (fire, gas leak, road flooding, and bridge safety), the selection, location deployment, and data collection and transmission link construction of smart light pole mounting equipment were completed. The specific steps are as follows: S101. Equipment Selection for Risk Adaptation Based on the monitoring needs of different risk types, select hardware devices with corresponding functions and match them with suitable edge computing units or auxiliary modules to ensure that the performance of the devices meets the accuracy, scope and real-time requirements of risk monitoring; at the same time, provide necessary auxiliary equipment to ensure the integrity and effectiveness of data collection. S102, Layered Point Layout Based on the equipment's data collection coverage, the characteristics of the monitoring area, and the risk distribution, determine the deployment location and spacing of various monitoring devices on smart light poles, prioritize locations with fewer monitoring blind spots and coverage of core risk areas, and add demonstration points or densify deployment for key structures or road sections prone to accidents, in combination with actual scenarios, to ensure no monitoring blind spots and coverage of key monitoring areas. S103, Data Acquisition and Transmission Link Setup The device collects data at a preset frequency. The collection frequency of risk monitoring data and urban appearance violation image data is based on the standard of fully recording the change process and meeting the real-time monitoring requirements. The edge device receives data through the local area network, completes temporary caching and format standardization processing, and establishes data association tags containing device identification, collection time, and location. The cloud receives data that needs to be stored for a long time or analyzed globally through a high-speed communication module and uploads it to the city lifeline cloud platform server to ensure that the data is traceable and accessible. S2. Perform data preprocessing on multi-source data, and deploy a lightweight model to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger graded alarms, output risk identification results and alarm information, and upload them to the cloud, as detailed below: Based on the edge computing unit mounted on the smart light pole, data preprocessing, deep learning model deployment, and risk identification and alarm are completed. The specific steps are as follows: S201, Multi-source data preprocessing Image data and monitoring data are processed separately. Image data is filtered to extract valid frames and reduce noise to remove environmental interference. Monitoring data is filtered to remove invalid data. All data is converted into a standard format compatible with edge-end algorithms and a data indexing system is established to ensure that the algorithm can efficiently access the data. S202, Lightweight Model Deployment Based on the identification requirements of different risk types, corresponding algorithm models are deployed. For image-related risks, models with feature extraction, target recognition, and tracking functions are deployed; for monitoring-related risks, models with data parsing, threshold judgment, and error correction functions are deployed to ensure that the models can quickly and accurately identify risks. S203, Real-time Risk Identification and Alarm Triggering Risks are categorized and monitored in real time with alarms. Image-related risks are analyzed in parallel using multi-threaded video streams, triggering alarms and marking risk areas when dual trigger conditions are met. Monitoring-related risks are analyzed in real time using sensor data, triggering alarms when preset safety thresholds are exceeded. Alarm information is synchronously transmitted to the cloud and backed up at the edge to prevent information loss. S204, Edge-end Optimization and Maintenance The algorithm model is lightweighted to reduce device resource consumption; containerization technology is adopted to improve the efficiency of algorithm updates and troubleshooting; the operating status of edge computing devices is monitored in real time, and maintenance reminders are triggered when anomalies occur to ensure stable device operation. Meanwhile, the parameters of the violation identification model are updated regularly to adapt to changes in violation behavior patterns. S3. Build a three-dimensional simulation and deduction platform in the cloud, integrate it to form a complete safety risk identification system, input the risk identification results and alarm information into the safety risk identification system, and output visualized situation, emergency plan and system-level management functions; Within the city lifeline cloud platform, data fusion, the construction of a 3D simulation platform, and the development of supporting systems are completed to achieve global decision support. The specific steps are as follows: S301, Multi-source data fusion processing It receives and verifies risk monitoring data and urban appearance violation identification data from the edge, and records the data transmission status; it performs consistency checks on related data and eliminates contradictory data; it uses a three-dimensional index storage of "device-time-space," has historical backtracking capabilities, and classifies data quality; it cross-verifies bridge data with water level data of flood-prone road sections, and data consistency is verified using the deviation rate formula: Data A is the bridge deck settlement-related water level data collected by the bridge monitoring unit, and data B is the water level data collected by the water level sensor in the flood-prone section during the same period. When the deviation rate is ≤5%, the data is judged to be consistent; when the deviation rate is >5%, it is marked as contradictory data and removed. At the same time, the collection time and device ID of the contradictory data are recorded for subsequent tracing and troubleshooting of equipment failure or environmental interference issues. S302, Construction of 3D Simulation and Deduction Platform The platform is built in phases. In the modeling phase, professional software is used to construct a 3D model containing various infrastructures to ensure consistency with the actual environment. In the rendering phase, the model is imported and professional technology is used to simulate real-time lighting and weather changes to enhance the realism of the simulation. In the functional development phase, dynamic simulation of risk scenarios and linkage response simulation are realized, and custom configuration of scene parameters is supported.

[0019] S303, Decision Support Application Development Develop a visual dashboard to intuitively present the safety situation and the distribution of urban appearance violations; develop a contingency plan generation function to provide emergency decision-making and violation handling references based on simulation results; develop a data export function to support the filtering and export of monitoring reports, violation identification results and simulation results by conditions, to meet the data application needs of management departments. S304, Construction of a Security Risk Identification System By integrating multiple units to form a complete system, the smart light pole monitoring unit realizes the integration of "data acquisition-transmission-edge processing" and provides front-end data support; the cloud platform functional unit integrates data services, model services and simulation modules to realize the collaborative application of data and models; the security management unit has the functions of access control, alarm handling and log auditing, and also adds a violation handling process management module to ensure the safe and stable operation of the system and the efficient promotion of violation handling.

[0020] The technical solution of the present invention will be further described below with reference to specific embodiments: Implementation Case 1: Fire Risk Identification Scenario The S101 uses a 1280×720 resolution surveillance camera with full-field image acquisition and high-definition imaging capabilities, paired with a Huawei Atlas500 smart station (with real-time data processing and algorithm deployment capabilities). The Huawei Atlas500 smart station has built-in hardware computing resources adapted to fire identification and supports multi-threaded data processing and model deployment. S102. Based on the coverage area of ​​the surveillance cameras, the characteristics and risk distribution of core fire risk areas such as industrial parks, the equipment will be deployed at the high-point smart light poles in the center of the industrial park to ensure that the field of view covers the entire park without any blind spots; additional monitoring points will be added for key sub-areas such as flammable material storage areas in the park, based on the actual scenario, to prioritize comprehensive monitoring of core risk areas. S103: The device acquires fire image data at a preset frequency of 25 frames per second. The acquisition frequency is designed to fully record the changes in flames and smoke and meet the requirements of real-time monitoring. The edge device receives data through a local area network, completes temporary caching and JPEG format standardization processing, and establishes data association tags containing device identification, acquisition time, and specific location in the industrial park. The cloud receives fire data that needs to be stored for a long time or analyzed globally through a high-speed communication module and uploads it to the city lifeline cloud platform server to ensure that the data is traceable and accessible. S201. Use the OpenCV module to extract the effective frames of the image, and remove environmental noise through Gaussian filtering. The formula is as follows: in, Image pixel coordinates, The standard deviation is the Gaussian kernel (based on the characteristics of fire images, this project uses the default value). This formula can effectively preserve the characteristics of the flame edge while reducing noise interference, convert the image to JPEG format and create a data index containing device ID, acquisition time and location; S202. Based on the fire risk identification requirements, deploy a convolutional neural network (CNN) fire detection model trained and optimized with 10,000 fire images, and load the pre-trained weight file; perform parameter pruning and precision quantization processing on the model by converting 32-bit floating-point to 16-bit integer to reduce the resource consumption of Huawei Atlas500 smart station and ensure that the model can quickly and accurately identify flame and smoke features. S203. By analyzing the video stream in parallel through multi-threading, when the dual conditions of "preset time interval (triggered once every fixed time) + flame / smoke features appearing in the image" are met, a fire alarm is triggered and the flame / smoke area is selected in the image; the alarm information is synchronously transmitted to the cloud-based city lifeline cloud platform and backed up locally on the Huawei Atlas500 smart station to avoid information loss; users can specify the location to save the alarm results for easy subsequent tracing and analysis. S204. Containerization technology is used to deploy the CNN fire detection model to improve the efficiency of algorithm updates and troubleshooting; the CPU utilization, memory usage, and communication link status of Huawei Atlas500 smart stations are monitored in real time, and maintenance reminders are triggered when the equipment is abnormal (such as excessive CPU utilization or insufficient memory) to ensure the stable operation of edge devices; the parameters of the fire detection model are updated regularly to adapt to the changes in flame / smoke characteristics of fires caused by different seasons and different combustibles. S301: Receive alarm data and image data uploaded from the edge terminal, record the transmission status through the data reception log, verify the data integrity, and store the data using a "device-time-space" three-dimensional index. S302. Use Blender software to build a 3D model of the industrial park, import it into the Unity engine and use PBR material textures to simulate the fire spread process in a fire scene, and support adjusting scene parameters such as wind force and temperature. S303. Mark the location and spread range of the fire on the visual dashboard, automatically generate fire rescue route plans, push early warning information to the fire control room, and support filtering and exporting fire monitoring reports by "fire occurrence time period". S304. The smart light pole monitoring unit integrates cameras and edge computing devices. The cloud platform data service module supports the query of fire images and alarm data. The safety management unit records the early warning viewing operation logs of personnel in the fire control room and divides operation permissions according to roles.

[0021] Implementation Case 2: Bridge Safety Monitoring Scenario S101. To address the monitoring needs of bridge safety risks (settlement, structural deformation), a chain camera with a maximum frame rate of 60Hz and a measurement accuracy within 0.1mm was selected, possessing high-speed image acquisition and high-precision displacement measurement capabilities. It is paired with an IMU inertial navigation unit capable of acquiring vibration data and compensating for vibration errors of the mounting platform, as well as a GPS synchronization module with time synchronization capabilities, ensuring that the equipment meets the accuracy and real-time requirements for bridge dynamic deformation monitoring. Simultaneously, auxiliary data transmission equipment is provided to ensure the complete acquisition of bridge structural images, tilt angles, and other data, providing foundational data for subsequent multi-source fusion and cleaning of bridge service data. S102. Based on the acquisition coverage of the chain camera and IMU inertial navigation unit, the structural characteristics of the Jiqingmen Bridge (mid-span and side-span are the key stress areas) and the risk distribution, the equipment will be deployed on the smart light poles on the north and south sides of the Jiqingmen Bridge, with 2 sets on each side. The equipment installation location covers the key stress areas such as the mid-span and side-span of the bridge to ensure that there are no blind spots in the monitoring. For key components such as bridge bearings that are prone to structural defects, no additional densification of equipment will be carried out (because the existing equipment can already cover the deformation monitoring needs of the key stress areas), and priority will be given to ensuring high-precision monitoring of key structures. S103. Data is collected at preset frequencies: chain cameras collect bridge structure image data at 30Hz, and IMU inertial navigation units collect tilt angle data at 10Hz. The collection frequency is based on the standard of fully recording the dynamic deformation process of the bridge and meeting the real-time monitoring requirements. The edge device receives data through the local area network, completes temporary caching and JSON format standardization processing, and establishes data association tags containing device identifiers (chain camera ID, IMU device ID), collection time, and bridge monitoring point locations (mid-span and side spans). The cloud receives bridge monitoring data that needs to be stored for a long time or analyzed globally through a high-speed communication module, uploads it to the city lifeline cloud platform server, and simultaneously synchronizes it to the bridge service data management platform to ensure data traceability and usability for subsequent fusion and cleaning. S201. For bridge structure image data acquired by chain cameras, a frame extraction algorithm is used to extract valid frames. Gaussian filtering is used to remove environmental interference (such as image blurring caused by reflection of vehicle lights on the bridge deck and atmospheric scattering). For tilt data acquired by IMU inertial navigation unit, a sliding window analysis is used to remove jump points (outliers caused by equipment vibration and electromagnetic interference). Referring to the anomaly identification logic in the multi-source fusion cleaning method of bridge service data, invalid data is filtered out. The processed image data and tilt data are converted into a standard format compatible with edge-end algorithms, and a data index containing device ID, acquisition time, and bridge monitoring point location is established for easy subsequent model calling. S202. Based on the requirements for bridge safety risk identification, deploy an ellipse fitting method sub-pixel circle center identification model and an IMU error compensation model. The core formulas are as follows: in, The sub-pixel center coordinates of the circular marker at different times. This is the IMU pole tilt angle after error compensation; the bridge settlement value can be accurately calculated using this formula. The subpixel circle center recognition model using circle fitting is used for high-precision positioning of circular markers in bridge structure images, improving the accuracy of settlement measurement; the actual displacement of the fixed point is calculated using the lamppost tilt angle output by the IMU, and the compensation formula is as follows: Calculation of the vertical height of the chain camera unit from the ground: in, Vertical height Mounting height for camera unit The horizontal tilt angle of the light pole is the angle of inclination towards the bridge. The angle of inclination of the light pole along the bridge direction.

[0022] Calculation of transverse displacement of the chain camera bridge: in, This represents the lateral displacement of the bridge. , The definition is the same as above. The IMU error compensation model is used to correct the error caused by the vibration of the light pole to the tilt angle data. Referring to the error correction approach in bridge service data cleaning, it ensures the accuracy of the monitoring data. Parameter pruning and precision quantization are performed on the two models to reduce the resource consumption of edge computing devices and ensure that the models can run quickly at the edge. S203. Calculate the bridge settlement value through the model. When the settlement value exceeds the preset safety threshold, trigger an alarm and record the bridge location and settlement data at the time of the alarm. S204. Lightweighting of the ellipse fitting model is performed to ensure that the bridge deformation measurement accuracy is <0.1mm. Containerization technology is used to deploy the model and the communication link status of the edge computing device is monitored in real time. Maintenance reminders are triggered when abnormalities occur. S301. Collect bridge settlement data, tilt data and alarm data uploaded from the edge terminal, and verify their consistency with water level data of flood-prone road sections (remove contradictory data). Store the data according to the "device-time-space" three-dimensional index and classify the data quality. S302. In the modeling stage, Blender software is used to construct a 3D model of the Jiqingmen Bridge, including details such as the bridge deck, piers, supports, and smart light poles, ensuring that the model is consistent with the actual bridge structure. In the rendering stage, the model is imported into the Unity engine, and PBR material mapping is used to simulate the material properties of different parts of the bridge (such as asphalt on the bridge deck and concrete on the piers), and to simulate the impact of different environmental conditions (such as vehicle load and temperature changes) on the bridge. In the functional development stage, dynamic simulation of bridge safety risk scenarios is implemented, which can simulate scenarios such as long-term settlement trends of the bridge and structural deformation caused by vehicle overloading. It supports the adjustment of parameters such as vehicle load and ambient temperature, and intuitively shows the degree of impact of different risk factors on bridge safety. S303. Display the bridge settlement change curve on the visualization dashboard, automatically generate bridge maintenance recommendation reports, and support exporting bridge safety monitoring results by "monthly / quarterly"; S304. The smart light pole monitoring unit integrates chain cameras and auxiliary monitoring equipment. The cloud platform model service module supports the calling and verification of ellipse fitting model. The safety management unit automatically assigns bridge maintenance tasks, records maintenance progress, and divides bridge data viewing permissions according to roles.

[0023] Implementation Case 3: Identification Scenario for Urban Appearance Violations (Illegal Parking of Non-motorized Vehicles) S101. To meet the needs of identifying urban appearance violations (illegal parking of non-motorized vehicles), a high-definition camera with all-weather image acquisition and dynamic target capture capabilities was selected, with a resolution of 2560×1504 to ensure clear capture of details of non-motorized vehicle parking. It was paired with an edge computing module with real-time data processing and algorithm deployment capabilities as an edge computing unit to ensure that the equipment meets the accuracy and real-time requirements for urban appearance violation image acquisition and to provide high-quality image data for subsequent violation identification. S102. Deploy the equipment at smart light poles at intersections of main urban roads and around commercial districts to ensure coverage of areas with high traffic and frequent parking of non-motorized vehicles. S103: Collect video data at 30 frames per second, transmit it to the edge computing device via the local area network, and after data caching and format standardization, upload it to the cloud as needed; S201. Use the OpenCV module to extract valid video frames, remove environmental noise through Gaussian filtering, convert the image to JPEG format, and create a data index containing device ID, acquisition time, and location. S202. Based on the requirements for identifying violations of urban appearance regulations, deploy a lightweight YOLOv11 object detection model trained and optimized using a dataset of violations of urban appearance regulations, and load the pre-trained weight file; perform parameter pruning and precision quantization on the model to reduce the resource consumption of the edge computing module and ensure that the model can quickly and accurately identify non-motorized vehicles and features of illegal parking (such as vehicles exceeding the parking line or occupying the sidewalk). S203. By analyzing the video stream frame by frame in parallel through multi-threading, when the characteristics of non-motorized vehicles being parked randomly are detected, an alarm is triggered and the illegal area is selected in the image; S204. Perform parameter pruning and precision quantization on the YOLOv11 model, deploy the model using containerization technology, monitor the operating status of edge computing devices in real time, and trigger maintenance reminders when abnormalities occur. S301: Receive violation alarm data and image data uploaded from the edge terminal, record the transmission status through the data reception log, verify the data integrity, and store the data using a "device-time-space" three-dimensional index; S302. Use Blender software to build a 3D model of the monitoring area, import it into the Unity engine to simulate the impact range of illegal parking, and support adjusting scene parameters such as pedestrian flow and vehicle flow. S303. Mark the location of illegally parked vehicles on the visual dashboard, automatically generate a handling reminder and push it to the mobile terminal of the urban management department, and support filtering and exporting violation monitoring reports by "time period of violation". S304. The smart light pole monitoring unit integrates high-definition cameras and edge computing devices. The cloud platform data service module supports querying violation images and alarm data. The security management unit records the violation handling operation logs of urban management personnel and divides operation permissions according to roles.

[0024] Example 2: This embodiment provides a safety risk identification system for urban lifeline engineering projects based on smart light poles, used to implement the safety risk identification method for urban lifeline engineering projects based on smart light poles described in Embodiment 1, including: The data receiving module is used to receive multi-source data from urban lifeline projects. The multi-source data includes image data or monitoring data of three types of risks: fire, urban appearance violations, and bridge safety. Data preprocessing is performed on multi-source data, and a lightweight model is deployed to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger hierarchical alarms, output risk identification results and alarm information, and upload them to the cloud. In the three types of risk identification, a CNN fire detection model is used to identify preprocessed fire image data, an ellipse fitting method for subpixel circle center identification and an IMU error compensation model are used to identify bridge safety data, a YOLOv11 model is used to identify urban appearance violation image data, and ... is used to identify road flooding image data; The platform construction and output module is used to build a 3D simulation and deduction platform in the cloud, integrate it to form a complete safety risk identification system, input the risk identification results and alarm information into the safety risk identification system, and output visualized situation, emergency plan and system-level management functions.

[0025] Example 3: The present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The memory stores the computer program capable of running on the processor. When the processor loads and executes the computer program, it adopts the urban lifeline engineering safety risk identification method based on smart light poles described in Embodiment 1.

[0026] It should be noted that the terminal device can be a computer device such as a desktop computer, a laptop computer, or a cloud server, and the terminal device includes, but is not limited to, a processor and a memory. For example, the terminal device may also include input / output devices, network access devices, and buses.

[0027] Furthermore, the processor can be a central processing unit (CPU). Of course, depending on the actual use, other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. can also be used. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it in this regard.

[0028] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0029] For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances. When an element is referred to as being "assembled on," "mounted on," "fixed to," or "set on" another element, it may be directly on the other element or there may be an intermediate element present. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be an intermediate element present. The terms "vertical," "horizontal," "upper," "lower," "left," "right," and similar expressions used herein are for illustrative purposes only and do not represent the only possible embodiments.

[0030] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0031] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this disclosure. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

Claims

1. A method for identifying safety risks in urban lifeline engineering based on smart light poles, which applies edge computing units and cloud computing for computation, characterized in that... Includes the following steps: Receive multi-source data from urban lifeline projects, including image data or monitoring data of three types of risks: fire, urban appearance violations, and bridge safety. Data preprocessing is performed on multi-source data, and a lightweight model is deployed to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger hierarchical alarms, output risk identification results and alarm information, and upload them to the cloud. In the three types of risk identification, the CNN fire detection model is used to identify the preprocessed fire image data, the ellipse fitting method for subpixel circle center identification and the IMU error compensation model are used to identify bridge safety data, and the YOLOv11 model is used to identify urban appearance violation image data. A 3D simulation platform is built in the cloud and integrated to form a complete safety risk identification system. The risk identification results and alarm information are input into the safety risk identification system, and the system outputs a visualized situation, emergency plan and system-level management functions.

2. The method for identifying safety risks in urban lifeline engineering based on smart light poles according to claim 1, characterized in that: The multi-source data is collected through chain cameras or surveillance cameras deployed on smart light poles, in conjunction with an IMU inertial navigation unit and a GPS synchronization module, as detailed below: (21) Risk-adaptive equipment selection: Different monitoring equipment is selected according to the monitoring needs of different risk types. For fire scenarios, a monitoring camera with a resolution of 1280×720 is used; for urban appearance violation scenarios, a monitoring camera with a resolution of 2560×1504 is used; and for bridge safety, a chain camera with a high frame rate of 60Hz and a measurement accuracy of 0.1mm is used, along with an IMU inertial navigation unit and a GPS synchronization module. (22) Layered point deployment: Based on the equipment collection coverage, monitoring area characteristics and risk distribution, determine the deployment location and interval of various monitoring devices on smart light poles, prioritize points with fewer monitoring blind spots and coverage of core risk areas, and add demonstration points or densify the deployment for key structures or road sections prone to accidents in combination with actual scenarios. (23) Data acquisition and transmission link construction: The edge computing unit receives data through the local area network, completes temporary caching and format standardization processing, and establishes data association tags containing device identification, acquisition time and location; The cloud receives data that needs to be stored long-term or analyzed globally through a high-speed communication module.

3. The method for identifying safety risks in urban lifeline engineering based on smart light poles according to claim 2, characterized in that: In step (23), fire monitoring collects images at 25 frames per second and converts them to JPEG format; bridge safety collects structural images at 30Hz and tilt angle data at 10Hz and converts them to JSON format; and urban appearance violations collect video data at 30 frames per second.

4. The method for identifying safety risks in urban lifeline engineering based on smart light poles according to claim 3, characterized in that: Multi-source data is preprocessed, and a lightweight model is deployed to extract features from image or monitoring data of three types of risks, identify them in real time, and trigger tiered alarms. The risk identification results and alarm information are output and uploaded to the cloud, as detailed below: (41) For fire risk identification scenarios: (41.1) The effective frames of the image are extracted using the OpenCV module, and environmental noise is removed by Gaussian filtering. The formula is as follows: in, Image pixel coordinates, The image was converted to JPEG format and a data index containing device ID, acquisition time, and location was created using a Gaussian kernel standard deviation. (41.2) Based on the fire risk identification requirements, deploy a CNN fire detection model trained and optimized with 10,000 fire images and load the pre-trained weight file; perform parameter pruning and precision quantization processing on the model by converting 32-bit floating-point to 16-bit integer to accurately identify flame and smoke features. (41.3) By analyzing the video stream in parallel through multi-threading, when the dual conditions of "preset time interval + appearance of flame / smoke features in the image" are met, a fire alarm is triggered and the flame / smoke area is selected in the image; the alarm information is synchronously transmitted to the cloud. (41.4) Containerization technology is used to deploy the CNN fire detection model, and the parameters of the fire detection model are updated regularly to adapt to the changes in flame / smoke characteristics of fires caused by different seasons and different combustibles; (42) For bridge safety monitoring scenarios: (42.1) For the bridge structure image data acquired by the chain camera, the effective frames are extracted by the frame extraction algorithm and the environmental interference is removed by Gaussian filtering noise reduction; for the tilt data acquired by the IMU inertial navigation unit, the jump points are removed by sliding window analysis, and the invalid data is filtered by referring to the anomaly identification logic in the multi-source fusion cleaning method of bridge service data; the processed image data and tilt data are converted into a standard format compatible with the edge-end algorithm, and a data index containing device ID, acquisition time and bridge monitoring point location is established to facilitate subsequent model calls; (42.2) Based on the requirements for bridge safety risk identification, an ellipse fitting method sub-pixel circle center identification model and an IMU error compensation model are deployed. The core formulas are as follows: in, The sub-pixel center coordinates of the circular marker at different times. The IMU pole tilt angle after error compensation; Indicates the bridge settlement value; The actual displacement of the fixed point is calculated using the lamp post tilt angle output by the IMU. The compensation formula is as follows: Calculation of the vertical height of the chain camera unit from the ground: in, Vertical height Mounting height for camera unit The horizontal tilt angle of the light pole is [missing information]. The angle of inclination of the light pole along the bridge direction; Calculation of transverse displacement of the chain camera bridge: in, The transverse bridge displacement is represented by an IMU error compensation model, which corrects the error caused by pole vibration on the tilt angle data. Parameter pruning and accuracy quantization are performed on both models. (42.3) Calculate the bridge settlement value through the model. When the settlement value exceeds the preset safety threshold, trigger an alarm and record the bridge location and settlement data at the time of the alarm. (42.4) The ellipse fitting model is lightweighted to ensure that the bridge deformation measurement accuracy is <0.1mm. The model is deployed using containerization technology, and the communication link status of the edge computing device is monitored in real time. When an anomaly occurs, a maintenance reminder is triggered. (43) Targeting violations of urban appearance regulations (43.1) The OpenCV module is used to extract the effective frames of the video, Gaussian filtering is used to remove environmental noise, the image is converted to JPEG format and a data index containing device ID, acquisition time and location is established; (43.2) Based on the requirements for identifying violations of urban appearance regulations, deploy a lightweight YOLOv11 object detection model trained and optimized by the urban appearance violation dataset, and load the pre-trained weight file; perform parameter pruning and precision quantization on the model to identify non-motorized vehicles and features of illegal parking, specifically including vehicles exceeding the parking line and parking on the sidewalk; (43.3) By analyzing the video stream frame by frame in parallel through multi-threading, when the characteristics of non-motorized vehicles being parked randomly are detected, an alarm is triggered and the illegal area is selected in the image; (43.4) Perform parameter pruning and precision quantization on the YOLOv11 model, deploy the model using containerization technology, monitor the operating status of edge computing devices in real time, and trigger maintenance reminders when abnormalities occur.

5. The method for identifying safety risks in urban lifeline engineering based on smart light poles according to claim 4, characterized in that: After being trained and optimized using fire images, the CNN fire detection model employs a fire risk confidence formula to quantify the reliability of the identification results and ensure alarm accuracy. The formula is as follows: in, A true positive result indicates that the actual fire was correctly identified. A false positive indicates that a non-fire incident was misdiagnosed. A false negative indicates that the fire has not been identified; an alarm is triggered when the confidence level is ≥85%.

6. The method for identifying safety risks in urban lifeline engineering based on smart light poles according to claim 5, characterized in that: A 3D simulation platform is built in the cloud, integrating it to form a complete safety risk identification system. The risk identification results and alarm information are input into the safety risk identification system, which outputs visualized situational awareness, emergency plans, and system-level management functions, as detailed below: (61) Multi-source data fusion processing: After receiving data from the edge, the integrity and validity are verified, the consistency of associated risk data is verified and contradictory data is removed, and a three-dimensional index of "device-time-space" is used for storage. It has historical backtracking function and graded data quality; bridge data and water level data of flood-prone road sections are cross-verified, and data consistency is verified through the deviation rate formula: Data A is the bridge deck settlement-related water level data collected by the bridge monitoring unit, and data B is the water level data collected by the water level sensor in the flood-prone section during the same period. When the deviation rate is ≤5%, the data is judged to be consistent; when the deviation rate is >5%, it is marked as contradictory data and removed. At the same time, the collection time and device ID of the contradictory data are recorded for subsequent tracing and troubleshooting of equipment failure or environmental interference issues. (62) Construction of a three-dimensional simulation and deduction platform: In the modeling stage, Blender is used to build a three-dimensional model of the infrastructure; in the rendering stage, PBR material textures are used to simulate lighting / weather; in the functional development stage, dynamic simulation of risk scenarios is realized, and wind force / temperature / vehicle load parameter configuration is supported. (63) Decision support application development: Visual dashboards present safety status and risk information, contingency plan generation function provides rescue route / maintenance suggestions, and data export function supports exporting reports by filtering according to conditions; (64) Construction of a safety risk identification system: integrating smart light pole monitoring unit, edge computing unit and cloud platform.

7. A safety risk identification system for urban lifeline engineering projects based on smart light poles, used to implement the safety risk identification method for urban lifeline engineering projects based on smart light poles as described in any one of claims 1 to 6, characterized in that, include: The data receiving module is used to receive multi-source data from urban lifeline projects. The multi-source data includes image data or monitoring data of three types of risks: fire, urban appearance violations, and bridge safety. Data preprocessing is performed on multi-source data, and a lightweight model is deployed to extract features from image data or monitoring data of three types of risks, identify them in real time, trigger hierarchical alarms, output risk identification results and alarm information, and upload them to the cloud. In the three types of risk identification, the CNN fire detection model is used to identify the preprocessed fire image data, the ellipse fitting method for subpixel circle center identification and the IMU error compensation model are used to identify bridge safety data, and the YOLOv11 model is used to identify urban appearance violation image data. The platform construction and output module is used to build a 3D simulation and deduction platform in the cloud, integrate it to form a complete safety risk identification system, input the risk identification results and alarm information into the safety risk identification system, and output visualized situation, emergency plan and system-level management functions.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that, When the processor loads and executes the computer program, it employs the urban lifeline engineering safety risk identification method based on smart light poles, as described in any one of claims 1 to 6.

Images