Positioning and visual data processing system and method for comprehensive pipe gallery operation

Abstract

The invention discloses a comprehensive pipe gallery operation positioning and visual data processing system and method, belongs to the technical field of intelligent monitoring and management, and aims to solve the problems of poor data synchronization, weak dynamic adaptation, high false alarm rate and the like. According to the system, positioning and visual data are calibrated at a millisecond level through the timestamp alignment module, and multi-source data synchronization is ensured; a graph neural network GNN is used for fusing multi-sensor data, a personnel-equipment-environment space-time correlation graph is constructed, a dynamic digital twinborn body is generated in combination with a BIM model, and real-time monitoring and fault early warning are achieved; the visual optimization module can dynamically correct false detection, significantly reduce false alarm rate and bandwidth occupation, and optimize data transmission; the data processing method based on the system comprises the following steps: calibrating multi-modal data, and compensating clock drift; constructing a space-time correlation graph, and distributing dynamic weights; analyzing a visual result in real time, training and updating model parameters, and compressing data to reduce bandwidth; and the method is suitable for power stations, urban pipe galleries and other scenes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent monitoring and management, and particularly relates to a positioning and visual data processing system and method for comprehensive pipe gallery operation. BACKGROUND

[0002] In the technical field of intelligent monitoring and management, as an important part of urban infrastructure, comprehensive pipe gallery undertakes the task of laying various pipelines such as power, communication, gas, water supply and drainage, and its safe operation is directly related to the normal operation of the city and the quality of life of residents. With the acceleration of urbanization, the scale of comprehensive pipe gallery is expanding, and the internal structure is becoming increasingly complex, which puts forward higher requirements for the intelligentization and refinement of operation and maintenance management.

[0003] Currently, the monitoring and inspection system of comprehensive pipe gallery mostly adopts a decentralized architecture, and each subsystem (such as positioning system, visual monitoring system, environmental monitoring system, etc.) operates independently, with non-uniform data formats, which leads to difficulties in multi-source data fusion and makes it difficult to achieve comprehensive, real-time and accurate monitoring of the state of the pipe gallery. For example, the positioning system and the visual monitoring system have poor time synchronization, which cannot accurately reflect the real-time position and state of personnel and equipment in the pipe gallery, bringing great challenges to fault warning and emergency response.

[0004] In the prior art, for example, CN113128473A proposes a patrol system for underground comprehensive pipe gallery, which uses a patrol robot for patrol, which improves the patrol efficiency to a certain extent, but the time synchronization of positioning data and visual data still depends on traditional methods, and it is difficult to achieve millisecond-level precision calibration. In a complex environment, such as the presence of multiple interference sources inside the pipe gallery, the data synchronization is poor, which affects the accuracy of fault warning. In addition, when processing visual data in a complex environment, the system may still be affected by interference factors such as shadows and reflections, resulting in a high false alarm rate. For example, when identifying equipment failure, normal equipment shadows may be misjudged as faults, with a false alarm rate of more than 15%, which seriously increases unnecessary operation and maintenance costs.

[0005] On the other hand, for example, CN119152637A proposes a method for monitoring and early warning of electrical fires in urban underground utility tunnels, which improves the accuracy of fire monitoring through multi-sensor fusion. However, this patent does not explicitly mention how to solve the time synchronization problem of multi-source data. In actual application, the sampling frequency and data transmission delay of different sensors may cause poor data synchronization, which in turn affects the timeliness and accuracy of fire warning. At the same time, this system mainly relies on fixed monitoring models and thresholds, and is difficult to automatically adjust the monitoring strategy according to the changes in the internal environment of the tunnel (such as light, temperature, humidity, etc.), and has insufficient dynamic adaptation capability. For example, in a high-temperature and high-humidity environment, the failure mode of electrical equipment may change, but the existing system may not be able to identify this change in time, resulting in ineffective warning.

[0006] In addition, the existing system also has obvious problems in data transmission efficiency. Whether it is the inspection robot system in CN113128473A or the electrical fire monitoring system in CN119152637A, when transmitting high-definition video data or multi-sensor data, they may face the challenge of large bandwidth occupation. When a large number of devices are deployed in the tunnel, the delay and packet loss rate of data transmission may increase, affecting the real-time performance and reliability of the system. Especially in emergency situations such as fire, timely transmission and processing of data is crucial to reduce losses.

[0007] In terms of data fusion and processing, existing technologies mostly use traditional data processing methods such as threshold comparison and rule matching, which are not sufficient in dealing with complex and variable tunnel environments. For example, visual detection systems are prone to false positives when identifying shadows, reflections and other interference factors, seriously affecting the reliability and practicality of the monitoring system. At the same time, the existing system also has the problem of large bandwidth occupation in data transmission, especially the transmission of high-definition video data, which puts high requirements on network bandwidth, increasing the difficulty of system deployment and operation and maintenance.

[0008] Although the rapid development of artificial intelligence, big data, and Internet of Things technologies in recent years has provided new possibilities for the intelligent upgrading of comprehensive tunnel monitoring systems, such as the introduction of advanced technologies such as graph neural networks (GNN), attention mechanisms, and LSTM, which can effectively improve data synchronization, reduce false detection rates, and optimize data transmission, there are still key technical problems to be solved, such as how to ensure the time synchronization accuracy of multi-modal data, how to dynamically allocate weights to different modal data, how to correct false detections in visual detection in real time, and how to optimize the transmission efficiency of visual data.

[0009] In summary, the existing comprehensive pipe gallery monitoring system has obvious deficiencies in data synchronization, dynamic adaptability, visual detection false alarm rate and data transmission efficiency, and it is difficult to meet the growing intelligent operation and maintenance requirements. Therefore, it has important practical significance and application value to develop an efficient and reliable positioning and visual data processing system and method for comprehensive pipe gallery operation. SUMMARY

[0010] The technical problem to be solved by the present application is to provide a positioning and visual data processing system and method for comprehensive pipe gallery operation, to solve the problems of time synchronization of positioning data and visual data, difficulty in multi-source data fusion, high visual monitoring false detection rate and large transmission bandwidth occupation in comprehensive pipe gallery operation, and to overcome the specific limitations of the prior art in precise synchronization of positioning and visual data, inaccurate spatio-temporal correlation graph construction, high visual detection model false alarm rate and low visual data transmission efficiency.

[0011] To achieve the above-mentioned target, the following technical solutions are adopted in the present application: The present application provides a positioning and visual data processing system and method for comprehensive pipe gallery operation, which is as follows: A positioning and visual data processing system for comprehensive pipe gallery operation, comprising: 1. A timestamp alignment module: used to obtain positioning modal and visual modal information in comprehensive pipe gallery operation, analyze modal conditions and extract material data, calculate the timestamp accuracy of each modal, and take the modal with the highest timestamp accuracy as the time reference target to dynamically calibrate the positioning data and visual data. The module further comprises: 1) A modal condition analysis unit: collects modal condition data of positioning modal and visual modal, including at least one of equipment error, network delay, and environmental temperature and humidity; 2) An accuracy calculation unit: calculates the timestamp accuracy of each modal based on the modal condition data; 3) A dynamic calibration unit: taking the modal with the highest timestamp accuracy as the reference, predicting and compensating the dynamic error of other modal through LSTM network.

[0012] 2. A data processing module: used to construct a spatio-temporal correlation graph of personnel-equipment-environment based on the timestamp-aligned data, and generate a pipe gallery monitoring model combined with a BIM model. The module further comprises: 1) A spatio-temporal correlation graph construction unit: fuses positioning data, visual data and environmental sensor data through graph neural network GNN to generate the spatio-temporal correlation relationship of personnel-equipment-environment; wherein the graph neural network GNN adopts an attention mechanism to assign dynamic weights to different modal data.

[0013] 2) Digital twin generation unit: combine BIM model and spatiotemporal correlation graph to generate a dynamically updated pipe gallery monitoring model, supporting real-time response to cable addition or removal or layout adjustment; the unit further includes: (1) Risk prediction subunit: simulate cable overload and equipment aging scenarios based on historical failure database to provide early warning of potential failures; use Monte Carlo simulation method, combined with real-time operating parameters of equipment (including current and temperature) and historical failure data, to generate a failure probability distribution map and mark high-risk areas; (2) Dynamic update subunit: automatically scan and update the digital twin after pipe gallery layout adjustment.

[0014] 3, Visual optimization module: for real-time analysis of visual monitoring result data, dynamic optimization of visual detection model, and transmission of optimized visual data before transmission. The module further includes: 1) Misjudgment judgment unit: when the monitoring result is abnormal but the feedback is normal, extract similar feature data (including shadow shape and light angle), train the visual detection model; identify the misjudgment type by optimizing the judgment model, and the expression of the optimization judgment model is: (1) In the formula, is the input data, is the visual monitoring result data, is the optimization standard (such as the feedback is feedback abnormal and is caused by incomplete model training); indicates that the corresponding visual monitoring result data meets the optimization standard; the output data is the optimization judgment value , the optimization judgment value is 1 or 0; when the set condition is met, the model optimization is triggered; 2) Model optimization unit: complete model parameter update to reduce false alarm rate; use transfer learning method, use pre-trained model parameters of similar pipe gallery scenes as initial value, fine-tune with current scene data to shorten model convergence time; 3) Transmission optimization unit: dynamically adjust visual data resolution and encoding format according to pipe gallery area requirements, including: compressing non-critical areas to reduce transmission bandwidth occupation; maintaining the quality of key areas for lossless transmission to ensure that data delay does not fall below the set parameter; the unit further includes: (1) Bandwidth adaptive subunit: real-time monitoring of network bandwidth, when the bandwidth decreases to the set value, automatically reduce the resolution of non-critical areas and switch the encoding; (2) Priority scheduling subunit: mark the highest priority for visual data of critical areas (including cable joints and equipment compartments) to ensure priority transmission in bandwidth competition.

[0015] Further, the application also provides a positioning and visual data processing method for comprehensive pipe gallery operation, comprising the following steps: Step 1: acquiring and calibrating positioning modal and visual modal data through a timestamp alignment module; Step 1.1: collecting modal condition data and calculating timestamp accuracy; Step 1.2: selecting the modal with the highest accuracy as the reference to dynamically calibrate other modal data; Step 1.3: compensating clock drift error through an LSTM network.

[0016] Step 2: constructing a space-time correlation graph and generating a digital twin through a data processing module.

[0017] Step 3: dynamically correcting false positives and optimizing visual data transmission through a visual optimization module; Step 3.1: real-time analysis of visual monitoring results, identification of false positive types and extraction of similar feature data; Step 3.2: dynamically training a visual detection model to complete parameter updating; Step 3.3: compressing visual data according to the requirements of the pipe gallery area to reduce transmission bandwidth occupation.

[0018] The positioning and visual data processing system and method for comprehensive pipe gallery operation provided by the application have the following beneficial effects: 1. The application effectively solves the problem of unsynchronized positioning data and visual data in comprehensive pipe gallery operation and the difficulty of multi-source data fusion, and through a high-precision timestamp alignment method (error ≤ 50 ms, 60%-80% higher than traditional methods), the time synchronization of multi-modal data is ensured, and the accuracy of data fusion is improved.

[0019] 2. The application successfully overcomes the technical limitations of the prior art, such as the difficulty of precise synchronization of positioning and visual data and the inaccuracy of space-time correlation graph construction, and through a graph neural network (GNN) and an attention mechanism, an accurate space-time correlation graph is constructed, a pipe gallery monitoring model is dynamically generated, and the accuracy and real-time performance of data fusion are improved.

[0020] 3. The model updating time after cable addition or reduction is shortened from 72 hours to 2 hours, the operation and maintenance efficiency is improved by 90%, and the laser scanning + ICP registration technology realizes automatic updating within 2 hours after layout adjustment, which is 12 times more efficient than manual updating.

[0021] 4. The application introduces an LSTM network for dynamic error compensation, captures nonlinear dependence, realizes higher-precision error compensation, and enhances the adaptability and stability of the system.

[0022] 5. This invention combines BIM models to generate dynamically updated utility tunnel monitoring models, supporting simulation and risk prediction, and significantly improving the level of intelligence in monitoring.

[0023] 6. This invention optimizes and analyzes visual monitoring results data in real time through a visual optimization module, dynamically adjusts the visual monitoring model, reduces the false alarm rate from 15% to 3%, reduces the workload of manual review by 80%, and solves the problem of false detection caused by changes in the pipe gallery in the existing visual detection model.

[0024] 7. This invention introduces an optimization judgment model to automatically identify visual monitoring result data that needs optimization, thereby improving optimization efficiency.

[0025] 8. This invention uses a transmission optimization unit to dynamically adjust the resolution and encoding format of visual data according to the needs of the utility tunnel area. It combines priority scheduling with ROI encoding, achieving a latency of ≤200ms in critical areas and reducing bandwidth usage in non-critical areas by 50%, thus balancing transmission efficiency and image quality and reducing bandwidth usage.

[0026] 9. This invention identifies transmission optimization points in real time and determines the optimal transmission optimization method, thereby improving the transmission efficiency and synchronization of visual data.

[0027] 10. The modular design of this invention enables each functional module to operate independently and work collaboratively, facilitating system maintenance and upgrades, and improving the system's flexibility and scalability.

[0028] 11. This invention effectively solves the problems of data synchronization, dynamic adaptation and intelligence in integrated utility tunnel monitoring, and has broad application prospects and economic value.

[0029] 12. The implementation effect of the present invention has been verified by theoretical derivation and simulation experiments. Key indicators such as time synchronization accuracy, model update speed and false alarm rate reduction are all superior to the existing technology. The annual fault response time is reduced by 120 hours and the annual direct economic loss is reduced by millions. It has significant economic benefits, strong practicality and high promotion value. Attached Figure Description

[0030] Figure 1 This is an architecture diagram of the system of the present invention. Detailed Implementation

[0031] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments: Example 1 like Figure 1 As shown, this embodiment provides a positioning and visual data processing system for integrated utility tunnel operations, including a timestamp module, a data processing module, and a visual optimization module: The timestamp module is used to align the positioning data and visual data with timestamps, determine the time reference target, and align the corresponding positioning data and visual data with timestamps according to the time reference target.

[0032] Furthermore, the determination of the time baseline target includes: Obtain the information modes corresponding to the integrated utility tunnel operation, and obtain the modal conditions that the information modes have; Obtain material data for the information modality based on modal conditions; calculate the timestamp accuracy of the information modality based on the material data; Determine the time reference target based on the timestamp accuracy.

[0033] Furthermore, information modalities include location modalities and visual modalities.

[0034] Furthermore, the time reference target is the information modality with the highest timestamp accuracy.

[0035] Furthermore, the time reference target is determined based on the timestamp accuracy, including: Each information modality is used as a time reference target to form a corresponding timestamp alignment method; timestamp alignment simulation is performed based on the timestamp alignment method to obtain the security accuracy and alignment duration corresponding to the timestamp alignment method. The priority of timestamp alignment methods is evaluated based on security accuracy and alignment duration, and the information modality corresponding to the highest priority timestamp alignment method is marked as the time reference target.

[0036] Furthermore, the priority of corresponding timestamp alignment methods is evaluated based on security accuracy and alignment duration, including: Set the priority formula, which is: (2) In the formula, Priority value; , These are two different proportionality coefficients, with a range of values ​​of [missing value]. , ; It is a natural constant; For safety and accuracy; For alignment duration; The priority of the corresponding timestamp alignment method is determined according to the priority value from high to low.

[0037] Furthermore, the timestamp precision of the corresponding information modality is calculated based on the source data, including: Determine the conditional precision of the information modality under the given modal conditions based on the source data; and statistically analyze the conditional share of the modal conditions in real time. Mark the modal conditions as , , The number of modal conditions; condition fraction and condition precision are respectively denoted as and ; The timestamp accuracy of the corresponding information modality is calculated based on the accuracy calculation formula.

[0038] Furthermore, the formula for calculating accuracy is: (3) In the formula, This refers to the timestamp precision.

[0039] The data processing module is used to process the positioning data and video data to obtain the integrated utility tunnel monitoring model; to display the monitoring model to the user in real time; and to monitor the integrated utility tunnel in real time based on the monitoring model.

[0040] The visual optimization module is used to perform real-time optimization analysis on visual data processing and to acquire visual monitoring result data in real time. The visual monitoring result data includes monitoring results, monitoring features, and result feedback. Real-time analysis of visual monitoring data to determine visual optimization points; Based on the visual optimization points, set the corresponding model optimization data, and optimize and adjust the preset visual monitoring model based on the model optimization data.

[0041] Furthermore, the visual monitoring results data are analyzed in real time, including: Set optimization criteria, and establish an optimization judgment model based on the optimization criteria. The expression of the optimization judgment model is: (1) In the formula, For input data, For visual monitoring results data, To optimize standards; This indicates that the corresponding visual monitoring results meet the optimization criteria; the output data is the optimization judgment value. The optimal judgment value is 1 or 0; By optimizing the judgment model, the visual monitoring result data and optimization standards are analyzed to obtain the optimized judgment value of the visual monitoring result data. When the optimization judgment value is 0, no corresponding operation is performed; When the optimization judgment value is 1, the visual optimization point is determined based on the visual monitoring result data.

[0042] Furthermore, the vision optimization module also includes a transmission optimization unit, which is used to perform transmission optimization analysis on the vision data, determine the transmission optimization method of the vision data, and process the vision data before transmission according to the transmission optimization method.

[0043] Furthermore, the visual data transmission optimization analysis is performed, including: Real-time acquisition of visual data processing records; real-time analysis of visual data processing records based on preset visual synchronization requirements to determine transmission optimization points; Based on the transmission optimization points, relevant historical visual data is obtained, and the historical visual data is marked as optimized transmission material. The optimized transmission material is analyzed according to the visual monitoring requirements to determine the candidate optimization methods for the transmission optimization points. The candidate optimization methods are then screened to determine the corresponding transmission optimization methods for the transmission optimization points.

[0044] Example 2 In another preferred embodiment, based on the above embodiment 1, this embodiment provides a positioning and visual data processing system for integrated utility tunnel operations, including a timestamp module, a data processing module and a visual optimization module; The timestamp module is used to timestamp align the positioning data and visual data to determine the time reference target, which serves as the reference target for subsequent timestamp alignment; and to timestamp align the corresponding positioning data and visual data according to the time reference target.

[0045] Furthermore, the determination of the time baseline target includes: The modal conditions of various information modalities in the current integrated utility tunnel operation are obtained. The information modalities include positioning modality and visual modality. Modal conditions, such as data acquisition equipment, environment, network, and other information related to timestamp accuracy, are determined based on the information modality. Based on the modal conditions, obtain the material data of the corresponding modal. The material data includes historical data or simulated data and timestamp error data under the corresponding modal conditions. The simulated data is data simulated under the corresponding modal conditions, such as simulated positioning data. Calculate the timestamp accuracy of the corresponding information modality based on the source data; determine the time reference target based on the timestamp accuracy.

[0046] Furthermore, information modalities also include other types of modalities that require timestamp alignment from integrated utility tunnel operations, such as information modalities corresponding to collected data on temperature, humidity, etc.

[0047] Furthermore, a time reference target is determined based on the timestamp accuracy, and the information mode with the highest timestamp accuracy is marked as the time reference target.

[0048] Furthermore, the time reference target is determined based on the timestamp accuracy, including: Timestamp alignment simulations are conducted, with each simulation serving as a time reference target under different timestamp alignment conditions. Based on the simulation results, the impact of the timestamp accuracy on the two scenarios is determined. Specifically, the impact of the current timestamp alignment on the safety accuracy of integrated utility tunnel operations is analyzed based on the simulation results. The safety accuracy and alignment time corresponding to the respective timestamp alignment method are obtained, where alignment time is the time spent processing the corresponding timestamp alignment. For example, historical data with information modalities containing safety issues are used to simulate timestamp alignment, obtaining simulated data after timestamp alignment adjustment. The simulated data is then analyzed to statistically determine the safety accuracy and alignment time corresponding to the respective timestamp alignment method.

[0049] The priority of the corresponding timestamp alignment method is evaluated based on the security accuracy and alignment duration, and then the information modality with the highest priority is marked as the time reference target.

[0050] Furthermore, the priority of the corresponding timestamp alignment method can be evaluated based on the security accuracy and alignment duration, and the evaluation can be based on the existing priority evaluation method.

[0051] Furthermore, the priority of corresponding timestamp alignment methods is evaluated based on security accuracy and alignment duration, including: Set the priority formula, which is: (2); Priority is determined according to the order of priority value from high to low.

[0052] Furthermore, the timestamp precision of the corresponding information modality is calculated based on the source data, including: Based on the statistical analysis of the material data, the timestamp accuracy of the information modality under the modal condition is marked as conditional accuracy. Mathematical statistical methods such as mean, mode, and variance can be used to statistically analyze the multiple timestamp error data corresponding to the material data. The proportion of the corresponding modal condition in the whole is determined based on historical data and marked as the condition share; Mark the modal conditions as , , The number of modal conditions; condition fraction and condition precision are respectively denoted as and ; The timestamp accuracy of the corresponding information modality is calculated using the accuracy calculation formula, which is: (3); Furthermore, other formulas can also be used to calculate accuracy, such as: (4); The mean, mode, etc. can also be used as formulas for calculating accuracy.

[0053] Furthermore, because timestamps can be inaccurate due to various reasons such as hardware clock drift, network transmission jitter, asynchronous clocks of multiple devices, and software processing delays, in order to improve timestamp alignment accuracy, the timestamps of positioning data and visual data are calibrated and adjusted accordingly based on the time reference target before timestamp alignment.

[0054] Furthermore, the timestamps of the positioning and visual data can be calibrated and adjusted accordingly based on the time reference target, which can be done based on existing calibration and adjustment methods.

[0055] For example: Based on relevant historical data, determine in real-time the time-influencing factors that affect the timestamps of positioning and visual data; collect data in real-time based on these time-influencing factors to obtain the timestamp errors under these factors and the corresponding timestamp alignment errors between them; organize this data to form training data; build an intelligent model based on the training data; train the model using the training data; and use the successfully trained intelligent model to calibrate and adjust the timestamps; utilize LSTM (Long Short-Term Memory) networks to learn the timestamp error patterns between devices, predict and compensate for dynamic errors; suitable for nonlinear error scenarios (such as wireless sensor networks).

[0056] Furthermore, timestamp calibration and adjustment are performed on multiple information modalities, including not only timestamp calibration and adjustment of positioning data and visual data.

[0057] The data processing module processes location data and video data to form a spatiotemporal correlation map of personnel, equipment, and environment. Specifically, it constructs this map based on timestamped location data and visual data. This map is then combined with a BIM (Building Information Modeling) model to overlay virtual and real-world scenes, resulting in a utility tunnel monitoring model. This model is generated using existing data processing methods and modeling techniques, such as a spatiotemporal data fusion model based on a graph neural network (GNN) to achieve collaborative analysis of location, visual, and environmental sensor data. Digital twin technology is introduced to construct a dynamic digital mirror of the utility tunnel, supporting simulation and risk prediction. The integrated utility tunnel is then monitored based on the monitoring model.

[0058] The visual optimization module is used to perform real-time optimization analysis of visual data processing. Existing visual data processing generally utilizes corresponding visual detection models to create intelligent models, but these models are prone to false detections due to changes in the utility tunnel, such as misidentifying equipment shadows as personnel, abandoned small tools, or other items or phenomena not observed during tunnel training. The module acquires visual monitoring result data in real time, including monitoring results, monitoring features, and result feedback. Monitoring results are analysis results based on the utility tunnel monitoring model to determine whether there are any anomalies. Monitoring features are those used to derive the monitoring result; features indicating normal monitoring are not extracted and are marked as "none." Result feedback is data provided by users based on actual conditions, such as monitoring anomalies requiring inspection or confirmation by staff. Feedback includes both "normal" and "abnormal" feedback, with "normal" feedback indicating correct monitoring results.

[0059] The visual monitoring results data are analyzed in real time to determine visual optimization points; corresponding model optimization data are set according to the visual optimization points. The model optimization data is collected according to the visual monitoring results data corresponding to the visual optimization points. Similar data are collected based on the monitoring characteristics of the visual monitoring results data as subsequent training materials and integrated into model optimization data; the visual monitoring model is optimized and adjusted according to the model optimization data.

[0060] Furthermore, real-time analysis of the visual monitoring results data is performed, including: Set optimization criteria. The optimization criteria is that the result feedback is abnormal, and the problem is determined to be caused by incomplete training of the visual detection model based on the monitoring features. For example, if the monitoring feature has not been trained, the normal situation will be analyzed as an abnormal result when the monitoring feature appears. The specific settings are determined by the platform. An optimization judgment model is established, and its expression is as follows: (1) In the formula, For input data, For visual monitoring results data, To optimize standards; This indicates that the corresponding visual monitoring results meet the optimization criteria; the output data is the optimization judgment value. The optimal judgment value is 1 or 0; By optimizing the judgment model, the visual monitoring result data and optimization standards are analyzed to obtain the optimized judgment value of the corresponding visual monitoring result data. Visual optimization points are determined based on the optimization judgment values. These visual optimization points indicate areas that require optimization and adjustment. They are determined based on visual monitoring data, which is equivalent to associating them with the corresponding visual monitoring data.

[0061] Furthermore, the optimized judgment model can also be built based on other existing intelligent algorithms, such as deep learning algorithms, and trained using a training set.

[0062] Furthermore, the vision optimization module also includes a transmission optimization unit, which is used to perform transmission optimization analysis on the vision data, determine the transmission optimization method of the vision data, process the vision data before transmission according to the transmission optimization method, and then transmit it to improve the transmission efficiency and synchronization of the vision data.

[0063] Furthermore, the transmission optimization analysis of visual data is performed, including: Determine visual monitoring requirements, mainly regarding the clarity and resolution requirements for different pipe corridors and equipment, and set them according to user needs; acquire visual data processing records in real time, that is, comprehensively process visual data, positioning data, etc., to achieve synchronous monitoring records, such as the processing records of visual data and positioning data corresponding to the pipe corridor monitoring model; identify transmission optimization points in real time based on visual data processing records, that is, optimization situations where data synchronization does not meet user requirements due to visual data transmission delays, etc. Historical visual data corresponding to the transmission optimization points are obtained and marked as optimized transmission materials. The optimized transmission materials are analyzed according to the visual monitoring requirements to determine the possible optimization methods. The possible optimization methods are then screened to determine the transmission optimization method corresponding to the transmission optimization point.

[0064] Furthermore, based on the visual data processing records, real-time identification of transmission optimization points includes: Set visual synchronization requirements according to user needs; perform real-time identification of visual data processing records according to visual synchronization requirements to identify visual data that does not meet visual synchronization requirements; identify the pipe gallery area and acquisition equipment corresponding to the visual data, and determine the visual optimization point based on the management area and acquisition equipment, that is, the visual data collected by the acquisition equipment corresponding to the pipe gallery area does not meet visual synchronization requirements. Specifically, visual data processing records can be identified and judged based on existing methods and visual synchronization requirements.

[0065] Furthermore, the optimized transmission materials are analyzed according to visual monitoring requirements. That is, under the time monitoring requirements, it is determined how the optimized transmission materials can be adjusted to meet the user's visual synchronization requirements, forming different alternative optimization methods. Optimization simulation analysis can be carried out based on intelligent models such as deep learning algorithms to estimate whether the adjusted and optimized transmission materials meet the visual synchronization requirements, and thus determine whether it is an alternative optimization method. In other words, intelligent analysis can be carried out using existing methods in the above manner.

[0066] Furthermore, the proposed optimization methods are screened using existing methods, such as using the display effect of visual data as a benchmark, or using transmission efficiency, data size, etc.; in particular, multiple screening factors can be used for screening.

[0067] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.

[0068] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

[0069] Example 3 In another preferred embodiment, based on the above embodiments 1 and 2, this embodiment provides a detailed description of the specific implementation of a positioning and visual data processing system for integrated utility tunnel operations.

[0070] I. Implementation Examples and System Overview This embodiment addresses the complex cable distribution and dynamic adjustment requirements within power plant cable corridors by designing and implementing a comprehensive positioning and visual data processing system for cable corridor operations. Through the collaborative work of a timestamp module, a data processing module, and a vision optimization module, this system achieves intelligent processing of positioning and visual data within the cable corridor, improving data synchronization and monitoring accuracy.

[0071] II. System Architecture and Module Functions The overall system architecture is as follows Figure 1 As shown, it mainly consists of three parts: timestamp module, data processing module, and visual optimization module.

[0072] 1. Timestamp module: 1) Function: Responsible for aligning the timestamps of positioning data and visual data, and determining the time reference target to improve the synchronization and accuracy of the data.

[0073] 2) Implementation details: (1) Information modality acquisition: Acquire the information modality corresponding to the integrated utility tunnel operation, including the positioning modality and the visual modality, while also considering other possible acquisition modalities (such as temperature, humidity, etc.). (2) Modal condition analysis: Analyze the impact of various modal conditions (such as acquisition equipment, environment, network, etc.) on timestamp accuracy; (3) Timestamp accuracy calculation: The timestamp accuracy of each information modality is calculated based on the source data (historical data or simulated data) using the following formula: (3) In the formula, For timestamp precision, As a conditional share, For conditional precision; (4) Determination of Time Baseline Target: Select the information modality with the highest timestamp accuracy as the time baseline target, or determine it after evaluating its priority through simulated alignment; the priority evaluation formula is: (2) In the formula, Priority value; , These are two different proportionality coefficients, with a range of values ​​of [missing value]. , ; It is a natural constant; For safety and accuracy; For alignment duration.

[0074] 2. Data Processing Module: 1) Function: Process location data and video data to construct a spatiotemporal correlation map of personnel, equipment and environment, generate a utility tunnel monitoring model, and realize real-time monitoring and simulation.

[0075] 2) Implementation details: (1) Spatiotemporal correlation map construction: The spatiotemporal correlation map is constructed by fusing the time-stamp-aligned location data and visual data using graph neural networks (GNNs); (2) BIM model overlay: Combine BIM model to realize the overlay of virtual and real scenes to generate a utility tunnel monitoring model; (3) Real-time monitoring and simulation: Real-time monitoring of the integrated utility tunnel is carried out based on the utility tunnel monitoring model, supporting simulation and risk prediction.

[0076] 3. Visual Optimization Module: 1) Function: Perform real-time optimization analysis on visual data, dynamically adjust the visual monitoring model, improve detection accuracy, and optimize the transmission of visual data.

[0077] 2) Implementation details: (1) Real-time optimization analysis: Set optimization criteria, establish an optimization judgment model, perform real-time analysis of visual monitoring results data, determine visual optimization points, and the expression of the optimization judgment model is: (1) In the formula, For input data, For visual monitoring results data, To optimize standards; This indicates that the corresponding visual monitoring results meet the optimization criteria; the output data is the optimization judgment value. The optimal judgment value is 1 or 0; (2) Model optimization and adjustment: The model optimization data is set according to the visual optimization points, and the visual monitoring model is optimized and adjusted; (3) Transmission optimization processing: The transmission optimization unit performs transmission optimization analysis on the visual data, determines the transmission optimization method, and improves the data transmission efficiency and synchronization.

[0078] III. Specific Implementation Steps 1. Hardware deployment: 1) Arrange UWB (Ultra Wide Band) base stations and cameras reasonably within the cable corridor to ensure coverage without blind spots; 2) Configure temperature and humidity sensors and network latency monitors to collect environmental parameters and network status in real time.

[0079] 2. Data Acquisition and Preprocessing: 1) Collect location data, visual data, environmental parameters, and network latency data; 2) Preprocess the collected data, including noise reduction, filtering, and outlier removal.

[0080] 3. Timestamp alignment: 1) Use the timestamp module to align the positioning data and visual data with timestamps to determine the time reference target; 2) The timestamps of the positioning and visual data are calibrated and adjusted according to the time reference target.

[0081] 4. Data Processing and Model Building: 1) Use the data processing module to process the timestamp-aligned positioning data and visual data to construct a spatiotemporal correlation map; 2) Combine the BIM model to generate a utility tunnel monitoring model to achieve real-time monitoring and simulation.

[0082] 5. Visual and transmission optimization: 1) Utilize the visual optimization module to perform real-time optimization analysis of visual data and dynamically adjust the visual monitoring model; 2) Optimize the transmission of visual data to improve data transmission efficiency and synchronization.

[0083] Example 4 In another preferred embodiment, based on embodiments 1, 2, and 3 above, this embodiment proposes a comprehensive positioning and visual data processing method for cable tunnel operations, addressing the complex operating environment and dynamically changing cable layout within power station cable tunnels. This method achieves efficient synchronization of positioning and visual data, accurate construction of spatiotemporal correlation maps, and dynamic optimization of the visual monitoring model through three main steps: timestamp alignment, data processing, and visual optimization. This improves the safety and management efficiency of cable tunnel operations.

[0084] The specific implementation steps are as follows: Step 1: Timestamp Alignment Module Processing Step 1.1: Collect modal condition data and calculate timestamp accuracy 1) Modal condition data acquisition: Using the modal condition analysis unit, modal condition data of positioning modality (such as UWB positioning system) and visual modality (such as high-definition camera) are collected, including equipment error (such as the accuracy of positioning tag), network latency (measured by ping command), and environmental temperature and humidity (using temperature and humidity sensors).

[0085] 2) Timestamp accuracy calculation: The accuracy calculation unit calculates the timestamp accuracy for each mode using formulas based on the collected modal condition data. (3) In the formula, For timestamp precision, This refers to the conditional share (i.e., the proportion of this modal condition among all conditions). This is the conditional precision (i.e., the average timestamp error under this modal condition).

[0086] Step 1.2: Select the mode with the highest accuracy as the benchmark and dynamically calibrate the data of other modes. 1) Based on the calculated timestamp accuracy, select the mode with the highest accuracy as the time reference target; 2) The dynamic calibration unit uses the selected time reference target as a benchmark and predicts and compensates for the dynamic errors of other modes through an LSTM network. Specific steps include: (1) Collect historical timestamp alignment data and train the LSTM network model.

[0087] (2) In actual operation, the trained LSTM network model is used to predict the timestamp error of each mode at the current time and to compensate for it.

[0088] Step 1.3: Compensate for clock drift error using an LSTM network 1) Utilize LSTM networks to capture the nonlinear characteristics of clock drift, and dynamically adjust the internal state through its gating mechanism (input gate, forget gate, output gate) to predict and compensate for clock drift errors; 2) Regularly update the LSTM network model parameters to adapt to changes in clock drift characteristics caused by factors such as equipment aging and environmental changes.

[0089] Step 2: Data processing module processing Step 2.1: Construct a spatiotemporal correlation map 1) The spatiotemporal correlation map construction unit fuses time-stamp-aligned positioning data, visual data, and environmental sensor data through a graph neural network (GNN); 2) GNN uses an attention mechanism to assign dynamic weights to data from different modalities. The weight calculation formula is as follows: ……………………(5; In the formula, , They are respectively , The eigenvectors of a mode, It is a trainable parameter matrix; This is the attention weight vector; 3) Based on the calculated dynamic weights, construct a spatiotemporal relationship map of personnel, equipment, and environment.

[0090] Step 2.2: Generate a digital twin 1) The digital twin generation unit combines the BIM model with the spatiotemporal correlation map to generate a dynamically updated utility tunnel monitoring model; 2) The risk prediction subunit uses the Monte Carlo simulation method, combining real-time equipment operating parameters (such as current and temperature) with historical fault data to generate a fault probability distribution map and mark high-risk areas; 3) Dynamically update sub-units automatically scan and update the digital twin after the layout of the utility tunnel is adjusted to ensure the accuracy and real-time performance of the monitoring model.

[0091] Step 3: Visual optimization module processing Step 3.1: Analyze the visual monitoring results in real time, identify false detection types, and extract similar feature data. 1) The false detection judgment unit analyzes the visual monitoring result data in real time. When the monitoring result is abnormal but the feedback is normal, it extracts similar feature data (such as shadow shape and illumination angle). 2) Identify false detection types by optimizing the judgment model. The expression for the optimized judgment model is: (1) In the formula, For input data, For visual monitoring results data, To optimize the criteria (e.g., the result feedback is abnormal and caused by incomplete model training); This indicates that the corresponding visual monitoring results meet the optimization criteria; the output data is the optimization judgment value. The optimal judgment value is 1 or 0; Step 3.2: Dynamically train the visual detection model and complete parameter updates. 1) The model optimization unit dynamically trains the visual detection model based on the extracted similar feature data; 2) The transfer learning method is adopted, using the pre-trained model parameters of similar utility tunnel scenarios as initial values, and then fine-tuning them with the current scenario data to shorten the model convergence time.

[0092] Step 3.3: Compress visual data according to the needs of the utility tunnel area to reduce transmission bandwidth usage. 1) The transmission optimization unit dynamically adjusts the visual data resolution and encoding format according to the needs of the utility tunnel area; 2) Compress non-critical areas to reduce transmission bandwidth usage; 3) Maintain lossless image quality transmission for critical areas (such as cable joints and equipment compartments) and ensure that data latency is not lower than the set parameters; 4) The bandwidth adaptive subunit monitors the network bandwidth in real time. When the bandwidth drops to the set value, it automatically reduces the resolution of non-critical areas and switches the encoding format. 5) The priority scheduling subunit marks the visual data of the critical area with the highest priority to ensure priority transmission during bandwidth contention.

[0093] Example 5 In another preferred embodiment, based on embodiments 1, 2, 3, and 4 above, this embodiment provides a positioning and visual data processing system for integrated utility tunnel operations. Through multimodal data time synchronization, dynamic digital twin modeling, and visual detection closed-loop optimization technology, it solves the problems of data timing misalignment, model update lag, and high visual false detection rate in existing utility tunnel monitoring, achieving high-precision, real-time, and low-false-alarm monitoring of utility tunnel operations. Its system architecture is as follows: 1. Timestamp Alignment Module This module achieves data time-series alignment through multimodal condition analysis and dynamic error compensation: 1) Modal Condition Analysis Unit: Collects modal condition data from positioning devices (UWB / Bluetooth), vision devices (cameras), and environmental sensors (thermometers, barometers), including parameters such as device clock drift rate, network latency fluctuation value, and ambient temperature and humidity, and constructs a modal condition database; 2) Precision Calculation Unit: Based on the conditional database, the weighted least squares method is used to calculate the timestamp precision of each modality. The weighting coefficients are inversely proportional to the device error and network stability. 3) Dynamic calibration unit: Based on the highest accuracy mode (such as high-precision UWB positioning), a dynamic error compensation model is established through LSTM network: the input is the historical timestamp deviation sequence, the output is the compensation coefficient, and the gradient descent method is used to iteratively optimize the model parameters so that the time alignment error of the multimodal data after compensation is ≤50ms, which is 80% higher than the traditional static calibration method (error ≥200ms).

[0094] 2. Data Processing Module This module builds a dynamic digital twin model based on time-aligned data: 1) Spatiotemporal correlation map construction unit: The graph neural network (GNN) is used to fuse the positioning trajectory, visual frame and environmental parameters to generate a spatiotemporal correlation map of personnel-equipment-environment; the GNN dynamically allocates weights through the attention mechanism to highlight key correlations (such as the risk weight of personnel approaching high-temperature equipment is increased by 30%). 2) Digital twin generation unit: Combines the BIM model with the spatiotemporal correlation map to generate a dynamically updated utility tunnel monitoring model; (1) Risk prediction subunit: The Monte Carlo simulation method is adopted. The real-time operating parameters (current, temperature) and historical fault data of the equipment are input to generate a heat map of the fault probability distribution in the next 48 hours and mark high-risk areas (such as triggering an early warning when the overload probability of the cable joint is ≥80%). (2) Dynamically update sub-units: When the layout of the utility tunnel is adjusted, the three-dimensional point cloud data of the changed area is obtained by laser scanner and registered with the BIM model by ICP (Iterative Closest Point) to identify the newly added / deleted cables or equipment. The model is automatically updated within 2 hours, which is 12 times more efficient than manual updates (cycle ≥ 24 hours).

[0095] 3. Visual Optimization Module This module improves the reliability of visual inspection through false detection closed-loop correction and adaptive transmission optimization. 1) False detection judgment unit: When the vision module detects an anomaly (such as excessive cable surface temperature) but the feedback system confirms that it is normal, the event is automatically marked as a potential false detection. Data of similar scenes (light intensity, shadow shape) are extracted as negative samples and combined with historical correct detection data to train and optimize the judgment model. 2) Model optimization unit: Using transfer learning, with a pre-trained ResNet-50 network as the backbone, the parameters of the first 3 convolutional layers are frozen, and only the fully connected layers and attention modules are fine-tuned. The model is updated within 48 hours, reducing the false positive rate from 15% to 3% and reducing the workload of manual review by 80%. 3) Transmission optimization unit: (1) Bandwidth adaptive subunit: Real-time monitoring of network bandwidth. When the bandwidth drops to a threshold (e.g., 70% of the original bandwidth), it automatically compresses the resolution of visual data in non-critical areas (e.g., ventilation ducts) from 4K to 720P and switches to H.264 encoding to reduce bandwidth usage by 50%. (2) Priority scheduling subunit: Data in key areas (such as transformer compartments and cable joints) are marked with the highest priority (priority=3), and TCP acceleration protocol is used to ensure priority transmission with a delay of ≤200ms to meet the real-time monitoring requirements.

[0096] Based on the above system, the method and steps for positioning and visual data processing of integrated utility tunnel operations are as follows: Step 1: Multimodal data timestamp alignment Step 1.1: Collect conditional data for the localization modality (UWB), visual modality (camera), and environmental modality (temperature and humidity) to construct a conditional database; Step 1.2: Calculate the timestamp accuracy of each modality and select the modality with the highest accuracy (such as UWB) as the benchmark; Step 1.3: Dynamically compensate for clock drift errors in other modes using an LSTM network to ensure that the time alignment error is ≤50ms.

[0097] Step 2: Construction of Dynamic Digital Twin Step 2.1: Use GNN to fuse temporally aligned data to generate a spatiotemporal correlation map of personnel-equipment-environment, and use an attention mechanism to dynamically allocate weights; Step 2.2: Combine the BIM model with the spatiotemporal map to generate a utility tunnel monitoring model; Step 2.3: Predict failure risks through Monte Carlo simulation and generate a probability distribution map; Step 2.4: When the layout is adjusted, the model is automatically updated by laser scanning + ICP registration, which is completed within 2 hours.

[0098] Step 3: Visual Inspection Closed-Loop Optimization and Transmission Adaptation Step 3.1: Analyze the visual monitoring results in real time, identify false detection types, and extract similar feature data (lighting, shadows). Step 3.2: The visual detection model was dynamically trained using transfer learning, and the parameters were updated within 48 hours, reducing the false alarm rate to 3%. Step 3.3: Compress visual data according to regional requirements: lossless transmission of critical areas (4K+H.265), and compressed transmission of non-critical areas (720P+H.264), reducing bandwidth usage by 50%.

[0099] In a preferred embodiment, the false detection judgment unit identifies the false detection type by optimizing the judgment model, and triggers model optimization when the set conditions are met. The above settings can automatically adjust the false detection judgment logic, reduce the false alarm rate, and at the same time, the unit can also record false detection cases to provide data support for subsequent model training and continuously improve detection accuracy and efficiency.

[0100] In the preferred embodiment, the model optimization unit adopts the transfer learning method, using the pre-trained model parameters of similar utility tunnel scenarios as initial values, and fine-tuning them in combination with the current scenario data to shorten the model convergence time. The above settings effectively improve the model's adaptability to new scenarios. At the same time, by integrating the prediction results of multiple sub-models through ensemble learning technology, the accuracy and robustness of the prediction are further enhanced, ensuring the efficiency and reliability of the solution.

[0101] In the preferred embodiment, the graph neural network (GNN) employs an attention mechanism in the step of constructing the spatiotemporal correlation map to assign dynamic weights to different modalities of data. This setting enables the model to automatically capture and emphasize key information while reducing the impact of noisy data, thereby effectively improving the accuracy and efficiency of constructing the spatiotemporal correlation map and laying a solid foundation for subsequent spatiotemporal data analysis.

[0102] In summary, the positioning and visual data processing system and method for integrated utility tunnel operations proposed in this invention have brought about many innovative breakthroughs in this field, effectively solved a series of existing core problems, and greatly improved the real-time performance, accuracy, and economy of utility tunnel operations.

[0103] Regarding time synchronization, existing technologies have not considered calculating timestamp accuracy by analyzing the modal conditions (such as device errors, network latency, and ambient temperature and humidity) of the positioning and visual modalities, and have used the modality with the highest accuracy as a benchmark for dynamic calibration. This invention, however, uses an LSTM network to predict and compensate for dynamic errors, successfully achieving high-precision time synchronization and effectively solving the problem of time asynchrony in multi-source data.

[0104] Regarding the construction of spatiotemporal correlation maps, existing technologies do not utilize graph neural networks (GNNs) combined with attention mechanisms to assign dynamic weights to different modalities of data to construct spatiotemporal correlation maps of personnel, equipment, and environment. This invention, however, significantly improves the accuracy and real-time performance of data fusion through dynamic weight allocation, enabling real-time responses to situations such as cable additions / reductions or layout adjustments. In terms of dynamic updates to digital twins, existing technologies lack methods for generating dynamically updated utility tunnel monitoring models that combine BIM models with spatiotemporal correlation maps, and support automatic updates for risk prediction and layout adjustments. This invention utilizes Monte Carlo simulation to provide early warnings of potential faults and automatically scans and updates the digital twin after layout adjustments, significantly improving the intelligence level of monitoring.

[0105] In the field of visual data transmission optimization, existing technologies do not propose a method for dynamically adjusting the resolution and encoding format of visual data according to the needs of the utility tunnel area. This invention, through bandwidth adaptation and priority scheduling, compresses non-critical areas while maintaining lossless image quality transmission in critical areas, effectively reducing transmission bandwidth consumption and ensuring that data latency is not lower than the set parameters.

[0106] From an overall methodological perspective, this invention proposes a multimodal data fusion method combining timestamp alignment, spatiotemporal correlation graph construction, and digital twin generation. This method not only overcomes the difficulties in data synchronization and fusion in existing technologies but also significantly improves the safety and efficiency of integrated utility tunnel operations through dynamic updates and risk prediction. Simultaneously, this invention achieves dynamic training and parameter updates of the visual detection model through a false alarm judgment unit and a model optimization unit. When the monitoring result is abnormal but the feedback is normal, the system can automatically extract similar feature data, optimize the judgment model, reduce the false alarm rate, and significantly improve the accuracy and reliability of visual monitoring. Moreover, the method proposed in this invention for dynamically adjusting the resolution and encoding format of visual data according to the needs of the utility tunnel area compresses data in non-critical areas while ensuring lossless transmission of image quality in critical areas, balancing transmission efficiency and image quality, and has significant application value in integrated utility tunnel operations.

[0107] Furthermore, this invention provides a systematic integrated utility tunnel operation positioning and visual data processing solution, comprehensively covering multiple aspects such as data synchronization, fusion, monitoring model optimization, and data transmission optimization. This solution adopts a modular design, enabling each functional module to operate independently and work collaboratively, greatly improving the system's flexibility and scalability.

Claims

1. A positioning and visual data processing system for integrated utility tunnel operations, characterized in that, include: The timestamp alignment module is used to acquire positioning and visual modal information in the integrated utility tunnel operation, analyze modal conditions and extract material data, calculate the timestamp accuracy of each modality, and use the modality with the highest timestamp accuracy as the time reference target to dynamically calibrate the positioning and visual data. The data processing module is used to construct a spatiotemporal correlation map of personnel, equipment and environment based on the timestamp-aligned data, and generate a utility tunnel monitoring model in combination with the BIM model; The visual optimization module is used to analyze visual monitoring results data in real time, dynamically optimize the visual detection model, and the transmission optimization unit compresses and encodes the visual data before transmission.

2. The positioning and visual data processing system for integrated utility tunnel operations according to claim 1, characterized in that, The timestamp alignment module includes: The modal condition analysis unit is used to collect modal condition data of positioning mode and visual mode, including at least one of equipment error, network latency, and ambient temperature and humidity; The precision calculation unit is used to calculate the timestamp precision of each mode based on the modal condition data. The dynamic calibration unit is used to predict and compensate for the dynamic errors of other modes using an LSTM network, based on the mode with the highest timestamp accuracy.

3. The positioning and visual data processing system for integrated utility tunnel operations according to claim 2, characterized in that, The data processing module includes: The spatiotemporal correlation map construction unit is used to generate spatiotemporal correlation relationships between people, equipment, and environment by fusing positioning data, visual data, and environmental sensor data through graph neural networks (GNNs). The digital twin generation unit is used to combine the BIM model with the spatiotemporal correlation map to generate a dynamically updated utility tunnel monitoring model, supporting real-time response to cable additions, reductions, or layout adjustments.

4. The positioning and visual data processing system for integrated utility tunnel operations according to claim 3, characterized in that, The digital twin generation unit further includes: The risk prediction subunit is used to simulate cable overload and equipment aging scenarios based on a historical fault database, and to provide early warning of potential faults. The dynamic update sub-unit is used to automatically scan and update the digital twin after the layout of the utility tunnel is adjusted.

5. The positioning and visual data processing system for integrated utility tunnel operations according to claim 4, characterized in that: The risk prediction subunit uses Monte Carlo simulation to generate a fault probability distribution map by combining real-time equipment operating parameters with historical fault data and marking high-risk areas.

6. The positioning and visual data processing system for integrated utility tunnel operations according to claim 5, characterized in that, The visual optimization module includes: The false detection judgment unit is used to extract similar feature data and train the visual detection model when the monitoring result is abnormal but the feedback is normal. The model optimization unit is used to update model parameters and reduce the false alarm rate.

7. The positioning and visual data processing system for integrated utility tunnel operations according to claim 6, characterized in that: The false detection judgment unit identifies the false detection type by optimizing the judgment model, and triggers model optimization when the set conditions are met.

8. The positioning and visual data processing system for integrated utility tunnel operations according to claim 7, characterized in that: The model optimization unit adopts the transfer learning method, which uses the pre-trained model parameters of similar utility tunnel scenarios as initial values ​​and fine-tunes them in combination with the current scenario data to shorten the model convergence time.

9. The positioning and visual data processing system for integrated utility tunnel operations according to claim 8, characterized in that, The transmission optimization unit in the vision optimization module dynamically adjusts the visual data resolution and encoding format according to the needs of the utility tunnel area, including: Compress non-critical areas to reduce transmission bandwidth usage; Maintain lossless image quality transmission in critical areas and ensure that data latency is not lower than the set parameters.

10. The positioning and visual data processing system for integrated utility tunnel operations according to claim 9, characterized in that, Its features are, The transmission optimization unit further includes: The bandwidth adaptive subunit is used to monitor network bandwidth in real time. When the bandwidth drops to a set value, it automatically reduces the resolution of non-critical areas and switches the encoding. The priority scheduling subunit is used to visually mark critical areas, including cable joints and equipment compartments, with the highest priority to ensure priority transmission during bandwidth contention.

11. A method for positioning and visual data processing in integrated utility tunnel operations, characterized in that, The method for data processing based on the positioning and visual data processing system for integrated utility tunnel operations as described in claim 10 includes the following steps: Step 1: Acquire and calibrate positioning modality and visual modality data through the timestamp alignment module; Step 2: Construct a spatiotemporal correlation map and generate a digital twin using the data processing module; Step 3: Dynamically correct false detections and optimize visual data transmission through the vision optimization module.

12. The positioning and visual data processing method for integrated utility tunnel operations according to claim 11, characterized in that, The timestamp alignment step in step 1 includes: Step 1.1: Collect modal condition data and calculate timestamp accuracy; Step 1.2: Select the mode with the highest accuracy as the benchmark and dynamically calibrate the data of other modes; Step 1.3: Compensate for clock drift error using an LSTM network.

13. The positioning and visual data processing method for integrated utility tunnel operations according to claim 12, characterized in that: In the step of constructing the spatiotemporal correlation map, the graph neural network (GNN) uses an attention mechanism to assign dynamic weights to data of different modalities.

14. The positioning and visual data processing method for integrated utility tunnel operations according to claim 13, characterized in that, The visual optimization steps in step 3 include: Step 3.1: Analyze the visual monitoring results in real time, identify false detection types, and extract similar feature data; Step 3.2: Dynamically train the visual detection model and update the parameters; Step 3.3: Compress visual data according to the needs of the utility tunnel area to reduce transmission bandwidth usage.

Images