Traffic engineering multi-source monitoring data fusion intelligent management and control system

Abstract

The invention discloses a traffic engineering multi-source monitoring data fusion intelligent management and control system, and the system comprises a multi-source data dynamic access module which is used for the access of multi-source heterogeneous data; a data processing module; the dynamic relation learning module is used for mining dynamic space-time association and potential laws in the multi-source data and establishing a time-varying mapping model among the data; the intelligent fusion decision module is used for generating a traffic control decision based on the processed multi-source data and an output result of the time-varying mapping model; an adaptive adjustment module; a data service module; through cooperation of the multi-source data dynamic access module, the data processing module, the dynamic relation learning module, the intelligent fusion decision-making module, the adaptive adjustment module and the data service module, full-process automation from multi-source data access to intelligent management and control decision-making can be realized, and self-adaption to changes of a complex traffic environment can be realized. And the fine management level of traffic engineering is improved.

Description

Technical Field

[0001] This invention belongs to the field of traffic engineering technology, specifically relating to an intelligent management and control system for multi-source monitoring data fusion in traffic engineering. Background Technology

[0002] Multi-source monitoring in traffic engineering integrates heterogeneous data from different sensors, systems, or platforms, such as GPS, cameras, and vehicle-mounted equipment, to achieve comprehensive perception and prediction of traffic conditions. Since the data formats and standards from different sources are not uniform, interconnection and interoperability are achieved through intelligent management and control to fuse and process the multi-source monitoring data in order to obtain traffic conditions more accurately.

[0003] Currently, the common approach to managing multi-source data fusion is to use a data fusion platform for centralized management. This platform integrates cross-source data through technologies such as relational projection and intelligent data orchestration. For example, Chinese patent application number 202111209444.2 discloses a multi-source data fusion method and system. This method includes: acquiring historical multi-source data; classifying the multi-source data according to its meaning; setting a data item for each category of data in the classification results; constructing association relationships between sub-data in the i-th category and mapping the corresponding association relationships to the corresponding i-th data item; constructing a data fusion structure based on all data items; acquiring input multi-source data; storing the input multi-source data into the corresponding data item category according to the data fusion structure; and acquiring current input multi-source data and storing it into the corresponding data item category according to the data fusion structure. This enables real-time data classification and fusion processing of the currently input business data, and orderly management of real-time multi-business data.

[0004] However, constructing a fusion structure by pre-setting data items and relationships makes it difficult to dynamically adapt to changes in new sensor types or data formats. For example, when introducing new equipment such as millimeter-wave radar, the data item classification logic needs to be manually adjusted, resulting in insufficient system scalability and affecting the quality of data fusion. Therefore, we need to propose an intelligent management and control system for multi-source monitoring data fusion in traffic engineering to solve the above-mentioned problems, enabling it to automatically discover and adaptively adjust relationships through entity linking technology, thereby improving the quality of data fusion. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent management and control system for multi-source monitoring data fusion in traffic engineering. By using entity linking technology to automatically discover and adaptively adjust relationships, the system improves the quality of data fusion and solves the problems mentioned in the background.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A traffic engineering multi-source monitoring data fusion intelligent management and control system includes:

[0008] The multi-source data dynamic access module is used to access multi-source heterogeneous data and generate raw data with unified entity identifiers.

[0009] The data processing module is used to clean, transform, and integrate the incoming raw data, eliminate noise, redundancy, and contradictions, and generate standardized, reusable multi-source data.

[0010] The dynamic relationship learning module is used to mine the dynamic spatiotemporal correlations and potential patterns in multi-source data and to establish a time-varying mapping model between data.

[0011] The intelligent fusion decision-making module generates traffic control decisions based on the processed multi-source data and the output of the time-varying mapping model.

[0012] The adaptive adjustment module monitors the operation status of the multi-source data dynamic access module, data processing module, dynamic relationship learning module, and intelligent fusion decision-making module in real time, and dynamically adjusts parameters according to the external environment and system performance to ensure system stability.

[0013] The data service module stores the processed data and decision results and provides standardized service interfaces to facilitate output in diverse formats.

[0014] The adaptive adjustment module is electrically connected to the multi-source data dynamic access module, the data processing module, the dynamic relationship learning module, and the intelligent fusion decision module, respectively.

[0015] Preferably, the multi-source data dynamic access module includes an interface adaptive adaptation unit, a dynamic load balancing unit, and an access status monitoring unit connected in sequence. The interface adaptive adaptation unit receives data through a standardized access interface and then uses a protocol conversion component to be compatible with the data formats of different devices.

[0016] The dynamic load balancing unit automatically expands access nodes through the Kubernetes container orchestration platform based on real-time data volume, while using a flow control algorithm to limit the access frequency of abnormal data sources.

[0017] The access status monitoring unit has a built-in heartbeat detection mechanism that provides real-time alarms for offline devices and automatically switches to a backup data source.

[0018] Preferably, the specific conversion process of the protocol conversion component is as follows:

[0019] A1. Receive raw data from different devices and automatically identify the source protocol type through a preset protocol feature library;

[0020] A2. Parse the data frame according to the format specification of the source protocol;

[0021] A3. Convert the extracted heterogeneous data into a system-compatible format and encapsulate the RTSP stream into a binary message body of the MQTT protocol;

[0022] A4. Encapsulate the data according to the specifications of the target protocol, which includes RESTful API and MQTT;

[0023] A5. Verify the integrity of the converted data, send the standardized data to the data processing module, and record the conversion log.

[0024] Preferably, the heartbeat detection mechanism is used to monitor whether the device is online in real time. The monitoring process of the heartbeat detection mechanism includes heartbeat packet definition, sending and receiving rules, offline judgment and alarm, and backup data source switching. Among them, the content of the heartbeat packet definition mechanism is the status message sent by the device periodically. The status message includes the device ID, current status, timestamp, and checksum. The sending and receiving rules include sending rules and receiving rules. The sending rules are that the device sends heartbeat packets to the system's receiving port at fixed intervals. The receiving rules are that the system records the last time of receiving the heartbeat packet and replies with an acknowledgment message to inform the device that it has been received.

[0025] The offline determination and alarm rules are determined as follows: if the system does not receive a heartbeat packet from a device within 3 cycles, it is determined to be offline, and the system automatically generates an alarm message and pushes it to the operation and maintenance platform; if the device is determined to be offline, the system automatically calls the backup strategy.

[0026] Preferably, the data processing module includes a data cleaning unit, a data conversion unit, and a data integration unit, which are connected in sequence. The data cleaning unit is used to perform noise filtering, anomaly detection, and missing value imputation on the original data.

[0027] The data conversion unit performs standardization and feature extraction processing on the cleaned data;

[0028] The data integration unit is used to perform multi-source alignment and remove redundant data from the converted data.

[0029] Preferably, the dynamic relationship learning module includes a spatiotemporal correlation modeling unit and a dynamic rule update unit. The spatiotemporal correlation modeling unit includes time dimension modeling and spatial dimension modeling. The time dimension modeling uses LSTM or Transformer time series model to analyze the time dependency of the data and combines a sliding window mechanism to update the time correlation rules in real time.

[0030] Spatial dimension modeling is based on graph neural networks to construct a road network topology graph, and the spatial influence weights of adjacent road segments and upstream and downstream road segments are quantified through graph attention mechanism.

[0031] The dynamic pattern update unit adopts an online framework to receive new data and update model parameters in real time, avoiding the failure of the relational model due to changes in traffic patterns.

[0032] Preferably, the intelligent fusion decision module includes a multi-source fusion unit and a decision generation unit. The multi-source fusion unit uses DS evidence theory to fuse sensor traffic flow, camera recognition data, and mobile phone signaling calculation data. Then, it constructs a multimodal fusion model based on the fused data and predicts and outputs traffic status through the multimodal fusion model.

[0033] When generating traffic control decisions, the decision generation unit presets basic strategies based on morning and evening peak hour rules, uses a reinforcement learning algorithm with the maximization of road network communication efficiency as the reward function, dynamically adjusts control measures, and pushes them in real time according to the control measures.

[0034] Preferably, the adaptive adjustment module includes a status monitoring unit, a decision analysis unit, a parameter execution unit, and a feedback verification unit connected in sequence. The status monitoring unit collects the operating indicators of each module in real time to provide data basis for adjustment.

[0035] The decision analysis unit determines whether adjustments are needed based on monitoring data and calculates the optimal adjustment strategy.

[0036] The parameter execution unit converts the adjustment strategy output by the decision analysis unit into specific instructions, sends them to the target module, and executes them.

[0037] The feedback verification unit monitors the adjusted module's operating status, verifies whether the adjustment effect meets the standard, and forms a closed-loop control.

[0038] Preferably, the decision analysis unit uses a combination of rule engine and adaptive algorithm for analysis. First, a threshold is preset, and then a reinforcement learning algorithm is used with failure rate and processing efficiency as reward functions to learn and dynamically adjust parameters.

[0039] Preferably, the data service module includes a data storage unit, an interface service unit, a data formatting unit, and a permission management unit. The input end of the data storage unit is connected to the data processing module and the intelligent fusion decision-making module, respectively. The output end of the data storage unit is connected to the interface service unit, and the interface service unit is connected to the data formatting unit and the permission management unit, respectively.

[0040] The intelligent management and control system for multi-source monitoring data fusion in traffic engineering proposed in this invention has the following advantages compared with existing technologies:

[0041] 1. When this invention accesses heterogeneous data through a multi-source data dynamic access module, entity linking technology first performs preliminary matching of cross-source entities to generate unified entity identifiers. During cleaning and integration, the data processing module verifies redundancy and contradictions based on the entity linking results, removes noise through entity attribute consistency rules, and ensures data standardization. When the dynamic relationship learning module mines spatiotemporal correlations, entity linking technology helps identify implicit correlations between cross-modal entities, and optimizes the time-varying mapping model by combining GNN and time-series models, making the association rules more accurate. The intelligent fusion decision module uses unified entity identifiers to integrate multi-source data features and generates decisions through weighted fusion, avoiding decision bias caused by entity ambiguity. The adaptive adjustment module monitors the entity linking accuracy in real time. If a decline in association quality is detected, dynamic adjustments are made: the entity recognition algorithm is optimized for the data access module, the entity consistency verification rules are updated for the processing module, and the association weight parameters are adjusted for the relationship learning module, forming a closed-loop optimization. The data service module stores fused data with unified entity identifiers and outputs it through a standardized interface to ensure that downstream applications use consistent entity association information, ultimately achieving full-process entity association reliability from data access to decision application and improving fusion quality.

[0042] 2. This invention, through the coordinated operation of a multi-source data dynamic access module, a data processing module, a dynamic relationship learning module, an intelligent fusion decision-making module, an adaptive adjustment module, and a data service module, can achieve full-process automation from multi-source data access to intelligent control and decision-making, and can adapt to changes in complex traffic environments, thereby improving the level of refined management of traffic engineering. Attached Figure Description

[0043] Figure 1 A system block diagram of an integrated intelligent control system according to an embodiment of the present invention is shown;

[0044] Figure 2 A block diagram of a multi-source data dynamic access module according to an embodiment of the present invention is shown;

[0045] Figure 3 A flowchart illustrating the conversion process of a protocol conversion component according to an embodiment of the present invention is shown.

[0046] Figure 4 A block diagram of an adaptive adjustment module according to an embodiment of the present invention is shown. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] This invention provides, for example Figure 1-4 The illustrated intelligent control system for multi-source monitoring data fusion in traffic engineering includes a multi-source data dynamic access module, a data processing module, a dynamic relationship learning module, an intelligent fusion decision-making module, an adaptive adjustment module, and a data service module. The output of the multi-source data dynamic access module is connected to the input of the data processing module, the output of the data processing module is connected to the input of the dynamic relationship learning module, the output of the dynamic relationship learning module is connected to the input of the intelligent fusion decision-making module, and the output of the intelligent fusion decision-making module is connected to the data service module. The adaptive adjustment module is electrically connected to the multi-source data dynamic access module, the data processing module, the dynamic relationship learning module, and the intelligent fusion decision-making module, respectively.

[0049] The multi-source data dynamic access module is used to access multi-source heterogeneous data and generate raw data with unified entity identifiers without the need for manual preset of interface parameters.

[0050] The multi-source data dynamic access module includes an interface adaptive adaptation unit, a dynamic load balancing unit, and an access status monitoring unit connected in sequence. The interface adaptive adaptation unit receives data through a standardized access interface and then uses a protocol conversion component to be compatible with the data formats of different devices. The data formats include structured data, unstructured data, and semi-structured data. Structured data includes traffic flow data acquired by vehicle sensors, unstructured data includes video streams captured by cameras, and semi-structured data includes XML logs. The protocol conversion component supports standardized protocols such as HTTP, CoAP, and Modbus. The standardized access interface includes RESTful API and MQTT protocol interface.

[0051] The specific conversion process of the protocol conversion component is as follows:

[0052] A1. Receive raw data from different devices and automatically identify the source protocol type through a preset protocol feature library; the raw data includes sensor values ​​of Modbus protocol, roadside device status of CoAP protocol, and camera captured images of HTTP protocol; the protocol feature library includes port number and frame header identifier: the frame header of Modbus is 0x01, and the UDP port of CoAP is 5683.

[0053] A2. Parse data frames according to the format specifications of the source protocol; specifically: extract key fields from structured data, including temperature, traffic flow, and speed, which are stored in the Modbus protocol via register addresses 0x0001-0x0003; parse frame tags from streaming protocols (such as RTSP) to extract video frame sequences; parse tag levels from semi-structured data to extract core information such as device ID and status, such as tag levels...<device_id> sensor_001< / device_id> , <status> online< / status> ).

[0054] A3. Convert the extracted heterogeneous data into a system-compatible format, encapsulate the RTSP stream into a binary message body of the MQTT protocol (transmitted through the payload field), and attach metadata such as frame timestamps and camera IDs;

[0055] For example, the Modbus hexadecimal value 0x0064 (corresponding to decimal 100) is converted into the JSON field {"device_id": "sensor_001", "traffic_flow": 100, "timestamp": "2025-07-16 08:30:00"};

[0056] A4. Encapsulate the data according to the specifications of the target protocol, which includes RESTful API and MQTT;

[0057] For example, a RESTful API might use an HTTP POST request to encapsulate JSON data into a request body for http: / / api.traffic-system.com / data.

[0058] MQTT: Publish data to the topic traffic / sensor / data, including QoS level (e.g., traffic flow data requires QoS=1 to ensure reliable transmission).

[0059] A5. Verify the integrity of the converted data, send the standardized data to the data processing module, and record the conversion log for fault tracing, such as the reason for the failure of a sensor data conversion; the data integrity check includes whether the fields of the structured data are missing and whether the frame sequence numbers of the video stream are continuous.

[0060] The dynamic load balancing unit automatically expands access nodes based on real-time data volume through the Kubernetes container orchestration platform to avoid data congestion, while using a flow control algorithm to limit the access frequency of abnormal data sources.

[0061] The core of the Kubernetes container orchestration platform is the scheduling algorithm. Its decision-making depends on multiple factors, but the core is the calculation of the node's comprehensive score. The higher the score, the higher the priority for scheduling.

[0062] The formula for the scheduling algorithm is:

[0063] ,

[0064] in, The overall score for each node. For the first The weights of each scheduling factor For the number of scheduling factors, For the first The scores for each scheduling factor range from 0 to 10, with 10 being the optimal score. , For the remaining resources of the node, Request resources for the Kubernetes container orchestration platform. If the remaining resources on the node are greater than or equal to the requested resources, the score is 10; otherwise, the score is 0.

[0065] For example, during the morning rush hour, if the camera processes requests using 2 CPU cores, and node A has 3 CPU cores remaining, then... =0, node B has 1 core remaining, then =0;

[0066] The flow control algorithm is set to the token bucket algorithm, and the algorithm formula is:

[0067] ,

[0068] in, Let be the number of tokens in the token bucket at time t. This is the maximum capacity of the token bucket. The initial number of tokens, For token generation rate, The elapsed time since the last token was generated;

[0069] For example, when a data source sends a data request, it needs to consume tokens. If the request requires m tokens (m is determined by the data size, such as 5 tokens for 1 frame of video), then:

[0070] like If yes, sending is allowed, and the token count is updated to... ;

[0071] like If so, rate limiting will be implemented, meaning requests will be rejected or cached and waited for.

[0072] The access status monitoring unit has a built-in heartbeat detection mechanism to provide real-time alarms for offline devices and automatically switch to a backup data source; the offline devices include faulty sensors, and the backup data source includes temporary data from sensors in adjacent road sections.

[0073] The heartbeat detection mechanism is used to monitor whether devices are online in real time. The monitoring process of the heartbeat detection mechanism includes heartbeat packet definition, sending and receiving rules, offline determination and alarm, and backup data source switching. Among them, the content of the heartbeat packet definition mechanism is the status message sent by the device periodically. The status message includes the device ID, current status, timestamp, and checksum. The sending and receiving rules include sending rules and receiving rules. The sending rule is that the device sends heartbeat packets to the system's receiving port at fixed intervals. The receiving rule is that the system records the last time it receives the heartbeat packet and replies with an acknowledgment message to inform the device that it has been received.

[0074] The offline determination and alarm rules are determined as follows: if the system does not receive a heartbeat packet from a device within 3 cycles, it is determined to be offline, and the system automatically generates an alarm message and pushes it to the operation and maintenance platform; if the device is determined to be offline, the system automatically calls the backup strategy.

[0075] For example, if the traffic flow sensor on road segment A goes offline, the system will switch to sensor data from the adjacent road segment B (temporary data from road segment A will be supplemented using a spatial interpolation algorithm). Once the original device is restored, the system will automatically switch back to the original data source.

[0076] The data processing module is used to clean, transform and integrate the incoming raw data, eliminate noise, redundancy and contradictions, and generate standardized and reusable multi-source data.

[0077] The data processing module includes a data cleaning unit, a data conversion unit, and a data integration unit, which are connected in sequence. The data cleaning unit is used to perform noise filtering, anomaly detection, and missing value imputation on the raw data.

[0078] When filtering noise, Kalman filters are used to process dynamic noise for sensor data; dehazing or deblurring algorithms are used for image data. Dehazing algorithms include dark channel prior algorithms, and deblurring algorithms include non-blind convolution algorithms.

[0079] During anomaly detection, outliers are identified based on isolated forest and DBSCAN clustering algorithms, and outliers are removed, such as jump values ​​caused by sensor failures and vehicle speeds that deviate significantly from the reasonable range.

[0080] When filling missing values, linear interpolation is used to fill short-term missing values, such as a sensor disconnection of 1 minute; long-term missing values ​​are filled by combining spatiotemporal correlation, such as a device failure of 1 hour, which is filled by using the same period data of the upstream road section and the historical average value.

[0081] The data conversion unit performs standardization and feature extraction processing on the cleaned data;

[0082] During standardized operations, the local time of different devices is converted to UTC+8 standard time, and GPS latitude and longitude and camera installation coordinates are converted to the national 2000 coordinate system. During feature extraction, key information is extracted from unstructured data, such as using YOLOv8 to identify vehicle type, number, and direction of travel from video; and using natural language processing (NLP) to extract controlled road sections and times from traffic control notices.

[0083] The data integration unit is used to perform multi-source alignment and remove redundant data from the converted data;

[0084] In multi-source alignment, different data in the same spatiotemporal dimension are associated, such as binding the traffic flow identified by the camera with the speed data of the sensors on the same road segment to form a traffic status package for that road segment; redundant data removal is to use a weighted average to remove duplicate data and retain data with high confidence, such as giving greater weight to sensor data with higher accuracy levels.

[0085] The dynamic relationship learning module is used to mine the dynamic spatiotemporal correlations and potential patterns in multi-source data and to establish a time-varying mapping model between data.

[0086] The dynamic relationship learning module includes a spatiotemporal correlation modeling unit and a dynamic rule update unit. The spatiotemporal correlation modeling unit includes time dimension modeling and spatial dimension modeling. The time dimension modeling uses LSTM or Transformer time series models to analyze the time dependence of data and combines a sliding window mechanism to update time correlation rules in real time, such as GRU capturing the impact of traffic flow in the previous 30 minutes on the current congestion probability.

[0087] Spatial dimension modeling is based on graph neural networks to construct a road network topology graph, and the spatial influence weights of adjacent road segments and upstream and downstream road segments are quantified through graph attention mechanism.

[0088] The dynamic pattern update unit adopts an online framework to receive new data and update model parameters in real time, thus avoiding the failure of the relational model due to changes in traffic patterns.

[0089] The intelligent fusion decision-making module generates traffic control decisions based on the processed multi-source data and the output of the time-varying mapping model, which facilitates the transformation from data to action.

[0090] The intelligent fusion decision module includes a multi-source fusion unit and a decision generation unit. The multi-source fusion unit uses DS evidence theory to fuse sensor traffic flow, camera recognition data, and mobile phone signaling calculation data to reduce the error of a single data source. Then, it constructs a multimodal fusion model based on the fused data and predicts and outputs the traffic status through the multimodal fusion model.

[0091] The formula for the multimodal fusion model is as follows:

[0092] ,

[0093] =1,

[0094] ,

[0095] ,

[0096] ,

[0097] in, Output values ​​for the model. For the input image data, For the input time series data, For the input spatial map data, A convolutional neural network used to extract visual features from images. Long Short-Term Memory (LSTM) networks are used to extract dynamic features from time-series data. For graph neural networks used to extract topological features from spatial data, , and These are the original features of image, temporal, and spatial modalities, respectively. , and These are the fusion weights for image, temporal, and spatial modalities, respectively.

[0098] When generating traffic control decisions, the decision generation unit presets basic strategies based on morning and evening peak rules, uses a reinforcement learning algorithm with the maximization of road network communication efficiency as the reward function, dynamically adjusts control measures, and pushes them in real time to ensure that the decision delay is less than 1 second, thus meeting the real-time requirements of traffic control.

[0099] For example, when dynamically adjusting control measures, such as when it is predicted that a certain road segment will be congested in 10 minutes, detour navigation can be used to find surrounding vehicles in advance;

[0100] The adaptive adjustment module monitors the operation status of the multi-source data dynamic access module, data processing module, dynamic relationship learning module and intelligent fusion decision module in real time, and dynamically adjusts parameters according to the external environment and system performance to ensure system stability.

[0101] The adaptive adjustment module includes a status monitoring unit, a decision analysis unit, a parameter execution unit, and a feedback verification unit. The status monitoring unit, decision analysis unit, parameter execution unit, and feedback verification unit are connected in sequence. The status monitoring unit collects the operating indicators of each module in real time to provide data basis for adjustment.

[0102] The status monitoring unit uses a distributed monitoring tool to collect real-time indicators through data collection points. The real-time indicators include resource indicators, performance indicators, and monitoring frequency. Resource indicators include CPU / memory utilization, network bandwidth usage, and storage IOPS; performance indicators include data access latency, data cleaning time, model inference accuracy, and error rate.

[0103] The decision analysis unit determines whether adjustments are needed based on monitoring data and calculates the optimal adjustment strategy.

[0104] The decision analysis unit uses a combination of rule engine and adaptive algorithm for analysis. First, a threshold is preset. For example, when the data access module delay is greater than 500ms, the number of access threads is increased. Then, a reinforcement learning algorithm is used, with failure rate and processing efficiency as reward functions, to learn to dynamically adjust parameters. The reinforcement learning algorithm is set to DQN.

[0105] The parameter execution unit converts the adjustment strategy output by the decision analysis unit into specific instructions, sends them to the target module, and executes them.

[0106] The parameter execution unit sends adjustment instructions to the target module through a standard API interface. Specifically, it adjusts the data source connection pool size and data cache queue length for the multi-source data dynamic access module; it adjusts the number of parallel computing nodes and the filter algorithm window size for the data processing module; it adjusts the GNN iteration count and learning rate for the dynamic relationship learning module; and it adjusts the fusion weights and output decision update frequency for the intelligent fusion decision module.

[0107] The feedback verification unit monitors the adjusted module's operating status, verifies whether the adjustment effect meets the standard, and forms a closed-loop control.

[0108] The feedback verification unit continuously monitors key indicators within 3-5 minutes after adjustment, and judges the effect by the difference between the actual value and the target value. If the difference is within the allowable range, the strategy is recorded as effective experience; if the difference is not within the allowable range, the decision analysis unit is triggered to recalculate the adjustment strategy to form a closed-loop control.

[0109] The data service module stores the processed data and decision results and provides standardized service interfaces to facilitate output in diverse formats.

[0110] The data service module includes a data storage unit, an interface service unit, a data formatting unit, and a permission management unit. The input end of the data storage unit is connected to the data processing module and the intelligent fusion decision-making module, respectively. The output end of the data storage unit is connected to the interface service unit, and the interface service unit is connected to the data formatting unit and the permission management unit, respectively.

[0111] The data storage unit is used to persistently store the processed data and the intelligent fusion decision results;

[0112] The data storage unit adopts a hybrid storage architecture to adapt to different types of data. The hybrid storage architecture mainly consists of a relational database for storing structured decision results, a time-series database for storing high-frequency monitoring data, an object storage for storing image or video data, and a cache database for caching hot data, so as to improve the interface response speed.

[0113] The interface service unit provides a standardized interface that supports external systems in querying or subscribing to data.

[0114] The interface service unit is built on an API gateway to construct a standardized interface layer, which includes RESTful API, WebSocket and gRPC interfaces. RESTful API supports query and batch export; WebSocket provides real-time push service, such as actively pushing alarm information to the command platform when the congestion level of a certain road segment exceeds the threshold; gRPC interface provides low-latency binary protocol communication for high-concurrency scenarios.

[0115] The data formatting unit converts the stored data into diverse output formats to meet the needs of different scenarios;

[0116] The data formatting unit is based on a template engine and visualization tools to achieve multi-format output, including structured format, report format and visualization format. The structured format defines the data structure through JSON Schema and supports XML conversion; the report format uses JasperReports to generate Excel / PDF reports; and the visualization format integrates ECharts / Chart.js to generate line charts (traffic trends) and heatmaps (congestion distribution), and returns chart data or HTML fragments through an interface.

[0117] The access control unit is used to control data access permissions and ensure data security.

[0118] The permission management unit implements identity authentication and authorization based on OAuth2.0+JWT. During identity authentication, the caller needs to pass a token (such as a JWT token) through the interface, and the permission management unit verifies the validity of the token (whether it has expired and whether the signature is correct).

[0119] During authorization, role-based access control (RBAC) models are used to define role permissions. For example, administrators can access all data, while ordinary users can only access public traffic status. The interface service unit filters data and returns results based on permissions.

[0120] When accessing heterogeneous data through the multi-source data dynamic access module, entity linking technology first performs preliminary matching of cross-source entities to generate unified entity identifiers. During the cleaning and integration process, the data processing module verifies redundancy (such as duplicate data collection of the same entity) and contradictions (such as different speed records of the same vehicle) based on the entity linking results. Noise is removed through entity attribute consistency rules (such as the binding relationship between vehicle ID and trajectory) to ensure data standardization. When the dynamic relationship learning module mines spatiotemporal correlations, entity linking technology helps identify implicit correlations between cross-modal entities (such as "the vehicle captured by camera A" and "the same vehicle in the GPS trajectory"). Combined with GNN and time series models, the time-varying mapping model is optimized to make the association rules more accurate.

[0121] The intelligent fusion decision-making module utilizes unified entity identifiers to integrate multi-source data features (such as traffic flow and weather data for the same road segment), and generates decisions through weighted fusion to avoid decision bias caused by entity ambiguity. The adaptive adjustment module monitors the accuracy of entity linking in real time (such as cross-source entity matching error rate). If a decline in association quality is detected, dynamic adjustments are made: the entity recognition algorithm is optimized for the data access module, the entity consistency verification rules are updated for the processing module, and the association weight parameters are adjusted for the relationship learning module, forming a closed-loop optimization.

[0122] The data service module stores fused data with unified entity identifiers and outputs it through standardized interfaces to ensure that downstream applications (such as traffic command systems) use consistent entity association information, thereby achieving end-to-end entity association reliability from data access to decision-making applications and improving fusion quality.

[0123] Through the coordinated operation of the multi-source data dynamic access module, data processing module, dynamic relationship learning module, intelligent fusion decision-making module, adaptive adjustment module, and data service module, the entire process from multi-source data access to intelligent control and decision-making can be automated. It can also adapt to changes in complex traffic environments and improve the level of refined management of traffic engineering.

[0124] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-source monitoring data fusion intelligent control system for traffic engineering, characterized in that: include: The multi-source data dynamic access module is used to access multi-source heterogeneous data and generate raw data with unified entity identifiers. The data processing module is used to clean, transform, and integrate the incoming raw data, eliminate noise, redundancy, and contradictions, and generate standardized, reusable multi-source data. The dynamic relationship learning module is used to mine the dynamic spatiotemporal correlations and potential patterns in multi-source data and to establish a time-varying mapping model between data. The intelligent fusion decision-making module generates traffic control decisions based on the processed multi-source data and the output of the time-varying mapping model. The adaptive adjustment module monitors the operation status of the multi-source data dynamic access module, data processing module, dynamic relationship learning module, and intelligent fusion decision-making module in real time, and dynamically adjusts parameters according to the external environment and system performance to ensure system stability. The data service module stores the processed data and decision results and provides standardized service interfaces to facilitate output in diverse formats. The adaptive adjustment module is electrically connected to the multi-source data dynamic access module, the data processing module, the dynamic relationship learning module, and the intelligent fusion decision module, respectively.

2. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 1, characterized in that: The multi-source data dynamic access module includes an interface adaptive adaptation unit, a dynamic load balancing unit, and an access status monitoring unit connected in sequence. The interface adaptive adaptation unit receives data through a standardized access interface and then uses a protocol conversion component to be compatible with the data formats of different devices. The dynamic load balancing unit automatically expands access nodes through the Kubernetes container orchestration platform based on real-time data volume, while using a flow control algorithm to limit the access frequency of abnormal data sources. The access status monitoring unit has a built-in heartbeat detection mechanism that provides real-time alarms for offline devices and automatically switches to a backup data source.

3. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 2, characterized in that: The specific conversion process of the protocol conversion component is as follows: A1. Receive raw data from different devices and automatically identify the source protocol type through a preset protocol feature library; A2. Parse the data frame according to the format specification of the source protocol; A3. Convert the extracted heterogeneous data into a system-compatible format and encapsulate the RTSP stream into a binary message body of the MQTT protocol; A4. Encapsulate the data according to the specifications of the target protocol, which includes RESTful API and MQTT; A5. Verify the integrity of the converted data, send the standardized data to the data processing module, and record the conversion log.

4. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 3, characterized in that: The heartbeat detection mechanism is used to monitor whether devices are online in real time. The monitoring process of the heartbeat detection mechanism includes heartbeat packet definition, sending and receiving rules, offline judgment and alarm, and backup data source switching. Among them, the content of the heartbeat packet definition mechanism is the status message sent by the device periodically. The status message includes the device ID, current status, timestamp, and checksum. The sending and receiving rules include sending rules and receiving rules. The sending rule is that the device sends heartbeat packets to the system's receiving port at fixed intervals. The receiving rule is that the system records the last time it receives the heartbeat packet and replies with an acknowledgment message to inform the device that it has been received. The offline determination and alarm rules are determined as follows: if the system does not receive a heartbeat packet from a device within 3 cycles, it is determined to be offline, and the system automatically generates an alarm message and pushes it to the operation and maintenance platform; if the device is determined to be offline, the system automatically calls the backup strategy.

5. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 1, characterized in that: The data processing module includes a data cleaning unit, a data conversion unit, and a data integration unit, which are connected in sequence. The data cleaning unit is used to perform noise filtering, anomaly detection, and missing value imputation on the raw data. The data conversion unit performs standardization and feature extraction processing on the cleaned data; The data integration unit is used to perform multi-source alignment and remove redundant data from the converted data.

6. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 5, characterized in that: The dynamic relationship learning module includes a spatiotemporal correlation modeling unit and a dynamic rule update unit. The spatiotemporal correlation modeling unit includes time dimension modeling and spatial dimension modeling. The time dimension modeling uses LSTM or Transformer time series model to analyze the time dependency of data and combines a sliding window mechanism to update the time correlation rules in real time. Spatial dimension modeling is based on graph neural networks to construct a road network topology graph, and the spatial influence weights of adjacent road segments and upstream and downstream road segments are quantified through graph attention mechanism. The dynamic pattern update unit adopts an online framework to receive new data and update model parameters in real time, avoiding the failure of the relational model due to changes in traffic patterns.

7. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 6, characterized in that: The intelligent fusion decision module includes a multi-source fusion unit and a decision generation unit. The multi-source fusion unit uses DS evidence theory to fuse sensor traffic flow, camera recognition data, and mobile phone signaling calculation data. Then, it constructs a multimodal fusion model based on the fused data and predicts and outputs traffic status through the multimodal fusion model. When generating traffic control decisions, the decision generation unit presets basic strategies based on morning and evening peak hour rules, uses a reinforcement learning algorithm with the maximization of road network communication efficiency as the reward function, dynamically adjusts control measures, and pushes them in real time according to the control measures.

8. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 7, characterized in that: The adaptive adjustment module includes a status monitoring unit, a decision analysis unit, a parameter execution unit, and a feedback verification unit connected in sequence. The status monitoring unit collects the operating indicators of each module in real time to provide data basis for adjustment. The decision analysis unit determines whether adjustments are needed based on monitoring data and calculates the optimal adjustment strategy. The parameter execution unit converts the adjustment strategy output by the decision analysis unit into specific instructions, sends them to the target module, and executes them. The feedback verification unit monitors the adjusted module's operating status, verifies whether the adjustment effect meets the standard, and forms a closed-loop control.

9. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 8, characterized in that: The decision analysis unit uses a combination of rule engine and adaptive algorithm for analysis. First, a threshold is preset, and then a reinforcement learning algorithm is used with failure rate and processing efficiency as reward functions to learn and dynamically adjust parameters.

10. The intelligent control system for multi-source monitoring data fusion in traffic engineering according to claim 9, characterized in that: The data service module includes a data storage unit, an interface service unit, a data formatting unit, and a permission management unit. The input end of the data storage unit is connected to the data processing module and the intelligent fusion decision-making module, respectively. The output end of the data storage unit is connected to the interface service unit, and the interface service unit is connected to the data formatting unit and the permission management unit, respectively.

Images