Multi-source data fusion-based rural highway traffic state safety monitoring method and system

Abstract

The present application belongs to the technical field of highway traffic safety monitoring, and particularly relates to a rural highway traffic state safety monitoring method and system based on multi-source data fusion. The method comprises the following steps: screening key monitoring points of rural highways, and obtaining original monitoring data of the key monitoring points; extracting traffic state features and safety risk features of the original monitoring data, and obtaining core fusion features after feature screening and fusion; inputting the core fusion features into a fuzzy neural network model, and performing comprehensive decision analysis on the core fusion features to obtain determination results of traffic state and safety risk; and adopting a dynamic early warning threshold self-adjusting mechanism based on road section features and real-time weather conditions to dynamically update the dynamic early warning threshold, and performing an alarm operation when the safety risk output by the fuzzy neural network model reaches an early warning condition. The present application is suitable for the traffic state safety monitoring scene of rural highways, and improves the accuracy of rural highway traffic state monitoring.

Description

Technical Field

[0001] This invention belongs to the field of highway traffic safety monitoring technology, and in particular relates to a method and system for monitoring the traffic condition and safety of rural roads by multi-source data fusion. Background Technology

[0002] Rural roads are the core carriers of rural transportation networks. However, due to limitations imposed by geographical environment, infrastructure conditions, and management level, there are still many problems in the current rural road traffic monitoring work. The most prominent issue is that existing traffic monitoring systems are mostly designed for urban roads and do not fully consider the unique characteristics of rural roads, such as dispersed road sections, mixed pedestrian and vehicle traffic, significant weather influences, and complex traffic behavior. As a result, they cannot accurately identify and analyze the unique safety risks of rural roads, and the accuracy of monitoring the traffic conditions of rural roads needs to be improved. Summary of the Invention

[0003] To overcome the shortcomings of the existing technologies, this invention provides a method and system for monitoring the traffic condition safety of rural roads using multi-source data fusion. Multi-source sensors are deployed at key monitoring points on rural roads to acquire raw monitoring data. A three-level fusion strategy combining data-level fusion, feature-level fusion, and decision-level fusion is employed to improve data effectiveness and analytical accuracy. Finally, automatic alarms are achieved based on dynamic early warning thresholds corrected by real-time weather conditions. This method is adapted to the traffic condition safety monitoring scenarios of rural roads, thereby improving the accuracy of rural road traffic condition monitoring.

[0004] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides a method for monitoring the traffic condition and safety of rural roads by fusing multi-source data.

[0005] A method for monitoring rural road traffic safety through multi-source data fusion includes the following steps: Select key monitoring points for rural roads and obtain raw monitoring data from these key monitoring points; Traffic status features and safety risk features are extracted from the original monitoring data. After feature screening and fusion, core fused features are obtained. The core fusion features are input into the fuzzy neural network model, and a comprehensive decision analysis is performed on the core fusion features to obtain the judgment results of traffic status and safety risks; A dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions is adopted to dynamically update the dynamic early warning threshold, and to execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

[0006] The second aspect of this invention provides a rural road traffic condition safety monitoring system based on multi-source data fusion.

[0007] A rural road traffic condition safety monitoring system based on multi-source data fusion includes: The data acquisition module is configured to: filter key monitoring points on rural roads and acquire the raw monitoring data of the key monitoring points; The feature extraction module is configured to: extract traffic status features and safety risk features from the original monitoring data, and after feature filtering and fusion, obtain the core fused features; The integrated decision-making module is configured to: input the core fusion features into the fuzzy neural network model, perform integrated decision analysis on the core fusion features, and obtain the judgment results of traffic status and safety risks; The alarm execution module is configured to: adopt a dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions to dynamically update the dynamic early warning threshold, and execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

[0008] The above one or more technical solutions have the following beneficial effects: This invention discloses a method and system for monitoring rural road traffic safety through multi-source data fusion. Starting from the scenario adaptability to rural roads, it first selects key monitoring points on rural roads and obtains their raw monitoring data. Then, based on a three-level fusion strategy of data-level fusion, feature-level fusion, and decision-level fusion, it obtains the core fusion features of rural roads and simultaneously determines the traffic status and safety risks of rural roads. Finally, it implements automatic alarms based on dynamic early warning thresholds corrected by real-time weather conditions. This invention, based on deep fusion of multi-source data, significantly improves data effectiveness and analytical accuracy, and enhances adaptability to rural road traffic safety monitoring scenarios.

[0009] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0010] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0011] Figure 1 This is a flowchart of the method in Example 1.

[0012] Figure 2 This is a layout diagram of the perception layer-transmission layer-fusion layer-application layer in Example 1.

[0013] Figure 3 This is a schematic diagram of the sensor layout in the sensing layer of Embodiment 1.

[0014] Figure 4 This is a schematic diagram of multi-source data hierarchical fusion for the feature extraction module in Example 2. Detailed Implementation

[0015] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0016] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0017] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0018] Example 1 In the field of rural road traffic condition monitoring, in addition to the insufficient adaptability to rural road scenarios and the inability to accurately identify and analyze the unique safety risks of rural roads mentioned in the background section, the following technical problems also exist: 1. Limited data collection dimensions: Existing monitoring methods mostly rely on single cameras or manual statistics, which cannot comprehensively capture multi-dimensional data such as traffic flow, vehicle speed, pedestrian trajectories, and weather conditions, making it difficult to reflect the complete status of rural road traffic operations.

[0019] 2. Difficulty in integrating multi-source data: Various types of data collected from the roadside have problems such as heterogeneous formats and inconsistent quality. There is a lack of effective integration methods, resulting in low data utilization and thus causing large deviations in traffic condition assessment results, which cannot meet the needs of precise management.

[0020] Furthermore, existing methods for fusing multi-source data are mostly designed for urban roads, failing to consider the technical challenges of rural roads: First, sensors use continuous acquisition modes, resulting in high power consumption and making them unsuitable for rural roads lacking stable power supply; second, multi-source data fusion only employs simple splicing methods, without designing spatiotemporal synchronization rules and feature selection rules for heterogeneous data from rural roads, leading to low fusion accuracy; third, data transmission uses a fixed mode, failing to consider the weak network environment of rural roads, easily causing data loss; fourth, warning thresholds are fixed values, failing to incorporate road segment characteristics and real-time weather conditions, resulting in low warning accuracy.

[0021] 3. Disconnect between monitoring and analysis: Existing systems can only display basic data in a simple way, and cannot identify abnormal traffic conditions and predict accident risks through data fusion, making it difficult to support the implementation of proactive safety management on rural roads.

[0022] In summary, current technologies lack a method specifically designed for rural road scenarios that enables efficient fusion of multi-source roadside data, accurate traffic condition monitoring, and integrated analysis of safety risks. Therefore, developing a multi-source data fusion-based traffic monitoring and safety analysis method adapted to the characteristics of rural roads has significant practical implications and application value.

[0023] This embodiment discloses a method for monitoring the traffic condition and safety of rural roads through multi-source data fusion. It is applicable to traffic data collection, fusion processing, condition assessment, and safety risk analysis in rural road scenarios, providing accurate data support and scientific decision-making basis for rural road traffic safety management.

[0024] like Figure 1 As shown, the method for monitoring rural road traffic safety through multi-source data fusion includes the following steps: Select key monitoring points for rural roads and obtain raw monitoring data from these key monitoring points; Traffic status features and safety risk features are extracted from the original monitoring data. After feature screening and fusion, core fused features are obtained. The core fusion features are input into the fuzzy neural network model, and a comprehensive decision analysis is performed on the core fusion features to obtain the judgment results of traffic status and safety risks; A dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions is adopted to dynamically update the dynamic early warning threshold, and to execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

[0025] The core objective of this invention is to address the problems of single data collection methods, difficulty in fusion of multi-source data, poor adaptability of status monitoring and safety analysis, high sensor power consumption, unstable data transmission in weak networks, and low adaptability of early warning thresholds in rural roads. It aims to achieve collaborative collection and hierarchical fusion of multi-dimensional traffic data for rural roads, accurately assess traffic operation status, efficiently identify traffic anomalies and safety risks, and provide real-time and accurate technical support for rural road traffic safety management, facility optimization, and decision-making.

[0026] like Figure 2 As shown, this embodiment adopts a four-level distributed architecture of "perception layer - transmission layer - fusion layer - application layer," and utilizes four proprietary technologies: low-power networking for rural road data collection, heterogeneous data fusion algorithm for rural roads, stable transmission protocol for weak network rural roads, and self-adjusting program for rural road risk thresholds. Each layer collaborates and operates in tandem, enabling bidirectional data flow and achieving a complete closed-loop process from on-site data collection on rural roads to traffic condition monitoring and safety analysis result output. Overall: The perception layer is used to realize the collaborative acquisition of multi-source roadside data. It integrates high-definition cameras, millimeter-wave radar, vehicle sensors, small weather stations and data acquisition terminals, and is equipped with rural road data acquisition low-power networking technology to realize low-power collaborative acquisition of sensors. The transport layer is used to realize multi-channel data transmission, adopting a multi-channel data transmission method of "wireless as the main method and wired as the auxiliary method", and carrying a rural road weak network stable transmission protocol to realize stable data transmission in weak network environments; The fusion layer is used to realize hierarchical fusion processing of multi-source data. It adopts a three-level fusion strategy of "data-level fusion - feature-level fusion - decision-level fusion" and is equipped with a rural road heterogeneous data fusion algorithm to achieve deep fusion of rural road heterogeneous multi-source data. The application layer is used to realize traffic condition monitoring and safety analysis display, including three functions: real-time traffic condition monitoring, intelligent analysis of safety risks, and data management and decision support. It is equipped with a rural road risk threshold self-adjustment program to realize dynamic self-adjustment of warning thresholds.

[0027] The above architecture is well adapted to the distribution characteristics of roadside infrastructure in rural roads, and has the advantages of flexible deployment, strong scalability, and low power consumption operation, achieving precise adaptation to rural road scenarios.

[0028] (I) Detailed implementation of the technical solution.

[0029] The method of this embodiment will now be explained in detail.

[0030] 1. Perception layer: Collaborative collection of multi-source roadside data.

[0031] The perception layer is used to acquire basic data sources and is adapted to the outdoor installation environment of rural roads. It integrates a variety of low-cost and highly adaptable roadside sensors. All sensors support low-power outdoor operation and can effectively resist interference from severe weather. The core is equipped with rural road data acquisition low-power networking technology to achieve low-power collaborative data acquisition of sensors.

[0032] like Figure 2 As shown, the perception layer specifically includes the following units: (1) Video acquisition unit: High-definition cameras are used to collect visual data such as vehicle type, driving speed, driving trajectory, pedestrian flow, pedestrian crossing behavior and non-motorized vehicle driving status within the rural road area; (2) Traffic sensing unit: Deploy vehicle sensors to collect basic traffic flow data such as traffic flow, headway, and lane occupancy rate, providing core parameters for traffic condition assessment; (3) Radar detection unit: Millimeter-wave radar is used to effectively compensate for the camera's acquisition defects in low visibility environments such as rain, fog, and night, and accurately capture the dynamic position and motion parameters of vehicles and pedestrians; (4) Meteorological monitoring unit: Deploy small meteorological stations to collect meteorological data such as rainfall, snowfall, visibility, wind speed, and road surface temperature that affect rural road traffic safety, and provide environmental parameters for safety risk analysis; (5) Data acquisition terminal: Seamlessly connected to various sensors to realize unified acquisition of multi-source data, preliminary filtering (removing obviously invalid data) and format standardization preprocessing, ensuring smooth data transmission and fusion in the future.

[0033] The low-power networking technology for rural road data collection adopts a time-slice polling collection strategy plus a trigger-based collaborative collection strategy, specifically including: Each sensor adopts a time-slice polling acquisition strategy, dynamically adjusting the sensor acquisition frequency according to the traffic flow characteristics of rural roads (off-peak / peak periods): during off-peak traffic periods (such as at night), the camera / radar acquires data at a low frequency of 5 frames / second, and the weather station acquires data once every 10 minutes; during peak traffic periods (such as market days at the village entrance), the acquisition frequency is automatically switched to 25 frames / second, and the weather station acquires data once every 1 minute.

[0034] Millimeter-wave radar and high-definition camera adopt a triggered collaborative acquisition: the radar first performs coarse detection of moving objects, and when a vehicle / pedestrian is detected to enter the monitoring range, the camera is triggered to capture high-definition images. When no moving object is detected, the camera enters sleep mode. Compared with the traditional continuous acquisition mode, the sensor power consumption is reduced by more than 60%, which is suitable for rural roads without stable power supply.

[0035] 2. Transport layer: Multi-channel data transmission.

[0036] To address the issue of weak network signals in some areas of rural roads, the transmission layer adopts a multi-channel data transmission method with "wireless as the primary method and wired as the secondary method" to ensure that the data collected by the perception layer can be stably and in real time transmitted to the fusion layer. The core of the protocol is the rural road weak network stable transmission protocol, which enables stable data transmission in weak network environments.

[0037] like Figure 2 As shown, the transport layer specifically includes the following units: (1) Wireless transmission unit: Supports multiple wireless communication methods such as 4G / 5G, LoRa, and NB-IoT. For remote rural road sections, LoRa / NB-IoT low-power wireless transmission mode is preferred to ensure the stability of data transmission in weak network environments; (2) Wired transmission unit: For road sections with good network signal around towns and villages, fiber optic and Ethernet wired transmission methods are adopted to greatly improve the transmission efficiency of large data such as video data; (3) Data caching unit: A local caching module is set up so that the collected data can be temporarily stored when the transmission network is interrupted and automatically retransmitted after the network is restored, which effectively avoids data loss and ensures data integrity.

[0038] Wherein, LoRa is a low-power local area network wireless standard, and NB-IoT is Narrow Band Internet of Things.

[0039] The stable transmission protocol for weak networks on rural roads includes a dynamic transmission mode switching rule, a cyclic caching strategy, and a resume transmission from break-point algorithm. Aiming at the network quality of different sections of rural roads, the dynamic transmission mode switching rule is designed to realize stable data transmission. The dynamic transmission mode switching rule is specifically as follows: The network quality detection module detects the network bandwidth (B) and packet loss rate (L) of the road section in real time. When B≥2Mbps and L≤5%, 4G / fiber is used to transmit high-definition video data, where Mbps means megabits per second; When 0.5Mbps≤B<2Mbps and 5%<L≤15%, the video data is subjected to H.265 lightweight compression (compression ratio 10:1) and then transmitted via NB-IoT; When B<0.5Mbps and L>15%, only the numerical core data of radar / meteorology is transmitted, and the video data is temporarily stored in the local cache unit.

[0040] The data cache unit adopts a cyclic caching strategy with a cache capacity of 64G, and overwrites according to the "time priority" principle. After the network is restored, the resume transmission from break-point algorithm is adopted to preferentially transmit high-risk data in the cache (such as video data where speeding / pedestrian crossing is detected), so as to ensure no loss of core safety data, and the data transmission success rate in a weak network environment reaches more than 98%.

[0041] 3. Fusion layer: Hierarchical fusion processing of multi-source data.

[0042] Aiming at the problems of heterogeneity and quality differentiation of multi-source data on rural roads, this embodiment innovatively adopts a three-level fusion strategy of "data-level fusion - feature-level fusion - decision-level fusion", and is centrally equipped with a rural road heterogeneous data fusion and connection algorithm to realize deep fusion of multi-source data, which significantly improves data validity and analysis accuracy.

[0043] On the whole, the data-level fusion adopts the spatio-temporal synchronization rule based on the radar timestamp and the standardized mapping rule for heterogeneous data; the feature-level fusion adopts the core feature screening rule of the mutual information method and the weighted fusion algorithm, retaining 12 dimensions of core fusion features for rural roads, and the weights of pedestrian crossing frequency and mixed pedestrian-vehicle traffic density are set to 0.2; the decision-level fusion adopts a fuzzy neural network model of "3-layer fuzzification + 3-layer neural network", and the overall data fusion accuracy reaches more than 92%.

[0044] As Figure 2 shown in the figure, the fusion layer is used to realize hierarchical fusion processing of multi-source data, including data-level fusion, feature-level fusion and decision-level fusion. The specific functions of each fusion level are as follows: (1) Data-level fusion: The raw data transmitted by the perception layer is cleaned, deduplicated, completed and synchronized. Noise data and abnormal data are removed. The raw data of different formats and acquisition frequencies (such as video frame data, radar numerical data and meteorological monitoring data) are uniformly converted into standardized data formats and synchronized alignment in time and space dimensions, laying the foundation for subsequent fusion processing.

[0045] Specifically, it includes: A unified spatiotemporal synchronization rule was established, using the timestamp of radar data as a benchmark. Video frame data was time-interpolated and aligned (interpolation step size 0.04s), and meteorological data was spatially interpolated and aligned (grid interpolation based on a monitoring point spacing of 500m). At the same time, a standardized mapping table for heterogeneous data was established, mapping the vehicle type identification results of the camera (small car / agricultural vehicle / non-motorized vehicle) to digital codes (1 / 2 / 3), normalizing the speed data of radar (km / h) to the [0,1] interval, and mapping the visibility (m) of meteorological data to (0.1 / 0.3 / 0.5 / 0.7 / 0.9) according to the level, so as to achieve standardization and unification of different types of data.

[0046] (2) Feature-level fusion: Traffic status features and safety risk features are extracted from the standardized data after data-level fusion, such as vehicle speeding features, pedestrian crossing features, visibility-vehicle speed coupling features in rainy and foggy weather, and density features of mixed pedestrian and vehicle traffic. Through feature association, feature screening, and feature fusion algorithms, a fusion feature set is constructed to provide core feature support for subsequent decision analysis.

[0047] Among them, traffic condition characteristics are parameter characteristics that reflect the basic traffic operation status of rural roads, mainly including: vehicle speeding characteristics, headway, lane occupancy, traffic flow, average vehicle speed, and vehicle type distribution (such as the proportion of agricultural vehicles).

[0048] Safety risk characteristics are abnormal behaviors and environmental coupling characteristics directly related to rural road traffic accident risks, mainly including: pedestrian crossing characteristics (pedestrian crossing frequency), pedestrian and vehicle mixed traffic density characteristics, non-motorized vehicle lane occupation time, nighttime driving ratio, and visibility-vehicle speed coupling characteristics in rainy and foggy weather.

[0049] Specifically, considering the characteristics of rural roads, such as "mixed traffic of pedestrians and vehicles, a large number of agricultural vehicles, and significant weather influences," this embodiment designs a selection rule that integrates feature sets. It uses the mutual information method to calculate the correlation between each feature and traffic risk, eliminates redundant features with a correlation of less than 0.3 (such as bus stop features on urban roads), and retains core features (12 dimensions in total): vehicle speeding value, proportion of agricultural vehicles, frequency of pedestrian crossings, density of mixed traffic of pedestrians and vehicles, visibility, road surface slipperiness, curve curvature, slope gradient, headway, lane occupancy rate, duration of non-motorized vehicle lane occupation, and proportion of nighttime driving.

[0050] Subsequently, a weighted fusion algorithm was used to analyze the core features after screening, with the weights of pedestrian crossing frequency and mixed pedestrian and vehicle density set to 0.2 (far higher than those on urban roads) to adapt to the safety risk characteristics of rural roads.

[0051] (3) Decision-level fusion: Based on the feature set formed by feature-level fusion, combined with the traffic operation characteristics of rural roads, a fuzzy neural network algorithm is introduced to conduct comprehensive decision analysis on the fused data, and finally output a unified and accurate traffic status and safety risk judgment result.

[0052] The fuzzy neural network introduced in this embodiment adopts a hybrid structure of "3-layer fuzzification + 3-layer neural network", specifically designed as follows: The input layer consists of 12-dimensional core fusion features of rural roads; The fuzzing layer divides each input feature into 5 fuzzy subsets (extremely low / low / medium / high / extremely high), and uses a Gaussian membership function for fuzzing. The parameters (mean / variance) of the membership function are obtained by training with historical accident data of rural roads (e.g., the mean of pedestrian crossing frequency is set to 2 times / hour, and the variance is set to 0.5). The rule layer sets 60 fuzzy rules, all designed for rural road scenarios (such as "if visibility is extremely low and the density of pedestrians and vehicles is high, then the traffic risk is extremely high"). The neural network layer uses a BP neural network with 20 hidden layer nodes, the activation function is Sigmoid, and the output layer is a 2D result (traffic status level / safety risk level). The training optimization used over 1000 historical data points from a street on a rural road. L2 regularization was added to prevent overfitting, and the training accuracy reached over 95%. The overall accuracy of data fusion reached over 92%.

[0053] 4. Application Layer: Traffic Condition Monitoring and Safety Analysis Display.

[0054] The application layer serves as the system's output and interaction terminal, targeting users such as rural road traffic management departments and maintenance units. It enables integrated display of data visualization, status monitoring, safety analysis, and result early warning. At its core, it is equipped with a self-adjusting program for rural road risk thresholds, achieving precise adaptation of early warning thresholds.

[0055] like Figure 2 As shown, the application layer is used for traffic condition monitoring and safety analysis visualization, and mainly includes: (1) Real-time traffic status monitoring: Through a visual interface, basic data such as traffic flow, average speed, vehicle type distribution, pedestrian flow, and weather conditions of each monitored section of rural roads are displayed in real time, and traffic operation status evaluation results (including four levels: smooth, basically smooth, slow and congested) are automatically generated. (2) Intelligent analysis of safety risks: Based on the decision analysis results of the fusion layer, it automatically identifies abnormal traffic conditions (such as speeding, pedestrians crossing at will, non-motorized vehicles occupying the road, and the impact of severe weather), analyzes the accident risk level of each monitored road section, accurately locates risk points (such as village entrance crossings, curves, slopes, etc.), and automatically generates risk analysis reports. (3) Data management and decision support: realize the storage, query and statistical analysis of multi-source fusion data, traffic status data and safety risk data, support historical data backtracking and trend analysis, and provide strong data support and decision basis for optimizing rural road traffic safety facilities, formulating traffic control measures and designing accident prevention plans.

[0056] This embodiment abandons the traditional fixed warning threshold and designs a dynamic self-adjustment mechanism for warning thresholds based on road segment characteristics and real-time weather. The specific implementation process of the rural road risk threshold self-adjustment program includes: 1. Set basic thresholds based on the static characteristics of road segments (curves / slopes / village entrances). For example, the basic threshold for pedestrian crossing warnings at village entrance crossings is set to 3 times / hour, and for straight road segments, it is set to 5 times / hour.

[0057] Based on the static characteristics of road segments (curves / slopes / village entrances), basic thresholds are set. The specific setting process includes: The static attributes of rural roads, such as road alignment and surrounding environment, are classified in advance. The classification types include curves, slopes, straight road sections and village entrance crossings. Based on historical traffic accident data, traffic flow patterns, and inherent risk levels for various road types, differentiated initial baseline values ​​for safety risk warnings are matched and set for different types of road sections: For special road sections with high accident risk and high probability of traffic conflict (such as village crossings, sharp bends, and steep slopes), a relatively low basic threshold is set to improve sensitivity. For straight road sections with relatively low risk, a relatively high base threshold is set to achieve accurate matching between the early warning benchmark and the actual physical characteristics of the road section.

[0058] 2. Set threshold correction coefficients based on real-time weather conditions. For example, when visibility is <200m, the correction coefficient is 0.6, lowering the base threshold by 60%; when the road surface is icy, the correction coefficient is 0.5.

[0059] Threshold correction coefficients are set based on real-time meteorological conditions, specifically including: Establish a mapping relationship between meteorological parameters and traffic safety risks in advance; It can acquire various meteorological data that affect traffic safety in real time (such as visibility, rainfall / snowfall, road surface temperature / slipperiness, wind speed, etc.) and divide them into multiple meteorological risk level ranges according to the severity of different meteorological conditions; For each meteorological risk level range, a corresponding threshold correction coefficient is configured: The more severe the weather conditions and the greater the adverse impact on driving safety, the smaller the correction coefficient should be set, so as to dynamically lower the basic warning threshold and thus improve the sensitivity of warnings under severe weather conditions.

[0060] 3. Dynamic early warning threshold = base threshold × correction coefficient.

[0061] The system updates the threshold every 5 minutes based on real-time meteorological data, achieving precise adaptation of the warning threshold and a warning accuracy of over 85%. Compared with the fixed threshold method, it reduces the false alarm rate and missed alarm rate by 40%.

[0062] The workflow of this embodiment revolves around a closed loop of "data acquisition-transmission-fusion-analysis-output," and utilizes four proprietary technologies to adapt to rural road scenarios at each stage. The specific steps are as follows: 1. The perception layer uses low-power networking technology for rural road data acquisition to enable low-power collaborative data acquisition by various sensors on rural road monitoring sections. It simultaneously collects raw data from multiple sources such as video, traffic flow, radar, and meteorology. After preliminary filtering and format preprocessing by the data acquisition terminal, the data is transmitted to the transmission layer. 2. The transmission layer uses the rural road weak network stable transmission protocol to flexibly select wireless or wired transmission methods according to the network environment of the monitored road section, and stably transmits the pre-processed raw data to the fusion layer; if the network is interrupted, the data buffer unit will temporarily store the data and automatically retransmit it through the breakpoint resume transmission algorithm after the network is restored. 3. The fusion layer uses a rural road heterogeneous data fusion algorithm to perform three-level fusion processing on the received multi-source data: data level, feature level, and decision level. It removes invalid data, extracts core features, and completes comprehensive decision analysis through a fuzzy neural network algorithm to output accurate traffic status and safety risk assessment results. 4. The application layer receives the decision analysis results from the fusion layer, displays the traffic operation status in real time through a visual interface, intelligently analyzes the safety risk level and risk points, generates monitoring reports and risk analysis reports, and realizes dynamic early warning through the rural road risk threshold self-adjustment program, supporting data storage, query and trend analysis; 5. The system enables bidirectional data flow at all levels. User operation commands at the application layer (such as selection of monitoring sections, data query conditions, and basic threshold settings for risk warning) can be transmitted back to each level, enabling flexible control of the system.

[0063] (ii) System deployment of the technical solution.

[0064] Taking a rural road as a specific implementation section, this section is a typical mountainous rural road characterized by mixed pedestrian and vehicular traffic, numerous curves, and significant impact from weather changes, making it suitable for the deployment and application of this system. The specific deployment plan is as follows: 1. Perception layer deployment: High-definition cameras, millimeter-wave radar, vehicle sensors, and small weather stations are deployed at key monitoring points such as village entrances, bends, and slopes along this road section. Each monitoring point is equipped with a data acquisition terminal, which is equipped with low-power networking technology for rural road data acquisition to achieve low-power local preliminary processing of multi-source data.

[0065] like Figure 3 The diagram shows the sensor layout of the perception layer. Example 3 illustrates the sensor deployment locations, quantity, coverage area, and specific application of low-power networking technology for data collection on rural roads at key monitoring points. Red dots represent monitoring points, and different symbols represent millimeter-wave radar, vehicle sensors, cameras, and small weather stations. Figure 3 The image shows three monitoring points: Monitoring Point 1, Monitoring Point 2, and Monitoring Point 3. Specifically, Monitoring Point 1 is the village entrance crossing, Monitoring Point 2 is a bend in the road, and Monitoring Point 3 is a ramp. At each monitoring point, millimeter-wave radar, vehicle sensors, cameras, and a small weather station are deployed. The cameras and small weather station are located on one side of the road, while the millimeter-wave radar and vehicle sensors are located on the other side.

[0066] 2. Transmission Layer Deployment: To address the issue of weak network signals in some areas of this road section, a dual-mode wireless transmission method of "NB-IoT + 4G" is adopted. NB-IoT / 4G transmission modules are deployed at each monitoring point, and local data caching terminals are set up, equipped with the rural road weak network stable transmission protocol to ensure stable data transmission.

[0067] 3. Fusion Layer Deployment: Build a cloud-based fusion processing platform, deploy heterogeneous data fusion algorithms and fuzzy neural network models for rural roads, and realize remote fusion processing and analysis of multi-source data.

[0068] 4. Application Layer Deployment: Build a visual operation platform for rural road traffic management departments, deploy core modules such as traffic status monitoring, safety analysis, and data management, and carry out a self-adjusting program for rural road risk thresholds to realize local display of monitoring results, dynamic early warning, and interactive operation.

[0069] (III) Specific Operation.

[0070] 1. The sensors in the perception layer continuously collect multi-source raw data such as traffic flow, vehicles, pedestrians, and weather on the implemented road sections 24 hours a day through the low-power networking technology of rural road data acquisition. The data acquisition terminal filters the raw data (removing blurry frames from cameras, invalid radar detection data, etc.) and standardizes the format, converting it into a unified numerical and video stream format.

[0071] 2. The transmission layer transmits the pre-processed raw data to the cloud-based converged processing platform via the rural road weak network stable transmission protocol and the NB-IoT / 4G module. If the network is interrupted, the data caching terminal will temporarily store the data for up to 72 hours. After the network is restored, the data will be automatically retransmitted through the breakpoint resume transmission algorithm to ensure that the data is not lost.

[0072] 3. The cloud-based fusion processing platform uses a heterogeneous data fusion algorithm for rural roads to perform three levels of fusion processing on the received data: First, it performs data-level fusion by cleaning, deduplication, and synchronization; then, it extracts core features such as speeding, pedestrian crossing, and visibility below 500 meters to complete feature-level fusion; finally, it completes decision-level fusion through a fuzzy neural network algorithm to output accurate traffic status and safety risk analysis results.

[0073] 4. The visualization operation platform displays basic data such as traffic flow, average vehicle speed, and weather conditions of the implemented road section in real time, automatically determines the traffic operation status (smooth / mostly smooth / slow / congested), accurately identifies abnormal traffic behaviors such as speeding and pedestrians crossing the road at will, dynamically adjusts the warning threshold through the self-adjustment program of rural road risk threshold, analyzes the accident risk level of each monitoring point, locates risk points such as village entrance crossings and bends, and automatically generates daily and weekly monitoring reports and risk analysis reports.

[0074] 5. Traffic management personnel can use the visualization platform to query historical data, adjust monitoring parameters, and set basic thresholds for risk warnings. The system can then adjust the work at each level based on user commands, enabling flexible and precise management of rural road traffic.

[0075] (iv) Implementation results.

[0076] The method described in this embodiment has achieved significant application results after being implemented on the aforementioned rural roads: it has enabled real-time collection and efficient fusion of multi-source traffic data for the road section through four proprietary technologies, achieving a traffic condition monitoring accuracy rate of over 95%, accurately identifying over 80% of abnormal traffic behaviors, effectively locating the unique safety risk points of rural roads, and providing precise data support for the optimization of traffic safety facilities on the road section.

[0077] Practical experience has shown that the implementation of this method can reduce the traffic accident rate on this road section by more than 30%, reduce sensor power consumption by more than 60%, achieve a data transmission success rate of more than 98% in weak network conditions, and achieve an early warning accuracy of more than 85%, significantly improving the level of intelligence and proactive safety capabilities of rural road traffic management.

[0078] This invention is specifically designed for the characteristics of rural roads, incorporating four proprietary technologies: low-power network for rural road data acquisition, heterogeneous data fusion algorithm for rural roads, stable transmission protocol for weak network rural roads, and self-adjusting procedure for rural road risk thresholds. Compared with existing technologies, it has the following significant advantages: 1. More comprehensive data collection: It integrates multiple roadside sensors such as cameras, vehicle sensors, radar, and weather stations to achieve collaborative collection of multi-dimensional data such as traffic flow, vehicles, pedestrians, and weather, effectively making up for the shortcomings of single data collection in existing rural roads; 2. More efficient data fusion: Adopting a three-level fusion strategy of "data level - feature level - decision level", the algorithm for fusion of heterogeneous rural road data solves the problems of heterogeneous data formats and quality differences from multiple sources. The data fusion accuracy reaches more than 92%, which is 30% higher than the traditional splicing and fusion method, perfectly adapting to the processing needs of multi-source data of rural roads. 3. Enhanced scenario adaptability: The system architecture adopts a low-power, multi-channel transmission, and distributed deployment design, which is suitable for the characteristics of rural roads, such as scattered road sections, weak network signals, and complex outdoor environments. At the same time, the safety analysis model combines the traffic characteristics of rural roads, such as "mixed pedestrian and vehicle traffic and significant weather influences," making the analysis results more in line with actual needs. 4. Integrated Functionality: It integrates traffic data collection, fusion processing, status monitoring, safety analysis, and data management, deeply combining data fusion with safety analysis. It can not only display basic data, but also accurately identify traffic anomalies and predict safety risks, providing strong support for proactive safety management of rural roads. 5. Practical and convenient operation: The application layer adopts a visual interface, which is simple and easy to understand. It supports real-time monitoring, historical data backtracking, and risk point location, which meets the actual needs of rural road traffic management departments. At the same time, the system has good scalability and can add monitoring sections and sensor types according to actual needs. 6. Lower sensor power consumption: Through the time-slice polling + trigger-based collaborative acquisition strategy of the rural road data acquisition low-power networking technology, the sensor power consumption is reduced by more than 60%, and no stable power supply is required, making it suitable for installation environments in rural roads where there is no stable power supply and the maintenance cost is high. 7. Enhanced stability in weak network transmission: Through dynamic transmission mode switching, lightweight compression, and breakpoint resume algorithm of the rural road weak network stable transmission protocol, the data transmission success rate in weak network environments reaches over 98%, with no loss of core security data, solving the pain point of weak network signal in rural roads. 8. Higher accuracy of early warning: Through the dynamic early warning threshold self-adjustment mechanism of the rural road risk threshold self-adjustment program, the early warning accuracy reaches more than 85%, which reduces the false alarm rate and missed alarm rate by 40% compared with the fixed threshold method, thus realizing the precision of rural road risk early warning.

[0079] Example 2 This embodiment discloses a rural road traffic condition safety monitoring system based on multi-source data fusion.

[0080] A rural road traffic condition safety monitoring system based on multi-source data fusion includes: The data acquisition module is configured to: filter key monitoring points on rural roads and acquire the raw monitoring data of the key monitoring points; The feature extraction module is configured to: extract traffic status features and safety risk features from the original monitoring data, and after feature filtering and fusion, obtain the core fused features; The integrated decision-making module is configured to: input the core fusion features into the fuzzy neural network model, perform integrated decision analysis on the core fusion features, and obtain the judgment results of traffic status and safety risks; The alarm execution module is configured to: adopt a dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions to dynamically update the dynamic early warning threshold, and execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

[0081] More specifically, the feature extraction module includes three sub-modules: a data-level fusion sub-module, a feature-level fusion sub-module, and a decision-level fusion sub-module. Figure 4 This is a schematic diagram illustrating the hierarchical fusion of multi-source data in the feature extraction module. (Example:) Figure 4 As shown: In the multi-source raw data input stage: video data, traffic flow data, radar detection data, and meteorological monitoring data collected by the perception layer are used as input; In the data-level fusion submodule: data cleaning, data deduplication, data completion, spatiotemporal synchronization alignment, and format standardization are implemented; In the feature-level fusion submodule: feature extraction, feature association, feature filtering, and feature fusion are implemented; In the decision-level fusion submodule: a fuzzy neural network model is constructed based on the fusion feature set to conduct multi-dimensional comprehensive decision analysis and finally obtain the traffic risk assessment result; In the output stage of the fusion results: the results of traffic operation status determination, safety risk level analysis, and abnormal behavior identification are output.

[0082] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0083] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for monitoring the traffic condition and safety of rural roads using multi-source data fusion, characterized in that, Includes the following steps: Select key monitoring points for rural roads and obtain raw monitoring data from these key monitoring points; Traffic status features and safety risk features are extracted from the original monitoring data. After feature screening and fusion, core fused features are obtained. The core fusion features are input into the fuzzy neural network model, and a comprehensive decision analysis is performed on the core fusion features to obtain the judgment results of traffic status and safety risks; A dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions is adopted to dynamically update the dynamic early warning threshold, and to execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

2. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, The key monitoring points for rural roads include village entrance crossings, bends, and slopes. The raw monitoring data includes video data, traffic sensing data, radar detection data, and meteorological monitoring data.

3. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, Raw monitoring data of key monitoring points is obtained by deploying video acquisition units, traffic sensing units, radar detection units, and meteorological monitoring units at these points. The video acquisition unit is used to collect visual data within the rural road area, including vehicle type, driving speed, driving trajectory, pedestrian flow, pedestrian crossing behavior, and non-motorized vehicle driving status. Traffic sensing units are used to collect basic traffic flow data, including traffic flow, headway, and lane occupancy. The radar detection unit is used to capture the dynamic position and motion parameters of vehicles and pedestrians; The meteorological monitoring unit is used to collect meteorological data including rainfall, snowfall, visibility, wind speed, and road surface temperature; The sensors in the video acquisition unit, traffic sensing unit, radar detection unit, and meteorological monitoring unit adopt a time-slice polling acquisition strategy, and dynamically adjust the acquisition frequency according to the traffic flow characteristics of rural roads. The radar detection unit and the video acquisition unit adopt a trigger-based collaborative acquisition method. When the radar detection unit does not detect a moving object, the video acquisition unit enters a sleep mode.

4. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, It also includes using dynamic transmission mode switching rules, a circular caching strategy, and a breakpoint resume algorithm to transmit raw monitoring data using a multi-channel data transmission method with wireless as the primary method and wired as a secondary method. The dynamic transmission mode switching rules are as follows: Real-time detection of network bandwidth and packet loss rate of road segment; when bandwidth is greater than or equal to the first preset threshold and packet loss rate is less than or equal to the second preset threshold, 4G / fiber optic transmission of video data is used. When the bandwidth is between the third preset threshold and the first preset threshold, and the packet loss rate is between the second preset threshold and the fourth preset threshold, the video data is compressed and then transmitted via NB-IoT. When the bandwidth is lower than the third preset threshold and the packet loss rate is greater than the fourth preset threshold, only the core numerical data of radar / meteorology is transmitted, and the video data is temporarily stored in the local cache unit.

5. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, It also includes using spatiotemporal synchronization rules based on radar timestamps and heterogeneous data standardization mapping rules to synchronize, align and standardize the original monitoring data to obtain standardized data.

6. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 5, characterized in that, The specific process for obtaining the core fusion features is as follows: Traffic condition features and safety risk features are extracted from standardized data, including vehicle speeding features, pedestrian crossing features, visibility-vehicle speed coupling features in rainy and foggy weather, and pedestrian-vehicle mixed traffic density features. The core feature selection rules of the mutual information method are adopted to calculate the correlation between each feature and traffic risk, and retain 12 core features of rural roads, including vehicle speeding value, proportion of agricultural vehicles, frequency of pedestrian crossing, density of mixed pedestrian and vehicle traffic, visibility, road surface slipperiness, curve curvature, slope gradient, headway, lane occupancy rate, duration of non-motorized vehicle lane occupation and nighttime driving ratio. The core fusion features are calculated using a weighted fusion algorithm based on the core features of the screened rural roads.

7. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, The fuzzy neural network model adopts a hybrid structure of fuzzification and neural network, and the specific data processing flow is as follows: The input layer of the fuzzy neural network model is the core fusion feature of rural roads; The fuzzing layer divides each input feature into 5 fuzzy subsets and uses a Gaussian membership function for fuzzing processing; The rule layer contains 60 fuzzy rules, all designed for rural road scenarios; The neural network layer uses a BP neural network, and the output layer is a 2D result, including traffic status level and safety risk level.

8. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 1, characterized in that, The dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions specifically includes: Determine the type of rural road sections to be monitored, including curves, slopes, straight sections, and village entrance crossings; Set basic thresholds based on different road segment types; Based on real-time weather conditions, a threshold correction coefficient is set; The product of the base threshold and the threshold correction coefficient is used as the dynamic early warning threshold.

9. The method for monitoring rural road traffic condition safety by multi-source data fusion as described in claim 8, characterized in that, Based on different road segment types, basic thresholds are set, specifically including: Rural roads are classified into various types, including curves, slopes, straight sections, and village entrance crossings. Based on historical traffic accident data, traffic flow patterns, and inherent risk levels for various road types, differentiated initial baseline values ​​for safety risk warnings are matched and set for different types of road sections: For special road sections with high accident risk and high probability of traffic conflicts, a first basic threshold is set; for straight road sections with relatively low risk, a second basic threshold is set, wherein the first basic threshold is lower than the second basic threshold. Based on real-time meteorological conditions, a threshold correction coefficient is set, specifically including: Establish a correlation mapping relationship between meteorological parameters and traffic safety risks; It can acquire various meteorological data that affect traffic safety in real time, including visibility, rainfall, snowfall, road surface temperature, road surface slipperiness and wind speed, and divide multiple meteorological risk level intervals according to the severity of different meteorological conditions; For each meteorological risk level range, a corresponding threshold correction coefficient is configured.

10. A rural road traffic condition safety monitoring system based on multi-source data fusion, characterized in that, include: The data acquisition module is configured to: filter key monitoring points on rural roads and acquire the raw monitoring data of the key monitoring points; The feature extraction module is configured to: extract traffic status features and safety risk features from the original monitoring data, and after feature filtering and fusion, obtain the core fused features; The integrated decision-making module is configured to: input the core fusion features into the fuzzy neural network model, perform integrated decision analysis on the core fusion features, and obtain the judgment results of traffic status and safety risks; The alarm execution module is configured to: adopt a dynamic early warning threshold self-adjustment mechanism based on road segment characteristics and real-time weather conditions to dynamically update the dynamic early warning threshold, and execute an alarm operation when the safety risk output by the fuzzy neural network model reaches the early warning condition.

Images