Traffic situation prediction method based on multi-source heterogeneous data fusion

Abstract

The invention relates to the field of traffic management, and discloses a traffic situation prediction method based on multi-source heterogeneous data fusion, which comprises the following steps of: firstly, acquiring traffic situation related data of a target area from a plurality of data sources, including traffic flow data, vehicle speed data, video image data, meteorological data and historical traffic statistical data; secondly, preprocessing the acquired traffic situation related data, including data cleaning, normalization or standardization processing, and performing time-space synchronization and matching; wherein the data cleaning comprises noise removal, abnormal value processing and missing value filling; and finally, inputting the preprocessed data into a pre-trained traffic situation prediction model, and outputting the predicted traffic jam degree of the target area. According to the invention, comprehensive analysis is carried out through the traffic-related situation data and the emergency data, and finally, the purpose of improving the prediction comprehensiveness through a multi-source cooperation mechanism is achieved.

Description

Technical Field

[0001] The present invention relates to the field of traffic management, and in particular to a traffic situation prediction method based on multi-source heterogeneous data fusion. Background Art

[0002] With the acceleration of urbanization and the surge in motor vehicle ownership, traffic congestion and frequent accidents are becoming increasingly serious, necessitating the urgent need for efficient traffic situation forecasting technology to support intelligent traffic management. Traditional traffic forecasting methods rely primarily on a single data source, employing statistical models or simple machine learning algorithms for short-term traffic flow prediction.

[0003] However, the above methods have obvious limitations: on the one hand, a single data source cannot fully reflect the complex traffic conditions, especially when abnormal events such as accidents and severe weather occur, the prediction accuracy drops sharply; on the other hand, the traffic system is affected by the coupling of multiple factors, including dynamic traffic flow, road topology, meteorological conditions and social media dissemination of emergencies. Traditional methods cannot effectively integrate the inherent connections between these heterogeneous data. Summary of the Invention

[0004] The purpose of the present invention is to provide a traffic situation prediction method based on multi-source heterogeneous data fusion to solve the above technical problems.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A traffic situation prediction method based on multi-source heterogeneous data fusion includes the following steps: S1. Collect traffic situation-related data of the target area from multiple data sources, including traffic flow data, vehicle speed data, video image data, meteorological data and historical traffic statistics; S2. Preprocess the collected traffic situation data, including data cleaning, normalization or standardization, and spatiotemporal synchronization and matching. Data cleaning includes noise removal, outlier processing, and missing value filling. S3. Input the pre-processed data into a pre-trained traffic situation prediction model and output the predicted traffic congestion level of the target area.

[0006] As a further technical solution, the time synchronization adopts interpolation or resampling methods, and the spatial matching adopts geographic coordinate alignment or grid processing; the video image data is extracted through a convolutional neural network to extract vehicle quantity, type and lane occupancy characteristics.

[0007] As a further technical solution, the traffic situation prediction model is constructed as follows: Quantifying the impact of meteorological data as visibility , rainfall intensity ; Analyze event data through the evaluation model to obtain the impact assessment value ; Traffic flow data, vehicle speed data, vehicle quantity, type and lane occupancy characteristics extracted from video image data are used as macro parameters and a sequence is constructed. ; The expression of the traffic situation prediction model is: ; in, is the predicted traffic congestion level in the target area, is the weight coefficient corresponding to each macro parameter, For the The reference value of a macro parameter is set based on the mean of historical data. For the The observed values ​​of the macro parameters, 、 is the weight factor corresponding to the meteorological parameter, , is the total number of macro parameters.

[0008] As a further technical solution, obtain the impact assessment value The process is: S31. Real-time collection of emergency data in the target area. Data sources include traffic control systems, social media sentiment, and roadside equipment reports. S32. Feature extraction is performed on the emergency event data to obtain the event type C, the topological distance D between the event location and the target area, the event duration T, and the propagation heat H. For example, the values ​​of different event types C are: traffic accident: C = 1.0, road construction: C = 0.8, large-scale event: C = 0.7, severe weather warning: C = 0.6, traffic control: C = 0.5, and other events: C = 0.3; S33. Calculate the impact assessment value through the assessment model: ; in, 、 、 is the preset weight coefficient and satisfies ; S43, the calculated impact assessment value Normalize to the interval [0,1] to obtain the final impact assessment value .

[0009] As a further technical solution, the normalization calculation method in S33 is: ;in, The value is 0. This is the highest valuation in history.

[0010] As a further technical solution, the method further includes: S4, receive emergency data in real time and predict the traffic congestion level in the target area Compare and analyze with the actual traffic status: If the relative deviation is within the threshold range, visualization information is generated; If the relative deviation exceeds the threshold and is associated with an emergency, the model dynamic correction module will be triggered to recalculate the predicted traffic congestion level in the target area based on the event characteristics; and issue a revised traffic situation warning and emergency diversion instructions.

[0011] As a further technical solution, the method for calculating the relative deviation between the predicted traffic congestion level and the actual observed value is as follows: ; is the actual observation value calculated from real-time traffic sensor data, ; in, For the Real-time traffic flow of each monitoring section, is the cross-sectional capacity, For the The average vehicle speed of each monitoring section, Design a maximum speed limit for the road, is the total number of monitoring sections.

[0012] As a further technical solution, if , threshold The default value is 0.15, and the following judgment is performed: Compare the input data missing rate with the data missing warning value. If it exceeds the value, it is marked as a data source anomaly and the data repair module is activated to use the pre-trained long short-term memory network (LSTM) neural network to fill in the missing values. For example, the formula for calculating the input data missing rate is: , is the input data missing rate like , it is recorded as data source abnormality; among them, is the amount of missing data, that is, the number of samples of traffic parameters (such as flow, speed, etc.) that are not collected normally or are invalid in the target area. is the total data volume, which is the total number of theoretical samples of traffic parameters that should be collected in the target area during the statistical period. The default value is 0.1. Compare the absolute value of the Pearson correlation coefficient between the impact intensity of the emergency and the deviation with the relevant warning value. If it exceeds the value, it is determined to be an event-dominated deviation and the model dynamic correction module is triggered. For example, the calculation method of the Pearson correlation coefficient is: ,like , it is determined to be an event-dominated deviation; among them, Indicates the impact assessment value of the emergency Relative deviation The covariance of Indicates the impact assessment value of the emergency The standard deviation of Relative deviation The standard deviation of The default is 0.7.

[0013] As a further technical solution, the method for providing revised traffic situation warning and emergency diversion instructions is as follows: According to the revised To perform a graded response: when ≥ the first warning value, conduct a first-level response: Issue a red alert and forcibly divert vehicles from surrounding roads; The linked traffic light system uses congestion relief mode, which means the cycle is extended by 30%; Push path avoidance to the navigation platform.

[0014] When the second warning value ≤ When the value is less than the first warning value, the second level response is carried out: Detour routes are provided through variable information boards.

[0015] Beneficial effects of the present invention: (1) The present invention constructs a global perception framework by deeply integrating five types of heterogeneous data: traffic, speed, video, meteorological, and historical statistics. The video data is extracted through a convolutional neural network (CNN) to extract lane features, vehicle types, and occupancy rates. Meteorological data is quantified into visibility and rainfall intensity influencing factors, and historical data provides macro-benchmark parameters. The multi-source collaborative mechanism significantly improves the comprehensiveness of the prediction, especially in emergency scenarios such as accidents and severe weather. It can simultaneously capture road condition images, social media public opinion, and meteorological changes, avoiding the prediction failure of traditional models due to information loss. (2) The present invention constructs multi-dimensional features based on event type, distance, duration and social media dissemination heat, and then quantifies the impact intensity through an exponential evaluation model. The model introduces a logarithmic function of dissemination heat to objectively reflect the amplifying effect of public opinion diffusion on traffic congestion; at the same time, a normalization mechanism is designed to ensure the comparability of evaluation values; and by comparing the deviation between the predicted value and the actual observation in real time, the event correlation analysis is automatically triggered, such as the Pearson correlation coefficient test. Once the event is determined to be the dominant deviation, the model correction module is immediately started to recalculate the congestion level. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 A diagram showing the steps of the method of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0019] See also Figure 1 As shown, the present invention is a traffic situation prediction method based on multi-source heterogeneous data fusion, comprising the following steps: S1. Collect traffic situation-related data of the target area from multiple data sources, including traffic flow data, vehicle speed data, video image data, meteorological data and historical traffic statistics; S2. Preprocessing the collected traffic situation-related data, including data cleaning, normalization or standardization, and spatiotemporal synchronization and matching; data cleaning includes noise removal, outlier processing, and missing value filling; the preprocessing methods used in step S2 are all mature methods in the prior art and will not be elaborated on here; S3. Input the pre-processed data into a pre-trained traffic situation prediction model and output the predicted traffic congestion level of the target area.

[0020] The time synchronization adopts interpolation or resampling methods, and the spatial matching adopts geographic coordinate alignment or grid processing; the video image data is extracted through a convolutional neural network to extract vehicle quantity, type and lane occupancy rate characteristics.

[0021] The traffic situation prediction model is constructed as follows: Quantifying the impact of meteorological data as visibility , rainfall intensity ; Analyze event data through the evaluation model to obtain the impact assessment value ; Traffic flow data, vehicle speed data, vehicle quantity, type and lane occupancy characteristics extracted from video image data are used as macro parameters and a sequence is constructed. ; The expression of the traffic situation prediction model is: ; in, is the predicted traffic congestion level in the target area, is the weight coefficient corresponding to each macro parameter, For the The reference value of a macro parameter is set based on the mean of historical data. For the The observed values ​​of the macro parameters, 、 is the weight factor corresponding to the meteorological parameter, , is the total number of macro parameters.

[0022] In this embodiment, in order to overcome the problem that traditional methods rely on isolated data sources such as only traffic or speed, which are difficult to capture the multi-dimensional status of complex traffic systems; by integrating five types of heterogeneous data such as traffic, speed, video, meteorological and historical statistics, a collaborative perception network is formed, and a convolutional neural network is used to extract lane characteristics, including vehicle type, number and lane occupancy status, to make up for the insufficient granularity of macro traffic data; visibility and rainfall intensity are converted into calculable influencing factors to objectively reflect the implicit constraints of the environment on road conditions; through geographic coordinate gridding and time interpolation technology, the spatiotemporal deviations of multi-source data collection are eliminated to ensure that heterogeneous information is coupled under a unified benchmark; and for sudden severe weather or local accident scenes, video image anomalies, meteorological mutations and historical patterns can be simultaneously integrated, significantly improving the completeness of situational awareness in complex scenes.

[0023] pass The impact assessment value of the emergency As a global multiplier factor, it objectively reflects the nonlinear amplification effect of accidents, control and other events on traffic congestion; the meteorological parameters (visibility and rainfall intensity ) is converted into the environmental constraint adjustment term in the denominator, and the weight factor , Dynamically balance the implicit attenuation effect of meteorological conditions on road capacity; based on historical benchmark values For macro parameter sequence Standardization is carried out through The system quantifies the degree of deviation of real-time traffic flow status from the historical norm, and ultimately constructs a physical driving architecture of event disturbance × (traffic flow dynamics / environmental constraints). This allows the prediction results to simultaneously bear the triple influence of sudden interference, environmental constraints, and the inherent laws of traffic flow, breaking through the limitation of traditional models that simplify complex coupling effects into linear superposition.

[0024] Obtain the impact assessment value The process is: S31. Real-time collection of emergency data in the target area. Data sources include traffic control systems, social media sentiment, and roadside equipment reports. S32. Feature extraction is performed on the emergency event data to obtain the event type C, the topological distance D between the event location and the target area, the event duration T, and the propagation heat H. For example, the values ​​of different event types C are: traffic accident: C = 1.0, road construction: C = 0.8, large-scale event: C = 0.7, severe weather warning: C = 0.6, traffic control: C = 0.5, and other events: C = 0.3; S33. Calculate the impact assessment value through the assessment model: ; in, 、 、 is the preset weight coefficient and satisfies ; S43, the calculated impact assessment value Normalize to the interval [0,1] to obtain the final impact assessment value .

[0025] By constructing a four-dimensional feature extraction framework of "event type-topological distance-duration-propagation heat" and designing an exponential evaluation model It integrates the inherent attributes of comprehensive events (type / distance / duration) and the dynamic diffusion of public opinion (dissemination heat), objectively reflects the amplifying effect of social media on traffic congestion, and realizes multi-dimensional coupling; the inverse of the topological distance reflects the spatial attenuation law, and the logarithm of the heat quantifies the marginal effect of information dissemination; the evaluation value is compressed to the standard range, eliminating the dimensional differences of different events and providing comparable input for the model.

[0026] The normalization calculation method in S33 is: ;in, The value is 0. This is the highest historical valuation. A linear normalization strategy uses historical maximum / minimum evaluation values ​​as a scale to avoid model output distortion caused by extreme events; the normalization interval is automatically expanded as the event library accumulates to ensure evaluation compatibility for emerging event types (such as large-scale events).

[0027] The method further comprises: S4, receive emergency data in real time and predict the traffic congestion level in the target area Compare and analyze with the actual traffic status: If the relative deviation is within the threshold range, visualization information is generated; If the relative deviation exceeds the threshold and is associated with an emergency, the model dynamic correction module will be triggered to recalculate the predicted traffic congestion level in the target area based on the event characteristics; and issue a revised traffic situation warning and emergency diversion instructions.

[0028] By comparing the predicted values ​​in real time It is associated with the actual traffic status and emergency event analysis to form: abnormal diagnosis trigger: when the deviation exceeds the threshold, it automatically distinguishes between data anomalies and event interference and starts the corresponding correction module; dynamic model update: event-related deviation triggers instant recalculation, avoiding the failure delay of traditional static models in emergency scenarios.

[0029] The method for calculating the relative deviation between the predicted traffic congestion level and the actual observed value is: ; is the actual observation value calculated from real-time traffic sensor data, ; in, For the Real-time traffic flow of each monitoring section, is the cross-sectional capacity, For the The average vehicle speed of each monitoring section, Design a maximum speed limit for the road, is the total number of monitoring sections.

[0030] based on The actual observation value formula realizes multi-parameter fusion characterization: integrated flow saturation rate and velocity decay rate Eliminate the one-sidedness of a single indicator; through the calculation of the mean value of the monitoring section, adapt to the heterogeneous topological structure of the urban road network and achieve spatial generalization capabilities.

[0031] like , threshold The default value is 0.15, and the following judgment is performed: Compare the input data missing rate with the data missing warning value. If it exceeds the value, it is marked as a data source anomaly and the data repair module is activated to use the pre-trained long short-term memory network (LSTM) neural network to fill in the missing values. For example, the formula for calculating the input data missing rate is: , is the input data missing rate like , it is recorded as data source abnormality; among them, is the amount of missing data, that is, the number of samples of traffic parameters (such as flow, speed, etc.) that are not collected normally or are invalid in the target area. is the total data volume, which is the total number of theoretical samples of traffic parameters that should be collected in the target area during the statistical period. The default value is 0.1. Compare the absolute value of the Pearson correlation coefficient between the impact intensity of the emergency and the deviation with the relevant warning value. If it exceeds the value, it is determined to be an event-dominated deviation and the model dynamic correction module is triggered. For example, the calculation method of the Pearson correlation coefficient is: ,like , it is determined to be an event-dominated deviation; among them, Indicates the impact assessment value of the emergency Relative deviation The covariance of Indicates the impact assessment value of the emergency The standard deviation of Relative deviation The standard deviation of The default is 0.7.

[0032] A dual diagnostic path is constructed through data missing rate detection and Pearson correlation coefficient analysis: when the missing rate exceeds the limit, the data source is automatically marked as abnormal, and LSTM repair is initiated to ensure input integrity and realize data quality monitoring; when the correlation coefficient exceeds the threshold, the event-dominant deviation is determined, and the model correction is accurately triggered instead of blind adjustment to realize event attribution decision.

[0033] The method for issuing revised traffic situation warning and emergency diversion instructions is as follows: According to the revised To perform a graded response: when ≥ the first warning value, conduct a first-level response: Issue a red alert and forcibly divert vehicles from surrounding roads; The linked traffic light system uses congestion relief mode, which means the cycle is extended by 30%; Push path avoidance to the navigation platform.

[0034] When the second warning value ≤ When the value is less than the first warning value, the second level response is carried out: Detour routes are provided through variable information boards.

[0035] In this embodiment, according to the modified Implement graded response, with the first-level red alert forcing diversion and extending the cycle of linked traffic lights. The predicted values ​​are directly converted into executable instructions such as signal control and route guidance to achieve a closed-loop management: match the response level according to the degree of congestion to avoid secondary congestion caused by excessive control.

[0036] It should be noted that the calculation formulas and various parameters involved in the calculations in the present invention have been dimensionally processed in advance, and the process of dimensionless processing is well known in the industry and will not be described here.

[0037] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A traffic situation prediction method based on multi-source heterogeneous data fusion, characterized in that: The following steps are involved: S1. Collect traffic situation-related data of the target area from multiple data sources, including traffic flow data, vehicle speed data, video image data, meteorological data and historical traffic statistics; S2. Preprocess the collected traffic situation data, including data cleaning, normalization or standardization, and spatiotemporal synchronization and matching. Data cleaning includes noise removal, outlier processing, and missing value filling. S3. Input the pre-processed data into a pre-trained traffic situation prediction model and output the predicted traffic congestion level of the target area.

2. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The time synchronization adopts interpolation or resampling methods, and the spatial matching adopts geographic coordinate alignment or grid processing; the video image data is extracted through a convolutional neural network to extract vehicle quantity, type and lane occupancy rate characteristics.

3. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 2 is characterized in that: The traffic situation prediction model is constructed as follows: Quantifying the impact of meteorological data as visibility , rainfall intensity ; Analyze event data through the evaluation model to obtain the impact assessment value ; Traffic flow data, vehicle speed data, vehicle quantity, type and lane occupancy characteristics extracted from video image data are used as macro parameters and a sequence is constructed. ; The expression of the traffic situation prediction model is: ； in, is the predicted traffic congestion level in the target area, is the weight coefficient corresponding to each macro parameter, For the Reference values ​​of macro parameters, For the The observed values ​​of the macro parameters, 、 is the weight factor corresponding to the meteorological parameter, , is the total number of macro parameters.

4. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 3 is characterized in that: Obtain the impact assessment value The process is: S31. Real-time collection of emergency data in the target area. Data sources include traffic control systems, social media sentiment, and roadside equipment reports. S32. Extract features from the emergency event data to obtain the event type C, the topological distance D between the event location and the target area, the event duration T, and the spread heat H; S33. Calculate the impact assessment value through the assessment model: ； in, 、 、 is the preset weight coefficient and satisfies ; S43, the calculated impact assessment value Normalize to the interval [0,1] to obtain the final impact assessment value .

5. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 4 is characterized in that: The normalization calculation method in S33 is: ;in, The value is 0. This is the highest valuation in history.

6. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 1 is characterized in that: The method further comprises: S4, receive emergency data in real time and predict the traffic congestion level in the target area Compare and analyze with the actual traffic status: If the relative deviation is within the threshold range, visualization information is generated; If the relative deviation exceeds the threshold and is associated with an emergency, the model dynamic correction module will be triggered to recalculate the predicted traffic congestion level in the target area based on the event characteristics; and issue a revised traffic situation warning and emergency diversion instructions.

7. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 6 is characterized in that: The method for calculating the relative deviation between the predicted traffic congestion level and the actual observed value is: ； is the actual observation value calculated from real-time traffic sensor data, ； in, For the Real-time traffic flow of each monitoring section, is the cross-sectional capacity, For the The average vehicle speed of each monitoring section, Design a maximum speed limit for the road, is the total number of monitoring sections.

8. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 7 is characterized in that: like , then perform the following judgment: Compare the input data missing rate with the data missing warning value. If it exceeds the value, it is marked as a data source anomaly and the data repair module is activated to use the pre-trained long short-term memory network (LSTM) neural network to fill in the missing values. For example, the formula for calculating the input data missing rate is: , is the input data missing rate like , it is recorded as data source abnormality; among them, is the amount of missing data, is the total data volume, The default value is 0.

1. Compare the absolute value of the Pearson correlation coefficient between the impact intensity of the emergency and the deviation with the relevant warning value. If it exceeds the value, it is determined to be an event-dominated deviation and the model dynamic correction module is triggered. For example, the calculation method of the Pearson correlation coefficient is: ,like , it is determined to be an event-dominated deviation; among them, Indicates the impact assessment value of the emergency Relative deviation The covariance of Indicates the impact assessment value of the emergency The standard deviation of Relative deviation The standard deviation of The default is 0.

7.

9. The traffic situation prediction method based on multi-source heterogeneous data fusion according to claim 6 is characterized in that: The method for issuing revised traffic situation warning and emergency diversion instructions is as follows: According to the revised To perform a graded response: when ≥ the first warning value, conduct a first-level response: Issue a red alert and forcibly divert vehicles from surrounding roads; The linked traffic light system activates congestion relief mode; Push path avoidance to the navigation platform. When the second warning value ≤ When the value is less than the first warning value, the second level response is carried out: Detour routes are provided through variable information boards.

Images