A Vehicle State Acquisition Method Based on Multi-Source Data Fusion from a Roadside Perspective

Abstract

The present invention discloses a method for collecting vehicle states based on multi-source data fusion from a roadside perspective. The present invention relates to the technical field of data analysis and includes the following steps: deploying various collection devices on the roadside to collect the original data of vehicle states and environmental conditions, and preprocessing the collected original data; performing time synchronization operations on the original data collected by different sensors after preprocessing. The method for collecting vehicle states based on multi-source data fusion from a roadside perspective analyzes the conditions of sensor data under different environmental conditions, analyzes the degree of interference with the collection of vehicle state data, and conducts quantitative analysis to implement corresponding data adjustment measures, thereby improving the accuracy of subsequent traffic management. By integrating data from different sources, real-time monitoring of vehicle states and dynamic traffic management are achieved, thus providing a scientific basis for traffic management departments to formulate targeted preventive measures.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and particularly to a method for collecting vehicle states based on multi-source data fusion from a roadside perspective. Background Art

[0002] With the rapid development of intelligent transportation systems, the accurate and real-time collection of vehicle states has become increasingly important. Intelligent transportation systems need to use vehicle state information for traffic flow management, traffic incident detection, traffic signal control, etc., to improve road traffic capacity and traffic safety levels. Therefore, the development of efficient and accurate vehicle state collection methods has become the key to the development of intelligent transportation systems.

[0003] In the prior art, multi-source data fusion involves a variety of collection devices and data sources. However, in different traffic environments and weather conditions, the performance and data quality of sensors will change, which will in turn affect the accuracy of collecting vehicle state data. Therefore, how to analyze the fusion state during multi-source data fusion, estimate data quality, and weaken the impact on the collection of vehicle state data is the problem we need to solve. For this reason, a method for collecting vehicle states based on multi-source data fusion from a roadside perspective is proposed. Summary of the Invention

[0004] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for collecting vehicle states based on multi-source data fusion from a roadside perspective, comprising the following steps:

[0005] Step 1: Deploy a variety of collection devices on the roadside to collect the original data of vehicle states and environmental conditions, and preprocess the collected original data;

[0006] Step 2: Perform time synchronization operations on the original data collected by different sensors after preprocessing, and use data fusion technology to fuse the original data of different sensors;

[0007] Step 3: Monitor data anomalies in sensors and data fusion results for changes in different environmental conditions, analyze the severity of the anomalies, and match corresponding recovery and adjustment measures;

[0008] Step 4: Verify and analyze the quality of the fused data after adjustment and recovery, control data errors, and adjust the application strategy of the data to meet traffic rule requirements;

[0009] Step 5: Use the processed data for vehicle state monitoring to implement targeted traffic management.

[0010] Preferably, in the above Step 1, the process of collecting and preprocessing the original data includes:

[0011] Deploy a variety of collection devices on the roadside to ensure coverage of the required traffic areas, including cameras, radars, lidars, and infrared sensors, for collecting raw data on vehicle status and environmental conditions;

[0012] Use the deployed variety of collection devices to collect vehicle status data such as the driving speed, position, and driving direction of the vehicle, and collect environmental condition data such as light intensity, weather conditions, temperature, humidity, and wind speed;

[0013] Transmit the collected raw data wirelessly to the data processing center and perform preprocessing operations. The preprocessing includes data cleaning, format conversion, and data calibration steps. Among them, data cleaning removes noise, duplicate values, and invalid values in the data to ensure the accuracy and integrity of the data. Format conversion converts the data collected by different sensors into a unified format for convenient subsequent processing and analysis. Data calibration calibrates the data collected by the sensors to eliminate the errors and biases of the sensors themselves;

[0014] Build a data warehouse and integrate the preprocessed data and store it in the data warehouse to ensure the accessibility and security of the data.

[0015] Preferably, in the second step, the time synchronization and fusion process of the raw data includes:

[0016] Retrieve the raw data collected by different sensors after preprocessing from the data warehouse, perform a time synchronization operation of software synchronization on it, traverse each data record, record the timestamps of each sensor data, use the lidar data as the benchmark for time synchronization, and extract its timestamp as the reference time point;

[0017] For the data of other sensors (such as cameras, radars, infrared sensors, etc.), compare its timestamp with the reference time point of the lidar, use the interpolation algorithm to calculate the data value of other sensors at the lidar reference time point, and through interpolation, obtain the data of all sensors at the lidar reference time point to achieve time synchronization from the software;

[0018] Perform time alignment and coordinate alignment on the data obtained by different sensors. Among them, time alignment is performed based on the results of the time synchronization operation to ensure the consistency of each sensor data in time, and the data of different sensors are converted into a unified coordinate system for coordinate alignment for subsequent data fusion;

[0019] A data fusion algorithm using the Kalman filtering method, defines a state equation to describe the change of sensor data over time, and defines an observation equation to describe the relationship between sensor data and the state, initializes the state estimate and covariance matrix. For each time step, uses the state equation to predict the next state, uses the observation equation and the observed data to update the state estimate and covariance matrix, repeats the Kalman filtering process until all sensor data are fused, and then obtains the fused data, integrating information from different sensors to improve the accuracy and reliability of the data;

[0020] Save the fused data back to the data warehouse for further analysis and processing.

[0021] Preferably, in the third step, the analysis process of the severity of abnormal situations includes:

[0022] Based on the fused data, analyze the data of various types of environmental conditions and monitor the changes in different environmental conditions;

[0023] Using time series analysis methods, comprehensively analyze the monitored data to identify abnormal data, and the abnormal data is manifested as sudden changes in data values and abnormal expansions in the data fluctuation range;

[0024] Integrate the abnormal data, determine the characteristics of the abnormal data, which are the number of abnormal values, the duration of the anomaly, and the fluctuation range of the abnormal data, calculate the abnormal interference index, and then quantitatively analyze the severity of the abnormal data;

[0025] According to the severity of the abnormal data, divide different interference levels, namely the minor interference level, the medium interference level, and the severe interference level. Combine the calculated abnormal interference index to match corresponding interference thresholds for each interference level, and match corresponding recovery adjustment measures for each interference level. The minor interference level corresponds to minor abnormal situations, and simple recovery adjustment measures such as adjusting the calibration parameters of the sensor and optimizing the data acquisition frequency are taken. At the same time, strengthen the monitoring and closely pay attention to the change trend of the abnormal data. The medium interference level corresponds to medium abnormal situations, and it is necessary to deeply analyze the abnormal reasons involving factors such as sensor failures, data transmission errors, or environmental changes. According to the analysis results, take corresponding recovery adjustment measures such as replacing faulty sensors, repairing data transmission links, and adjusting monitoring strategies. During the recovery process, the severe interference level corresponds to severe abnormal situations, and it is necessary to immediately activate the emergency plan, take emergency recovery adjustment measures such as shutting down faulty equipment and starting the standby system, deeply analyze to find out the abnormal reasons, formulate a long-term recovery plan, and strengthen safety monitoring during the recovery process.

[0026] Preferably, the process of identifying abnormal data includes:

[0027] Using the time series analysis method, decompose the time series of the monitoring data into three parts: trend, seasonality, and residuals. According to the performance state of the abnormal data, such as sudden changes in data values and abnormal expansion of the data fluctuation range, combine with the three-sigma principle based on the normal distribution to set the abnormal detection threshold T;

[0028] Analyze the residual part of the time series decomposition of the monitoring data, calculate the residuals at each time point. For each time point t, calculate the median absolute deviation of the residuals at the w time points before and after it to measure the volatility of the data;

[0029] Calculate the absolute value of the difference between the actual observed value at time point t and the reference value at time point t, and combine it with the median absolute deviation to construct an abnormal detection index;

[0030] Calculate the abnormal detection index for each time point, and compare it with the set abnormal detection threshold. If the abnormal detection index is greater than the abnormal detection threshold T, the data at time point t is abnormal data.

[0031] Preferably, the calculation expression of the abnormal detection index is:

[0032]

[0033] Wherein, is the abnormal detection index, is the actual observed value at time point t, is the reference value at time point t, which is used to reflect the normal state of the data, is the median absolute deviation of the residuals at the w time points before and after time point t, which is used to measure the volatility of the data, are the residuals at the w time points before and after time point t.

[0034] Preferably, the process of obtaining the abnormal interference index includes:

[0035] Integrate the abnormal data within a fixed time period from different sensors, including speed, position, driving direction, light intensity, weather conditions, temperature, humidity, and wind speed, to form an abnormal data set. The abnormal data set contains the time stamps and abnormal value information of each abnormal data point;

[0036] Analyze the abnormal data set to determine the abnormal data characteristics such as the number of abnormal values, the duration of the abnormality, and the fluctuation range of the abnormal data, and evaluate the impact of the abnormal data on vehicle state recognition, including the decrease in recognition accuracy and the increase in false alarm rate;

[0037] Based on the evaluation results of the impact of the abnormal data on vehicle state recognition, comprehensively analyze the abnormal data characteristics to obtain the abnormal interference index, and quantify the severity of the interference of the abnormal data on vehicle state recognition.

[0038] Preferably, the calculation expression of the abnormal interference index is as follows:

[0039]

[0040] where is the abnormal interference index, is the total number of abnormal values within a fixed time period, is the total time length during which the abnormal values persist, is the maximum deviation degree of the abnormal value from its reference value, is the number of baseline abnormal values, representing the number of abnormal values expected under normal conditions, is an adjustment parameter used to control the influence degree of the number of abnormal values on the index, is the maximum value of the abnormal duration during the observation period, is the maximum value of the abnormal fluctuation range during the observation period;

[0041] The three interference levels correspond to three interference thresholds, where the interference thresholds include an upper threshold and a lower threshold;

[0042] The three interference levels and the three interference thresholds satisfy the following relationship:

[0043] Slight interference level ;

[0044] Medium interference level ;

[0045] Severe interference level ;

[0046] where is the abnormal interference index, is the lower threshold corresponding to the medium interference level and the upper threshold corresponding to the slight interference level, is the lower threshold corresponding to the severe interference level and the upper threshold corresponding to the medium interference level.

[0047] Preferably, in step four, the process of adjusting the data application strategy corresponding to the traffic rule requirements includes:

[0048] Verify the restored fusion data, and check the integrity, accuracy, consistency, and timeliness of the data. The integrity check ensures that all necessary data fields are fully populated without missing values, and checks whether the number of data records matches the expectation to ensure no data loss. The accuracy check verifies the accuracy of the data by comparing with the data source or using known correct values, and uses statistical methods to evaluate the reasonableness of the data. The consistency check ensures that the data between different data sources is logically consistent, and checks whether the data follows the expected format and units. The timeliness check verifies the timeliness of the data to ensure that the data is up-to-date and relevant to the current traffic rule requirements, and analyzes the data quality to identify outliers in the data;

[0049] According to the results of the quality analysis, identify the error sources and perform error correction to control data errors. Among them, analyze the links in the data collection, processing, transmission, and storage processes that may introduce errors, identify common error sources such as sensor failures, data loss, or data conversion errors, use data smoothing techniques to correct random errors in the data, and for systematic errors, adjust the data collection and processing methods, regularly evaluate the changing trend of data errors, and take corresponding corrective measures;

[0050] Analyze traffic rule requirements, including speed limits, traffic signal control, and lane allocation, adjust the precision requirements of the corresponding data according to traffic rule requirements, synchronously adjust the data application strategy, implement and monitor the strategy effect, corresponding to traffic rule requirements.

[0051] Preferably, in step five, the process of implementing targeted traffic management includes:

[0052] Using real-time data processing technology, perform real-time analysis on the collected vehicle status data, monitor the basic information of the vehicle's speed, position, and driving direction. According to the real-time traffic flow and vehicle status data, dynamically adjust the control strategy of traffic lights, and improve traffic mobility and reduce traffic congestion through the intelligent signal light system;

[0053] In case of traffic accidents or road congestion, use vehicle status data to quickly locate the problem area, and through traffic guidance and diversion strategies, provide drivers with real-time traffic guidance and information, guiding vehicles to avoid congested sections and balancing the road network load;

[0054] Using the analysis results of cameras and vehicle status data, conduct real-time monitoring and recording of traffic violations, and upload the violation data to the traffic management department for subsequent punishment and education work;

[0055] Regularly analyze and evaluate the vehicle status monitoring and traffic management effects. Using data visualization technology, display the changing trends of key indicators such as traffic flow, congestion conditions, and violation behaviors, and adjust and optimize traffic management strategies according to the analysis results.

[0056] The present invention provides a method for collecting vehicle status based on multi-source data fusion from a roadside perspective. It has the following beneficial effects:

[0057] 1. For the method for collecting vehicle status based on multi-source data fusion from a roadside perspective, by analyzing the status of sensor data under different environmental conditions and analyzing the interference degree on the collection of vehicle status data, conduct quantitative analysis to implement corresponding data adjustment measures, thereby improving the accuracy of subsequent traffic management. By integrating data from different sources, realize real-time monitoring of vehicle status and dynamic traffic management, thus providing a scientific basis for traffic management departments to formulate targeted preventive measures.

[0058] 2. For the method for collecting vehicle status based on multi-source data fusion from a roadside perspective, by integrating data from different sensors, obtain basic information such as the speed, position, and driving direction of vehicles in real time and accurately, greatly improving the efficiency and accuracy of traffic management. During the traffic peak period, the system can quickly identify and handle traffic bottlenecks, and by dynamically adjusting the signal control strategy, effectively relieve traffic congestion and improve road capacity. At the same time, accurate vehicle status monitoring can also provide more reliable data support for traffic planning and decision-making, helping to continuously optimize the urban traffic system. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 is the method flow chart of a method for collecting vehicle status based on multi-source data fusion from a roadside perspective of the present invention;

[0060] Figure 2 is the analysis flow chart of the severity of abnormal situations of the present invention;

[0061] Figure 3 is the acquisition flow chart of the abnormal interference index of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0062] The present invention will be further described in detail below in conjunction with the drawings and specific embodiments. The embodiments of the present invention are given for the purpose of illustration and description, and are not exhaustive or limit the present invention to the disclosed form. Many modifications and variations are obvious to those of ordinary skill in the art. The embodiments are selected and described to better illustrate the principles and practical applications of the present invention, and enable those of ordinary skill in the art to understand the present invention and thus design various embodiments with various modifications suitable for specific purposes.

[0063] The first embodiment, asFigure 1 , Figure 2 As shown in Figure 2 , the present invention provides a technical solution: a method for collecting vehicle states based on multi-source data fusion from a roadside perspective, comprising the following steps:

[0064] Step 1: Deploy collection devices on the roadside to collect the original data of vehicle states and environmental conditions, and preprocess the collected original data. Deploy collection devices on the roadside to ensure that the required traffic areas can be covered, including cameras, radars, lidars, and infrared sensors, for collecting the original data of vehicle states and environmental conditions. Use the deployed cameras, radars, lidars, and infrared sensors to collect vehicle state data such as the driving speed, position, and driving direction of the vehicle, and collect environmental condition data such as light intensity, weather conditions, temperature, humidity, and wind speed. Transmit the collected original data to the data processing center wirelessly and perform preprocessing operations. The preprocessing includes data cleaning, format conversion, and data calibration steps. Among them, data cleaning removes noise, duplicate values, and invalid values in the data to ensure the accuracy and integrity of the data. Format conversion converts the data collected by different sensors into a unified format for convenient subsequent processing and analysis. Data calibration calibrates the data collected by the sensors to eliminate the errors and biases of the sensors themselves. Build a data warehouse and integrate the preprocessed data and store it in the data warehouse to ensure the accessibility and security of the data;

[0065] Step 2: Perform time synchronization operations on the original data collected by different sensors after preprocessing, and use data fusion technology to fuse the original data of different sensors. Retrieve the original data collected by different sensors after preprocessing from the data warehouse, perform a time synchronization operation of software synchronization on it, traverse each data record, record the timestamps of the data of each sensor, use the lidar data as the benchmark for time synchronization, extract its timestamp as the reference time point, and for the data of other sensors (such as cameras, radars, infrared sensors, etc.), compare their timestamps with the reference time point of the lidar, use the interpolation algorithm to calculate the data values of other sensors at the lidar reference time point, and through interpolation, obtain the data of all sensors at the lidar reference time point, achieve time synchronization from the software, align the time and coordinates of the data obtained by different sensors. Among them, perform time alignment based on the results of the time synchronization operation to ensure the consistency of the data of each sensor in time, and convert the data of different sensors into a unified coordinate system for coordinate alignment in order to perform subsequent data fusion. Use the data fusion algorithm of the Kalman filter method, define the state equation to describe the change of sensor data over time, and define the observation equation to describe the relationship between sensor data and the state. Initialize the state estimate and covariance matrix. For each time step, use the state equation to predict the next state, use the observation equation and observation data to update the state estimate and covariance matrix, repeat the Kalman filter process until all sensor data are fused, and then obtain the fused data, synthesize the information from different sensors, improve the accuracy and reliability of the data, store the fused data back in the data warehouse for further analysis and processing;

[0066] Step 3: For changes in different environmental conditions, monitor data anomalies in the sensor and data fusion results, analyze the severity of the anomalies, match corresponding recovery and adjustment measures. Based on the fused data, analyze environmental condition data of various types, monitor changes in different environmental conditions, use time series analysis methods to comprehensively analyze the monitored data, identify abnormal data. Abnormal data is manifested as sudden changes in data values and abnormal expansion of the data fluctuation range. Integrate the abnormal data, determine the characteristics of the abnormal data, namely the number of outliers, the duration of the anomaly, and the fluctuation range of the abnormal data. Analyze the interference of the abnormal data on vehicle state recognition, calculate the abnormal interference index, and then quantitatively analyze the severity of the abnormal data. According to the severity of the abnormal data, divide different interference levels, namely the minor interference level, the medium interference level, and the severe interference level. Combine the calculated abnormal interference index, match corresponding interference thresholds for each interference level, and match corresponding recovery and adjustment measures for each interference level. The minor interference level corresponds to minor abnormal conditions, and simple recovery and adjustment measures such as adjusting the calibration parameters of the sensor and optimizing the data acquisition frequency are taken. At the same time, strengthen the monitoring and closely pay attention to the change trend of the abnormal data. The medium interference level corresponds to medium abnormal conditions, and it is necessary to deeply analyze the abnormal reasons involving factors such as sensor failures, data transmission errors, or environmental changes. According to the analysis results, take corresponding recovery and adjustment measures such as replacing faulty sensors, repairing the data transmission link, and adjusting the monitoring strategy. During the recovery process, the severe interference level corresponds to severe abnormal conditions, and it is necessary to immediately activate the emergency plan and take emergency recovery and adjustment measures such as shutting down faulty equipment and starting the standby system. Deeply analyze, find out the abnormal reasons, and formulate a long-term recovery plan. During the recovery process, strengthen safety monitoring;

[0067] Further, the process of abnormal data identification includes:

[0068] Using the time series analysis method, decompose the time series of the monitored data into three parts: trend, seasonality, and residuals. According to the manifestation states of sudden changes in the data values of the abnormal data and abnormal expansion of the data fluctuation range, combined with the three-sigma principle based on the normal distribution to set the abnormal detection threshold T, analyze the residual part of the time series decomposition of the monitored data, calculate the residual of each time point. For each time point t, calculate the median absolute deviation of the residuals of the w time points before and after it to measure the volatility of the data. Calculate the absolute value of the difference between the actual observed value at time point t and the reference value at time point t, and combine it with the median absolute deviation to construct the abnormal detection index. Calculate the abnormal detection index of each time point, and compare it with the set abnormal detection threshold. If the abnormal detection index is greater than the abnormal detection threshold T, the data at time point t is abnormal data;

[0069] Even further, the calculation expression of the abnormal detection index is:

[0070]

[0071] Among them, is the anomaly detection index, is the actual observed value at time point t, is the reference value at time point t, which is used to reflect the normal state of the data, is the median absolute deviation of the residuals of w time points before and after time point t, which is used to measure the volatility of the data, is the residual of w time points before and after time point t. The residual is the result of subtracting the reference value from the original data;

[0072] Step Four: Verify and perform quality analysis on the adjusted and restored fusion data, control data errors, and adjust the application strategy of the data to meet traffic rule requirements;

[0073] Step Five: Use the processed data for vehicle status monitoring to implement targeted traffic management.

[0074] Second Embodiment, based on the First Embodiment, please refer to Figure 3 As shown, the process of obtaining the anomaly interference index includes:

[0075] Integrate the anomaly data within a fixed time period from different sensors, including speed, position, driving direction, light intensity, weather conditions, temperature, humidity, and wind speed, to form an anomaly data set. The anomaly data set contains the time stamps and anomaly value information of each anomaly data point. Analyze the anomaly data set to determine the anomaly data characteristics such as the number of anomaly values, the duration of anomalies, and the fluctuation range of anomaly data. Evaluate the impact of the anomaly data on vehicle status recognition, including the decrease in recognition accuracy and the increase in false alarm rate. Based on the evaluation results of the impact of the anomaly data on vehicle status recognition, comprehensively analyze and obtain the anomaly interference index by combining the anomaly data characteristics to quantify the severity of the interference of the anomaly data on vehicle status recognition;

[0076] Furthermore, the calculation expression of the anomaly interference index is:

[0077]

[0078] Among them, is the anomaly interference index, is the total number of anomaly values within a fixed time period, is the total time length of the duration of the anomaly values, is the maximum deviation degree of the anomaly value from its reference value, is the baseline number of anomaly values, representing the number of anomaly values expected under normal conditions, is an adjustment parameter used to control the influence degree of the number of anomaly values on the index, is the maximum value of the abnormal duration during the observation period. is the maximum value of the abnormal fluctuation range during the observation period. is the sigmoid function in the shape of S, which converts the number of abnormal values into a value between 0 and 1. When the number of abnormal values is much higher than the baseline, this value approaches 1. Adjust the exponent through the radical and the maximum duration. When the abnormal duration is long, the value of this term will decrease. directly reflects the size of the abnormal fluctuation range, normalizes it to between 0 and 1, and the value range is between 0 and 1, where 0 means no abnormal interference and 1 means extreme abnormal interference. As the severity of the abnormal data increases, the value of will also increase, so as to quantitatively analyze the severity of the abnormal data, evaluate the interference of the abnormal data on the vehicle state recognition, and take corresponding measures accordingly.

[0079] The three interference levels correspond to three interference thresholds. Among them, the interference threshold includes an upper threshold and a lower threshold.

[0080] The three interference levels and the three interference thresholds satisfy the following relationship:

[0081] Slight interference level ;

[0082] Medium interference level ;

[0083] Severe interference level ;

[0084] Among them, is the abnormal interference index. is the lower threshold corresponding to the medium interference level and the upper threshold corresponding to the slight interference level. is the lower threshold corresponding to the severe interference level and the upper threshold corresponding to the medium interference level. By analyzing the abnormal conditions of the sensor data under different environmental conditions, and analyzing the interference degree of the abnormal data on the vehicle state data collection, quantitative analysis is carried out to implement corresponding data adjustment measures, and then improve the accuracy of subsequent traffic management. By integrating data from different sources, real-time monitoring of the vehicle state and dynamic traffic management are realized, so as to provide a scientific basis for the traffic management department to formulate targeted preventive measures.

[0085] In step four, the process of adjusting the data application strategy according to the traffic rule requirements includes:

[0086] Verify the restored fused data, checking the integrity, accuracy, consistency, and timeliness of the data. Integrity checks ensure that all necessary data fields are fully populated without missing values, and check whether the number of data records matches the expectation to ensure no data loss. Accuracy checks verify the accuracy of the data by comparing with the data source or using known correct values, and use statistical methods to evaluate the reasonableness of the data. Consistency checks ensure that the data between different data sources is logically consistent and check whether the data follows the expected format and units. Timeliness checks verify the timeliness of the data to ensure that the data is up-to-date and relevant to the current traffic rule requirements, and analyze the data quality, identifying outliers in the data. According to the results of the quality analysis, identify the sources of errors and perform error correction to control data errors. Among them, analyze the links in the data collection, processing, transmission, and storage processes that may introduce errors, identify common error sources such as sensor failures, data loss, or data conversion errors, use data smoothing techniques to correct random errors in the data, and for systematic errors, adjust the data collection and processing methods, regularly evaluate the changing trend of data errors, and take corresponding corrective measures. Analyze traffic rule requirements, including speed limits, traffic signal control, and lane allocation, adjust the precise requirements of the corresponding data according to traffic rule requirements, synchronously adjust the data application strategy, implement and monitor the strategy effect, corresponding to the traffic rule requirements;

[0087] In step five, the process of implementing targeted traffic management includes:

[0088] Using real-time data processing technology, conduct real-time analysis on the collected vehicle status data, monitor the basic information of the vehicle's speed, position, and driving direction. According to the real-time traffic flow and vehicle status data, dynamically adjust the control strategy of traffic lights. Through the intelligent signal light system, improve traffic mobility and reduce traffic congestion. When a traffic accident or road congestion occurs, use the vehicle status data to quickly locate the problem area, and through traffic guidance and diversion strategies, provide real-time traffic guidance and information to drivers, guiding vehicles to avoid congested sections and balancing the road network load. Using the analysis results of cameras and vehicle status data, conduct real-time monitoring and recording of traffic violations, upload the violation data to the traffic management department for subsequent punishment and education work. Regularly analyze and evaluate the vehicle status monitoring and traffic management effects, use data visualization technology to display the changing trends of key indicators such as traffic flow, congestion conditions, and violation behaviors, and according to the analysis results, adjust and optimize the traffic management strategy.

[0089] Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art and related fields without creative efforts shall fall within the scope of protection of the present invention. Structures, devices, and operation methods not specifically described and explained in the present invention shall be implemented by conventional means in the art unless otherwise specified and limited.

Claims

1. A vehicle state acquisition method based on multi-source data fusion from a roadside perspective, characterized in that, Including the following steps: Step 1: Deploy various collection devices on the roadside to collect the original data of vehicle status and environmental conditions, and preprocess the collected original data; Step 2: Perform time synchronization operations on the original data collected by different sensors after preprocessing, and use data fusion technology to fuse the original data of different sensors; Step 3: In response to changes in different environmental conditions, monitor data anomalies in the sensors and data fusion results, analyze the severity of the anomalies, and match corresponding recovery and adjustment measures. The analysis process of the severity of the anomalies includes: Based on the fused data, analyze the data of various types of environmental conditions and monitor changes in different environmental conditions; Use time series analysis methods to comprehensively analyze the monitored data and identify abnormal data. The abnormal data is manifested as sudden changes in data values and abnormal expansions of data fluctuation ranges; Integrate the abnormal data to determine the characteristics of the abnormal data, namely the number of abnormal values, the duration of the anomaly, and the fluctuation range of the abnormal data. Analyze the interference of the abnormal data on vehicle status recognition, calculate the abnormal interference index, and then quantitatively analyze the severity of the abnormal data; According to the severity of the abnormal data, divide different interference levels, namely the minor interference level, the medium interference level, and the severe interference level. Combine the calculated abnormal interference index to match corresponding interference thresholds for each interference level, and match corresponding recovery and adjustment measures for each interference level; The process of obtaining the abnormal interference index includes: Integrate the abnormal data within a fixed time period from different sensors, including speed, position, driving direction, light intensity, weather conditions, temperature, humidity, and wind speed, to form an abnormal data set. The abnormal data set contains the time stamps and abnormal value information of each abnormal data point; Analyze the abnormal data set to determine the abnormal data characteristics of the number of abnormal values, the duration of the anomaly, and the fluctuation range of the abnormal data, and evaluate the impact of the abnormal data on vehicle status recognition, including the decrease in recognition accuracy and the increase in false alarm rate; Based on the evaluation results of the impact of the abnormal data on vehicle status recognition, comprehensively analyze and obtain the abnormal interference index by combining the abnormal data characteristics, and quantitatively analyze the severity of the interference of the abnormal data on vehicle status recognition; The calculation expression of the abnormal interference index is: Wherein, A is the abnormal interference index, N is the total number of outliers within a fixed time period, D is the total duration of the outliers, B is the maximum deviation of the outliers from their reference values, N0 is the number of baseline outliers, representing the number of outliers expected under normal conditions, τ is an adjustment parameter used to control the influence degree of the number of outliers on the index, D max is the maximum value of the abnormal duration during the observation period, B max is the maximum value of the abnormal fluctuation range during the observation period; The three interference levels correspond to three interference thresholds. Among them, the interference thresholds include upper thresholds and lower thresholds; The three interference levels and the three interference thresholds satisfy the following relationship: Slight interference level 0 < A ≤ A M ; Medium interference level AM < A ≤ A L ; Severe interference level A L <A < 1; where A is the abnormal interference index, and A M is the lower threshold corresponding to the medium interference level and the upper threshold corresponding to the slight interference level, and AL is the lower threshold corresponding to the severe interference level and the upper threshold corresponding to the medium interference level; Step 4: Verify and analyze the quality of the fused data after adjustment and recovery, control data errors, and adjust the application strategy of the data to meet traffic rule requirements; Step 5: Use the processed data to monitor vehicle status to implement targeted traffic management.

2. The vehicle state acquisition method based on multi-source data fusion from a roadside perspective according to claim 1, wherein: In the said Step 1, the process of collecting and preprocessing the original data includes: Deploy various collection devices on the roadside, including cameras, radars, lidars, and infrared sensors, to collect the original data of vehicle status and environmental conditions; Using a variety of deployed acquisition devices, collect vehicle status data such as the driving speed, position, and driving direction of the vehicle, and collect environmental condition data such as light intensity, weather conditions, temperature, humidity, and wind speed; Transmit the collected raw data to the data processing center wirelessly and perform preprocessing operations, which include data cleaning, format conversion, and data calibration steps; Construct a data warehouse and integrate the preprocessed data and store it in the data warehouse.

3. The vehicle state acquisition method based on multi-source data fusion from a roadside perspective according to claim 2, wherein: In the second step, the time synchronization and fusion process of the raw data includes: Retrieve the raw data collected by different sensors after preprocessing from the data warehouse, perform a time synchronization operation of software synchronization on it, traverse each data record, record the timestamp of each sensor data, use the lidar data as the benchmark for time synchronization, and extract its timestamp as the reference time point; For the data of other sensors, compare its timestamp with the reference time point of the lidar, use the interpolation algorithm to calculate the data value of other sensors at the lidar reference time point, and through interpolation, obtain the data of all sensors at the lidar reference time point to achieve time synchronization from software; Perform time alignment and coordinate alignment on the data obtained by different sensors. Among them, perform time alignment based on the result of the time synchronization operation, and convert the data of different sensors into a unified coordinate system for coordinate alignment; Using the data fusion algorithm of the Kalman filter method, define the state equation to describe the change of sensor data over time, and define the observation equation to describe the relationship between sensor data and the state, initialize the state estimate and covariance matrix, for each time step, use the state equation to predict the next state, use the observation equation and observation data to update the state estimate and covariance matrix, repeat the Kalman filter process until all sensor data are fused, and then obtain the fused data; Redeposit the fused data into the data warehouse for further analysis and processing.

4. A vehicle state acquisition method based on multi-source data fusion from a roadside perspective according to claim 1, characterized in that: The process of abnormal data identification includes: Using the time series analysis method, decompose the time series of the monitoring data into three parts: trend, seasonality, and residual. According to the manifestation states of sudden changes in the data value of abnormal data and abnormal expansion of the data fluctuation range, combine the three-sigma principle based on the normal distribution to set the abnormal detection threshold T; Analyze the residual part of the time series decomposition of the monitoring data, calculate the residual of each time point, and for each time point t, calculate the median absolute deviation of the residuals of the w time points before and after it to measure the volatility of the data; Calculate the absolute value of the difference between the actual observed value at time point t and the reference value at time point t, and combine it with the median absolute deviation to construct an abnormal detection index; Calculate the abnormal detection index of each time point and compare it with the set abnormal detection threshold. If the abnormal detection index is greater than the abnormal detection threshold T, the data at time point t is abnormal data.

5. The vehicle state acquisition method based on multi-source data fusion from a roadside perspective according to claim 4, characterized in that: The calculation expression of the abnormal detection index is: Among them, Z t is the anomaly detection index, Y t is the actual observed value at time point t, BV t is the reference value at time point t, MAD(R t-w , …, R t+w ) is the median absolute deviation of the residuals of w time points before and after time point t, (R t-w , …, R t+w ) are the residuals of w time points before and after time point t.

6. The vehicle state acquisition method based on multi-source data fusion from the roadside perspective according to claim 5, wherein: In the fourth step, the process of adjusting the data application strategy according to the traffic rule requirements includes: Verify the restored fused data and check the integrity, accuracy, consistency, and timeliness of the data; Identify the error sources based on the results of quality analysis and perform error correction to control data errors; Analyze traffic rule requirements, including speed limits, traffic signal control, and lane allocation. Adjust the precision requirements of the corresponding data according to traffic rule requirements, synchronously adjust the data application strategy, implement and monitor the strategy effect, corresponding to the traffic rule requirements.

7. A method for collecting vehicle states based on multi-source data fusion from a roadside perspective according to claim 6, characterized in that: In step five, the process of implementing targeted traffic management includes: Using real-time data processing technology, perform real-time analysis on the collected vehicle status data, monitor the basic information of the vehicle's speed, position, and driving direction. According to the real-time traffic flow and vehicle status data, dynamically adjust the control strategy of traffic lights; In the event of a traffic accident or road congestion, use vehicle status data to quickly locate the problem area, and through traffic guidance and diversion strategies, provide drivers with real-time traffic guidance and diversion information to guide vehicles to avoid congested sections and balance the road network load; Using the analysis results of cameras and vehicle status data, conduct real-time monitoring and recording of traffic violations, and upload the violation data to the traffic management department; Regularly analyze and evaluate vehicle status monitoring and traffic management effects, use data visualization technology to display the change trends of key indicators such as traffic flow, congestion conditions, and violations, and adjust and optimize traffic management strategies according to the analysis results.

Images