Intelligent management method for identifying driving risk perception based on multidimensional data analysis

Abstract

This invention discloses an intelligent management method for driving risk perception and identification based on multidimensional data analysis, relating to the field of risk identification. The method includes the following steps: acquiring multidimensional vehicle status information and extracting the data to obtain multidimensional feature data; constructing a multidimensional driving profile based on the multidimensional feature data and performing spatiotemporal fusion of the multidimensional feature data to obtain multidimensional fused feature data; matching risk behaviors based on the multidimensional fused feature data to obtain the identification intervals of corresponding types of risk behaviors; classifying the identification intervals of each multidimensional fused feature data into risk levels to obtain risk assessment levels, and performing correlation and comparison analysis to obtain correlation and comparison feature data; verifying and matching the relationships between each multidimensional feature data based on the correlation and comparison feature data to obtain corresponding perception decision information; and obtaining corresponding risk intervention information for intelligent management based on the perception decision information. This invention improves the accuracy of driving risk perception and identification.

Description

Technical Field

[0001] This invention relates to the field of risk identification technology, and in particular to an intelligent management method for driving risk perception and identification based on multidimensional data analysis. Background Technology

[0002] With the continuous expansion of global civilian vehicle ownership and commercial vehicle operation scale, road traffic has become a core hub of modern society. However, the accompanying road traffic safety issues are becoming increasingly prominent. Statistics show that most road traffic accidents are directly or indirectly related to factors such as human driving behavior, abnormal vehicle condition, and insufficient adaptability to the road environment. High-risk driving behaviors such as fatigued driving, distracted driving, speeding, sudden acceleration and deceleration, and following too closely, as well as external conditions such as severe weather, complex road conditions, and traffic congestion, together constitute the main causes of driving risks. Therefore, the perception, identification, and safety management of driving risks have become an urgent need in the field of intelligent transportation safety.

[0003] A search revealed Chinese invention patent CN112270114A, which discloses a method for identifying personalized risk behaviors in vehicles. The method involves a vehicle-to-everything (V2X) information platform collecting vehicle speed data, preprocessing the collected speed data, calculating driving acceleration data based on the preprocessed speed, standardizing the driving acceleration data using Z-normalization, classifying the standardized acceleration data using PAA (Parallel Autoencoder) method, extracting frequent driving behavior patterns (typical driving behavior patterns), determining driver driving behavior preferences based on the extracted typical driving behavior patterns, building a personalized driving behavior assessment model based on typical driving behaviors, and using the constructed assessment model to evaluate the driver's driving risk. This invention specifically helps drivers improve their driving behavior, establishes a personalized driving risk identification method, and proposes a new approach to improve vehicle safety and reduce accidents.

[0004] Compared with existing technologies, the Chinese invention patent with patent number CN112270114A can identify risks based on the driver's driving behavior, thereby improving vehicle safety.

[0005] However, in actual use, the above methods only identify risks in the driver's driving behavior, ignoring the environmental information and other external influences in which the driver's driving behavior occurs, thus reducing the accuracy of the driving risk identification process to some extent. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of insufficient accuracy in existing technologies by proposing an intelligent management method for driving risk perception and identification based on multidimensional data analysis.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The intelligent management method for driving risk perception and identification based on multidimensional data analysis includes the following steps:

[0009] Step S1: Set up an intelligent risk management platform, obtain multi-dimensional data status information through the intelligent risk management platform, preprocess the obtained multi-dimensional data status information, extract data features according to the data preprocessing results, and obtain the corresponding multi-dimensional feature data.

[0010] Step S2: Construct a corresponding multi-dimensional driving profile from the obtained multi-dimensional feature data, and perform spatiotemporal fusion of the obtained multi-dimensional feature data based on the constructed multi-dimensional driving profile to obtain multi-dimensional fused feature data;

[0011] Step S3: Perform risk behavior matching on the obtained multidimensional fusion feature data to obtain the identification range of the multidimensional fusion feature data corresponding to each type of risk behavior;

[0012] Step S4: Divide the risk level according to the identification range of the multidimensional fusion feature data corresponding to each risk behavior, obtain the corresponding risk assessment level, and perform correlation and comparison analysis on the multidimensional feature data within each risk assessment level to obtain correlation and comparison feature data.

[0013] Step S5: Verify and match the relationships between the various multidimensional feature data based on the correlation and comparison feature data, and obtain the perceived decision information of the corresponding risk assessment level based on the verification and matching results;

[0014] Step S6: Obtain corresponding risk intervention information based on the obtained perception and decision information, and perform intelligent management based on the obtained risk intervention information.

[0015] The above technical solution further includes: the process of acquiring multidimensional data state information includes:

[0016] An intelligent risk management platform is set up, which is used to obtain the basic information of the corresponding vehicle and set up an intelligent management terminal based on the basic information of the vehicle.

[0017] The intelligent management terminal is used to communicate with the multi-dimensional data acquisition terminal in the corresponding vehicle, and to acquire the corresponding multi-dimensional data status information through the multi-dimensional data acquisition terminal. The multi-dimensional data status information includes driver status information, vehicle operation status information, road environment status information, and traffic flow status information.

[0018] The obtained multidimensional data state information is preprocessed separately, and the data preprocessing process includes data cleaning, data standardization and data alignment.

[0019] Furthermore, the process of obtaining the corresponding multidimensional feature data includes:

[0020] The multidimensional data status information that has completed data preprocessing is sequentially subjected to data feature extraction based on the corresponding driver status information, vehicle operation status information, road environment status information and traffic flow status information. The data feature extraction includes basic feature extraction and change feature extraction.

[0021] Based on the data feature extraction results, obtain the multidimensional feature data corresponding to the state information of the corresponding type of multidimensional data.

[0022] Furthermore, the process of acquiring multidimensional fusion feature data includes:

[0023] The obtained multidimensional feature data is classified into profile levels according to the multidimensional data status information to which it belongs. Traffic flow status information, road environment status information, driver status information and vehicle operation status are set into corresponding multidimensional levels. The multidimensional levels include traffic level, road level, driving level and vehicle level, and each level is sequentially connected.

[0024] A label system is constructed in each multidimensional level. The label system is used to describe the quantitative data of multidimensional feature data corresponding to different types of multidimensional data state information. Each multidimensional feature data is compared and matched with the corresponding label system. A hierarchical multidimensional profile is constructed based on the comparison and matching results. The obtained hierarchical multidimensional profiles are connected sequentially between levels to construct the corresponding multidimensional driving profile.

[0025] Based on the constructed multidimensional driving profile, the obtained multidimensional feature data is spatiotemporally fused to obtain multidimensional fused feature data.

[0026] Furthermore, the process of matching the obtained multi-dimensional fused feature data with risk behaviors includes:

[0027] The multi-dimensional driving profiles are classified according to the corresponding label system within the hierarchical multi-dimensional profiles to obtain the hierarchical label combination type of the multi-dimensional driving profiles. Based on the hierarchical label combination type, the corresponding historical multi-dimensional fusion feature data and the corresponding historical risk behavior accident data are obtained. The obtained historical multi-dimensional fusion feature data and historical risk behavior accident data are set as the corresponding historical datasets.

[0028] Statistical analysis was performed on the historical datasets corresponding to the corresponding label combination types at each level to obtain the initial risk identification baseline at the corresponding label position within each label combination type.

[0029] Each level of label combination type is combined horizontally and vertically. Based on the horizontal and vertical combination processing results, the changes in the initial risk identification baseline at the corresponding label position within and between the level multidimensional profiles are statistically analyzed to obtain an initial identification baseline change comparison table.

[0030] The obtained multidimensional fusion feature data are compared and matched sequentially with the label position information of the corresponding multidimensional driving profile based on the initial label baseline change comparison table. The initial risk label baseline is adaptively adjusted according to the comparison and matching results, and the corresponding label interval is set according to the adjusted initial risk label baseline.

[0031] Furthermore, the process of obtaining the associated comparison feature data includes:

[0032] The risk levels of the identification intervals within the hierarchical multidimensional profiles corresponding to historical multidimensional fusion feature data and historical risk behavior accident data are divided to obtain the risk assessment level within each identification interval of each hierarchical multidimensional profile.

[0033] Based on the risk assessment level within each identifier interval of each level of multidimensional profile, cross-combinations are performed within and between levels. Based on the cross-combination results, historical cross datasets are set up corresponding to the multidimensional feature data and historical risk behavior accident data within the corresponding historical multidimensional fusion feature data.

[0034] The obtained historical cross datasets are subjected to correlation analysis to obtain the cross-correlation data corresponding to each historical cross dataset. The obtained cross-correlation data is then filtered. Based on the filtering results, the probability of the simultaneous occurrence of multidimensional feature data corresponding to historical risk behavior accident data in the historical cross dataset is statistically analyzed to obtain the corresponding cross-correlation probability. Based on the obtained cross-correlation probability, the corresponding correlation control feature data of the historical cross dataset is set.

[0035] Furthermore, the process of obtaining the perceived decision information corresponding to the risk assessment level includes:

[0036] Obtain the multidimensional feature data corresponding to each multidimensional fusion feature data, compare and match the risk assessment level corresponding to each multidimensional feature data within the adjusted label interval, and obtain the comparison and matching results corresponding to each multidimensional feature data.

[0037] The comparison and matching results corresponding to each multidimensional feature data are matched with the corresponding associated comparison feature data to verify the conditions. It is determined whether the comparison and matching results corresponding to each multidimensional feature data meet the verification conditions. If they all meet the verification conditions, the corresponding multidimensional feature data is marked normally. If there are verification conditions that do not meet the conditions, the corresponding multidimensional feature data is marked abnormally. Perceptual decision information is generated based on the comparison and matching results of each multidimensional feature data and the marking results of the corresponding verification conditions.

[0038] Furthermore, the process of obtaining corresponding risk intervention information based on the acquired perception and decision-making information includes:

[0039] Obtain the comparison and matching results corresponding to each multi-dimensional feature data in the perception and decision information, as well as the labeling results of the verification conditions. Then, sequentially perform risk traversal on the comparison and matching results corresponding to each multi-dimensional feature data to determine whether there is any risky behavior.

[0040] If no risky behavior exists, no risk intervention information is generated. If risky behavior exists but there is no verification condition for anomaly markers, statistical analysis is performed on the risky behavior, a risk intervention control database is preset, and the corresponding risk intervention information in the risk intervention control database is obtained based on the statistical analysis results. Intelligent management is then performed based on the risk intervention information. If risky behavior exists and verification condition for anomaly markers exists, the corresponding risk intervention information is obtained based on the existing risky behavior, and the multidimensional feature data corresponding to the corresponding anomaly marker results are re-analyzed until there are no anomaly marker results in the perception decision information. Finally, the corresponding risk intervention information is output for intelligent management.

[0041] The present invention has the following beneficial effects:

[0042] 1. In this invention, multi-dimensional data state information corresponding to driver status information, vehicle operation status information, road environment status information, and traffic flow status information is obtained and extracted separately. Based on the data extraction results, a multi-dimensional profile corresponding to each multi-dimensional level is constructed. The multi-dimensional profiles at each level are then progressively connected to construct a multi-dimensional driving profile. Based on the constructed multi-dimensional driving profile, driving risks during the driving process are perceived and identified sequentially from the traffic level, road level, driving level, and vehicle level, which improves the accuracy of risk perception and identification to a certain extent.

[0043] 2. In this invention, the labels of the multi-dimensional profiles at each level within the multi-dimensional driving profile are combined and statistically analyzed to obtain the corresponding initial risk identification baseline. The obtained initial risk identification baseline is then adjusted according to the labels of the corresponding multi-dimensional feature data. Based on the adjustment, the identification range of the corresponding risk behavior is obtained, which can improve the accuracy and efficiency of the risk identification process to a certain extent.

[0044] 3. In this invention, the risk assessment levels within each identifier interval of each level of multidimensional profile are cross-combined within and between levels. Based on the historical cross dataset corresponding to the cross-combination results, correlation comparison statistics are performed to obtain the corresponding correlation comparison feature data. Based on the obtained correlation comparison feature data, the matching results of each multidimensional feature data are verified and analyzed, which can improve the accuracy of the risk perception and identification process to a certain extent. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the intelligent management method for driving risk perception and identification based on multidimensional data analysis proposed in this invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Example 1

[0048] like Figure 1 As shown, the intelligent management method for driving risk perception and identification based on multi-dimensional data analysis proposed in this invention includes the following steps:

[0049] Step S1: Set up an intelligent risk management platform, obtain multi-dimensional data status information through the intelligent risk management platform, preprocess the obtained multi-dimensional data status information, extract data features according to the data preprocessing results, and obtain the corresponding multi-dimensional feature data.

[0050] Step S2: Construct a corresponding multi-dimensional driving profile from the obtained multi-dimensional feature data, and perform spatiotemporal fusion of the obtained multi-dimensional feature data based on the constructed multi-dimensional driving profile to obtain multi-dimensional fused feature data;

[0051] Step S3: Perform risk behavior matching on the obtained multidimensional fusion feature data to obtain the identification range of the multidimensional fusion feature data corresponding to each type of risk behavior;

[0052] Step S4: Divide the risk level according to the identification range of the multidimensional fusion feature data corresponding to each risk behavior, obtain the corresponding risk assessment level, and perform correlation and comparison analysis on the multidimensional feature data within each risk assessment level to obtain correlation and comparison feature data.

[0053] Step S5: Verify and match the relationships between the various multidimensional feature data based on the correlation and comparison feature data, and obtain the perceived decision information of the corresponding risk assessment level based on the verification and matching results;

[0054] Step S6: Obtain corresponding risk intervention information based on the obtained perception and decision information, and perform intelligent management based on the obtained risk intervention information.

[0055] In the specific implementation process, the step S1 of setting up an intelligent risk management platform, obtaining multi-dimensional data status information through the intelligent risk management platform, preprocessing the obtained multi-dimensional data status information, and extracting data features according to the data preprocessing results to obtain the corresponding multi-dimensional feature data includes:

[0056] S11: Set up an intelligent risk management platform, which is used to obtain basic vehicle information of the corresponding vehicle and set up an intelligent management terminal based on the basic vehicle information, wherein:

[0057] The basic information of the driving vehicle includes the vehicle identification information, performance parameter information and driving status information of the corresponding vehicle. The intelligent management terminal is used to conduct intelligent risk assessment of the driving process of the corresponding vehicle according to the risk identification process in the intelligent risk management platform.

[0058] S12: The intelligent management terminal is used to communicate with the multi-dimensional data acquisition terminal in the corresponding vehicle, and to acquire corresponding multi-dimensional data status information through the multi-dimensional data acquisition terminal. The multi-dimensional data status information includes driver status information, vehicle operating status information, road environment status information, and traffic flow status information, wherein:

[0059] Driver status information includes physiological signals, eye features, and driving operation information;

[0060] Vehicle operating status information includes speed, acceleration, yaw rate, tire pressure, etc.

[0061] Road environment status information includes road condition information, weather information, traffic sign information, etc.

[0062] Traffic flow status information includes time information, relative speed information, vehicle density information, historical accident information, etc.

[0063] S13: Perform data preprocessing on the obtained multidimensional data state information respectively. The data preprocessing process includes data cleaning, data standardization and data alignment.

[0064] S14: The multidimensional data state information that has completed data preprocessing is sequentially subjected to data feature extraction based on the corresponding driver state information, vehicle operation state information, road environment state information and traffic flow state information. The data feature extraction includes basic feature extraction and change feature extraction. Based on the data feature extraction results, the multidimensional feature data corresponding to the corresponding type of multidimensional data state information is obtained.

[0065] S141: The basic feature extraction process is as follows: the collected multidimensional data state information is used to extract multidimensional feature data that can be directly extracted under the current situation, such as physiological characteristics, driving behavior, etc.

[0066] S142: The process of extracting change features is as follows: the multidimensional data status information collected can be used to extract multidimensional feature data based on the changes or frequency of occurrence within the time period set in advance, such as the frequency of accelerator / brake operation, the standard deviation of following distance, and the frequency of lane change.

[0067] In the specific implementation process, the step S2, which involves constructing a corresponding multi-dimensional driving profile from the obtained multi-dimensional feature data and performing spatiotemporal fusion on the obtained multi-dimensional feature data based on the constructed multi-dimensional driving profile, to obtain multi-dimensional fused feature data, includes:

[0068] S21: The obtained multidimensional feature data is classified according to its respective multidimensional data state information, and traffic flow state information, road environment state information, driver state information, and vehicle operation state are set into corresponding multidimensional levels. The multidimensional levels include traffic level, road level, driver level, and vehicle level, and each level is sequentially and progressively connected, that is, the traffic level is connected to the road level, the road level is connected to the driver level, and so on.

[0069] Connect to the vehicle level;

[0070] S22: Construct a label system in each multidimensional level. The label system is used to describe the quantized data of multidimensional feature data corresponding to different types of multidimensional data state information. The label system sets corresponding labels according to the quantized data of multidimensional feature data in each multidimensional level.

[0071] S23: Compare and match each multidimensional feature data with the corresponding label system, construct a hierarchical multidimensional profile based on the comparison and matching results, and construct the corresponding multidimensional driving profile by sequentially connecting the obtained hierarchical multidimensional profiles according to the hierarchical levels.

[0072] S24: Based on the constructed multidimensional driving profile, the obtained multidimensional feature data is spatiotemporally fused to obtain multidimensional fused feature data, which is the quantification of various types of multidimensional feature data within the same time period.

[0073] In the specific implementation process, the step S3 of matching the obtained multi-dimensional fusion feature data with risk behaviors and obtaining the identification interval of the multi-dimensional fusion feature data corresponding to each type of risk behavior includes:

[0074] S31: Classify the multi-dimensional driving profiles according to the corresponding label system within the hierarchical multi-dimensional profiles, obtain the hierarchical label combination type of the multi-dimensional driving profiles, obtain the corresponding historical multi-dimensional fusion feature data and the corresponding historical risk behavior accident data according to the hierarchical label combination type, and set the obtained historical multi-dimensional fusion feature data and historical risk behavior accident data as the corresponding historical datasets.

[0075] S32: Perform statistical analysis on the historical datasets corresponding to the corresponding level label combination types, and set the initial risk identification baseline at the corresponding label position within each level label combination type. The initial risk identification baseline is the minimum standard corresponding to the multidimensional feature data when historical risk behavior accident data occurs.

[0076] S33: Perform horizontal and vertical combinations on the label combination types of each level, and statistically analyze the changes in the initial risk identification baseline at the corresponding label positions within and between the level multidimensional profiles based on the horizontal and vertical combination processing results, and obtain a comparison table of initial identification baseline changes.

[0077] S331: In the process of horizontal and vertical combination of the label combination types of each level, the quantization level of the corresponding label in the corresponding level label combination type is adjusted sequentially based on the single variable method. Among them, horizontal combination means that the label combination types in other multi-dimensional levels are consistent, and the quantization level of the label corresponding to the label combination type in the same multi-level gradually increases. Vertical combination means that the label combination types in the same multi-dimensional level are consistent, and the quantization level of the label corresponding to the label combination type in adjacent multi-level levels is changed one by one.

[0078] S331: Based on the results of the hierarchical label combination type, the changes in the minimum standards corresponding to the multidimensional feature data of the historical risk behavior accident data involved are statistically analyzed, and the obtained statistical results are used to set up an initial identification baseline change comparison table. The initial identification baseline change comparison table includes the changes of different hierarchical label combination types on each corresponding initial risk identification baseline.

[0079] S34: The obtained multidimensional fusion feature data are compared and matched sequentially based on the label position information of the corresponding multidimensional driving profile and the initial label baseline change comparison table. The initial risk label baseline is adaptively adjusted according to the comparison and matching results. The corresponding label interval is set according to the adjusted initial risk label baseline. The label interval is the quantitative situation when there may be risky behavior.

[0080] In the specific implementation process, the step S4, which involves classifying risk levels based on the identifier range of the multidimensional fusion feature data corresponding to each risk behavior, obtaining the corresponding risk assessment level, and performing correlation and comparison analysis on the multidimensional feature data within each risk assessment level to obtain the correlation and comparison feature data, includes:

[0081] S41: Divide the identification intervals within the hierarchical multidimensional profiles corresponding to the historical multidimensional fusion feature data and historical risk behavior accident data into risk levels, and obtain the risk assessment level within each identification interval of each hierarchical multidimensional profile.

[0082] S411: Map the occurrence of historical risk behavior incident data to the corresponding positions within the identifier intervals of each multidimensional feature data in the historical multidimensional fusion feature data;

[0083] S412: Perform a percentage analysis of the occurrence rate of historical risk behavior incidents at corresponding locations within each identification interval, and classify the identification intervals into risk levels based on the percentage analysis results;

[0084] S413: Obtain the risk assessment level of the corresponding multidimensional feature data based on the risk level classification results. The risk assessment level includes three levels: high risk, medium risk, and low risk.

[0085] S42: Based on the risk assessment level of each identifier interval in the multidimensional profile of each level, perform cross-combination within and between levels respectively, and set up the historical cross dataset corresponding to the multidimensional feature data and the historical risk behavior accident data in the corresponding historical multidimensional fusion feature data according to the cross-combination results.

[0086] S421: The process of cross-combining within a hierarchy is as follows: combine the historical risk behavior accident data corresponding to the multidimensional feature data at the risk assessment level between each identifier interval in the hierarchical multidimensional profile. The resulting historical cross dataset consists of historical risk behavior accident data and historical multidimensional fusion feature data at different risk assessment levels belonging to different multidimensional feature data within the same multidimensional hierarchy.

[0087] S422: The process of cross-combining between levels is as follows: the historical risk behavior accident data corresponding to the multidimensional feature data at the risk assessment level between each identifier interval in the multidimensional profile of different levels are combined. The resulting historical cross dataset contains historical risk behavior accident data and historical multidimensional fusion feature data at different risk assessment levels belonging to different multidimensional feature data in different multidimensional levels.

[0088] S43: Perform correlation analysis on the obtained historical cross datasets to obtain the cross-correlation data corresponding to each historical cross dataset. Filter the obtained cross-correlation data. Based on the filtering results, perform statistical analysis on the probability of the simultaneous occurrence of multidimensional feature data corresponding to historical risk behavior accident data in the historical cross datasets to obtain the corresponding cross-correlation probability. Set the corresponding correlation control feature data for the historical cross datasets based on the obtained cross-correlation probability.

[0089] S431: The process of obtaining cross-correlation data includes: labeling different types of multidimensional feature data within the historical cross dataset as x and y, and labeling the cross-correlation data as r, where:

[0090] Where i is the data label for x and y, and These are the mean data for different types of multidimensional feature data;

[0091] S432: Set the cross-association threshold, compare the difference between the obtained cross-association data and the cross-association threshold, and select the historical cross-association dataset if the cross-association data is greater than or equal to the cross-association threshold.

[0092] S433: Perform statistical analysis on the probability of simultaneous occurrence of multidimensional feature data at the corresponding risk assessment level when historical risk behavior incidents occur in the selected historical cross dataset to obtain the cross-correlation probability.

[0093] S434: Set the correlation control threshold, compare the difference between the obtained cross-association probability and the correlation control threshold. If the cross-association probability is greater than or equal to the correlation control threshold, set the historical cross-association dataset as the control group, and associate the corresponding cross-association probability with the control group.

[0094] S435: Integrate the obtained historical cross datasets and their corresponding cross-association probabilities with the association labeling results of the control group to generate association control feature data.

[0095] In the specific implementation process, the step S5, which involves verifying and matching the relationships between various multidimensional feature data based on the correlation comparison feature data, and obtaining the perceived decision information corresponding to the risk assessment level based on the verification and matching results, includes:

[0096] S51: Obtain the multi-dimensional feature data corresponding to each multi-dimensional fusion feature data, compare and match the risk assessment level corresponding to each multi-dimensional feature data within the adjusted label interval, and obtain the comparison and matching results corresponding to each multi-dimensional feature data.

[0097] S52: The comparison matching results corresponding to each multidimensional feature data are matched with the corresponding associated comparison feature data to determine whether the comparison matching results corresponding to each multidimensional feature data meet the verification conditions.

[0098] S521: In the process of determining whether the verification conditions are met, it is determined whether the corresponding multidimensional feature data in the associated control feature data appear simultaneously or not at the same time. If the multidimensional feature data in the corresponding risk assessment level appear simultaneously or not at the same time, the verification conditions are met; otherwise, the verification conditions are not met.

[0099] S522: If all verification conditions are met, the corresponding multidimensional feature data will be labeled normally.

[0100] S523: If there are verification conditions that do not meet the requirements, the corresponding multidimensional feature data will be marked as anomalies, and perceptual decision information will be generated based on the comparison and matching results of each multidimensional feature data and the marking results of the corresponding verification conditions.

[0101] In the specific implementation process, the step S6 of obtaining corresponding risk intervention information based on the obtained perception and decision information, and performing intelligent management based on the obtained risk intervention information includes:

[0102] S61: Obtain the comparison and matching results and the labeling results of the verification conditions corresponding to each multi-dimensional feature data in the perception decision information, and sequentially perform risk traversal on the comparison and matching results corresponding to each multi-dimensional feature data to determine whether there is any risky behavior.

[0103] S611: In the process of determining whether there is a risky behavior, it is determined whether the risk level intervals to which each multidimensional feature data belongs are all in low risk. If the risk level intervals to which the multidimensional feature data belongs are all in low risk, then it is not judged as a risky behavior; otherwise, it is judged as a risky behavior.

[0104] S612: If no risky behavior exists, no risk intervention information will be generated;

[0105] S613: If there are abnormal markers for risky behaviors without verification conditions, perform statistical analysis on the risky behaviors, obtain the risk level range of the multidimensional feature data corresponding to the corresponding type of risky behavior, set risk weights according to the risk level range, and perform weighted processing on the risk weights and occurrence of each multidimensional feature data to obtain comprehensive evaluation data of the corresponding type of risky behavior.

[0106] S614: Set up corresponding risk intervention information based on the comprehensive assessment data of the corresponding types of risk behaviors. The risk intervention information includes various types such as sound and light warnings, seat vibration, and voice prompts. Integrate the risk intervention information set under the comprehensive assessment data corresponding to each type of risk behavior to generate a risk intervention comparison database.

[0107] S615: Obtain the corresponding risk intervention information from the risk intervention comparison database based on the comprehensive assessment data corresponding to the statistical analysis results, and perform intelligent management based on the risk intervention information;

[0108] S616: If there are abnormal markers for risky behaviors and verification conditions, obtain the corresponding risk intervention information based on the existing risky behaviors, and re-analyze the multi-dimensional feature data corresponding to the abnormal marker results until there are no abnormal marker results in the perception decision information, and output the corresponding risk intervention information for intelligent management.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A driving risk perception and identification intelligent management method based on multidimensional data analysis, characterized in that, Includes the following steps: Step S1: Set up an intelligent risk management platform, obtain multi-dimensional data status information through the intelligent risk management platform, preprocess the obtained multi-dimensional data status information, extract data features according to the data preprocessing results, and obtain the corresponding multi-dimensional feature data. Step S2: Construct a corresponding multi-dimensional driving profile from the obtained multi-dimensional feature data, and perform spatiotemporal fusion of the obtained multi-dimensional feature data based on the constructed multi-dimensional driving profile to obtain multi-dimensional fused feature data; Step S3: Perform risk behavior matching on the obtained multidimensional fusion feature data to obtain the identification range of the multidimensional fusion feature data corresponding to each type of risk behavior; Step S4: Divide the risk level according to the identification range of the multidimensional fusion feature data corresponding to each risk behavior, obtain the corresponding risk assessment level, and perform correlation and comparison analysis on the multidimensional feature data within each risk assessment level to obtain correlation and comparison feature data. Step S5: Verify and match the relationships between the various multidimensional feature data based on the correlation and comparison feature data, and obtain the perceived decision information of the corresponding risk assessment level based on the verification and matching results; Step S6: Obtain corresponding risk intervention information based on the obtained perception and decision information, and perform intelligent management based on the obtained risk intervention information.

2. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 1, characterized in that, The process of acquiring multidimensional data state information includes: An intelligent risk management platform is set up, which is used to obtain the basic information of the corresponding vehicle and set up an intelligent management terminal based on the basic information of the vehicle. The intelligent management terminal is used to communicate with the multi-dimensional data acquisition terminal in the corresponding vehicle, and to acquire the corresponding multi-dimensional data status information through the multi-dimensional data acquisition terminal. The multi-dimensional data status information includes driver status information, vehicle operation status information, road environment status information, and traffic flow status information. The obtained multidimensional data state information is preprocessed separately, and the data preprocessing process includes data cleaning, data standardization and data alignment.

3. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 2, characterized in that, The process of obtaining the corresponding multidimensional feature data includes: The multidimensional data status information that has completed data preprocessing is sequentially subjected to data feature extraction based on the corresponding driver status information, vehicle operation status information, road environment status information and traffic flow status information. The data feature extraction includes basic feature extraction and change feature extraction. Based on the data feature extraction results, obtain the multidimensional feature data corresponding to the state information of the corresponding type of multidimensional data.

4. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 3, characterized in that, The process of acquiring multidimensional fusion feature data includes: The obtained multidimensional feature data is classified into profile levels according to the multidimensional data status information to which it belongs. Traffic flow status information, road environment status information, driver status information and vehicle operation status are set into corresponding multidimensional levels. The multidimensional levels include traffic level, road level, driving level and vehicle level, and each level is sequentially connected. A label system is constructed in each multidimensional level. The label system is used to describe the quantitative data of multidimensional feature data corresponding to different types of multidimensional data state information. Each multidimensional feature data is compared and matched with the corresponding label system. A hierarchical multidimensional profile is constructed based on the comparison and matching results. The obtained hierarchical multidimensional profiles are connected sequentially between levels to construct the corresponding multidimensional driving profile. Based on the constructed multidimensional driving profile, the obtained multidimensional feature data is spatiotemporally fused to obtain multidimensional fused feature data.

5. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 4, characterized in that, The process of matching risk behaviors using the obtained multidimensional fusion feature data includes: The multi-dimensional driving profiles are classified according to the corresponding label system within the hierarchical multi-dimensional profiles to obtain the hierarchical label combination type of the multi-dimensional driving profiles. Based on the hierarchical label combination type, the corresponding historical multi-dimensional fusion feature data and the corresponding historical risk behavior accident data are obtained. The obtained historical multi-dimensional fusion feature data and historical risk behavior accident data are set as the corresponding historical datasets. Statistical analysis was performed on the historical datasets corresponding to the corresponding label combination types at each level to obtain the initial risk identification baseline at the corresponding label position within each label combination type. Each level of label combination type is combined horizontally and vertically. Based on the horizontal and vertical combination processing results, the changes in the initial risk identification baseline at the corresponding label position within and between the level multidimensional profiles are statistically analyzed to obtain an initial identification baseline change comparison table. The obtained multidimensional fusion feature data are compared and matched sequentially with the label position information of the corresponding multidimensional driving profile based on the initial label baseline change comparison table. The initial risk label baseline is adaptively adjusted according to the comparison and matching results, and the corresponding label interval is set according to the adjusted initial risk label baseline.

6. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 5, characterized in that, The process of obtaining the associated control feature data includes: The risk levels of the identification intervals within the hierarchical multidimensional profiles corresponding to historical multidimensional fusion feature data and historical risk behavior accident data are divided to obtain the risk assessment level within each identification interval of each hierarchical multidimensional profile. Based on the risk assessment level within each identifier interval of each level of multidimensional profile, cross-combinations are performed within and between levels. Based on the cross-combination results, historical cross datasets are set up corresponding to the multidimensional feature data and historical risk behavior accident data within the corresponding historical multidimensional fusion feature data. The obtained historical cross datasets are subjected to correlation analysis to obtain the cross-correlation data corresponding to each historical cross dataset. The obtained cross-correlation data is then filtered. Based on the filtering results, the probability of the simultaneous occurrence of multidimensional feature data corresponding to historical risk behavior accident data in the historical cross dataset is statistically analyzed to obtain the corresponding cross-correlation probability. Based on the obtained cross-correlation probability, the corresponding correlation control feature data of the historical cross dataset is set.

7. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 6, characterized in that, The process of obtaining the perceived decision information corresponding to the risk assessment level includes: Obtain the multidimensional feature data corresponding to each multidimensional fusion feature data, compare and match the risk assessment level corresponding to each multidimensional feature data within the adjusted label interval, and obtain the comparison and matching results corresponding to each multidimensional feature data. The comparison and matching results corresponding to each multidimensional feature data are matched with the corresponding associated comparison feature data to verify the conditions. It is determined whether the comparison and matching results corresponding to each multidimensional feature data meet the verification conditions. If they all meet the verification conditions, the corresponding multidimensional feature data is marked normally. If there are verification conditions that do not meet the conditions, the corresponding multidimensional feature data is marked abnormally. Perceptual decision information is generated based on the comparison and matching results of each multidimensional feature data and the marking results of the corresponding verification conditions.

8. The intelligent management method for driving risk perception and identification based on multidimensional data analysis according to claim 7, characterized in that, The process of obtaining corresponding risk intervention information based on the acquired perception and decision-making information includes: Obtain the comparison and matching results corresponding to each multi-dimensional feature data in the perception and decision information, as well as the labeling results of the verification conditions. Then, sequentially perform risk traversal on the comparison and matching results corresponding to each multi-dimensional feature data to determine whether there is any risky behavior. If no risky behavior exists, no risk intervention information is generated. If risky behavior exists but there is no verification condition for anomaly markers, statistical analysis is performed on the risky behavior, a risk intervention control database is preset, and the corresponding risk intervention information in the risk intervention control database is obtained based on the statistical analysis results. Intelligent management is then performed based on the risk intervention information. If risky behavior exists and verification condition for anomaly markers exists, the corresponding risk intervention information is obtained based on the existing risky behavior, and the multidimensional feature data corresponding to the corresponding anomaly marker results are re-analyzed until there are no anomaly marker results in the perception decision information. Finally, the corresponding risk intervention information is output for intelligent management.

Images