A highway driving risk discrimination method based on speed risk potential field

Abstract

The application relates to a highway driving risk discrimination method based on a speed risk potential field, which comprises the following steps: 1) data acquisition and processing: acquiring and processing gantry bay road section monitoring data, highway accident data and map POI data; 2) analyzing the data acquired and processed in step 1) to obtain speed data and traffic accident data; 3) establishing a speed risk potential field model according to the data analysis result in step 2), determining the speed potential V i,j , the speed potential field intensity and the speed risk potential field intensity realize the highway driving risk discrimination based on the speed risk potential field. The method can solve the problem that the current driving risk discrimination is difficult to realize risk classification in a macroscopic level.

Description

Technical Field

[0001] This invention belongs to the field of road traffic safety technology, and relates to a method for identifying highway driving risks in the transportation industry, and more particularly to a method for identifying highway driving risks based on a speed risk potential field. Background Technology

[0002] Current research on the impact of vehicle speed on driving risk can be divided into two aspects: risk mechanism and accident correlation. Research on risk mechanism attempts to explain the inherent laws governing the impact of speed on driving risk from the perspectives of "people, vehicles, and roads," proposing theories such as "sight distance theory," "safe distance theory," "vehicle dynamics," and "traffic conflict." Regarding accident correlation, statistical methods are used to analyze the correlation between speed and traffic accidents, with relevant statistical indicators including absolute speed, speed dispersion, and operating speed. Assessment methods based on risk mechanism can effectively identify risk types, but the consideration of risk causes is relatively singular, resulting in incomplete risk assessment results. Assessment methods based on speed statistical indicators can more comprehensively characterize driving risk, but due to limitations in data collection methods, there is limited research on the distribution of speed in two-dimensional space.

[0003] The speed risk potential field is based on the trend of speed change, considers the impact of road alignment on driving risk, and uses potential field theory to characterize the distribution pattern of road driving risk within the road domain. The speed risk potential field consists of two parts: speed potential energy and speed risk potential field strength. When a vehicle approaches a risk source, its speed decreases; when it leaves a risk source, its speed increases. This change in motion can be considered as the vehicle being acted upon by an "external force," that is, the vehicle being affected by the risk field.

[0004] Driving risk identification includes spatial location, risk classification, and quantification of high-risk road sections. Existing driving risk identification methods have difficulty achieving risk classification at the macro level.

[0005] Given the aforementioned shortcomings of existing technologies, there is an urgent need to research a new method for identifying highway driving risks. Summary of the Invention

[0006] The purpose of this invention is to achieve spatial positioning, classification and quantification of driving risks at a macro scale. It proposes a method for identifying highway driving risks based on speed risk potential field. This method analyzes the spatial distribution characteristics of speed, comprehensively considers the impact of speed changes and road alignment on driving risks, introduces the safety potential field theory, establishes a speed risk potential field model, and proposes a macroscopic identification method for highway driving risks.

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

[0008] A method for identifying highway driving risks based on a speed risk potential field, characterized by the following steps:

[0009] 1) Data Acquisition and Processing: Acquire and process gantry checkpoint road section monitoring data, highway accident data, and map POI data;

[0010] 2) Analyze the data collected and processed in step 1) to obtain speed data and traffic accident data;

[0011] 3) Based on the data analysis results in step 2), establish a velocity risk potential field model and determine the velocity potential energy V. i,j Velocity potential energy field strength and velocity risk potential field strength Achieve highway driving risk identification based on speed risk potential field.

[0012] Preferably, in step 1), the collected and processed gantry checkpoint road section monitoring data includes license plate, vehicle type, driving direction, time, speed, and lane; the collected and processed highway accident data includes time, station number, driving direction, accident vehicle type, and accident type; and the collected and processed map POI data includes coordinates, monitoring time, road segment duration, and travel time.

[0013] Preferably, in step 2), obtaining the speed data specifically involves: statistically analyzing the vehicle speeds in the gantry checkpoint road cross-section monitoring data within the same time period to obtain the 75th, 50th, and 25th percentile values ​​and the average value, and comparing them with the vehicle speeds in the map POI data. Simultaneously, the collected cross-sectional vehicle speeds are categorized and statistically analyzed according to lane and time. The obtained traffic accident data includes comparing the speed gradient with the distribution pattern of traffic accidents, comparing the average longitudinal slope of the road with the distribution pattern of traffic accidents, and comparing the impact of road radius on traffic accidents.

[0014] Preferably, when obtaining the impact of the comparative road radius on traffic accidents, the minimum radius R of the circular curve without superelevation is taken. * The ratio of the standard value to the road radius R is the curvature parameter α. α is positively correlated with driving risk. When R ≥ R * When the distance is considered similar to that of a straight segment, the impact on driving safety is minimal, so α is set to 1.

[0015]

[0016] Preferably, in step 3), the velocity potential energy V is determined. i,j At that time, the average speed of each road segment within the road area is used as the velocity potential energy V. i,j The velocity potential energy V at the roadside guardrail i,jIf the value is 0, and the vehicle speeds of each lane on the road segment are known, the velocity potential energy can be calculated directly using the speed data of each lane. If lane-level speed data is unavailable, the velocity potential energy can be calculated using the following formula:

[0017]

[0018] In the formula: V i,j The velocity potential energy at the j-th lane, with the statistical unit at the i-th position, is expressed in km / h. The lateral distribution coefficient is related to the number of lanes and lateral position. The specific values ​​are as follows: For a two-way four-lane road, the left lane is 0.90, the inner lane is 1.00, the lane dividing line is 0.925, the outer lane is 0.85, and the right lane is 0.80; For a two-way six-lane road, the left lane is 0.90, the inner lane is 1.00, the lane dividing line between the inner and middle lanes is 0.95, the middle lane is 0.90, the lane dividing line between the middle and outer lanes is 0.825, the outer lane is 0.75, and the right lane is 0.70.

[0019] Preferably, in step 3), the velocity risk potential field strength is determined by the following method:

[0020] U = V a，b -V c，d

[0021]

[0022]

[0023]

[0024]

[0025] In the formula: E V Let E be the spatial velocity risk potential field strength, which has magnitude and direction. x and E y These refer to the lateral and longitudinal velocity risk potential field intensities, respectively; η is the velocity risk potential field intensity coefficient; β is the longitudinal risk influence factor; i is the road slope; k is the risk direction adjustment coefficient; d x and d y These refer to lateral and longitudinal distances, respectively, with the direction pointing from the lower potential energy position to the higher potential energy position, where d x The units are m and d. y The unit is km; U is the velocity potential energy difference; V a,b and V c,dLet be the velocity potential energy at lateral positions b and d, respectively, for station a; and let α be the curvature parameter, which is the lateral risk factor.

[0026] The calculation method for the velocity risk potential field intensity coefficient η is as follows:

[0027]

[0028] When constructing the spatial velocity risk potential field strength, a risk direction adjustment coefficient k is introduced to achieve effective assessment of spatial driving risk. The calculation methods for the risk direction adjustment coefficient and the spatial velocity risk potential field strength threshold are as follows:

[0029]

[0030]

[0031] In the formula: e x ,e y ,e xy These are the risk thresholds for the lateral, longitudinal, and spatial velocity risk potential field strengths, respectively.

[0032] Compared with the prior art, the highway driving risk identification method based on speed risk potential field of the present invention has one or more of the following beneficial technical effects: The present invention analyzes the spatial distribution law of vehicle speed, combines the influence of road alignment and speed gradient on driving risk, introduces potential field theory to propose speed risk potential field theory, which can better realize driving risk identification at the macro level. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the implementation of the highway driving risk identification method based on speed risk potential field of the present invention.

[0034] Figure 2 Comparison results of vehicle speed data from different sources.

[0035] Figure 3 This is the result of the cross-sectional vehicle speed analysis.

[0036] Figure 4 The results are the analysis of speed gradient and traffic accident distribution.

[0037] Figure 5 The results show the analysis of road longitudinal slope and traffic accident distribution.

[0038] Figure 6 The results are the analysis of curvature parameters and traffic accident distribution.

[0039] Figure 7 This is the statistical result of the time distribution of the cross-sectional velocity variation coefficient.

[0040] Figure 8 The results show the potential field distribution of traffic accident and lateral speed risk.

[0041] Figure 9 The results show the potential field distribution of traffic accident and longitudinal velocity risk.

[0042] Figure 10 The results show the potential field distribution of traffic accident and spatial velocity risk.

[0043] Figure 11 The results show the potential field distribution of traffic accident and lateral speed risk.

[0044] Figure 12 The results show the potential field distribution of traffic accident and longitudinal velocity risk.

[0045] Figure 13 The results show the potential field distribution of traffic accident and spatial velocity risk. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments. The content of the embodiments is not intended to limit the scope of protection of the present invention.

[0047] Figure 1 A flowchart illustrating the implementation of the highway driving risk identification method based on speed risk potential field of the present invention is shown. Figure 1 As shown, the highway driving risk identification method based on speed risk potential field of the present invention includes the following steps:

[0048] I. Data Acquisition and Processing.

[0049] Specifically, it is necessary to collect and process gantry checkpoint road section monitoring data, highway accident data, and map POI data.

[0050] When collecting and processing road section monitoring data at gantry checkpoints, a certain monitoring duration (in days) is required. Each set of monitoring data includes license plate number, vehicle type, driving direction, time, speed, and lane information. When collecting and processing highway accident data, a certain statistical duration (in years) is required. Each set of accident data includes time, station number, driving direction, accident vehicle type, and accident type. Python programs can be used to collect Baidu Maps POI data; each set of data includes coordinates, monitoring time, road segment duration, and travel time.

[0051] Second, analyze the data collected and processed in step one to obtain speed data and traffic accident data.

[0052] This involves analyzing speed data, statistically analyzing vehicle speeds acquired by gantry checkpoints within the same time period, obtaining the 75th, 50th, and 25th percentiles and the average value, and comparing these with speeds obtained from Baidu Maps POI data. Additionally, the collected cross-sectional vehicle speeds are categorized and statistically analyzed according to lane and time.

[0053] It is necessary to analyze traffic accident data, including comparing the distribution patterns of speed gradients and traffic accidents, comparing the distribution patterns of average road longitudinal slope and traffic accidents, and comparing the impact of road radius on traffic accidents.

[0054] When comparing the impact of road radius on traffic accidents, the minimum radius R of a circular curve without superelevation is taken. * The ratio of the standard value to the road radius R is the curvature parameter α, as shown in equation (1). α is positively correlated with driving risk; when R ≥ R... * When the distance is considered similar to that of a straight segment, the impact on driving safety is minimal, so α is set to 1.

[0055]

[0056] III. Based on the data analysis results in step two, establish a velocity risk potential field model and determine the velocity potential energy V. i,j Velocity potential energy field strength and velocity risk potential field strength Achieve highway driving risk identification based on speed risk potential field.

[0057] Among them, velocity potential energy V i,j The determination of the velocity potential energy (Vi,j) requires the characteristic that the speeds of most vehicles tend to be consistent within the same time period and range. Therefore, the velocity selection trend of traffic flow at a certain location is proposed as the velocity potential energy (Vi,j). The field strength of the velocity potential energy field is the velocity gradient, which can describe the degree of velocity change in each direction within the road space. The average speed of each road segment within the road space is used as the velocity potential energy (Vi,j). i,j The velocity potential energy at the roadside guardrail is 0. When the vehicle speeds of each lane in a road section are known, the velocity potential energy can be calculated directly using the speed data of each lane; when lane-level speed data cannot be obtained, the velocity potential energy of the entire area can be calculated according to equation (2):

[0058]

[0059] In the formula: V i,j Let i be the velocity potential energy (km / h) at the j-th lane, with the statistical unit at point i. The lateral distribution coefficient is related to the number of lanes and lateral position. The specific values ​​are shown in Table 1.

[0060] Table 1. Values ​​of Lateral Distribution Coefficient

[0061]

[0062] Furthermore, based on the gradient changes of velocity potential energy along the lateral and longitudinal directions, the influence of road longitudinal slope and horizontal radius on driving risk is superimposed to form the velocity risk potential field intensity. This describes the magnitude and direction of vehicle driving risk at this location, characterizing the distribution of driving risk on the road segment. Speed ​​risk potential field intensity. The calculation method is shown in equations (3) to (7); where the curvature parameter is used as the lateral risk impact factor α, and the calculation method is shown in equation (1).

[0063] U = V a，b -V c，d (3)

[0064]

[0065]

[0066]

[0067]

[0068] In the formula: E V Let E be the spatial velocity risk potential field strength, which has magnitude and direction. x and E y These refer to the lateral and longitudinal velocity risk potential field intensities, respectively; η is the velocity risk potential field intensity coefficient; β is the longitudinal risk influence factor; i is the road slope; k is the risk direction adjustment coefficient; d x and d y These refer to lateral and longitudinal distances, respectively, with the direction pointing from the lower potential energy position to the higher potential energy position, where d x The units are m and d. y The unit is km; U is the velocity potential energy difference; V a,b and V c,d Let be the velocity potential energy (km / h) at lateral positions b and d of station a and c, respectively.

[0069] The velocity risk potential field strength coefficient η is calculated as shown in equation (8).

[0070]

[0071] In the formula: SD is the standard deviation of velocity, and CV is the coefficient of variation of velocity. The two together describe the velocity dispersion. The average speed. The coefficient of variation (CV) refers to the difference between the standard deviation of the speed and the average speed. The larger the ratio, the greater the dispersion. The greater the dispersion, the smaller the velocity risk potential field strength. Therefore, the reciprocal of the velocity variation coefficient CV is used as the velocity risk potential field strength coefficient η.

[0072] When constructing the spatial velocity risk potential field strength, a risk direction adjustment coefficient k is introduced to achieve effective assessment of spatial driving risk. The calculation methods of the risk direction adjustment coefficient and the spatial velocity risk potential field strength threshold are shown in equations (9) and (10).

[0073]

[0074]

[0075] In the formula: e x ,e y ,e xy These are the risk thresholds for the lateral, longitudinal, and spatial velocity risk potential field strengths, respectively.

[0076] The present invention will now be described in detail with reference to a specific embodiment, so that those skilled in the art can better implement the present invention.

[0077] 1) A total of 124,216 sets of road cross-section monitoring data from gantry checkpoints and 439 sets of highway accident data were collected. Four road cross-sections were measured at the gantry checkpoints: K2084, K2088, K2110, and K2114. The total monitoring period was 31 days. Each set of monitoring data included license plate number, vehicle type, driving direction, time, speed, and lane information. The selected cross-section monitoring data is shown in Table 2.

[0078] Table 2 Road section monitoring data

[0079]

[0080] 2) Collect highway accident data for a period of 4 years. Each set of accident data includes time, station number, direction of travel, vehicle type, and accident type. The selected accident data is shown in Table 3.

[0081] Table 3 Highway Accident Data

[0082]

[0083] 3) A total of 91,955 sets of Baidu Map POI data were collected using Python. Each set of data included coordinates, monitoring time, road segment length, and travel time. The average vehicle speed for each road segment was further calculated based on the road segment length and travel time. To verify the accuracy of the monitoring data, road segment lengths were set to 30km, 16km, 8km, 4km, 2km, 1km, and 0.5km. One 30km segment had its start and end points set at two gantries. Comparison with actual measured data showed that the travel time obtained for a 30km segment had an error of 16s compared to the average travel time obtained from gantry monitoring, meeting practical requirements. Subsequently, using 30km as the standard, road segment monitoring data was divided, and the travel times for 16km, 8km, 4km, and 2km segments were recursively corrected to finally obtain the speed data for a 2km segment.

[0084] 4) Based on the data collected in steps 1), 2), and 3), and considering the actual road section's speed limit of 80 km / h for both large and small vehicles, the implementation of passenger and freight separation control measures, and the outer lane being a dedicated freight lane, the following gantry checkpoints were established: K2084, located on a small-radius curve (R = 340 m); K2088, located on a continuous S-shaped curve (R = 620 m); K2110, located on a straight section; and K2114, located on a curved section (R = 870 m). The vehicle speeds obtained from the gantry checkpoints within the same time period were statistically analyzed to obtain the 75th, 50th, and 25th percentiles and the average value. These values ​​were then compared and analyzed with the speed data obtained from Baidu POI. Figure 2 As shown in the figure, the results indicate that the vehicle speeds obtained from Baidu POI are quite close to the average vehicle speeds in the inner lanes obtained from the gantry checkpoints.

[0085] The cross-sectional vehicle speeds collected at checkpoints were categorized and statistically analyzed according to lane and time, as shown in the following figures. Figure 3 As shown in Table 4, comparing the vehicle speed statistics at the four cross-sections reveals the following:

[0086] (1) The speed change trends at the four cross-sections were the same in the early morning and evening. From 7:00 to 9:00, the speed showed a clear increasing trend; from 18:00 to 20:00, the speed showed a clear decreasing trend; at other times, the speed changes at each cross-section did not show the same trend over time.

[0087] (2) The average speed and speed dispersion of the inner and outer lanes at the same cross section have basically the same trend over time. At cross section K2084, the average speed of the inner and outer lanes at 15:00 is at the peak and valley, and the period with the largest speed dispersion is at 8:00. At cross section K2088, the average speed and speed dispersion of the inner and outer lanes have the same trend from 12:00 to 17:00. At cross section K2110, except for 14:00 and 17:00, the average speed and speed dispersion of the inner and outer lanes have basically the same trend over the other time periods. At cross section K2114, except for 13:00 and 15:00, the average speed and speed dispersion of the inner and outer lanes have basically the same trend over the other time periods.

[0088] (3) The speed dispersion is higher on straight sections than on curved sections, and the speed dispersion of the inner lane is higher than that of the outer lane. The speed dispersion of the inner and outer lanes is similar on curved sections. Section K2110 is located on a straight section, and the speed standard deviation of the inner and outer lanes is much greater than that of other sections, and the speed standard deviation of the inner lane is higher than that of the outer lane.

[0089] (4) The average speed ratio of the inner and outer lanes is basically the same, close to 0.85.

[0090] Table 4. Vehicle Speed ​​Statistics at Cross-Sections

[0091]

[0092] 5) Based on the data collected in steps 1), 2), and 3), and the cross-sectional vehicle speed statistics obtained in step 4), compare the speed gradient with the distribution pattern of traffic accidents, such as... Figure 4 As shown, the speed gradient and traffic accident distribution have a certain degree of convergence, but the correlation is low in the range of K2085 to K2105.

[0093] 6) Based on the data collected in steps 1), 2), and 3), compare the average longitudinal slope of the road with the distribution pattern of traffic accidents, such as... Figure 5 As shown in the figure, the results indicate that road sections with an average longitudinal slope of less than 2% had significantly fewer accidents than other road sections. Furthermore, the section with the highest number of accidents was located in the middle of the downhill section from K2118 to K2130.

[0094] 7) Based on the data collected in steps 1), 2), and 3), compare the impact of road radius on traffic accidents, and compare the curvature parameter α with the distribution pattern of traffic accidents, such as... Figure 6 As shown in the figure, it can be seen that it has a high correlation.

[0095] 8) Based on the cross-sectional monitoring data collected in step 1), select the measured data from 10:00 to 20:00 over 28 days at cross-section K2084, and calculate the velocity variation coefficient (CV) in hourly units. The calculation results are as follows: Figure 7As shown in the figure. According to the results, the speed variation coefficient CV is basically between 0.05 and 0.17. Finally, the speed variation coefficient is taken as 0.10, and the speed risk potential field intensity coefficient η = 10 is calculated. The design speed of this section is 80km / h, and the minimum radius of the circular curve without superelevation is 2500m. The risk impact factor is calculated according to equations (1) and (6), and the calculation results are shown in Table 5.

[0096] Table 5 Calculation of Risk Impact Factors

[0097]

[0098] 9) Based on the data collected in steps 1) and 2), the speed risk potential field is calculated for the section from K2064+140 to K2094+140. The design index of this expressway is the lateral distribution coefficient, which is taken from Table 1. The speed potential energy is obtained from the Baidu speed data obtained above. A speed potential energy matrix is ​​constructed, and the speed potential energy distribution and speed risk potential field intensity distribution are calculated.

[0099] Based on the collected accident data, the section from K2064+140 to K2094+140 was further screened, with a total of 157 recorded accidents. These can be categorized into three groups based on vehicle type: 89 accidents involving small vehicles, 60 accidents involving large vehicles, and 8 accidents involving both large and small vehicles. They can also be categorized into three groups based on accident type: 48 accidents related to lateral risks (collisions with guardrails, side-impact accidents, improper operation), 29 accidents related to longitudinal risks (rear-end collisions), and 80 accidents with unclear types. Furthermore, lane information is missing from the accident data. Based on on-site investigation, passenger and freight vehicle separation control measures are implemented on this section. Generally, large vehicles travel in the outer lane, and small vehicles travel in the inner lane. Considering that vehicles generally travel close to the center line, and the common vehicle width is 1.8m, the accident points were located 0.9m to the left and right of the lane center line. Only small vehicle accidents occurred in the inner lane, while accidents involving large vehicles occurred in the outer lane. Speed ​​risk potential field intensity and related accident distribution in the K2064+140~K2094+140 section, such as Figures 8-10 As shown.

[0100] Depend on Figure 7 It can be seen that traffic accidents cluster at the peak field strength locations such as K2068, K2085, and K2088. Statistical analysis of the lateral velocity risk potential field strength Ex at the locations of 48 accidents related to lateral risk (all statistics related to velocity risk potential field strength are absolute values) shows a maximum value of 244.2, a minimum value of 16.6, and a mean of 75.5. A total of 16 accidents, accounting for 33%, had Ex values ​​greater than the mean at their locations. Figure 8It can be seen that traffic accidents cluster at the two peak field strengths at K2065 and K2088. Statistical analysis of the longitudinal velocity risk potential field strength Ey at the locations of 29 accidents related to longitudinal risk shows a maximum value of 141.1, a minimum value of 5.12, and an average value of 74.6. A total of 15 accidents, accounting for 51.72%, had Ey values ​​greater than the average.

[0101] The greater the speed risk potential field strength, the higher the probability of an accident. To effectively identify driving risks, the average speed risk potential field strength at the accident location is selected as the risk threshold, i.e., the risk threshold e of the lateral speed risk potential field strength. x The risk threshold e for the longitudinal velocity risk potential field strength is 75.5. y The value is 74.6. When the velocity risk potential field strength at a given location exceeds the risk threshold, it indicates a higher probability of a corresponding risk-related accident, i.e., a high-risk lateral or longitudinal road section. In the K2064+140~K2094+140 section, 13.85% of the section was identified as a high-risk lateral section, with 33.33% of lateral risk accidents occurring within this area; 16.26% of the section was identified as a high-risk longitudinal section, with 51.72% of longitudinal risk accidents occurring within this area.

[0102] Both lateral and longitudinal velocity risk potential fields can identify road segment driving risks from a single perspective. To better achieve road segment driving risk assessment, both lateral and longitudinal driving risks are considered comprehensively. The spatial velocity risk potential field intensity is calculated according to equation (7) and further used as a macro-risk evaluation index for the road segment. Combining equation (9), the risk direction adjustment coefficient k is taken as 1.01, and the spatial velocity risk potential field intensity E v and the distribution of related accidents, such as Figure 9 As shown.

[0103] According to equation (10), the risk threshold of the spatial velocity risk potential field is 106.8. Further statistics show that 45.86% of historical traffic accidents, 42.42% of accidents with unclear morphology, 60.07% of longitudinal risk accidents, and 40% of lateral risk accidents occurred in areas exceeding the risk threshold. The high-risk road sections identified by the spatial velocity risk potential field cover the high-risk road sections identified by the lateral and longitudinal velocity risk potential fields, and can further effectively identify high-risk road sections with complex risk structures.

[0104] Therefore, it can be seen that the speed risk potential field strength is highly correlated with traffic accidents and can spatially characterize highway driving risks to a certain extent. The comprehensive use of lateral, longitudinal and spatial speed risk potential fields can identify driving risks.

[0105] 10) Based on the risk potential field and conclusions obtained in step 9), the road segment from K2094+140 to K2130+640 was selected as the effectiveness verification segment. The above method was used to identify road segment risks and verify the effectiveness of the method. The speed risk potential field intensity coefficient η = 9.5. The total number of accidents on this segment was 282, including 56 accidents related to lateral risks, 76 accidents related to longitudinal risks, and 150 accidents with unclear morphology. The speed risk potential field intensity and related accident distribution are as follows: Figures 11-13 As shown. By Figure 11-13 It can be seen that the distribution of accidents and the distribution of speed risk potential field intensity have certain regularities, among which the distribution of lateral risk-related traffic accidents and the distribution of lateral speed risk potential field intensity have obvious regularities.

[0106] The risk thresholds were based on the aforementioned statistical results, with the thresholds for lateral, longitudinal, and spatial speed risk potential fields being 75.5, 74.6, and 106.8, respectively. A statistical comparison of the identification effectiveness of the speed risk potential field intensity was conducted, and the results are shown in Table 6. The identification benefit ratio is the ratio of the overall accident identification rate to the proportion of road sections exceeding the risk threshold. Specifically, the spatial speed risk potential field identified 24.33% of high-risk road sections, 43.26% of accidents, and a identification benefit ratio of 1.78.

[0107] 11) Based on the analysis in step 9) and the relationship between traffic accidents and risk potential fields obtained in step 10), the total number of accidents in the K2094+140~K2130+640 section within the time period was 282, with a maximum spatial speed risk potential field intensity of 297; the total number of accidents in the K2064+140~K2094+140 section was 157, with a maximum spatial speed risk potential field intensity of 260. The speed risk potential field intensity in the K2094+140~K2130+640 section is generally greater than that in the K2064+140~K2094+140 section, consistent with the overall accident volume trend.

[0108] Comparing the relevant statistical data of the validity verification road sections and the case study road sections, the results show that:

[0109] (1) The speed risk potential field strength and the distribution of related accidents have certain regularities;

[0110] (2) The three types of speed risk potential fields have a high identification rate for both road sections, and the selected risk threshold can effectively classify and identify the driving risks on the highway.

[0111] (3) Furthermore, the extreme values ​​of the speed risk potential field intensity can characterize the overall safety level of a road segment. In summary, the speed risk potential field can identify different forms of driving risks and quantify the severity of highway driving risks. Compared with traditional one-dimensional identification methods, it achieves effective characterization of driving risks in two-dimensional space.

[0112] Table 6 Comparison of the identification effect of velocity risk potential field strength

[0113]

[0114] The above embodiments of the present invention are merely examples for clearly illustrating the present invention and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. All obvious variations or modifications derived from the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A highway driving risk discrimination method based on speed risk potential field, characterized in that, Comprise the following steps: 1) data acquisition and processing: acquisition and processing of gantry road section monitoring data, highway accident data and map POI data; 2) analysis of the data collected and processed in step 1) to obtain speed data and traffic accident data; 3) According to the data analysis result in step 2), a speed risk potential field model is established to determine the speed potential V i,j , the speed potential field strength and the speed risk potential field strength The highway driving risk discrimination based on the speed risk potential field is realized. The speed potential energy V in step 3) is determined as follows i,j The average speed of each road section in the road domain is used as the speed potential energy V i,j The speed potential energy V at the roadside guardrail is 0 i,j When the speed of each lane on the road section is known, the speed potential energy is directly calculated using the speed data of each lane; when the lane-level speed data cannot be obtained, the speed potential energy can be calculated according to the following formula: wherein: V i,j is the speed potential energy at the i-th unit for the j-th lane, in km / h; is the lateral distribution coefficient, which is related to the number of lanes and the lateral position, and the lateral distribution coefficient The specific values are as follows: for a two-way four-lane, 0.90 at the left edge line, 1.00 at the inner lane, 0.925 at the lane boundary line, 0.85 at the outer lane, and 0.80 at the right edge line; for a two-way six-lane, 0.90 at the left edge line, 1.00 at the inner lane, 0.95 at the lane boundary line between the inner lane and the middle lane, 0.90 at the middle lane, 0.825 at the lane boundary line between the middle lane and the outer lane, 0.75 at the outer lane, and 0.70 at the right edge line. In step 3), the speed risk potential field intensity is determined by U = V a,b - V c,d wherein: E V is the spatial velocity risk potential field intensity, with magnitude and direction, E x and E y denote the lateral and longitudinal velocity risk potential field intensity, respectively; η is the velocity risk potential field intensity coefficient; β is the longitudinal risk influence factor; i is the road slope rate; k is the risk direction adjustment coefficient; d x and d y denote the lateral and longitudinal distance, respectively, with direction from low potential to high potential location, wherein d x is in m, d y is in km; U is the velocity potential difference; V a,b and V c,d are the velocity potential at lateral location b at a stake number and at lateral location d at c stake number, respectively, in km / h; α is the curvature parameter, i.e., the lateral risk influence factor, The calculation method of the speed risk potential field intensity coefficient η is as follows: wherein: SD is the standard deviation of speed, and CV is the coefficient of variation of speed, both of which together describe the speed dispersion; is the average speed; When constructing the spatial speed risk potential field intensity, the risk direction adjustment coefficient k is introduced to realize effective evaluation of the spatial driving risk, and the risk direction adjustment coefficient and the calculation method of the spatial speed risk potential field intensity threshold value are as follows: where: e x ,e y ,e xy are the risk threshold values for the lateral, longitudinal and spatial velocity risk potential field strength, respectively.

2. The highway driving risk discrimination method based on velocity risk potential field according to claim 1, characterized in that, In the step 1), the collected and processed gantry road section monitoring data include license plate, vehicle type, driving direction, time, speed and lane; the collected and processed highway accident data include time, stake number, driving direction, accident vehicle type and accident type; and the collected and processed map POI data include coordinates, monitoring time, road section duration and travel time.

3. The highway driving risk discrimination method based on velocity risk potential field according to claim 2, characterized in that, In the step 2), the obtained speed data are specifically: the vehicle speeds in the gantry road section monitoring data within the same time are counted to obtain 75%, 50% and 25% quantile values and an average value, and are compared with the vehicle speeds in the map POI data, and the collected cross section speeds are classified and counted according to lanes and time; and the obtained traffic accident data include comparison of speed gradient and distribution of traffic accidents, comparison of road average longitudinal slope and distribution of traffic accidents and comparison of road radius and influence on traffic accidents.

4. The highway driving risk discrimination method based on velocity risk potential field according to claim 3, characterized in that, When the influence of the comparative road radius on the traffic accidents is obtained, the minimum radius R of the circular curve without super-elevation is taken * The ratio of the specification value to the road radius R is the curvature parameter α, and α is positively correlated with the driving risk. When R≥R * , it is considered to be similar to the straight section and has little influence on the driving safety, and α is taken as 1.

Images