Left-turn human-vehicle interaction risk assessment method and system based on unmanned aerial vehicle video

Abstract

The invention discloses a left-turn human-vehicle interaction risk assessment method and system based on an unmanned aerial vehicle video, and relates to the technical field of traffic safety. Starting from three aspects of a man-vehicle blind area field, a potential energy field and a future possible operation space field, the safety risk of a man-vehicle interaction scene in the left turning process of a vehicle is comprehensively evaluated; video data are collected in real time through the unmanned aerial vehicle technology, and the traffic data collection efficiency and accuracy are improved. A target detection and tracking algorithm is utilized to more accurately extract the feature information of the vehicle and the pedestrian, so that the evaluation precision of the left-turn human-vehicle interaction risk is improved, and the human-vehicle interaction condition is fed back in real time; based on the blind area field risk quantity, the potential energy field risk quantity and the future possible operation space field risk quantity, a left-turn human-vehicle interaction risk assessment result is determined, mutual influence and action mechanisms of different factors are comprehensively considered, scientific basis and suggestions can be provided for traffic management departments and related decision makers, the traffic accident rate is reduced, and the risk assessment efficiency is improved. And the driving safety is improved.

Description

Technical Field

[0001] The present invention relates to the field of traffic safety technology, and in particular to a left-turn human-vehicle interaction risk assessment method based on drone video and a system thereof. Background Art

[0002] With the acceleration of urbanization and the diversification of transportation, traffic safety issues are becoming increasingly serious. According to a 2023 report by the World Health Organization, about 1.2 million people die in road traffic accidents each year, more than half of whom are vulnerable road users, and the losses caused by traffic accidents account for 3% of the GDP of most countries.

[0003] Speeding, poor enforcement of traffic laws, distracted driving, unsafe road traffic facilities, etc. are important causes of traffic accidents. Vehicle left turn, as a common traffic behavior, involves multiple factors and is complex and diverse.

[0004] Although there are studies on left-turn vehicle driving behavior and risk quantification assessment, there are deficiencies in the research on human-vehicle interaction models and risk quantification assessment. The industry's human-vehicle interaction risk assessment methods are mainly divided into deterministic assessment methods and probability-based assessment methods. Deterministic assessment methods are usually based on a series of preset conditions and rules, and use logical judgment to determine whether a risk exists or not, but do not fully consider the dynamic changes and uncertainties in actual traffic flows, and the mutation and randomness of pedestrian movement states. Probability-based methods have become the focus of research. Although uncertainty is taken into account, they may not be able to respond to immediate changes in the traffic environment in real time. Existing research on human-vehicle interaction models based on "field" theory has limitations, and there is insufficient research on risk assessment for the superposition of multiple fields. There is little research on the comprehensive consideration of pedestrian characteristics and driver characteristics. Summary of the invention

[0005] The purpose of the present invention is to provide a left-turn human-vehicle interaction risk assessment method and system based on drone video to improve the above technical problems.

[0006] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0007] A left-turn human-vehicle interaction risk assessment method based on drone video, comprising:

[0008] S1. Construct a vehicle state model and a pedestrian state model, and obtain the initial parameters of the vehicle and the initial parameters of the pedestrian;

[0009] S2, inputting the initial parameters of the vehicle into the vehicle state model, and outputting corresponding vehicle driving data;

[0010] S3, inputting the initial parameters of the pedestrian into the pedestrian state model, and outputting corresponding pedestrian action data;

[0011] S4, constructing a human-vehicle blind spot field coordinate system, and calculating a corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data;

[0012] S5, constructing a protection zone model, and calculating the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data;

[0013] S6. Collect pedestrian-vehicle interaction video data through drones, generate corresponding possible future operation space fields, and calculate the corresponding possible future operation space field risk amount;

[0014] S7. Determine a left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

[0015] Furthermore, the formula of the vehicle state model is:

[0016] ;

[0017] in, , They represent the vehicle at time in the road coordinate system. The X and Y coordinates of Indicates time The vehicle yaw angle, Indicates time The arc length of the vehicle's motion trajectory, Indicates the vehicle's speed. Indicates the vehicle running acceleration, represents the curvature of the vehicle turning trajectory, Indicates time The straight-line distance corresponding to the arc length of the vehicle's motion trajectory, represents pi, represents the sine function, Indicates time The angle corresponding to the arc length of the vehicle's motion trajectory, represents the cosine function, , Respectively indicate time The X and Y coordinates of the vehicle in the road coordinate system, Indicates the turning radius, Indicates time The speed of the vehicle, Indicates straight-line distance The angle with the positive direction of the X axis of the road coordinate system.

[0018] Furthermore, the formula of the pedestrian state model is:

[0019] ;

[0020] in, Indicates the time difference of pedestrian movement. , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time To time The speed difference between the X and Y directions.

[0021] Furthermore, the initial parameters of the vehicle include the initial state of the vehicle and turning radius The initial parameters of the pedestrian include the position information and speed information of the pedestrian in the X and Y directions in the initial state; the vehicle driving data includes the time Vehicle position information in the road coordinate system ,speed and yaw angle ; The pedestrian action data is the pedestrian status .

[0022] Furthermore, the constructing of the human-vehicle blind spot field coordinate system and calculating the corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data includes:

[0023] S4-1. Blind spots of pedestrians during left-turn human-vehicle interaction and the driver's blind spot , determine the overlap area of ​​human-vehicle blind spot interaction ;

[0024] S4-2. Based on the time Vehicle position information in the road coordinate system , determine the distance factor risk function according to the Gaussian distribution form ;

[0025] S4-3, based on the speed and visibility , determine the risk function for vehicle speed and visibility ;

[0026] S4-4. Risk function based on the distance factor and the risk function of vehicle speed and visibility , determine the comprehensive risk function ;

[0027] S4-5. Based on the comprehensive risk function , determine the risk amount of the blind area .

[0028] Furthermore, the formula of the distance factor risk function is:

[0029] ;

[0030] in, represents the amplitude, , represents the mean, , represents the standard deviation, represents a natural constant;

[0031] The formula for the risk function of vehicle speed and visibility is:

[0032] ;

[0033] in, , represents a constant, Indicates visibility;

[0034] The formula of the comprehensive risk function is:

[0035] ;

[0036] in, represents the adjustment parameter;

[0037] The formula for the blind spot risk is:

[0038] ;

[0039] in, , represents the double integral, , They represent the length from the risk point to the origin of the polar coordinates and the angle between the risk point and the X-axis. The comprehensive risk function The polar coordinate expression of .

[0040] Furthermore, the construction of the protection zone model and calculation of the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data include:

[0041] S5-1, setting vehicle protection zone parameters during left-turn human-vehicle interaction, and creating a vehicle protection zone model based on a rectangle and a semicircle;

[0042] S5-2. Based on the vehicle protection zone model, limit the vehicle potential field generation range corresponding to the vehicle, and according to the formula:

[0043] ;

[0044] Get the vehicle potential field risk within the vehicle potential field generation range ;in, Indicates that the vehicle is at time The actual driving speed is , Indicates that the vehicle is at time The upper speed limit, represents the logarithmic function with the natural constant e as the base, represents the vehicle field strength coefficient, Indicates that the vehicle is at time The angle between the velocity vector and the vehicle's center of mass, represents the pseudo distance, represents the absolute value, represents the rate of change of the control potential field strength with distance, Indicates visibility;

[0045] S5-3, setting the parameters of the pedestrian protection zone during the left-turn human-vehicle interaction process, and creating a pedestrian protection zone model;

[0046] S5-4. Based on the pedestrian protection area model, limit the pedestrian potential field generation range corresponding to the pedestrian, and according to the formula:

[0047] ;

[0048] Obtain the pedestrian potential field risk within the pedestrian potential field generation range ;

[0049] S5-5, according to the formula:

[0050] ;

[0051] Get the potential energy field risk ;

[0052] The protection zone model includes the vehicle protection zone model and the pedestrian protection zone model.

[0053] Furthermore, the collecting of pedestrian-vehicle interaction video data by drone, generating a corresponding possible future operation space field, and calculating the corresponding possible future operation space field risk amount include:

[0054] S6-1, collecting pedestrian-vehicle interaction video data through a drone, and dividing it into a vehicle image sequence and a pedestrian image sequence of multiple frames respectively;

[0055] S6-2, using Kalman filtering to continuously track the vehicles and pedestrians in the vehicle image sequence and the pedestrian image sequence, respectively, and assign corresponding IDs;

[0056] S6-3, based on each of the IDs outputted in S6-2, the vehicle image sequence and the pedestrian image sequence are processed respectively by a target detection and tracking algorithm to obtain corresponding vehicle information and pedestrian information;

[0057] The vehicle information includes the vehicle speed, the vehicle direction and the vehicle lane; the pedestrian information includes the pedestrian position information, the pedestrian crossing speed and the pedestrian movement trajectory;

[0058] S6-4, based on the vehicle information, generating a corresponding possible future operation space of the vehicle;

[0059] S6-5, based on the pedestrian information, generating a corresponding possible future movement space of the pedestrian;

[0060] S6-6, recording the overlapping area of ​​the possible future running space of the vehicle and the possible future running space of the pedestrian as the possible future running area , calculate the possible future operating area The plane projection area , and obtain the possible future operation space field risk ;

[0061] The possible future operating space field includes the possible future operating space of vehicles and the possible future operating space of pedestrians.

[0062] Furthermore, based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount, the left turn human-vehicle interaction risk assessment result is determined, and the formula is:

[0063] ;

[0064] in, , , represents the weight coefficient, , , They represent the blind area risk, potential energy risk and future possible operation space risk respectively. Indicates the risk assessment result of left-turn human-vehicle interaction.

[0065] A left-turn human-vehicle interaction risk assessment system based on drone video is provided, which includes a pedestrian-vehicle data acquisition module, a vehicle status module, a pedestrian status module, a blind spot field risk quantity calculation module, a potential energy field risk quantity calculation module, a future possible operation space field risk quantity calculation module and a left-turn human-vehicle interaction risk assessment module; wherein:

[0066] A pedestrian-vehicle data acquisition module, used to obtain the initial parameters of the vehicle and the initial parameters of the pedestrian, and to shoot the pedestrian-vehicle interaction video data through a drone;

[0067] A vehicle state module, used for processing the initial parameters of the vehicle to obtain the vehicle driving data;

[0068] A pedestrian state module, used for processing the initial parameters of the pedestrian to obtain the pedestrian action data;

[0069] A blind spot risk calculation module, used to construct the human-vehicle blind spot coordinate system and calculate the corresponding blind spot risk based on the pedestrian action data and the vehicle driving data;

[0070] A potential energy field risk amount calculation module, used to construct the protection zone model and calculate the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data;

[0071] A future possible operation space field risk amount calculation module is used to process pedestrian-vehicle interaction video data through Kalman filtering and target detection and tracking algorithms to generate the corresponding future possible operation space field and calculate the corresponding future possible operation space field risk amount;

[0072] The left-turn human-vehicle interaction risk assessment module is used to calculate the left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

[0073] The beneficial effects of the present invention are:

[0074] 1. The present invention comprehensively evaluates the safety risk of vehicles during left turns from three aspects: the blind spot field between people and vehicles, the potential energy field, and the possible future operation space field, providing a comprehensive perspective for safety risk assessment;

[0075] 2. The present invention determines the overlapped area of ​​human-vehicle blind spot interaction based on the blind spot vision of pedestrians and vehicle drivers during the left-turn human-vehicle interaction process, calculates the risk function of distance factor, the risk factors of vehicle speed and visibility, and comprehensively selects the comprehensive risk function to assist in improving the accuracy of the left-turn human-vehicle interaction risk assessment;

[0076] 3. The present invention proposes a protection zone model that integrates rectangles and semicircles, limits the potential field generation range of vehicles and pedestrians, and calculates the potential field risk, further improving the accuracy of left-turn human-vehicle interaction risk assessment;

[0077] 4. The present invention uses drone technology to collect video data in real time, thereby improving the efficiency and accuracy of traffic data collection; and uses target detection and tracking algorithms to more accurately extract feature information of vehicles and pedestrians, improve the accuracy of left-turn human-vehicle interaction risk assessment, and provide real-time feedback on human-vehicle interaction conditions;

[0078] 5. The present invention determines the left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operating space field risk amount, and comprehensively considers the mutual influence and action mechanism of different factors, which can provide scientific basis and suggestions for traffic management departments and relevant decision makers, reduce the incidence of traffic accidents and improve driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0080] Figure 1 is a flow chart of a method in an embodiment of the present invention;

[0081] Figure 2 A road coordinate system diagram in an embodiment of the present invention;

[0082] Figure 3 is a schematic diagram of a vehicle motion trajectory in an embodiment of the present invention;

[0083] Figure 4 Schematic diagram of the human-vehicle blind spot interaction overlap area in an embodiment of the present invention;

[0084] Figure 5 A schematic diagram of a vehicle protection zone model in an embodiment of the present invention;

[0085] Figure 6 Schematic diagram of a pedestrian protection zone model in an embodiment of the present invention;

[0086] Figure 7 4 is a system structure diagram in an embodiment of the present invention. DETAILED DESCRIPTION

[0087] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0088] See also Figure 1 , this embodiment provides a left-turn human-vehicle interaction risk assessment method based on drone video, which includes:

[0089] S1. Construct a vehicle state model and a pedestrian state model, and obtain the initial parameters of the vehicle and the initial parameters of the pedestrian;

[0090] like Figure 2 As shown in FIG. 1 , a road coordinate system is established based on a fixed pedestrian crossing scene, and a constant turning curvature and CAA acceleration model are selected to describe vehicle motion, that is, to define the vehicle state space.

[0091] The vehicle at time The state space is defined as Among them, the main factors affecting the vehicle state include vehicle acceleration, steering and braking operations. Corresponding to the CAA acceleration model, the turning radius and vehicle acceleration As the input of CAA acceleration model. When it is greater than 0, it indicates that the vehicle is in the acceleration process; when the vehicle is running at an acceleration When it is less than 0, it indicates that the vehicle is in the process of deceleration.

[0092] During vehicle driving, at The curvature of the vehicle turning trajectory ,speed and yaw rate The formula is:

[0093] ;

[0094] Thus, from the moment To time During the vehicle driving process, the state space of the corresponding vehicle in the continuous motion state is:

[0095] ;

[0096] in, represents the integral function, Indicates time The state space.

[0097] like Figure 3 As shown, the vehicle travels from point A to point B, that is, time To time The vehicle turns left during driving, then the corresponding time The arc length of the vehicle's motion trajectory The formula is:

[0098] ;

[0099] time The arc length of the vehicle's motion trajectory Corresponding angle The formula is:

[0100] ;

[0101] time The arc length of the vehicle's motion trajectory Corresponding straight-line distance The formula is:

[0102] ;

[0103] Straight line distance The positive angle with the X axis of the road coordinate system The formula is:

[0104] ;

[0105] Based on the above parameters, the time can be determined The formulas for the X and Y coordinates of the vehicle in the road coordinate system are:

[0106] ;

[0107] Based on the above, the formula of the vehicle state model is:

[0108] ;

[0109] in, , They represent the vehicle at time in the road coordinate system. The X and Y coordinates of Indicates time The vehicle yaw angle, Indicates time The arc length of the vehicle's motion trajectory, Indicates the vehicle's speed. Indicates the vehicle running acceleration, represents the curvature of the vehicle turning trajectory, Indicates time The straight-line distance corresponding to the arc length of the vehicle's motion trajectory, represents pi, represents the sine function, Indicates time The angle corresponding to the arc length of the vehicle's motion trajectory, represents the cosine function, , Respectively indicate time The X and Y coordinates of the vehicle in the road coordinate system, Indicates the turning radius, Indicates time The speed of the vehicle, Indicates straight-line distance The angle with the positive direction of the X axis of the road coordinate system.

[0110] It can be seen that the input of the vehicle model is the initial state of the vehicle and turning radius , the output is the time Vehicle position information in the road coordinate system ,speed and yaw angle .

[0111] The state of the pedestrian in the motion process is considered to include the pedestrian's position information and pedestrian speed information. The uncertainty of the pedestrian's motion process is manifested as the change in the magnitude and direction of the pedestrian's speed in the X and Y directions, which are independent of each other. Therefore, the pedestrian state can be described as:

[0112] ;

[0113] On the X-axis, from time To time The motion expression of the pedestrian state is:

[0114] ;

[0115] ;

[0116] ;

[0117] ;

[0118] ;

[0119] in, represents the time constant adjustment coefficient of the first-order Markov process in the X direction, It represents the average speed of pedestrians moving in the X direction, in m / s, and follows a Gaussian distribution. The speed will vary up and down. Indicates the mean in the X direction is 0, and the variance is Gaussian white noise, in m / s, It means the mean is 0 and the variance is Gaussian white noise distribution, Indicates the mean , the variance is Gaussian distribution, Represents the variance of the Gaussian distribution.

[0120] On the Y axis, from time To time The motion expression of the pedestrian state is:

[0121] ;

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] in, represents the time constant adjustment coefficient of the first-order Markov process in the Y direction, It represents the average speed of pedestrians moving in the Y direction, in m / s, and obeys Gaussian distribution. The speed will vary up and down. Indicates the mean in the Y direction is 0, and the variance is Gaussian white noise, in m / s, It means the mean is 0 and the variance is Gaussian white noise Sound distribution, represents the variance of the Gaussian distribution, Indicates the mean , a Gaussian distribution with variance .

[0127] Therefore, the formula of the pedestrian state model is:

[0128] ;

[0129] in, Indicates the time difference of pedestrian movement. , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time To time The speed difference between the X and Y directions.

[0130] The initial parameters of the vehicle include the initial state of the vehicle and turning radius The initial parameters of the pedestrian include the position information and speed information of the pedestrian in the X and Y directions in the initial state.

[0131] S2, inputting the initial parameters of the vehicle into the vehicle state model, and outputting corresponding vehicle driving data;

[0132] The vehicle driving data includes time Vehicle position information in the road coordinate system ,speed and yaw angle .

[0133] S3, inputting the initial parameters of the pedestrian into the pedestrian state model, and outputting corresponding pedestrian action data;

[0134] The pedestrian action data is the pedestrian status The unit of speed information is m / s.

[0135] S4, constructing a human-vehicle blind spot field coordinate system, and calculating a corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data;

[0136] The constructing of a human-vehicle blind spot field coordinate system and calculating a corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data includes:

[0137] S4-1. Blind spots of pedestrians during left-turn human-vehicle interaction and the driver's blind spot , determine the overlap area of ​​human-vehicle blind spot interaction ;

[0138] During the left-turn human-vehicle interaction, both the driver and the pedestrian have blind spots. Figure 4 As shown, the human-vehicle blind spot interaction overlap area is determined by establishing a coordinate system ; Blind spots of pedestrian vision and the driver's blind spot They are all fan-shaped areas, and their corresponding coordinates are shown in Table 1:

[0139] Table 1

[0140]

[0141] in, , Respectively represent rectangular coordinates, , represents the rectangular coordinates of pedestrians, , represents the vehicle coordinates, It represents the radius of the circle obtained by converting the pedestrian's rectangular coordinates to the limit coordinates. It represents the length from the pedestrian position in polar coordinates to the origin of polar coordinates. It represents the angle between the pedestrian position and the X-axis in polar coordinates. , They represent the minimum and maximum angles between the pedestrian position and the X-axis in polar coordinates, respectively. It represents the radius of the circle obtained by converting the vehicle rectangular coordinates to the limit coordinates. Indicates the length from the risk point to the origin of the polar coordinates. It represents the angle between the risk point and the X-axis in polar coordinates. , They represent the minimum and maximum values ​​of the angle between the risk point and the X-axis in polar coordinates. The risk point is the position where pedestrians are prone to accidents when the vehicle turns left.

[0142] When the vehicle turns left, the vehicle's own structure blocks the driver's line of sight. When the vehicle turns left, the A-pillar on the left side of the front of the vehicle may block the driver's line of sight within a certain angle range on the left side, making it impossible for the driver to directly see the blocked area. This blocked area is the blind spot caused by the line of sight being blocked by the A-pillar, that is, the driver's field of vision blind spot .

[0143] The height, standing position and sight direction of pedestrians will affect the formation of their blind spots. Turning their heads or lowering their heads will also change their sight range, causing vehicles to enter their blind spots without them noticing. This is the pedestrian's blind spot. .

[0144] S4-2. Based on the time Vehicle position information in the road coordinate system , determine the distance factor risk function according to the Gaussian distribution form ;

[0145] The formula of the distance factor risk function is:

[0146] ;

[0147] in, represents the amplitude, , represents the mean, , represents the standard deviation, represents a natural constant;

[0148] S4-3, based on the speed and visibility , determine the risk function for vehicle speed and visibility ;

[0149] The formula for the risk function of vehicle speed and visibility is:

[0150] ;

[0151] in, , represents a constant, Indicates visibility; when speed As , the risk function As visibility increases As , the risk function Then it decreases;

[0152] As a combined effect of speed and visibility, high speed and low visibility significantly increase the risk. , The term represents the linear effect of speed on risk, where the higher the speed, the greater the risk. If speed has a greater impact on risk than visibility, increase If visibility has a greater impact on risk, increase The value of .

[0153] S4-4. Risk function based on the distance factor and the risk function of vehicle speed and visibility , determine the comprehensive risk function ;

[0154] The formula of the comprehensive risk function is:

[0155] ;

[0156] in, represents the adjustment parameter;

[0157] S4-5. Based on the comprehensive risk function , determine the risk amount of the blind area .

[0158] The formula for the blind spot risk is:

[0159] ;

[0160] in, , represents the double integral, , They represent the length from the risk point to the origin of the polar coordinates and the angle between the risk point and the X-axis. The comprehensive risk function The polar coordinate expression of .

[0161] S5, constructing a protection zone model, and calculating the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data;

[0162] The image sensor uses drones to collect video, computers and communication technology to collect traffic information such as traffic flow, vehicle speed, and vehicle interval at a designated location. It is suitable for analyzing and processing continuous video images to obtain traffic flow information. A protection zone based on speed direction is established based on the image sensor test distance to limit the generation range of the potential field, that is, the construction of the protection zone model, and based on the pedestrian action data and the vehicle driving data, the corresponding potential field risk is calculated, including:

[0163] S5-1. Set the vehicle protection zone parameters during the left-turn human-vehicle interaction process, and Figure 5 As shown in the figure, from the perspective of vehicle operation risk, the risk in the vehicle speed direction is significantly higher than the rear risk, so a rectangle and a semicircle are selected to create a vehicle protection zone model;

[0164] Among them, the vehicle protection area parameters include three parameters: a1=1.8m, b1=4.0m, and R1=0.9m.

[0165] S5-2. Based on the vehicle protection zone model, limit the vehicle potential field generation range corresponding to the vehicle, and according to the formula:

[0166] ;

[0167] Get the vehicle potential field risk within the vehicle potential field generation range ;in, Indicates that the vehicle is at time The actual driving speed is , Indicates that the vehicle is at time The upper speed limit, represents the logarithmic function with the natural constant e as the base, represents the vehicle field strength coefficient, Indicates that the vehicle is at time The angle between the velocity vector and the vehicle's center of mass, represents the pseudo distance, represents the absolute value, represents the rate of change of the control potential field strength with distance, Indicates visibility;

[0168] In practice, the potential field generated by vehicles and pedestrians is not only related to the distance, but also to their motion state. It is generally believed that there is a higher potential field strength in the positive direction of the vehicle and pedestrian's velocity. In order to better describe this behavior of following the velocity, a pseudo-distance is introduced. The concept of pseudo-range The calculation formula is:

[0169] ;

[0170] In order to eliminate the magnitude error caused by different plane measurement standards of vehicles and pedestrians, a standard distance is added to each direction of the pseudo-distance. Make corrections. , Respectively represent the coordinates of the vehicle or pedestrian in the X direction and the Y direction, , Represents the planar position of the center of mass of the potential field source vehicles and pedestrians.

[0171] If the plane point is outside the influence area, then In the protection zone model, a "plane point" refers to any position in two-dimensional space, which is used to represent the position of a vehicle or a pedestrian, and may be located inside or outside the protection zone. The "influence zone" refers to the area covered by the potential field protection zone generated by the vehicle or pedestrian. By judging whether a plane point is located in the influence zone, it can be determined whether there is a potential risk.

[0172] S5-3, set the parameters of the pedestrian protection zone during the left-turn human-vehicle interaction process, such as Figure 6 As shown, create a pedestrian protection area model;

[0173] Among them, the pedestrian protection area parameters include two parameters: a2=0.4m, b2=0.6m.

[0174] S5-4. Based on the pedestrian protection area model, limit the pedestrian potential field generation range corresponding to the pedestrian, and according to the formula:

[0175] ;

[0176] Obtain the pedestrian potential field risk within the pedestrian potential field generation range ; If the plane point is outside the influence area, then .

[0177] S5-5, according to the formula:

[0178] ;

[0179] Get the potential energy field risk ;

[0180] The protection zone model includes the vehicle protection zone model and the pedestrian protection zone model.

[0181] S6. Collect pedestrian-vehicle interaction video data through drones, generate corresponding possible future operation space fields, and calculate the corresponding possible future operation space field risk amount;

[0182] The method of collecting pedestrian-vehicle interaction video data through a drone, generating a corresponding possible future operation space field, and calculating the corresponding possible future operation space field risk amount includes:

[0183] S6-1, collecting pedestrian-vehicle interaction video data through a drone, and dividing it into a vehicle image sequence and a pedestrian image sequence of multiple frames respectively;

[0184] S6-2, using Kalman filtering to continuously track the vehicles and pedestrians in the vehicle image sequence and the pedestrian image sequence respectively, and assign corresponding IDs to ensure that the target can be accurately tracked;

[0185] S6-3, based on each of the IDs outputted in S6-2, the vehicle image sequence and the pedestrian image sequence are processed respectively by a target detection and tracking algorithm to obtain corresponding vehicle information and pedestrian information;

[0186] The vehicle information includes the vehicle speed, the vehicle direction and the vehicle lane; the pedestrian information includes the pedestrian position information, the pedestrian crossing speed and the pedestrian movement trajectory;

[0187] The target detection and tracking algorithm may adopt any one of the existing target detection and target tracking algorithms.

[0188] S6-4, based on the vehicle information, generating a corresponding possible future operation space of the vehicle;

[0189] Assume that the vehicle follows an arc during a left turn. The length of the arc traveled by the vehicle at the steering angle is the yaw angle and turning radius The product of yaw angle The unit is radians.

[0190] Set the research vehicle to start from the initial position Start with yaw angle Turn left, the vehicle at time Location The expression is:

[0191] ;

[0192] in, Indicates the vehicle's speed.

[0193] If the vehicle is at time If the steering is performed at , the new direction angle is , the corresponding position for:

[0194] ;

[0195] The possible future operation space of the vehicle is represented as a fan-shaped area, with the center at the initial position of the vehicle. , the radius is (Maximum time The initial direction is , assuming the maximum steering angle range of the vehicle is , then the future operation space of the vehicle can be expressed as:

[0196] ;

[0197] in, Represents the inverse tangent function.

[0198] Based on the vehicle information, the possible future operation space of pedestrians is modeled, and the operation space can be represented as a sector area, where the sector angle range is .

[0199] S6-5, based on the pedestrian information, generating a corresponding possible future movement space of the pedestrian;

[0200] Pedestrians at the moment The maximum walking distance is , which is the radius of the sector area. The direction of the sector is determined by the current speed direction. and allowable angular offset OK, that is, pedestrians can Walk in any direction between. Indicates the pedestrian crossing speed. , Pedestrian crossing speed The speed in the Y and X directions.

[0201] At any time, the possible coordinates of the pedestrian are affected by the angle and radius The pedestrian is affected by The possible positions within can be expressed as:

[0202] ;

[0203] The possible future positions of pedestrians must satisfy:

[0204] ;

[0205] in, , represents the initial position of the pedestrian, represents the maximum moment, Indicates the direction of the pedestrian's initial motion speed.

[0206] Based on the above process, the possible future operation space of pedestrians is defined as a is the vertex and the radius is , the direction range is sector area.

[0207] S6-6, recording the overlapping area of ​​the possible future running space of the vehicle and the possible future running space of the pedestrian as the possible future running area , calculate the possible future operating area The plane projection area , and obtain the possible future operation space field risk ;

[0208] The possible future operating space field includes the possible future operating space of vehicles and the possible future operating space of pedestrians.

[0209] S7. Determine a left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

[0210] The left-turn human-vehicle interaction risk assessment result is determined based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount, and the formula is:

[0211] ;

[0212] in, , , represents the weight coefficient, , , They represent the blind area risk, potential energy risk and future possible operation space risk respectively. Indicates the risk assessment result of left-turn human-vehicle interaction.

[0213] like Figure 7 As shown, a left-turn human-vehicle interaction risk assessment system based on drone video includes a pedestrian-vehicle data acquisition module, a vehicle status module, a pedestrian status module, a blind spot field risk quantity calculation module, a potential energy field risk quantity calculation module, a future possible operation space field risk quantity calculation module and a left-turn human-vehicle interaction risk assessment module; wherein:

[0214] A pedestrian-vehicle data acquisition module, used to obtain the initial parameters of the vehicle and the initial parameters of the pedestrian, and to shoot the pedestrian-vehicle interaction video data through a drone;

[0215] A vehicle state module, used for processing the initial parameters of the vehicle to obtain the vehicle driving data;

[0216] A pedestrian state module, used for processing the initial parameters of the pedestrian to obtain the pedestrian action data;

[0217] A blind spot risk calculation module, used to construct the human-vehicle blind spot coordinate system and calculate the corresponding blind spot risk based on the pedestrian action data and the vehicle driving data;

[0218] A potential energy field risk amount calculation module, used to construct the protection zone model and calculate the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data;

[0219] A future possible operation space field risk amount calculation module is used to process pedestrian-vehicle interaction video data through Kalman filtering and target detection and tracking algorithms to generate the corresponding future possible operation space field and calculate the corresponding future possible operation space field risk amount;

[0220] The left-turn human-vehicle interaction risk assessment module is used to calculate the left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

[0221] In summary, the present invention starts from three aspects: human-vehicle blind spot field, potential energy field and future possible operation space field, comprehensively evaluates the safety risk of vehicles during left turns, and provides a comprehensive safety risk assessment perspective; based on the blind spot field of pedestrians and vehicle drivers during left-turn human-vehicle interaction, the human-vehicle blind spot interaction overlap area is determined, the distance factor risk function, the vehicle speed and visibility risk factors are calculated, and the comprehensive risk function is comprehensively selected to assist in improving the accuracy of left-turn human-vehicle interaction risk assessment; a protection zone model that integrates rectangles and semicircles is proposed to limit the potential field generation range of vehicles and pedestrians, and the potential energy field risk amount is calculated to further improve the accuracy of left-turn human-vehicle interaction risk assessment. Accuracy; real-time video data is collected through drone technology to improve the efficiency and accuracy of traffic data collection; and the target detection and tracking algorithm is used to more accurately extract the characteristic information of vehicles and pedestrians, improve the accuracy of left-turn human-vehicle interaction risk assessment, and provide real-time feedback on human-vehicle interaction conditions; based on the blind spot field risk amount, the potential energy field risk amount and the future possible operating space field risk amount, the left-turn human-vehicle interaction risk assessment result is determined, and the mutual influence and action mechanism of different factors are comprehensively considered, which can provide scientific basis and suggestions for traffic management departments and relevant decision makers, and can be applied to different traffic environments and conditions, predict potential risk points, reduce the incidence of traffic accidents, and improve driving safety.

[0222] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A left-turn human-vehicle interaction risk assessment method based on drone video, characterized in that: include: S1. Construct a vehicle state model and a pedestrian state model, and obtain the initial parameters of the vehicle and the initial parameters of the pedestrian; S2, inputting the initial parameters of the vehicle into the vehicle state model, and outputting corresponding vehicle driving data; S3, inputting the initial parameters of the pedestrian into the pedestrian state model, and outputting corresponding pedestrian action data; S4, constructing a human-vehicle blind spot field coordinate system, and calculating a corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data; S5, constructing a protection zone model, and calculating the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data; S6. Collect pedestrian-vehicle interaction video data through drones, generate corresponding possible future operation space fields, and calculate the corresponding possible future operation space field risk amount; S7. Determine a left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

2. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 1, characterized in that: The formula of the vehicle state model is: ； in, , They represent the vehicle at time in the road coordinate system. The X and Y coordinates of Indicates time The vehicle yaw angle, Indicates time The arc length of the vehicle's motion trajectory, Indicates the vehicle's speed. Indicates the vehicle running acceleration, represents the curvature of the vehicle turning trajectory, Indicates time The straight-line distance corresponding to the arc length of the vehicle's motion trajectory, represents pi, represents the sine function, Indicates time The angle corresponding to the arc length of the vehicle's motion trajectory, represents the cosine function, , Respectively indicate time The X and Y coordinates of the vehicle in the road coordinate system, Indicates the turning radius, Indicates time The speed of the vehicle, Indicates straight-line distance The angle with the positive direction of the X axis of the road coordinate system.

3. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 2, characterized in that: The formula of the pedestrian state model is: ； in, Indicates the time difference of pedestrian movement. , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The position information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time The speed information of the pedestrian in the X and Y directions, , Respectively indicate time To time The speed difference between the X and Y directions.

4. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 3, characterized in that: The initial parameters of the vehicle include the initial state of the vehicle and turning radius The initial parameters of the pedestrian include the position information and speed information of the pedestrian in the X direction and Y direction in the initial state; The vehicle driving data includes time Vehicle position information in the road coordinate system ,speed and yaw angle ; The pedestrian action data is the pedestrian status .

5. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 4, characterized in that: The constructing of a human-vehicle blind spot field coordinate system and calculating a corresponding blind spot field risk amount based on the pedestrian action data and the vehicle driving data includes: S4-1. Blind spots of pedestrians during left-turn human-vehicle interaction and the driver's blind spot , determine the overlap area of ​​human-vehicle blind spot interaction ; S4-2. Based on the time Vehicle position information in the road coordinate system , determine the distance factor risk function according to the Gaussian distribution form ; S4-3, based on the speed and visibility , determine the risk function for vehicle speed and visibility ; S4-4. Risk function based on the distance factor and the risk function of vehicle speed and visibility , determine the comprehensive risk function ; S4-5. Based on the comprehensive risk function , determine the risk amount of the blind area .

6. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 5, characterized in that: The formula of the distance factor risk function is: ； in, represents the amplitude, , represents the mean, , represents the standard deviation, represents a natural constant; The formula for the risk function of vehicle speed and visibility is: ； in, , represents a constant, Indicates visibility; The formula of the comprehensive risk function is: ； in, represents the adjustment parameter; The formula for the blind spot risk is: ； in, , represents the double integral, , They represent the length from the risk point to the origin of the polar coordinates and the angle between the risk point and the X-axis. The comprehensive risk function The polar coordinate expression of .

7. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 5, characterized in that: The step of constructing a protection zone model and calculating a corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data includes: S5-1, setting vehicle protection zone parameters during left-turn human-vehicle interaction, and creating a vehicle protection zone model based on a rectangle and a semicircle; S5-2. Based on the vehicle protection zone model, limit the vehicle potential field generation range corresponding to the vehicle, and according to the formula: ； Obtain the vehicle potential field risk within the vehicle potential field generation range ;in, Indicates that the vehicle is at time The actual driving speed is , Indicates that the vehicle is at time The upper speed limit, represents the logarithmic function with the natural constant e as the base, represents the vehicle field strength coefficient, Indicates that the vehicle is at time The angle between the velocity vector and the vehicle's center of mass, represents the pseudo distance, represents the absolute value, represents the rate of change of the control potential field strength with distance, Indicates visibility; S5-3, setting the parameters of the pedestrian protection zone during the left-turn human-vehicle interaction process, and creating a pedestrian protection zone model; S5-4. Based on the pedestrian protection area model, limit the pedestrian potential field generation range corresponding to the pedestrian, and according to the formula: ； Obtain the pedestrian potential field risk within the pedestrian potential field generation range ; S5-5, according to the formula: ； Get the potential energy field risk ; The protection zone model includes the vehicle protection zone model and the pedestrian protection zone model.

8. A left-turn human-vehicle interaction risk assessment method based on drone video according to claim 7, characterized in that: The method of collecting pedestrian-vehicle interaction video data through a drone, generating a corresponding possible future operation space field, and calculating the corresponding possible future operation space field risk amount includes: S6-1, collecting pedestrian-vehicle interaction video data through a drone, and dividing it into a vehicle image sequence and a pedestrian image sequence of multiple frames respectively; S6-2, using Kalman filtering to continuously track the vehicles and pedestrians in the vehicle image sequence and the pedestrian image sequence, respectively, and assign corresponding IDs; S6-3, based on each of the IDs outputted in S6-2, the vehicle image sequence and the pedestrian image sequence are processed respectively by a target detection and tracking algorithm to obtain corresponding vehicle information and pedestrian information; The vehicle information includes the vehicle speed, the vehicle direction and the vehicle lane; the pedestrian information includes the pedestrian position information, the pedestrian crossing speed and the pedestrian movement trajectory; S6-4, based on the vehicle information, generating a corresponding possible future operation space of the vehicle; S6-5, based on the pedestrian information, generating a corresponding possible future movement space of the pedestrian; S6-6, recording the overlapping area of ​​the possible future running space of the vehicle and the possible future running space of the pedestrian as the possible future running area , calculate the possible future operating area The plane projection area , obtain the possible future operation space field risk amount; The possible future operating space field includes the possible future operating space of vehicles and the possible future operating space of pedestrians.

9. A method and system for assessing left-turn human-vehicle interaction risk based on drone video according to claim 8, characterized in that: The left-turn human-vehicle interaction risk assessment result is determined based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount, and the formula is: ； in, , , represents the weight coefficient, , , They represent the blind area risk, potential energy risk and future possible operation space risk respectively. Indicates the risk assessment result of left-turn human-vehicle interaction.

10. A left-turn human-vehicle interaction risk assessment system based on drone video, used to implement a left-turn human-vehicle interaction risk assessment method based on drone video according to any one of claims 1 to 9, characterized in that: It includes pedestrian-vehicle data acquisition module, vehicle status module, pedestrian status module, blind spot field risk quantity calculation module, potential energy field risk quantity calculation module, future possible operation space field risk quantity calculation module and left turn human-vehicle interaction risk assessment module; among which: A pedestrian-vehicle data acquisition module, used to obtain the initial parameters of the vehicle and the initial parameters of the pedestrian, and to shoot the pedestrian-vehicle interaction video data through a drone; A vehicle state module, used for processing the initial parameters of the vehicle to obtain the vehicle driving data; A pedestrian state module, used for processing the initial parameters of the pedestrian to obtain the pedestrian action data; A blind spot risk calculation module, used to construct the human-vehicle blind spot coordinate system and calculate the corresponding blind spot risk based on the pedestrian action data and the vehicle driving data; A potential energy field risk amount calculation module, used to construct the protection zone model and calculate the corresponding potential energy field risk amount based on the pedestrian action data and the vehicle driving data; A future possible operation space field risk amount calculation module is used to process pedestrian-vehicle interaction video data through Kalman filtering and target detection and tracking algorithms to generate the corresponding future possible operation space field and calculate the corresponding future possible operation space field risk amount; The left-turn human-vehicle interaction risk assessment module is used to calculate the left-turn human-vehicle interaction risk assessment result based on the blind spot field risk amount, the potential energy field risk amount and the future possible operation space field risk amount.

Images