A method for predicting the probability of abnormal pedestrian movement

Abstract

The present invention discloses a method for predicting the probability of abnormal pedestrian crossing, which belongs to the technical field of traffic behavior analysis and can solve the problem that the existing technology cannot analyze and predict the probability of abnormal pedestrian crossing, resulting in a greater traffic safety risk. The method includes: S1, establishing a relationship model between the probability of abnormal pedestrian crossing and its multiple influencing factors; S2, collecting the actual behavior and multiple influencing factors of each pedestrian crossing abnormally among multiple pedestrians in the target section, and determining the coefficient of each influencing factor in the relationship model according to the actual behavior, influencing factors and relationship model of each pedestrian; S3, determining the correlation between the corresponding influencing factor and the behavior probability according to each coefficient, and predicting the probability of abnormal pedestrian crossing according to the correlation. The present invention is used to predict the probability of abnormal pedestrian crossing.

Description

Technical Field

[0001] The invention relates to a method for predicting the probability of abnormal pedestrian crossing, and belongs to the technical field of traffic behavior analysis. Background Art

[0002] There are many sections without traffic lights or zebra crossings in urban living areas. The demand for pedestrians crossing the street is often high, resulting in a large number of abnormal pedestrian crossing behaviors. Compared with the normal crossing behavior of pedestrians following traffic lights and crossing along zebra crossings, the abnormal crossing behavior of pedestrians is difficult to accurately analyze and predict due to the influence of traffic flow, which poses a greater traffic safety risk.

[0003] Therefore, it is necessary to study and analyze the abnormal crossing behavior of pedestrians, which is of great significance for scientifically and rationally transforming urban sections without signal control and zebra crossings, implementing pedestrian protection measures, improving road traffic conditions, and reducing traffic safety risks. Summary of the invention

[0004] The present invention provides a method for predicting the probability of abnormal pedestrian crossing behavior, which can solve the problem that the prior art is unable to analyze and predict the probability of abnormal pedestrian crossing behavior, resulting in a greater traffic safety risk.

[0005] The present invention provides a method for predicting the probability of abnormal pedestrian crossing, the method comprising:

[0006] S1. Establish a relationship model between the probability of pedestrians' abnormal crossing and its multiple influencing factors;

[0007] S2, obtaining the actual path and multiple influencing factors of each pedestrian's abnormal walking among multiple pedestrians in the target road section, and determining the coefficient of each influencing factor in the relationship model according to the actual path, influencing factor and the relationship model of each pedestrian;

[0008] S3. Determine the correlation between the corresponding influencing factor and the behavior probability according to each coefficient, and predict the behavior probability of abnormal pedestrian crossing according to the correlation.

[0009] Optionally, the probability of abnormal pedestrian movement includes the probability of pedestrians walking from a preset starting position, the probability of pedestrians walking along a preset walking angle, and the probability of pedestrians walking along a preset walking trajectory; the walking angle is the angle between the direction of travel of the pedestrian when entering the road and the direction of travel of the road.

[0010] Optionally, the relationship model includes a travel position sub-model, a travel angle sub-model and a travel trajectory sub-model.

[0011] Optionally, the travel position sub-model and the travel angle sub-model are both binary logistic regression models; and the travel trajectory sub-model is a multivariate logistic regression model.

[0012] Optionally, the method for determining the coefficient of each influencing factor in S2 is a maximum likelihood estimation method.

[0013] Optionally, the correlation between the corresponding influencing factor and the behavior probability is determined according to each coefficient in S3, specifically:

[0014] When the coefficient is greater than zero, it is determined that the corresponding impact factor is positively correlated with the behavior probability;

[0015] When the coefficient is less than zero, it is determined that the corresponding impact factor is negatively correlated with the behavior probability;

[0016] When the coefficient is equal to zero, it is determined that the corresponding impact factor is independent of the behavior probability.

[0017] Optionally, the multiple influencing factors include pedestrian attribute factors, walking process factors and built environment factors.

[0018] Optionally, the preset starting position is a road opening, the preset travel angle is a travel angle greater than 60°, and the preset travel trajectory includes at least one of a single-curved trajectory, a hyperbolic trajectory and a complex trajectory.

[0019] The beneficial effects that the present invention can produce include:

[0020] The present invention establishes a relationship model between the probability of abnormal pedestrian crossing and its influencing factors, calculates the coefficient of each influencing factor, determines the correlation between each influencing factor and the probability of abnormal pedestrian crossing, and achieves a qualitative explanation of the influencing factors of abnormal pedestrian crossing, so that the probability of abnormal pedestrian crossing can be analyzed and predicted, which provides a scientific basis for planning crossing facilities and implementing pedestrian protection measures in urban sections without signal control and zebra crossings, and is conducive to improving road traffic conditions and reducing traffic safety risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A flow chart of a method for predicting the probability of abnormal pedestrian movement provided by an embodiment of the present invention;

[0022] Figure 2 A schematic diagram of a pedestrian's walking trajectory provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0023] The present invention is described in detail below in conjunction with embodiments, but the present invention is not limited to these embodiments.

[0024] The embodiment of the present invention provides a method for predicting the probability of abnormal pedestrian movement. Figure 1 As shown, the method includes:

[0025] S1. Establish a relationship model between the probability of pedestrians' abnormal crossing and its multiple influencing factors;

[0026] S2, obtaining the actual path and multiple influencing factors of each pedestrian's abnormal walking among multiple pedestrians in the target road section, and determining the coefficient of each influencing factor in the relationship model according to the actual path, influencing factor and relationship model of each pedestrian;

[0027] S3. Determine the correlation between the corresponding influencing factor and the behavior probability according to each coefficient, and predict the behavior probability of abnormal pedestrian crossing according to the correlation.

[0028] Among them, abnormal pedestrian crossing is the opposite of normal pedestrian crossing, and includes the complex crossing behavior of pedestrians under the influence of traffic flow on roads without signal control and zebra crossing guidance.

[0029] Specifically, the target road section is a section in the city without signal control and zebra crossing, with no physical separation facilities on both sides of the road section and no lane separation facilities in the middle of the road, and pedestrians can enter the road from any position on both sides of the road and walk through it. This embodiment obtains specific data of the actual path of abnormal pedestrian walking and multiple influencing factors from the target road section. The length of the target road section can be specifically set according to the data acquisition requirements. For example, the length of the target road section can be 100 meters to 150 meters.

[0030] Specifically, the probability of abnormal pedestrian movement may include the probability of the pedestrian moving from a preset starting position, the probability of the pedestrian moving along a preset moving angle, and the probability of the pedestrian moving along a preset moving trajectory.

[0031] Correspondingly, the actual path of the pedestrian may include the actual starting position, actual walking angle and actual walking trajectory of the pedestrian.

[0032] The starting position is the position where the pedestrian enters the road from one side and starts to cross the road; the crossing angle is the angle between the pedestrian's direction of travel at the starting position and the road's direction of travel; the crossing trajectory is the complete trajectory of the pedestrian entering the road from one side and leaving the road from the other side. Figure 2 As shown, the travel trajectory may include a single-curved trajectory, a hyperbolic trajectory, and a complex trajectory.

[0033] Specifically, the preset starting position Enter of this embodiment is the road opening, the preset traveling angle Angle is a traveling angle greater than 60°, and the preset traveling trajectory Shape includes at least one of a single-curved trajectory, a hyperbolic trajectory, and a complex trajectory.

[0034] Specifically, this embodiment first establishes a plane rectangular coordinate system on the target road section with the road travel direction as the horizontal axis, and determines the coordinate data of the preset starting position Enter, the preset travel angle Angle, and the preset travel trajectory Shape; then, video data of each pedestrian walking through the target road section is collected by deploying drones above the road, cameras on both sides of the road, etc., and the coordinate data of the actual starting position, actual travel angle, and actual travel trajectory of each pedestrian when walking are extracted from the video data; finally, the coordinate data of each pedestrian is compared and analyzed with the coordinate data of the preset starting position Enter, the preset travel angle Angle, and the preset travel trajectory Shape, so as to determine whether each pedestrian walks from the preset starting position Enter, whether he walks along the preset travel angle Angle, and whether he walks along the preset travel trajectory Shape, and then the probability of each pedestrian walking from the preset starting position, the probability of walking along the preset travel angle, and the probability of walking along the preset travel trajectory can be determined.

[0035] Specifically, in this embodiment, the value of the above probability is defined as follows:

[0036] 1. The probability of a pedestrian passing through from a preset starting position

[0037] When the pedestrian enters from the preset starting position, the probability is 1; otherwise, the probability is 0.

[0038] 2. The probability of pedestrians walking along the preset walking angle

[0039] When the pedestrian walks along the preset walking angle Angle, the probability value is 1; otherwise, the probability value is 0.

[0040] 3. The probability of pedestrians walking along the preset walking trajectory

[0041] The preset trajectory Shape includes single-curved trajectory, hyperbolic trajectory and complex trajectory;

[0042] When the pedestrian walks along a single curved trajectory, the probability is 0;

[0043] When the pedestrian walks along a hyperbolic trajectory, the probability is 1;

[0044] When the pedestrian walks along a complex trajectory, the probability is 2.

[0045] Specifically, the multiple influencing factors may include pedestrian attribute factors, walking process factors and built environment factors.

[0046] Specifically, pedestrian attribute factors may include:

[0047] Pedestrian gender information G: If the pedestrian is male, G=1, otherwise G=0;

[0048] Pedestrian carrying luggage L: If the pedestrian carries luggage that affects walking, L=1, otherwise L=0;

[0049] Pedestrians use mobile phones while walking: M = 1 if pedestrians use mobile phones while walking, otherwise M = 0;

[0050] Specifically, the travel process factors may include:

[0051] Pedestrian waiting time before crossing W;

[0052] The average speed of pedestrians S;

[0053] The conflict situation V between pedestrians and vehicles when crossing: if there is no conflict with vehicles, V=0; if there is a conflict with a small vehicle, V=1; if there is a conflict with a large vehicle, V=2; the conflict situations between pedestrians and vehicles include the pedestrian crossing process being suspended due to the movement of vehicles, the walking trajectory being changed due to the movement of vehicles, etc.

[0054] The driving state VS of the vehicle when there is a conflict with the pedestrian: VS=0 if there is no conflict, VS=1 if the vehicle is moving at a constant speed when there is a conflict, VS=2 if the vehicle is accelerating when there is a conflict, and VS=3 if the vehicle is decelerating when there is a conflict;

[0055] The situation P of pedestrians being blocked by roadside parking when entering the road: if blocked, P=1, if not blocked, P=0; it is worth noting that the blocking situation should be observed from the perspective of the driver of the vehicle traveling on the road to see whether the pedestrians in front of the driving direction are blocked by roadside parking when entering the road.

[0056] The specific data of the above-mentioned pedestrian attribute factors and walking process factors can be obtained by analyzing the aforementioned collected video data of pedestrians walking through the target road section.

[0057] Specifically, built environment factors may include:

[0058] The setting of the zebra crossing in front of the pedestrian's starting position is C: if there is a zebra crossing, C=1, otherwise C=0;

[0059] The number of road openings CN between the pedestrian's starting position and the ending position, where the pedestrian completes the crossing and leaves the road.

[0060] The horizontal distance D between the pedestrian's starting position and the nearest road opening.

[0061] The specific data of the above-mentioned built environment factors can be obtained by analyzing the environmental information of the target road section and the aforementioned collected video data of pedestrians passing through the target road section.

[0062] By arranging the above data, we can obtain the data collection statistics table of the target section, as shown in Table 1.

[0063] Table 1 Data collection statistics of target road sections

[0064]

[0065] Specifically, the relationship model may include a walking position sub-model, a walking angle sub-model and a walking trajectory sub-model, which respectively correspond to the probability of a pedestrian walking from a preset starting position, the probability of a pedestrian walking along a preset walking angle and the probability of a pedestrian walking along a preset walking trajectory.

[0066] In this embodiment, since the probability of a pedestrian passing through a preset starting position and the probability of a pedestrian passing through a preset passing angle both have only two values, 0 and 1, the passing position sub-model and the passing angle sub-model can both adopt a binary logistic regression model; since the probability of a pedestrian passing through a preset passing trajectory has three values, 0, 1, and 2, the passing trajectory sub-model can adopt a multivariate logistic regression model.

[0067] Specifically, the formula of the travel position sub-model is:

[0068] (1)

[0069] (2)

[0070] In formula (1) and formula (2), is the probability that a pedestrian walks through from the preset starting position Enter at the road opening, and its value corresponds to the data in the Enter column in Table 1.

[0071] They are respectively G , L , M , W , S , V , VS , P , C , CN , D The coefficients of the influencing factors, which represent the influence of each influencing factor on the pedestrian from the preset starting position Enter The correlation of the probability of traveling.

[0072] Specifically, the formula of the travel angle sub-model is:

[0073] (3)

[0074] (4)

[0075] In formula (3) and formula (4), It is the probability that the pedestrian walks along the preset walking angle Angle greater than 60°, and its value corresponds to the data in the Angle column in Table 1.

[0076] are the coefficients of the influencing factors G, L, M, W, S, V, VS, P, C, CN, and D in formula (4), respectively, representing the correlation between each influencing factor and the probability of pedestrians walking along a preset walking angle Angle greater than 60°.

[0077] Specifically, the formula of the travel trajectory sub-model is:

[0078] (5)

[0079] (6)

[0080] (7)

[0081] (8)

[0082] In formula (5) to formula (8), and They are the probabilities of pedestrians walking along the hyperbolic and complex preset walking trajectories Shape, respectively, and their values ​​correspond to the data in the Shape column in Table 1.

[0083] are the coefficients of the influencing factors G, L, M, W, S, V, VS, P, C, CN, and D in formula (6), respectively, representing the correlation between each influencing factor and the probability of pedestrians walking along the hyperbolic preset walking trajectory Shape.

[0084] are the coefficients of the influencing factors G, L, M, W, S, V, VS, P, C, CN, and D in formula (8), respectively, representing the correlation between each influencing factor and the probability of pedestrians walking along the complex preset walking trajectory Shape.

[0085] The data of influencing factors such as G, L, M, W, S, V, VS, P, C, CN, and D corresponding to multiple pedestrians in Table 1 are used as independent variables, and the data of Enter, Angle, and Shape corresponding to multiple pedestrians in Table 1 are used as dependent variables, and are input into formulas (1) to (8) respectively. The corresponding coefficients of each influencing factor in the walking position sub-model, the walking angle sub-model, and the walking trajectory sub-model can be calculated, as shown in Table 2. According to the coefficient of each influencing factor in Table 2, the correlation between the corresponding influencing factor and the path probability can be determined.

[0086] Table 2 Relationship model impact factor coefficient table

[0087]

[0088] Specifically, the method for determining the coefficient of each influencing factor in S2 may be the maximum likelihood estimation method. This method is based on the data in Table 1, as well as the travel position sub-model, the travel angle sub-model, and the travel trajectory sub-model, and obtains the corresponding coefficient of each influencing factor in each sub-model by maximizing the likelihood value of each group of data.

[0089] Specifically, in S3, the correlation between the corresponding influencing factor and the behavior probability is determined according to each coefficient, which can be specifically:

[0090] When the coefficient is greater than zero, it is determined that the corresponding impact factor is positively correlated with the behavior probability;

[0091] When the coefficient is less than zero, it is determined that the corresponding impact factor is negatively correlated with the behavior probability;

[0092] When the coefficient is equal to zero, it is determined that the corresponding impact factor is independent of the behavior probability.

[0093] Specifically, when When a certain coefficient is greater than zero, it means that the larger the value of the corresponding influencing factor is, the more inclined the pedestrian is to enter from the preset starting position; when a certain coefficient is less than zero, it means that the larger the value of the corresponding influencing factor is, the more inclined the pedestrian is not to enter from the preset starting position; when a certain coefficient is equal to zero, it means that the corresponding influencing factor has nothing to do with whether the pedestrian enters from the preset starting position.

[0094] Specifically, when When a certain coefficient is greater than zero, it means that the larger the value of the corresponding influencing factor is, the more inclined the pedestrian is to walk along the preset walking angle Angle; when a certain coefficient is less than zero, it means that the larger the value of the corresponding influencing factor is, the more inclined the pedestrian is not to walk along the preset walking angle Angle; when a certain coefficient is equal to zero, it means that the corresponding influencing factor has nothing to do with whether the pedestrian walks along the preset walking angle Angle.

[0095] Specifically, when When a certain coefficient in is greater than zero, it means that the larger the value of the corresponding influencing factor is, the more the pedestrian tends to walk along the hyperbolic preset walking trajectory Shape compared to the monobolic trajectory; when a certain coefficient is less than zero, it means that the larger the value of the corresponding influencing factor is, the more the pedestrian tends not to walk along the hyperbolic preset walking trajectory Shape compared to the monobolic trajectory; when a certain coefficient is equal to zero, it means that the corresponding influencing factor has nothing to do with whether the pedestrian walks along the hyperbolic preset walking trajectory Shape.

[0096] Specifically, when When a certain coefficient in is greater than zero, it means that the larger the value of the corresponding influencing factor is, the more the pedestrian tends to walk along the complex preset walking trajectory Shape compared to the single curved trajectory; when a certain coefficient is less than zero, it means that the larger the value of the corresponding influencing factor is, the more the pedestrian tends not to walk along the complex preset walking trajectory Shape compared to the single curved trajectory; when a certain coefficient is equal to zero, it means that the corresponding influencing factor has nothing to do with whether the pedestrian walks along the complex preset walking trajectory Shape.

[0097] For example, if If it is a positive value, it means that male pedestrians are more inclined to choose to enter from the preset starting position at the road opening. This can achieve a qualitative explanation of the factors affecting abnormal pedestrian crossing, so that the probability of abnormal pedestrian crossing can be analyzed and predicted, providing a scientific basis for planning crossing facilities and implementing pedestrian protection measures in urban sections without signal control and zebra crossings, which is conducive to improving road traffic conditions and reducing traffic safety risks.

[0098] The above are only a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application is disclosed as above with preferred embodiments, it is not intended to limit the present application. Any technician familiar with the profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A method for predicting the probability of abnormal pedestrian movement, characterized in that: The method comprises: S1. Establish a relationship model between the probability of pedestrians' abnormal crossing and its multiple influencing factors; S2, obtaining the actual path and multiple influencing factors of each pedestrian's abnormal walking among multiple pedestrians in the target road section, and determining the coefficient of each influencing factor in the relationship model according to the actual path, influencing factor and the relationship model of each pedestrian; S3. Determine the correlation between the corresponding influencing factor and the behavior probability according to each coefficient, and predict the behavior probability of the pedestrian's abnormal crossing according to the correlation; The abnormal pedestrian crossing behavior probability includes the probability of the pedestrian crossing from a preset starting position, the probability of the pedestrian crossing along a preset crossing angle, and the probability of the pedestrian crossing along a preset crossing trajectory; the relationship model includes a crossing position sub-model, a crossing angle sub-model, and a crossing trajectory sub-model, which respectively correspond to the probability of the pedestrian crossing from a preset starting position, the probability of the pedestrian crossing along a preset crossing angle, and the probability of the pedestrian crossing along a preset crossing trajectory; the multiple influencing factors include the pedestrian's gender information, the pedestrian's luggage, the pedestrian's use of mobile phones when crossing, the waiting time before the pedestrian crosses, the average speed of the pedestrian crossing, the conflict between the pedestrian and the vehicle when the pedestrian crosses, the driving state of the vehicle when the conflict occurs with the pedestrian, the pedestrian being blocked by the roadside parking when entering the road, the setting of the zebra crossing in front of the pedestrian's starting position, the number of road openings between the pedestrian's starting position and the ending position, and the horizontal distance between the pedestrian's starting position and the nearest road opening; In S3, the correlation between the corresponding influencing factor and the behavior probability is determined according to each coefficient, specifically: When the coefficient is greater than zero, it is determined that the corresponding impact factor is positively correlated with the behavior probability; When the coefficient is less than zero, it is determined that the corresponding impact factor is negatively correlated with the behavior probability; When the coefficient is equal to zero, it is determined that the corresponding impact factor is independent of the behavior probability.

2. The method according to claim 1, characterized in that: The travel angle is the angle between the direction of travel of the pedestrian when entering the road and the direction of travel of the road.

3. The method according to claim 2, characterized in that The travel position sub-model and the travel angle sub-model are both binary logistic regression models; the travel trajectory sub-model is a multivariate logistic regression model.

4. The method according to claim 1, characterized in that: The method for determining the coefficient of each influencing factor in S2 is the maximum likelihood estimation method.

5. The method according to claim 2, characterized in that: The preset starting position is a road opening, the preset traveling angle is a traveling angle greater than 60°, and the preset traveling trajectory includes at least one of a single-curved trajectory, a hyperbolic trajectory, and a complex trajectory.

Images