Method for Identifying Multiple Abnormal State Points on Special Road Sections Based on Spatial Autocorrelation

Abstract

The present invention relates to the field of intelligent transportation technology, and particularly to a method for identifying multiple occurrence points of abnormal states on special road sections based on spatial autocorrelation. By constructing a feature category set of abnormal operating states and performing spatial autocorrelation analysis on it, the spatial correlation characteristics of the obtained abnormal state spatial correlation data become more obvious, reducing the influence of abnormal operating state data that only affects the local traffic flow state and has weak spatial correlation. Furthermore, the accuracy of identifying multiple occurrence points is higher.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation, and particularly to a method for identifying multiple abnormal state points in special sections based on spatial autocorrelation. Background Art

[0002] With the rapid development of the construction of smart transportation and transportation power in China, technologies such as artificial intelligence and big data are widely used in the transportation field. Traffic perception facilities are becoming increasingly rich, and the macroscopic perception of traffic flow speed, density, flow, various traffic events, etc. and the microscopic perception accuracy of vehicle coordinates, trajectories, etc. are also getting higher and higher. How to apply mathematical statistics, data mining and other methods to make full and efficient use of traffic data is the main way to effectively improve the outdoor layout position of intelligent roadside facilities on smart highways and improve the traffic operation risks in special sections in the future. And accurately identifying the accident-prone points in special sections of expressways is the basis for controlling the overall risk development trend and effectively implementing risk early warning. Therefore, there is an urgent need for a method for identifying multiple abnormal state points in special sections. Summary of the Invention

[0003] Aiming at the deficiencies of the existing technology, the present invention proposes a method for identifying multiple abnormal state points in special sections based on spatial autocorrelation, so as to improve the accuracy of identifying accident-prone points in special sections.

[0004] In a first aspect, the present invention provides a method for identifying multiple abnormal state points in special sections based on spatial autocorrelation.

[0005] In a first implementable manner, a method for identifying multiple abnormal state points in special sections based on spatial autocorrelation includes:

[0006] Constructing an abnormal operation state data set for special sections;

[0007] Obtaining an abnormal state feature category set according to the abnormal operation state data set;

[0008] Performing spatial autocorrelation analysis according to the abnormal state feature category set to obtain abnormal state spatial correlation data;

[0009] Identifying multiple abnormal operation state points in special sections according to the abnormal state spatial correlation data.

[0010] Combined with the first implementable manner, in a second implementable manner, constructing an abnormal operation state data set for special sections includes:

[0011] Using roadside cameras and roadside radars to provide traffic operation state data within special sections;

[0012] Constructing an abnormal operation state data set according to the traffic operation state data.

[0013] Combined with the second implementation manner, in the third implementation manner, an abnormal state feature category set is obtained according to the abnormal operation state data set, including:

[0014] Determine the recognition features of multiple abnormal behavior types;

[0015] Identify the abnormal operation state data set according to each recognition feature to obtain an abnormal state feature category set, and the abnormal state feature category set includes multiple abnormal behavior types.

[0016] Combined with the second implementation manner, in the fourth implementation manner, spatial autocorrelation analysis is performed according to the abnormal state feature category set to obtain abnormal state spatial correlation data, including:

[0017] Obtain the spatial weights of each abnormal behavior type according to the abnormal state feature category set and the length of the special section;

[0018] Construct a spatial adjacency matrix according to the abnormal operation state data set, and the spatial adjacency matrix includes multiple spatial objects;

[0019] Assign values to each spatial object according to each spatial weight according to the principle of distance decay to obtain each abnormal state spatial correlation data.

[0020] Combined with the fourth implementation manner, in the fifth implementation manner, the spatial weight is obtained through the following formula:

[0021]

[0022] In the above formula, w ij is the unit value of the i-th row and j-th column in the spatial weight matrix, φ1 is the coordinate of the first abnormal operation state, φ2 is the coordinate of the second abnormal operation state, Δτ is the longitudinal coordinate difference between the first abnormal operation state and the second abnormal operation state, and L is the length of the special section.

[0023] Combined with the fourth implementation manner, in the sixth implementation manner, the abnormal state spatial correlation data is obtained through the following formula:

[0024]

[0025] Among them, I is the abnormal state spatial correlation data, z i is the deviation between the abnormal operation state data of the i-th spatial object and the average value, σ ij is the equivalent property damage only revision coefficient, w ij is the unit value of the i-th row and j-th column in the spatial weight matrix, m is the total number of abnormal operation states, and S0 is the aggregation of all spatial weights.

[0026] Combined with the first implementation, in the seventh implementation, identifying multiple occurrence points of abnormal operating states on special sections based on abnormal state space correlation data includes:

[0027] Performing clustering analysis on each abnormal state space correlation data to obtain an aggregation index;

[0028] Identifying abnormal operating state points based on the aggregation index.

[0029] Combined with the first implementation, in the eighth implementation, identifying multiple occurrence points of abnormal operating states on special sections based on abnormal state space correlation data includes:

[0030] Performing density estimation on each abnormal state space correlation data to obtain the density distribution of the abnormal state;

[0031] Identifying abnormal operating state points based on the density distribution.

[0032] In a second aspect, the present invention provides an apparatus for identifying multiple occurrence points of abnormal states on special sections based on spatial autocorrelation.

[0033] In the ninth implementation, an apparatus for identifying multiple occurrence points of abnormal states on special sections based on spatial autocorrelation includes:

[0034] A construction module configured to construct a dataset of abnormal operating states of special sections;

[0035] An acquisition module configured to obtain a set of abnormal state feature categories based on the dataset of abnormal operating states;

[0036] A spatial autocorrelation analysis module configured to perform spatial autocorrelation analysis based on the set of abnormal state feature categories to obtain abnormal state space correlation data;

[0037] A multiple occurrence point identification module configured to identify multiple occurrence points of abnormal operating states on special sections based on the abnormal state space correlation data.

[0038] As can be seen from the above technical solutions, the beneficial technical effects of the present invention are as follows:

[0039] 1. Since not all abnormal operating states within special sections will evolve into traffic accidents, and some behaviors such as slow traffic and speeding only affect the local traffic flow state, therefore, this solution constructs a set of feature categories of abnormal operating states and performs spatial autocorrelation analysis on it, so that the spatial correlation characteristics of the obtained abnormal state space correlation data are more obvious, reducing the influence of abnormal operating state data that only affects the local traffic flow state and has weak spatial correlation, and thus making the accuracy of identifying multiple occurrence points higher.

[0040] 2. Automatically identify potential hazard points (sections) in highway merging and diverging areas and tunnel entrances and exits through abnormally state-space associated data, providing technical support for the precise setting of roadside perception warning facilities in special highway sections, which is of great significance for improving the traffic safety level of special highway sections and reducing the cost of intelligent highway facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0042] Figure 1 Schematic diagram of a method for identifying multiple potential hazard points in special sections based on spatial autocorrelation provided by the present invention;

[0043] Figure 2 Structural diagram of the classification of feature category sets provided by the present invention;

[0044] Figure 3 Schematic diagram of a method for identifying potential hazard points in special sections provided by the present invention;

[0045] Figure 4 Schematic diagram of the principle of density distribution provided by the present invention;

[0046] Figure 5 Structural diagram of a device for identifying multiple potential hazard points in special sections based on spatial autocorrelation provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following will describe in detail the embodiments of the technical solutions of the present invention with reference to the drawings. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention, and thus are only examples and cannot be used to limit the protection scope of the present invention.

[0048] It should be noted that unless otherwise specified, the technical terms or scientific terms used in this application should have the ordinary meanings understood by those skilled in the art to which the present invention belongs.

[0049] Combined with Figure 1 As shown, this embodiment provides a method for identifying multiple potential hazard points in special sections based on spatial autocorrelation, including:

[0050] Step S01, construct a dataset of abnormal operating states of special sections;

[0051] Step S02, obtain a set of abnormal state feature categories based on the dataset of abnormal operating states;

[0052] Step S03: Perform spatial autocorrelation analysis based on the abnormal state feature category set to obtain abnormal state spatial correlation data;

[0053] Step S04: Identify the frequent occurrence points of abnormal operation status in special road sections based on the abnormal status spatial correlation data.

[0054] Optionally, constructing an abnormal operation status dataset of a special road section includes: using roadside cameras and roadside radars to provide traffic operation status data in the special road section; and constructing an abnormal operation status dataset based on the traffic operation status data.

[0055] In some embodiments, roadside intelligent cameras are used to capture lane-by-lane video information for special sections of highways, such as merge zones and tunnel entrances and exits. Based on this video information, all-weather lane-level traffic status data is extracted. This lane-level traffic status data includes information such as lane occupancy, event location, and event type. Roadside radar is used to capture vehicle operating parameters within each lane of a special section, including real-time location, trajectory, instantaneous speed, average speed, and acceleration. The radar directly outputs information such as event location, event type, and the vehicle's real-time location, trajectory, instantaneous speed, average speed, and acceleration.

[0056] In some embodiments, traffic operation status data obtains parameters such as collision time, rear intrusion time and braking distance ratio, and uses parameters such as collision time, rear intrusion time and braking distance ratio as vehicle collision risk parameters to identify inflection points and mutation points of traffic operation status within a certain time and space period, and establish an abnormal operation status data set such as wrong-way driving, slow driving, speeding, and continuous lane changing.

[0057] In some embodiments, the predicted collision time is calculated by dividing the distance between two vehicles by their relative speed. The post-intrusion time (PTT) is the time difference between two vehicles reaching a specific intersection. It refers to the conflict time between straight-moving and right-turn merging. The PTT is a time-based traffic conflict identification indicator and also serves as an indicator of conflict severity.

[0058] Optionally, obtaining an abnormal state feature category set based on the abnormal operation state data set includes: determining identification features of multiple abnormal behavior types; identifying the abnormal operation state data set based on each identification feature to obtain an abnormal state feature category set, wherein the abnormal state feature category set includes multiple abnormal behavior types.

[0059] In some embodiments, Figure 2 Divide the structure diagram for the feature category set, combined with Figure 2As shown in the figure, after analyzing the abnormal operation state data sets of special sections such as the merging and diverging areas and the tunnel entrances and exits, the abnormal behavior types can be mainly divided into reverse driving, slow driving, speeding, dangerous lane-changing, parking, and pedestrian intrusion. Taking the abnormal behavior types as the first-level categories, the main recognition features of various abnormal behaviors are screened from the perspectives of traffic flow state and vehicle dynamics, including the abnormal growth frequency of traffic density, driving direction, vehicle speed change rate, time headway, etc., to clarify the theoretical requirements for identifying abnormal operation states, and taking the recognition features corresponding to each type as the second-level categories. For example Figure 2 As shown in the figure, the second-level categories corresponding to reverse driving are driving direction, speed, and acceleration; the second-level categories corresponding to slow driving and speeding are coordinate transformation rate and vehicle speed transformation rate; the second-level categories corresponding to dangerous lane-changing are lateral speed and lateral coordinate transformation rate; the second-level categories corresponding to parking are no coordinate transformation and speed being zero; the second-level categories corresponding to pedestrian intrusion are detecting pedestrians.

[0060] In some embodiments, the abnormal operation states include reverse driving, slow driving, speeding, dangerous lane-changing, parking, and pedestrian intrusion. The recognition feature of reverse driving is: the driving direction is opposite, and the speed, acceleration, and coordinate transformation are opposite; the recognition feature of slow driving is: the coordinate transformation rate is small, and the speed transformation rate is small; the recognition feature of speeding is: the coordinate transformation rate is large, and the speed transformation rate is large; the recognition feature of dangerous lane-changing is: the lateral speed transformation rate increases, and the vertical and lane coordinate transformation rates increase; the recognition feature of parking is: no coordinate transformation, and the detected speed is zero; the recognition feature of pedestrian intrusion is: pedestrians are detected within the monitoring range.

[0061] Optionally, after obtaining the abnormal state feature category set according to the abnormal operation state data set, it further includes: performing recognition processing on the abnormal operation state data set according to the recognition features of each abnormal operation state type to obtain the occurrence times of each abnormal operation state, and converting the occurrence times of the abnormal state into equivalent event times; performing identification of abnormal operation state points according to the equivalent event times.

[0062] In some embodiments, in combination with Figure 3 As shown in the figure, after performing spatial clustering on the abnormal operation state data, three spatial point distribution situations of random, uniform, and aggregated are obtained. Then, the Gaussian kernel function is selected to estimate the spatial distribution density of abnormal state points in the plane space, convert the occurrence times of the abnormal operation state into equivalent event times, and refer to the idea of the Equivalent Property Damage Only (EPDO) method to evaluate the weights of each abnormal state point, and establish a model for identifying the frequent occurrence locations of abnormal operation states in special sections of expressways based on spatial autocorrelation. According to the spatial distribution density and spatial point weights, automatically identify the potential hazard points (sections) of the merging and diverging areas and tunnel entrances and exits of expressways.

[0063] Optionally, spatial autocorrelation analysis is performed based on the abnormal state feature category set to obtain abnormal state spatial association data, including: obtaining the spatial weight of each abnormal behavior type based on the abnormal state feature category set and the length of the special road section; constructing a spatial adjacency matrix based on the abnormal operation state data set, the spatial adjacency matrix including multiple spatial objects; assigning values to each spatial object according to each spatial weight according to the principle of distance attenuation to obtain each abnormal state spatial association data.

[0064] Optionally, the spatial weight is obtained by the following formula:

[0065]

[0066] In the above formula, w ij is the cell value in the i-th row and j-th column of the spatial weight matrix, φ1 is the coordinate of the first abnormal operating state, φ2 is the coordinate of the second abnormal operating state, Δτ is the difference in longitudinal coordinates between the first and second abnormal operating states, and L is the length of the special road section.

[0067] Optionally, the abnormal state space correlation data is obtained by the following formula:

[0068]

[0069] Among them, I is the abnormal state space correlation data, z i is the deviation between the abnormal operating state data of the i-th spatial object and the average value, σ ij is the equivalent property loss correction factor, w ij is the cell value in the i-th row and j-th column of the spatial weight matrix, m is the total number of abnormal operating states, and S0 is the aggregation of all spatial weights.

[0070] In some embodiments, the revision coefficient for parking is 1.0, the revision coefficient for driving against traffic is 0.9, the revision coefficient for pedestrian intrusion is 0.8, the coefficients for slowing down and speeding are 0.7 respectively, and the revision coefficient for dangerous lane change is 0.6.

[0071] Optionally, by calculating Obtain the deviation between the abnormal operation data of the i-th spatial object and the average value; where E[I] = -1 / (m-1), V[I] = E[I 2 ]-E[I] 2 .

[0072] Optionally, by calculating Get the aggregation of all spatial weights.

[0073] Optionally, the spatial adjacency matrix is Each variable in the matrix is a spatial object, which includes abnormal operation status data and the location where it occurs.

[0074] In some embodiments, constructing an r×r normalized spatial adjacency matrix can effectively represent the location of r spatial objects or the proximity relationship of their affiliated regions, which is a key step in spatial autocorrelation analysis. This matrix needs to define the mutual adjacency relationship between spatial objects before performing spatial autocorrelation analysis, revealing the spatial correlation between geographical objects.

[0075] Optionally, based on the abnormal state spatial correlation data, identify the multiple occurrence points of the abnormal operation state of special sections, including: performing clustering analysis on each abnormal state spatial correlation data to obtain an aggregation index; identifying the abnormal operation state points according to the aggregation index.

[0076] Optionally, by calculating obtain the aggregation index; where x is the abnormal state spatial correlation data of region R, is the average value of the correlation data; w ij is the spatial weight matrix. The value range of the aggregation index I is between [-1, 1]. I less than 0 indicates negative correlation, I greater than 0 indicates positive correlation, and I equal to 0 indicates that the spatial object units within the study area i are independent of each other. The closer the value of I is to 1, the more significant the agglomeration effect of a certain attribute of the research object in spatial distribution; the closer the value of I is to -1, the more significant the divergence of a certain attribute of the research object in spatial distribution.

[0077] Optionally, based on the abnormal state spatial correlation data, identify the multiple occurrence points of the abnormal operation state of special sections, including: performing density estimation on each abnormal state spatial correlation data to obtain the density distribution of the abnormal state; identifying the abnormal operation state points according to the density distribution.

[0078] Optionally, by calculating obtain the density distribution of the abnormal state; as Figure 4 shown, K() is the kernel function, h is the window width, also known as the domain threshold for abnormal operation state clustering, m' is the number of points within the research area R, that is, the equivalent number of occurrences of abnormal operation state events, n' is the dimension of the spatial adjacency data, x - x i is the distance from the preset point position x to the abnormal state spatial correlation data, that is, the distance to the event x i at.

[0079] In some embodiments, since there are many kernel functions, in this embodiment, Python programming is used. Based on the machine learning method, the best kernel function is determined to be the Gaussian kernel function. Through this step, the accuracy of clustering fitting for the abnormal operation state of special sections of expressways is effectively ensured, and for different special scenarios, the best kernel function type can be adaptively obtained through prior data accumulation.

[0080] In some embodiments, when n = 2, the density distribution of the two-dimensional plane space is calculated by the following formula:

[0081]

[0082] In some embodiments, since geographical phenomena and events can occur at any location in the plane space, but the probabilities of occurrence vary depending on the location, the probability of events occurring in areas with dense points is high, and vice versa for areas with sparse points. Therefore, density estimation of the spatial correlation data of each abnormal state is particularly useful in analyzing and displaying point data. The density distribution is the highest at the center of each point and continuously decreases outward, and the density is 0 when the distance from the center reaches a certain threshold range (the edge of the window). Therefore, the locations corresponding to the distribution density greater than the preset threshold are determined as accident-prone points, so as to identify the accident-prone points in the characteristic road sections and provide technical support for the precise setting of roadside perception warning facilities in special road sections of expressways. Combined with Figure 5 As shown in the figure, a device for identifying accident-prone points of abnormal states in special road sections based on spatial autocorrelation includes: a construction module 101 configured to construct a dataset of abnormal operating states of special road sections; an acquisition module 102 configured to obtain a set of abnormal state characteristic categories according to the dataset of abnormal operating states; a spatial autocorrelation analysis module 103 configured to perform spatial autocorrelation analysis according to the set of abnormal state characteristic categories to obtain spatial correlation data of abnormal states; and an accident-prone point identification module 104 configured to identify accident-prone points of abnormal operating states in special road sections according to the spatial correlation data of abnormal states.

[0083] In some embodiments, not all abnormal operating states will evolve into traffic accidents in special road sections such as highway merging and diverging areas and tunnel entrances and exits. Some behaviors such as slow traffic and speeding only affect the local traffic flow state. Therefore, traditional accident statistics and risk analysis methods are not applicable to such road sections. Based on the distribution law of abnormal operating states in special road sections of expressways, this application uses the spatial autocorrelation method to weaken the accident type and strengthen the identification ability of abnormal points. The occurrence locations of abnormal operating states on special road sections are abstracted as a series of irregular random points distributed in the space of special road sections of expressways. The central tendency and dispersion degree of this space are described by the spatial point pattern statistic (point pattern), and the low-order and high-order statistical properties of the point distribution in the abnormal operating space are explored; a density function is selected to estimate the density of accident points in the plane space, and a model for identifying accident-prone locations of abnormal operating states in special road sections of expressways based on spatial autocorrelation is established, so as to realize the automatic identification of accident-prone points of abnormal operating states in special road sections.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention, and they should all be covered within the scope of the claims and the description of the present invention.

Claims

1. A method for identifying multiple abnormal state points on special road sections based on spatial autocorrelation, characterized in that Including: Constructing an abnormal operation state dataset for special sections, including: Using roadside cameras and roadside radars to provide traffic operation state data within special sections; Constructing an abnormal operation state dataset based on the traffic operation state data; Obtaining an abnormal state feature category set according to the abnormal operation state dataset; Performing spatial autocorrelation analysis according to the abnormal state feature category set to obtain abnormal state spatial association data, including: Obtaining the spatial weights of each abnormal behavior type according to the abnormal state feature category set and the length of the special section; Constructing a spatial adjacency matrix based on the abnormal operation state dataset, where the spatial adjacency matrix includes multiple spatial objects; Assigning values to each of the spatial objects according to each of the spatial weights according to the principle of distance decay to obtain each abnormal state spatial association data; Identifying multiple occurrence points of the abnormal operation state of the special section according to the abnormal state spatial association data.

2. The method according to claim 1, wherein Obtaining an abnormal state feature category set according to the abnormal operation state dataset, including: Determining the recognition features of multiple abnormal behavior types; Identifying the abnormal operation state dataset according to each of the recognition features to obtain an abnormal state feature category set, where the abnormal state feature category set includes multiple abnormal behavior types.

3. The method according to claim 1, wherein The spatial weight is obtained through the following formula: ； In the above formula, is the cell value of the i-th row and j-th column in the spatial weight matrix, is the coordinate of the first abnormal operating state, is the coordinate of the second abnormal operating state, is the difference in the longitudinal coordinates between the first abnormal operating state and the second abnormal operating state, and L is the length of the special section.

4. The method according to claim 1, characterized in that, The abnormal state spatial association data is obtained through the following formula: ； Among them, is the associated data of the abnormal state space, is the deviation between the abnormal operation state data of the i-th spatial object and the average value, is the equivalent only property damage revision factor, is the cell value of the i-th row and j-th column in the spatial weight matrix, and m is the total number of abnormal operation states, is the aggregation of all spatial weights.

5. The method according to claim 1, characterized in that Identifying multiple occurrence points of the abnormal operation state of the special section according to the abnormal state spatial association data, including: Performing clustering analysis on each of the abnormal state spatial association data to obtain an aggregation index; Identifying the points of abnormal operation state according to the aggregation index.

6. The method according to claim 1, characterized in that Identifying multiple occurrence points of the abnormal operation state of the special section according to the abnormal state spatial association data, including: Performing density estimation on each of the abnormal state spatial association data to obtain the density distribution of the abnormal state; Identifying the points of abnormal operation state according to the density distribution.

7. An identification device for multiple abnormal state points on special road sections based on spatial autocorrelation, characterized in that, Including: A construction module configured to construct an abnormal operation state dataset for special sections, including: Using roadside cameras and roadside radars to provide traffic operation state data within special sections; Constructing an abnormal operation state dataset based on the traffic operation state data; An acquisition module configured to obtain an abnormal state feature category set according to the abnormal operation state dataset; A spatial autocorrelation analysis module configured to perform spatial autocorrelation analysis according to the abnormal state feature category set to obtain abnormal state spatial association data, including: Obtaining the spatial weights of each abnormal behavior type according to the abnormal state feature category set and the length of the special section; Constructing a spatial adjacency matrix based on the abnormal operation state dataset, where the spatial adjacency matrix includes multiple spatial objects; Assigning values to each of the spatial objects according to each of the spatial weights according to the principle of distance decay to obtain each abnormal state spatial association data; A multiple occurrence point identification module configured to identify multiple occurrence points of the abnormal operation state of the special section according to the abnormal state spatial association data.

Images