CN108091131B - Traffic incident identification method and device - Google Patents
Traffic incident identification method and device Download PDFInfo
- Publication number
- CN108091131B CN108091131B CN201611035883.5A CN201611035883A CN108091131B CN 108091131 B CN108091131 B CN 108091131B CN 201611035883 A CN201611035883 A CN 201611035883A CN 108091131 B CN108091131 B CN 108091131B
- Authority
- CN
- China
- Prior art keywords
- traffic
- parameters
- road
- predicted
- classification
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
- G08G1/0116—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from roadside infrastructure, e.g. beacons
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
- G08G1/0133—Traffic data processing for classifying traffic situation
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/042—Detecting movement of traffic to be counted or controlled using inductive or magnetic detectors
Abstract
The invention discloses a traffic incident identification method, which comprises the following steps: receiving traffic parameters of a road to be predicted, which are sent by a road side unit; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted; classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result. The invention also discloses a traffic incident recognition device.
Description
Technical Field
The invention relates to the technology of Internet of things, in particular to a traffic incident identification method and device.
Background
The traffic event refers to an occasional event affecting traffic safety and smoothness on a highway, such as a vehicle traffic accident, a fault parking, a control and the like. Traffic events have a serious impact on road traffic, and reports have shown that 60% of traffic congestion is caused by aperiodic congestion caused by traffic events, and has increased dramatically year by year. Therefore, the traffic event is detected, and the change of the traffic flow caused by the traffic event is predicted, so that the improvement of the road traffic efficiency is greatly influenced.
However, the existing traffic incident detection is mainly performed based on video files, and therefore a plurality of cameras need to be arranged on a road, each camera transmits the video files to a platform, and then visual analysis is performed according to the video files, so that traffic incidents are obtained, and the method is high in cost; and when extreme weather occurs, the identification of traffic events is inefficient.
Disclosure of Invention
In order to solve the existing technical problem, embodiments of the present invention provide a method and an apparatus for identifying a traffic event.
The technical scheme of the embodiment of the invention is realized as follows:
the embodiment of the invention provides a traffic incident identification method, which comprises the following steps:
receiving traffic parameters of a road to be predicted, which are sent by a Road Side Unit (RSU); the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted;
classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result;
and identifying the traffic incident on the road to be predicted by using the classification result.
In the above scheme, the classifying the traffic parameters by using the traffic parameters and by using a hierarchical classification manner to obtain a classification result includes:
carrying out clustering analysis by using the traffic parameters to obtain a first clustering result;
and performing secondary classification by using the primary clustering result and combining the traffic parameters to obtain the classification result.
In the above scheme, performing cluster analysis by using the traffic parameters to obtain a first clustering result includes:
and performing cluster analysis by using the traffic parameters and combining at least two first traffic flows and a cluster algorithm.
In the above scheme, the traffic parameters include: traffic flow, traffic flow average speed and time occupancy; the time occupancy is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length;
the cluster analysis is carried out by utilizing the traffic parameters and combining at least two first traffic flows and a cluster algorithm, and comprises the following steps:
dividing the average speed and the time occupancy of the traffic flow in the traffic parameters according to the at least two first traffic flows and the traffic flow in the traffic parameters to obtain divided data;
and obtaining the first clustering result by using the divided data and combining a K-means (K-means) method.
In the above scheme, the performing the second classification by using the first clustering result and combining the traffic parameter to obtain the classification result includes:
dividing the traffic parameters which are not classified in the first clustering result by using at least two set second traffic flows to obtain each divided group of data; the second traffic flow is not equal to the first traffic flow;
and classifying the divided groups of data based on the established model to obtain the classification result.
In the above scheme, the method further comprises:
sending a traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event.
The embodiment of the invention also provides a traffic incident recognition device, which comprises:
the third receiving unit is used for receiving the traffic parameters of the road to be predicted, which are sent by the road side unit; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted;
the identification unit is used for classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result.
In the foregoing solution, the identification unit is specifically configured to:
carrying out clustering analysis by using the traffic parameters to obtain a first clustering result;
and performing secondary classification by using the primary clustering result and combining the traffic parameters to obtain the classification result.
In the foregoing solution, the identification unit is specifically configured to:
and performing cluster analysis by using the traffic parameters and combining at least two first traffic flows and a cluster algorithm.
In the above scheme, the apparatus further comprises:
the third sending unit is used for sending the traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event.
The traffic incident identification method and the device provided by the embodiment of the invention receive the traffic parameters of the road to be predicted, which are sent by the road side unit; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted; classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result. The traffic event identification is carried out by utilizing the traffic flow information acquired by the sensing coil, a large amount of data does not need to be transmitted, and the network burden is reduced; meanwhile, the traffic incident is identified by adopting a hierarchical classification mode, so that the identification accuracy can be greatly improved.
Drawings
In the drawings, which are not necessarily drawn to scale, like reference numerals may describe similar components in different views. Like reference numerals having different letter suffixes may represent different examples of similar components. The drawings illustrate generally, by way of example, but not by way of limitation, various embodiments discussed herein.
Fig. 1 is a flow chart illustrating a method for identifying a traffic event at an RSU side according to an embodiment of the present invention;
FIG. 2 is a flow chart illustrating a method for identifying a traffic incident on a platform according to an embodiment of the present invention;
FIG. 3 is a system architecture diagram according to a second embodiment of the present invention;
FIG. 4 is a flowchart illustrating a second method for identifying a traffic event according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a clustering result of 500 initial traffic flows per hour according to a second embodiment of the present invention;
FIG. 6 is a schematic diagram of a clustering result under different initial traffic flows according to a second embodiment of the present invention;
FIG. 7a is a schematic diagram of the trend of the average speed of the second embodiment of the present invention with the initial traffic flow less than 1500 vehicles/hour;
FIG. 7b is a schematic diagram of the trend of the time occupancy change when the initial traffic flow is less than 1500 vehicles/hour according to the second embodiment of the present invention;
FIG. 8a is a schematic diagram illustrating the trend of the average speed of the second embodiment of the present invention with the initial traffic flow greater than 2400 vehicles/hour;
FIG. 8b is a schematic diagram illustrating a trend of a time occupancy change with an initial traffic flow greater than 2400 vehicles/hour according to an embodiment of the present invention;
fig. 9 is a schematic structural diagram of a traffic incident recognition apparatus according to a third embodiment of the present invention;
fig. 10 is a schematic structural diagram of another traffic event recognition device according to the embodiment of the invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
The existing traffic incident detection and identification method is mainly based on video file identification, and the method needs to arrange a plurality of cameras on a road, and each camera transmits a video file to a platform for analysis. The method needs to arrange a large number of cameras and needs to transmit a large number of video files, so that the cost is high. Moreover, in the video processing process, when extreme weather such as heavy fog and strong light occurs, the efficiency of video detection is reduced, the efficiency of identifying traffic incidents is reduced, and the identification effect is not ideal.
Therefore, the inventor finds out in the process of implementing the invention that: data collection can be carried out based on the road sensing coil, and traffic events can be identified. The basic idea is as follows: and analyzing the historical traffic state data, finding out the traffic state parameter characteristics when the traffic incident occurs, comparing the traffic state parameter characteristics with the current traffic state parameters, and analyzing whether the traffic incident occurs at present, thereby identifying the traffic incident.
However, the inventors have found that this method has certain drawbacks: when the data collected by the sensing coil is used for classifying traffic events of different categories, if the traffic events of all the categories are to be distinguished at one time, the situation of wrong classification often occurs, and the reason for the situation is that the training of the model is concentrated, the characteristic difference of some categories is not obvious, and the classification effect is poor.
Based on this, in various embodiments of the invention: the platform receives traffic parameters of a road to be predicted, which are sent by the RSU; the traffic parameters are obtained according to traffic flow information collected by each sensing coil laid on the road to be predicted; classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result.
Example one
The traffic incident identification method of the embodiment of the invention is applied to an RSU (road side unit), and comprises the following steps as shown in figure 1:
step 101: receiving traffic flow information collected by each sensing coil paved on a road to be predicted;
step 102: determining the traffic parameters of the road to be predicted by using the current traffic flow information and the historical traffic flow information;
step 103: and sending the determined traffic parameters to the platform.
Here, the traffic parameters may include a traffic flow, a traffic flow average speed, and a time occupancy; the time occupancy rate is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length.
Wherein, the traffic volume is as follows: the number of vehicles passing the sensing coil per unit time. The unit is vehicle/h.
The average speed of the traffic flow is as follows: the average speed of the vehicle over the sensing coil.
The determined traffic parameters are used for traffic event identification. Specifically, the platform identifies the traffic events on the road to be predicted by using the traffic parameters in a hierarchical classification mode.
In practical application, when the traffic event recognition result represents that a traffic event occurs on the road to be predicted, the platform can send the traffic event to the RSU so as to give an early warning to related vehicles, avoid entering a traffic jam road section and improve user experience.
Based on this, in an embodiment, the method may further include:
receiving a traffic event identification result on a road to be predicted, which is sent by the platform; the traffic event identification result comprises the category and the occurrence place of the traffic event;
and sending the type and the occurrence place of the traffic incident, and early warning the related vehicles.
Here, the relevant vehicle refers to a vehicle that can communicate with the RSU.
Correspondingly, the embodiment also provides a traffic event identification method, which is applied to a platform, and as shown in fig. 2, the method includes the following steps:
step 201: receiving traffic parameters of a road to be predicted, which are sent by a road RSU;
here, the traffic parameter is a traffic parameter obtained from traffic flow information collected by each sensing coil on the road to be predicted.
Step 202: classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result;
specifically, the traffic parameters are utilized to perform clustering analysis to obtain a first clustering result;
and performing secondary classification by using the primary clustering result and combining the traffic parameters to obtain the classification result.
The method for performing cluster analysis by using the traffic parameters to obtain a first clustering result comprises the following steps:
and performing cluster analysis by using the traffic parameters and combining at least two first traffic flows and a cluster algorithm.
Here, in practical use, the traffic parameters may include: traffic flow, traffic flow average speed and time occupancy; the time occupancy is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length;
the cluster analysis is carried out by utilizing the traffic parameters and combining at least two first traffic flows and a cluster algorithm, and comprises the following steps:
dividing the average speed and the time occupancy of the traffic flow in the traffic parameters according to the at least two first traffic flows and the traffic flow in the traffic parameters to obtain divided data;
and obtaining the first clustering result by using the divided data and combining a K-means method.
In practical application, the first traffic flow may be set as required, for example, the corresponding data may be divided every 100 vehicles/hour.
In practical applications, the first traffic flow may also be referred to as an initial traffic flow.
And performing secondary classification by using the first clustering result and combining the traffic parameters to obtain a classification result, wherein the classification result comprises the following steps:
dividing the traffic parameters which are not classified in the first clustering result by using at least two set second traffic flows to obtain each divided group of data; the second traffic flow is not equal to the first traffic flow;
and classifying the divided groups of data based on the established model to obtain the classification result.
Specifically, the traffic parameters which are not classified in the first clustering result are classified for the second time, and specifically, the selected data range is narrowed or expanded according to the traffic parameters which are not classified, and the traffic events on the road to be predicted are further determined through a machine learning algorithm.
In the second classification process, machine learning algorithms such as a random forest, a decision tree and a neural network can be selected to determine the traffic incident on the road to be predicted.
The second traffic flow needs to be set as required.
The determined traffic event outcome on the road to be predicted may include the category and the location of occurrence of the traffic event.
And the position corresponding to the point where the traffic parameter changes is the occurrence place of the traffic incident.
Step 203: and identifying the traffic incident on the road to be predicted by using the classification result.
That is, what type of traffic event the classified data each corresponds to is found.
When the traffic event recognition result represents that a traffic event occurs on the road to be predicted, the platform sends the traffic event recognition result to the RSU; the traffic event recognition result comprises the category and the occurrence place of the traffic event, so that the RSU gives an alarm to the relevant vehicle.
It should be noted that: in actual application, the RSU and the platform carry out information interaction through the base station.
In the traffic incident identification method provided by the embodiment of the invention, an RSU receives traffic flow information collected by each sensing coil laid on a road to be predicted; determining the traffic parameters of the road to be predicted by using the current traffic flow information and the historical traffic flow information; sending the determined traffic parameters to the platform; and the platform identifies the traffic events on the road to be predicted by using the traffic parameters and adopting a hierarchical classification mode. The traffic event identification is carried out by utilizing the traffic flow information acquired by the sensing coil, a large amount of data does not need to be transmitted, and the network burden is reduced; meanwhile, the traffic incident is identified by adopting a hierarchical classification mode, so that the identification accuracy can be greatly improved.
In addition, when the traffic event identification result represents that a traffic event occurs on the road to be predicted, the platform sends the traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event; and the RSU sends the type and the occurrence place of the traffic incident to early warn relevant vehicles, and when the traffic incident occurs, the vehicles can avoid congested road sections in time, so that the passing efficiency is improved.
Example two
On the basis of the first embodiment, the present embodiment describes in detail a process of traffic event identification.
Fig. 3 is a system architecture diagram according to an embodiment of the invention. As can be seen from fig. 3, the scheme of the embodiment of the present invention is: the method comprises the steps that a sensing coil collects vehicle motion data, the collected data (traffic flow information) are transmitted to an RSU, the RSU calculates traffic parameters of a road section and sends the traffic parameters to a cloud platform through a base station, the cloud platform analyzes the collected data for a period of time, related information of a traffic incident is analyzed, a traffic incident classification result and traffic incident occurrence place information are sent to the RSU through the base station, and the RSU broadcasts the information to a vehicle to remind the vehicle of avoiding the traffic incident occurrence place.
Fig. 4 is a schematic flow chart illustrating the traffic event recognition according to the embodiment. With reference to fig. 4, the process of identifying a traffic event in this embodiment includes the following steps:
step 401: the sensing coil collects traffic flow information of a road and sends the traffic flow information to the RSU;
at present, a road is generally paved with a sensing coil, when a vehicle passes through the sensing coil, a signal of the sensing coil is triggered, the sensing coil records information that the vehicle passes through, and can record time used by the vehicle to pass through the sensing coil, so that traffic flow information is obtained; and then the moving speed and the time occupancy of the vehicle can be calculated.
The sensing coils are each equipped with a communication device by which the recorded information is transmitted to the RSU.
Step 402: the RSU determines traffic parameters of the road to be predicted based on the received sensing coil data and transmits the traffic parameters to the cloud platform;
that is, the RSU collates the data based on the received sensor coil data, calculates traffic parameters (e.g., every 30 seconds), and transmits the traffic parameters (e.g., transmits the calculated traffic parameters to the cloud platform every 5 minutes).
Here, after receiving the data of the sensing coil, the RSU calculates the traffic volume, average speed, and time occupancy parameters, specifically:
(1) traffic volume:
the traffic volume refers to the actual number of vehicles passing through a certain place or a certain section of a road in unit time, and in the embodiment of the invention, the traffic volume refers to the number of vehicles passing through the sensing coil. The traffic volume is divided into daily traffic volume, hourly traffic volume, etc., and in the present embodiment, the unit of the traffic volume is vehicle/h.
(2) Average speed of traffic flow
The traffic flow average speed is generally divided into a time average speed at a specific point and a section average speed on a specific link. The time average speed is the arithmetic average of the instantaneous speeds of all vehicles passing through a section in the observation time, and the interval average speed is the quotient of the observation distance and the average travel time for the vehicles to pass through the observation distance. The former represents the operation condition of the traffic flow at a specific observation place, and the latter represents the operation condition of the traffic flow on a specific link space. When the two speed values are obviously lower than the normal value, the traffic accident happens at the traffic of the observation place or the observation road section.
In the present embodiment, the traffic flow average speed refers to the average speed at which the vehicle passes over the sensor coil.
(3) Time occupancy
The time occupancy is determined at a certain observation timeTThe ratio of the total time of the sensing coil occupied by the vehicle to the observation time length is calculated by the formula
。
Wherein the content of the first and second substances,
is as followsiThe time that the vehicle occupies the sensing coil,Tis the length of the observation period.
The size of the time occupancy can reflect the traffic running state. Specifically, in the case where the traffic flow is small, the number of vehicles passing through the sensor coil per unit time is small, and the time occupancy is low because the vehicle speed is high. As the traffic flow increases, the number of vehicles passing through the sensor coil per unit time increases and the vehicle speed decreases, so that the time taken by the sensor coil by the vehicles increases and the time occupancy rate significantly increases. When a traffic event occurs, the traffic volume through the sensing coil may be reduced, but the time occupancy is still at a higher level due to the significantly reduced vehicle speed. For each road segment, the time occupancy on all its sensing coils is calculated.
And after calculating the traffic flow, the average speed and the time occupancy parameters at the sensing coil, sending the calculated traffic parameters to the cloud platform side every 5 minutes, and sending a timestamp of each piece of data at the same time so that the cloud platform can distinguish the time corresponding to the data.
Step 403: the cloud platform carries out clustering based on the received traffic parameters, and further classifies the traffic parameters in different categories to obtain a traffic event identification result;
that is, the cloud platform processes and learns the data, and analyzes whether a traffic event occurs, the type of the traffic event, and the occurrence location at present.
Here, when classifying the traffic incident, if the original data is directly classified once, the experimental effect is analyzed and known: when data are collected on a road section with a small traffic flow, the vehicle running speed caused by a traffic event is often not obvious when the vehicle running speed is reduced, namely the characteristics of the traffic event are not obvious, so that the situations of no traffic event and the situations of traffic events cannot be distinguished, and the classification effect is poor. Meanwhile, when data are collected on a road section with a large traffic flow, traffic jam can be easily caused by traffic events, the traffic operation parameter characteristics under the two conditions of the traffic events and the traffic jam can be not distinguished remarkably, the conditions of the traffic events are mistakenly divided into the conditions of the traffic jam, and therefore the classification effect is poor.
Therefore, when the initial traffic flows (which can be set) of the roads are different, the classification of the cases with traffic incidents is easy to be wrong, which is specifically represented by: when the initial traffic flow is small, the traffic flow is easily wrongly classified as no traffic; it is susceptible to misclassification as traffic congestion when the initial traffic flow is large. Meanwhile, since different initial traffic flows may cause erroneous classification of the occurrence of a traffic accident, it is necessary to identify a critical point of the traffic flow in order to prepare for subsequent further classification.
Based on this, in order to solve the problems of poor classification effect and traffic flow critical point identification, the embodiment adopts a hierarchical classification method to identify the traffic events. The overall classification method comprises the following steps:
(1) dividing the training set according to different initial traffic flows so as to divide data corresponding to the different initial traffic flows;
such as: and dividing every 100 vehicles/hour according to the initial traffic flow to obtain divided data.
Wherein the training set is the received traffic parameters.
(2) And respectively carrying out cluster analysis on the divided data corresponding to different initial traffic flows.
In the method, the K-means method is adopted for clustering, so that whether the characteristics among data are obvious or not can be better reflected. The K-means algorithm is used for clustering samples into K clusters (cluster), and the specific algorithm is described as follows:
1. randomly selecting k cluster centroids (cluster centroids) as
。
2. The following process is repeated until convergence
{ for each sample i, calculate the class to which it should belong
For each class j, the centroid of the class is recalculated
Because the data corresponding to each initial traffic flow comprises the data of three conditions of occurrence of traffic events, traffic jams and no-traffic events, if the data characteristics of the three conditions are obvious, three types can be clustered, namely the three types of occurrence of the traffic events, the traffic jams and the no-traffic events. If the features are not obvious, then there are less than three types clustered.
Taking data with an initial traffic flow of 500 vehicles/hour as an example to illustrate the clustering effect, as shown in fig. 5, the training set has 104 pieces of data, wherein 41 pieces of data with traffic accidents, 31 pieces of traffic jam data and 64 pieces of data without traffic incidents are clustered by using a K-means method to obtain two types of data. The number of the first type data is 31, and the number of the second type data is 73. According to data comparison and analysis, all data with traffic incidents and data without the traffic incidents are clustered into the same class, and traffic jam data is of another class, so that the data characteristics of the traffic incidents and the data without the traffic incidents are not obviously different under the traffic flow.
(3) Analyzing the clustering results under different initial traffic flows to obtain the clustering result shown in fig. 6;
as can be seen from fig. 6, when the road traffic flow is small, the data difference between the traffic event and the traffic-free event is small, when the road traffic flow is large, the data difference between the traffic event and the traffic jam is small, and when the traffic flow is moderate, the data difference between the traffic event, the traffic jam and the traffic-free event is large.
Specifically, when the initial traffic flow of the road is between 1500-2400 vehicles/hour, 3 classes can be clustered, and the accuracy of the clustered classes is high, so that the data in the range can be directly classified to identify the traffic event, when the initial traffic flow is not in the range, 2 classes can be clustered, and as can be seen from data comparison, the characteristics of the traffic event are not obvious in the two cases, so that further classification is needed, and the identification accuracy is improved.
(4) Based on the clustering result, further classifying;
that is, the data of which the initial traffic flow is not in the range of 1500-2400 vehicles/hour is used for further classification.
Through analysis, the following results can be found: when the initial traffic flow of the road is less than 1500 vehicles/hour, the traffic jam data can be easily screened out after the clustering is carried out by the K-means method, and the data of traffic events and traffic-free events are not easy to screen. As can be seen from comparative analysis of the raw data of the two cases, the traffic event data is mainly different from the data of the traffic-free event in the range of about 100 meters around the traffic accident occurrence point. In this range, as shown in fig. 7a-b, the average speed is decreased and then increased, and the time occupancy is increased and then decreased in the data with the occurrence of the traffic event, while the data without the occurrence of the traffic event is not significantly changed. Therefore, the selected data range is narrowed (namely the second traffic flow is smaller than the initial traffic flow) during identification, data in the range of every 100 meters are compared and analyzed, when the change trends of two groups of data (the data with traffic events and the data without traffic events) are found to be different, the traffic events can be judged to be in the range, and the data with the changed average speed and time occupancy is the traffic accident data correspondingly, so that the identification process is completed. Wherein, in figure 7a,
indicating the trend of the average speed change at the occurrence of a traffic event,
indicating the trend of the average speed change when no traffic event occurs. In the context of figure 7b of the drawings,
represents the time occupancy change trend when the traffic incident occurs,
indicating the trend of time occupancy change when no traffic events occur.
When the traffic flow is larger than 2400 vehicles/hour, the data without traffic incidents can be screened out after the clustering by the K-means method, and the trafficEvent and traffic congestion data are not easily screened. As can be seen from the comparative analysis of the raw data in the two cases, the traffic event data and the traffic jam data are different in that, as shown in fig. 8a-b, the traffic event data shows the process that the average speed is decreased and then increased, and the time occupancy and the traffic flow are increased and then decreased, while the average speed in the traffic jam data is gradually decreased to be stable, and the time occupancy and the traffic flow are gradually increased to be stable. Therefore, when the classification is performed, the data selection range needs to be expanded (namely the second traffic flow is larger than the initial traffic flow), the change conditions of different traffic parameters in a larger range are contrasted and analyzed, and when the data meet different change trends, the data correspond to traffic events and traffic jams, so that the identification process is completed. Wherein, in figure 8a,
indicating the trend of the average speed change at the occurrence of a traffic event,
indicating the trend of the average speed change when no traffic event occurs. In the case of figure 8b of the drawings,
represents the time occupancy change trend when the traffic incident occurs,
indicating the trend of time occupancy change when no traffic events occur.
Based on the analysis, in practical application, machine learning algorithms such as random forests, decision trees, neural networks and the like can be selected, average speed, time occupancy and traffic flow in a certain range are used as input, the type of the traffic event is used as output, and then a classification result can be obtained, so that a traffic event recognition result is finally obtained. In addition, the point at which the general traffic parameter changes is the place at which the traffic event occurs.
Step 404: and the cloud platform issues the classification result of the traffic incident to the RSU, and the RSU broadcasts the classification result to inform surrounding vehicles.
Here, when the cloud platform recognizes that a traffic event occurs, the cloud platform issues the type of the traffic event and the location information of the occurrence to the RSU, and the RSU broadcasts the information to inform surrounding vehicles.
The RSU may transmit the category of the traffic event and the location information of the occurrence to the surrounding vehicles through a vehicle-to-vehicle (V2X) device.
In practical application, when the RSU broadcasts the traffic event information, if the traffic event information of other road sections sent by other RSUs is received, the information can also be sent to the relevant vehicles.
Specifically, when the RSU receives the traffic jam or the traffic event in front of the current road section, the RSU broadcasts the traffic event information of the road in front in time to remind the vehicle of avoiding. Meanwhile, traffic information of surrounding smooth roads can be sent to the vehicles, the vehicles are reminded of being capable of driving on other roads, and the vehicles are prevented from entering a traffic jam road section.
In summary, in the solution provided by the embodiment of the present invention, the RSU mainly utilizes the road sensing coil (the vehicle sensing coil transmits information to the RSU through the V2X device), collects traffic flow information, and sends traffic parameters obtained based on the data of the sensing coil to the cloud platform; the cloud platform adopts a hierarchical classification method, namely, current data characteristics are firstly analyzed through clustering, a plurality of types with obvious characteristics are classified, then, according to characteristics of initial traffic flow and traffic parameters, a data range selected for classification is determined, namely, data in different ranges are selected, and secondary classification is carried out through machine learning methods such as random forests, decision trees, neural networks and the like, so that traffic events are identified.
Compared with a mode of carrying out traffic incident identification based on video and a mode of directly carrying out classification based on sensing coil data, the scheme provided by the embodiment of the invention is more targeted, can cover different traffic conditions, and can obviously improve the accuracy of traffic incident identification.
In addition, the traffic event information is transmitted through facilities such as RSUs and base stations, and the traffic event information in a wider range can be acquired, so that the traffic event information has a remarkable effect of improving the safety and efficiency of road traffic.
EXAMPLE III
To implement the method of the embodiment of the present invention, the embodiment provides a traffic event recognition device, which is disposed in an RSU, as shown in fig. 9, and the device includes:
the first receiving unit 91 is used for receiving traffic flow information collected by each sensing coil laid on a road to be predicted;
a determining unit 92, configured to determine a traffic parameter of the road to be predicted by using the current traffic flow information and the historical traffic flow information;
a first sending unit 93, configured to send the determined traffic parameter to the platform; the determined traffic parameters are used for traffic event identification.
Here, the traffic parameters may include a traffic flow, a traffic flow average speed, and a time occupancy; the time occupancy rate is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length.
Wherein, the traffic volume is as follows: the number of vehicles passing the sensing coil per unit time. The unit is vehicle/h.
The average speed of the traffic flow is as follows: the average speed of the vehicle over the sensing coil.
The determined traffic parameters are used for traffic event identification. Specifically, the platform identifies the traffic events on the road to be predicted by using the traffic parameters in a hierarchical classification mode.
In practical application, when the traffic event recognition result represents that a traffic event occurs on the road to be predicted, the platform can send the traffic event to the RSU so as to give an early warning to related vehicles, avoid entering a traffic jam road section and improve user experience.
Based on this, in an embodiment, the apparatus may further include:
the second receiving unit is used for receiving the traffic event identification result on the road to be predicted, which is sent by the platform;
and the second sending unit is used for sending the traffic incident and early warning relevant vehicles when the traffic incident identification result represents that the traffic incident occurs on the road to be predicted.
In practical applications, the first receiving unit 91 and the second sending unit may be implemented by a V2X device in the traffic event recognition device; the determination Unit 92 may be implemented by a Central Processing Unit (CPU), a Microprocessor (MCU), a Digital Signal Processor (DSP), or a Programmable logic Array (FPGA) in the traffic event recognition device; the first sending unit 93 and the second receiving unit may be implemented by a transceiver in the traffic event recognition device.
In order to implement the method according to the embodiment of the present invention, the embodiment further provides a traffic event recognition device, which is disposed on the platform, as shown in fig. 10, and includes:
a third receiving unit 101, configured to receive traffic parameters of a road to be predicted, where the traffic parameters are sent by an RSU; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted;
the identification unit 102 is configured to classify the traffic parameters by using the traffic parameters in a hierarchical classification manner to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result.
The identification unit 102 is specifically configured to:
carrying out clustering analysis by using the traffic parameters to obtain a first clustering result;
and performing secondary classification by using the primary clustering result and combining the traffic parameters to obtain the classification result.
The method for performing cluster analysis by using the traffic parameters to obtain a first clustering result comprises the following steps:
the recognition unit 102 performs cluster analysis by using the traffic parameters in combination with at least two first traffic flows and a clustering algorithm.
Here, in practical use, the traffic parameters may include: traffic flow, traffic flow average speed and time occupancy; the time occupancy is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length;
the cluster analysis is carried out by utilizing the traffic parameters and combining at least two first traffic flows and a cluster algorithm, and comprises the following steps:
the identification unit 102 divides the average speed and the time occupancy of the traffic flow in the traffic parameter according to the at least two first traffic flows and the traffic flow in the traffic parameter to obtain divided data;
and obtaining the first clustering result by using the divided data and combining a K-means method.
In practical application, the first traffic flow may be set as required, for example, the corresponding data may be divided every 100 vehicles/hour.
In practical applications, the first traffic flow may also be referred to as an initial traffic flow.
And performing a second classification by using the first clustering result and combining the traffic parameters to obtain a classification result, wherein the classification result comprises the following steps:
the identification unit 102 divides the traffic parameters which are not classified in the first clustering result by using at least two set second traffic flows to obtain each divided group of data; the second traffic flow is not equal to the first traffic flow;
the identification unit 102 classifies each divided group of data based on the established model to obtain the classification result.
Specifically, the traffic parameters which are not classified in the first clustering result are classified for the second time, and specifically, the selected data range is narrowed or expanded according to the traffic parameters which are not classified, and the traffic events on the road to be predicted are further determined through a machine learning algorithm.
In the second classification process, machine learning algorithms such as a random forest, a decision tree and a neural network can be selected to determine the traffic incident on the road to be predicted.
The second traffic flow needs to be set as required.
The determined traffic event outcome on the road to be predicted may include the category and the location of occurrence of the traffic event.
And the position corresponding to the point where the traffic parameter changes is the occurrence place of the traffic incident.
When the traffic event recognition result represents that a traffic event occurs on the road to be predicted, the traffic event recognition result can be sent to the RSU; the traffic event recognition result comprises the category and the occurrence place of the traffic event, so that the RSU gives an alarm to the relevant vehicle.
Based on this, in an embodiment, the apparatus may further include:
the third sending unit is used for sending the traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event.
It should be noted that: in actual application, the RSU and the platform carry out information interaction through the base station.
In practical applications, the third receiving unit 101 and the third transmitting unit may be implemented by a transceiver in the traffic event recognition device; the identification unit 102 may be implemented by a CPU, MCU, DSP or FPGA in the traffic event identification device.
According to the scheme provided by the embodiment of the invention, the RSU receives traffic flow information collected by each sensing coil paved on a road to be predicted; determining the traffic parameters of the road to be predicted by using the current traffic flow information and the historical traffic flow information; sending the determined traffic parameters to the platform; the platform classifies the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; and identifying the traffic incident on the road to be predicted by using the classification result. The traffic event identification is carried out by utilizing the traffic flow information acquired by the sensing coil, a large amount of data does not need to be transmitted, and the network burden is reduced; meanwhile, the traffic incident is identified by adopting a hierarchical classification mode, so that the identification accuracy can be greatly improved.
In addition, when the traffic event identification result represents that a traffic event occurs on the road to be predicted, the platform sends the traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event; and the RSU sends the type and the occurrence place of the traffic incident to early warn relevant vehicles, and when the traffic incident occurs, the vehicles can avoid congested road sections in time, so that the passing efficiency is improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.
Claims (4)
1. A traffic event identification method, the method comprising:
receiving traffic parameters of a road to be predicted, which are sent by a road side unit; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted;
classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; the traffic parameters include: traffic flow, traffic flow average speed and time occupancy; the time occupancy is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length;
identifying the traffic incident on the road to be predicted by using the classification result;
in the process of classifying the traffic parameters in a hierarchical classification mode, when a preset first traffic flow is not in a preset range, clustering to obtain traffic parameters with unobvious characteristics of traffic events; carrying out secondary classification aiming at traffic parameters with unobvious characteristics of traffic events to obtain a classification result;
obtaining traffic parameters with unobvious characteristics of the traffic events through clustering; and performing secondary classification aiming at traffic parameters with unobvious characteristics of the traffic events to obtain a classification result, wherein the classification result comprises the following steps:
dividing the average speed and the time occupancy of the traffic flow in the traffic parameters according to at least two first traffic flows and the traffic flow in the traffic parameters to obtain divided data;
obtaining a first clustering result by using the divided data and combining a K clustering method;
dividing the traffic parameters which are not classified in the first clustering result by using at least two set second traffic flows to obtain each divided group of data; the second traffic flow is not equal to the first traffic flow;
and classifying the divided groups of data based on the established model to obtain the classification result.
2. The method of claim 1, further comprising:
sending a traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event.
3. A traffic event recognition apparatus, the apparatus comprising:
the third receiving unit is used for receiving the traffic parameters of the road to be predicted, which are sent by the road side unit; the traffic parameters are obtained according to traffic flow information acquired by each sensing coil on the road to be predicted;
the identification unit is used for classifying the traffic parameters by using the traffic parameters in a hierarchical classification mode to obtain a classification result; identifying the traffic incident on the road to be predicted by using the classification result; the traffic parameters include: traffic flow, traffic flow average speed and time occupancy; the time occupancy is the ratio of the sum of the time occupied by the vehicle by the sensor coil on the road to be predicted to the observation time length;
in the process of classifying the traffic parameters in a hierarchical classification mode, when a preset first traffic flow is not in a preset range, clustering to obtain traffic parameters with unobvious characteristics of traffic events; carrying out secondary classification aiming at traffic parameters with unobvious characteristics of traffic events to obtain a classification result;
the identification unit is specifically configured to: dividing the average speed and the time occupancy of the traffic flow in the traffic parameters according to at least two first traffic flows and the traffic flow in the traffic parameters to obtain divided data; obtaining a first clustering result by using the divided data and combining a K clustering method; dividing the traffic parameters which are not classified in the first clustering result by using at least two set second traffic flows to obtain each divided group of data; the second traffic flow is not equal to the first traffic flow; and classifying the divided groups of data based on the established model to obtain the classification result.
4. The apparatus of claim 3, further comprising:
the third sending unit is used for sending the traffic event identification result to the road side unit; the traffic event identification result comprises the category and the occurrence place of the traffic event.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611035883.5A CN108091131B (en) | 2016-11-22 | 2016-11-22 | Traffic incident identification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611035883.5A CN108091131B (en) | 2016-11-22 | 2016-11-22 | Traffic incident identification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108091131A CN108091131A (en) | 2018-05-29 |
| CN108091131B true CN108091131B (en) | 2021-06-04 |
Family
ID=62169956
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611035883.5A Active CN108091131B (en) | 2016-11-22 | 2016-11-22 | Traffic incident identification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108091131B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109285349B (en) * | 2018-11-02 | 2020-09-22 | 浙江海康智联科技有限公司 | Method for detecting highway traffic incident under cooperative vehicle and road environment and early warning system |
| CN111404976B (en) * | 2019-01-02 | 2023-05-09 | 中国移动通信有限公司研究院 | Data processing method, device and system, vehicle-mounted terminal and medium |
| CN110097753B (en) * | 2019-04-04 | 2021-04-02 | 安徽中交汇能网络科技有限公司 | Secondary accident early warning method, device and system and road monitoring device |
| TWI774984B (en) * | 2019-09-05 | 2022-08-21 | 中華電信股份有限公司 | Traffic incident detection system and method |
| CN111341098B (en) * | 2020-02-18 | 2021-07-27 | 北京中旖新能源有限公司 | Congestion state prediction method and device |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100433072C (en) * | 2007-03-01 | 2008-11-12 | 上海三全科技有限公司 | The single-loop and single-section onsite detection method for traffic situation and traffic accident |
| CN101188064B (en) * | 2007-12-20 | 2011-01-19 | 北京交通大学 | 3D integrated freeway traffic event automatic detection method |
| CN102436740A (en) * | 2011-09-29 | 2012-05-02 | 东南大学 | Automatic detection method of traffic incident on highway |
| CN102592453B (en) * | 2012-02-27 | 2014-02-26 | 东南大学 | Real-time traffic condition judging method based on time window |
| CN103810849B (en) * | 2012-11-15 | 2015-10-28 | 北京掌城科技有限公司 | A kind of traffic flow variation tendency extracting method based on floating car data |
| US9368027B2 (en) * | 2013-11-01 | 2016-06-14 | Here Global B.V. | Traffic data simulator |
-
2016
- 2016-11-22 CN CN201611035883.5A patent/CN108091131B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108091131A (en) | 2018-05-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108091131B (en) | Traffic incident identification method and device | |
| Arun et al. | A systematic mapping review of surrogate safety assessment using traffic conflict techniques | |
| CN108091132B (en) | Traffic flow prediction method and device | |
| US10957201B2 (en) | System and method for relative positioning based safe autonomous driving | |
| CN108492555B (en) | Urban road network traffic state evaluation method and device | |
| Pande et al. | Comprehensive analysis of the relationship between real-time traffic surveillance data and rear-end crashes on freeways | |
| AU2005264151B2 (en) | System and method for monitoring driving | |
| CN103562978B (en) | Vehicle data analysis method and vehicle data analysis system | |
| EP3251023A1 (en) | Vehicle driver assist system | |
| CN113870564B (en) | Traffic jam classification method and system for closed road section, electronic device and storage medium | |
| CN112835790A (en) | Test method, device, equipment and medium for automatic driving software | |
| Ansariyar et al. | Statistical analysis of vehicle-vehicle conflicts with a LIDAR sensor in a signalized intersection. | |
| Zhou et al. | A vision-based abnormal trajectory detection framework for online traffic incident alert on freeways | |
| Deng et al. | Classifying human behaviors: Improving training of conventional algorithms | |
| GB2565345A (en) | A method for use in a vehicle | |
| CN108388947B (en) | Improved machine learning method suitable for vehicle-mounted short-distance communication network safety early warning | |
| Das et al. | Multivariate analysis on dynamic car-following data of non-lane-based traffic environments | |
| CN112447042B (en) | Traffic incident detection system and method | |
| CN114822044B (en) | Driving safety early warning method and device based on tunnel | |
| Banerjee et al. | A modern intersection data analytics system for pedestrian and vehicular safety | |
| Guyonvarch et al. | Evaluation of safety critical event triggers in the UDrive data | |
| Detzer et al. | Analysis of traffic safety for cyclists: The automatic detection of critical traffic situations for cyclists | |
| US10415981B2 (en) | Anomaly estimation apparatus and display apparatus | |
| Oh et al. | Hazardous driving event detection and analysis system in vehicular networks: Methodology and field implementation | |
| JP2019207620A (en) | Vehicle proposal system, driving feature generation device, and vehicle proposal device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |