KR101123967B1 - Traffic congestion prediction system, prediction method and recording medium thereof - Google Patents

Abstract

PURPOSE: A traffic congestion prediction system, a prediction method, and a recording medium thereof are provided to easily apply traffic congestion variables by modeling general characteristics of a road. CONSTITUTION: A vehicle detector(10) detects a traffic situation of an intersection of a road. A vehicle number calculation part(20) calculates the number of outflow vehicles and the number of inflow vehicles per each lane based on the traffic situation. An intersection number calculation part(30) calculates the number of intersections per each unit length of the road based on the length of the load. A road property value providing part(40) provides a road property value corresponding to a downtown road, an outer road, a main road, or a highway. A traffic congestion information calculation part calculates traffic congestion information of the road through a network modeling process.

Description

TRAFFIC CONGESTION PREDICTION SYSTEM, PREDICTION METHOD AND RECORDING MEDIUM THEREOF

The present invention relates to a traffic jam prediction system. More specifically, the present invention relates to a traffic congestion prediction system, a prediction method, and a recording medium capable of estimating traffic congestion by grasping the number of vehicles flowing in and out of an actual road.

ITS (Intelligent Transportaion System) is a system that applies advanced electronic, information, and communication technologies to existing traffic components such as roads and vehicles to efficiently operate traffic facilities and provide useful information to road pedestrians. This can help to reduce traffic congestion by increasing the efficiency and capacity of the traffic system, improve driver mobility, convenience and safety by improving traveler services, and improve the safety of the traffic system. ITS is a high-level concept that includes the Advanced Traffic Management System (ATMS) and the Advanced Traveler Information System (ATIS).

ATMS installs a system that can detect traffic information such as vehicle characteristics and speed on the road, and analyzes traffic conditions in real time. Based on this, it manages road traffic and implements an optimal signal system. In addition, ATIS is a system that provides the driver with various traffic information such as traffic conditions, road conditions, the shortest route from the starting point to the destination, the time required, and the parking lot condition, to support the safe and smooth optimal traffic.

In constructing these two systems, there is a need for a study on utilizing statistical information of a vehicle detection system (VDS) that detects vehicles on the road.

On the other hand, the conventional congestion prediction model is known to predict the driving time for any route specified by the user, to predict the driving time during long-distance driving through a similar driving time pattern or a prediction method for a predetermined form of road section. . However, such a conventional congestion prediction model has a disadvantage in that the prediction is inaccurate, or the dynamic change such as population change or inflow and outflow of a vehicle, etc. cannot cope with traffic conditions.

Accordingly, there is a need for a study of a traffic congestion prediction system that can solve the disadvantages of the aforementioned conventional congestion prediction model and adapt to change.

The present invention has been made in view of the above necessity, and an object of the present invention is to provide a prediction model reflecting the characteristics of a road route, and to predict a traffic congestion system, a prediction method, and a new road pattern. The recording medium is provided.

An object of the present invention as described above is a vehicle detection unit for detecting a vehicle flowing in and out at the intersection of the road; A vehicle number calculator configured to calculate the number of inflow vehicles per lane and the number of outflow vehicles per lane based on the detected vehicles; Intersection number calculation unit for calculating the number of intersections per unit length of the road based on the road length of the road; A road characteristic value providing unit for providing a road characteristic value of a road corresponding to any one selected from a city road, an outer road, a main road, and a highway which are characteristics of a road; And a congestion information calculating unit for calculating congestion information of the road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the road characteristics. Can be.

The vehicle detection unit may include an imaging camera that captures the inflow and outflow of the vehicle.

The vehicle detection unit preferably detects the vehicle based on the image of the image of the vehicle.

The vehicle detector may include a loop coil embedded in the lower part of the road.

The congestion information calculating unit may obtain vehicle number information corresponding to the number of inflow vehicles per lane and the number of outflow vehicles per lane in real time.

The road characteristic value providing unit may further provide a road curvature value and a road slope value of the road.

The congestion information is preferably at least one of the degree of congestion of the road, the frequency of congestion, the speed of the section and the time required for the section.

Neural network modeling is preferably input by the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the characteristics of the road as input variables, the congestion information as target variables, and modeled by error backward propagation (BP). Do.

The target variable is preferably any one of at least three levels of congestion information.

The apparatus may further include a weather characteristic value providing unit that provides weather characteristic value information corresponding to at least one of snow, rain, and fog based on weather condition information on a road.

On the other hand, an object of the present invention as another category, the vehicle detecting unit detecting a vehicle flowing in and out at the intersection of the road (S10); Calculating the number of inflow vehicles per lane and the number of outflow vehicles per lane based on the number of vehicles detected by the vehicle count calculator (S20); And calculating, by the congestion information calculating unit, congestion information of the road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length of the road, and the road characteristic values (S30). It can be achieved by providing a traffic congestion prediction method.

In the congestion information calculating step S30 of the congestion information calculating unit, the number of intersections per unit length is preferably information calculated by the intersection number calculating unit based on the road length of the road.

In the congestion information calculating step S30 of the congestion information calculating unit, the road characteristic value is information provided by the characteristic value providing unit corresponding to any one selected from the city road, the outer road, the main road, and the highway which are characteristics of the road.

In the congestion information calculating step (S30) of the congestion information calculating unit, the congestion information is preferably at least one of the degree of congestion, the congestion frequency, the section speed and the section time required.

In the congestion information calculating step (S30) of the congestion information calculating unit, the neural network modeling is an input variable, the number of inflow vehicles per lane, the number of outgoing vehicles per lane, the number of intersections per unit length, and the road characteristic value, and the congestion information as a target variable. Modeled by reverse propagation is preferred.

In addition, an object of the present invention can be achieved by providing a computer readable recording medium having recorded thereon a program capable of executing a traffic jam prediction method.

According to one embodiment of the present invention as described above, by modeling the general characteristics of the road there is an effect that can easily apply a variable that affects the congestion.

In addition, it is possible to accurately predict the congestion by reflecting the actual characteristic data of the road.

In addition, it can be applied to a new situation, such as environmental changes in the road situation, there is an effect that can predict the level of congestion even on roads without congestion information or construction.

1 is a block diagram showing a configuration according to an embodiment of the present inventors traffic congestion prediction system,
2 is a view briefly showing a target road to explain an embodiment of the present inventors traffic congestion prediction system;
FIG. 3 is a table showing an example of information related to input and output variables used by the congestion information calculating unit for pattern learning in one embodiment of the present invention according to the present invention. FIG.
4 is a table illustrating an example of input and output variables used for actual pattern learning based on the information indicated by the table of FIG. 3 in one embodiment of the present traffic congestion prediction system;
5 is a table showing an example of weights calculated between nodes and input variables of a hidden layer as a result of repetitive learning through a neural network model of a traffic congestion prediction system according to an embodiment of the present invention;
FIG. 6 is a table showing an example of weights calculated between nodes and output variables of a hidden layer as a result of repetitive learning through a neural network model of a traffic congestion prediction system according to an embodiment of the present invention;
7 is a flowchart sequentially showing an embodiment of the present inventors traffic congestion prediction method.

Traffic Congestion Prediction System

1 is a block diagram showing a configuration according to an embodiment of the present inventors traffic congestion prediction system. As shown in FIG. 1, one embodiment of the present inventors traffic congestion prediction system includes a vehicle detecting unit 10, a vehicle number calculating unit 20, an intersection number calculating unit 30, a road characteristic value providing unit 40, and a congestion. An information calculating unit 50 is included.

The present embodiment detects the number of vehicles of the inflow and outflow vehicles of the predicted road through the vehicle detection unit 10, and the number of intersections per unit length of the road together with the number of inflow vehicles per lane and the number of outflow vehicles per lane calculated based on this. And it is possible to accurately predict the degree of congestion through pattern learning by modeling the neural network with the road characteristic values as four input variables and the degree of congestion as output variables.

The vehicle detecting unit 10 serves to detect a vehicle flowing in and out at an intersection of the road. The vehicle detecting unit 10 may include an imaging camera 110 for capturing the incoming and outgoing vehicles, and detects the vehicle through image processing based on the image of the vehicle captured by the imaging camera 110. The vehicle detecting unit 10 may detect a traffic vehicle through a loop coil embedded in the lower part of the road in addition to the image.

The vehicle number calculator 20 calculates the number of inflow vehicles per lane and the number of outflow vehicles per lane based on the detected vehicles. The vehicle number calculating unit 20 preferably calculates the number of vehicles for the same specific lane, but may be limited to a road in one direction and may be divided by the number of lanes of the corresponding road and may use an average value.

The intersection number calculator 30 calculates the number of intersections per unit length of the road based on the road length of the road. The number of intersections per unit length of the road can be calculated based on the length of the road and the number of intersections, and is used as an input variable for identifying the congestion degree by the intersection.

The road characteristic value providing unit 40 serves to provide the road characteristic value of the road in response to any one selected from the city road, the outer road, the main road, and the highway which are characteristics of the road. As roads can generally have different congestion characteristics, they are generally divided into city roads, outer roads, main roads, and highways, and the road characteristic values are numerically calculated for them. can do. In this embodiment, road characteristics are given as 0.3 for general roads and outer roads, 0.6 for main roads, and 0.9 for highways. In addition, the road characteristic value providing unit 40 may further provide a road curvature value and a road slope value in addition to a more accurate prediction.

The congestion information calculating unit 50 calculates congestion information of a road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and road characteristic values. Here, the neural network modeling is based on the error propagation method (Backpropagation, BP) using the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the road characteristic as four input variables and the congestion information as a target variable. Model so that learning can take place. Vehicle number information corresponding to the number of inflow vehicles per lane and the number of outflow vehicles per lane obtained by the congestion information calculating unit 50 may be obtained in real time, or may be obtained based on previously accumulated data.

In addition, the congestion information may be used at least one of the degree of congestion of the road, the frequency of congestion, the section speed and the time required for the section. Here, the degree of congestion, the section speed and the section duration time may be a value corresponding to the average traveling speed of the vehicle and the average driving time of the vehicle, and in the case of the congestion frequency, it may be defined as the number of times below a certain section speed.

In addition to the above-described configuration of the present embodiment, the apparatus may further include a weather characteristic value providing unit that provides weather characteristic value information corresponding to at least one of snow, rain, and fog, based on weather state information on the road, and in this case, congestion information considering the weather Can be an output variable.

2 is a diagram briefly showing a target road to explain an embodiment of the present inventors traffic congestion prediction system. Here, an arrow (→ or ↔) indicated on each road indicates a possible traveling direction of the vehicle. As shown in FIG. 2, the road generally has a constant road length L, and is generally provided with intersections C1 and C2 therebetween. Therefore, in the present exemplary embodiment, the vehicle detecting unit 10 detects the vehicle based on the captured image by placing the imaging camera 110 at the inlet 210 and the outlet 220, respectively, and based on the number of the inlet vehicle and the outlet vehicle. You can count the numbers.

In addition, since the prediction target road shown in FIG. 2 includes three lanes in one direction, the vehicle number calculating unit 20 divides the total number of inflow vehicles and the total number of outflow vehicles into three lanes for a predetermined time, and inflow vehicles per lane. The number and number of spilled vehicles per lane can be calculated. In addition, the intersection number calculating unit 30 calculates the number of intersections per unit length by dividing the number of intersections for the corresponding road by the road length L of the corresponding road. In addition, it is obvious that traffic congestion can be predicted using the same system for lanes in one direction and the opposite direction.

Neural network using test data modelling  Result>

In order to verify the accuracy of the above-mentioned traffic congestion prediction system, some data are based on data from eight cities of Santa Clara, SanFrancisco, Napa, Solano, San Benito, San Bruno, Sonoma, and Yolo, cities near Bay Area, California, USA. Was selected to train the neural network model and some other data were selected to verify the learning results. Hereinafter, the neural network modeling result will be described based on the data for the city.

3 is a diagram showing an example of information related to input variables and output variables used by the congestion information calculating unit for pattern learning in one embodiment of the present inventors traffic congestion prediction system, and FIG. 4 is a congestion prediction system according to the present invention. FIG. 3 is a diagram illustrating one example of an input variable and an output variable used for actual pattern learning based on the information represented by the table of FIG. 3.

Referring to FIG. 3 and FIG. 4, the data given as a table in FIG. 3 are processed and four input variables (inflows per lane, outflows per lane, number of intersections per length, characteristics) as shown in the table of FIG. Four output variables (congestion diagram) are provided to the congestion information calculating unit 50, and the neural network used is a network having one hidden layer (hidden layer) so that the hidden layer has ten nodes. The maximum number of iterations was 50,000 times, and the input pattern used four input variables (inflow per lane, outflow per lane, number of intersections per length, characteristics) as described above. In the case of congestion, the number of congestion was divided into five stages, and based on the annual average congestion data, 0 (none), 0.25 (low) for 1-49, 0.5 (medium) for 50-150, and 151-200 In the case of, we set 0.75 (many), and in case of 201 or more, 1 (very much).

FIG. 5 is a table illustrating an example of weights calculated between nodes and input variables of a hidden layer as a result of repetitive learning through a neural network model of a traffic congestion prediction system according to the present invention. FIG. In one embodiment of the inventors traffic congestion prediction system is a table showing an example of the weights calculated between the node and the output variable of the hidden layer as a result of repeated learning through the neural network model.

The combined weight between each node of the hidden layer (Node # 1 ~ Node # 10) and the input variable (Input Layer) is output as shown in the table shown in Figure 5, each node of the hidden layer (Node # 1 ~ Node # 10) ) And the combined weight between the output variables (Output Layer) is output as shown in the table shown in FIG.

The results given in the tables of FIG. 5 and FIG. 6 show an RMS error of 0.0261 and an actual learning rate of about 94.3%. As a result of verifying the learning results through some other data in the city, the RMS error is 0.0973 and the predicted success rate is about. 87.11%. The predicted success rate by road type was 100% for general roads and outer roads, 80% for main roads, and 86.7% for highways.

Traffic congestion prediction method

7 is a flowchart sequentially showing an embodiment of the present inventors traffic congestion prediction method. Referring to FIG. 7, first, the vehicle detecting unit 10 detects a vehicle flowing in and out at an intersection of a road (S10).

Next, the vehicle number calculating unit 20 calculates the number of inflow vehicles per lane and the number of outflow vehicles per lane based on the detected vehicles (S20).

Finally, the congestion information calculating unit 50 calculates congestion information of the road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length of the road, and the road characteristic values (S30). An embodiment of the traffic congestion prediction method may be performed.

In particular, in the congestion information calculating step S30 of the congestion information calculating unit 50, the number of intersections per unit length is information calculated by the intersection number calculating unit 30 based on the road length of the road, and the road characteristic value provides a characteristic value. Information provided corresponding to any one selected from the city road, the outer road, the main road, and the highway, which are characteristics of the additional road.

In addition, in the congestion information calculating step S30 of the congestion information calculating unit 50, the congestion information may predict traffic congestion using at least one of congestion degree, congestion frequency, section speed, and section duration time.

In addition, neural network modeling is modeled so that the number of inflow vehicles per lane, the number of outgoing vehicles per lane, the number of intersections per unit length, and the road characteristics are input variables, the congestion information is a target variable, and the pattern can be learned by error backward propagation. Can be used.

On the other hand, learning by neural network modeling of the above-described traffic congestion prediction method may be implemented as a separate program capable of executing this, and the traffic congestion prediction method may also be implemented by providing a computer-readable recording medium on which such a program is recorded.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, It will be understood that the invention may be practiced. Therefore, the embodiments described above are to be understood as illustrative and not restrictive in all aspects. In addition, the scope of the present invention is indicated by the appended claims rather than the detailed description above. Also, it is to be construed that all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts are included in the scope of the present invention.

C1, C2: intersection
L: road length
10: vehicle detection unit
110: imaging camera
20: vehicle number calculation unit
210: inlet
220: outlet
30: intersection number calculation unit
40: road characteristic value providing unit
50: identity information calculation unit

Claims (16)

A vehicle detecting unit detecting a vehicle flowing in and out at an intersection of the roads;
A vehicle number calculating unit configured to calculate the number of inflow vehicles per lane and the number of outflow vehicles per lane based on the detected vehicles;
An intersection number calculator configured to calculate an intersection number per unit length of the road based on a road length of the road;
A road characteristic value providing unit providing a road characteristic value of the road corresponding to any one selected from among a city road, an outer road, a main road, and a highway which are characteristics of the road; And
Traffic congestion prediction system comprising; a congestion information calculating unit for calculating congestion information of the road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the road characteristic value .
The method of claim 1,
The vehicle detecting unit includes a traffic jam camera for capturing the inflow and outflow of the vehicle.
The method of claim 2,
The vehicle detection unit detects the traffic jam based on the image of the captured vehicle.
The method of claim 1,
The vehicle detecting unit includes a loop coil embedded in the lower part of the road.
The method of claim 1,
The congestion information calculating unit, the traffic congestion prediction system, characterized in that to obtain in real time vehicle number information corresponding to the number of inflow vehicles per lane and the number of outflow vehicles per lane.
The method of claim 1,
The road characteristic value providing unit further provides a road curvature value and a road slope value of the road.
The method of claim 1,
The congestion information is traffic congestion prediction system, characterized in that at least one of the road congestion degree, the congestion frequency, section speed and section time required.
The method of claim 1,
In the neural network modeling, the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the road characteristic are input variables, and the congestion information is a target variable.
Traffic congestion prediction system, characterized by error backward propagation.
The method of claim 8,
The target variable is a traffic congestion prediction system, characterized in that any one of at least three levels of congestion information.
The method of claim 1,
And a weather characteristic value providing unit for providing weather characteristic value information corresponding to at least one of snow, rain, and fog based on weather condition information on the road.
Detecting the vehicle flowing in and out at the intersection of the vehicle detecting unit at step S10;
A vehicle number calculating unit calculating a number of inflow vehicles per lane and a number of outflow vehicles per lane based on the detected vehicles (S20); And
Calculating, by the congestion information calculating unit, the congestion information of the road through neural network modeling based on the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length of the road, and the road characteristic value of the road (S30) Traffic congestion prediction method comprising;
12. The method of claim 11,
In the congestion information calculating step (S30) of the congestion information calculating unit,
The number of intersections per unit length is traffic congestion prediction method, characterized in that the information calculated by the intersection number calculation unit based on the road length of the road.
12. The method of claim 11,
In the congestion information calculating step (S30) of the congestion information calculating unit,
The road characteristic value traffic congestion prediction method characterized in that the characteristic value providing unit provides information corresponding to any one selected from the road, the outer road, the main road and the expressway characteristics of the road.
12. The method of claim 11,
In the congestion information calculating step (S30) of the congestion information calculating unit,
The congestion information is traffic congestion prediction method, characterized in that at least one of the degree of congestion, the frequency of congestion, section speed and the time required section.
12. The method of claim 11,
In the congestion information calculating step (S30) of the congestion information calculating unit,
In the neural network modeling, the number of inflow vehicles per lane, the number of outflow vehicles per lane, the number of intersections per unit length, and the road characteristic are input variables, and the congestion information is a target variable.
Traffic congestion prediction method, characterized by the error reverse propagation method.
A computer-readable recording medium having recorded thereon a program capable of executing the traffic congestion prediction method according to any one of claims 11 to 15.

Images