KR20230080741A - System and Method for predicting real-time lane queue using an image detector, and a recording medium recording a computer readable program for executing the method - Google Patents

Abstract

A real-time lane queuing prediction system and method using an image detector, and a recording medium recording a computer readable program for executing the method are disclosed. A real-time lane queue prediction system using an image detector includes an image acquisition unit, a moving flow traffic counting unit, a queue prediction unit, and a queue length output unit. The image acquisition unit acquires an image of moving objects on the upstream intersection, the moving flow traffic counting unit counts the moving flow traffic volume, which is the number of moving objects for each moving flow within the upstream intersection, from the acquired image, and the queue predicting unit calculates the moving flow traffic volume from information on the moving flow traffic volume on the downstream side. The length of the queue for each lane of the mobile object at the side intersection is predicted, and the queue length output unit outputs the predicted length of the queue for each lane.

Description

System and method for predicting real-time lane queue using an image detector, and a recording medium recording a computer-readable program for executing the method {System and Method for predicting real-time lane queue using an image detector, and a recording medium recording a computer readable program for executing the method }

The present invention relates to traffic control-related technology, and more particularly, to a system and method for predicting the length of a queue of cars waiting at a traffic light at an intersection.

The technology of estimating and predicting the length of the queue of vehicles waiting for traffic lights at an intersection is essential for intersection signal operation, and is a technology that is essential for driving safety and convenience for mechanical drivers such as autonomous driving in the future.

Conventionally, in order to predict the queue by lane, a method of predicting the queue by lane was mainly used by using the aggregated traffic volume recognized for a certain period of time with a detection technology such as an inductive loop detector (ILD) installed for each lane. .

However, such a conventional general vehicle detection technology (ILD) is difficult to accurately detect traffic volume, high installation cost, difficult to maintain, and requires installation of a plurality of ILDs for each lane for accurate queuing prediction.

In addition, since the aggregated traffic volume is used, not only is it impossible to predict the queue by lane in real time, but also a method of displaying prediction information of queue by lane in real time has not been proposed.

KR 100459476 B1

The present invention has been made to solve the above-mentioned conventional problems, and an object of the present invention is to provide a system and method capable of providing more accurate vehicle queue prediction with low installation and maintenance costs.

In addition, an object of the present invention is to provide a system and method capable of performing real-time prediction of queues by lanes and effectively providing information on prediction of queues by lanes in real time.

To achieve the above object, a real-time lane queue prediction system using an image detector according to the present invention includes an image acquisition unit, a moving flow traffic counting unit, a queue estimation unit, and a queue length output unit. The image acquisition unit acquires an image of moving objects on the upstream intersection, the moving flow traffic counting unit counts the moving flow traffic volume, which is the number of moving objects for each moving flow within the upstream intersection, from the acquired image, and the queue predicting unit calculates the moving flow traffic volume from information on the moving flow traffic volume on the downstream side. The length of the queue for each lane of the mobile object at the side intersection is predicted, and the queue length output unit outputs the predicted length of the queue for each lane.

According to this configuration, since separate facilities such as an ILD (Inductive Loop Detector) are not installed for each lane, more accurate vehicle queue prediction can be provided with low installation and maintenance costs.

In addition, real-time queuing prediction for each lane is performed, and real-time queuing prediction information for each lane can be effectively provided.

In this case, the moving flow traffic counting unit may include a moving object information extracting unit that extracts moving object information from an image, and a moving flow traffic amount information calculating unit that calculates the moving flow traffic amount information by counting the moving flow traffic amount from the moving object information.

In addition, the moving flow traffic volume information may include at least one of unique information about an upstream intersection, reference time, counting time interval, traffic volume information, and traffic signal information.

In addition, the information on the moving object may include one or more information of time, object type, unique number, whether or not it is new, location, speed, direction, and reliability of the moving object.

Also, the moving flow traffic information calculation unit may determine whether the moving object enters or exits by using the location and time data of the moving object in the moving object information.

In addition, when it is determined that a moving object has entered, the moving flow traffic information calculation unit may assign a unique access road number and add it to the intersection object list.

In addition, when it is determined that the moving object has entered, the moving flow traffic calculation unit may update the moving flow traffic information after assigning an exit route unique number and a moving flow unique number, deleting them from the intersection object list.

In addition, the moving flow traffic information calculation unit may perform entry/exit of a moving object and classification of inside/outside of the intersection according to whether the moving object exists in the intersection object list.

In addition, the moving flow traffic information calculation unit may further perform a spatial search for the intersection to determine whether the moving object is present in the upstream intersection.

In addition, the moving flow traffic information calculation unit may determine that the moving object is entering the upstream intersection when it is determined that the moving object exists in the intersection object list and is located in the upstream intersection.

Also, the moving flow traffic information calculation unit may determine that the moving object is located within the upstream intersection when it is determined that the moving object exists in the intersection object list and is located outside the upstream intersection.

In addition, the moving flow traffic information calculation unit may determine that the moving object enters the upstream intersection when it is determined that the moving object does not exist in the intersection object list and is located within the intersection.

In addition, the moving flow traffic information calculation unit may determine that the moving object is located outside the intersection when it is determined that the moving object does not exist in the intersection object list and is located outside the intersection.

In addition, the queue prediction unit calculates the upstream traffic volume of the road section between the upstream intersection and the downstream intersection from the moving flow traffic information, and the upstream traffic management unit calculates the arrival time and arrival time of the upstream traffic at the downstream intersection. A downstream arrival traffic volume prediction unit that predicts the traffic volume, a lane traffic volume ratio prediction unit that predicts the distribution state of the traffic volume arriving at the downstream intersection by lane, and the length of the queue by lane is calculated using the distribution state by lane and traffic signal information. A lane length prediction unit may be included.

In addition, the queue length output unit may output each time length of the queue corresponding to the lane on the road shape dividing the lane.

In addition, the queue length output unit may sequentially output the time-by-time length of the queue corresponding to the lane on the road shape dividing the lane.

In addition, the queue length output unit may output the length of the queue by time, respectively, with respect to the current state and the state when the user's vehicle arrives at the queue.

In addition, the real-time lane queue prediction method using an image detector according to the present invention is a queue prediction method performed by a real-time lane queue prediction system using an image detector, and an image acquisition step of acquiring an image of a moving object on an upstream intersection, From the image, a moving flow traffic volume counting step of counting the moving flow traffic volume, which is the number of moving objects by moving flow in the upstream intersection, a queue prediction step of predicting the length of the queue for each lane of moving objects at the downstream intersection from information on the moving flow traffic volume, and a queue length output step of outputting the length of the queue for each lane.

In addition, a recording medium recording a computer readable program for executing the method is disclosed together.

According to the present invention, since separate facilities such as an ILD (Inductive Loop Detector) are not installed for each lane, more accurate vehicle queue prediction can be provided with low installation and maintenance costs.

In addition, real-time queuing prediction for each lane is performed, and real-time queuing prediction information for each lane can be effectively provided.

1 is a schematic block diagram of a real-time lane queue prediction system using an image detector according to an embodiment of the present invention.
FIG. 2 is a conceptual diagram for explaining a prediction process performed in the lane queue prediction system of FIG. 1;
FIG. 3 is a view showing a state of use of an actual implementation of the moving flow traffic counting unit of FIG. 1;
4 is a diagram showing the configuration of an actual implementation example of a queue prediction unit;
5 to 7 are diagrams showing examples of output screens output from the queue length output unit of FIG. 1;

Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings.

1 is a schematic block diagram of a real-time lane queuing prediction system using an image detector according to an embodiment of the present invention. In FIG. 1, the system for predicting lane queues in real time using an image detector includes an image acquisition unit 110, a moving flow traffic counting unit 120, a queue prediction unit 130, and a queue length output unit 140. .

The moving flow traffic counting unit 120 includes a moving object information extraction unit 122 and a moving flow traffic amount information calculating unit 124, and the queue predicting unit 130 includes the upstream traffic management unit 132 and the downstream arriving traffic It includes a predictor 134, a road traffic ratio predictor 136, and a queue length predictor 138.

The image acquisition unit 110 acquires an image of a moving object on an upstream intersection. FIG. 2 is a conceptual diagram for explaining a prediction process performed in the lane queuing prediction system of FIG. 1 . In FIG. 2 , it can be confirmed that the image acquisition unit 110 is implemented as a camera installed at an intersection.

In addition, in FIG. 2, the present invention uses the automatically recognized moving object (vehicle) information at the intersection to (1) count the traffic flow at the intersection, and (2) predict the queue for each lane using the counted traffic. After that, (3) it can be confirmed that this is to provide a technology for expressing it to the vehicle driver.

The moving flow traffic counting unit 120 counts the moving flow traffic amount, which is the number of moving objects for each moving flow in the upstream intersection, from the obtained image. FIG. 3 is a diagram showing a state of use of an actual implementation of the moving flow traffic counting unit of FIG. 1 .

To this end, the moving object information calculation unit 122 extracts moving object information from the image. At this time, the moving object information is at least one of the time of the moving object, object type, unique number, whether it is new, location, speed, direction, or reliability. information may be included.

In addition, the moving flow traffic amount information calculation unit 124 calculates the moving flow traffic amount information by counting the moving flow traffic amount from the moving object information. In this case, the moving flow traffic volume information may include one or more of information unique to an upstream intersection, a reference time, an aggregated time interval, traffic volume information, and traffic signal information. In addition, the moving flow traffic information calculation unit 124 may determine whether the moving object enters or exits by using the location and time data of the moving object in the moving object information.

To this end, the moving flow traffic information calculation unit 124 may assign a unique entry number to the intersection object list when it is determined that the moving object has entered, and if it is determined that the moving object has entered, the unique entry number and movement After assigning a flow-specific number and deleting it from the intersection object list, moving flow traffic information can be updated, and moving objects can be entered, exited, and classified inside and outside the intersection depending on whether or not they exist in the intersection object list.

In addition, the moving flow traffic information calculation unit 124 may further perform a spatial search to determine whether the moving object exists in the upstream intersection, and when it is determined that the moving object exists in the intersection object list and is located in the upstream intersection. , It can be determined that the moving object is entering the upstream intersection, and if it is determined that the moving object exists in the intersection object list and is located outside the upstream intersection, it can be determined that the moving object is located within the upstream intersection.

In addition, when the moving flow traffic information calculation unit 124 determines that the moving object does not exist in the intersection object list and is located within the intersection, it may be determined that the moving object enters the upstream intersection, and the moving object is included in the intersection object list. When it is determined that the moving object does not exist and is located outside the intersection, it may be determined that the moving object is located outside the intersection.

The queue predictor 130 predicts the length of the queue for each lane of a moving object at a downstream intersection from information on moving traffic. To this end, the upstream traffic management unit 132 calculates the upstream traffic volume of the road section between the upstream intersection and the downstream intersection from the moving flow traffic information, and the downstream arrival traffic volume prediction unit 134 calculates the upstream traffic volume on the downstream side The arrival time at the intersection and the traffic volume according to the arrival time are predicted, the lane traffic ratio prediction unit 136 predicts the distribution state of the traffic volume arriving at the downstream intersection by lane, and the queue length prediction unit 138 predicts the traffic volume by lane Using the distribution status and traffic signal information, the length of the queue for each lane is calculated. 4 is a diagram showing the configuration of an actual implementation of a queue prediction unit.

The queue length output unit 140 outputs the predicted queue length for each lane. At this time, the queue length output unit 140 may output the time-specific lengths of the queues corresponding to the lanes on the road shapes dividing the lanes.

In addition, the queue length output unit 140 may sequentially output the time-by-time length of the queue corresponding to the lane on the road shape dividing the lanes, and display the time-by-hour length of the queue to the current state and the user's vehicle. You can print each one about the status when it arrives at this queue. 5 to 7 are diagrams illustrating examples of output screens output from the queue length output unit of FIG. 1 .

To describe the real-time lane queue prediction method using an image detector according to an embodiment of the present invention in more detail, the present invention is largely composed of three parts. The first is a method of counting traffic flow at an intersection using vehicle information automatically recognized from a video, the second is a method of predicting the queue length in real time using the counted traffic, and the last is a method of predicting the This is a method of expressing the length of the queue to the vehicle driver.

Intersection movement flow traffic count method

Movement Traffic (MT) at an intersection is defined as the number of vehicles for each movement flow within an intersection observed during a given time period. , a moving flow with the same entry and exit roads at an intersection is unique.

Traffic volume in a general road section is the number of vehicles observed for a given time at a point, and since the vehicle group subject to observation consists of one movement stream, there is no need to distinguish the movement stream.

However, in the case of an intersection, since numerous moving flows exist simultaneously in the same space, the moving flows must be distinguished, and for this, a more complex counting method than the existing method is required.

The traffic volume counting method of the intersection movement proposed in the present invention is largely composed of three subsystems (input, coefficient, and output), the configuration of which is shown in FIG. 3. In FIG. 3, the input system that provides input data for counting is based on video, and automatically recognizes moving object (CDO, Camera Detected Object) information from the video is encoded into a CD message and transmitted to the counting system. . The CDO message consists of time stamp, object type, unique number, whether it is new, location (latitude/longitude), speed, direction, and reliability.

The counting system decodes the received CDO message and uses it to perform an intersection traffic flow counting algorithm. MTR, MT Record) is stored in the database, and at the same time, it is encoded as an MT message and transmitted to the output system. The MT message consists of an intersection identification number, reference time, aggregate time interval, traffic volume information (array), and traffic signal information (array).

The MT message is the input data of LQS (Lane-by-lane Queue prediction System), and LQS predicts the queue by lane. The intersection traffic flow counting algorithm uses the location and time stamp data of each moving object in the CDO message as input data to determine whether to enter/exit. It is added to the object list (Lt), and if it is determined that it is exiting, an exit route identification number and a movement flow identification number are given, deleted from Lt, and MTR is updated. If the moving object exists outside the intersection, the corresponding data is discarded.

Here, spatial search can be used when determining whether a moving object is present in an intersection, and three spatial search methods are available. The first method is a method of GeoFencing using a polygon that includes the entire spatial range of an intersection, the second method is a method of creating a separate polygon in the entry/exit section of each intersection, and the third method is a method near an intersection. This method uses the distance and angle based on a specific point (location of a landmark) of In addition to this, when a precision road map (or electronic road map) is used, a unique number of a map-matched link may be used.

Mobile objects are classified into entry/exit/inside intersection/outside intersection depending on whether or not they exist within Lt. When a moving object exists within Lt, if it is judged to be inside the intersection through space search, it is judged to be still inside the intersection, and if it is judged to be outside the intersection, it is determined that the moving object has entered the intersection.

When the moving object does not exist within Lt, if it is judged to be inside the intersection through space search, it is determined that the moving object has entered the intersection, and if it is judged to be outside the intersection, it is determined that it has not yet entered the intersection.

Method for predicting queue length by lane in real time

The method for estimating the length of the queue by lane in real time consists of five models that are sequentially connected as shown in FIG. 4 . The upstream traffic management model decodes the MT message and plays a role in calculating the upstream traffic volume of the road section. The downstream arrival traffic volume prediction model is a model that predicts when and how much upstream traffic will arrive downstream, and Robertson's Platoon Dispersion Model can be used.

The lane traffic ratio prediction model is a model that predicts how the traffic volume arriving downstream is distributed to each lane. Myopic-type model simply uses the traffic volume ratio by lane observed in the previous traffic signal cycle, and the traffic volume ratio by multiple observed lanes in the past. A Kalman Filter algorithm or a probability-based model that predicts a future ratio using ? can be used.

The lane queue length prediction model is a model that predicts the queue length for each lane by using the downstream traffic volume and traffic signal appearance by lane, and a traffic shockwave-based prediction model can be used. The PVD (Probe Vehicle Data) data-based correction model is a model that corrects the predicted queue length for each lane using the position, speed, and visual data of the probe vehicle data (PVD).

How to provide queue length information by lane in the future

The predicted future queue length information for each lane must be delivered to the vehicle driver in an appropriate way, and for this purpose, an effective visual display method is required. The expression method proposed in the present invention is divided into three types, the first is a static expression method, the second is a dynamic expression method, and the last is a static expression method based on the position of an ego-vehicle.

The static display method is a method of indicating the length of a queue by time zone using a change in color and a road shape that distinguishes lanes. As shown in FIG. 5, the same color and gradation effect may be used, or different colors may be used for each time period.

The dynamic display method uses the road shape and color change that distinguishes the lanes, and introduces the display delay time to intuitively express the increase and decrease in the queue by time zone. expressed in the form

The static display method based on the location of the host vehicle compares the speed of the host vehicle and the rate of increase or decrease in the queue and displays the current queue status and the queue status at the time the vehicle reaches the end of the queue using the screen. , when the vehicle (red + black) enters the end of the queue after 10 to 15 seconds, it can be expressed as shown in FIG. 7.

In summary, the present invention (1) counts the real-time intersection moving traffic volume using the intersection mobile object (vehicle) data recognized through the video detector, and (2) predicts the queue at the intersection in real time using the resulting value. (3) It is an invention related to a system and method for expressing it to a driver.

According to the present invention, it is expected to alleviate congestion at an intersection in the city center and reduce socio-economic costs accordingly by using it for an intersection signal optimization operation. In addition, user convenience can be improved by utilizing real-time queue length information when selecting a route and lane for an autonomous vehicle. In addition, it can be expected to secure technology to prevent safety accidents by enabling autonomous vehicles to change lanes in advance according to the length of the queue in real time.

Although the present invention has been described by some preferred embodiments, the scope of the present invention should not be limited thereto, but should also extend to modifications or improvements of the above embodiments supported by the claims.

110: image acquisition unit
120: moving flow traffic counting unit
122: mobile body information extraction unit
124: moving flow traffic information calculation unit
130: queue prediction unit
132: Upstream Traffic Management Department
134: downstream arrival traffic prediction unit
136: lane traffic ratio prediction unit
138: Queue length prediction unit
140: queue length output unit

Claims (19)

an image acquiring unit acquiring an image of a moving object on an upstream intersection;
a moving flow traffic counting unit counting the moving flow traffic amount, which is the number of moving objects for each moving flow within the upstream intersection, from the image;
a queue prediction unit that predicts the length of the queue for each lane of the moving object at the downstream intersection from the information on the traffic volume of the moving stream; and
Lane queue prediction system, characterized in that it comprises a queue length output unit for outputting the length of the queue for each lane.
The method according to claim 1, wherein the moving flow traffic counting unit,
a moving object information extraction unit extracting moving object information from the image; and
and a moving flow traffic amount information calculation unit for counting the moving flow traffic amount from the information of the moving object and calculating the moving flow traffic amount information.
The method of claim 2,
The lane queue prediction system, characterized in that the moving flow traffic information includes at least one of unique information about the upstream intersection, reference time, aggregated time interval, traffic volume information, and traffic signal information.
The method of claim 3,
The lane queue prediction system, characterized in that the information of the moving object includes one or more information of time, object type, unique number, whether it is new, location, speed, direction, or reliability of the moving object.
The method of claim 4,
The lane queue prediction system of claim 1 , wherein the moving flow traffic information calculation unit determines whether the moving object enters or exits by using the location and time data of the moving object in the moving object information.
The method of claim 5,
The lane queue prediction system, characterized in that, when it is determined that the moving object enters, the moving flow traffic information calculation unit assigns a unique access road number and adds it to an intersection object list.
The method of claim 6,
Wherein the moving flow traffic information calculation unit determines that the moving object has entered, after assigning an exit route identification number and a movement flow identification number, deleting the intersection object list, and then updating the moving flow traffic information. prediction system.
The method of claim 7,
The lane queue prediction system according to claim 1 , wherein the moving flow traffic information calculation unit performs entry/exit of the moving object and classification of entry/exit of the intersection according to whether the moving object exists in the intersection object list.
The method of claim 8,
The lane queue prediction system of claim 1 , wherein the moving flow traffic information calculation unit further performs a spatial search for the upstream intersection to determine whether the moving object exists in the upstream intersection.
The method of claim 9,
Wherein the moving flow traffic information calculation unit determines that the moving object is located in the upstream intersection when the moving object is present in the intersection object list and is determined to be located in the upstream intersection, lane queue prediction, characterized in that system.
The method of claim 9,
Wherein the moving flow traffic information calculation unit determines that the moving object has entered the upstream intersection when it is determined that the moving object exists in the intersection object list and is located outside the upstream intersection, lane queue prediction system, characterized in that .
The method of claim 11,
Wherein the moving flow traffic information calculator determines that the moving object has not entered the intersection when it is determined that the moving object does not exist in the intersection object list and is located within the intersection, Lane Queue Prediction System.
The method of claim 12,
Wherein the moving flow traffic information calculation unit determines that the moving object has not entered the intersection when it is determined that the moving object does not exist in the intersection object list and is located outside the intersection, Lane Queue Prediction System.
The method according to claim 13, wherein the queue prediction unit,
an upstream traffic management unit calculating an upstream traffic volume of a road section between the upstream intersection and the downstream intersection from the moving flow traffic information;
a downstream arrival traffic volume predicting unit for predicting the arrival time of the upstream traffic at the downstream intersection and the traffic volume according to the arrival time;
a lane traffic volume ratio prediction unit for predicting a distribution state of each lane of the traffic volume arriving at the downstream intersection; and
and a lane queue length estimation unit for calculating the length of the queue for each lane using the distribution state for each lane and the traffic signal information.
The method of claim 14,
The lane queue prediction system, characterized in that the queue length output unit outputs the length of each time zone of the queue corresponding to the lane on the road shape dividing the lane.
The method of claim 15
The lane queue prediction system, characterized in that the queue length output unit sequentially outputs the time-specific length of the queue for each time period.
The method of claim 16
The lane queue prediction system, characterized in that the queue length output unit outputs the length of the queue for each time period for a current state and a state when the user's vehicle arrives at the queue.
A queue prediction method performed by a real-time lane queue prediction system using an image detector,
An image acquisition step of acquiring an image of a moving object on an upstream intersection;
a moving flow traffic counting step of counting a moving flow traffic amount, which is the number of moving objects for each moving flow in the upstream intersection, from the image;
a queue prediction step of estimating the length of the queue for each lane of the moving object at the downstream intersection from the information on the traffic volume of the moving flow; and
A real-time lane queue prediction method using an image detector, characterized in that it comprises a queue length outputting step of outputting the length of the queue for each lane.
A recording medium recording a computer readable program for executing the method of claim 18.

Images