Method for identifying lane-changing behavior of vehicle on closed road based on limited cross-section observation information

Abstract

The application discloses a lane-closing vehicle lane-changing behavior recognition method based on limited section observation information, utilizes fixed-point arrangement of sensors along a closed road to record license plates, passing time, speed and lane information of each vehicle, and constructs a vehicle information vector. When lane information of a vehicle passing through two continuous sensors is inconsistent, the vehicle lane-changing is directly determined. Otherwise, vehicles on the same lane are sorted according to the time of passing through the sensors, vehicles with changed position relationship between the two sensors are found, and an inverse sequence pair is constructed. When the road satisfies the free flow condition and a time difference of passing residual time of two vehicles in the inverse sequence pair is less than a minimum lane-changing cost, it is determined that the vehicle with smaller passing residual time lane-changes. Otherwise, vehicle flow density, speed and headway of the current lane and adjacent lanes are comprehensively considered, overtake motivation of the two vehicles in the inverse sequence pair is calculated, and the vehicle with larger overtake motivation lane-changes.

Description

Technical Field

[0001] This invention belongs to the field of road traffic monitoring technology, specifically relating to a method for identifying vehicle lane-changing behavior on closed roads based on limited cross-sectional observation information. Background Technology

[0002] Lane changing refers to the lateral movement of a vehicle while driving, changing from its current lane to another lane. Lane changes are necessary for overtaking, exiting auxiliary lanes, avoiding obstacles, or preparing to turn. Since each lane change encroaches on the driving space of other vehicles, frequent lane changes increase the risk of collisions or reduce road efficiency. This is especially true in tunnel sections, where the human eye needs to adapt to the significant difference in light levels upon entering and exiting a tunnel. Changing lanes at tunnel entrances and exits can easily lead to a "light and dark hole" effect, making it difficult for drivers to judge road conditions ahead, thus increasing the probability of accidents.

[0003] Traditional lane change monitoring methods involve burying geomagnetic induction coils under the road surface. The changes in the magnetic field caused by vehicles passing over the coils are used to sense vehicle speed and headway, thereby inferring whether there are frequent lane changes. However, this method is difficult to identify the specific lane change trajectory of the vehicles.

[0004] With the development of high-definition cameras and computer vision technology, existing monitoring methods utilize roadside cameras to capture images of vehicles in motion, and then employ deep learning algorithms for real-time analysis. This approach can map the continuous movement trajectory of each vehicle, automatically determine whether a vehicle has changed lanes, and achieve a high accuracy rate. However, this method requires deploying a large number of cameras for video recording to achieve lane-changing behavior monitoring across the entire road segment. Not only is the initial hardware investment high, but the subsequent processing of large amounts of image data also consumes significant resources. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention proposes a method for recognizing lane-changing behavior of vehicles in closed lanes based on limited cross-sectional observation information. This method utilizes limited sensor resources on the road to identify vehicle information and identifies lane-changing behavior by constructing a simple numerical calculation model, thereby reducing hardware installation and maintenance costs as well as the resource consumption of deep learning models.

[0006] The method for recognizing vehicle lane-changing behavior in closed lanes based on limited cross-sectional observation information includes the following steps:

[0007] Step 1: Deploy sensors at fixed points along the closed road to record the license plate of each vehicle. and the time it takes to pass through the sensor. ,speed With the driving lane Construct vehicle information vectors The superscripts in and out represent the information recorded by the first and second sensors in two adjacent sensors along the traffic flow direction, respectively.

[0008] Step 2: Read vehicle information vector Lane information in the middle, if ≠ Determine the vehicle A lane change occurred.

[0009] Step 3: If the lane information of all vehicles recorded by two consecutive sensors remains unchanged, then sort the vehicles in the same lane according to the time they passed the sensor to obtain the entry sequence. and leaving the sequence .

[0010] Step 4: Traverse into the sequential sequence With leaving the sequence If it exists and vehicles , The change in the relative positions between the two adjacent sensors indicates that a vehicle overtaking maneuver occurred on the road segment between them, changing lanes and then returning to its original lane. This is recorded as (…). , ) is an inversion pair.

[0011] Step 5: Define the free flow velocity Minimum lane change time cost and time measurement error .

[0012] For vehicles in the inversion pair, calculate their travel time residual between two adjacent sensors. :

[0013]

[0014]

[0015] in, Indicates the actual travel time. This represents the travel time in free-flow state obtained based on vehicle dynamics and speed limit constraints.

[0016] When satisfied , , If the travel time residuals of the two vehicles in the reverse pair satisfy... If yes, proceed to step 6; otherwise, proceed to step 7.

[0017] Step 6: Determine the vehicle that overtook the other vehicle in the reverse pair based on the travel time residual between the two vehicles in the reverse pair.

[0018] ①When That is, when both cars accelerate, the judgment is made. Change lanes to overtake.

[0019] ②When ,Right now The vehicle passage time was significantly shorter than expected, therefore, it was determined that... Change lanes to overtake.

[0020] ③ When 0 Under the premise that the free flow assumption holds, Vehicle travel time less than ,determination Change lanes to overtake.

[0021] Step 7, when , , Cannot be true at the same time, or At that time, based on the motive for changing lanes, determine the vehicle that made the lane change:

[0022]

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030]

[0031] in, Indicates the feasibility of changing lanes at the cross section. The weight, Indicates lane change benefits The weight. , These represent the feasibility of lane changing at the observation positions of the two sensors, respectively. Indicates the safe headway. , This indicates the distance between the front of the vehicle and the adjacent lane when the vehicle passes the sensor. The closest distance to the rear of the vehicle. , This indicates the distance between the rear of the vehicle and the adjacent lane when the vehicle passes the sensor. The closest distance to the front of the vehicle. K and K represent adjacent lanes, respectively. Its traffic density index and speed advantage , This indicates the corresponding weight. , These represent the lanes the vehicle is currently traveling in. Adjacent lane The number of vehicles on it. , These represent the lanes the vehicle is currently traveling in. Adjacent lane Traffic flow speed. Indicates the speed of traffic flow in a lane. This represents the congestion density given in the Greenshields model. The value represents the vehicle density of the lane, L represents the road length between the two sensors, and n represents the number of vehicles in the lane.

[0032] when > At that time, determine the original following vehicle Actively change lanes and overtake, when > At that time, determine the original vehicle in front. After changing lanes, the vehicle failed to overtake and returned to its original lane.

[0033] As a preferred option, calculate the current vehicle's driving lane. Adjacent lane Dynamic traffic density , Indicators for measuring traffic density :

[0034]

[0035]

[0036]

[0037] in, This indicates the average headway of a vehicle in a lane. Indicates the length of the vehicle. This indicates the average speed of vehicles traveling within the lane.

[0038] The present invention has the following beneficial effects:

[0039] This method constructs a simple numerical model that can identify vehicles that change lanes by using limited observation resources obtained from sensors fixed at fixed points on closed roads. This not only reduces the cost of hardware installation and maintenance, but also reduces the resources consumed by deep learning models. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of a vehicle entering a tunnel in the embodiment;

[0041] Figure 2 This is a schematic diagram of a vehicle leaving the tunnel in the embodiment;

[0042] Figure 3 This is a schematic diagram of a vehicle explicitly changing lanes in the embodiment;

[0043] Figure 4 This is a schematic diagram illustrating how a vehicle returns to its original lane after successfully changing lanes and overtaking another vehicle, as shown in the example.

[0044] Figure 5 A diagram illustrating the safe time intervals required for lane-changing operations;

[0045] Figure 6 This is a schematic diagram illustrating how a vehicle returns to its original lane after failing to change lanes to overtake in the example. Detailed Implementation

[0046] The present invention will be further explained below with reference to the accompanying drawings;

[0047] Taking a closed multi-lane tunnel on a highway as an example, this paper illustrates a method for recognizing lane-changing behavior of vehicles in closed lanes based on limited cross-sectional observation information:

[0048] Step 1, as follows Figure 1 , Figure 2 As shown, a sensor is installed at both the entrance and exit of the tunnel.

[0049] Sensors are used to record the license plates of every vehicle entering and exiting the tunnel. and the time it takes to pass through the sensor. ,speed With the driving lane Construct vehicle information vectors The superscripts "in" and "out" indicate the location sensors at the tunnel entrance and exit, respectively.

[0050] As a preferred embodiment, when the length of the target identification section exceeds 1000 meters, in order to ensure identification accuracy, sensors can be installed every 500 meters along the road.

[0051] Step 2: Read vehicle information vector Lane information in the middle, if ≠ Determine the vehicle A lane change occurred. For example... Figure 3 As shown, vehicle A was traveling in lane 2 when entering the tunnel and in lane 1 when exiting the tunnel. The lane information recorded by the sensors when entering and exiting the tunnel is inconsistent, which can be determined as vehicle A changing lanes inside the tunnel.

[0052] Step 3: If the lane information of all vehicles remains unchanged when entering and exiting the tunnel, then sort the vehicles in the same lane according to the time they pass the sensor to obtain the entry sequence. and leaving the sequence .

[0053] Step 4: Traverse into the sequential sequence With leaving the sequence If it exists and If this happens, the FIFO (First-In, First-Out) rule is broken, indicating that the vehicle... , The relative positions of the two vehicles changed, therefore, an overtaking maneuver must have occurred inside the tunnel, in which the following vehicle changed lanes and then returned to its original lane. Record this as (…). , () represents an inversion pair. For example... Figure 4 As shown, vehicle A changes lanes from lane 2 to lane 1, passes vehicles C, D, and E, and then returns to lane 2. Therefore, vehicle A's lane information when entering and exiting the tunnel is consistent, but its positional relationship relative to vehicles C, D, and E has changed.

[0054] Step 5: Define the free flow velocity Minimum lane change time cost and time measurement error .

[0055] For vehicles in the inversion pair, calculate their travel time residuals within the tunnel. :

[0056]

[0057]

[0058] in, Indicates the actual travel time. This represents the travel time in a free-flow state, determined based on vehicle dynamics and speed limit constraints. Free-flow refers to a state where vehicles are not directly interfered with by other vehicles or traffic control facilities, and drivers freely choose their speed based on road geometry, vehicle performance, and personal preferences.

[0059] Travel time residual The physical meaning of lane change loss is the additional time loss in actual vehicle travel time compared to ideal free travel time, which can be considered as the loss caused by lane changes. Vehicle interaction loss and measurement error The combination of these three factors led to this.

[0060] Therefore, when , , When this condition is met, the road conditions inside the tunnel satisfy the free-flow assumption. In this case, if the number of lane changes is 0, then the travel time between different vehicles is entirely due to time measurement errors. If it exists This indicates that a vehicle's travel time is abnormal, which must have been due to a lane change. Proceed to step 6.

[0061] When the road conditions inside the tunnel cannot meet the free flow assumption, there may be high traffic density, causing the following vehicle to fail to change lanes and overtake or the preceding vehicle to slow down when changing lanes, proceeding to step 7.

[0062] Step 6: Under the free-flow assumption, vehicle interaction loss That is, travel time residual Mainly caused by vehicles changing lanes, then , This indicates the number of lane changes, and vehicles that have changed lanes can be uniquely identified based on the travel time residual. Based on the minimum lane change principle, inversion pairs are determined (…). , Vehicles that overtake in the following situations:

[0063] ①When That is, when both cars accelerate, the judgment is made. Change lanes to overtake.

[0064] ②When ,Right now The vehicle passage time was significantly shorter than expected, therefore, it was determined that... Change lanes to overtake.

[0065] ③ When 0 Under the premise that the free flow assumption holds, Vehicle travel time less than ,determination Change lanes to overtake.

[0066] Step 7: Define the lane change motivation function :

[0067]

[0068] in, Indicates the feasibility of changing lanes at the cross section. The weight, Indicates lane change benefits The weight.

[0069] The feasibility of the cross-section lane change Based on sensor observations at the tunnel entrance and exit locations, determine whether there is space for lane changing:

[0070]

[0071]

[0072]

[0073] in, , These indicate the feasibility of lane changing at the tunnel entrance and exit locations, respectively. For example... Figure 5 As shown, for a vehicle to complete an overtaking maneuver, the distance between it and the adjacent vehicle must meet the safe distance requirement. =1.5~2.0s, indicating the set safe headway. , This indicates the distance between the front of the vehicle and the adjacent lane when the vehicle enters and exits the tunnel. The closest distance to the rear of the vehicle. , This indicates the distance between the rear of the vehicle and the adjacent lane when the vehicle enters and exits the tunnel. The closest distance to the front of the vehicle. When When this occurs, it indicates that there is no space for lane changing at either the tunnel entrance or exit. This indicates that there is space for lane changing at the tunnel exit or entrance. This indicates that there is space for lane changing at both the tunnel entrance and exit.

[0074] Because sensor data inside the tunnel is missing, a lane-changing benefit function is constructed by comprehensively considering the traffic density index P of the adjacent lane and the speed advantage K. To assess whether a lane change occurred:

[0075]

[0076] in, , This indicates the corresponding weight. When... At that time, changing lanes would yield greater benefits, leading to the perception that there was an incentive to overtake. At that time, the benefit of changing lanes is less than that of maintaining the original position, so it is believed that there is no obvious motivation to overtake.

[0077] The traffic flow density index P is defined as the current lane. Number of vehicles Adjacent lane Number of vehicles The ratio:

[0078]

[0079] When P < 1, it indicates that the traffic density in the current lane is lower, and the driver is more likely to not change lanes. When P > 1, it indicates that the traffic density in the current lane is higher than that in the adjacent lane, and the driver is more likely to change lanes.

[0080] As a preferred embodiment, the dynamic traffic flow density within the lane is calculated. To measure traffic density index P:

[0081]

[0082]

[0083]

[0084] in, The average headway of a lane is represented by L, the tunnel length is represented by n, and the number of vehicles in the lane is represented by n. Indicates the length of the vehicle. This indicates the average speed of vehicles traveling within the lane.

[0085] The speed advantage K is calculated based on the speed-density relationship:

[0086]

[0087]

[0088]

[0089] in, Indicates the speed of traffic flow in a lane. This represents the congestion density given in the Greenshields model. This indicates the vehicle density of the lane. When the current lane... Traffic speed Larger than the adjacent lane Traffic speed At that time, the current lane has a speed advantage, and vehicles will choose to return to the current lane after changing lanes to an adjacent lane. For example... Figure 6 As shown, after vehicle A changed lanes to lane 1, it failed to overtake due to the speed advantage in lane 2 and returned to lane 2.

[0090] Calculate inversion pairs ( , The lane-changing mechanisms of the two vehicles in the diagram, when > At that time, determine the original following vehicle Actively change lanes and overtake, when > At that time, determine the original vehicle in front. After changing lanes, the overtaking attempt failed, and the vehicle returned to its original lane.

Claims

1. A method for recognizing lane-changing behavior of vehicles in closed lanes based on limited cross-sectional observation information, characterized in that: The specific steps are as follows: Step 1: Deploy sensors at fixed points along the closed road to record the license plate of each vehicle, as well as the time, speed and lane information when passing the sensors, and construct a vehicle information vector. Step 2: Compare the lane information when the vehicle passes two consecutive sensors. If they are inconsistent, it is determined that the vehicle has changed lanes. Step 3: If the lane information of all vehicles recorded by two consecutive sensors remains unchanged, then sort the vehicles in the same lane according to the time they passed the sensor to obtain the entry sequence. and leaving the sequence ; Step 4: Traverse into the sequential sequence With leaving the sequence If the vehicle , If the relative positions of the elements change, then denote ( , ) is an inversion pair; Step 5: Define the free flow velocity Minimum lane change time cost and time measurement error For vehicles in the reverse pair, calculate their travel time residual between two adjacent sensors. : in, Indicates the actual travel time. This represents the travel time in free-flow state obtained based on vehicle dynamics and speed limit constraints; when the following conditions are met... , , If the travel time residuals of the two vehicles in the reverse pair satisfy... If yes, proceed to step 6; otherwise, proceed to step 7. Step 6, when , or 0 At that time, the judgment Changing lanes to overtake; Step 7, when , , Cannot be true at the same time, or At that time, calculate the overtaking motivation of the vehicle in the reverse alignment: in, Indicates the feasibility of changing lanes at the cross section. The weight, Indicates lane change benefits The weight; when > At that time, determine the original following vehicle Actively change lanes and overtake, when > At that time, determine the original vehicle in front. After changing lanes, the vehicle failed to overtake and returned to its original lane.

2. The method for recognizing lane-changing behavior of vehicles in a closed lane based on limited cross-sectional observation information as described in claim 1, characterized in that: The feasibility of the cross-section lane change Based on sensor observations, determine whether there is space for lane changing at the cross-section location: in, , These represent the lane change feasibility at the observation positions of the two sensors, with in and out representing the preceding and following sensors along the driving direction, respectively. This indicates the set safe headway. , This indicates the distance between the front of the vehicle and the adjacent lane when the vehicle passes the sensor. The closest distance to the rear of the vehicle. , This indicates the distance between the rear of the vehicle and the adjacent lane when the vehicle passes the sensor. The closest distance to the front of the vehicle.

3. The method for recognizing lane-changing behavior of vehicles in a closed lane based on limited cross-sectional observation information as described in claim 2, characterized in that: Setting safe headway =1.5~2.0s.

4. The method for recognizing lane-changing behavior of vehicles in a closed lane based on limited cross-sectional observation information as described in claim 1, characterized in that: The lane change benefit function Taking into account both the traffic density index P of the adjacent lane and the speed advantage K: in , Indicates the corresponding weight; , These represent the lanes the vehicle is currently traveling in. Adjacent lane The number of vehicles on the road; , These represent the lanes the vehicle is currently traveling in. Adjacent lane Traffic flow speed.

5. The method for recognizing lane-changing behavior of vehicles in a closed lane based on limited cross-sectional observation information as described in claim 4, characterized in that: By calculating the dynamic traffic density within the lane To measure traffic density index P: in, The average headway of a vehicle in the lane is represented by L, the road length between two adjacent sensors is represented by n, and the number of vehicles in the lane is represented by n. Indicates the length of the vehicle. This indicates the average speed of vehicles traveling within the lane.

6. The method for recognizing lane-changing behavior of vehicles in a closed lane based on limited cross-sectional observation information as described in claim 4, characterized in that: Calculate traffic flow speed based on the Greenshields model : in, This represents the congestion density given in the Greenshields model. The value represents the vehicle density of the lane, L represents the road length between the two sensors, and n represents the number of vehicles in the lane.

7. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1 to 6.

Images