CN117576948A - Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics - Google Patents

Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics Download PDF

Info

Publication number
CN117576948A
CN117576948A CN202311531226.XA CN202311531226A CN117576948A CN 117576948 A CN117576948 A CN 117576948A CN 202311531226 A CN202311531226 A CN 202311531226A CN 117576948 A CN117576948 A CN 117576948A
Authority
CN
China
Prior art keywords
motor vehicle
vehicle
track
vehicles
motor
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.)
Pending
Application number
CN202311531226.XA
Other languages
Chinese (zh)
Inventor
郑玉冰
马羊
冯忠祥
黄从俊
蒋旭
刘宗族
黄河
方诗圣
张卫华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hefei Urban Planning & Design Institute
Original Assignee
Hefei Urban Planning & Design Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hefei Urban Planning & Design Institute filed Critical Hefei Urban Planning & Design Institute
Priority to CN202311531226.XA priority Critical patent/CN117576948A/en
Publication of CN117576948A publication Critical patent/CN117576948A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/16—Anti-collision systems
    • G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention provides a multi-mode collision risk identification method for motor vehicle flows at an intersection taking queue characteristics into consideration, which belongs to the technical field of urban road traffic running risk identification and state monitoring and comprises the steps of acquiring real-time tracks of motor vehicles at the intersection and generating a motor vehicle track belt; rasterizing an intersection, and assigning values and binarizing all grids in a polygonal area occupied by all motor vehicle track belts; extracting motor vehicle queues through a communication area identification algorithm, principal component analysis and outlier rejection and sequencing; respectively identifying rear-end collision, angle collision and scraping collision of motor vehicles in the queue; dynamically outputting traffic conflict risk conditions in motor vehicle flows in the intersection in a thermodynamic diagram form; on the basis of the running characteristics of the motor vehicles with the queue running characteristics in the excavator motor vehicle flow, the mutual interference characteristics among the vehicles in the motor vehicle team and between the motor vehicle team and other external vehicles are considered, and the dynamic situation of traffic conflict risks in the motor vehicle flow is fully identified.

Description

Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics
Technical Field
The invention relates to the technical field of urban road traffic running risk identification and state monitoring, in particular to a multi-mode collision risk identification method for motor vehicle flows at intersections, which considers queue characteristics.
Background
At urban signal control intersections, the motor vehicle flows are large and mutual interference is frequently generated in various motor vehicles. The state of interaction risk between vehicles is often described in the prior art as a traffic collision, i.e. two or more traffic participants are in close temporal and spatial proximity to each other, wherein at least one of the traffic participants has to adjust the state of motion to avoid traffic accidents. At intersections, identifying traffic conflict conditions within a motor vehicle flow has a critical role in assessing the operational safety level of the traffic flow as a whole.
For the urban intersections adopting signal control, motor vehicle flows in different driving directions can pass through according to the signal phases in sequence. In any green light phase, a motor vehicle running in the same direction can have obvious following phenomenon, namely, a rear vehicle follows a front vehicle to adjust the running route and the movement state of the rear vehicle, and the situation that the motor vehicle running in the same direction can spontaneously form a queue to run through an intersection is reflected on a macroscopic level. Along with the continuous change of the vehicle position and the movement state, the queues in the vehicle flow are dynamically generated, changed and disappeared along with the movement process of the vehicles, so that the mutual interference process between the vehicles in the motor vehicle flow is more complex, and the interaction risks can be generated between the motor vehicles in the queues and the motor vehicles outside the queues in the longitudinal direction and the transverse direction, thereby causing traffic collision. For example, for a motor vehicle flow that is turning in a train, the lead vehicle may collide with the opposing straight vehicle flow at an angle, and the lead vehicle may collide with the rear vehicle at a rear end, and the motor vehicles in the train may collide with the motor vehicle flow traveling straight out of the train at an angle. It follows that if one wants to evaluate the running state and the safety level of the motor vehicle flow at the signal control intersection accurately, one has to consider the dynamic queue characteristics in the motor vehicle flow in different driving directions to fully mine the traffic collision risk between the motor vehicles in the queues, between the different queues and between the queues and other single motor vehicles. However, in the prior art, the recognition of the motor vehicle traffic conflict at the signal control intersection mainly starts from a single motor vehicle layer, and a certain specific form of traffic conflict risk is concerned, so that a multi-mode traffic conflict risk recognition analysis method established by considering the queue characteristics in each motor vehicle flow is rarely available.
The existing urban intersection motor vehicle traffic conflict risk scheme generally only focuses on risks of a certain specific form, such as rear-end collision and angle collision, and is insufficient in consideration of interweaving and complex interaction states among multidirectional motor vehicle flows in actual scenes, such as mutual interference between straight motor vehicles and co-rightward motor vehicles, straight motor vehicles and opposite-leftward motor vehicles, right-turning motor vehicles and opposite-leftward motor vehicles; in addition, in the prior art, collision risks of motor workshops at intersections are analyzed, and traffic collision is generally judged only by starting from the running state of a single motor vehicle according to whether the motor vehicle crosses tracks of other motor workshops or not. However, in an actual urban signal control intersection, particularly for an intersection with a large traffic flow, a motor vehicle workshop generally runs in a green light stage in a queue form, and collision disturbance on motor vehicles in a fleet may affect the running states of other motor vehicles in the fleet; furthermore, the location of the motor vehicle within the fleet affects the collision risk situation it is exposed to. The dynamic queue characteristics in the motor vehicle flow at the intersection are not fully considered in the prior art, and microscopic interaction characteristics of a motor vehicle under the influence of the queue driving state are not concerned. Based on the method, the invention provides a multi-mode collision risk identification method for the motor vehicle flows at the intersection in consideration of the queue characteristics.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a multi-mode collision risk identification method for motor vehicle flows at intersections, which considers the characteristics of queue, considers the characteristics of mutual interference among vehicles in motor vehicle teams and between the motor vehicle teams and other external vehicles on the basis of the characteristics of queue running in the motor vehicle flows, fully identifies rear-end collision, angle collision and scraping collision in the motor vehicle flows, and dynamically outputs the occurrence frequency and distribution position of multi-mode collision events in a thermodynamic diagram mode.
In order to achieve the above purpose, the present invention provides the following technical solutions:
1. the intersection motor vehicle flow multi-mode conflict risk identification method considering the queue characteristics is characterized by comprising the following steps of:
(1) Acquiring a real-time track of the motor vehicle in the signal control intersection, and generating a motor vehicle track belt according to the real-time track of the motor vehicle;
(2) By (X) r ,Y r ) T Converting the signal control intersection area into an m multiplied by n grid area for the datum point in the geodetic coordinate system, assigning i to all grids in the polygonal area occupied by the motor vehicle i track area, and assigning i to all grids in the polygonal area occupied by all motor vehicle track areas to obtain a three-dimensional matrixWherein N is in Is the number of motor vehicles in the intersection; will->Summing along a third dimension and summing to obtain a matrix M m×n Binarization is carried out to generate a binary image;
(3) Analyzing whether motor vehicle track bands in the binary image are overlapped or not through a communication area identification algorithm, finding out motor vehicles with overlapped track bands, and searching through grid positions to obtain IDs of the corresponding motor vehicles to obtain motor vehicles belonging to the same set; after removing vehicles to be separated from a motorcade, arranging vehicles in the motorcade according to the fact that an included angle between a connecting line vector of mass center points of two vehicles and the advancing direction of the motor vehicle j is an obtuse angle for any motor vehicle i behind the motor vehicle j;
(4) Establishing a track pool of the track of the turning motor vehicle, predicting the track of the turning motor vehicle by adopting a track matching method, and predicting the track of the straight motor vehicle by adopting an extrapolation method based on the current running direction of the straight motor vehicle to assist in identifying traffic conflicts in subsequent motor vehicle flows;
(5) The method comprises the steps of considering the positions and the running directions of motor vehicles in a queue, identifying the angle conflict and the scraping conflict between a head vehicle and other motor vehicles, identifying the rear-end collision conflict and the scraping conflict of a rear vehicle, and calculating the positions of corresponding traffic collision events;
(6) In the rasterized intersection area, counting the total number of conflict frequencies of all grids in the Deltat by taking the Deltat as a time interval, generating the thermodynamic distribution of the conflict risk of the motor vehicle flow in the intersection area according to the conflict frequency, obtaining a conflict frequency thermodynamic diagram, and updating in real time along with the dynamic change of the traffic operation process.
The method for generating the motor vehicle track belt according to the real-time track of the motor vehicle comprises the following steps:
for a motor vehicle with ID i, at the current time t, toCharacterization of N in the past t Historical trace of time, where P t Characterizing a track point of the motor vehicle i at a moment t; respectively shifting all track points along the track tangent line to two sides by w i /2，w i Obtaining a historical track band of the motor vehicle i for the width of the vehicle body as shown in a formula (1); when the track belt of the motor vehicle is constructed, the positions of the head and the tail of the motor vehicle at different moments are contained in the track belt;
wherein:and->The coordinates are the mass center of the motor vehicle i at the moment (t-1), the ground coordinates of the left boundary point and the right boundary point, X is a transverse coordinate, and Y is a longitudinal coordinate; />For the azimuth angle of motor vehicle i at time (t-1).
The specific method for rejecting vehicles to be removed from the fleet is as follows:
for any motorcade, taking the position of a motor vehicle as input at t time, obtaining the main direction u of the motorcade by using a principal component analysis method, projecting all motor vehicles to a new u-v coordinate system, selecting three motor vehicles from the queue, and constructing a spline curve f (u) in a u-v plane to fit the position of the selected motor vehicle; computer vehicle track point (u) t ,v t ) T Offset distance |v from fitting spline curve t -f (u) | and calculating an average offset distance value for all trajectory pointsIf the average offset distance is +.>Less than threshold d v Three motor vehicles all belong to the fleet; repeatedly selecting different three vehicles to combine until the average offset distance of the three vehicles is less than d v Obtaining initial three inner peripheral values; on the basis of this, the other motor vehicles continue to be selected in the queue if their average offset distance value is greater than d v Removing the waste residue; if the average offset distance is less than d v Adding the internal value as an internal value, performing spline curve fitting on the positions of the motor vehicles divided into the internal value again, and repeating the process until all the motor vehicles in the communication area are analyzed; so far, all motor vehicle queues and corresponding constituent vehicles in the intersection at the current moment are obtained.
The track prediction of the turning motor vehicle by using the track matching method is specifically as follows:
firstly, collecting a large number of motor vehicle tracks at the peak hour of an intersection, establishing a track pool, only reserving turning motor vehicle tracks, numbering each turning track, dividing track points on each track into a plurality of grid units, and establishing a path search matrix with m x n according to the distribution of the track points in each grid unit, wherein each element in the matrix stores the turning track number occupying the grid unit;
when the motor vehicle enters the intersection area, judging whether the motor vehicle is in a turning state or not by using a linear regression model; for a turning motor vehicle, rasterizing the historical track according to the same method of track processing, and further expanding the grid corresponding to the motor vehicle track by 1 grid unit size by adopting an image processing technology; establishing an occupied frequency histogram of all grid units according to a turning motor vehicle track pool; aiming at grid units occupied by historical tracks of a motor vehicle to be predicted, determining the track number of the existing motor vehicle with highest frequency according to the occupied frequency, and taking the track as the predicted track of the motor vehicle; and updating along with the time steps, repeating the track matching step, and dynamically predicting the track of the turning motor vehicle.
The identification of the angle conflict is specifically as follows:
assuming that the motor vehicle is only concerned about the risk of collision with the vehicle in view, a length of use is L i Width W i Is used for representing the motor vehicle in a rectangular simplified manner and is in an angle theta i Radius is R i Is representative of the field of view of the driver of the motor vehicle;
it is assumed that at the current instant t, the position and velocity vectors of the motor vehicle i are respectivelyAnd->First, the position of the motor vehicle i is taken as the origin, and the current advancing direction is +.>An axis perpendicular to +.>Rotated clockwise by 90 DEG to +.>The shaft establishes a local rectangular coordinate system, and converts the global coordinate and the speed vector into a local space of the motor vehicle i through a rectangular coordinate conversion method;
when the individual motor vehicles j with candidate conflict are in a straight running state, calculating the angle conflict by adopting an extrapolation method, quantifying the severity of the conflict by using TTC indexes, and calculating a TTC value by using formulas (2) - (4);
wherein:
wherein:the coordinates of the motor vehicle j in the partial space of the motor vehicle i at the time t; />Is the direction of travel of the vehicle j in the partial space of the motor vehicle i; />A speed scalar for vehicle i at time t; epsilon is a characteristic TTC i Or TTC j A sufficiently large constant;
identification of angular conflicts using vector fork multiplication when the vehicle j is corneringStarting point, predicted driving path of motor vehicle jDiameter->The upper point sets up a vector for the end point>Then pair->With the forward direction of the vehicle i>Vector cross-product calculation is carried out; when->At->When on the left side, the person is left>And->The cross value of (2) is positive, when +.>At->On the right side, the cross value is negative; />The position where the positive and negative polarities of the cross product change is a conflict point, if the conflict point does not exist, the conflict point indicates that the motor vehicle i and the motor vehicle j do not exist; when the motor vehicle i is a turning vehicle and the candidate collision vehicle is a straight-going vehicle, the angle collision judgment is also carried out by adopting the flow;
when both the vehicle i and the vehicle j are in the turning state, it is assumed thatIs the track of advance of motor vehicle i, then calculateAnd->Is to obtain a pair-wise distance matrix, i.e. calculate +.>And->A matrix established by Euclidean distance between different track points; on the basis, searching the minimum distance value in the matrix, and if the minimum distance value is smaller than a preset critical distance delta, considering that the angle conflict exists between the two vehicles; after the angle conflict is identified, the distances from the motor vehicle i to the conflict point of the motor vehicle j are respectively calculated along the two forward paths, and TTC indexes are calculated.
The identification of the rear-end collision is specifically as follows:
when the motor vehicle i is straight, converting the positions and directions of other motor vehicles in the visual field range into a local coordinate space of the motor vehicle i; then, consider only that in front of it and offset from its lateral position by less than w i Motor vehicle j of/2; calculating the rear-end collision between the motor vehicles i and j according to the formulas (5) to (6);
wherein:
wherein: l (L) i 、L j The lengths of the motor vehicle i and the motor vehicle j are respectively;respectively a speed scalar of the motor vehicle i and the motor vehicle j at a time t; a, a i 、a j Acceleration of the motor vehicle i and the motor vehicle j at the time t are respectively; ΔD of ij Between vehicles i and j in the direction of advance of the motor vehicleIs a relative distance of (2); deltav ij The relative speed between vehicle i and vehicle j; Δa ij The relative acceleration between the vehicle i and the vehicle j;
fitting a spline curve based on the history path of the motor vehicle i and the position of the motor vehicle j when the motor vehicle i turns; based on this, a curve distance Δd from motor vehicle i to motor vehicle j is calculated along a spline curve ij And then, judging the rear-end collision according to the rear-end collision calculation thought when the two vehicles run straight.
The scratch collision is identified as follows:
when the motor vehicle i is a straight-going vehicle, carrying out conflict recognition in a local coordinate system of the motor vehicle i; establishing an interest area to eliminate other vehicles which cannot have side collision conflict risks with the vehicle i, wherein the length of the interest area is 2L i Lane width 2W twice r The method comprises the steps of carrying out a first treatment on the surface of the Then, using the lateral velocity and the lateral clearance distance to calculate the vehicle-to-vehicle i-side event TTC in the region of interest ij Specific calculation is carried out according to formula (7);
wherein:for the angle between the direction of advance of motor vehicle i and motor vehicle j at time t, when +.>When (I)>Otherwise, let->W j Width of motor vehicle j;
when the motor vehicle i is in a turning state, the centroid of the motor vehicle i is taken as an origin, and a curve path is established along the centroidA shaft toLateral distance establishment of a candidate vehicle path>An axis, constructing a Frenet coordinate system; then, by +.>And->Establishing a new rectangular coordinate system for the coordinate components; by this process, the scratch collision analysis of the turning vehicle can be converted into a problem similar to that in which the vehicle is in a straight-running state, and then the scratch collision index is calculated according to equation (7).
For angle conflict, recording the position of a conflict point; for rear-end collision, recording the position of the tail of the front vehicle; for scratch conflicts, the midpoint position of the line connecting the centroids of the two vehicles is recorded.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention cuts in from the motor vehicle flow queue operation characteristics of the urban signal control intersection, considers the mutual interference characteristics of motor vehicles in each direction of the motorcade and motor vehicles outside the motorcade on the basis of dynamically identifying the motor vehicle set which runs according to the queue, respectively establishes the identification flow and risk index calculation method of rear-end collision conflict, angle conflict and scraping conflict for the straight motor vehicle flow and the turning motor vehicle flow, combines the thermodynamic diagram to realize the dynamic output of the occurrence frequency and the space position of various conflicts, can more fully and intuitively reveal the dynamic operation risk condition of the motor vehicle flow at the signal control intersection, and has positive significance for the real-time monitoring and management and control of the urban road traffic operation state.
2. Based on the track data of the motor vehicle, the dynamic identification and extraction of the motor vehicle queue are realized through a plurality of processes of track band generation, occupied area binarization, communication area generation and abnormal value elimination; and a track matching method is provided for track prediction of the turning motor vehicles in the queue running, so that a technical basis is provided for accurately identifying the multi-mode traffic collision risk among vehicles.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a flow chart of the motor vehicle track communication zone generation of the present invention.
FIG. 3 is a schematic illustration of outlier rejection in a motor vehicle queue according to the present invention.
FIG. 4 is a schematic view of the view field and the local coordinate system of the motor vehicle according to the present invention.
FIG. 5 is a schematic diagram of the calculation of the angular conflict of the motor vehicle according to the present invention.
Fig. 6 is a schematic diagram of a rear-end collision calculation in a motor vehicle according to the present invention.
FIG. 7 is a schematic diagram of a scratch conflict calculation for a motor vehicle of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments.
1. Technical overview
The invention provides a technical framework of a multi-mode collision risk identification method for motor vehicle flows at an urban signal control intersection in consideration of queue characteristics, which is shown in fig. 1. Aiming at urban signal control intersection scenes, based on motor vehicle track data, a plurality of processes of track belt generation, intersection rasterization, track belt occupation area binarization, motor vehicle track communication area generation and abnormal value elimination in a vehicle team set are adopted to identify motor vehicle sets with queue operation characteristics in motor vehicle flows in different running directions at the current moment. Then, for the motor vehicles running in a queue, track prediction is performed on the left-turning motor vehicle and the right-turning motor vehicle respectively by adopting a track matching method, and for the motor vehicles running in a straight line, track prediction is performed by adopting an extrapolation method based on the current running direction of the motor vehicles. On this basis, the recognition and calculation of the rear-end collision, the angle collision and the scratch collision are respectively performed in consideration of the difference in the positions of the motor vehicles in the train (the head car or the rear car) and the difference in the traveling direction (straight running or turning). And finally, dynamically outputting the traffic conflict risks in the motor vehicle flows in the intersections in a thermodynamic diagram form according to the distribution positions, the occurrence frequency and the like of the multi-mode traffic conflict events. The following will describe each step in the technical scheme of the present invention.
2. Motor vehicle fleet partitioning based on grid map
At the urban signal control intersection, each motor vehicle flow has obvious characteristic of running in a queue, and mutual interference between motor vehicles is also influenced by the queue state. For a signalized intersection, therefore, it is necessary to first analyze the queue conditions within the motor vehicle flow in assessing the traffic collision risk status of the motor vehicle flow.
In a vehicle-road cooperative environment, the position and motion state information of all motor vehicles at any moment in an intersection can be obtained; in a non-vehicle-road cooperative environment, dynamic tracking and information extraction of the motor vehicle can be performed by utilizing YOLOv5 in combination with deep start algorithm based on a monitoring video image at an intersection. The motor vehicles are numbered according to the time sequence of entering the intersection, and the motor vehicle with the ID of i is provided with the following steps at the current time tCharacterization of N in the past t Historical trace of time, where P t The position of motor vehicle i at time t (i.e. the locus point) is characterized. Respectively shifting all track points along the track tangent line to two sides by w i /2(w i Vehicle body width) as shown in formula (1), a history trace band of the motor vehicle i is obtained. It should be noted that in constructing a motor vehicle track belt, the positions of the head and tail of the motor vehicle at different moments should be included in the track belt.
Wherein:and->The geodetic coordinates of the centroid, the left boundary point and the right boundary point of the motor vehicle i at the moment (t-1) are respectively, X is a transverse coordinate, Y is a longitudinal coordinateCoordinates; />For the azimuth angle of motor vehicle i at time (t-1).
By (X) r ,Y r ) T The intersection area is converted into an m×n grid area as a reference point in the geodetic coordinate system, and the grid sizes along X, Y axes are respectively delta X And delta Y . Based on the history track zone T of the motor vehicle i i And assigning i to all grids in the polygonal area occupied by the track band.
As shown in FIG. 2, at any time, all the vehicles in the intersection are processed according to the above flow, so as to obtain a three-dimensional matrixWherein N is in Is the number of vehicles in the intersection. Based on this, will->Summing along a third dimension to obtain a matrix M m×n . Then, all non-zero elements in the matrix are assigned to 1, and a binary mask is generated. And analyzing whether the track bands of the motor vehicles in the binary image are overlapped by using a connected region identification algorithm, if the track bands of the two motor vehicles are overlapped, considering that the two motor vehicles belong to the same set, and outputting all grid positions in the connected region. By means of the grid positions obtained, in the matrix +.>The ID of the corresponding motor vehicle is retrieved. If there is more than one vehicle in a certain communication area, these vehicles are considered to belong to the same collection.
And then, further judging the motor vehicle set to remove abnormal values. In any motor vehicle train line, it is assumed that motor vehicle j is the lead vehicle and its heading isFor any motor vehicle i behind motor vehicle j, the line vector of the two centroid points is then equal to +.>The included angle of (1) is an obtuse angle, i.e.)>If motor vehicle j is a tail car, there is +.>According to the above method, the order of the vehicles in the collection can be determined.
It is considered that in a practical scenario, some vehicles may quickly get out of the fleet, where they no longer belong to the fleet, and therefore need to be rejected. For any fleet, at time t, the principal direction u of the fleet is obtained using principal component analysis with the vehicle position as input, and all vehicles are projected to a new u-v coordinate system, as shown in fig. 3. On the basis, three vehicles are selected from the queue at will, and a spline curve f (u) is constructed in the u-v plane to fit the position of the selected vehicle. Computer vehicle track point (u) t ,v t ) T Offset distance |v from fitting spline curve t -f (u) | and calculating an average offset distance value for all trajectory pointsIf for three selected motor vehicles, the average offset distanceLess than threshold d v Three vehicles all belong to the fleet. Repeatedly selecting different three vehicles to combine until the average offset distance of the three vehicles is less than d v The initial three inner values are obtained. On the basis of this, the other motor vehicles continue to be selected in the queue if their average offset distance value is greater than d v Removing the waste residue; if the average offset distance is less than d v Adding the internal value as the internal value, performing spline curve fitting on the motor vehicle position divided into the internal value again, and repeating the above process until all the connected areas are in the connected areaAnd (5) finishing analysis of the motor vehicle. So far, all motor vehicle queues and corresponding constituent vehicles in the intersection at the current moment are obtained.
3. Motor vehicle trajectory prediction
The track method based on deep learning has significantly reduced accuracy along with the increase of a prediction time window (such as more than 1 s), and cannot effectively and accurately identify traffic collision. In the signal control intersection, the movement of the straight motor vehicle and the turning motor vehicle follow a certain rule, so the invention utilizes the track matching method to predict the track of the turning motor vehicle so as to assist the identification of the traffic conflict in the follow-up motor vehicle flow. Firstly, collecting a large number of motor vehicle tracks at the peak hour of the intersection, establishing a track pool, only reserving the turning motor vehicle tracks, and numbering each turning track. According to the divided intersection grid units, track points on each track are divided into a plurality of grid units, and an m-by-n path search matrix is established according to the distribution of the track points in each grid unit, and each element in the matrix stores a turning track number occupying the grid unit.
When the motor vehicle enters the intersection area, a linear regression model is used for judging whether the motor vehicle is in a turning state or not. For a turning motor vehicle, the historical track is rasterized according to the same method of track processing, and the grids corresponding to the motor vehicle track are further expanded by 1 grid unit size by adopting an image processing technology. And establishing an occupied frequency histogram of all grid units according to the turning motor vehicle track pool. Aiming at grid units occupied by historical tracks of the motor vehicle to be predicted, determining the track number of the existing motor vehicle with highest frequency according to the occupied frequency, and taking the track as the predicted track of the motor vehicle. And updating along with the time steps, repeating the track matching step, and dynamically predicting the track of the turning motor vehicle. And the subsequent traffic conflict recognition for the turning motor vehicles is carried out based on the predicted track.
4. Motor vehicle angle conflict calculation
Within an intersection, the head car and the rear car in the same queue may face different collision risks. In particular, the head gear may be involved with other maneuversThe workshops are subject to angular and scraping conflicts, and then the vehicle may face rear-end collisions and scraping conflicts. Therefore, the risk of collision faced by the motor vehicles in the fleet needs to be considered according to the positions and the running directions of the motor vehicles. To simplify the collision calculation process, let the motor vehicle pay attention only to the collision risk between the motor vehicle and the vehicles in the visual field, the use length is L i Width W i Is used for representing the motor vehicle in a rectangular simplified manner and is in an angle theta i Radius is R i Is representative of the field of view of the driver of the motor vehicle, as shown in fig. 4 (a).
For angle conflicts, it is assumed that at the current instant t, the position and velocity vectors of motor vehicle i are respectivelyAnd->First, the position of the motor vehicle i is taken as the origin, and the current advancing direction is +.>An axis perpendicular to +.>Rotated clockwise by 90 DEG to +.>The shaft establishes a local rectangular coordinate system as shown in fig. 4 (b), and the global coordinate and the velocity vector can be converted into a local space of the motor vehicle i by a rectangular coordinate conversion method.
When the individual motor vehicle j candidate for collision is in a straight-ahead state, as in fig. 5 (a), the angular collision is calculated by extrapolation, and the severity of the collision is quantified by a Time To Collision (TTC) index, and the TTC value is calculated by using equations (2) to (4).
Wherein:
wherein:the coordinates of the motor vehicle j in the partial space of the motor vehicle i at the time t; />Is the direction of travel of the vehicle j in the partial space of the motor vehicle i; />A speed scalar for vehicle i at time t; epsilon is a characteristic TTC i Or TTC j A sufficiently large constant.
When the motor vehicle j turns, the vector fork multiplication is adopted to recognize the angle conflict. As in FIG. 5 (b) toStarting point, & gt, with predicted driving path of motor vehicle j>The upper point sets up a vector for the end point>Then pair->With the forward direction of the vehicle i>Vector cross product calculation is performed. When->At->When on the left side, the person is left>And->The cross value of (2) is positive, when +.>At->On the right, the cross value is negative. />The position where the positive and negative polarities of the cross product change is the conflict point, and if the conflict point does not exist, the conflict point indicates that the motor vehicle i and the motor vehicle j do not exist. When the motor vehicle i is a turning vehicle and the collision vehicle candidate is a straight-running vehicle, the above-described flow is also employed to perform the angle collision judgment.
When both the vehicle i and the vehicle j are in the turning state, as shown in fig. 5 (c), it is assumed thatIs the forward trajectory of motor vehicle i, then +.>And->Is to obtain a pair-wise distance matrix, i.e. calculate +.>And->And a matrix established by Euclidean distances among different track points. On the basis, searching the minimum distance value in the matrix, and if the minimum distance value is smaller than delta, considering that the angle conflict exists between the two vehicles. After the angle conflict is identified, the distances from the motor vehicle i to the conflict point of the motor vehicle j are respectively calculated along the two forward paths, and TTC indexes are calculated.
5. Rear-end collision calculation for motor vehicle
Similar to the angle collision, the rear-end collision is also classified into a straight vehicle or a cornering vehicle as the current vehicle.
When the motor vehicle i is straight, the positions and directions of other motor vehicles in the visual field range are converted into the local coordinate space of the motor vehicle i. Then, consider only that in front of it and offset from its lateral position by less than w i Motor vehicle j of/2. As shown in fig. 6 (a), the calculation of the rear-end collision between the motor vehicles i and j is performed according to equations (5) to (6).
Wherein:
wherein: l (L) i 、L j The lengths of the motor vehicle i and the motor vehicle j are respectively;respectively a speed scalar of the motor vehicle i and the motor vehicle j at a time t; a, a i 、a j Acceleration of the motor vehicle i and the motor vehicle j at the time t are respectively; ΔD of ij Is the relative distance between vehicle i and vehicle j in the direction of motor vehicle travel; deltav ij The relative speed between vehicle i and vehicle j; Δa ij The relative acceleration between the vehicle i and the vehicle j.
When vehicle i is turning, as in fig. 6 (b), a spline curve is fitted based on the historic path of vehicle i and the position of vehicle j. Based on this, a curve distance Δd from motor vehicle i to motor vehicle j is calculated along a spline curve ij And then, judging the rear-end collision according to the rear-end collision calculation thought when the two vehicles run straight.
6. Motor vehicle scratch conflict identification
When the motor vehicle i is a straight-traveling vehicle, collision recognition is performed in a local coordinate system thereof. As shown in FIG. 7 (a), a region of interest (ROI) is established to reject other non-regionsVehicles that would risk side collision with vehicle i, the length of the ROI is 2L i Lane width 2W twice r . Then, using the lateral velocity and the lateral clearance distance to calculate the vehicle-to-vehicle i-side event TTC within the ROI ij The specific calculation is performed according to formula (7).
Wherein:for the angle between the direction of advance of motor vehicle i and motor vehicle j at time t, when +.>When (I)>Otherwise, let->W j Width of motor vehicle j. The meaning of the remaining symbols is consistent with the above.
When the motor vehicle i is in a turning state, as shown in fig. 7 (b), the centroid of the motor vehicle i is taken as the origin, and a curve path is establishedAxle, set up +_ with lateral distance to candidate vehicle path>And (5) an axis, and constructing a Frenet coordinate system. Then, by +.>And->For the coordinate components, a new rectangular coordinate system is established. By this process, the scratch collision analysis of the turning vehicle can be converted into a problem similar to that of the vehicle in a straight-running state, and then the scratch collision is calculated according to equation (7)And (5) an index.
7. Output of a thermodynamic diagram of motor vehicle flow traffic collision risk
And at any moment, for all motor vehicles in the intersection, based on the conflict automatic calculation process, adopting a parallel calculation idea, synchronously calculating to obtain the traffic conflict state between the motor vehicles and the surrounding motor vehicles, and calculating the position of the corresponding traffic conflict event. For angle conflict, recording the position of a conflict point; for rear-end collision, recording the positions of the head and the tail of the vehicle; for scratch conflicts, the midpoint position of the line connecting the centroids of the two vehicles is recorded. So far, the position information of all traffic conflict events in the motor vehicle flow at the moment is obtained.
And calculating the traffic collision frequency of each grid by taking Deltat as a time interval in the grid intersection area. If the location of the traffic conflict event is within a certain grid area, the event is considered to be distributed in the grid. Counting the total number of conflict frequencies of all grids in delta t, characterizing low frequencies by blue according to the conflict frequencies, characterizing high frequencies by red, generating thermodynamic distribution of motor vehicle flow conflict risks in an intersection area, obtaining a conflict frequency thermodynamic diagram, and updating in real time along with dynamic changes of traffic running processes. In addition, different colors can be used for respectively representing different types of traffic conflict events, and dynamic time-space distribution of the different types of conflict events is respectively displayed in the grid area of the intersection.
The foregoing description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical solution of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.

Claims (8)

1. The intersection motor vehicle flow multi-mode conflict risk identification method considering the queue characteristics is characterized by comprising the following steps of:
(1) Acquiring a real-time track of the motor vehicle in the signal control intersection, and generating a motor vehicle track belt according to the real-time track of the motor vehicle;
(2) By (X) r ,Y r ) T Converting the signal control intersection area into an m multiplied by n grid area for the datum point in the geodetic coordinate system, assigning i to all grids in the polygonal area occupied by the motor vehicle i track area, and assigning i to all grids in the polygonal area occupied by all motor vehicle track areas to obtain a three-dimensional matrixWherein N is in Is the number of motor vehicles in the intersection; will->Summing along a third dimension and summing to obtain a matrix M m×n Binarization is carried out to generate a binary image;
(3) Analyzing whether motor vehicle track bands in the binary image are overlapped or not through a communication area identification algorithm, finding out motor vehicles with overlapped track bands, and searching through grid positions to obtain IDs of the corresponding motor vehicles to obtain motor vehicles belonging to the same set; after removing vehicles to be separated from a motorcade, sorting the vehicles in the motorcade according to the fact that the included angle between the connecting line vector of the centroid points of two vehicles and the advancing direction of the motor vehicle j is an obtuse angle for any motor vehicle i behind the motor vehicle j;
(4) Establishing a track pool of the track of the turning motor vehicle, predicting the track of the turning motor vehicle by adopting a track matching method, and predicting the track of the straight motor vehicle by adopting an extrapolation method based on the current running direction of the straight motor vehicle to assist in identifying traffic conflicts in subsequent motor vehicle flows;
(5) The method comprises the steps of considering the positions and the running directions of motor vehicles in a queue, identifying the angle conflict and the scraping conflict between a head vehicle and other motor vehicles, identifying the rear-end collision conflict and the scraping conflict of a rear vehicle, and calculating the positions of corresponding traffic collision events;
(6) In the rasterized intersection area, counting the total number of conflict frequencies of all grids in the Deltat by taking the Deltat as a time interval, generating the thermodynamic distribution of the conflict risk of the motor vehicle flow in the intersection area according to the conflict frequency, obtaining a conflict frequency thermodynamic diagram, and updating in real time along with the dynamic change of the traffic operation process.
2. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking into consideration queue characteristics according to claim 1, wherein the generation of motor vehicle track bands according to real-time tracks of motor vehicles is specifically as follows:
for a motor vehicle with ID i, at the current time t, toCharacterization of N in the past t Historical trace of time, where P t Characterizing a track point of the motor vehicle i at a moment t; respectively shifting all track points along the track tangent line to two sides by w i /2，w i Obtaining a historical track band of the motor vehicle i for the width of the vehicle body as shown in a formula (1); when the track belt of the motor vehicle is constructed, the positions of the head and the tail of the motor vehicle at different moments are contained in the track belt;
wherein:and->The coordinates are the mass center of the motor vehicle i at the moment (t-1), the ground coordinates of the left boundary point and the right boundary point, X is a transverse coordinate, and Y is a longitudinal coordinate; />For the azimuth angle of motor vehicle i at time (t-1).
3. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking into consideration queue characteristics according to claim 1, wherein the specific method for eliminating vehicles coming off from the fleet is as follows:
for any motorcade, taking the position of a motor vehicle as input at t time, obtaining the main direction u of the motorcade by using a principal component analysis method, projecting all motor vehicles to a new u-v coordinate system, selecting three motor vehicles from the queue, and constructing a spline curve f (u) in a u-v plane to fit the position of the selected motor vehicle; computer vehicle track point (u) t ,v t ) T Offset distance |v from fitting spline curve t -f (u) | and calculating an average offset distance value for all trajectory pointsIf the average offset distance is +.>Less than threshold d v Three motor vehicles all belong to the fleet; repeatedly selecting different three vehicles to combine until the average offset distance of the three vehicles is less than d v Obtaining initial three inner peripheral values; on the basis of this, the other motor vehicles continue to be selected in the queue if their average offset distance value is greater than d v Removing the waste residue; if the average offset distance is less than d v Adding the internal value as an internal value, performing spline curve fitting on the positions of the motor vehicles divided into the internal value again, and repeating the process until all the motor vehicles in the communication area are analyzed; so far, all motor vehicle queues and corresponding constituent vehicles in the intersection at the current moment are obtained.
4. The method for identifying the multi-mode collision risk of motor vehicle flows at intersections taking queue characteristics into consideration according to claim 1, wherein the track prediction of the turning motor vehicles by utilizing a track matching method is specifically as follows:
firstly, collecting a large number of motor vehicle tracks at the peak hour of an intersection, establishing a track pool, only reserving turning motor vehicle tracks, numbering each turning track, dividing track points on each track into a plurality of grid units, and establishing a path search matrix with m x n according to the distribution of the track points in each grid unit, wherein each element in the matrix stores the turning track number occupying the grid unit;
when the motor vehicle enters the intersection area, judging whether the motor vehicle is in a turning state or not by using a linear regression model; for a turning motor vehicle, rasterizing the historical track according to the same method of track processing, and further expanding the grid corresponding to the motor vehicle track by 1 grid unit size by adopting an image processing technology; establishing an occupied frequency histogram of all grid units according to a turning motor vehicle track pool; aiming at grid units occupied by existing tracks of the current motor vehicle, determining the track number of the existing motor vehicle in a track pool with highest frequency according to the occupied frequency, and taking the track as a predicted track of the current motor vehicle; and updating along with the time steps, repeating the track matching step, and dynamically predicting the track of the turning motor vehicle.
5. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking into consideration queue characteristics according to claim 1, wherein the identification of the angular collision is specifically as follows:
assuming that the motor vehicle is only concerned about the risk of collision with the vehicle in view, a length of use is L i Width W i Is used for representing the motor vehicle in a rectangular simplified manner and is in an angle theta i Radius is R i Is representative of the field of view of the driver of the motor vehicle;
it is assumed that at the present instant t, the position and velocity vectors of motor vehicle i are (x i t ,y i t ) T And (3) withFirst, the position of the motor vehicle i is taken as the origin, and the current advancing direction is +.>An axis perpendicular toIn->Rotated clockwise by 90 DEG to +.>The shaft establishes a local rectangular coordinate system, and converts the global coordinates and the speed vector into a local space of the motor vehicle i through a rectangular coordinate conversion method;
when the individual motor vehicles j with candidate conflict are in a straight running state, calculating the angle conflict by adopting an extrapolation method, quantifying the severity of the conflict by using TTC indexes, and calculating a TTC value by using formulas (2) - (4);
wherein:
wherein:the coordinates of the motor vehicle j in the partial space of the motor vehicle i at the time t; />Is the direction of travel of the vehicle j in the partial space of the motor vehicle i; />A speed scalar for vehicle i at time t; epsilon is a characteristic TTC i Or TTC j A sufficiently large constant;
identification of angular conflicts using vector fork multiplication when the vehicle j is corneringStarting point, & gt, with predicted driving path of motor vehicle j>The upper point sets up a vector for the end point>Then pair->With the forward direction of the vehicle i>Vector cross-product calculation is carried out; when->At->When on the left side, the person is left>And->The cross value of (2) is positive, when +.>At->On the right side, the cross value is negative; />The position where the positive and negative polarities of the cross product change is a conflict point, if the conflict point does not exist, the conflict point indicates that the motor vehicle i and the motor vehicle j do not exist; when the motor vehicle i is a turning vehicle and the candidate collision vehicle is a straight-going vehicle, the angle collision judgment is also carried out by adopting the flow;
when both the vehicle i and the vehicle j are in the turning state, it is assumed thatIs the forward trajectory of motor vehicle i, then +.>And (3) withIs to obtain a pair-wise distance matrix, i.e. calculate +.>And->A matrix established by Euclidean distance between different track points; on the basis, searching the minimum distance value in the matrix, and if the minimum distance value is smaller than a preset critical distance delta, considering that the angle conflict exists between the two vehicles; after the angle conflict is identified, the distances from the motor vehicle i to the conflict point of the motor vehicle j are respectively calculated along the two forward paths, and TTC indexes are calculated.
6. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking queue characteristics into consideration according to claim 5, wherein the identification of rear-end collision is specifically as follows:
when the motor vehicle i is straight, converting the positions and directions of other motor vehicles in the visual field range into a local coordinate space of the motor vehicle i; then, consider only that in front of it and offset from its lateral position by less than w i Motor vehicle j of/2; calculating the rear-end collision between the motor vehicles i and j according to the formulas (5) to (6);
wherein:
wherein: l (L) i 、L j The lengths of the motor vehicle i and the motor vehicle j are respectively;respectively a speed scalar of the motor vehicle i and the motor vehicle j at a time t; a, a i 、a j Acceleration of the motor vehicle i and the motor vehicle j at the time t are respectively; ΔD of ij Is the relative distance between vehicle i and vehicle j in the direction of motor vehicle travel; deltav ij The relative speed between vehicle i and vehicle j; Δa ij The relative acceleration between the vehicle i and the vehicle j;
fitting a spline curve based on the history path of the motor vehicle i and the position of the motor vehicle j when the motor vehicle i turns; based on this, a curve distance Δd from motor vehicle i to motor vehicle j is calculated along a spline curve ij And then, judging the rear-end collision according to the rear-end collision calculation thought when the two vehicles run straight.
7. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking queue characteristics into consideration according to claim 5, wherein the identification of scraping collisions is specifically as follows:
when the motor vehicle i is a straight-going vehicle, carrying out conflict recognition in a local coordinate system of the motor vehicle i; taking the current vehicle mass center point as the center and the current vehicle advancing direction as the symmetry axis, establishing an interest area to eliminate other vehicles which cannot have side collision conflict risks with the vehicle i, wherein the length of the interest area is 2L i Width ofDouble lane width 2W r The method comprises the steps of carrying out a first treatment on the surface of the Then, using the lateral velocity and the lateral clearance distance to calculate the time distance TTC from vehicle j to the side of vehicle i in the region of interest ij Specific calculation is carried out according to formula (7);
wherein:for the angle between the direction of advance of motor vehicle i and motor vehicle j at time t, when +.>When (I)>Otherwise, let->W j Width of motor vehicle j;
when the motor vehicle i is in a turning state, the centroid of the motor vehicle i is taken as an origin, and a curve path is established along the centroidAxle, set up +_ with lateral distance to candidate vehicle path>An axis, constructing a Frenet coordinate system; then, by +.>And->Establishing a new rectangular coordinate system for the coordinate components; through the processThe scratch collision analysis of the turning vehicle may be converted into a problem similar to the vehicle being in a straight-ahead state, and then the scratch collision index may be calculated according to equation (7).
8. The method for identifying the risk of multi-mode collision of motor vehicle flows at intersections taking into account queue characteristics according to claim 1, wherein for angular collisions, the position of the conflict point is recorded; for rear-end collision, recording the position of the tail of the front vehicle; for scratch conflicts, the midpoint position of the line connecting the centroids of the two vehicles is recorded.
CN202311531226.XA 2023-11-16 2023-11-16 Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics Pending CN117576948A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311531226.XA CN117576948A (en) 2023-11-16 2023-11-16 Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311531226.XA CN117576948A (en) 2023-11-16 2023-11-16 Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics

Publications (1)

Publication Number Publication Date
CN117576948A true CN117576948A (en) 2024-02-20

Family

ID=89894848

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311531226.XA Pending CN117576948A (en) 2023-11-16 2023-11-16 Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics

Country Status (1)

Country Link
CN (1) CN117576948A (en)

Similar Documents

Publication Publication Date Title
Chandra et al. Forecasting trajectory and behavior of road-agents using spectral clustering in graph-lstms
CN110488802B (en) Decision-making method for dynamic behaviors of automatic driving vehicle in internet environment
US11131993B2 (en) Methods and systems for trajectory forecasting with recurrent neural networks using inertial behavioral rollout
CN111382768B (en) Multi-sensor data fusion method and device
CN111445699B (en) Intersection traffic conflict discrimination method based on real-time vehicle track
US11003928B2 (en) Using captured video data to identify active turn signals on a vehicle
JP7119365B2 (en) Driving behavior data generator, driving behavior database
CN110531661A (en) A kind of vehicle is automatically with control method of speeding, device and equipment
CN113112805B (en) Intersection monitoring and early warning method based on base station communication and intersection camera positioning
Zhang et al. A framework for turning behavior classification at intersections using 3D LIDAR
Ghahremannezhad et al. Real-time accident detection in traffic surveillance using deep learning
CN113665570A (en) Method and device for automatically sensing driving signal and vehicle
CN112833903A (en) Trajectory prediction method, apparatus, device and computer readable storage medium
Trentin et al. Interaction-aware intention estimation at roundabouts
CN117576948A (en) Intersection motor vehicle flow multi-mode conflict risk identification method considering queue characteristics
Cao et al. Vehicle motion analysis based on a monocular vision system
CN114529883A (en) Method and electronic device for detecting traffic light
CN116529141A (en) Motion planning in autonomous vehicle curve coordinate system
CN114429622A (en) Automatic vehicle driving method and device, vehicle and storage medium
CN106097751A (en) Vehicle travel control method and device
Schömer et al. What if? behavior estimation by predicting traffic scenes from state histories
CN116451084B (en) Training sample preprocessing method for driving style recognition model
US20230252638A1 (en) Systems and methods for panoptic segmentation of images for autonomous driving
Cheung et al. Classifying driver behaviors for autonomous vehicle navigation
CN115421161B (en) Unmanned mine car control method based on laser radar ranging

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