A trajectory matching and fusion method for multi-radar detection at intersections

Abstract

The present invention discloses a method for matching and fusing multiple radar detection trajectories at an intersection. The method obtains and calculates the reference trajectories of vehicles in each lane of the intersection and the actual trajectories of the radar equipment detecting and updating the position points in real time. By matching the actual trajectories detected by the radar with the preset reference trajectories, multiple trajectory fragments of the same vehicle detected within the same radar and between different radars are matched and fused in real time, thereby outputting the complete driving trajectory of the same vehicle passing through the intersection, achieving a good real-time matching and fusion effect. The present invention solves the problem that the detection range of a single radar cannot cover the entire area of ​​the intersection, as well as the problem that the detection trajectory of the same radar frequently jumps and the detection trajectory of different radars is difficult to connect in real time in actual applications. The method has the advantages of simple implementation, high accuracy, fast matching, strong real-time performance, and low computational complexity.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent transportation multi-sensor fusion, and in particular relates to a multi-radar detection trajectory matching and fusion method for intersections. Background Art

[0002] Radar: A traffic radar microwave detector installed at one entrance to an intersection. It periodically monitors the real-time location of vehicles in the center and oncoming entrances. Traffic radar is used to detect the real-time location of vehicles in a target area. The effective detection range of a radar device installed at one entrance to an intersection is typically a sector-shaped area, encompassing part of the center of the intersection and the oncoming entrances. For an intersection, four radars are required at each of the four entrances to create a complete detection area covering the center and all four entrances. During a single vehicle's passage through the intersection, multiple consecutive position data points at different times can be linked using the same detection target ID to generate a track segment. Due to the limited detection range of a single radar, the complete track of a vehicle through the intersection requires combining multiple track segments generated by multiple radars. The overlapping segments within the shared detection area are then compared and matched to form the complete track of the vehicle through the intersection.

[0003] In actual applications, due to the limitations of the radar equipment's detection capabilities, the radar equipment cannot detect vehicles that are moving slowly or stopped, or are blocked by other vehicles during the detection process. The detection trajectory of the same radar for the same vehicle during a single passage through an intersection may jump, and the target ID may change, forming different trajectories. The detection trajectories of different radars may be lost or offset, resulting in large offsets of multiple radar trajectory fragments in the common detection area, making it difficult to match and connect the trajectories detected by each radar.

[0004] For example, a Chinese patent document, "A Multi-radar Fusion Track State Estimation Method Based on Variational Theory," has the publication number "CN111679269A." This invention relates to a multi-radar fusion track state estimation method based on variational theory, belonging to the field of multi-sensor target tracking and data fusion. To overcome the problem that radar intelligence processing systems often oscillate back and forth between multiple original tracks when fusing tracks, resulting in a jagged pattern that affects track quality, and the problem that conventional methods are prone to turning "bumps" or even tracking failure, this invention uniformly manages observation information from multiple radars, constructs a variational model that includes the target motion velocity transformation curve, and provides a fast solution method for the variational model. The solved motion velocity curve is used to extrapolate the tracks of all observing radars, and finally estimates and updates the target state estimate. This invention maximizes the correlation of observation information, reduces the adverse effects of relative track deviation and sudden target maneuvers on target estimation, improves the accuracy of target motion velocity and fused position estimation, enhances track quality, and achieves reliable target tracking. However, all of the aforementioned solutions are open-space, non-specific trajectory target detection trajectory fusion methods. This patent matches and connects multiple radar detection trajectories based on the reference trajectory of each lane, solving the problem of incomplete detection range of a single radar and frequent trajectory interruptions. Summary of the Invention

[0005] This invention aims to address the problem of a single radar's detection range failing to fully cover an intersection, as well as the frequent jumps in the same radar's detection trajectory and the difficulty in real-time merging different radar detection trajectories in practical applications. A method for matching and fusing multiple radar detection trajectories at intersections is provided. This method is used to match and merge vehicle trajectories detected by multiple radars at different intersection entrances, outputting the complete trajectory of the same vehicle passing through the same intersection in a single pass. The proposed method offers the advantages of simple implementation, high accuracy, fast matching, strong real-time performance, and minimal computational effort, achieving excellent real-time matching and fusion results.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for matching and fusing multi-radar detection trajectories at intersections, comprising the following steps:

[0007] S1, install radar equipment near the stop line at each entrance of the intersection;

[0008] S2, calibrating the area corresponding to each lane at the intersection;

[0009] S3, calibrating the area range corresponding to each exit of the intersection;

[0010] S4, obtaining the vehicle driving reference trajectory TR of each lane at the intersection ref ;

[0011] S5, the radar device detects and updates each track of the position point in real time as TR cur , if TR cur The number of location points contained is greater than the threshold τ valid , then mark the trajectory as a valid trajectory TR valid , calculate TR valid and all reference trajectories S for all lanes ref Each reference trajectory TR in ref The correlation matching degree θ between rel ;

[0012] S6, select the correlation matching degree θ rel Highest TR ref Recorded as TR match , the above TR valid Join TR match The corresponding matching trajectory set S match , S match Contains all matches to TR match TR valid , S ref Each TR in ref Each has its own matching trajectory set S match ;

[0013] S7, add a new S match TR valid Recorded as TR new , with the same S match Other TRs in valid For comparison, if TR new The first position point of is in the lane frame, and compared with other TR valid In terms of distance TR match The first position point is closest, then mark TR new is the main trajectory TR of a car that has just entered the intersection main , and TR new Remove S match ;

[0014] S8, regularly check the TR of each vehicle main , if a car's TR main In the preset time range T expire If there is no updated position point within , the connection matching process will be entered. Among them, the trajectory TR is a series of position coordinate points formed by the real-time position of the same vehicle detected by the same radar in chronological order. The position data output by the radar detection contains the detection target ID. Ideally, when the same vehicle passes through the intersection once, the detection target ID in each position data output by the same radar is the same. Main trajectory TR mainIt refers to the output of the latest position point in the main trajectory as the current position point of the moving vehicle. The radar is specifically a traffic radar microwave detector, which is installed at a certain entrance of the intersection and is used to detect the real-time position information of vehicles in the middle and opposite entrances of the intersection at fixed periods. The present invention discloses a method for matching and fusion of multi-radar detection trajectories at intersections in the field of intelligent transportation. By matching the actual trajectory detected by the radar with a preset reference trajectory, multiple trajectory fragments of the same vehicle detected within the same radar and between different radars are matched and fused in real time, thereby outputting the complete driving trajectory of the same vehicle passing through the intersection.

[0015] As an example, the vehicle driving reference trajectory TR of each lane in step S4 is ref The method for obtaining includes the following steps:

[0016] S41, prepare all radar detection trajectory data of the intersection within a period of time T to form a radar detection trajectory dataset S;

[0017] S42, filter S according to the number of real-time position points contained in each trajectory, and remove those that are less than or equal to the position point threshold τ filter After the trajectory of , we can get S′;

[0018] S43, the filtered radar detection trajectory dataset S′ is divided according to the lane entry frames of the intersection and classified into the lane entry frames of the intersection. The last intersecting lane entry frame is taken as the lane entry to which the radar detection trajectory belongs. After the classification is completed, a trajectory set S′ is associated with each lane entry i. i ;

[0019] S44: After completing the lane entry trajectory division, the remaining radar detection trajectories that do not belong to any lane entry are divided according to the exit frames of the intersection. The last intersecting exit frame is taken as the exit to which the radar detection trajectory belongs. After the classification is completed, a trajectory set S′ is associated with each exit o. o ;

[0020] S45: traverse all the entry lanes of the intersection, determine the flow type of each entry lane i, and select the destination exit trajectory set S′ of the entry lane according to the flow type. o , denoted as S′ dst； For the lane entry trajectory set S′ i Each trajectory TR i , calculate TR i and the destination exit trajectory set S′ dst Each trajectory TR o The degree of connection between link；

[0021] S46, θ link The calculation method is as follows: take TRo The first position coordinate point P o1 , traverse TR i All position coordinate points, select the distance P o1 The nearest position coordinate point P i1 , then P i1 and P o1 As the starting point, traverse TR in chronological order i and TR o points until TR i or TR o The last position coordinate point, a total of n points, to build a point pair (P i1 , P o1 ), (P i2 , P o2 ),……,(P in , P on )Set S pair ,but

[0022]

[0023] In the formula, ||(P ik , P ok )|| is P ik With P ok The distance metric between them, α is the weight coefficient, n is S pair The number of point pairs in the set;

[0024] S47, for each entry lane i, if it is a single flow direction, that is, there is only one destination exit, then select S′ i and S′ dst Between θ link The highest pair of trajectories (TR imax , TR omax ) as the reference trajectory for the lane; if it is a mixed flow, that is, there is more than one destination exit, then select S′ i S′ corresponding to each destination exit dst Internal θ link The highest pair of trajectories (TR imax , TR omax ) as the reference trajectory set for the lane; the total reference trajectory set for all lanes is denoted as S ref ;

[0025] S48, S ref Each reference trajectory in the set is a pair of trajectories (TR imax , TR omax ), need to be fused into one trajectory to obtain the reference trajectory TR ref ;

[0026] S49: Store the reference trajectory or reference trajectory set associated with each lane i at the intersection for subsequent use. The flow direction types in step S45 include left turn, right turn, straight ahead, left turn and straight ahead, and right turn and straight ahead.

[0027] As an example, the correlation matching degree θ in step S5 is rel The calculation method is: take TR valid The first position coordinate point P v1 , traverse TR ref All position coordinate points, select the distance P v1 The nearest position coordinate point P r1 , then P v1 and P r1 As the starting point, traverse TR in chronological order valid and TR ref points until TR valid or TR ref The last position coordinate point, a total of m points, to build a point pair (P v1 , P r1 ), (P v2 , P r2 ),……,(P vm , P rm ) set, then

[0028]

[0029] In the formula, ||(P vk , P rk )|| is P vk With P rk The distance metric between them, m is the number of point pairs.

[0030] Preferably, the connection matching process in step S8 is specifically as follows:

[0031] S81, if the interrupted TR main The last position point is within the exit box and is within the distance of TR match The last position point of the vehicle is closest, then the vehicle has gone out of the detection range of each radar, and the corresponding TR main The connection matching process is completed;

[0032] S82, otherwise, calculate the interrupted TR main With the previous S match Other TR valid The degree of connection between link , choose θ link Highest TR valid As a new TR main and move it out of S match, completing the connection and matching process.

[0033] As an example, the fusion method in step S48 is: omax Delete all the position coordinate points belonging to S pair The coincident coordinate points {P o1 , P o2 ,...P on}, the remaining coordinate points retain their original order and are spliced ​​in TR imax At the end of the point sequence, a complete reference trajectory TR is formed ref .

[0034] Preferably, the detection range of the radar device in step S1 can cover the middle of the intersection and the opposite entrance area, and the union of the detection ranges of all radar devices can cover the entire area of ​​the intersection and each entrance.

[0035] Preferably, the area range in step S2 is for each lane entry of each entrance, and the lane entry frame is calibrated in the latitude and longitude coordinate system or the local coordinate system of the intersection; the lane entry frame starts from the stop line of the lane and extends backward along the lane marking lines on both sides to the farthest distance that the radar can detect.

[0036] Preferably, the area range in step S3 is an exit frame calibrated for each exit in a latitude and longitude coordinate system or a local intersection coordinate system. The exit frame starts from the starting position of the exit and extends forward along the exit markings on both sides to the maximum distance detectable by the radar. The exit is no longer subdivided by lane, but the entire lane area of ​​the exit is regarded as an exit frame.

[0037] Therefore, the present invention has the following beneficial effects:

[0038] 1. A multi-radar trajectory matching and fusion method for intersections. This method matches the actual trajectory detected by radar with a preset reference trajectory. It then performs real-time matching and fusion of multiple trajectory fragments of the same vehicle detected within the same radar and between different radars. This method outputs the complete trajectory of the same vehicle passing through the intersection, achieving excellent real-time matching and fusion results.

[0039] 2. The proposed method for matching and fusion of multi-radar detection trajectories at intersections offers the advantages of simplicity, high accuracy, fast matching, strong real-time performance, and minimal computational effort. This low computational effort makes it suitable for deployment at resource-constrained roadside edge computing nodes, shortening the distance between data and computation and avoiding network transmission delays associated with cloud computing. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a schematic diagram of the detection range and area calibration of the intersection radar equipment of the present invention;

[0041] Figure 2 This is a schematic diagram of the reference trajectory of the left turn flow direction of the south entrance lane 1 of the present invention;

[0042] Figure 3 This is a schematic diagram of the track connection and matching effect of multiple radar detections at an intersection according to the present invention;

[0043] In the figure, 1, the left turn flow into lane 1 is directed to the reference track 2, the first vehicle A trajectory segment 3, the second vehicle A trajectory segment 4, and the third vehicle A trajectory segment. DETAILED DESCRIPTION

[0044] Example

[0045] This embodiment proposes a method for matching and fusing multi-radar detection trajectories at an intersection, including the following steps:

[0046] S1, install radar equipment near the stop line at each entrance of the intersection; refer to Figure 1 , so that the detection range of the radar device can cover the middle of the intersection and the opposite entrance area, and the union of the detection ranges of all radar devices can cover the entire area of ​​the intersection and each entrance;

[0047] S2, calibrate the area range corresponding to each lane entry at the intersection; for each lane entry at each entrance, calibrate the lane entry area polygonal box in the latitude and longitude coordinate system or the local coordinate system of the intersection, which is the lane entry box. The lane entry box starts from the stop line of the lane and extends backward along the lane markings on both sides to the farthest distance that the radar can detect, referring to Figure 1 Lane entry frame 1 and lane entry frame 2;

[0048] S3, calibrate the area range corresponding to each exit of the intersection; for each exit, calibrate the exit area polygonal box in the latitude and longitude coordinate system or the local coordinate system of the intersection, which is the exit box. The exit box starts from the starting position of the exit and extends forward along the exit markings on both sides to the farthest distance that the radar can detect. For the exit, it is no longer subdivided by lane, but the entire lane area of ​​the exit is regarded as an exit box, with reference to Figure 1 The exit box 1;

[0049] S4, obtaining the vehicle driving reference trajectory TR of each lane at the intersection ref；

[0050] S41, prepare all radar detection trajectory data of the intersection within a period of time T to form a radar detection trajectory dataset S;

[0051] S42, filter S according to the number of real-time position points contained in each trajectory, and remove those that are less than or equal to the position point threshold τ filterAfter filtering the trajectory, we can get S′; ensure that each detection trajectory after filtering has good quality. filter Configurable, typically 10;

[0052] S43, the filtered radar detection trajectory dataset S′ is divided according to the lane entry frames of the intersection and classified into the lane entry frames of the intersection. The last intersecting lane entry frame is taken as the lane entry to which the radar detection trajectory belongs. After the classification is completed, a trajectory set S′ is associated with each lane entry i. i ;

[0053] S44: After completing the lane entry trajectory division, the remaining radar detection trajectories that do not belong to any lane entry are divided according to the exit frames of the intersection. The last intersecting exit frame is taken as the exit to which the radar detection trajectory belongs. After the classification is completed, a trajectory set S′ is associated with each exit o. o ;

[0054] S45: traverse all the entry lanes of the intersection, determine the flow type of each entry lane i, and select the destination exit trajectory set S′ of the entry lane according to the flow type. o , denoted as S′ dst； For the lane entry trajectory set S′ i Each trajectory TR i , calculate TR i and the destination exit trajectory set S′ dst Each trajectory TR o The degree of connection between link ;

[0055] S46, θ link The calculation method is as follows: take TR o The first position coordinate point P o1 , traverse TR i All position coordinate points, select the distance P o1 The nearest position coordinate point P i1 , then P i1 and P o1 As the starting point, traverse TR in chronological order i and TR o points until TR i or TR o The last position coordinate point, a total of n points, to build a point pair (P i1 , P o1 ), (P i2 , P o2 ),……,(P in , P on )Set S pair ,but

[0056]

[0057] In the formula, ||(P ik , P ok )|| is P ik With P ok The distance metric between them, α is the weight coefficient, n is S pair The number of point pairs in the set; S47, for each entry lane i, if it is a single flow direction, that is, there is only one destination exit, then select S′ i and S′ dst Between θ link The highest pair of trajectories (TR imax , TR omax ) as the reference trajectory for the lane; if it is a mixed flow, that is, there is more than one destination exit, then select S′ i S′ corresponding to each destination exit dst Internal θ link The highest pair of trajectories (TR imax , TR omax ) as the reference trajectory set for the lane; the total reference trajectory set for all lanes is denoted as S ref ;

[0058] S48, S ref Each reference trajectory in the set is a pair of trajectories (TR imax , TR omax ), need to be fused into one trajectory to obtain the reference trajectory TR ref ; The fusion method is: TR omax Delete all the position coordinate points belonging to S pair The coincident coordinate points {P o1 , P o2 ,…P on}, the remaining coordinate points retain their original order and are spliced ​​in TR imax At the end of the point sequence, a complete reference trajectory TR is formed ref ;

[0059] S49, store the reference trajectory or reference trajectory set associated with each lane i at the intersection for subsequent use. Figure 2 Incoming lane 1 turns left and flows to reference track 1;

[0060] S5, the radar device detects and updates each track of the position point in real time as TR cur , if TR cur The number of location points contained is greater than the threshold τ valid , then mark the trajectory as a valid trajectory TR valid , calculate TR valid With S refEach reference trajectory TR in ref The correlation matching degree θ between rel ; Correlation matching degree θ rel The calculation method is: take TR valid The first position coordinate point P v1 , traverse TR ref All position coordinate points, select the distance P v1 The nearest position coordinate point P r1 , then P v1 and P r1 As the starting point, traverse TR in chronological order valid and TR ref points until TR valid or TR ref The last position coordinate point, a total of m points, to build a point pair (P v1 , P r1 ), (P v2 , P r2 ),……,(P vm , P rm ) set, then

[0061]

[0062] In the formula, ||(P vk , P rk )|| is P vk With P rk The distance metric between them, m is the number of point pairs;

[0063] S6, select the correlation matching degree θ rel Highest TR ref Recorded as TR match , the above TR valid Join TR match The corresponding matching trajectory set S match , S match Contains all matches to TR match TR valid , S ref Each TR in ref Each has its own matching trajectory set S match ;

[0064] S7, add a new S match TR valid Recorded as TR new , with the same S match Other TRs in valid For comparison, if TR new The first position point of is in the lane frame, and compared with other TR valid In terms of distance TRmatch The first position point is closest, then mark TR new is the main trajectory TR of a car that has just entered the intersection main , and TR new Remove S match；

[0065] S8, regularly check the TR of each vehicle main , if a car's TR main In the preset time range T expire If there is no updated location point within , the connection matching process will begin;

[0066] S81, if the interrupted TR main The last position point is within the exit box and is within the distance of TR match The last position point of the vehicle is closest, then the vehicle has gone out of the detection range of each radar, and the corresponding TR main The connection matching process is completed;

[0067] S82, otherwise, calculate the interrupted TR main With the previous S match Other TR valid The degree of connection between link , choose θ link Highest TR valid As a new TR main and move it out of S match , complete this connection matching process, refer to Figure 3 The first vehicle A trajectory segment 2, the second vehicle A trajectory segment 3 and the third vehicle A trajectory segment 4.

[0068] refer to Figures 1 to 3 The radar device in the step is a traffic radar microwave detector, which is installed at a certain entrance of the intersection and is used to detect the real-time position information of vehicles in the middle and opposite entrances of the intersection every fixed period.

[0069] A trajectory (TR) is a series of position coordinates connected in chronological order from the real-time position of the same vehicle detected by the same radar. The position data output by the radar includes the detected target ID. Ideally, during a single vehicle pass through the intersection, the detected target ID is the same in all the position data output by the same radar.

[0070] Main track TR main The latest position point in the main trajectory is output as the current position point of the traveling vehicle.

[0071] In step S41, the length of T can be adjusted based on the size of S, ensuring that S can successfully complete subsequent reference trajectory acquisition. Typically, T can be selected from 0:00 to 24:00 on any given day. In step S43, considering the possibility of a vehicle changing lanes at an entrance, a radar-detected trajectory may intersect with multiple entry lane frames. The last intersecting entry lane frame is used as the entry lane to which the radar-detected trajectory belongs for classification. In step S44, trajectories that do not belong to any entry lane or any exit are not subsequently used.

[0072] The method for matching and fusion of multiple radar detection trajectories at intersections proposed in the present invention is used to match and connect vehicle trajectories detected by multiple radars at different entrances to an intersection, outputting the complete trajectory of the same vehicle passing through the same intersection in a single pass, thereby solving the problem that the detection range of a single radar cannot cover the entire area of ​​the intersection, as well as the problem that the same radar detection trajectory frequently jumps and different radar detection trajectories are difficult to connect in real time in practical applications. The method proposed in the present invention has the advantages of simple implementation, high accuracy, fast matching, strong real-time performance, and low computational complexity. Due to the low computational complexity, it is suitable for deployment on resource-constrained roadside edge computing nodes, shortening the distance between data and computing, and avoiding network transmission delays in cloud computing, thereby achieving good real-time matching and fusion effects.

[0073] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for matching and fusion of multi-radar detection trajectories at intersections, characterized by: include: S1, install radar equipment near the stop line at each entrance of the intersection; S2, calibrating the area corresponding to each lane entry at the intersection and determining the lane entry frame; S3, calibrating the area corresponding to each exit of the intersection and determining the exit frame; S4, obtaining the vehicle driving reference trajectory TR of each lane at the intersection ref ; S5, the radar device detects and updates each track of the position point in real time as TR cur , if TR cur The number of location points contained is greater than the threshold τ valid , then mark the trajectory as a valid trajectory TR valid , calculate TR valid and all reference trajectories S for all lanes ref Each reference trajectory TR in ref The correlation matching degree θ between rel ; S6, select the correlation matching degree θ rel Highest TR ref Recorded as TR match , the above TR valid Join TR match The corresponding matching trajectory set S match , S match Contains all matches to TR match TR valid , S ref Each TR in ref Each has its own matching trajectory set S match ; S7, add a new S match TR valid Recorded as TR new , with the same S match Other TRs in valid For comparison, if TR new The first position point of is in the lane frame, and compared with other TR valid In terms of distance TR match The first position point is closest, then mark TR new is the main trajectory TR of a car that has just entered the intersection main , and TR new Remove S match ; S8, regularly check the TR of each vehicle main , if a car's TR main In the preset time range T expire There is no updated position point within TR main Interrupt and enter the connection matching process, which includes: If the interrupted TR main The last position point is within the exit box and is within the distance of TR match The last position point of the vehicle is closest, then the vehicle has gone out of the detection range of each radar, corresponding to TR main The connection matching process ends.

2. The method for matching and fusing multi-radar detection trajectories at an intersection according to claim 1, wherein the vehicle driving reference trajectory TR of each lane in step S4 is ref The method for obtaining includes the following steps: S41, prepare all radar detection trajectory data of the intersection within a period of time T to form a radar detection trajectory dataset S; S42, filter S according to the number of real-time position points contained in each trajectory, and remove those that are less than or equal to the position point threshold τ filter After the trajectory of S ′ ; S43, the filtered radar detection trajectory dataset S ′ The lanes are divided into different lanes at the intersection. Considering the possibility of vehicles changing lanes at the entrance, a radar detection track may intersect with multiple lane frames. The last intersecting lane frame is taken as the lane to which the radar detection track belongs. After the classification is completed, a trajectory set S is associated with each lane i. i ′ ; S44, after completing the lane entry trajectory division, the remaining radar detection trajectories that do not belong to any lane entry are divided according to the exit frames of the intersection. A radar detection trajectory may intersect with multiple exit frames, and the exit frame that intersects the last one is taken as the exit to which the radar detection trajectory belongs. After the classification is completed, a trajectory set S is associated with each exit o. ′ o ; S45, traverse all the entry lanes of the intersection, determine the flow type of each entry lane i, and select the destination exit trajectory set S of the entry lane according to the flow type. ′ o , denoted as S ′ dst ;for Lane entry trajectory set S i ′ Each trajectory TR i , calculate TR i and the destination exit trajectory set S ′ dst Each trajectory TR o The degree of connection between link ; S46, θ link The calculation method is as follows: take TR o The first position coordinate point P o1 , traverse TR i All position coordinate points, select the distance P o1 The nearest position coordinate point P i1 , then P i1 and P o1 As the starting point, traverse TR in chronological order i and TR o points until TR i or TR o The last position coordinate point, a total of n points, to build a point pair (P i1 , P o1 ), (P i2 , P o2 ),……,(P in , P on )Set S pair ,but In the formula, ||(P ik ,P ok )|| is P ik With P ok The distance metric between them, α is the weight coefficient, n is S pair The number of point pairs in the set; S47, for each lane i, if it is a single flow direction, that is, there is only one destination exit, then select S i ′ With S ′ dst Between θ link The highest pair of trajectories (TR imax ,TR omax ) as the reference trajectory for the lane entry; If the flow is mixed, that is, there is more than one destination outlet, select S i ′ S corresponding to each destination exit ′ dst Internal θ link The highest pair of trajectories (TR imax ,TR omax ) as the reference trajectory set for the lane; the total reference trajectory set for all lanes is denoted as S ref ; S48, S ref Each reference trajectory in the set is a pair of trajectories (TR imax ,TR omax ), need to be fused into one trajectory to obtain the reference trajectory TR ref ; S49: Store the reference trajectory or reference trajectory set associated with each incoming lane i at the intersection for subsequent use.

3. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 1 is characterized in that: The correlation matching degree θ in step S5 rel The calculation method is: take TR valid The first position coordinate point P v1 , traverse TR ref All position coordinate points, select the distance P v1 The nearest position coordinate point P r1 , then P v1 and P r1 As the starting point, traverse TR in chronological order valid and TR ref points until TR valid or TR ref The last position coordinate point, a total of m points, to build a point pair (P v1 , P r1 ), (P v2 , P r2 ),……,(P vm , P rm ) set, then In the formula, ||(P vk ,P rk )|| is P vk With P rk The distance metric between them, m is the number of point pairs.

4. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 1, wherein: The connection matching process in step S8 further includes: If the interrupted TR main The last position point is not in the exit box, and the TR is interrupted. main With the previous S match Other TR valid The degree of connection between link , choose θ link Highest TR valid As a new TR main and move it out of S match , completing the connection and matching process.

5. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 2 is characterized in that: The fusion method in step S48 is: omax Delete all the position coordinate points belonging to S pair The coincident coordinate points {P o1 ,P o2 ,…P on }, the remaining coordinate points retain their original order and are spliced ​​in TR imax At the end of the point sequence, a complete reference trajectory TR is formed ref .

6. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 1, characterized in that: The detection range of the radar device in step S1 can cover the middle of the intersection and the opposite entrance area, and the union of the detection ranges of all radar devices can cover the entire area of ​​the intersection and each entrance.

7. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 1, characterized in that: The area range in step S2 is for each lane entry at each entrance, and the lane entry frame is calibrated in the latitude and longitude coordinate system or the local coordinate system of the intersection; the lane entry frame starts from the stop line of the lane and extends backward along the lane marking lines on both sides to the farthest distance that the radar can detect.

8. The method for matching and fusion of multiple radar detection trajectories at intersections according to claim 1, characterized in that: The area range in step S3 is to calibrate the exit frame in the latitude and longitude coordinate system or the local coordinate system of the intersection for each exit. The exit frame starts from the starting position of the exit and extends forward along the exit markings on both sides to the farthest distance that the radar can detect.

Images