All-time-space vehicle trajectory reconstruction method, system and readable storage medium

Abstract

The application relates to the technical field of data processing, in particular to a full-time-space vehicle trajectory reconstruction method, model, system and readable storage medium. Known data collected by a radar-camera integrated machine is used to form a single-source and bidirectional fragmented detector data environment, then high-resolution transverse and longitudinal speed data of vehicles in a blind area range is predicted, and finally, based on the high-resolution transverse and longitudinal speed data, transverse and longitudinal candidate trajectories are reconstructed from the upstream and downstream respectively, and the candidate trajectories are fused, so that a full-time-space individual vehicle high-time-space-resolution transverse and longitudinal trajectory is obtained. The application aims to solve the problem of predicting a full-time-space vehicle trajectory by using single-source detector data.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, model, system, and readable storage medium for reconstructing vehicle trajectories in all time and space. Background Technology

[0002] High-resolution, all-time-space individual vehicle lateral and longitudinal micro-motion trajectory data accurately records the micro-motion characteristics of vehicles in continuous time and space dimensions. It can accurately characterize the dynamic interaction patterns between individual vehicles and traffic flow groups, and is a fundamental data source for analyzing complex multi-lane traffic systems.

[0003] Currently, on highways, the monitoring range of integrated radar-visual cameras is approximately 400 meters, capable of collecting multi-dimensional trajectory parameters such as vehicle ID, timestamp, longitudinal and lateral position, and speed within its monitoring range. However, fixed-point detectors are constrained by practical factors such as equipment cost, installation conditions, and privacy protection, resulting in low actual coverage; the distance between two adjacent fixed-point detectors typically exceeds 500 meters. This low coverage characteristic leads to significant detection blind spots on highways, causing integrated radar-visual cameras to acquire only fragmented vehicle trajectory data, which is significantly different from full-time and spatial vehicle lateral and longitudinal trajectory data.

[0004] In view of this, this application proposes a full-time and space-time vehicle trajectory reconstruction method, which aims to realize full-time and space-time vehicle trajectory reconstruction for highways with detection blind spots. Summary of the Invention

[0005] The main purpose of this application is to provide a method for reconstructing vehicle trajectories in all time and space, which aims to solve the problem of predicting vehicle trajectories in all time and space using single-source detector data.

[0006] To achieve the above objectives, this application provides a method for reconstructing vehicle trajectories across all times and spaces, the method comprising:

[0007] S10, acquire the first vehicle driving data collected by the upstream integrated radar vision machine and the second vehicle driving data collected by the downstream integrated radar vision machine;

[0008] S20, determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data, wherein the set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle;

[0009] S30, the set of horizontal and vertical positions is input into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, and a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0010] S40, the first longitudinal prediction trajectory and the second longitudinal prediction trajectory are weighted and summed to obtain a longitudinal fusion trajectory, and the first lateral prediction trajectory and the second lateral prediction trajectory are weighted and summed to obtain a lateral fusion trajectory, wherein the weight sum of the first longitudinal prediction trajectory and the second longitudinal prediction trajectory is 1, and the weight sum of the first lateral prediction trajectory and the second lateral prediction trajectory is 1.

[0011] Optionally, the mathematical expression for the lateral fusion trajectory is:

[0012]

[0013] in, , , .

[0014] In the formula, For horizontal fusion trajectory, The weight value corresponding to the first horizontal predicted trajectory. This is the first lateral predicted trajectory. This is the second lateral predicted trajectory. The weight value corresponding to the second horizontal predicted trajectory;

[0015] The mathematical expression for the longitudinal fusion trajectory is:

[0016]

[0017] in, , , .

[0018] In the formula, For vertical fusion trajectory, The weight value corresponding to the first longitudinal predicted trajectory. This is the first longitudinal predicted trajectory. This is the second longitudinal predicted trajectory. The weight value is the value corresponding to the second longitudinal predicted trajectory.

[0019] Optionally, the expression for the first lateral predicted trajectory is:

[0020]

[0021] The expression for the second lateral predicted trajectory is:

[0022]

[0023] In the formula, Let u be the x-coordinate of the position u of the target vehicle when it leaves the upstream radar-vision integrated machine. The x-coordinate of the position d of the target vehicle when it leaves the downstream radar-visual integrated machine; t i This represents the i-th time point when the vehicle is within the blind spot. Let represent the predicted lateral speed of the vehicle at time point i, and n represent the total number of time points within the blind zone for that vehicle.

[0024] Optionally, the expression for the first longitudinal predicted trajectory is:

[0025]

[0026] The expression for the second longitudinal predicted trajectory is:

[0027]

[0028] In the formula, Let u be the ordinate of the position u of the target vehicle when it leaves the upstream radar-visual integrated machine. The vertical coordinate of position d when the target vehicle leaves the downstream radar-visual integrated machine; t i This represents the i-th time point when the vehicle is within the blind spot. Let represent the predicted longitudinal speed of the vehicle at the i-th time point, and n represent the total number of time points within the blind zone of the vehicle.

[0029] Optionally, the preset model satisfies the following steps during execution:

[0030] S21, Extract the multidimensional spatiotemporal features from the set of horizontal and vertical positions;

[0031] S22, Discretize the multidimensional spatiotemporal features;

[0032] S23, Introducing step size As a learnable parameter, the discretized multidimensional spatiotemporal features of the input are processed according to the step size. The size of the sample is used to obtain a discretized input velocity sequence. A selective mechanism is then used to transform the discrete parameters in the discretized input velocity sequence into dependent input parameters, wherein the step size is... The learning process is adjusted based on the speed of the discretized multidimensional spatiotemporal features.

[0033] Optionally, S21 includes:

[0034] S211, uncover the hidden state of the previous time step t-1. Multidimensional spatiotemporal characteristics of vehicle lateral / longitudinal velocity The output shows the current state. , the current state and multidimensional spatiotemporal features Mapping to output response The multidimensional spatiotemporal feature sequence as output:

[0035]

[0036]

[0037] in, , These represent multidimensional time series data, respectively, as input and output. Represents the current state of the transportation system, and indicates the relationship with... The result of combined action; matrix Used to adjust the dynamic state changes of the observation equation; and These are the input projection matrix and the output projection matrix, respectively; Let R be the residual, and R be a real number.

[0038] Optionally, S22 includes:

[0039] S221, convert matrices A and B into discrete parameters. and The sequence value is maintained each time a discrete signal is received until the next new discrete signal is received:

[0040]

[0041]

[0042] in, and The matrix is ​​respectively and The parameters after discretization, Δ t A is the increment of the state matrix, Δ t B represents the increment of the input matrix, and I represents the identity matrix.

[0043] Furthermore, to achieve the above objectives, this application also provides a spatiotemporal vehicle trajectory reconstruction model, which includes:

[0044] The data acquisition module is used to acquire the first vehicle driving data collected by the upstream radar-view integrated machine and the second vehicle driving data collected by the downstream radar-view integrated machine;

[0045] The lateral and longitudinal position extraction module is used to determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data. The set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle.

[0046] The horizontal and vertical trajectory prediction module is used to input the horizontal and vertical position set into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, as well as a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0047] The trajectory fusion module is used to perform a weighted sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory to obtain a longitudinal fused trajectory, and to perform a weighted sum of the first lateral predicted trajectory and the second lateral predicted trajectory to obtain a lateral fused trajectory, wherein the weight sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory is 1, and the weight sum of the first lateral predicted trajectory and the second lateral predicted trajectory is 1.

[0048] In addition, to achieve the above objectives, this application also provides a computer system, the computer system comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the all-time-space vehicle trajectory reconstruction method as described in any of the preceding claims.

[0049] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the all-time-space vehicle trajectory reconstruction method as described in any of the preceding claims.

[0050] This application has at least the following beneficial effects:

[0051] 1. Analyze the characteristics of data collected by upstream and downstream integrated radar and vision devices, comprehensively consider the time sequence of bidirectional location data collected by single-source detectors, and unify the ID of individual vehicles when they are collected by upstream and downstream integrated radar and vision devices to obtain known information about the vehicle's location at both upstream and downstream locations, thus expanding the universality.

[0052] 2. By deeply mining the spatiotemporal correlation characteristics of data collected by upstream and downstream integrated radar-visual systems, a "forward-backward" bidirectional learning prediction model is proposed to construct a bidirectional collaborative perception mechanism. This application not only achieves high-resolution prediction of vehicle lateral and longitudinal speeds within the blind zone and full-time spatiotemporal reconstruction of individual vehicle lateral and longitudinal positions, but also effectively solves the problem of result instability in single-path prediction through bidirectional information complementarity.

[0053] 3. Using fragmented vehicle trajectory data collected by a two-way radar-visual integrated machine, and employing a deep learning framework with known data, the long-range spatiotemporal dependencies between data are effectively captured through state space modeling, and the lateral and longitudinal velocities of all vehicles within the blind zone are predicted. Based on the predicted velocities, the upstream and downstream candidate trajectory fusion method is used to reconstruct the high spatiotemporal resolution lateral and longitudinal positions of individual vehicles in all time and space, providing a general and practical deep learning prediction method for perceiving the motion state of vehicles within the blind zone of the detector in complex highway environments.

[0054] 4. The radar-visual integrated machine can report detection data in real time. This invention can be used in both online and offline environments, thereby expanding its application in traffic information prediction. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating the all-time and all-space vehicle trajectory reconstruction method according to an embodiment of this application;

[0056] Figure 2 This is a framework diagram of the Mamba model used in this application to predict the lateral and longitudinal velocities of a vehicle.

[0057] Figure 3 This is a schematic diagram of the highway environment involved in the embodiments of this application;

[0058] Figure 4 This is a partial prediction diagram of the lateral and longitudinal speeds of vehicles in a real road environment involved in the embodiments of this application;

[0059] Figure 5 This is a comparison diagram of the actual and predicted longitudinal trajectories of vehicles within the blind zone area involved in the embodiments of this application;

[0060] Figure 6 This is a comparison diagram of the actual and predicted longitudinal trajectories of vehicles within another blind zone area involved in the embodiments of this application;

[0061] Figure 7 This is a comparison diagram of the actual and predicted longitudinal trajectories of vehicles within another blind zone area involved in the embodiments of this application;

[0062] Figure 8 This is a comparison diagram of the actual and predicted longitudinal trajectories of vehicles within another blind zone area involved in the embodiments of this application;

[0063] Figure 9 This is a contour map comparing the actual and predicted longitudinal speeds of vehicles within the blind spot area involved in the embodiments of this application;

[0064] Figure 10 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0065] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0066] To better understand the above technical solutions, exemplary embodiments of this disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art.

[0067] First Embodiment

[0068] Reference Figure 1 This embodiment provides a method for reconstructing vehicle trajectories across all times and spaces, the method comprising the following steps:

[0069] S10, acquire the first vehicle driving data collected by the upstream integrated radar vision machine and the second vehicle driving data collected by the downstream integrated radar vision machine;

[0070] In this embodiment, the radar-visual integrated machines installed on adjacent upstream and downstream sections of expressways are used to collect multi-dimensional vehicle driving data of individual vehicles within their detection range, including but not limited to ID, timestamp, and horizontal and vertical position and speed data, forming single-source, bidirectional, fragmented and low-coverage known data, and the area between the upstream and downstream radar-visual integrated machines is regarded as the blind zone.

[0071] S20, determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data, wherein the set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle;

[0072] In this embodiment, the set of lateral and longitudinal positions of the target vehicle that needs trajectory reconstruction is determined from the original vehicle driving data when it passes the upstream and downstream integrated radar-viewing machines. This set includes the target vehicle's lateral coordinates, longitudinal coordinates, lateral velocity, and longitudinal velocity. Subsequent trajectory prediction is then performed based on these parameters.

[0073] As some alternative implementation methods, since the time format of the detection data from adjacent integrated radar-view cameras is not uniform, it is necessary to unify the collected data according to the time order. The time unit of the data collected by the integrated radar-view cameras is set to a frame, and the position unit is set to feet, as shown in the following formula:

[0074]

[0075]

[0076] in, It is time, measured in frames. and These are the horizontal and vertical distances, respectively, in feet; and These are the horizontal and vertical speeds, respectively, in feet per second; 'a' is the conversion factor between frames and seconds.

[0077] In this embodiment, a = 0.1.

[0078] In the process of predicting vehicle lateral and longitudinal speeds, blind zone ranges are determined using integrated radar-visual cameras installed at upstream and downstream locations. The time range of the blind zone is determined by the time each vehicle leaves the upstream radar-visual camera's acquisition range and enters the downstream radar-visual camera's acquisition range, while the spatial range of the blind zone is determined by the interval between the upstream and downstream radar-visual cameras. The vehicle data collected by the integrated radar-visual cameras includes vehicle ID, timestamp, lateral and longitudinal speeds, and position, i.e. ( , , , , , ), of which, the raw data when the vehicle passes the upstream radar-visual integrated machine, i.e. ( , , , , , (), , , , , , (), , , , , , ), , , , , and These are, respectively, the ID, timestamp, lateral speed, longitudinal speed, lateral position, and longitudinal position of the i-th vehicle within the range collected by the upstream radar-view integrated machine; and the raw data of the vehicle when it passes the downstream radar-view integrated machine, i.e. ( , , , , , (), , , , , , (), , , , , , ), , , , , and These are the ID, timestamp, lateral speed, longitudinal speed, lateral position, and longitudinal position of the i-th vehicle within the range of the downstream radar-visual integrated machine.

[0079] S30, the set of horizontal and vertical positions is input into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, and a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0080] In this embodiment, after obtaining the set of horizontal and vertical positions, the set of horizontal and vertical positions is input into a preset model, which is a trained deep learning model. The preset model must satisfy the following requirements: it must be able to generate a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, and a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0081] To achieve the above, this embodiment describes the setup architecture or implementation steps of the preset model:

[0082] S21, Extract the multidimensional spatiotemporal features from the set of horizontal and vertical positions;

[0083] S22, Discretize the multidimensional spatiotemporal features;

[0084] S23, Introducing step size As a learnable parameter, the discretized multidimensional spatiotemporal features of the input are processed according to the step size. The size of the sample is used to obtain a discretized input velocity sequence. A selective mechanism is then used to transform the discrete parameters in the discretized input velocity sequence into dependent input parameters, wherein the step size is... During the learning process, adjustments are made based on the speed of the discretized multidimensional spatiotemporal features.

[0085] Furthermore, and optionally, in S21, the lateral and longitudinal velocities of individual vehicles are modeled temporally by mining the hidden state of the previous time step t-1. Vehicle lateral and longitudinal velocity characteristics Output the current state Then through and Mapping to output response :

[0086] S211, uncover the hidden state of the previous time step t-1. Multidimensional spatiotemporal characteristics of vehicle lateral / longitudinal velocity The output shows the current state. , the current state and multidimensional spatiotemporal features Mapping to output response The multidimensional spatiotemporal feature sequence as output:

[0087]

[0088]

[0089] in, , These represent multidimensional time series data, respectively, as input and output. Represents the current state of the transportation system, and indicates the relationship with... The result of combined action; matrix Used to adjust the dynamic state changes of the observation equation; and These are the input projection matrix and the output projection matrix, respectively; Let R be the residual, and R be a real number.

[0090] Further and optionally, in S22, the continuous parameter sum is converted into a discrete parameter sum using a zero-order hold technique, maintaining the value of the sequence each time a discrete signal is received until the next new discrete signal is received. This discretization method not only ensures the stability of the computation but also preserves the time dependence on the continuous signal, i.e.:

[0091] S221, convert matrices A and B into discrete parameters. and The sequence value is maintained each time a discrete signal is received until the next new discrete signal is received:

[0092]

[0093]

[0094] in, and The matrix is ​​respectively and The parameters after discretization, Δ t A is the increment of the state matrix, Δ tB represents the increment of the input matrix, and I represents the identity matrix.

[0095] Furthermore, and optionally, in S23, a selective mechanism and a time-varying module are used to learn key information from the bidirectional data and extract local and global temporal features, thereby outputting high spatiotemporal resolution lateral and longitudinal velocity results for vehicles within the blind spot. This is achieved by introducing a new learnable parameter—step size. To determine the time to save this value, the input continuous velocity timing data is based on... The magnitude of the input velocity is sampled to obtain a discretized input velocity sequence; in addition, a selective mechanism is added to adjust the discrete parameters. and By transforming the parameters into input-dependent parameters, the parameters can be dynamically adjusted based on the input speed data during deep learning, resulting in better performance on long sequence modeling tasks.

[0096] For example, incorporating a selective mechanism and a time-varying module is represented as follows:

[0097]

[0098]

[0099] in, This represents the hidden state within time interval l; and These represent the horizontal and vertical velocity sequences of the input and output within time interval l, respectively.

[0100] Furthermore, and optionally, let the lateral and longitudinal positions of the upstream radar-visual integrated machine at the individual vehicle be respectively... and The horizontal and vertical positions of the downstream integrated radar screen are respectively and The lateral and longitudinal velocity sequences of vehicles within the blind spot area, predicted based on the preset model, are as follows: and Then we have:

[0101] The upstream and downstream lateral reconstruction trajectories of individual vehicles are predicted from upstream to downstream and from downstream to upstream, respectively:

[0102]

[0103]

[0104] In the formula, Let u be the x-coordinate of the position u of the target vehicle when it leaves the upstream radar-vision integrated machine. The x-coordinate of the position d of the target vehicle when it leaves the downstream radar-visual integrated machine; ti This represents the i-th time point when the vehicle is within the blind spot. Let represent the predicted lateral speed of the vehicle at time point i, and n represent the total number of time points within the blind zone for that vehicle.

[0105] The upstream and downstream longitudinal reconstruction trajectories of individual vehicles are predicted from upstream to downstream and from downstream to upstream, respectively, i.e.:

[0106]

[0107]

[0108] In the formula, Let u be the ordinate of the position u of the target vehicle when it leaves the upstream radar-visual integrated machine. The vertical coordinate of position d when the target vehicle leaves the downstream radar-visual integrated machine; t i This represents the i-th time point when the vehicle is within the blind spot. Let represent the predicted longitudinal speed of the vehicle at the i-th time point, and n represent the total number of time points within the blind zone of the vehicle.

[0109] S40, the first longitudinal prediction trajectory and the second longitudinal prediction trajectory are weighted and summed to obtain a longitudinal fusion trajectory, and the first lateral prediction trajectory and the second lateral prediction trajectory are weighted and summed to obtain a lateral fusion trajectory, wherein the weight sum of the first longitudinal prediction trajectory and the second longitudinal prediction trajectory is 1, and the weight sum of the first lateral prediction trajectory and the second lateral prediction trajectory is 1.

[0110] In this embodiment, after obtaining the trajectories in both the horizontal and vertical directions, there are two reconstructed trajectories for the individual vehicle in both directions throughout the entire time and space. Within the blind zone, the individual vehicle trajectory reconstruction is obtained based on velocity integrals. Therefore, a phenomenon may occur where the vehicle trajectory does not coincide with the actual trajectory when reaching the last time point. Figure 2 As shown. To overcome this phenomenon, different weights are assigned to the two trajectories in each direction. Regardless of the horizontal or vertical position, the weight of the trajectory predicted from upstream to downstream gradually changes from 1 to 0. Since the sum of the two weights is 1, the weight of the trajectory predicted from downstream to upstream gradually changes from 0 to 1.

[0111] Furthermore, and optionally, the mathematical expression for the lateral fusion trajectory is:

[0112]

[0113] in, , , .

[0114] In the formula, For horizontal fusion trajectory, The weight value corresponding to the first horizontal predicted trajectory. This is the first lateral predicted trajectory. This is the second lateral predicted trajectory. The weight value corresponding to the second horizontal predicted trajectory;

[0115] The mathematical expression for the longitudinal fusion trajectory is:

[0116]

[0117] in, , , .

[0118] In the formula, For vertical fusion trajectory, The weight value corresponding to the first longitudinal predicted trajectory. This is the first longitudinal predicted trajectory. This is the second longitudinal predicted trajectory. The weight value is the value corresponding to the second longitudinal predicted trajectory.

[0119] In some alternative implementations, The value is 10.

[0120] In the technical solution provided in this embodiment, the known data collected by the radar-vision integrated machine is used to form a single-source, bidirectional fragmented detector data environment. Then, the high-resolution lateral and longitudinal velocity data of the vehicle within the blind zone are predicted. Finally, based on the high-resolution lateral and longitudinal velocity data, the lateral and longitudinal candidate trajectories are reconstructed from upstream and downstream respectively, and the candidate trajectories are fused to obtain the high spatiotemporal resolution lateral and longitudinal trajectory of the individual vehicle in all time and space.

[0121] Second Embodiment

[0122] Based on the first embodiment, this embodiment provides an experimental method using highways from the NGSIM dataset as the experimental object. Two integrated radar-visual cameras are installed at locations 200 feet upstream and 1600 feet downstream, each with a detection range of 400 feet. The spatial interval of the blind zone between adjacent integrated radar-visual cameras is 1000 feet. Using single-source, fragmented vehicle data collected by the integrated radar-visual cameras as input, the high-resolution lateral and longitudinal velocities of vehicles within the blind zone are predicted. This allows for the reconstruction of the full-time, high-spatiotemporal resolution lateral and longitudinal trajectories of individual vehicles, resulting in the following... Figure 3 The diagram shows the environment of a highway section. In this embodiment, the preset model used is the Mamba model.

[0123] First, some of the original data required for this invention in the NGSIM dataset are shown in Table 1:

[0124] Table 1. Partial raw trajectory data from the NGSIM dataset

[0125]

[0126] The formulas for calculating the lateral and longitudinal velocities of vehicles collected by the radar-visual integrated machine are as follows:

[0127]

[0128]

[0129] After standardizing the data format and determining the time sequence of the collected data, the time range is determined by the time the vehicle leaves the upstream detection range and enters the downstream detection range, which, together with the spatial range of the radar-visual integrated machine, forms the blind zone range. As shown in Table 1, the spatial range of the blind zone in this embodiment is between 400 feet and 1400 feet, and the time range of the vehicle with ID=2 within the blind zone is between 103 frames and 321 frames.

[0130] Next, the state-space model in the Mamba model is used to capture the long-range temporal correlation between bidirectional known fragmented data. Then, a selection mechanism and time-varying module are used to learn the key information of the bidirectional known fragmented data and extract local and global temporal features, thereby predicting the high-precision lateral and longitudinal velocities of vehicles within the blind zone. The state dimension is 16, the convolution kernel width is 4, the expansion factor is 2, and the Adam optimizer is used. The prediction results for the lateral and longitudinal velocities of some vehicles are as follows: Figure 4 As shown.

[0131] Depend on Figure 4 As can be seen, the blue curve represents the actual lateral and longitudinal speeds of vehicles within the blind zone, while the red curve represents the predicted lateral and longitudinal speeds of vehicles within the blind zone. The method involved in this embodiment, by learning from the vehicle driving data collected by the upstream and downstream radar-visual integrated machines, can predict the high-resolution lateral and longitudinal vehicle speeds within the blind zone, and the predicted speed change trend is similar to the actual motion trend.

[0132] According to the steps in this embodiment, the high spatiotemporal resolution horizontal and vertical reconstructed trajectory results of individual vehicles within the blind zone are as follows: Figures 5 to 9 As shown. By Figures 5 to 8 As can be seen, the blue dashed line represents the upstream candidate trajectory, the red dashed line represents the downstream candidate trajectory, and the black solid line represents the vehicle's actual trajectory; the left side shows the longitudinal reconstructed trajectory corresponding to some vehicles, and the right side shows the lateral reconstructed trajectory corresponding to some vehicles; the colors of the lateral and longitudinal reconstructed trajectories represent the reconstruction error. Figure 9As can be seen from the longitudinal trajectory on the left, the method of this embodiment can reconstruct the longitudinal trajectory of the vehicle within the blind spot area with high accuracy, and when approaching the upstream and downstream integrated radar-visual equipment, the accuracy of the fused trajectory is approximately 0, basically coinciding with the actual trajectory; Figure 9 As can be seen from the right-hand lateral trajectory, the method in this embodiment can effectively reflect the lateral movement state of the vehicle within the blind spot. Furthermore, by utilizing the upstream and downstream candidate trajectory fusion method, the lateral trajectory reconstruction error of the vehicle is significantly reduced, and the start and end points are identical to the true values, thus enabling better prediction of the vehicle's lane-changing spatiotemporal points. Figure 9 It can be seen that this embodiment can reproduce the trend of speed decrease and the congestion fluctuation path well, indicating that the prediction of this model on large-scale velocity field structure (such as congestion wave propagation, acceleration and deceleration sections) is basically accurate, and the fluctuation pattern and transmission trend are highly consistent.

[0133] Since the predicted lateral and longitudinal trajectory data are microscopic data, the mean absolute error (MAE) and root mean square error (RMSE) are used to quantify the lateral and longitudinal trajectory reconstruction accuracy of all individual vehicles within the blind zone in this invention. Within this blind zone, the spatial range of the blind zone is 1000 feet, and the detection range of the radar-visual integrated machine is 400 feet. The quantification indices MAE and RMSE for the longitudinal reconstructed trajectory are 3.67 feet and 5.71 feet, respectively; the quantification indices MAE and RMSE for the lateral reconstructed trajectory are 1.08 feet and 1.35 feet, respectively.

[0134] Furthermore, as an implementation scheme, this application also proposes a spatiotemporal vehicle trajectory reconstruction model, which includes:

[0135] Data acquisition module 100 is used to acquire the first vehicle driving data collected by the upstream radar-view integrated machine and the second vehicle driving data collected by the downstream radar-view integrated machine;

[0136] The lateral and longitudinal position extraction module 200 is used to determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data. The set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle.

[0137] The horizontal and vertical trajectory prediction module 300 is used to input the horizontal and vertical position set into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, as well as a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0138] The trajectory fusion module 400 is used to perform a weighted sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory to obtain a longitudinal fused trajectory, and to perform a weighted sum of the first lateral predicted trajectory and the second lateral predicted trajectory to obtain a lateral fused trajectory, wherein the weight sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory is 1, and the weight sum of the first lateral predicted trajectory and the second lateral predicted trajectory is 1.

[0139] As one implementation scheme, Figure 10 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0140] like Figure 10 As shown, the computer system may include: a processor 1001, such as a CPU; a memory 1005; a user interface 1003; a network interface 1004; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0141] Those skilled in the art will understand that Figure 10 The computer system architecture shown does not constitute a limitation on the computer system and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0142] like Figure 10 As shown, the memory 1005, as a storage medium, may include an operating system, a network communication module, a user interface module, and computer programs. The operating system is a program that manages and controls the hardware and software resources of the computer system, as well as the operation of the computer programs and other software or programs.

[0143] exist Figure 10 In the computer system shown, the user interface 1003 is mainly used to connect to the terminal and communicate with the terminal; the network interface 1004 is mainly used to communicate with the backend server; and the processor 1001 can be used to call the computer program stored in the memory 1005.

[0144] In this embodiment, the computer system includes: a memory 1005, a processor 1001, and a computer program stored in the memory and executable on the processor, wherein:

[0145] When processor 1001 calls a computer program stored in memory 1005, it performs the following operations:

[0146] S10, acquire the first vehicle driving data collected by the upstream integrated radar vision machine and the second vehicle driving data collected by the downstream integrated radar vision machine;

[0147] S20, determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data, wherein the set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle;

[0148] S30, the set of horizontal and vertical positions is input into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, and a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream.

[0149] S40, the first longitudinal prediction trajectory and the second longitudinal prediction trajectory are weighted and summed to obtain a longitudinal fusion trajectory, and the first lateral prediction trajectory and the second lateral prediction trajectory are weighted and summed to obtain a lateral fusion trajectory, wherein the weight sum of the first longitudinal prediction trajectory and the second longitudinal prediction trajectory is 1, and the weight sum of the first lateral prediction trajectory and the second lateral prediction trajectory is 1.

[0150] Furthermore, those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in a computer system to implement the process steps of the embodiments of the above methods.

[0151] Therefore, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the various steps of the all-time-space vehicle trajectory reconstruction method described in the above embodiments.

[0152] The computer-readable storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0153] It should be noted that, since the storage medium provided in the embodiments of this application is the storage medium used to implement the methods of the embodiments of this application, those skilled in the art can understand the specific structure and variations of the storage medium based on the methods described in the embodiments of this application, and therefore will not be repeated here. All storage media used in the methods of the embodiments of this application fall within the scope of protection of this application.

[0154] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0155] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0156] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0157] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0158] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended technical solutions are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0159] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations of this application fall within the scope of the technical solution of this application and its equivalents, this application also intends to include these modifications and variations.

Claims

1. A method for reconstructing vehicle trajectories across all time and space, characterized in that, The method includes the following steps: S10, acquire the first vehicle driving data collected by the upstream integrated radar vision machine and the second vehicle driving data collected by the downstream integrated radar vision machine; S20, determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data, wherein the set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle; S30, the set of horizontal and vertical positions is input into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, and a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream. S40, the first longitudinal prediction trajectory and the second longitudinal prediction trajectory are weighted and summed to obtain a longitudinal fusion trajectory, and the first lateral prediction trajectory and the second lateral prediction trajectory are weighted and summed to obtain a lateral fusion trajectory, wherein the weight sum of the first longitudinal prediction trajectory and the second longitudinal prediction trajectory is 1, and the weight sum of the first lateral prediction trajectory and the second lateral prediction trajectory is 1. The expression for the first lateral predicted trajectory is: ； The expression for the second lateral predicted trajectory is: ； In the formula, Let u be the x-coordinate of the position u of the target vehicle when it leaves the upstream radar-vision integrated machine. The x-coordinate of the position d of the target vehicle when it leaves the downstream radar-visual integrated machine; t i This represents the i-th time point when the vehicle is within the blind spot. This represents the predicted lateral velocity of the vehicle at the i-th time point, and n represents the total number of time points in which the vehicle is within the blind spot. The expression for the first longitudinal predicted trajectory is: ； The expression for the second longitudinal predicted trajectory is: ； In the formula, Let u be the ordinate of the position u of the target vehicle when it leaves the upstream radar-visual integrated machine. The vertical coordinate of position d when the target vehicle leaves the downstream radar-visual integrated machine; t i This represents the i-th time point when the vehicle is within the blind spot. This represents the predicted longitudinal speed of the vehicle at the i-th time point, and n represents the total number of time points in which the vehicle is within the blind zone. The preset model satisfies the following steps during execution: S21, Extract the multidimensional spatiotemporal features from the set of horizontal and vertical positions; S22, Discretize the multidimensional spatiotemporal features; S23, Introducing step size As a learnable parameter, the discretized multidimensional spatiotemporal features of the input are processed according to the step size. The size of the sample is used to obtain a discretized input velocity sequence. A selective mechanism is then used to transform the discrete parameters in the discretized input velocity sequence into dependent input parameters, wherein the step size is... During the learning process, adjustments are made based on the speed of the discretized multidimensional spatiotemporal features.

2. The method as described in claim 1, characterized in that, The mathematical expression for the lateral fusion trajectory is: ； in, , , ; In the formula, For horizontal fusion trajectory, The weight value corresponding to the first horizontal predicted trajectory. This is the first lateral predicted trajectory. This is the second lateral predicted trajectory. The weight value corresponding to the second horizontal predicted trajectory; The mathematical expression for the longitudinal fusion trajectory is: ； in, , , ; In the formula, For vertical fusion trajectory, The weight value corresponding to the first longitudinal predicted trajectory. This is the first longitudinal predicted trajectory. This is the second longitudinal predicted trajectory. The weight value is the value corresponding to the second longitudinal predicted trajectory.

3. The method as described in claim 1, characterized in that, S21 includes: S211, uncover the hidden state of the previous time step t-1. Multidimensional spatiotemporal characteristics of vehicle lateral / longitudinal velocity The output shows the current state. , the current state and multidimensional spatiotemporal features Mapping to output The multidimensional spatiotemporal feature sequence as output: ； ； in, , These represent the multidimensional spatiotemporal features of the input and output, respectively. Represents the current state of the transportation system, and indicates the relationship with... The result of combined action; matrix Used to adjust the dynamic state changes of the observation equation; and These are the input projection matrix and the output projection matrix, respectively; Let R be the residual, and R be a real number.

4. The method as described in claim 1, characterized in that, S22 includes: S221, convert matrices A and B into discrete parameters. and It maintains the current state each time a discrete signal is received. The value is retained until the next new discrete signal is received: ； ； in, and The matrix is ​​respectively and The parameters after discretization, Δ t A is the increment of the state matrix, Δ t B represents the increment of the input matrix, and I represents the identity matrix.

5. A system for implementing the all-time-space vehicle trajectory reconstruction method as described in claim 1, characterized in that, The system includes: The data acquisition module is used to acquire the first vehicle driving data collected by the upstream radar-view integrated machine and the second vehicle driving data collected by the downstream radar-view integrated machine; The lateral and longitudinal position extraction module is used to determine the set of lateral and longitudinal positions of the target vehicle in the first vehicle driving data and the second vehicle driving data. The set of lateral and longitudinal positions includes the lateral coordinates, longitudinal coordinates, lateral velocity and longitudinal velocity of the target vehicle. The horizontal and vertical trajectory prediction module is used to input the horizontal and vertical position set into a preset model so that the preset model generates a first vertical prediction trajectory and a first horizontal prediction trajectory generated from downstream to upstream, as well as a second vertical prediction trajectory and a second horizontal prediction trajectory generated from upstream to downstream. The trajectory fusion module is used to perform a weighted sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory to obtain a longitudinal fused trajectory, and to perform a weighted sum of the first lateral predicted trajectory and the second lateral predicted trajectory to obtain a lateral fused trajectory, wherein the weight sum of the first longitudinal predicted trajectory and the second longitudinal predicted trajectory is 1, and the weight sum of the first lateral predicted trajectory and the second lateral predicted trajectory is 1.

6. A computer system, characterized in that, The computer system includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the all-time-space vehicle trajectory reconstruction method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the all-time-space vehicle trajectory reconstruction method as described in any one of claims 1 to 4.

Images