Traffic restriction information extraction model training method and device

Abstract

This disclosure provides a training method and apparatus for a traffic restriction information extraction model. The method includes: acquiring vehicle travel trajectories at multiple intersections and multiple travel link pairs corresponding to each intersection; wherein, each travel link pair includes an inbound link and an outbound link matching the inbound link; matching the trajectory points of the vehicle travel trajectories with the multiple travel link pairs to generate matched link pair travel trajectories; and training a model based on the extracted detour features, link pair topology features, and link flow features in the link pair travel trajectories to obtain a traffic restriction information extraction model. This disclosure can reduce the dependence on trajectory sampling coverage, solve the problem of missing matching on short roads, and detect intersection restrictions using trajectory detour features, avoiding detour interference.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of model training technology, and in particular to a training method and apparatus for a traffic restriction information extraction model. Background Technology

[0002] With urban development, city road networks are becoming increasingly complex, placing immense pressure on traffic management. While traffic restrictions effectively promote rational traffic flow, ensure traffic safety, and improve efficiency, they also complicate route selection. The accuracy of traffic restriction information directly impacts user experience; discrepancies between navigation routes and actual conditions at intersections can lead to traffic violations and even accidents. Currently, traffic restriction information displayed on electronic maps is typically collected in the field by map providers' personnel. Given the frequent changes at intersections, this process is time-consuming and costly, making it impossible to guarantee the real-time nature of traffic restriction information.

[0003] With the widespread use of GPS (Global Positioning System) embedded devices, various types of moving objects generate massive amounts of GPS trajectory data, providing a new opportunity for traffic restriction analysis. Through these massive trajectories, comprehensive road information can be extracted, enabling efficient, real-time, and accurate traffic restriction intelligence at extremely low cost.

[0004] Existing technologies typically analyze vehicle trajectories and traffic flow at road intersections to quickly detect changes in traffic restrictions, improving data real-time performance and providing a safer and faster driving environment. This technology requires massive amounts of data to cover all intersections. If this is met, road combinations without matching trajectories are considered to have traffic restrictions. However, acquiring such massive amounts of data is difficult and cannot guarantee sufficient coverage of all intersections, or even complete intersections. Furthermore, the presence of traffic restrictions can lead users to take detours, with detour distances varying significantly across different road networks. In some cases where traffic restrictions exist, users can reach their desired destination after a brief detour and U-turn. In such situations, road combinations with traffic restrictions are mistakenly identified as not having traffic restrictions, reducing the recall rate of traffic restriction detection. Summary of the Invention

[0005] The embodiments of this disclosure provide a training method and apparatus for a traffic restriction information extraction model, which reduces the dependence on trajectory sampling coverage, solves the problem of missing matching of short roads, and uses trajectory detour features to detect traffic restrictions and avoid detour interference.

[0006] According to a first aspect of the embodiments of this disclosure, a method for training a traffic restriction information extraction model is provided, comprising:

[0007] Obtain vehicle travel trajectories at multiple intersections, and multiple travel link pairs corresponding to each of the multiple intersections; wherein, each travel link pair includes: an inbound link and an outbound link matching the inbound link;

[0008] The trajectory points of the vehicle's driving trajectory are matched with the multiple driving link pairs to generate a matching link pair driving trajectory;

[0009] Based on the extracted detour features, link topology features, and link traffic features in the driving trajectory, a traffic restriction information extraction model is obtained through model training.

[0010] Optionally, the process of matching the trajectory points of the vehicle's driving trajectory with the multiple driving link pairs to generate a matched link pair driving trajectory includes:

[0011] The trajectory points of the vehicle's driving trajectory are differentially processed to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0012] The differential trajectory points are matched with the multiple driving link pairs to obtain the matched link pair trajectories;

[0013] The link pair trajectory is preprocessed according to a preset trajectory threshold to obtain the link pair travel trajectory.

[0014] Optionally, the step of performing differential processing on the trajectory points of the vehicle's driving trajectory to generate differential trajectory points corresponding to the vehicle's driving trajectory includes:

[0015] The trajectory points are sorted based on the timestamps corresponding to the trajectory points to obtain sorted trajectory points;

[0016] Interpolation is performed on the sorted trajectory points to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0017] Optionally, matching the differential trajectory points with the multiple driving link pairs to obtain the matched link pair trajectories includes:

[0018] Based on the entry and exit links of the differential trajectory points, determine the target driving link pair among the multiple driving link pairs;

[0019] Based on the differential trajectory points and the target driving link pair, the trajectory of the link pair is determined.

[0020] Optionally, the step of preprocessing the link pair trajectory according to a preset trajectory threshold to obtain the link pair travel trajectory includes:

[0021] Obtain the number of vehicle trajectories at the multiple intersections;

[0022] Identify the initial intersections among the plurality of intersections where the number of driving trajectories is greater than a first trajectory number threshold;

[0023] Obtain the detour trajectories with detour characteristics from the vehicle driving trajectories at the initial intersection;

[0024] Obtain the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection;

[0025] Identify the target intersections in the initial intersections whose ratio is greater than a set ratio threshold;

[0026] Obtain target driving link pairs for each driving link pair at the target intersection where the number of driving link trajectories is greater than a second trajectory number threshold, and use the link pair trajectory corresponding to the target driving link pair as the link pair driving trajectory.

[0027] Optionally, the traffic restriction information extraction model is trained based on the extracted detour features, link pair topology features, and link flow features in the driving trajectory of the link pairs to obtain the model, including:

[0028] The driving trajectory of the link pair is input into the initial traffic restriction information extraction model;

[0029] The initial traffic restriction information extraction model extracts detour features, link pair topology features, and link flow features from the link pair's travel trajectory, and outputs the sample accuracy value corresponding to the link pair's travel trajectory based on the detour features, link pair topology features, and link flow features.

[0030] The target link trajectory in the link pair's driving trajectory is obtained based on the sample precision value;

[0031] The target link trajectory and the link pair driving trajectory are differentially processed to generate a differential link trajectory;

[0032] The initial traffic restriction information extraction model is trained a second time based on the differential link trajectory until the traffic restriction information extraction model is obtained.

[0033] Optionally, after training the model based on the extracted detour features, link pair topology features, and link flow features in the link pair's driving trajectory to obtain the traffic restriction information extraction model, the method further includes:

[0034] When it is necessary to extract traffic restriction information of the intersection to be predicted, the historical vehicle travel trajectory of the intersection to be predicted is obtained.

[0035] The historical vehicle driving trajectory is input into the traffic restriction information extraction model, and the traffic restriction information of the intersection to be predicted and the confidence level corresponding to the traffic restriction information are obtained from the traffic restriction information extraction model.

[0036] Based on the confidence level and confidence threshold, the target traffic restriction information in the traffic restriction information is determined.

[0037] Optionally, determining the target traffic restriction information in the traffic restriction information based on the confidence level and the confidence threshold includes:

[0038] Obtain traffic restriction information with a confidence level greater than a first confidence threshold from the traffic restriction information, and use the traffic restriction information with a confidence level greater than the first confidence threshold as the target traffic restriction information;

[0039] Obtain traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold from the traffic restriction information, so that business personnel can determine the target traffic restriction information from the traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold;

[0040] The traffic restriction information with a confidence level less than the second confidence threshold is obtained, and the traffic restriction information extraction model is iteratively trained based on the traffic restriction information with a confidence level less than the second confidence threshold.

[0041] Secondly, embodiments of this disclosure provide a training apparatus for a traffic restriction information extraction model, comprising:

[0042] The vehicle trajectory acquisition module is used to acquire vehicle trajectories at multiple intersections and multiple driving link pairs corresponding to the multiple intersections; wherein, the driving link pair includes: an inbound link and an outbound link that matches the inbound link;

[0043] The link pair driving trajectory generation module is used to match the trajectory points of the vehicle driving trajectory with the multiple driving link pairs to generate a matched link pair driving trajectory.

[0044] The traffic restriction extraction model acquisition module is used to train a model based on the detour features, link pair topology features, and link flow features in the extracted link pair driving trajectory to obtain a traffic restriction information extraction model.

[0045] Optionally, the link-to-trajectory generation module includes:

[0046] The differential trajectory point generation unit is used to perform differential processing on the trajectory points of the vehicle's driving trajectory to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0047] The link pair trajectory acquisition unit is used to match the differential trajectory points with the multiple driving link pairs to obtain the matched link pair trajectory;

[0048] The link pair travel trajectory acquisition unit is used to preprocess the link pair trajectory according to a preset trajectory threshold to obtain the link pair travel trajectory.

[0049] Optionally, the differential trajectory point generation unit includes:

[0050] The sorted trajectory point acquisition subunit is used to sort the trajectory points based on the timestamps corresponding to the trajectory points to obtain sorted trajectory points;

[0051] The differential trajectory point generation subunit is used to perform interpolation processing on the sorted trajectory points to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0052] Optionally, the link pair trajectory acquisition unit includes:

[0053] The target link pair determination subunit is used to determine the target driving link pair among the multiple driving link pairs based on the driving in link and driving out link of the differential trajectory point;

[0054] The link pair trajectory determination subunit is used to determine the link pair trajectory based on the differential trajectory points and the target driving link pair.

[0055] Optionally, the link pair driving trajectory acquisition unit includes:

[0056] The driving trajectory number acquisition subunit is used to acquire the number of driving trajectories of vehicles at the multiple intersections;

[0057] An initial intersection acquisition subunit is used to acquire initial intersections among the plurality of intersections where the number of driving trajectories is greater than a first trajectory number threshold.

[0058] The detour trajectory acquisition subunit is used to acquire detour trajectories with detour characteristics from the vehicle driving trajectories at the initial intersection.

[0059] The trajectory ratio acquisition subunit is used to acquire the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection;

[0060] The target intersection acquisition subunit is used to acquire target intersections in the initial intersections whose ratio is greater than a set ratio threshold.

[0061] The link pair driving trajectory acquisition subunit is used to acquire target driving link pairs at the target intersection whose number of driving link trajectories is greater than the second trajectory number threshold, and to use the link pair trajectory corresponding to the target driving link pair as the link pair driving trajectory.

[0062] Optionally, the intersection limit extraction model acquisition module includes:

[0063] The link pair driving trajectory input unit is used to input the link pair driving trajectory into the initial traffic restriction information extraction model;

[0064] The sample precision value extraction unit is used to extract the detour features, link pair topology features, and link flow features in the link pair travel trajectory through the initial traffic restriction information extraction model, and the initial traffic restriction information extraction model outputs the sample precision value corresponding to the link pair travel trajectory based on the detour features, the link pair topology features, and the link flow features;

[0065] The target link trajectory acquisition unit is used to acquire the target link trajectory in the link pair driving trajectory based on the sample precision value;

[0066] A differential link trajectory generation unit is used to perform differential processing on the target link trajectory and the link pair driving trajectory to generate a differential link trajectory;

[0067] The traffic restriction extraction model acquisition unit is used to perform secondary training on the initial traffic restriction information extraction model based on the differential link trajectory until the traffic restriction information extraction model is obtained.

[0068] Optionally, the device further includes:

[0069] The historical driving trajectory acquisition module is used to acquire the historical vehicle driving trajectory of the intersection to be predicted when it is necessary to extract traffic restriction information of the intersection to be predicted.

[0070] The confidence level acquisition module is used to input the historical vehicle driving trajectory into the traffic restriction information extraction model, and to obtain the traffic restriction information of the intersection to be predicted output by the traffic restriction information extraction model, and the confidence level corresponding to the traffic restriction information;

[0071] The target traffic restriction information determination module is used to determine the target traffic restriction information in the traffic restriction information based on the confidence level and the confidence threshold.

[0072] Optionally, the target intersection limit information determination module includes:

[0073] The first target traffic restriction information acquisition unit is used to acquire traffic restriction information with a confidence level greater than a first confidence threshold from the traffic restriction information, and to use the traffic restriction information with a confidence level greater than the first confidence threshold as the target traffic restriction information;

[0074] The second target traffic restriction information acquisition unit is used to acquire traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold, so that business personnel can determine the target traffic restriction information among the traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold;

[0075] An iterative training unit is used to acquire traffic restriction information with a confidence level less than the second confidence threshold from the traffic restriction information, and to iteratively train the traffic restriction information extraction model based on the traffic restriction information with a confidence level less than the second confidence threshold.

[0076] According to a third aspect of the embodiments of this disclosure, an electronic device is provided, comprising:

[0077] A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the training method for the traffic restriction information extraction model described in any of the preceding claims.

[0078] According to a fourth aspect of the embodiments of the present disclosure, a readable storage medium is provided, wherein when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the training method for the traffic restriction information extraction model described in any of the preceding claims.

[0079] This disclosure provides a training method and apparatus for a traffic restriction information extraction model. It acquires vehicle travel trajectories at multiple intersections and multiple travel link pairs corresponding to each intersection. Each travel link pair includes an inbound link and an outbound link matching the inbound link. The method matches the trajectory points of the vehicle travel trajectories with the multiple travel link pairs to generate matched link pair travel trajectories. Based on the extracted detour features, link pair topology features, and link flow features in the link pair travel trajectories, the model is trained to obtain the traffic restriction information extraction model. This disclosure utilizes massive trajectory data to mine road network traffic restriction elements, uses differentially differentiated trajectories for road binding to address the problem of missing short road matching, and extracts detour features, road network topology features, and flow features to train the traffic restriction mining model, thereby avoiding detour interference. Attached Figure Description

[0080] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments of this disclosure will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0081] Figure 1 A flowchart illustrating the steps of a training method for a traffic restriction information extraction model provided in an embodiment of this disclosure;

[0082] Figure 2 A schematic diagram of a driving link provided for an embodiment of this disclosure;

[0083] Figure 3 A schematic diagram of a detour driving trajectory provided for an embodiment of this disclosure;

[0084] Figure 4 This is a schematic diagram of the structure of a training device for a traffic restriction information extraction model provided in an embodiment of the present disclosure. Detailed Implementation

[0085] The technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without creative effort are within the protection scope of the embodiments of this disclosure.

[0086] Example 1

[0087] Reference Figure 1 The flowchart illustrates the steps of a training method for a traffic restriction information extraction model provided by an embodiment of this disclosure, as shown below. Figure 1 As shown, the training method for this traffic restriction information extraction model may include the following steps:

[0088] Step 101: Obtain the vehicle driving trajectory at multiple intersections, and the multiple driving link pairs corresponding to the multiple intersections; wherein, the driving link pair includes: an inbound link and an outbound link matching the inbound link.

[0089] The embodiments disclosed herein can be applied to scenarios where traffic restriction information extraction models are trained by combining differential driving trajectories and detour features at intersections.

[0090] Vehicle trajectory refers to the historical trajectory of a vehicle at multiple intersections, that is, the trajectory generated when a vehicle passes through multiple intersections.

[0091] A driving link pair refers to the link pair formed when a vehicle enters or exits an intersection. In this example, a driving link pair can include: an entering link and an exit link that matches the entering link. Driving link pairs can be combined with... Figure 2 The following is a detailed description.

[0092] Reference Figure 2 The diagram illustrates a driving link provided by an embodiment of this disclosure, such as... Figure 2 As shown, taking a single intersection as an example, let the intersection ID be C1, and the link IDs of the links constituting this intersection be R1, R2, R3, R4, R5, R6, R7, and R8, respectively. Based on the traffic direction of the eight links, it can be determined that the entry links to intersection C1 are R2, R4, R6, and R8, and the exit links are R1, R3, R5, and R7. Taking R2 as the entry point, under the condition of no traffic restrictions, exit links are R1, R3, R5, and R7. Similarly, we can list all the possible combinations of link pairs at this intersection, namely: R2-R1, R2-R3, R2-R5, R2-R7, R4-R1, R4-R3, R4-R5, R4-R7, R6-R1, R6-R3, R6-R5, R6-R7, R8-R1, R8-R3, R8-R5, and R8-R7.

[0093] It is understood that the above examples are merely examples listed to better understand the technical solutions of the embodiments of this disclosure, and are not intended to be the only limitation of this embodiment.

[0094] After obtaining the vehicle driving trajectories at multiple intersections and the multiple driving link pairs corresponding to each intersection, step 102 is executed.

[0095] Step 102: Match the trajectory points of the vehicle's driving trajectory with the multiple driving link pairs to generate a matched link pair driving trajectory.

[0096] After obtaining the vehicle driving trajectory at multiple intersections and the multiple driving link pairs corresponding to each intersection, the trajectory points of the vehicle driving trajectory can be matched with the multiple driving link pairs to generate the matched link pair driving trajectory. Specifically, the following detailed implementation method can be described in detail.

[0097] In one specific implementation of this disclosure, step 102 may include:

[0098] Sub-step S1: Perform differential processing on the trajectory points of the vehicle's driving trajectory to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0099] After obtaining the vehicle trajectories at multiple intersections, the vehicle trajectories at each intersection can be differentially processed to generate differential trajectory points corresponding to the vehicle trajectories, thereby obtaining a denser number of trajectories. Specifically, the differential processing of the trajectory points of the vehicle trajectories can be described in detail in conjunction with the following specific implementation method.

[0100] In one specific implementation of an embodiment of this disclosure, the sub-step S1 described above may include:

[0101] Sub-step A1: Sort the trajectory points based on the timestamps corresponding to the trajectory points to obtain sorted trajectory points.

[0102] In the embodiments of this disclosure, after obtaining the vehicle's driving trajectory, the trajectory points can be sorted based on the timestamps corresponding to the trajectory points, thereby obtaining sorted trajectory points.

[0103] After obtaining the sorted trajectory points, execute sub-step A2.

[0104] Sub-step A2: Perform interpolation on the sorted trajectory points to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0105] After obtaining the sorted trajectory points, interpolation can be performed on the sorted trajectory points to generate the differential trajectory points corresponding to the vehicle's driving trajectory.

[0106] The embodiments of this disclosure interpolate the trajectory points to obtain a denser pair of trajectory points, which can solve the problem of missing trajectory matching for short roads.

[0107] After performing differential processing on the trajectory points of the vehicle's driving trajectory to generate differential trajectory points corresponding to the vehicle's driving trajectory, sub-step S2 is executed.

[0108] Sub-step S2: Match the differential trajectory points with the multiple driving link pairs to obtain the matched link pair trajectories.

[0109] A link pair trajectory is a link pair trajectory obtained by matching differential trajectory points with a driving link pair trajectory. In other words, a link pair trajectory includes a driving link pair and the differential trajectory points corresponding to that driving link pair trajectory.

[0110] After generating differential trajectory points corresponding to the vehicle's driving trajectory by differential processing of the trajectory points, the differential trajectory points can be matched with multiple driving link pairs to obtain the matched link pair trajectory. Specifically, the following detailed implementation method can be described in detail.

[0111] In another specific implementation of the embodiments of this disclosure, the above sub-step S2 may include:

[0112] Sub-step B1: Based on the entry and exit links of the differential trajectory points, determine the target driving link pair among the multiple driving link pairs.

[0113] In embodiments of this disclosure, a target travel link pair refers to a travel link pair at an intersection that matches the differential trajectory points, for example, such as Figure 2 As shown, if the driving link pair matched by the differential trajectory points is R2-R1, then the driving link pair R2-R1 is taken as the target driving link pair.

[0114] After obtaining the differential trajectory points, the target driving link pair among multiple driving link pairs can be determined based on the entry and exit links corresponding to the differential trajectory points.

[0115] After the target driving link pair is determined, proceed to sub-step B2.

[0116] Sub-step B2: Determine the trajectory of the link pair based on the differential trajectory points and the target driving link pair.

[0117] After obtaining the target driving link pair, the link pair trajectory can be determined based on the differential trajectory points and the target driving link pair.

[0118] After matching the differential trajectory points with multiple driving link pairs to obtain the matched link pair trajectory, sub-step S3 is executed.

[0119] Sub-step S3: Preprocess the link pair trajectory according to the preset trajectory threshold to obtain the link pair driving trajectory.

[0120] Preset trajectory thresholds refer to thresholds set in advance to filter driving trajectories used for training traffic restriction information extraction models. In this example, preset trajectory thresholds may include: trajectory number thresholds, detour ratio thresholds, etc.

[0121] After matching the differential trajectory points with multiple driving link pairs to obtain the matched link pair trajectories, the link pair trajectories can be preprocessed according to a preset trajectory threshold to obtain the link pair driving trajectory, that is, to filter out the driving trajectories that meet the model training conditions. Specifically, the following detailed implementation method can be described in detail.

[0122] In another specific implementation of the embodiments of this disclosure, the above sub-step S3 may include:

[0123] Sub-step C1: Obtain the number of vehicle trajectories at the multiple intersections.

[0124] In this embodiment, the number of driving trajectories refers to the number of vehicle driving trajectories at multiple intersections, that is, the number of vehicle driving trajectories at each intersection.

[0125] After obtaining the vehicle driving trajectories at multiple intersections, the number of vehicle driving trajectories at multiple intersections can be obtained, and then sub-step C2 can be executed.

[0126] Sub-step C2: Obtain the initial intersections among the multiple intersections where the number of driving trajectories is greater than the first trajectory number threshold.

[0127] The first trajectory number threshold refers to the pre-set threshold used to determine whether the number of vehicle driving trajectories at an intersection meets the model training conditions.

[0128] An initial intersection refers to an intersection where the number of vehicle trajectories among multiple intersections exceeds a first threshold for the number of trajectories.

[0129] After obtaining the number of vehicle trajectories at multiple intersections, the intersections with a number of trajectories greater than a first trajectory number threshold can be selected as initial intersections. For example, the multiple intersections are: intersection 1, intersection 2, intersection 3, and intersection 4. The number of vehicle trajectories at intersection 1 is 100, the number of vehicle trajectories at intersection 2 is 80, the number of vehicle trajectories at intersection 3 is 88, and the number of vehicle trajectories at intersection 4 is 50. The first trajectory number threshold is 68. In this case, intersection 1, intersection 2, and intersection 3 can be selected as initial intersections.

[0130] After obtaining the initial intersections where the number of driving trajectories is greater than the first trajectory number threshold, sub-step C3 is executed.

[0131] Sub-step C3: Obtain the detour trajectory with detour characteristics from the vehicle driving trajectory of the initial intersection.

[0132] A detour route refers to a vehicle's travel trajectory that exhibits detour characteristics, for example, such as... Figure 3 As shown, due to the no-left-turn restriction at intersection C1 (R4 to R1), drivers need to take a detour to enter R1 from R4. As shown in trajectory T1, the vehicle enters intersection C1 from R4, detours to R5 and R6, and then exits from R1. This detour is the vehicle's trajectory.

[0133] After obtaining the initial intersections where the number of driving trajectories exceeds the first trajectory number threshold, it is possible to obtain the detour driving trajectories with detour characteristics among the vehicle driving trajectories of the initial intersections.

[0134] After obtaining the detour trajectory with detour characteristics from the vehicle driving trajectory at the initial intersection, execute sub-step C4.

[0135] Sub-step C4: Obtain the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection.

[0136] After obtaining the detour trajectories with detour characteristics in the vehicle driving trajectory of the initial intersection, the ratio of the number of detour trajectories to the number of driving trajectories corresponding to the initial intersection can be obtained, that is, the proportion of detour trajectories at each initial intersection.

[0137] After obtaining the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection, sub-step C5 is executed.

[0138] Sub-step C5: Obtain the target intersections in the initial intersections whose ratio is greater than the set ratio threshold.

[0139] The target intersection refers to the intersection from the initial intersections where the proportion of detour routes exceeds a certain threshold.

[0140] Setting a ratio threshold refers to a pre-set threshold used to filter vehicle trajectories with detour characteristics at intersections.

[0141] After obtaining the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection, target intersections with a ratio greater than a set ratio threshold can be selected from the initial intersections, that is, intersections with a detour feature ratio greater than the threshold can be selected.

[0142] After identifying the target intersections in the initial intersections whose ratio is greater than the set ratio threshold, sub-step C6 is executed.

[0143] Sub-step C6: Obtain the target driving link pairs for each driving link pair at the target intersection whose number of driving link trajectories is greater than the second trajectory number threshold, and use the link pair trajectory corresponding to the target driving link pair as the link pair driving trajectory.

[0144] After obtaining the target intersection, we can obtain target driving link pairs for each driving link pair at the target intersection whose number of driving link trajectories is greater than the second trajectory number threshold, and use the link pair trajectory corresponding to the target driving link pair as the link pair driving trajectory for subsequent training of the traffic restriction information extraction model.

[0145] In this embodiment, according to business logic, traffic restriction information exists at complex intersections with high total traffic volume and low combined traffic volume for link pairs with traffic restrictions. The core characteristic is the presence of detour behavior. If there is no corresponding matching trajectory for the current link pair, no judgment is made to avoid misjudgment of traffic restrictions due to insufficient trajectory coverage. The recall module filters traffic by setting a minimum traffic volume threshold for the entire intersection, a minimum route-binding trajectory threshold for each link pair, and a minimum detour ratio threshold, according to the above logic. The recall module performs secondary cleaning on the entire sample space to ensure that the subsequent model meets business expectations in terms of key features, reducing the bad case rate; alleviating the problem of imbalance between positive and negative traffic restriction samples to a certain extent; and increasing robustness to the current traffic restriction label.

[0146] After matching the trajectory points of the vehicle's driving trajectory with multiple driving link pairs to generate a matching link pair driving trajectory, step 103 is executed.

[0147] Step 103: Based on the extracted detour features, link pair topology features, and link traffic features in the driving trajectory of the link pairs, a model is trained to obtain a traffic restriction information extraction model.

[0148] After obtaining the link pair travel trajectory, a traffic restriction information extraction model can be trained based on the link pair travel trajectory. Specifically, the link pair travel trajectory can be input into the traffic restriction information extraction model to be trained, and the traffic restriction information extraction model to be trained can extract the detour features, link pair topology features, and link flow features in the link pair travel trajectory for model training to obtain the traffic restriction information extraction model. Specifically, the model training process can be described in detail in conjunction with the following specific implementation method.

[0149] In another specific implementation of the embodiments of this disclosure, step 103 may include:

[0150] Sub-step D1: Input the driving trajectory of the link pair into the initial traffic restriction information extraction model.

[0151] In the embodiments of this disclosure, after obtaining the link pair travel trajectory, the link pair travel trajectory can be input into the initial traffic restriction information extraction model.

[0152] After inputting the link pair driving trajectory into the initial traffic restriction information extraction model, sub-step D2 is executed.

[0153] Sub-step D2: Extract the detour features, link pair topology features, and link flow features from the link pair's travel trajectory using the initial traffic restriction information extraction model, and output the sample accuracy value corresponding to the link pair's travel trajectory based on the detour features, link pair topology features, and link flow features.

[0154] After inputting the link pair's travel trajectory into the initial traffic restriction information extraction model, the detour features, link pair topology features, and link flow features in the link pair's travel trajectory can be extracted using the initial traffic restriction information extraction model.

[0155] The detour characteristics may include: the ratio of the number of trajectories with a trajectory length between inbound and outbound link pairs that is within 20 meters, 50 meters, 100 meters, 200 meters, and 500 meters to the total number of matched trajectories; the ratio of the number of trajectories with a trajectory point number between inbound and outbound link pairs that is within 5, 10, 20, and 40 points to the total number of matched trajectories; the ratio of the ratio of the trajectory length to the link length matched on inbound and outbound link pairs to the link length that is within 1.0, 1.5, 2.0, 3.0, and 5.0 to the total number of matched trajectories; the ratio of the length of the unmatched trajectory between inbound and outbound link pairs to the interval length between link pairs that is within 1.0, 1.5, 2.0, 3.0, and 5.0 to the total number of matched trajectories; and the ratio of the number of detour trajectories between inbound and outbound link pairs to the total number of matched trajectories.

[0156] Link pair topology characteristics may include: left turn, straight, right turn, U-turn type, and the link length of each incoming and outgoing link pair.

[0157] Link traffic characteristics can include inbound and outbound link traffic characteristics, the traffic of each inbound and outbound link pair, the common traffic matched between inbound and outbound link pairs, the traffic proportion of the current inbound and outbound link pair combination in the entire intersection, and the diversion and convergence proportions of the current inbound and outbound link pair within the intersection.

[0158] After extracting the detour features, link topology features, and link flow features from the link pair's travel trajectory, the initial traffic restriction information extraction model can perform corresponding model processing on the detour features, link topology features, and link flow features to obtain the sample accuracy value corresponding to the link pair's travel trajectory. Thus, the initial traffic restriction information extraction model can output the sample accuracy value corresponding to the link pair's travel trajectory.

[0159] After obtaining the sample accuracy value corresponding to the link pair's driving trajectory, execute sub-step D3.

[0160] Sub-step D3: Obtain the target link trajectory in the link pair's driving trajectory based on the sample precision value.

[0161] After obtaining the sample precision value corresponding to the link pair's driving trajectory, the target link trajectory in the link pair's driving trajectory can be obtained based on the sample precision value. Specifically, the link pair driving trajectory with a sample precision value greater than the precision threshold can be selected from the link pair driving trajectory based on the sample precision value as the target link trajectory.

[0162] After obtaining the target link trajectory in the link pair driving trajectory based on the sample precision value, sub-step D4 is executed.

[0163] Sub-step D4: Perform differential processing on the target link trajectory and the link pair driving trajectory to generate a differential link trajectory.

[0164] After obtaining the target link trajectory, differential processing can be performed on the target link trajectory and the link pair driving trajectory to generate a differential link trajectory.

[0165] After generating the differential link trajectory, execute sub-step D5.

[0166] Sub-step D5: Based on the differential link trajectory, perform secondary training on the initial traffic restriction information extraction model until the traffic restriction information extraction model is obtained.

[0167] After generating the differential link trajectory, the initial traffic restriction information extraction model can be trained again based on the differential link trajectory until the traffic restriction information extraction model is obtained.

[0168] After training the traffic restriction information extraction model, traffic restriction information can be predicted and deployed based on the model. Specifically, the following detailed implementation method can be described in detail.

[0169] In another specific implementation of the embodiments of this disclosure, after step 105 above, the following may be included:

[0170] Step F1: When it is necessary to extract traffic restriction information of the intersection to be predicted, obtain the historical vehicle travel trajectory of the intersection to be predicted.

[0171] In the embodiments of this disclosure, when it is necessary to extract traffic restriction information of the intersection to be predicted, the historical vehicle travel trajectories of the intersection to be predicted can be obtained.

[0172] After obtaining the historical vehicle travel trajectories of the intersection to be predicted, step F2 is executed.

[0173] Step F2: Input the historical vehicle driving trajectory into the traffic restriction information extraction model, and obtain the traffic restriction information of the intersection to be predicted output by the traffic restriction information extraction model, and the confidence level corresponding to the traffic restriction information.

[0174] After obtaining the historical vehicle driving trajectories of the intersection to be predicted, the historical vehicle driving trajectories can be input into the traffic restriction information extraction model, and the traffic restriction information of the intersection to be predicted and the corresponding confidence level of the traffic restriction information can be obtained from the traffic restriction information extraction model.

[0175] Step F3: Based on the confidence level and confidence threshold, determine the target traffic restriction information in the traffic restriction information.

[0176] After obtaining the traffic restriction information and the corresponding confidence level of the intersection to be predicted, the target traffic restriction information in the traffic restriction information can be determined based on the confidence level and the confidence threshold. The target traffic restriction information is the information used to indicate that there is a traffic restriction at the intersection to be predicted.

[0177] Specifically, it can be divided into the following three situations:

[0178] 1. Obtain traffic restriction information with a confidence level greater than the first confidence threshold, and use the traffic restriction information with a confidence level greater than the first confidence threshold as the target traffic restriction information. For example, if the first confidence threshold is 80%, traffic restriction information with a confidence level greater than 80% can be uploaded online to be broadcast to vehicles.

[0179] 2. Obtain traffic restriction information with a confidence level less than or equal to a first confidence threshold and greater than a second confidence threshold. Business personnel can then determine the target traffic restriction information among these traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold. For example, if the first confidence threshold is 80% and the second confidence threshold is 60%, then the traffic restriction information with a confidence level between 60% and 80% from the traffic restriction information of the intersection to be predicted can be manually verified to assist in its online deployment.

[0180] 3. Obtain traffic restriction information with a confidence level less than the second confidence threshold, and iteratively train the traffic restriction information extraction model based on the traffic restriction information with a confidence level less than the second confidence threshold.

[0181] The training method for the traffic restriction information extraction model provided in this disclosure involves acquiring vehicle travel trajectories at multiple intersections and multiple travel link pairs corresponding to each intersection. Each travel link pair includes an inbound link and an outbound link matching the inbound link. The method matches the trajectory points of the vehicle travel trajectories with the multiple travel link pairs to generate matched link pair travel trajectories. Based on the extracted detour features, link pair topology features, and link flow features in the link pair travel trajectories, the model is trained to obtain the traffic restriction information extraction model. This disclosure utilizes massive trajectory data to mine road network traffic restriction elements, uses differentially differentiated trajectories for road binding to address the problem of missing short road matching, and extracts detour features, road network topology features, and flow features to train the intersection restriction mining model, thereby avoiding detour interference.

[0182] Example 2

[0183] Reference Figure 4 The diagram shows a schematic representation of the structure of a training device for a traffic restriction information extraction model provided in an embodiment of this disclosure. Figure 4 As shown, the training device 400 for the traffic restriction information extraction model may include the following modules:

[0184] The vehicle trajectory acquisition module 410 is used to acquire vehicle trajectories at multiple intersections and multiple driving link pairs corresponding to the multiple intersections; wherein, the driving link pair includes: an inbound link and an outbound link that matches the inbound link.

[0185] The link pair driving trajectory generation module 420 is used to match the trajectory points of the vehicle driving trajectory with the multiple driving link pairs to generate a matched link pair driving trajectory.

[0186] The traffic restriction extraction model acquisition module 430 is used to train a model based on the detour features, link pair topology features and link flow features in the extracted link pair driving trajectory to obtain a traffic restriction information extraction model.

[0187] Optionally, the link-to-trajectory generation module includes:

[0188] The differential trajectory point generation unit is used to perform differential processing on the trajectory points of the vehicle's driving trajectory to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0189] The link pair trajectory acquisition unit is used to match the differential trajectory points with the multiple driving link pairs to obtain the matched link pair trajectory;

[0190] The link pair travel trajectory acquisition unit is used to preprocess the link pair trajectory according to a preset trajectory threshold to obtain the link pair travel trajectory.

[0191] Optionally, the differential trajectory point generation unit includes:

[0192] The sorted trajectory point acquisition subunit is used to sort the trajectory points based on the timestamps corresponding to the trajectory points to obtain sorted trajectory points;

[0193] The differential trajectory point generation subunit is used to perform interpolation processing on the sorted trajectory points to generate differential trajectory points corresponding to the vehicle's driving trajectory.

[0194] Optionally, the link pair trajectory acquisition unit includes:

[0195] The target link pair determination subunit is used to determine the target driving link pair among the multiple driving link pairs based on the driving in link and driving out link of the differential trajectory point;

[0196] The link pair trajectory determination subunit is used to determine the link pair trajectory based on the differential trajectory points and the target driving link pair.

[0197] Optionally, the link pair driving trajectory acquisition unit includes:

[0198] The driving trajectory number acquisition subunit is used to acquire the number of driving trajectories of vehicles at the multiple intersections;

[0199] An initial intersection acquisition subunit is used to acquire initial intersections among the plurality of intersections where the number of driving trajectories is greater than a first trajectory number threshold.

[0200] The detour trajectory acquisition subunit is used to acquire detour trajectories with detour characteristics from the vehicle driving trajectories at the initial intersection.

[0201] The trajectory ratio acquisition subunit is used to acquire the ratio of the number of detour driving trajectories to the number of driving trajectories corresponding to the initial intersection;

[0202] The target intersection acquisition subunit is used to acquire target intersections in the initial intersections whose ratio is greater than a set ratio threshold.

[0203] The link pair driving trajectory acquisition subunit is used to acquire target driving link pairs at the target intersection whose number of driving link trajectories is greater than the second trajectory number threshold, and to use the link pair trajectory corresponding to the target driving link pair as the link pair driving trajectory.

[0204] Optionally, the intersection limit extraction model acquisition module includes:

[0205] The link pair driving trajectory input unit is used to input the link pair driving trajectory into the initial traffic restriction information extraction model;

[0206] The sample precision value extraction unit is used to extract the detour features, link pair topology features, and link flow features in the link pair travel trajectory through the initial traffic restriction information extraction model, and the initial traffic restriction information extraction model outputs the sample precision value corresponding to the link pair travel trajectory based on the detour features, the link pair topology features, and the link flow features;

[0207] The target link trajectory acquisition unit is used to acquire the target link trajectory in the link pair driving trajectory based on the sample precision value;

[0208] A differential link trajectory generation unit is used to perform differential processing on the target link trajectory and the link pair driving trajectory to generate a differential link trajectory;

[0209] The traffic restriction extraction model acquisition unit is used to perform secondary training on the initial traffic restriction information extraction model based on the differential link trajectory until the traffic restriction information extraction model is obtained.

[0210] Optionally, the device further includes:

[0211] The historical driving trajectory acquisition module is used to acquire the historical vehicle driving trajectory of the intersection to be predicted when it is necessary to extract traffic restriction information of the intersection to be predicted.

[0212] The confidence level acquisition module is used to input the historical vehicle driving trajectory into the traffic restriction information extraction model, and to obtain the traffic restriction information of the intersection to be predicted output by the traffic restriction information extraction model, and the confidence level corresponding to the traffic restriction information;

[0213] The target traffic restriction information determination module is used to determine the target traffic restriction information in the traffic restriction information based on the confidence level and the confidence threshold.

[0214] Optionally, the target intersection limit information determination module includes:

[0215] The first target traffic restriction information acquisition unit is used to acquire traffic restriction information with a confidence level greater than a first confidence threshold from the traffic restriction information, and to use the traffic restriction information with a confidence level greater than the first confidence threshold as the target traffic restriction information;

[0216] The second target traffic restriction information acquisition unit is used to acquire traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold, so that business personnel can determine the target traffic restriction information among the traffic restriction information with a confidence level less than or equal to the first confidence threshold and greater than the second confidence threshold;

[0217] An iterative training unit is used to acquire traffic restriction information with a confidence level less than the second confidence threshold from the traffic restriction information, and to iteratively train the traffic restriction information extraction model based on the traffic restriction information with a confidence level less than the second confidence threshold.

[0218] The training device for the traffic restriction information extraction model provided in the embodiments of this disclosure acquires vehicle driving trajectories at multiple intersections and multiple driving link pairs corresponding to each intersection. Each driving link pair includes an inbound link and an outbound link matching the inbound link. The device matches the trajectory points of the vehicle driving trajectories with the multiple driving link pairs to generate matched link pair driving trajectories. Based on the extracted detour features, link pair topology features, and link flow features in the link pair driving trajectories, the device trains a model to obtain the traffic restriction information extraction model. The embodiments of this disclosure mine road network traffic restriction elements from massive trajectory data, use differentially differentiated trajectories for road binding to solve the problem of missing short road matching, and extract detour features, road network topology features, and flow features to train the intersection restriction mining model, thereby avoiding detour interference.

[0219] Embodiments of this disclosure also provide an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the training method of the traffic restriction information extraction model of the foregoing embodiments.

[0220] Embodiments of this disclosure also provide a readable storage medium that, when instructions in the storage medium are executed by a processor of an electronic device, enables the electronic device to perform the training method for the traffic restriction information extraction model of the foregoing embodiments.

[0221] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0222] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of this disclosure are not directed to any particular programming language. It should be understood that the embodiments of this disclosure described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the embodiments of this disclosure.

[0223] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this disclosure may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0224] Similarly, it should be understood that, in order to simplify this disclosure and aid in understanding one or more of the various inventive aspects, in the foregoing description of exemplary embodiments of this disclosure, various features of embodiments of this disclosure are sometimes grouped together in a single embodiment, figure, or description thereof. However, this approach to disclosure should not be construed as reflecting an intention that the claimed embodiments of this disclosure require more features than expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the present disclosure.

[0225] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature serving the same, equivalent, or similar purpose.

[0226] The various component embodiments of this disclosure can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the motion picture generation device according to the embodiments of this disclosure. Embodiments of this disclosure can also be implemented as device or apparatus programs for performing some or all of the methods described herein. Such programs implementing embodiments of this disclosure can be stored on a computer-readable medium or can take the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0227] It should be noted that the above embodiments are illustrative of embodiments of this disclosure and not restrictive of embodiments of this disclosure, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of this disclosure can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0228] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0229] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the embodiments of the present disclosure. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the embodiments of the present disclosure should be included within the protection scope of the embodiments of the present disclosure.

[0230] The above description is merely a specific implementation of the embodiments of this disclosure, but the protection scope of the embodiments of this disclosure is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the embodiments of this disclosure should be included within the protection scope of the embodiments of this disclosure. Therefore, the protection scope of the embodiments of this disclosure should be determined by the protection scope of the claims.

Claims

1. A training method for a traffic restriction information extraction model, characterized in that, The method comprises the following steps: acquiring vehicle driving trajectories at multiple intersections, and multiple driving link pairs respectively corresponding to the multiple intersections; wherein the driving link pair comprises an entry link and an exit link matched with the entry link; performing matching processing on trajectory points of the vehicle driving trajectories and the multiple driving link pairs to generate matched link pair driving trajectories; performing model training based on extracted detour features, link pair topological features and link flow features in the link pair driving trajectories to obtain a traffic restriction information extraction model; after the model training based on the extracted detour features, link pair topological features and link flow features in the link pair driving trajectories to obtain the traffic restriction information extraction model, the method further comprises the following steps: when it is necessary to extract traffic restriction information of a to-be-predicted intersection, acquiring historical vehicle driving trajectories of the to-be-predicted intersection; inputting the historical vehicle driving trajectories into the traffic restriction information extraction model, and acquiring traffic restriction information of the to-be-predicted intersection output by the traffic restriction information extraction model and a confidence degree corresponding to the traffic restriction information; based on the confidence degree and a confidence degree threshold, determining target traffic restriction information in the traffic restriction information.

2. The method of claim 1, wherein, The matching processing on the trajectory points of the vehicle driving trajectories and the multiple driving link pairs to generate matched link pair driving trajectories comprises the following steps: performing differential processing on the trajectory points of the vehicle driving trajectories to generate differential trajectory points corresponding to the vehicle driving trajectories; matching the differential trajectory points with the multiple driving link pairs to obtain matched link pair trajectories; performing preprocessing on the link pair trajectories according to a preset trajectory threshold to obtain the link pair driving trajectories.

3. The method of claim 2, wherein, The differential processing on the trajectory points of the vehicle driving trajectories to generate differential trajectory points corresponding to the vehicle driving trajectories comprises the following steps: sorting the trajectory points based on time stamps corresponding to the trajectory points to obtain sorted trajectory points; performing interpolation processing on the sorted trajectory points to generate the differential trajectory points corresponding to the vehicle driving trajectories.

4. The method of claim 2, wherein, The matching of the differential trajectory points with the multiple driving link pairs to obtain matched link pair trajectories comprises the following steps: determining a target driving link pair in the multiple driving link pairs according to an entry link and an exit link of the differential trajectory points; based on the differential trajectory points and the target driving link pair, determining the link pair trajectories.

5. The method of claim 2, wherein, The preprocessing on the link pair trajectories according to a preset trajectory threshold to obtain the link pair driving trajectories comprises the following steps: acquiring a driving trajectory quantity of the vehicle driving trajectories at the multiple intersections; acquiring initial intersections in the multiple intersections, in which the driving trajectory quantity is greater than a first trajectory quantity threshold; acquiring detour driving trajectories with detour features in the vehicle driving trajectories of the initial intersections; acquiring a ratio of the quantity of the detour driving trajectories to the driving trajectory quantity corresponding to the initial intersections; acquiring target intersections in the initial intersections, in which the ratio is greater than a set ratio threshold; acquire a target link pair of each driving link pair of the target intersection, the number of driving link trajectories of which is greater than a second trajectory number threshold, and take the link pair trajectory corresponding to the target link pair as a link pair driving trajectory.

6. The method of claim 1, wherein, The model training is performed based on the detour features, the link pair topological features and the link flow features in the extracted link pair driving trajectory, to obtain a traffic restriction information extraction model, which includes: inputting the link pair driving trajectory into an initial traffic restriction information extraction model; extracting the detour features, the link pair topological features and the link flow features in the link pair driving trajectory by the initial traffic restriction information extraction model, and outputting a sample precision value corresponding to the link pair driving trajectory by the initial traffic restriction information extraction model according to the detour features, the link pair topological features and the link flow features; acquiring a target link trajectory in the link pair driving trajectory according to the sample precision value; performing differential processing on the target link trajectory and the link pair driving trajectory to generate a differential link trajectory; performing secondary training on the initial traffic restriction information extraction model based on the differential link trajectory, until the traffic restriction information extraction model is obtained.

7. The method of claim 1, wherein, The target traffic restriction information in the traffic restriction information is determined based on the confidence and a confidence threshold, which includes: acquiring traffic restriction information in the traffic restriction information whose confidence is greater than a first confidence threshold, and taking the traffic restriction information whose confidence is greater than the first confidence threshold as the target traffic restriction information; acquiring traffic restriction information in the traffic restriction information whose confidence is less than or equal to the first confidence threshold and greater than a second confidence threshold, to determine the target traffic restriction information in the traffic restriction information whose confidence is less than or equal to the first confidence threshold and greater than the second confidence threshold by a business staff; acquiring traffic restriction information in the traffic restriction information whose confidence is less than the second confidence threshold, and performing iterative training on the traffic restriction information extraction model based on the traffic restriction information whose confidence is less than the second confidence threshold.

8. A training device for a traffic restriction information extraction model, characterized in that, It includes: a vehicle driving trajectory acquisition module, configured to acquire vehicle driving trajectories at a plurality of intersections, and a plurality of driving link pairs corresponding to the plurality of intersections respectively; wherein the driving link pair includes an entry link and an exit link matched with the entry link; a link pair driving trajectory generation module, configured to perform matching processing on trajectory points of the vehicle driving trajectories and the plurality of driving link pairs, to generate matched link pair driving trajectories; a traffic restriction extraction model acquisition module, configured to perform model training based on detour features, link pair topological features and link flow features in the extracted link pair driving trajectory, to obtain a traffic restriction information extraction model; after the model training based on the detour features, the link pair topological features and the link flow features in the extracted link pair driving trajectory to obtain the traffic restriction information extraction model, it further includes: when it is needed to extract traffic restriction information of a to-be-predicted intersection, acquiring historical vehicle driving trajectories of the to-be-predicted intersection; inputting the historical vehicle driving trajectory into the traffic restriction information extraction model, and obtaining traffic restriction information of the to-be-predicted intersection output by the traffic restriction information extraction model, and a confidence degree corresponding to the traffic restriction information; determining target traffic restriction information in the traffic restriction information based on the confidence degree and a confidence degree threshold.

9. An electronic device, comprising: comprise: a processor, a memory, and a computer program stored on the memory and executable on the processor, and the processor implements the training method of the traffic restriction information extraction model in any one of claims 1 to 7 when executing the program.

10. A readable storage medium, characterized by, When the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute the training method of the traffic restriction information extraction model in any one of claims 1 to 7.

Images