Path restoration method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a path restoration method and device, equipment and a storage medium, and relates to the technical field of intelligent transportation, and the specific implementation scheme is as follows: determining a candidate point location group of missing point locations based on a first point location known to be passed by a target vehicle; point location data of each first point location and point location data of each second point location are obtained, and the point location data of the first point locations comprise point location feature data of the first point locations, time data of the target vehicle passing through the first point locations and situation data of the first point locations when the target vehicle passes through the first point locations; the point location data of the second point location comprises point location feature data of the second point location and situation data of the second point location when the target vehicle passes through the first point location; generating path features of each passable path based on the point location data of the point locations in each passable path; and determining a restoration path of the target vehicle from the passable paths based on the path characteristics. By applying the scheme provided by the embodiment of the invention, the driving path can be restored.

Description

Technical Field

[0001] The present application relates to the field of intelligent transportation technology, and in particular to a path restoration method, device, equipment and storage medium. Background Art

[0002] Cameras, ETC (Electronic Toll Collection) antennas and other collection equipment are installed on the roadside or on the gantry. When a vehicle passes through a road with collection equipment, the corresponding vehicle data can be obtained, and the driving path of the vehicle can be determined based on the obtained vehicle data. The determined path can be used to find the location of the vehicle, billing, etc.

[0003] The above-mentioned collection equipment may miss the vehicle data of passing vehicles during operation. For example, the collection equipment may miss the vehicle data due to high-density traffic flow, weather factors, etc., or due to special circumstances such as loss of vehicle-mounted electronic tag information or damage to the collection equipment. In this way, the traffic camera or ETC antenna as the collection equipment cannot obtain the vehicle data of the above-mentioned vehicle. In this case, the driving path determined based on the vehicle data is incomplete.

[0004] This requires a solution to restore the incomplete driving path to determine the road the vehicle actually traveled. Summary of the invention

[0005] The purpose of the embodiments of the present application is to provide a path restoration method, device, equipment and storage medium to achieve driving path restoration. The specific technical solution is as follows:

[0006] The present application provides a path restoration method, the method comprising:

[0007] Based on the first points that the target vehicle is known to have passed, determine a candidate point group of missing points in the driving path of the target vehicle, wherein each first point and the second point in each candidate point group form a passable path for the target vehicle;

[0008] Obtaining point data of each first point and point data of each second point, wherein the point data of the first point includes: point feature data of the first point, time data of the target vehicle passing the first point, and situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes: point feature data of the second point and situation data of the second point when the target vehicle passes the first point, and the situation data is: data characterizing the traffic status of the point;

[0009] Based on the point data of the points in each passable path, the path characteristics of each passable path are generated;

[0010] Based on the path characteristics of each navigable path, a restoration path of the target vehicle is determined from each navigable path.

[0011] The present application also provides a path restoration device, the device comprising:

[0012] A point group determination module is used to determine a candidate point group of missing points in the driving path of the target vehicle based on a first point that the target vehicle is known to have passed, wherein each first point and a second point in each candidate point group form a passable path for the target vehicle;

[0013] The point data acquisition module is used to obtain the point data of each first point and the point data of each second point, wherein the point data of the first point includes: the point feature data of the first point, the time data of the target vehicle passing the first point and the situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes: the point feature data of the second point and the situation data of the second point when the target vehicle passes the first point, and the situation data is: data characterizing the traffic status of the point;

[0014] A path feature generation module, used to generate path features of each passable path based on the point data of each passable path;

[0015] The restoration path determination module is used to determine the restoration path of the target vehicle from each passable path based on the path characteristics of each passable path.

[0016] An embodiment of the present application also provides an electronic device, including a memory for storing computer programs; and a processor for implementing the above-mentioned path restoration method when executing the program stored in the memory.

[0017] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium has a computer program, and when the computer program is executed by a processor, the above-mentioned path restoration method is implemented.

[0018] The embodiment of the present application also provides a computer program product including instructions, which, when executed on a computer, enables the computer to execute the above-mentioned path restoration method.

[0019] Beneficial effects of the embodiments of the present application:

[0020] In the solution provided in the embodiment of the present application, the point data of the second point of each candidate point group and the point data of the first point that the target vehicle is known to have passed are determined to generate the path characteristics of each traversable path. Thus, a traversable path is selected as the restoration path according to the path characteristics of each traversable path, thereby realizing the restoration of the driving path. In the above process, the point feature data of the first point, the time data of the target vehicle passing the first point, and the situation data of the first point when the target vehicle passes the first point are used, and the point feature data and situation data of the second point are used. When determining the traversable path for path restoration, the reference data is relatively rich and comprehensive, so that the obtained path features are relatively accurate, and accordingly, the accuracy of the driving path restoration is also improved.

[0021] Of course, implementing any product or method of the present application does not necessarily require achieving all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application, and for ordinary technicians in this field, other embodiments can also be obtained based on these drawings.

[0023] Figure 1 A schematic diagram of a first path restoration method provided in an embodiment of the present application;

[0024] Figure 2 A schematic diagram of the first point distribution provided in the embodiment of the present application;

[0025] Figure 3 A schematic diagram of the second point distribution provided in the embodiment of the present application;

[0026] Figure 4 A schematic diagram of a second path restoration method provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of a path restoration process provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of a time prediction process provided in an embodiment of the present application;

[0029] Figure 7 A schematic diagram of the structure of a path restoration device provided in an embodiment of the present application;

[0030] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field based on the present application belong to the scope of protection of the present application.

[0032] In order to restore the driving path of a vehicle, an embodiment of the present application provides a path restoration method, an apparatus, a device and a storage medium, which are described below respectively.

[0033] First, some concepts involved in the embodiments of the present application are explained.

[0034] 1. Point

[0035] The point represents the location of a collection device deployed on the road for collecting vehicle information. The above-mentioned collection device can be a camera, a radio frequency reader, etc.

[0036] 2. Bayonet

[0037] The road area where the above-mentioned collection equipment is deployed can be called a checkpoint.

[0038] For example, a road area where an ETC toll station equipped with a radio frequency reader is located can be called a checkpoint, and a road area where a surveillance camera is installed can be called a checkpoint, etc.

[0039] It can be seen that the position of the collection equipment deployed at the checkpoint is the above-mentioned point, that is, there is a corresponding relationship between the checkpoint and the point.

[0040] 3. Adjacent points, upstream points, and downstream points

[0041] The points corresponding to two acquisition devices arranged adjacently on the road can be called adjacent points.

[0042] For a road, the corresponding points of each collection device can be sorted according to their distribution in the travel direction of the road. The point adjacent to a point and ranked higher can be called the upstream point of the point; similarly, the point adjacent to and ranked lower can be called the downstream point of the point.

[0043] It should be noted that if the road is bidirectional, you can choose either direction.

[0044] The path restoration method provided in the embodiment of the present application is described below.

[0045] See also Figure 1, is a flow chart of a first path restoration method provided in an embodiment of the present application, the method comprising the following steps S101-S104.

[0046] Step S101: Based on a first point that the target vehicle is known to have passed, a candidate point group of missing points in the driving path of the target vehicle is determined.

[0047] Each first point and the second point in each candidate point group form a passable path for the target vehicle.

[0048] The first point and the missing point are explained below respectively.

[0049] For the first point:

[0050] In the application scenario of the embodiment of the present application, each collection device can continuously collect vehicle information of passing vehicles and send the collected vehicle information to the electronic device serving as the executor of this solution, so that the electronic device can determine the vehicle corresponding to the vehicle information collected by each collection device based on the received vehicle information, and then determine the point corresponding to the collection device that collects the vehicle information of the target vehicle as the first point.

[0051] The electronic device may determine the vehicle corresponding to the vehicle information collected by the collection device based on the vehicle identification included in the received vehicle information. The vehicle identification may be a license plate number or the like.

[0052] Specifically, if the acquisition device is a camera, the vehicle information received by the electronic device may be an image taken by the acquisition device, and then the license plate number contained in the image may be identified based on the image recognition algorithm, and the vehicle may be identified based on the license plate number; if the acquisition device is a radio frequency reader, the vehicle information received by the electronic device may directly include the vehicle identification read by the radio frequency reader for the electronic tag set in the vehicle, and then the vehicle may be identified based on the vehicle identification.

[0053] For missing points:

[0054] Missing points can be understood as points where the target vehicle passed but the collection equipment failed to successfully collect the corresponding vehicle information.

[0055] Specifically, based on the above description of the first point, when the target vehicle is traveling, the point corresponding to the collection device that collects the vehicle information of the target vehicle is the first point. The execution subject can obtain the vehicle information collected by the collection device corresponding to the first point, so the first point can also be called a known point in the driving path of the target vehicle.

[0056] However, in cases of bad weather, loss of vehicle-mounted electronic tag information, or damage to the collection device, the collection device will not successfully collect vehicle information when the target vehicle passes by the collection device, that is, the vehicle information will be missed. At this time, the executing entity cannot obtain the vehicle information of the above-mentioned collection device. For the executing entity, the point corresponding to the above-mentioned collection device is unknown or missing. Therefore, the point corresponding to the above-mentioned collection device can be called the missing point in the driving path of the target vehicle.

[0057] The above-mentioned candidate point group is further described.

[0058] The candidate point group is a set of missing points included in a possible path of the target vehicle passing through the first point. The missing points in the candidate point group can be referred to as second points.

[0059] It is understandable that there may be multiple possible paths passing through the first point, and therefore there may be multiple candidate point groups.

[0060] Furthermore, the second point included in the candidate point group may be a point belonging to one road section or may be a point belonging to multiple road sections.

[0061] Combine the following Figure 2 , a more intuitive explanation of the above candidate point group is given.

[0062] See also Figure 2 , which is a schematic diagram of the first point distribution provided in an embodiment of the present application.

[0063] Figure 2 The solid circle p1-p3 represents the first point, the dotted circle p4-p6 represents the second point, and the line segment between the points represents the path that the target vehicle passes through. Figure 2 It can be seen that there are four possible paths for the target vehicle to pass through the first point, namely P1: p1→p4→p2→p3, P2: p1→p4→p2→p6→p3, P3: p1→p5→p2→p3 and P4: p1→p5→p2→p6→p3.

[0064] Among them, for possible path P1, point p4 other than the first point in the path is the missing point in the path, and the candidate point group corresponding to P1 is [p4]. At this time, the second point p4 included in the selected point group belongs to one section; for possible path P2, points p4 and p6 other than the first point in the path are the missing points in the path, and the candidate point group corresponding to P2 is [p4, p6]. At this time, the second points p4 and p6 included in the selected point group belong to two different sections. The candidate point groups corresponding to other possible paths are similar, and no longer given examples one by one.

[0065] The following is an introduction to the method of determining the candidate point group.

[0066] Specifically, the candidate point group can be determined based on the road network topology information.

[0067] Among them, the above-mentioned road network topology information may include the location information of each collection device set up on the roads in the preset area, and the above-mentioned location information may include the latitude and longitude coordinates, the checkpoint number of the checkpoint where it is located, etc.

[0068] It can be seen that based on the above location information contained in the road network topology information, the positions of each point on the road can be determined.

[0069] In one implementation, possible paths passing through the first point may be determined based on the road network topology information, and then for each possible path, a set of points in the possible path except the first point may be determined as a candidate point group.

[0070] A method for determining possible paths passing through the first point is introduced below.

[0071] For example, based on the road network topology information, the road to which each first point belongs can be determined, and then based on the road to which each first point belongs, the path to which adjacent first points belong can be determined, and then the section between adjacent first points in the determined road can be determined as a possible sub-section between the two. In this way, the possible sub-sections between each pair of adjacent first points can be determined, and then based on the combination of adjacent sub-paths, the possible path composed of each sub-section can be obtained.

[0072] Combine the following Figure 3 Provide a more intuitive explanation.

[0073] See also Figure 3 , which is a schematic diagram of the second point distribution provided in an embodiment of the present application.

[0074] Among them, p1-p3 is the first point, p4-p6 is the second point, Figure 3 The solid line and the dashed line represent road A and road B, respectively.

[0075] It can be seen that for the adjacent first points p1 and p2, p1 belongs to section A and section B, and p2 also belongs to section A and section B. Then the section between p1 and p2 in road A (p1→p4→p2) is a possible sub-section, and the section between p1 and p2 in road B (p1→p5→p2) is also a possible sub-section. Similarly, for the adjacent first points p2 and p3, two possible sub-sections can also be determined: (p2→p6→p3) and (p2→p3). Therefore, by randomly combining the possible sub-paths between the adjacent first points p1 and p2 and the adjacent first points p2 and p3, four possible paths can be obtained: p1→p4→p2→p6→p3, p1→p4→p2→p3, p1→p5→p2→p6→p3, and p1→p5→p2→p3.

[0076] In this way, the set of points except the first point in the possible path can be determined as a candidate point group, such as the set of points [p4, p6] except the first point p1-p3 in the possible path p1→p4→p2→p6→p3 is a candidate point group.

[0077] In another implementation, based on the previous implementation, after determining each possible sub-segment, the road network topology information can be combined to determine whether the possible sub-segments are connected, and then the interconnected possible sub-segments can be combined to obtain a possible path, so that the set of points in the possible path except the first point can be determined as candidate points.

[0078] Combine the following Figure 3 A more intuitive explanation of how to obtain possible paths is given.

[0079] It is assumed that the road A shown by the solid line and the road B shown by the dotted line are not connected to each other, for example, the road A and the road B are not connected ramps, interchange roads that are not at the same vertical height, etc.

[0080] From the aforementioned Figure 3 It can be seen from the description that, for the adjacent first points p1 and p2, the section between p1 and p2 in road A (p1→p4→p2) and the section between p1 and p2 in road B (p1→p5→p2) are possible sub-sections; for the adjacent first points p2 and p3, two possible sub-sections can also be determined in road A and road B respectively: (p2→p3) and (p2→p6→p3).

[0081] Since Road A and Road B are not connected to each other, there is no connectivity between the possible sub-segments (p1→p4→p2) in Road A and the possible sub-segments (p2→p6→p3) in Road B, and they cannot be combined into a possible path; there is no connectivity between the possible sub-segments (p1→p5→p2) in Road B and the possible sub-segments (p2→p3) in Road A, and they cannot be combined into a possible path. Therefore, we can only combine the possible sub-segments (p1→p4→p2) and (p2→p3) in Road A to obtain the possible path: p1→p4→p2→p3, and combine the possible sub-segments (p1→p5→p2) and (p2→p6→p3) in Road B to obtain the possible path: p1→p5→p2→p6→p3.

[0082] Step S102: obtaining point data of each first point and point data of each second point.

[0083] Among them, the point data of the first point includes: the point feature data of the first point, the time data of the target vehicle passing the first point and the situation data of the first point when the target vehicle passes the first point; the point data of the second point includes: the point feature data of the second point and the situation data of the second point when the target vehicle passes the first point; the situation data characterizes the traffic status and traffic trend of the point.

[0084] The point data is data recording point-related information, and the point data may include multiple different types of point-related data, and different types of point data record different information. The following first describes each type of data included in the point data of the first point.

[0085] 1. Point feature data of the first point

[0086] The point feature data is data that describes the characteristics of the point, and may include at least two of the attribute data of the point, the passing time data of the point, the vehicle data of the target vehicle, and the situation data of the point when the target vehicle passes the point. This allows the point feature data to contain different types of data related to the point, so that the point feature data can represent more information.

[0087] The following is an introduction to each type of data that can be included in point feature data.

[0088] Attribute data for points:

[0089] The attribute data of a point may include at least one of the following information: a geographical location corresponding to the point, a point identifier, and topological information of the point in the road network.

[0090] The above-mentioned geographical location indicates the geographical location of the collection device corresponding to the point, which can be the absolute location of the collection device, such as longitude and latitude coordinates, etc.; it can also be the relative location of the collection device, such as the number of the collection device on the road, etc.

[0091] The above-mentioned point identification can be any identification such as the identification of the collection device corresponding to the point, the number of the collection device corresponding to the point, the pile number of the checkpoint corresponding to the point, etc.

[0092] The topological information of the above-mentioned point in the road network may be the point identifiers of the upstream point and the downstream point of the point.

[0093] The passing time data for the point:

[0094] The vehicle passing time data of a point may include at least one of the following information:

[0095] The time characteristics of the time when a vehicle passes through the point in a preset first time unit, and the time characteristics of the time when a vehicle passes a preset distance in the road section where the point is located.

[0096] Among them, the first time unit is a preset time period, which can be one day, one hour, etc.; the preset distance can be determined manually, for example, it can be 50 meters, 100 meters, etc., and the embodiment of the present application is not limited to this.

[0097] Specifically, the passing time of each vehicle passing the point in the first time unit can be obtained as the time characteristic of the vehicle's passing time in the first time unit; and for each vehicle passing the point, the time consumed by the vehicle to pass a preset distance on the path where the point is located can be determined, as the time characteristic of the time required for the vehicle to pass the preset distance in the road section where the point is located.

[0098] Vehicle data for the target vehicle:

[0099] The above-mentioned vehicle data may include at least one of the following information: license plate number, historical time when the target vehicle passes through the point, and historical trajectory data of the target vehicle.

[0100] Since the collection device can record the vehicle identification and time of the vehicle passing through the collection device, the point corresponding to the collection device passed by the target vehicle and the historical time when it passed the above point can be determined based on the information recorded by each collection device.

[0101] In this way, the historical trajectory data of the target vehicle can be determined based on each point passed by the target vehicle and the historical time when the target vehicle passed each point.

[0102] Situation data of the point when the target vehicle passes through the point:

[0103] Situation data is data reflecting the traffic status and traffic trend of a point, and specifically may be data reflecting the flow characteristics of vehicles passing through the point, and may include data on vehicle speed and vehicle quantity.

[0104] In one case, the situation data may include at least one of the following information: the flow rate and vehicle speed of the point within the time period to which the target vehicle passes the point.

[0105] Among them, the flow rate of a point within a certain time period can be the number of vehicles passing through the point, and the vehicle speed of a point within a certain time period can be a representation value of the instantaneous speed of each vehicle passing through the point. The above representation value can be an average value, a maximum value, etc.

[0106] The above-mentioned various types of data may be pre-stored in a collection device and sent to an electronic device by the collection device; or may be pre-stored in an electronic device.

[0107] From the above, it can be seen that the solution provided in the embodiment of the present application can obtain the attribute data of the point, the passing time data of the point, the vehicle data of the target vehicle and the situation data of the point when the target vehicle passes the point based on information from multiple dimensions, so that the point feature data obtained based on the above-mentioned attribute data, passing time data, vehicle data and situation data contains richer and more comprehensive information, which makes the information used for reference in the subsequent restoration path obtained based on the obtained point features richer, which is conducive to obtaining a more accurate restoration path.

[0108] From the above content, it can be known that there can be multiple data used to obtain point feature data, such as point attribute data, vehicle passing time data, and vehicle data. In one case, the above data can be converted into discrete vectors, and the above point feature data can be obtained based on the above discrete vectors and the point prediction model. The specific implementation method is detailed in the subsequent embodiments, which will not be described in detail here.

[0109] 2. Time data of the target vehicle passing the first point

[0110] The collection device corresponding to the first point can record the collection time of the vehicle information after collecting the vehicle information, so that the above collection time can be obtained as the time data when the target vehicle passes the first point.

[0111] The above-mentioned collection time may include the collection date, the collection time accurate to the second, etc.

[0112] 3. Situation data of the first point when the target vehicle passes the first point

[0113] The above situation data may include the instantaneous speed of the target vehicle when passing the point, the average speed of vehicles passing the point within a preset time period before the target vehicle passes the first point, the number of vehicles passing within the above preset time period, etc.

[0114] The following describes each type of data included in the point data of the second point.

[0115] 1. The point feature data of the second point

[0116] The point feature data of the second point is similar to the point feature data of the first point, and the only difference is the corresponding point type, which will not be repeated here.

[0117] 2. Situation data of the second point when the target vehicle passes the first point

[0118] Since the second point is a point that has not yet been determined to be actually passed by the target vehicle, the time when the target vehicle passes the second point cannot be determined. Considering that the first point and the second point are both on a passable path of the target vehicle, the time and position are relatively close. If the target vehicle actually passes the second point, the situation data of the second point when the target vehicle passes the first point can also better fit the situation data of the second point when the target vehicle passes the second point. Therefore, when selecting the situation data for the second point, the time corresponding to the situation data is the time when the target vehicle passes the first point.

[0119] The situation data of the second point is similar to the situation data of the second point, the only difference is that the corresponding point types are different, which will not be repeated here.

[0120] Step S103: Generate path features of each traversable path based on the point data of each traversable path.

[0121] In the solution provided in the embodiment of the present application, the traversable path includes various points, each of which corresponds to point data of the point, so that the path characteristics of the traversable path can be generated based on the point data of each point.

[0122] Specifically, the path features of each passable path may be generated in the following manner.

[0123] In one implementation, for each of the first point and the second point, data included in the point data of the point can be fused to obtain the fused data of the point, and then the fused data of the points in each passable path can be spliced ​​according to the order of the points in each passable path to obtain the path characteristics of each passable path.

[0124] For details on the specific implementation method, please refer to the following Figure 4Steps S403 to S404 in the illustrated embodiment are not described in detail here.

[0125] In another implementation, the data included in each point data may be fused according to the order of the points in each passable path according to the data category to obtain fused data of each category of data, and then the fused data of each category of data may be spliced ​​to obtain the path characteristics of the passable path.

[0126] For example, the point feature data of the first point and the second point are fused to obtain the first fused data, the time data of the first point is fused to obtain the second fused data, the situation data of the first point and the second point are fused to obtain the third fused data, and then the first fused data, the second fused data and the third fused data are spliced ​​to obtain the path characteristics of the passable path.

[0127] Step S104: Based on the path characteristics of each navigable path, determine the restoration path of the target vehicle from each navigable path.

[0128] Each traversable path is a possible path for the target vehicle when it passes the first point. In this step, the path actually passed by the target vehicle can be determined from the traversable paths. The above path can be called a restored path.

[0129] In one embodiment, for each passable path, based on the path characteristics of the passable path, the probability that the second point in the candidate point group corresponding to the passable path is the point through which the target vehicle passes is obtained, and according to the probabilities corresponding to each passable path, the restored path of the target vehicle is determined from each passable path.

[0130] First, a method for obtaining the probability that the second point is a point where the target vehicle passes is described.

[0131] Among them, for each passable path, the point data of the second point in the candidate point can be input into the pre-trained first path restoration model to obtain the probability that the second point output by the above model is the point where the target vehicle passes.

[0132] The above-mentioned model may be a model trained by taking the second point of the sample in the sample candidate point group as input and whether the second point of the sample is a passing point of the target vehicle as a training annotation.

[0133] If the second point of the sample is the passing point of the target vehicle, it is marked as 1; if the second point of the sample is not the passing point of the target vehicle, it is marked as 0.

[0134] After obtaining the probabilities corresponding to the various traversable paths, the restoration path of the target vehicle may be determined from the various traversable paths in the following manner.

[0135] In the first method, the sum of the probabilities that the second point included in each passable path is the point through which the target vehicle passes can be calculated, and the passable path with the highest corresponding sum value can be determined as the restoration path.

[0136] The second method is to calculate the average value of the probability that the second point included in each passable path is the point where the target vehicle passes, and determine the passable path with the highest corresponding average value as the restoration path.

[0137] The third method is to set corresponding weights for the second points of different pass orders in the passable paths, and then perform weighted calculation on the probability that the second point included in each passable path is the passing point of the target vehicle according to the above weights, and the probability that the second point included in each passable path with the largest corresponding calculation result is the passing point of the target vehicle.

[0138] For example, a higher weight is set for the second point at the front of the path sequence, and a lower weight is set for the second point at the back of the path sequence, and so on.

[0139] The sequence of the pathways is described below.

[0140] Since the position and passing time of the first points are known, the order in which the vehicles pass through the first points can be determined, and the travel direction of the vehicles on the road when passing through the first points can be determined, so based on the position distribution of the first points and the second points in the passable path along the travel direction, the order in which the vehicles pass through the second points in the passable path can be determined. In this way, the passage order of each point in the passable path is determined.

[0141] It can be seen that the probability that the second point in the candidate point group corresponding to the passable path is the point where the target vehicle passes is first obtained. In this way, the restoration path can be determined from the passable paths according to the probabilities corresponding to each passable path. For example, the passable path with the highest probability of the target vehicle passing can be determined from all the passable paths as the restoration path, which improves the accuracy of the determined restoration path.

[0142] In another implementation, the path features of each traversable path may be input into a pre-trained second path restoration model to obtain a restored path output by the above model.

[0143] The above-mentioned model may be a model obtained by training using the path features of the sample path as input and whether the sample path is the actual path passed by the target vehicle as a training annotation.

[0144] As can be seen from the above, when the solution provided in the embodiment of the present application is used to perform path restoration, a candidate point group of missing points in the target vehicle's driving path can be determined based on the first point that the target vehicle is known to have passed. Since each first point and the second point in each candidate point group form a passable path for the target vehicle, the path characteristics of each path can be generated based on the point data of the first point and the second point, and then the restored path of the target vehicle can be determined from each passable path based on the path characteristics of each passable path, thereby realizing the restoration of the vehicle's driving path.

[0145] Among them, the point data of the first point includes point feature data, time data of the target vehicle passing the first point, and situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes point feature data and situation data of the second point when the target vehicle passes the first point. It can be seen that the point data of the first point and the second point both contain data of multiple dimensions associated with the points, so that the point data of the first point and the second point can more comprehensively represent various information related to the points, so that the path characteristics of the traversable path generated based on the point data of the first point and the second point can represent more comprehensive and rich path information, so that the information referenced when determining the restored path based on the above path characteristics is richer and more comprehensive, and the accuracy of the determined restored path characteristics is improved.

[0146] The following steps A to C are used to illustrate the method of obtaining the above-mentioned point feature data based on discrete vectors and point prediction models.

[0147] Step A: Convert each data used to generate point data into a discrete vector.

[0148] The data used to generate the point data may be the attribute data of the point, the passing time data of the point, the vehicle data of the target vehicle, and the situation data of the point when the target vehicle passes the point.

[0149] Specifically, each data used to generate the point data may be encoded based on an encoding method such as one-hot encoding to obtain a discrete vector representing each data.

[0150] Step B: splicing the obtained discrete vectors to obtain spliced ​​data.

[0151] Before concatenating the discrete vectors, the discrete vectors may be subjected to dimensionality reduction processing. For example, an embedding method may be used to obtain low-dimensional vectors of each discrete vector, and then the obtained low-dimensional vectors are concatenated to obtain concatenated data.

[0152] Step C: Input the spliced ​​data into a pre-trained point prediction model for point prediction, and obtain the target features for point prediction generated by the point prediction model based on the spliced ​​data during the point prediction process as feature data.

[0153] Among them, the point prediction model is: a prediction model trained by a cyclic prediction method using the description data of the third point in the sample driving path of the vehicle as input data and the fourth point in the sample driving path as supervision information, the third point and the fourth point are partial points in the sample driving path, and the description data of the third point is data used to generate feature data of the third point.

[0154] During the inference process, the above model generates target features for point prediction based on the spliced ​​data, and predicts other points in the path based on the target features. Therefore, the above target features can be obtained as feature data of the point.

[0155] It can be seen from the above that the various data used to generate point data are converted into discrete vectors, and the discrete vectors are spliced ​​to obtain spliced ​​data. After the spliced ​​data is input into the point prediction model, the target features generated by the model during the reasoning process for point prediction can be easily obtained, thereby using the above target features as the feature data of the point, thereby improving the efficiency of obtaining the feature data of the point.

[0156] In addition, the above-mentioned model is a point prediction model trained based on point data. The target features used for point prediction generated in the reasoning process have strong representation capabilities and a more standardized format. Therefore, using the target features as feature data of the point can improve the data quality of the obtained feature data.

[0157] In one embodiment of the present application, after determining the restoration path of the target vehicle, it is also possible to predict the time when the target vehicle passes through each second point in the restoration path based on the point data of the upstream point of the second point in the first point, the point data of the downstream point, and the point data of the second point.

[0158] The above time may be the number of minutes of the elapsed time relative to the whole day, such as the 300th minute, the 400th minute, etc.

[0159] Specifically, the time when the target vehicle passes the second point can be predicted based on the following method.

[0160] In one embodiment, the time when the target vehicle passes the second point can be predicted by a neural network based on the point data of the upstream point, the point data of the downstream point and the point data of the second point. Figure 6 The illustrated embodiment will not be described in detail here.

[0161] In another embodiment, based on the point data of the upstream point and the downstream point of the second point, the first time and the first speed of the target vehicle passing the upstream point, as well as the second time and the second speed of the target vehicle passing the downstream point can be determined, and the speed change trend of the target vehicle between the upstream point and the downstream point can be obtained based on the above information obtained, and the time when the target vehicle passes the second point can be predicted based on the obtained speed change trend and the distance between the second point and the upstream and downstream points.

[0162] The above speed change trend may be acceleration, jerk, etc., which is not limited in the embodiments of the present application.

[0163] For example, based on the above-mentioned first time, first speed, second time and second speed, the acceleration of the target vehicle between the upstream point and the downstream point can be predicted. Based on the determined acceleration, first speed and the distance between the second node and the upstream node, the time when the target vehicle passes the second point can be determined by solving the variable acceleration motion problem.

[0164] As can be seen from the above, after obtaining the restored path in this way, the time when the target vehicle passes the second point can be further predicted based on the point data of the upstream and downstream points of the second point in the restored path and the point data of the second point, so that the information represented by the restored path is not only the position of the second point, but also the time when the target vehicle passes the second point, thereby increasing the amount of information represented by the restored path.

[0165] In addition, when predicting the above time, the point data of the upstream and downstream points are combined for prediction. Since the upstream and downstream points are closely related to the second point, the point data of the upstream and downstream points can more accurately predict the time when the target vehicle passes the second point.

[0166] In one case, if the reference point of the second point does not exist in the first point, the point data of the reference point is determined to be the point data of a preset virtual point, and the time when the target vehicle passes the second point is predicted based on the point data of the reference point and the point data of the second point.

[0167] Among them, the reference point is: upstream point or downstream point.

[0168] For example, if the second point in the candidate point group is located at the beginning or end of the road where the drivable path is located, in this case, there is no upstream point or downstream point for the second point, that is, there is no reference point for the second point. At this time, the time when the target vehicle passes the second point can be predicted based on the reference point.

[0169] The virtual point may be set by the staff based on experience, for example, it may be a point that is a preset data distance from the second point.

[0170] It should be noted that in the point data of the virtual point, the situation data can be set to empty.

[0171] In this case, the specific method of predicting the time when the target vehicle passes the second point is similar to the method of predicting the above time based on the point data of the upstream and downstream points introduced in the previous embodiment. The only difference is that the missing upstream point or downstream point is replaced by a reference point, which will not be repeated here.

[0172] It can be seen from the above that by setting a virtual point, when there are no upstream or downstream points at the second point, the time for the target vehicle to pass the point can be normally predicted based on the point data of the virtual point, thereby improving the applicability of the predicted time of the present application scheme.

[0173] The following combination Figure 4 The flowchart of the second driving path restoration method shown in FIG. 1 is a detailed description of step S103. In this embodiment, steps S403-S404 are the implementation methods of step S103.

[0174] In one embodiment of the present application, see Figure 4 , a flow chart of a driving path restoration method is provided, the method comprising the following steps S401-S405.

[0175] Step S401: Based on a first point that the target vehicle is known to have passed, a candidate point group of missing points in the driving path of the target vehicle is determined.

[0176] Step S402: Obtaining point data of each first point and point data of each second point.

[0177] The above steps S401-S402 are the same as the above steps Figure 1 In the illustrated embodiment, steps S101 and S102 are the same and will not be described in detail here.

[0178] Step S403: for each of the first point and the second point, data fusion is performed on the data included in the point data of the point to obtain fused data of the point.

[0179] In this step, feature encoding may be performed on the data included in the point data to obtain data features, and then feature fusion may be performed on the obtained data features, and the obtained fusion results may be used as fusion data of the point.

[0180] Among them, the above-mentioned feature encoding can be implemented based on encoding methods such as one-hot encoding; the above-mentioned feature fusion can be implemented by splicing, superposition, weighted summation, etc. of the features. Of course, if the dimension of the data features obtained by encoding is too high, the data features can be reduced in dimension through pooling operation and then fused. The embodiments of the present application are not limited to this.

[0181] The following describes the method of fusing the data included in the point data of the first point and the second point respectively.

[0182] For the first point, the point feature data of the first point, the time data when the target vehicle passes the first point, and the data features corresponding to the situation data of the first point when the target vehicle passes the first point are obtained, and the data features are fused, and the fusion result is the fused data.

[0183] For the second point, the point feature data of the second point and the data features corresponding to the situation data of the second point when the target vehicle passes the first point are taken, and the data features are fused, and the fusion result is the fused data.

[0184] Step S404: according to the order of the points in each passable path, the fused data of the points in each passable path are spliced ​​to obtain the path characteristics of each passable path.

[0185] The order of points refers to the order in which the points on the traversable path are passed when the vehicle travels along the traversable path.

[0186] In this way, the fused data of each point can be spliced ​​according to the order of each point to obtain the path features.

[0187] Step S405: Based on the path characteristics of each traversable path, determine the restoration path of the target vehicle from each traversable path.

[0188] The above step S405 is the same as the above Figure 1 Step S104 in the illustrated embodiment is the same and will not be described again here.

[0189] As can be seen from the above, in the solution provided by the embodiment of the present application, the point data of each point is fused, and the fused data obtained by splicing is used to obtain the path characteristics of each pass path, so that the path characteristics reflect the information represented by the multiple point data of multiple points, and the information used is more comprehensive and more effective in characterizing the characteristics of the path. In addition, the information loss when obtaining the path characteristics based on the above data fusion and data splicing method is small and the efficiency is high.

[0190] Combine the following Figure 5 , this paper provides an overall description of a path restoration process provided in an embodiment of the application.

[0191] See also Figure 5 , which is a schematic diagram of a path restoration process provided in an embodiment of the present application.

[0192] exist Figure 5 In the path restoration process shown, the vehicle ID, historical trajectory characteristics, historical situation characteristics, and road network topology characteristics are all relevant information of the first point.

[0193] The following describes the processing of the first point related information and the processing of the second point related information in the candidate point group respectively.

[0194] For the first point:

[0195] First, based on the vehicle ID, the license plate Embedding (embedded feature) of the first point can be obtained, based on the historical trajectory features, the position Embedding and time Embedding of the first point can be obtained, based on the historical situation features, the speed Embedding and flow Embedding of the first point can be obtained, and then, the above Embeddings are input into the hierarchical embedding layer (Hierarchy Embedding Layer) for feature dimension reduction and feature fusion processing to obtain the point feature data, situation data and time data of the first point, and the obtained point feature data, situation data and time data are input into the RNN pooling layer 2 (RNN Pooling2) for pooling and fusion processing to obtain the fused data of the first point.

[0196] Among them, the point feature data of the first point is obtained based on the road network topology characteristics and position Embedding, the situation data of the first point is obtained based on the above speed Embedding and flow Embedding, and the time data of the first point is obtained based on the above time Embedding.

[0197] For the second point in the candidate point group:

[0198] The point feature data and situation data of the second point are input into the RNN pooling layer 1 (RNN Pooling1) for pooling and fusion processing to obtain the fusion data of the second point.

[0199] The point feature data and situation data of the second point may also be obtained by processing the vehicle ID, historical situation features, road network topology features, etc. corresponding to the second point by the Hierarchy Embedding Layer.

[0200] After processing the relevant information of the first point and the second point, the first fused data and the second fused data can be spliced ​​to obtain the path characteristics of the passable path, so that the obtained path characteristics can be input into MLP (Multilayer Perceptron) for prediction to obtain the probability of whether the second point in the candidate point group belongs to the vehicle passing point, and then the target candidate point group can be determined based on the obtained probability, so that the restored path can be obtained by the first point and the second point of the determined target candidate point group.

[0201] Next, combine Figure 6 , the aforementioned process of predicting the time when the target vehicle passes the second point is generally described.

[0202] See also Figure 6 , which is a schematic diagram of a time prediction process provided in an embodiment of the present application.

[0203] exist Figure 6 In the time prediction process shown, the vehicle ID, historical trajectory characteristics, historical situation characteristics and road network topology characteristics are relevant information of the second point and the upstream and downstream points of the second point.

[0204] The following describes the process of processing the relevant information of upstream and downstream points.

[0205] and Figure 5 Similarly, first, the license plate Embedding (embedded feature) can be obtained based on the vehicle ID, the position Embedding and time Embedding can be obtained based on the historical trajectory features, and the speed Embedding and flow Embedding can be obtained based on the historical situation features. Then, the above Embeddings are input into the Hierarchy Embedding Layer for feature dimensionality reduction processing to obtain upstream point-to-point feature data, downstream point-to-point feature data, upstream point situation data, downstream point situation data, and time data. The obtained time data is input into the RNN layer for pooling processing to obtain the time fusion data of the obtained time data.

[0206] Among them, the point feature data of the upstream and downstream points are obtained based on the road network topology characteristics and position Embedding, the situation data of the upstream and downstream points are obtained based on the above speed Embedding and flow Embedding, and the time data of the upstream and downstream points are obtained based on the above time Embedding.

[0207] The following describes the process of processing the information related to the second point.

[0208] The processing is similar to the above processing for the upper and lower points, except that the second point does not contain time embedding, which will not be described here. In this way, the feature data of the second point and the situation data of the second point can be obtained.

[0209] Then, the point feature data of the upstream point, the point feature data of the downstream point, the upstream point situation data, the downstream point situation data, the second point point feature data, and the second point situation data can be fused to obtain the third fused data, and the upstream point time data and the downstream point time data can be input into the RNN layer for pooling and fusion processing to obtain the fourth fused data. The obtained third fused data and the fourth fused data can be input into the MLP layer for time prediction to obtain the time when the target vehicle passes the second point.

[0210] Depend on Figure 5 and Figure 6 As can be seen from the processes shown, the Hierarchy Embedding Layer in both processes is used to perform feature dimensionality reduction and feature fusion processing on various data of the points to obtain point feature data, situation data and other data of the points. That is, in both processing flows, the data processing results of the Hierarchy Embedding Layer can be shared, thereby improving the execution efficiency of the path restoration process and the time prediction process.

[0211] Corresponding to the above method embodiment, the embodiment of the present application also provides a path restoration device.

[0212] See also Figure 7 , an embodiment of the present application provides a schematic diagram of the structure of a path restoration device, the device comprising:

[0213] The point group determination module 701 is used to determine a candidate point group of missing points in the driving path of the target vehicle based on the first points that the target vehicle is known to have passed, wherein each first point and the second point in each candidate point group form a passable path for the target vehicle;

[0214] The point data acquisition module 702 is used to obtain the point data of each first point and the point data of each second point, wherein the point data of the first point includes: the point feature data of the first point, the time data of the target vehicle passing the first point, and the situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes: the point feature data of the second point and the situation data of the second point when the target vehicle passes the first point, and the situation data represents the traffic state and traffic trend of the point;

[0215] A path feature generation module 703, for generating path features of each passable path based on the point data of each passable path;

[0216] The restoration path determination module 704 is used to determine the restoration path of the target vehicle from each passable path based on the path characteristics of each passable path.

[0217] As can be seen from the above, when the solution provided in the embodiment of the present application is used to perform path restoration, a candidate point group of missing points in the target vehicle's driving path can be determined based on the first point that the target vehicle is known to have passed. Since each first point and the second point in each candidate point group form a passable path for the target vehicle, the path characteristics of each path can be generated based on the point data of the first point and the second point, and then the restored path of the target vehicle can be determined from each passable path based on the path characteristics of each passable path, thereby realizing the restoration of the vehicle's driving path.

[0218] Among them, the point data of the first point includes point feature data, time data of the target vehicle passing the first point, and situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes point feature data and situation data of the second point when the target vehicle passes the first point. It can be seen that the point data of the first point and the second point both contain multi-dimensional data associated with the points, so that the point data of the first point and the second point can more comprehensively represent various information related to the points, so that the path characteristics of the traversable path generated based on the point data of the first point and the second point can represent more comprehensive and rich path information, so that the information referenced when determining the restored path based on the above path characteristics is richer and more comprehensive, and the accuracy of the determined restored path characteristics is improved.

[0219] In one embodiment of the present application, the path feature generation module 703 is specifically used to perform data fusion on the data included in the point data of each point in the first point and the second point to obtain the fused data of the point; and splice the fused data of the points in each passable path according to the order of the points in each passable path to obtain the path features of each passable path.

[0220] As can be seen from the above, in the solution provided by the embodiment of the present application, the point data of each point is fused, and the fused data obtained by splicing is used to obtain the path characteristics of each pass path, so that the path characteristics reflect the information represented by the multiple point data of multiple points, and the information used is more comprehensive and more effective in characterizing the characteristics of the path. In addition, the information loss when obtaining the path characteristics based on the above data fusion and data splicing method is small and the efficiency is high.

[0221] In one embodiment of the present application, the restored path determination module 704 is specifically used to obtain, for each passable path, based on the path characteristics of the passable path, the probability that the second point in the candidate point group corresponding to the passable path is the point where the target vehicle passes; and determine the restored path of the target vehicle from each passable path according to the probabilities corresponding to each passable path.

[0222] It can be seen that the probability that the second point in the candidate point group corresponding to the passable path is the point where the target vehicle passes is first obtained. In this way, the restoration path can be determined from the passable paths according to the probabilities corresponding to each passable path. For example, the passable path with the highest probability of the target vehicle passing can be determined from all the passable paths as the restoration path, which improves the accuracy of the determined restoration path.

[0223] In one embodiment of the present application, the point feature data of the point is data generated based on at least two of the following data:

[0224] Attribute data of the point, passing time data of the point, vehicle data of the target vehicle, and situation data of the point when the target vehicle passes the point.

[0225] In this way, the point feature data can contain different types of point-related data, so that the point feature data can represent more information.

[0226] In one embodiment of the present application, the attribute data includes at least one of the following information: a geographical location corresponding to the point, a point identifier, and topological information of the point in the road network; and / or

[0227] The vehicle passing time data includes at least one of the following information: time characteristics of the vehicle passing time of the point in a preset first time unit, time characteristics of the vehicle passing time of a preset distance in the road section where the point is located;

[0228] The vehicle data includes at least one of the following information: a license plate number, a historical time when the target vehicle passes a point, and historical track data of the target vehicle; and / or

[0229] The situation data includes at least one of the following information: the flow rate and vehicle speed of the point within the time period to which the target vehicle passes the point.

[0230] From the above, it can be seen that the solution provided in the embodiment of the present application can obtain the attribute data of the point, the passing time data of the point, the vehicle data of the target vehicle and the situation data of the point when the target vehicle passes the point based on information from multiple dimensions, so that the point feature data obtained based on the above-mentioned attribute data, passing time data, vehicle data and situation data contains richer and more comprehensive information, which makes the information used for reference in the subsequent restoration path obtained based on the obtained point features richer, which is conducive to obtaining a more accurate restoration path.

[0231] In one embodiment of the present application, the feature data of the point is obtained by:

[0232] Convert each data used to generate point data into a discrete vector;

[0233] splicing the obtained discrete vectors to obtain spliced ​​data;

[0234] The spliced ​​data is input into a pre-trained point prediction model for point prediction, and the target features for point prediction generated based on the spliced ​​data during the point prediction process by the point prediction model are obtained as the feature data, wherein the point prediction model is a prediction model trained by a cyclic prediction method using the description data of the third point in the sample driving path of the vehicle as input data and the fourth point in the sample driving path as supervision information, the third point and the fourth point are partial points in the sample driving path, and the description data of the third point is data used to generate feature data of the third point.

[0235] It can be seen from the above that the various data used to generate point data are converted into discrete vectors, and the discrete vectors are spliced ​​to obtain spliced ​​data. After the spliced ​​data is input into the point prediction model, the target features generated by the model during the reasoning process for point prediction can be easily obtained, thereby using the above target features as the feature data of the point, thereby improving the efficiency of obtaining the feature data of the point.

[0236] In addition, the above-mentioned model is a point prediction model trained based on point data. The target features used for point prediction generated in the reasoning process have strong representation capabilities and a more standardized format. Therefore, using the target features as feature data of the point can improve the data quality of the obtained feature data.

[0237] In one embodiment of the present application, the device further includes:

[0238] The first time prediction module is used to predict the time when the target vehicle passes through each second point in the restored path based on the point data of the upstream point, the downstream point and the point data of the second point in the first point.

[0239] As can be seen from the above, after obtaining the restored path in this way, the time when the target vehicle passes the second point can be further predicted based on the point data of the upstream and downstream points of the second point in the restored path and the point data of the second point, so that the information represented by the restored path is not only the position of the second point, but also the time when the target vehicle passes the second point, thereby increasing the amount of information represented by the restored path.

[0240] In addition, when predicting the above time, the point data of the upstream and downstream points are combined for prediction. Since the upstream and downstream points are closely related to the second point, the point data of the upstream and downstream points can more accurately predict the time when the target vehicle passes the second point.

[0241] In one embodiment of the present application, the device further includes:

[0242] The second time prediction module is used to determine the point data of the reference point as the point data of a preset virtual point if there is no reference point of the second point in the first point, and predict the time when the target vehicle passes the second point based on the point data of the reference point and the point data of the second point, wherein the reference point is: an upstream point or a downstream point.

[0243] It can be seen from the above that by setting a virtual point, when there are no upstream or downstream points at the second point, the time for the target vehicle to pass the point can be normally predicted based on the point data of the virtual point, thereby improving the applicability of the predicted time of the present application scheme.

[0244] The present application also provides an electronic device, such as Figure 8 As shown, including:

[0245] Memory 801, used for storing computer programs;

[0246] The processor 802 is used to implement the above path restoration method steps when executing the program stored in the memory 801.

[0247] Furthermore, the electronic device may further include a communication bus and / or a communication interface, and the processor 802, the communication interface, and the memory 801 communicate with each other via the communication bus.

[0248] The communication bus mentioned in the above electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0249] The communication interface is used for communication between the above electronic device and other devices.

[0250] The memory may include a random access memory (RAM) or a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0251] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0252] In another embodiment provided by the present application, a computer-readable storage medium is provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above path restoration methods are implemented.

[0253] In another embodiment provided by the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any path restoration method in the above embodiments.

[0254] In the solutions provided in the embodiments of the present application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are all carried out under the premise of the user's knowledge and authorization, and comply with the relevant laws and regulations and do not violate public order and good morals.

[0255] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or other media (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0256] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0257] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment, and storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0258] The above description is only a preferred embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application are included in the protection scope of the present application.

Claims

1. A path restoration method, characterized in that: The method comprises: Based on the first points that the target vehicle is known to have passed, determine a candidate point group of missing points in the driving path of the target vehicle, wherein each first point and the second point in each candidate point group form a passable path for the target vehicle; Obtaining point data of each first point and point data of each second point, wherein the point data of the first point includes: point feature data of the first point, time data of the target vehicle passing the first point, and situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes: point feature data of the second point and situation data of the second point when the target vehicle passes the first point, and the situation data represents the traffic state and traffic trend of the point; Based on the point data of the points in each passable path, the path characteristics of each passable path are generated; Based on the path characteristics of each navigable path, a restoration path of the target vehicle is determined from each navigable path.

2. The method according to claim 1, characterized in that The generating of path features of each passable path based on the point data of the points in each passable path includes: For each of the first point and the second point, data fusion is performed on the data included in the point data of the point to obtain fused data of the point; According to the order of the points in each passable path, the fused data of the points in each passable path are spliced ​​to obtain the path characteristics of each passable path.

3. The method according to claim 1, characterized in that The determining the restoration path of the target vehicle from each passable path based on the path characteristics of each passable path includes: For each passable path, based on the path characteristics of the passable path, obtain the probability that the second point in the candidate point group corresponding to the passable path is the point where the target vehicle passes; According to the probabilities corresponding to the various traversable paths, the restoration path of the target vehicle is determined from the various traversable paths.

4. The method according to claim 1, characterized in that: The point feature data of a point is data generated based on at least two of the following data: Attribute data of the point, passing time data of the point, vehicle data of the target vehicle, and situation data of the point when the target vehicle passes the point.

5. The method according to claim 4, characterized in that The attribute data includes at least one of the following information: the geographical location corresponding to the point, the point identification, and the topological information of the point in the road network; and / or The vehicle passing time data includes at least one of the following information: time characteristics of the vehicle passing time of the point in a preset first time unit, time characteristics of the vehicle passing time of a preset distance in the road section where the point is located; and / or The vehicle data includes at least one of the following information: a license plate number, a historical time when the target vehicle passes a point, and historical track data of the target vehicle; and / or The situation data includes at least one of the following information: the flow rate and vehicle speed of the point within the time period to which the target vehicle passes the point.

6. The method according to claim 4, characterized in that The characteristic data of the point is obtained by: Convert each data used to generate point data into a discrete vector; splicing the obtained discrete vectors to obtain spliced ​​data; The spliced ​​data is input into a pre-trained point prediction model for point prediction, and the target features for point prediction generated based on the spliced ​​data during the point prediction process by the point prediction model are obtained as the feature data, wherein the point prediction model is a prediction model trained by a cyclic prediction method using the description data of the third point in the sample driving path of the vehicle as input data and the fourth point in the sample driving path as supervision information, the third point and the fourth point are partial points in the sample driving path, and the description data of the third point is data used to generate feature data of the third point.

7. The method according to any one of claims 1 to 6, characterized in that After determining the restoration path of the target vehicle from each passable path based on the path features of each passable path, the method further includes: For each second point in the restored path, the time when the target vehicle passes the second point is predicted based on the point data of the upstream point, the downstream point and the point data of the second point in the first point.

8. The method according to claim 7, characterized in that The method further comprises: If the reference point of the second point does not exist in the first point, the point data of the reference point is determined to be the point data of a preset virtual point, and the time when the target vehicle passes the second point is predicted based on the point data of the reference point and the point data of the second point, wherein the reference point is: an upstream point or a downstream point.

9. A path restoration device, characterized in that: The device comprises: A point group determination module is used to determine a candidate point group of missing points in the driving path of the target vehicle based on a first point that the target vehicle is known to have passed, wherein each first point and a second point in each candidate point group form a passable path for the target vehicle; A point data acquisition module, used to obtain the point data of each first point and the point data of each second point, wherein the point data of the first point includes: point feature data of the first point, time data of the target vehicle passing the first point, and situation data of the first point when the target vehicle passes the first point, and the point data of the second point includes: point feature data of the second point and situation data of the second point when the target vehicle passes the first point, and the situation data represents the traffic status and traffic trend of the point; A path feature generation module, used to generate path features of each passable path based on the point data of each passable path; The restoration path determination module is used to determine the restoration path of the target vehicle from each passable path based on the path characteristics of each passable path.

10. The device according to claim 9, characterized in that The path feature generation module is specifically used to perform data fusion on the data included in the point data of each of the first point and the second point to obtain the fused data of the point; and splice the fused data of the points in each passable path according to the order of the points in each passable path to obtain the path features of each passable path; or The restoration path determination module is specifically used to obtain, for each passable path, the probability that the second point in the candidate point group corresponding to the passable path is the point where the target vehicle passes based on the path characteristics of the passable path; and determine the restoration path of the target vehicle from each passable path according to the probabilities corresponding to each passable path; or The point feature data of the point is data generated based on at least two of the following data: attribute data of the point, time data of the vehicle passing the point, vehicle data of the target vehicle, and situation data of the point when the target vehicle passes the point; or The attribute data includes at least one of the following information: the geographical location corresponding to the point, the point identification, and the topological information of the point in the road network; and / or the vehicle passing time data includes at least one of the following information: the time characteristics of the vehicle passing time of the point in a preset first time unit, and the time characteristics of the time when the vehicle passes a preset distance in the road section where the point is located; and / or the vehicle data includes at least one of the following information: license plate number, historical time when the target vehicle passes through the point, and historical trajectory data of the target vehicle; and / or the situation data includes at least one of the following information: flow rate and vehicle speed of the point within the time period to which the target vehicle passes through the point; or The feature data of the point is obtained in the following manner: converting each data used to generate the point data into a discrete vector; splicing the obtained discrete vectors to obtain spliced ​​data; inputting the spliced ​​data into a pre-trained point prediction model for point prediction, and obtaining the target features for point prediction generated based on the spliced ​​data during the point prediction process of the point prediction model as the feature data, wherein the point prediction model is a prediction model trained by a cyclic prediction method using the description data of the third point in the sample driving path of the vehicle as input data and the fourth point in the sample driving path as supervision information, the third point and the fourth point are partial points in the sample driving path, and the description data of the third point is data used to generate the feature data of the third point; or The device further includes: a first time prediction module, which is used to predict the time when the target vehicle passes through each second point in the restoration path after the restoration path determination module determines the restoration path, based on the point data of the upstream point of the second point in the first point, the point data of the downstream point, and the point data of the second point; or The device also includes: a second time prediction module, which is used to determine that the point data of the reference point is the point data of a preset virtual point if there is no reference point of the second point in the first point, and predict the time when the target vehicle passes the second point based on the point data of the reference point and the point data of the second point, wherein the reference point is: an upstream point or a downstream point.

11. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, for implementing any of the methods described in claims 1-8 when executing a program stored in a memory.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images