A method and system for detecting road opening status

Abstract

An embodiment of the present specification provides a method and system for detecting the opening status of a road, the method comprising: determining a local road network where a candidate closed road is located, the local road network comprising the candidate closed road, upstream roads of the candidate closed road, and downstream roads of the candidate closed road; determining vehicle driving trajectories corresponding to the roads in the local road network; determining associated feature information based on the vehicle driving trajectories, the associated information features being related to the association relationship between the roads in the local road network and the vehicle driving trajectories; and using an opening probability determination model to process the associated feature information to determine whether the candidate closed road is open, the opening probability determination model being a trained machine learning model.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and in particular to a method and system for detecting the road opening status. Background Art

[0002] Route navigation plays an important role in transportation. The navigation system will plan multiple travel routes in the road network for users to choose based on the starting and ending information of the trip. The optimal route with the shortest distance and / or the shortest time is the recommended first choice. However, the roads in the road network are complicated. Whether the road is a closed road or whether the road is open (passable) will affect the route planning of the trip. At the same time, the actual road conditions are often more complicated. Situations such as traffic accidents, road construction, and temporary control will also affect whether the road is passable. Therefore, in the planning of navigation routes, it is particularly important to detect the open status of roads in the road network.

[0003] At present, the use of manual reporting or manual inspection and review to check whether a closed road has been opened requires a lot of manpower and material costs, and is also relatively lacking in timeliness. If the change in the opening status of the road cannot be discovered in time, the navigation route cannot be corrected in time, resulting in users taking detours and wasting time and money.

[0004] Therefore, a method and system for detecting the opening status of roads are provided to automatically detect the opening status of roads in the road network, provide an effective basis for planning the optimal travel route of the navigation system, reduce labor costs and improve the user's travel experience. Summary of the invention

[0005] One of the embodiments of the present specification provides a method for detecting the opening status of a road, including determining a local road network where a candidate closed road is located, the local road network including the candidate closed road, upstream roads of the candidate closed road, and downstream roads of the candidate closed road; determining a vehicle driving trajectory corresponding to the road in the local road network; based on the vehicle driving trajectory, determining associated feature information, the associated information feature is related to the relationship between the road in the local road network and the vehicle driving trajectory; using an opening probability determination model to process the associated feature information to determine whether the candidate closed road is open, the opening probability determination model being a trained machine learning model.

[0006] One of the embodiments of the present specification provides a road opening status detection system, including a road network information determination module, used to determine a local road network where a candidate closed road is located, the local road network including the candidate closed road, the upstream road of the candidate closed road and the downstream road of the candidate closed road; a vehicle trajectory determination module, used to determine the vehicle driving trajectory corresponding to the road in the local road network; an associated feature determination module, used to determine associated feature information based on the vehicle driving trajectory, the associated information feature is related to the association between the road in the local road network and the vehicle driving trajectory; a road status determination module, used to process the associated feature information using an opening probability determination model to determine whether the candidate closed road is open, the opening probability determination model is a trained machine learning model.

[0007] One of the embodiments of the present specification provides a road opening status detection device, including a processor, wherein the processor is used to execute the above-mentioned road opening status detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0009] Figure 1 is a schematic diagram of a scene of a road opening status detection system according to some embodiments of this specification;

[0010] Figure 2 is a schematic diagram of a module of a road opening status detection system according to some embodiments of this specification;

[0011] Figure 3 is an exemplary flow chart of a method for detecting a road opening state according to some embodiments of this specification;

[0012] Figure 4a It is an exemplary process of the method for determining whether a vehicle driving trajectory passes through a candidate closed road according to some embodiments of this specification;

[0013] Figure 4b is an exemplary schematic diagram of trajectory points and local road networks shown in some embodiments of this specification;

[0014] Figure 4c is an exemplary schematic diagram of the association feature information of another trajectory point and a local road network shown in some embodiments of this specification;

[0015] Figure 5a is an exemplary schematic diagram of a process of determining whether a candidate closed road is open according to some embodiments of this specification;

[0016] Figure 5b is an exemplary schematic diagram of multiple reference determination models according to some embodiments of this specification;

[0017] Figure 6 is an exemplary flow chart of determining whether a candidate closed road is open based on multiple first opening probabilities according to some embodiments of this specification;

[0018] Figure 7 It is an exemplary schematic diagram of a method for verifying the result of opening a candidate closed road according to some embodiments of this specification. DETAILED DESCRIPTION

[0019] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0020] It should be understood that the "system", "device", "unit" and / or "module" used herein are a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0021] Unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0022] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0023] Figure 1 It is a scene schematic diagram of a road opening status detection system according to some embodiments of this specification.

[0024] like Figure 1As shown, the application scenario 100 of the road opening status detection system may include a traffic information database 110 , a network 120 , a terminal 130 , a processing device 140 and a storage device 150 .

[0025] The traffic information database 110 refers to a data source for providing traffic information. For example, it may include a third-party traffic-related service platform (such as a server or database of a map operator) and / or a third-party road network database. In some embodiments, the traffic information database 110 may be used to provide a local road network corresponding to a target area (such as an online car-hailing operation area) and its road network-related data (such as road information, connection information between roads, etc.). In some embodiments, the traffic information database 110 may also provide vehicle driving trajectory information. For example, it may provide a vehicle information trajectory of a local road network corresponding to a candidate closed road. In some embodiments, the traffic information database 110 may interact with other components in the application scenario 100 through the network 120. For example, it may send traffic-related information (such as road network data, vehicle driving trajectory, etc.) to the processing device 140 through the network 120, so that the processing 140 performs the detection of the road opening status. For another example, the traffic information database 110 may send traffic-related information to the storage device 150 through the network 120.

[0026] The network 120 may include any suitable network that facilitates the exchange of information and / or data for the application scenario 100. In some embodiments, one or more components of the application scenario 100 (e.g., the traffic information database 110, the terminal 130, the processing device 140, or the storage device 150) may transmit information and / or data to one or more other components of the application scenario 100 via the network 120. For example, the processing device 140 may obtain multiple target trajectory points of a candidate closed road from the traffic information database 110 via the network 120. In some embodiments, the network 120 may be any one or more of a wired network or a wireless network. In some embodiments, the network may be a point-to-point, shared, centralized, or other topological structure or a combination of multiple topological structures.

[0027] The terminal 130 may include a mobile device 130-1, a tablet computer 130-2, a laptop computer 130-3, etc., or any combination thereof. In some embodiments, the terminal 130 may interact with other components in the application scenario 100 through the network 120. In some embodiments, the terminal 130 may receive information and / or instructions input by a user (e.g., a system administrator), and send the received information and / or instructions to the processing device 140 via the network 120. For example, a user may input an operation instruction to the traffic information database 110 (e.g., checking the opening status of each road in the traffic information database 110) through the terminal 130. In some embodiments, the terminal 130 may display the vehicle driving trajectory of the road, as well as the trajectory point distribution information generated by the processing device 140.

[0028] The processing device 140 may process data and / or information obtained from the traffic information database 110, the terminal 130, and the storage device 150. For example, the processing device 140 may obtain relevant information (such as road type, number, whether it is one-way, etc.) of the candidate closed roads in the traffic information database 110, the vehicle driving trajectory (such as its corresponding multiple trajectory points), etc. For another example, the processing device 140 may determine the opening status of the candidate closed roads by processing the associated features of the vehicle driving trajectory and the candidate closed roads.

[0029] In some embodiments, the processing device 140 can be a single server or a server group. The server group can be centralized or distributed. In some embodiments, the processing device 140 can be local or remote. The processing device 140 can be directly connected to the medical scanning device 110, the terminal 130 and the storage device 150 to access the stored or acquired information and / or data. In some embodiments, the processing device 140 can be implemented on a cloud platform. By way of example only, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any combination thereof.

[0030] The storage device 150 may store data and / or instructions. In some embodiments, the storage device 150 may store data obtained from the traffic information database 110, the terminal 130 and / or the processing device 140. For example, the storage device 150 may store vehicle driving trajectory data (such as characteristic information of trajectory points) of each road of the local road network 110, etc. In some embodiments, the storage device 150 may store data and / or instructions used by the processing device 140 to execute the exemplary methods described in this specification. For example, the storage device 150 may store instructions for the processing device 140 to execute the methods shown in each flowchart. In some embodiments, the storage device 150 may include a large-capacity storage device, a removable storage device, a volatile read-write memory, a read-only memory (ROM), etc., or any combination thereof. In some embodiments, the storage device 150 may be implemented on a cloud platform. In some embodiments, the storage device 150 may be part of the processing device 140.

[0031] The above description is for illustrative purposes only, and actual application scenarios may vary.

[0032] It should be noted that the application scenario 100 is provided for illustrative purposes only and is not intended to limit the scope of the present application. For those of ordinary skill in the art, various modifications or variations can be made according to the description of this specification. However, these variations and modifications will not deviate from the scope of the present application.

[0033] Figure 2 It is a module schematic diagram of a road opening status detection system according to some embodiments of this specification.

[0034] like Figure 2 As shown, the road opening status detection system 200 (hereinafter referred to as the detection system) may include a road network information determination module 210, a vehicle trajectory determination module 220, an associated feature determination module 230, and a road status determination module 240. In some embodiments, the road network information determination module 210, the vehicle trajectory determination module 220, the associated feature determination module 230, and the road status determination module 240 may be implemented by the processing device 140.

[0035] The road network information determination module 210 is used to determine the local road network where the candidate closed road is located, and the local road network includes the candidate closed road, the upstream road of the candidate closed road, and the downstream road of the candidate closed road.

[0036] In some embodiments, the road network information determination module 210 may also be used to: determine an initial vehicle driving trajectory corresponding to a candidate closed road; determine an extended vehicle driving trajectory including the initial vehicle driving trajectory; and determine a local road network based on the extended vehicle driving trajectory.

[0037] The vehicle trajectory determination module 220 is used to determine the vehicle driving trajectory corresponding to the road in the local road network.

[0038] The associated feature determination module 230 is used to determine associated feature information based on the vehicle driving trajectory, where the associated information feature is related to the association relationship between the roads in the local road network and the vehicle driving trajectory.

[0039] In some embodiments, the associated feature determination module 230 can also be used to: obtain trajectory point features of each trajectory point among multiple trajectory points on the vehicle's driving trajectory, the trajectory point features at least including location information and vehicle driving information; determine associated feature information based on the trajectory point features of each trajectory point and the road network data of the local road network, the associated feature information including at least one of the following information: the distance between each trajectory point and the road on which it is located, the driving speed of each trajectory point on the road on which it is located, and whether the vehicle's driving trajectory passes through a candidate closed road.

[0040] In some embodiments, the associated feature determination module 230 can also be used to: determine multiple target trajectory points from multiple trajectory points based on the distance between each trajectory point and the road on which it is located, and the distance between each target trajectory point and the road on which it is located is less than a preset distance threshold; determine the trajectory point distribution information of the multiple target trajectory points on each road in the local road network based on the position information of the multiple target trajectory points; and determine whether the vehicle's driving trajectory passes through the candidate closed road based on the trajectory point distribution information.

[0041] In some embodiments, the preset distance threshold is related to the positioning accuracy of the vehicle's driving trajectory.

[0042] The road status determination module 240 is used to process the associated feature information using the opening probability determination model to determine whether the candidate closed road is open. The opening probability determination model is a trained machine learning model.

[0043] In some embodiments, the activation probability determination model is generated by training using a random dropout mechanism.

[0044] In some embodiments, the road state determination module 240 can also be used to: perform multiple neuron deactivation processes on the opening probability determination model based on a preset neuron deactivation probability to obtain multiple reference determination models; use each reference determination model to process associated feature information to determine the first opening probability of the candidate closed road; and determine whether the candidate closed road is open based on multiple first opening probabilities corresponding to multiple reference determination models.

[0045] In some embodiments, the road status determination module 240 can also be used to: determine a first comparison result based on multiple first opening probabilities and opening probability thresholds; determine a second comparison result based on the divergence of multiple first opening probabilities and the divergence thresholds; determine whether the candidate closed road is open based on the first comparison result and the second comparison result.

[0046] In some embodiments, the road status determination module 240 can also be used to: obtain a road image of a candidate closed road; process the road image based on a road opening verification model to determine a second opening probability of the candidate closed road, where the road opening verification is a machine learning model; based on the second opening probability, verify the determination result of whether the candidate closed road is open to obtain a verification result of whether the candidate closed road is open.

[0047] In some embodiments, the road status determination module 240 can also be used to: process road images based on a road opening verification model to determine road abnormality features of candidate closed roads, where the road abnormality features include road construction features and closure notice features; and determine a second opening probability of the candidate closed road based on the road abnormality features.

[0048] It should be noted that the above description of the road opening status detection system 200 and its modules is only for the convenience of description, and this specification cannot be limited to the scope of the embodiments. It can be understood that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine the various modules without deviating from this principle, or to form a subsystem connected with other modules. For example, the road network information determination module 210, the vehicle trajectory determination module 220, the associated feature determination module 230, and the road state determination module 240 can be different modules in the system, or one module can realize the functions of two or more of the above modules. For example, each module can share a storage module, or each module can have its own storage module. Such variations are all within the protection scope of this specification.

[0049] Figure 3 It is an exemplary flow chart of a method for detecting a road opening status according to some embodiments of this specification.

[0050] In some embodiments, process 300 may be performed by a detection system. Figure 3 As shown, process 300 includes the following steps.

[0051] Step 310, determining a local road network where the candidate closed road is located, the local road network including the candidate closed road, an upstream road of the candidate closed road, and a downstream road of the candidate closed road.

[0052] A candidate closed road refers to a suspected closed road that requires road status detection. Among them, the road status of a candidate closed road may include an open state and an unopened state (or a closed state). When a vehicle can pass from one end of a candidate closed road to the other end, it means that its road status is open, otherwise it is closed. A candidate closed road may be one or more roads in a road network (such as an urban road network or a highway network) corresponding to a target area (such as an online car-hailing operation area). It should be understood that in some cases, the road status of a candidate closed road may also include a semi-open state. For example, a lane of a candidate closed road is impassable due to construction, while other lanes are passable. At this time, vehicles can only pass through the candidate closed road at a relatively slow speed.

[0053] In some embodiments, candidate closed roads can be determined based on information reported by users (such as passengers and drivers). For example, a driver may report feedback information that a certain road is impassable to the online car-hailing platform, and the detection system can determine that the road is a candidate closed road based on the feedback information. In some embodiments, candidate closed roads can be determined based on road opening information recorded on a third-party transportation service platform. For example, a closed road recorded on a map service platform can be determined as a candidate closed road.

[0054] In some embodiments, the detection system may determine candidate closed roads based on traffic flow information of multiple roads in the target area. For example, traffic flow information within a preset time period (such as the past week or three days) up to the current time point may be obtained, and when the traffic flow of a certain road is lower than a traffic flow threshold (such as the average traffic flow within a preset time period), the road is determined as a candidate closed road.

[0055] In some embodiments, the detection system may also determine a candidate closed road based on the difference between the actual driving route of the vehicle and the preset driving route (such as the navigation route planned by the system). For example, when the actual driving routes of multiple vehicles deviate from a road in the preset driving route, the road may be determined as a candidate closed road. The actual driving route of the vehicle may be determined based on the actual trip corresponding to the vehicle order (or taxi order).

[0056] The road network refers to the network structure composed of roads in a certain area (such as the target area). In the road network, roads can be represented in the form of straight lines or curves. The local road network refers to the road network related to the candidate closed road in the target area, which includes multiple connecting roads connected to the candidate closed road.

[0057] When a vehicle can enter a candidate closed road via one or more roads, these roads can be referred to as upstream roads of the candidate closed road; when a vehicle can enter one or more roads via a candidate closed road, these roads can be referred to as downstream roads of the candidate closed road. The upstream road, the candidate closed road, and the downstream road can constitute a driving route or a portion thereof.

[0058] It should be noted that the determination of the upstream road and / or the downstream road is related to the preset driving direction (such as one-way driving or two-way driving) of the candidate closed road, which can be obtained based on the local road network data.

[0059] The local road network data may include, but is not limited to, the number of roads in the local road network, the name of each road, the type (highway, urban road, etc.), number or ID (such as Link ID of the road), whether it is one-way, length and width, location, etc. The local road network data also includes the connection information between each road, for example, the connection nodes between the candidate closed road and the upstream road and the downstream road and their node locations (such as longitude and latitude), etc. The local road network data can be determined based on the map data or road network database of a third party.

[0060] In some embodiments, the detection system may obtain an initial vehicle driving trajectory corresponding to the candidate closed road, and determine an extended vehicle driving trajectory including the initial vehicle driving trajectory based on the initial vehicle driving trajectory, and further determine a local road network based on the extended vehicle driving trajectory.

[0061] The initial vehicle driving trajectory corresponding to the candidate closed road may include the vehicle driving trajectory that passes through the candidate closed road within the first preset time period. For example, it may be the vehicle driving trajectory that passes through the candidate closed road in the past 3 days.

[0062] The extended vehicle driving trajectory refers to a vehicle driving trajectory generated by matching the trajectory points of the upstream road and / or the downstream road using the initial vehicle driving trajectory. The operation of matching the trajectory points of the upstream road (upstream trajectory points) and / or the trajectory points of the downstream road (downstream trajectory points) can be referred to as trajectory point matching. In some embodiments, the upstream trajectory points and / or the downstream trajectory points are trajectory points related to the trajectory points in the initial vehicle driving trajectory (such as trajectory points extended from the trajectory points of a vehicle).

[0063] In some embodiments, the detection system may perform trajectory point matching after expanding (e.g., forward and / or backward) a preset time step (e.g., 1 minute) in the time dimension. For example, the detection system may first determine multiple starting trajectory points at the starting endpoint of the candidate closed road, and the time corresponding to the multiple starting trajectory points (e.g., 8:00am), and then expand the preset time step (e.g., 1 minute) forward in the time dimension to obtain a first extended time period (e.g., 7:59 to 8:00), and obtain the upstream trajectory points within the first extended time period to determine the first extended vehicle driving trajectory, and the upstream road corresponding to the first extended vehicle driving trajectory.

[0064] At the same time, multiple terminal trajectory points at the terminal end point of the candidate closed road and the times corresponding to the multiple terminal trajectory points (such as 8:05am) can also be determined, and then the preset time step (such as 1min) is extended backward in the time dimension to obtain a second extended time period (such as 8:05 to 8:06), and the downstream trajectory points within the second extended time period are obtained to determine the second extended vehicle driving trajectory and the downstream road corresponding to the second extended vehicle driving trajectory.

[0065] In some embodiments, the detection system can also extend in the spatial dimension (toward the upstream road direction and / or the downstream road direction) at a preset distance step (such as 3m) to perform trajectory point matching to obtain upstream trajectory points and / or downstream trajectory points in the extended space, thereby obtaining the extended vehicle driving trajectory and its corresponding upstream road and / or downstream road.

[0066] It should be noted that the extended vehicle driving trajectory can reflect the supplement or extension of the initial vehicle driving trajectory on the upstream road and / or downstream road of the candidate closed road. A candidate closed road may have multiple upstream roads and downstream roads. Compared with the traditional local road network determined by map or road network data, some embodiments of this specification determine the candidate closed road and its corresponding upstream road and downstream road by extending the vehicle driving trajectory, so that the determined local road network is more relevant and targeted to the candidate closed road, and the subsequent result of determining whether the candidate closed road is open is more accurate. At the same time, it avoids the erroneous judgment of the road opening state caused by the lack of spatial and temporal correlation (such as continuity) between the trajectory points of the upstream road and / or downstream road and the trajectory points on the candidate closed road. For example, when the connection between the candidate closed road and the downstream road is closed, the trajectory points of the downstream road are not the trajectory points when the vehicle of the candidate closed road leaves the candidate closed road, but the trajectory points left by the vehicle of the downstream road after turning around here. In some embodiments of this specification, by extending the vehicle driving trajectory, the local road network can be determined more targeted.

[0067] Step 320, determining the vehicle driving trajectory corresponding to the road in the local road network.

[0068] The vehicle driving trajectory includes a vehicle driving route passing through at least one road (at least one of a candidate closed road, an upstream road, and a downstream road) in the local road network within a second preset time. The vehicle driving trajectory may include multiple trajectory points, each trajectory point corresponding to a position point of the vehicle in the local road network at a certain moment in history.

[0069] In this specification, as long as at least one track point in the vehicle's driving trajectory is on a certain road, the vehicle's driving trajectory can be considered to pass through the road; when the distance between the position of the track point and the position of the road is less than a preset distance threshold, the track point can be considered to be on the road. The second preset time can be a period of time close to the current time (such as a day or a week), which is the same as or different from the first preset time above.

[0070] Step 330, determining associated feature information based on the vehicle driving trajectory, where the associated information feature is related to the association relationship between the roads in the local road network and the vehicle driving trajectory.

[0071] The associated feature information can reflect the relationship between the vehicle's driving trajectory and each road in the local road network. It can be determined according to actual needs. For example, the associated feature information can include but is not limited to distance relationship (such as deviation distance), relative orientation relationship (such as east, west, longitudinal, etc.), etc.

[0072] In some embodiments, the detection system can obtain trajectory point features of each of the multiple trajectory points on the vehicle's driving trajectory, the trajectory point features including at least location information and vehicle driving information, and determine associated feature information based on the trajectory point features of each trajectory point and the road network data of the local road network, wherein the associated feature information includes at least one of the following information: the distance between each trajectory point and the road on which it is located, the driving speed of each trajectory point on the road on which it is located, and whether the vehicle's driving trajectory passes through the candidate closed road.

[0073] The track point features include the location information of the track point, which may include the longitude and latitude information of the track point, which may be determined based on the vehicle-mounted GPS positioning device. The location information may also include the road to which the track point belongs in the local road network (such as the Link ID of the road), which side of the road it belongs to, and other information.

[0074] The vehicle driving information may include the vehicle's driving speed and driving direction. The vehicle driving information may also include information such as the length of time the vehicle stays at the location corresponding to the track point.

[0075] In some embodiments, the detection system can determine the distance between each trajectory point and the road on which it is located based on the location information of the trajectory point and the location information of the road in the road network data, which can be the vertical distance between the trajectory point and the road. Figure 4b , Figure 4b The arrows in represent roads in the local road network, and the dots represent trajectory points, wherein the vertical distance H between the trajectory point and the road 412 on which it is located is the distance between the trajectory point and the road 412 .

[0076] The driving speed of the track point on the road on which it is located refers to the driving speed of the vehicle when it passes the track point on the road. The detection system can determine the driving speed of the track point on the road on which it is located based on the vehicle driving information of the track point.

[0077] The vehicle driving trajectory intersecting the candidate closed road means that multiple trajectory points of the vehicle driving trajectory corresponding to the same vehicle cover the candidate closed road and its upstream road and downstream road. This means that the vehicle can enter the candidate closed road via the upstream road and drive to the downstream road.

[0078] In some embodiments, the detection system can determine whether the vehicle's driving trajectory passes through a candidate closed road based on the distribution information of multiple trajectory points in the vehicle's driving trajectory in the local road network. Figure 4a and its description.

[0079] Step 340, using the opening probability determination model to process the associated feature information to determine whether the candidate closed road is open, the opening probability determination model is a trained machine learning model.

[0080] When the candidate closed road is in an open state, vehicles can pass through the candidate closed road. Specifically, vehicles can enter the candidate closed road from the upstream road of the candidate closed road, and after passing through the candidate closed road, they can enter its downstream road. When the candidate closed road is in an unopened state (i.e., closed state), the candidate closed road is impassable and / or partially impassable.

[0081] It should be noted that the opening status of the candidate closed roads is time-sensitive. Specifically, its opening status may change with changes in actual conditions (such as road construction, traffic accidents, road damage, etc.). In some embodiments, the detection system can obtain one or more candidate closed roads based on a preset time period (such as every day, every week, etc.) and detect the opening status to achieve real-time or regular monitoring of the road status, thereby providing users with more accurate route navigation.

[0082] In some embodiments, the detection system can process the associated feature information based on a trained opening probability determination model to determine whether the candidate closed road is open. The opening probability determination model refers to a model used to determine whether a candidate closed road is open. In some embodiments, the opening probability determination model is a trained machine learning model. For example, it can be a multi-layer perceptron (Multi-Layer Perception, MLP), a long short-term memory neural network (Long Short-Term Memory, LSTM) or other custom deep learning network models. For more information about the opening probability determination model, see Figure 5a and its description.

[0083] In some embodiments, the detection system may also verify the result of determining whether the candidate closed road is open. For example, the detection system may verify the open status of the candidate closed road based on the road image of the candidate closed road. For more information on verifying the result of determining whether the candidate closed road is open, see Figure 7 and its description.

[0084] In some embodiments of the present specification, by determining the correlation feature information between each road in the local road network and the corresponding vehicle driving trajectory, the distribution of the trajectory points of the vehicle driving trajectory on the candidate closed road and its upstream road and / or downstream road can be analyzed. In addition, by processing the correlation information through the trained opening probability determination model, the distribution law of the trajectory points and its relationship with the road opening status can be automatically learned using an artificial intelligence algorithm, saving a lot of manpower and material costs, and also making the result of determining whether the candidate road is open more accurate.

[0085] Figure 4a This is an exemplary process of a method for determining whether a vehicle driving trajectory passes through a candidate closed road according to some embodiments of this specification.

[0086] In some embodiments, process 400 may be performed by a detection system. Figure 4a As shown, process 400 includes the following steps.

[0087] Step 410, based on the distance between each trajectory point and the road on which it is located, multiple target trajectory points are determined from the multiple trajectory points, and the distance between each target trajectory point and the road on which it is located is less than a preset distance threshold.

[0088] The target trajectory point refers to a trajectory point in the vehicle's driving trajectory that meets a condition. The condition includes that the distance between the trajectory point and the road on which it is located is less than a preset distance threshold. The distance between the trajectory point and the road on which it is located can be determined based on the vertical distance between the trajectory point and the road.

[0089] In some embodiments, the preset distance threshold is related to the positioning accuracy of the vehicle's driving trajectory. The positioning accuracy can be determined based on the positioning accuracy of a positioning device (such as a vehicle-mounted GPS positioning device), which can be used to reflect the degree of deviation between the vehicle's driving trajectory in the local road network and the actual vehicle's driving trajectory.

[0090] The positioning accuracy may include a positioning deviation range. For example, 3m to 5m. In some embodiments, the detection system may determine a preset distance threshold based on the maximum value of the deviation range corresponding to the positioning accuracy (e.g., 5m), and obtain multiple trajectory points within a range where the distance (e.g., vertical distance) between the target vehicle and the road is less than the preset distance threshold (e.g., 5m) according to the vehicle driving trajectory corresponding to the target vehicle, as target trajectory points.

[0091] In some embodiments of the present specification, positioning accuracy is introduced so that the setting of the preset distance threshold is more in line with the actual situation of the vehicle position, thereby making the target trajectory point more accurate, providing a more accurate data basis for the subsequent detection of the opening status of the candidate closed roads.

[0092] Step 420, based on the location information of the plurality of target trajectory points, determining the trajectory point distribution information of the plurality of target trajectory points on each road of the local road network.

[0093] The trajectory point distribution information can be used to characterize the spatial distribution of the target trajectory points on each road in the local road network.

[0094] In some embodiments, the detection system may determine the road (e.g., candidate closed road, upstream road, downstream road) to which each target trajectory point belongs, and count the number of trajectory points at different positions and / or sub-segments in each road as trajectory point distribution information. In some embodiments, the detection system may connect multiple target trajectory points in chronological order to obtain a trajectory line as trajectory point distribution information.

[0095] Step 430 : determining whether the vehicle driving trajectory runs through the candidate closed road based on the trajectory point distribution information.

[0096] In some embodiments, the detection system may determine whether the vehicle driving trajectory runs through the candidate closed road based on the trajectory point distribution information corresponding to multiple sub-road sections in the candidate closed road.

[0097] For example, the detection system may divide the candidate closed road into a plurality of sub-segments according to relevant information of the candidate closed road (such as length, width, curvature, etc.). Exemplarily, the candidate closed road may be divided into k sub-segments according to the direction from the upstream road to the downstream road, and each sub-segment may be numbered in sequence to generate a sub-segment sequence corresponding to the candidate closed road.

[0098] In some embodiments, the detection system can also spatially extend the candidate closed road with a preset extension distance (such as 3m) in the spatial dimension to obtain an upstream sub-segment and a downstream sub-segment, and expand the above sub-segment sequence according to the travel direction of the candidate closed road (such as one-way) to generate a sub-segment sequence corresponding to the candidate closed road. It can be understood that when the target trajectory point corresponding to the upstream sub-segment is continuous with the trajectory point on the sub-segment on the candidate closed road and the target trajectory point corresponding to the downstream sub-segment, the possibility that the candidate closed road is in an open state is greater.

[0099] Furthermore, the detection system can determine the trajectory point quantity sequence corresponding to the sub-segment sequence based on the trajectory point distribution information and the sub-segment sequence. The trajectory point quantity sequence can be used to determine the uniformity of the trajectory points on the candidate closed road, which can reflect the degree of continuity of the vehicle's driving trajectory on the candidate closed road. It can further reflect the probability that the vehicle's driving trajectory runs through the candidate closed road. The higher the uniformity, the greater the probability that the vehicle's driving trajectory runs through the candidate closed road.

[0100] In some embodiments, the track point uniformity can be determined based on the variance value of the number of track points corresponding to the multiple sub-road sections. When the variance value is smaller, it means that the fluctuation (or difference) of the number of track points corresponding to each sub-road section is smaller, and the track point uniformity is greater. When the track point uniformity is less than the uniformity threshold, it is determined that the vehicle driving track runs through the candidate closed road.

[0101] For another example, the detection system can determine the degree of overlap between the trajectory line obtained by connecting the target trajectory points in time sequence and the candidate closed road, the upstream road, and the downstream road. If the trajectory line can completely cover the upstream road (or a part thereof), the candidate closed road, and the downstream road (or a part thereof), it means that the vehicle's driving trajectory passes through the candidate closed road.

[0102] Figure 4b It is an exemplary schematic diagram of trajectory points and local road networks shown in some embodiments of this specification.

[0103] like Figure 4b As shown, the local road network 410 includes an upstream road 411, a candidate closed road 412, a downstream road 413, and a downstream road 414, and multiple trajectory points are shown in circles in the figure. The detection system can determine the longitudinal vertical distance (as marked H in the figure) between each trajectory point and each road based on the location information of each trajectory point (such as trajectory point 415) and the road information of each road as the corresponding target distance. When the target distance is less than the preset distance threshold, the trajectory point can be set as the target trajectory point. Based on the distance between each trajectory point and the road on which it is located, the target trajectory point (shown in a white hollow circle in the figure) and the non-target trajectory point (shown in a gray circle in the figure) can be determined.

[0104] Since multiple target trajectory points are evenly distributed along the upstream road 411 , the candidate closed road 412 , and the downstream road 413 , it means that the vehicle driving trajectory covers the upstream road 411 , the candidate closed road 412 , and the downstream road 413 , further indicating that the vehicle driving trajectory runs through the candidate closed road 412 .

[0105] Figure 4c is an exemplary schematic diagram of another trajectory point and associated characteristic information of a local road network shown in some embodiments of this specification. Figure 4b The difference is, Figure 4c The target trajectory points in the figure (shown as white hollow circles) are distributed on the upstream road 421 and a part of the candidate closed road 422, but are not distributed on the other part of the candidate closed road 422 and the downstream road 423, indicating that the vehicle driving trajectory does not pass through the candidate closed road 422.

[0106] In some embodiments of the present specification, by dividing the candidate closed road into multiple sub-segments, the distribution of target trajectory points of the candidate closed road can be analyzed more finely, so that the determination result of whether the vehicle driving trajectory passes through the candidate closed road is more accurate.

[0107] It should be understood that Figure 4a The method in is only one embodiment of determining whether the vehicle driving trajectory runs through the candidate closed road, and there may be many variations. For example, it may be determined whether the vehicle driving trajectory runs through the candidate closed road based on all trajectory points instead of the target trajectory point.

[0108] Figure 5a The figure is an exemplary schematic diagram of a process of determining whether a candidate closed road is open according to some embodiments of the present specification.

[0109] In some embodiments, the detection system may process the associated feature information using the opening probability determination model to determine whether the candidate closed road is open. For example, the input of the opening probability determination model includes the associated feature information, and the output of the opening probability model includes the probability that the candidate closed road has been opened, or the judgment result of whether it is opened.

[0110] In some embodiments, the detection system can perform multiple neuron deactivation processes on the opening probability determination model based on a preset neuron deactivation probability to obtain multiple reference determination models, and use each reference determination model to process associated feature information to determine the first opening probability of the candidate closed road, and determine whether the candidate closed road is open based on multiple first opening probabilities corresponding to multiple reference determination models.

[0111] The reference determination model can be generated by evolving the activation probability determination model. Figure 5aAs shown, after the activation probability determination model 510 undergoes multiple (eg, n) neuron inactivation processes, reference determination models 520-1, reference determination models 520-2, ..., and reference determination models 520-n are obtained respectively.

[0112] Neuron inactivation processing can make some neurons in the activation probability determination model 510 in an inactivated state (failed or unactivated state). It can be achieved in many ways, for example, by randomly deleting or shutting down some neurons.

[0113] In some embodiments, the detection system may perform neuron inactivation processing on the activation probability determination model 510 based on the random inactivation matrix. The random inactivation matrix may be a matrix of multiple rows and columns composed of multiple elements, each of which corresponds to a neuron in multiple hidden layers (intermediate layers) of the activation probability determination model 510, and the value of each element may be 0 or 1. The value of each element in the initial random inactivation matrix may be 1, and the detection system may set the value of some elements from 1 to 0 according to the preset inactivation probability before the neuron inactivation processing, so as to obtain the random inactivation matrix corresponding to the current neuron inactivation processing.

[0114] The preset inactivation probability refers to the probability that each neuron is set to an inactivated state. It can be a value in the interval (0, 1) determined according to actual needs, such as 0.5. The larger the value, the higher the probability that each neuron becomes an inactivated neuron, or the lower the probability that it remains an activated neuron.

[0115] In some embodiments, the preset deactivation probability is related to the preset number of reference determination models. When the preset number is larger, the preset deactivation probability can be set smaller. In some embodiments, the preset deactivation probability can be set to a value within the deactivation probability threshold range. Exemplarily, the deactivation probability threshold range can be an interval of [0.2, 0.5]. The detection system can adjust the preset deactivation probability according to the value of the preset number. The smaller the value of the preset number, the more it can be adjusted towards the maximum value of the deactivation probability threshold range, otherwise, it can be adjusted towards the minimum value.

[0116] In some embodiments of the present specification, the preset inactivation probability is adaptively adjusted according to the preset number of reference determination models, so that multiple reference determination models have certain differences. In addition, the setting of the inactivation probability threshold range can avoid the inability to accurately obtain the first activation probability due to too many neurons in the inactivated state in each reference determination model, resulting in a serious decline in the model's expression ability; at the same time, it also avoids the situation where there are too few neurons in the inactivated state, resulting in small differences between different reference determination models and affecting the diversity of the first activation probability prediction.

[0117] In some embodiments, after determining the random inactivation matrix corresponding to the current neuron inactivation process, the detection system can multiply the value of each element in the matrix by the current weight of the corresponding neuron. When the value of the element is 0, the weight of the corresponding neuron is set to 0, and the neuron becomes an inactivated neuron; when the value of an element is 1, the weight of the neuron corresponding to the element remains unchanged, and the neuron remains an activated neuron.

[0118] For multiple neurons in one or more intermediate layers (hidden layers) in the activation probability determination model, the detection system can set the weights of some neurons to 0 through a random deactivation matrix in one neuron deactivation process to obtain the corresponding reference determination model after the neuron deactivation process. After n neuron deactivation processes, n reference determination models are obtained.

[0119] In some embodiments of the present specification, a random deactivation matrix is ​​constructed based on a preset deactivation probability, and a neuron deactivation process is performed on an activation probability determination model based on the random deactivation matrix, so that when using the random deactivation mechanism, requirements such as the number of reference determination models can be taken into account to obtain a reference determination model that better meets the requirements.

[0120] Combination Figure 5b , Figure 5b is an exemplary schematic diagram of multiple reference determination models according to some embodiments of the present specification.

[0121] like Figure 5b As shown, the turn-on probability determination model 510 includes a plurality of activated neurons (shown as hollow circles) in an activated state.

[0122] After the first neuron inactivation process is performed on the activation probability determination model 510, two of the multiple activated neurons are set to inactivated neurons in an inactivated state (shown by gray solid circles), and a reference determination model 520-1 is obtained. In the subsequent reference determination model processing of the associated feature information, the weight of the inactivated neuron is 0 (or the output result is 0), that is, the inactivated neuron has no effect on the calculation result in the forward propagation calculation process.

[0123] Similarly, after the second neuron inactivation process is performed on the activation probability determination model 510, three of the multiple activated neurons are set to be inactivated neurons in an inactivated state, and a reference determination model 520-2 is obtained. Based on this principle, after the nth neuron inactivation process is performed, a reference determination model 520-n (not shown in the figure) is obtained.

[0124] It can be understood that the reference determination model 520 - 1 and the reference determination model 520 - 2 have different activated neurons and different numbers of activated neurons, so the output first activation probabilities may also be different.

[0125] In some embodiments, the reference determination model can be used to process the associated feature information to determine a first open probability of the candidate closed road, wherein a larger value of the first open probability indicates a greater probability that the candidate closed road predicted by the reference determination model is open.

[0126] like Figure 5a As shown, the inputs of reference determination model 520-1, reference determination model 520-2, ... reference determination model 520-n are all associated feature information 521, and the outputs are first activation probability 522-1, first activation probability 522-2, ... first activation probability 522-n.

[0127] It can be understood that since the neurons in the reference determination model 520-1, the reference determination model 520-2, ... the reference determination model 520-n are in an activated state and their numbers are different, the first activation probability 522-1, the first activation probability 522-2, ... the first activation probability 522-n may be different.

[0128] In some embodiments, the detection system may determine the statistical values ​​of multiple first opening probabilities to determine the road state of the candidate closed road, that is, the determination result 531 of whether it is open. The statistical value includes but is not limited to the average value, maximum value and / or minimum value of the multiple first opening probabilities; the determination result 531 of whether it is open can be represented by a binary classification value, such as 0 indicating that the road state is a closed state, and 1 indicating that the road state is an open state. For example only, when the statistical value (such as the average value) is greater than a preset opening probability threshold (such as 0.85), the road state is determined to be an open state.

[0129] In some embodiments, the detection system can also determine whether the candidate closed road is open based on the probability distribution corresponding to multiple first opening probabilities. The probability distribution can reflect the quantitative distribution of the first opening probabilities within multiple probability intervals, wherein the numerical range of the opening probability can be divided into m equal parts according to the preset number of intervals m (such as 5), and m (such as m1, m2, ..., m5) probability intervals are obtained. The detection system can determine the number of first opening probabilities falling within different probability intervals, and use the probability interval with the largest number as the target probability interval, and calculate the statistical value (such as the average value) of the first opening probability within the target probability interval. When the statistical value (such as the average value) is greater than the preset opening probability threshold, the road state is determined to be an open state, otherwise it is a closed state.

[0130] In some embodiments of the present specification, probability distributions corresponding to a plurality of first opening probabilities are considered, and a minority obeys majority mechanism is adopted to make the result of determining whether a candidate closed road is open more accurate.

[0131] In some embodiments, the detection system may also determine whether the candidate closed road is open based on a plurality of first opening probabilities and their divergences. Figure 6 and its description.

[0132] In some embodiments, the activation probability determination model may be generated through training based on a random deactivation mechanism. The activation probability determination model may be generated through training by a detection system or other systems (such as a supplier system of the activation probability determination model).

[0133] The training samples of the opening probability determination model include sample-related feature information, which can take a road with a known road state (such as closed state or open state) as a sample candidate closed road, and determine the local road network corresponding to the sample candidate closed road as the sample local road network based on the road network data. The detection system can further determine the sample vehicle driving trajectory based on the vehicle driving trajectory corresponding to each road in the sample local road network, thereby determining the sample-related feature information based on the sample vehicle driving trajectory. The training label of the opening probability determination model can be determined based on the actual road state corresponding to the sample candidate closed road. If the actual road state is a closed state, the value of its label is 0, otherwise it is 1.

[0134] The initial activation probability determination model can be iterated multiple times using training samples and training labels until the training termination condition is met, thereby obtaining the activation probability determination model. In each round of iteration, the detection system can use a random deactivation mechanism to randomly deactivate some neurons in the initial activation probability determination model in this round of iteration with a preset deactivation probability. The detection system can input the sample-associated feature information into the initial activation probability determination model after the random deactivation process of this round of iteration, and determine the loss function based on the sample first activation probability output by the initial activation probability determination model and the training label, and update the parameters of the initial activation probability determination model based on the value of the loss function until the preset training condition is met. Among them, the preset training condition can be that the value of the loss function is less than a threshold, converges, or the number of iterations reaches a threshold.

[0135] In some embodiments of the present specification, a model is determined through multiple references to determine the first opening probability of a candidate closed road, so that the result of whether the candidate closed road is opened has higher fault tolerance and accuracy, providing a more accurate data basis for the planning of subsequent navigation routes.

[0136] Figure 6 The present invention is an exemplary flowchart of determining whether a candidate closed road is open based on a plurality of first opening probabilities according to some embodiments of the present specification.

[0137] In some embodiments, process 600 may be performed by a detection system. Figure 6 As shown, process 600 includes the following steps.

[0138] Step 610: Determine a first comparison result based on a plurality of first activation probabilities and activation probability thresholds.

[0139] The multiple first activation probabilities can be determined based on multiple reference determination models. For relevant contents of the reference determination model, see Figure 5a and Figure 5b and its description.

[0140] The first comparison result may characterize the probability that the candidate closed road is in an open state. In some embodiments, the first comparison result may be a comparison result between a plurality of statistical values ​​(such as mean, maximum value, minimum value, etc.) of first open probabilities and an open probability threshold. For example, the first comparison result may be the difference between the statistical value of the first open probability and the open probability threshold. For another example, the first comparison result may be a judgment result of whether the statistical value of the first open probability is greater than the open probability threshold. When the statistical value is greater than the open probability threshold, it indicates that the candidate closed road has a higher probability of being in an open state; and the larger the statistical value, the greater the probability that the candidate closed road is in an open state.

[0141] Step 620: Determine a second comparison result based on the divergence of the plurality of first activation probabilities and the divergence threshold.

[0142] The divergence degree can be used to reflect the degree of difference between the multiple first activation probabilities. When the difference between any two first activation probabilities among the multiple first activation probabilities is small, it means that the divergence degree is smaller.

[0143] The divergence may be a value in various forms, for example, it may be in the form of a difference between a maximum value and a minimum value among a plurality of first opening probabilities, or in the form of a difference between a maximum value / minimum value and an average value of a plurality of first opening probabilities (or its absolute value). In some embodiments, the detection system may determine the divergence based on a variance value of a plurality of first opening probabilities. The variance value may reflect the fluctuation of a plurality of first opening probabilities, and a larger variance value indicates a larger divergence, indicating a greater uncertainty in the final state of the candidate closed road.

[0144] The second comparison result can reflect the credibility of the output results of multiple reference determination models (i.e., multiple first activation probabilities). The second comparison result can be a comparison result between the divergence and the divergence threshold. For example, the second comparison result can be the difference between the divergence and the divergence threshold. For another example, the second comparison result can be a judgment result of whether the divergence is less than the divergence threshold. When the divergence is less than the divergence threshold, it means that the multiple first activation probabilities are relatively consistent, and the credibility of the output result is higher; and the smaller the divergence, the greater the credibility of the output result.

[0145] Step 630: Determine whether the candidate closed road is open based on the first comparison result and the second comparison result.

[0146] In some embodiments, when the first comparison result shows that the statistical value of the first opening probability is greater than the opening probability threshold, and the second comparison result shows that the divergence is less than the divergence threshold, the detection system may determine that the detection system determines that the candidate closed road is in an open state. When the first comparison result shows that the statistical value of the first opening probability is less than the opening probability threshold, and the second comparison result shows that the divergence is less than the divergence threshold, the detection system determines that the candidate closed road is in a closed state.

[0147] In some embodiments, when the second comparison result shows that the divergence is greater than the divergence threshold, it may be difficult to determine whether the candidate closed road is open based on the first opening probability, and the detection system may perform Figure 7 In the method, whether the candidate closed road is open is determined based on the road image. Alternatively, when the divergence is greater than the divergence threshold, the detection system can preliminarily determine the open state based on multiple first open probabilities, and then use Figure 7 The method in is based on road images to verify the preliminary determination results.

[0148] In some embodiments of the present specification, a comprehensive analysis and evaluation of the road status of a candidate closed road (such as whether the corresponding opening status is opened) is performed through multiple first opening probabilities and the divergence of the multiple first opening probabilities, so that the result of whether the candidate closed road is opened can be made more accurate.

[0149] Figure 7 It is an exemplary schematic diagram of a method for verifying the result of opening a candidate closed road according to some embodiments of this specification.

[0150] In some embodiments, when the candidate closed road is determined to be open in step 340, the Figure 7 The method in the above example is used to verify the judgment result to prevent the judgment result from being wrong and causing a bad experience for the driver and passengers. In some embodiments, when the candidate closed road is determined to be closed in step 340, the following method can also be used: Figure 7 The method in is used to perform secondary verification.

[0151] In some embodiments, the detection system can obtain a road image of a candidate closed road; process the road image based on a road opening verification model to determine a second opening probability of the candidate closed road, where the road opening verification is a machine learning model; and based on the second opening probability, verify the determination result of whether the candidate closed road is open to obtain a verification result of whether the candidate closed road is open.

[0152] The road image may be used to reflect the actual conditions of the candidate closed road, which may include but is not limited to environmental information of the candidate closed road (such as obstacles such as gravel), traffic information (such as traffic accidents), road information (such as road signs), etc.

[0153] The road image can be taken by the vehicle-mounted equipment of the vehicle passing through the candidate closed road or the user's mobile terminal. For example, the driver may upload the road photo of the candidate closed road taken by the vehicle's driving recorder or mobile phone to the traffic platform. The road image can also be taken by the monitoring equipment installed on the candidate closed road.

[0154] In some embodiments, the detection system may divide the candidate closed road into multiple sub-segments; determine the target sub-segment based on the trajectory point distribution information of the candidate closed road; and obtain the road image of the target sub-segment. Figure 4a and its description.

[0155] The target sub-segment may refer to a suspicious segment that makes vehicles in the candidate closed road impassable. For example, if a sub-segment has serious road damage or is under construction (such as road repair or subway construction), making the sub-segment impassable, then the sub-segment may be the target sub-segment.

[0156] In some embodiments, the detection system may determine the target sub-segment based on the change information of the number or density of target trajectory points corresponding to multiple sub-segments of the candidate closed road. For example, the number of target trajectory points of multiple consecutive sub-segments (sub-segment sequences) may be counted in sequence from the upstream road to the downstream road. When the number of target trajectory points corresponding to a certain sub-segment and its subsequent sub-segments is less than the average number of target trajectory points of the candidate closed road, the detection system may determine that the sub-segment is the target sub-segment, and determine the corresponding road image based on the position information of the target sub-segment.

[0157] In some embodiments of the present specification, by determining a target sub-road section and acquiring its corresponding road image, the acquired road image is made more targeted and the consumption of necessary computing resources is reduced.

[0158] In some embodiments, the detection system may process the road image based on the road opening verification model to determine a second opening probability of the candidate closed road.

[0159] The road opening verification model can be used to process the road image, and its output can be used to determine the opening probability of the candidate closed road. In some embodiments, the road opening verification model is a trained machine learning model. For example, it can be a convolutional neural network (CNN) model, a YOLO model, or other customized deep learning network models or a combination thereof.

[0160] In some embodiments, Figure 7 As shown, the input of the road opening verification model 710 may include a road image 711, and the output of the road opening verification model 710 may include a second opening probability 731 corresponding to the candidate closed road. The second opening probability may be used to reflect the possibility that the road state corresponding to the candidate closed road is an open state. In some embodiments, the second opening probability may be used to verify the determination result of whether the candidate closed road is open to further determine whether the candidate closed road is open. At this time, the training sample of the road opening verification model 710 may include a sample road image of a sample candidate closed road, and the training label may include the opening state of the sample candidate closed road.

[0161] In some embodiments, the detection system may process the road image based on the road opening verification model to determine the road abnormality features of the candidate closed road; and determine the second opening probability of the candidate closed road based on the road abnormality features.

[0162] The road abnormality feature may be related to various abnormal factors that affect the passage of vehicles on the candidate closed road, which may reflect the reason for the road closure.

[0163] like Figure 7 As shown, the input of the road opening verification model 710 may include a road image 711 , and the road opening verification model 710 may extract features from the road image 711 to obtain road abnormality features 721 .

[0164] The road abnormality feature 721 may include multiple abnormal road features, which may include but are not limited to road construction features and closed notice features. The road construction feature is related to abnormal factors that are likely to occur during road construction (such as cones, forklifts, road maintenance workers, etc.). The closed notice feature is related to abnormal factors that are likely to occur when the road is prohibited from passing (such as a no-passing notice).

[0165] In some embodiments, the road opening verification model 710 may identify abnormal factors in the road image 711 based on the road abnormality features 721. For example, the road opening verification model 710 may output the probability of various abnormal factors (no-entry signs, cones, etc.) being included in the road image 711, or output information such as the category and location of the abnormal factors included in the road image 711. At this time, the training samples of the road opening verification model 710 include sample road images (such as road construction images) of sample candidate closed roads, and the training labels include abnormal factors marked in the sample road images.

[0166] The detection system may determine the second opening probability 731 based on the output of the road opening verification model 710. For example, the higher the probability that the road image 710 contains abnormal factors, the lower the second opening probability 731.

[0167] In some embodiments, Figure 7 As shown, the detection system can determine the verification result 741 according to the second opening probability 731 and the opening determination result 732. Exemplarily, when the opening determination result 732 indicates an opening state, if the second opening probability 731 is greater than the preset opening probability, the verification result 741 is opening, otherwise it is closed.

[0168] It should be noted that the determination result of whether the candidate closed road is open may also be verified through various methods such as manual review.

[0169] In some embodiments of the present specification, the results of determining whether a candidate road is open are verified by a road opening verification model, thereby reducing errors in automated opening status detection. At the same time, road opening status detection is performed based on multimodal data (vehicle trajectory data and road images), further making the results of determining the opening status of candidate closed roads more accurate.

[0170] It should be noted that the above description of the relevant process is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0171] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0172] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0173] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0174] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0175] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0176] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0177] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A method for detecting a road opening status, characterized in that: include: Determine a local road network where the candidate closed road is located, the local road network including the candidate closed road, an upstream road of the candidate closed road, and a downstream road of the candidate closed road; Determining vehicle driving trajectories corresponding to roads in the local road network; Based on the vehicle driving trajectory, determining associated feature information, wherein the associated feature information is related to an associated relationship between a road in the local road network and the vehicle driving trajectory; Processing the associated feature information using an opening probability determination model to determine whether the candidate closed road is open, wherein the opening probability determination model is a trained machine learning model; The using the opening probability determination model to process the associated feature information to determine whether the candidate closed road is open includes: Based on a preset neuron inactivation probability, performing multiple neuron inactivation processes on the activation probability determination model to obtain multiple reference determination models; Processing the associated feature information using each of the reference determination models to determine a first opening probability of the candidate closed road; and Determining whether the candidate closed road is open based on the multiple first opening probabilities corresponding to the multiple reference determination models, wherein determining whether the candidate closed road is open based on the multiple first opening probabilities corresponding to the multiple reference determination models comprises: Determining a first comparison result based on the multiple first activation probabilities and the activation probability thresholds; Determining a second comparison result based on the divergence degrees and divergence thresholds of the plurality of first activation probabilities; Based on the first comparison result and the second comparison result, it is determined whether the candidate closed road is open.

2. The method according to claim 1, characterized in that The determining of the associated feature information based on the vehicle driving trajectory includes: Acquire a trajectory point feature of each of the plurality of trajectory points on the vehicle driving trajectory, wherein the trajectory point feature includes at least position information and vehicle driving information; Based on the trajectory point features of each of the trajectory points and the road network data of the local road network, the associated feature information is determined, and the associated feature information includes at least one of the following information: the distance between each of the trajectory points and the road on which it is located, the driving speed of each of the trajectory points on the road on which it is located, and whether the vehicle driving trajectory runs through the candidate closed road.

3. The method according to claim 2, characterized in that Whether the vehicle driving trajectory passes through the candidate closed road is determined based on the following steps: Based on the distance between each of the trajectory points and the road on which it is located, determining a plurality of target trajectory points from the plurality of trajectory points, wherein the distance between each of the target trajectory points and the road on which it is located is less than a preset distance threshold; Based on the location information of the multiple target trajectory points, determining the trajectory point distribution information of the multiple target trajectory points on each road of the local road network; as well as Based on the trajectory point distribution information, it is determined whether the vehicle driving trajectory runs through the candidate closed road.

4. The method according to claim 1, characterized in that: The activation probability determination model is generated by training using a random deactivation mechanism.

5. The method according to claim 1, characterized in that The method further comprises: Acquire a road image of the candidate closed road; Processing the road image based on a road opening verification model to determine a second opening probability of the candidate closed road, wherein the road opening verification is a machine learning model; Based on the second opening probability, the determination result of whether the candidate closed road is open is verified to obtain a verification result of whether the candidate closed road is open.

6. The method according to claim 5, characterized in that Processing the road image based on the road opening verification model to determine the second opening probability of the candidate closed road includes: Processing the road image based on the road opening verification model to determine abnormal road features of the candidate closed road, wherein the abnormal road features include road construction features and closure notice features; Based on the abnormal road feature, a second opening probability of the candidate closed road is determined.

7. A road opening status detection system, characterized in that: include: A road network information determination module, used to determine a local road network where the candidate closed road is located, wherein the local road network includes the candidate closed road, an upstream road of the candidate closed road, and a downstream road of the candidate closed road; A vehicle trajectory determination module, used to determine a vehicle driving trajectory corresponding to a road in the local road network; An associated feature determination module, configured to determine associated feature information based on the vehicle driving trajectory, wherein the associated feature information is related to an associated relationship between a road in the local road network and the vehicle driving trajectory; A road status determination module, used to process the associated feature information using an opening probability determination model to determine whether the candidate closed road is open, wherein the opening probability determination model is a trained machine learning model; The road state determination module is further used to perform multiple neuron deactivation processes on the opening probability determination model based on a preset neuron deactivation probability to obtain multiple reference determination models; and use each of the reference determination models to process the associated feature information to determine a first opening probability of the candidate closed road; and determining whether the candidate closed road is open based on a plurality of first opening probabilities corresponding to the plurality of reference determination models; The road state determination module is further used to determine a first comparison result based on the multiple first opening probabilities and the opening probability threshold; Determining a second comparison result based on the divergence degrees and divergence thresholds of the plurality of first activation probabilities; Based on the first comparison result and the second comparison result, it is determined whether the candidate closed road is open.

8. A road opening status detection device, characterized in that: The method comprises a processor, wherein the processor is used to execute the method for detecting the road opening status according to any one of claims 1 to 6.

Images