Road network data updating method and device

Abstract

The invention provides a road network data updating method and device, and relates to the technical field of big data, in particular to the technical field of map navigation. A specific embodiment of the method comprises the steps of generating a track peak graph based on crowd source track data; generating a road center line and a road boundary line based on the track peak map; and updating the road network data based on the road center line and the road boundary line.

Description

Technical Field

[0001] The present disclosure relates to the field of big data technology, and in particular to the field of map navigation technology. Background Art

[0002] Urban digital road network data is a true reflection of the real world road network and is the basis for electronic navigation systems to provide convenient travel services. Accurate and real-time road network data is crucial for various road network-based information services.

[0003] With the rapid development of cities, new construction, reconstruction, temporary closure and other road changes are constantly occurring. Failure to update road-level (SD) road network data in a timely manner can easily lead to users violating traffic regulations and even cause serious traffic accidents. Lane-level (LD) road network data has an important impact on smart transportation, autonomous driving and other fields. Rapid and real-time construction of road network data and timely updating of road network data by sensing changes in the real world road network have become key and difficult issues that must be overcome in the field of map making.

[0004] At present, there are two main methods for discovering and updating road network data:

[0005] First, data is collected using proprietary equipment: a collection vehicle equipped with special collection equipment collects road network information on-site to obtain accurate road network data.

[0006] Second, remote sensing imaging technology: As an important basic geographic information data, remote sensing image data is characterized by massive, multi-source, and high resolution. It is a reliable data support for high-precision road extraction. Roads are mainly extracted through the texture, color, and spectrum characteristics of roads in image data. Summary of the invention

[0007] The embodiments of the present disclosure provide a road network data updating method, apparatus, device, storage medium and program product.

[0008] In a first aspect, an embodiment of the present disclosure proposes a road network data updating method, comprising: generating a trajectory peak graph based on crowd-source trajectory data; generating a road centerline and a road boundary line based on the trajectory peak graph; and updating the road network data based on the road centerline and the road boundary line.

[0009] In the second aspect, an embodiment of the present disclosure proposes a road network data updating device, comprising: a first generation module, configured to generate a trajectory peak graph based on crowd-source trajectory data; a second generation module, configured to generate a road center line and a road boundary line based on the trajectory peak graph; an update module, configured to update the road network data based on the road center line and the road boundary line.

[0010] In a third aspect, an embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in the first aspect.

[0011] In a fourth aspect, an embodiment of the present disclosure proposes a non-transitory computer-readable storage medium storing computer instructions, where the computer instructions are used to cause a computer to execute the method described in the first aspect.

[0012] In a fifth aspect, an embodiment of the present disclosure proposes a computer program product, including a computer program, which implements the method described in the first aspect when executed by a processor.

[0013] The key or important features of the embodiments of the present disclosure are not intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Other features, purposes and advantages of the present disclosure will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure. Among them:

[0015] Figure 1 is a flow chart of an embodiment of a road network data updating method according to the present disclosure;

[0016] Figure 2 It is a schematic diagram of the structure of the UNet network;

[0017] Figure 3 It is a schematic diagram of the structure of CBAM;

[0018] Figure 4 is a flow chart of another embodiment of the road network data updating method according to the present disclosure;

[0019] Figure 5 is a flow chart of an embodiment of a lane-level road network data updating method according to the present disclosure;

[0020] Figure 6 is a structural schematic diagram of an embodiment of a road network data updating device according to the present disclosure;

[0021] Figure 7 It is a block diagram of an electronic device used to implement the road network data updating method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0022] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0023] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0024] Figure 1 A process 100 of an embodiment of a road network data updating method according to the present disclosure is shown. The road network data updating method comprises the following steps:

[0025] Step 101, generating a trajectory peak graph based on crowd-source trajectory data.

[0026] In this embodiment, the execution body of the road network data updating method can generate a trajectory peak graph based on multi-source trajectory data.

[0027] The execution subject of the road network data updating method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, used to provide distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0028] With the development of satellite positioning technology and the popularization of location-aware devices, users generate a large amount of trajectory data with rich dynamic information when using devices for navigation. Crowd-source trajectory data can be collected from navigation users' devices. Due to the large number of devices, low collection cost, fast dynamic update, strong generalization, and high road network coverage, crowd-source trajectory data can contain rich road spatiotemporal information, record vehicle location, movement direction, and status, and can perceive the geometry, topology, and traffic status of the road in real time. Therefore, crowd-source trajectory data can serve as an important data source for rapid generation and perception update of road networks. Crowd-source trajectory data can include road-level trajectory data and lane-level trajectory data.

[0029] Based on the large-scale high-quality multi-source trajectory data accumulated over multiple days, a trajectory peak map can be automatically generated. In some embodiments, based on the spatial distribution information of the multi-source trajectory data, Gaussian blur can be used to highlight the overall distribution of the trajectory, and the trajectory can be projected to generate a trajectory peak map. Among them, the trajectory peak map can distinguish the uplink and downlink trajectories according to the trajectory direction, which helps to improve the road boundary judgment.

[0030] Step 102: Generate a road centerline and a road boundary line based on the trajectory peak graph.

[0031] In this embodiment, the execution entity may generate a road centerline and a road boundary line based on the trajectory peak graph.

[0032] Based on the trajectory peak graph, a road centerline can be automatically generated, wherein the road centerline generation may refer to fitting a road centerline representing the current road or lane from a trajectory cluster having the same turning rule.

[0033] In some embodiments, based on the skeleton extraction model, the road centerline can be extracted from the trajectory peak map. The skeleton extraction model can be an improved UNet network. Figure 2 As shown, the UNet network has a symmetrical U-shaped encoding and decoding structure, which may include an encoder and a decoder. The trajectory peak map is input into the encoder to obtain a feature map of the trajectory peak map. Then, the feature map is input into the decoder to obtain the center line of the road. The encoder may correspond to the downsampling process of the trajectory peak map, and the decoder may correspond to the upsampling process of the feature map. The encoder may obtain the feature map of the trajectory peak map through a CNN (Convolutional Neural Networks) unit. CNN units may include but are not limited to: convolutional layers, BN (Batch Normalization) layers, and ReLU (Rectified Linear Unit) layers, etc. The encoder may use a maximum pooling operation to reduce the feature map, thereby extracting higher-level features to obtain scene information. The decoder may use an upsampling operation to restore the original trajectory peak map size. Upsampling may combine the information of each downsampled layer and the input information of the upsampled layer to restore the detail information. As shown Figure 3 As shown in the figure, the decoder can introduce CBAM (Convolutional Block Attention Module, feedforward convolutional neural network attention module) to infer the attention of the feature map along the channel dimension and the spatial dimension in turn to obtain the attention map. Then, the attention map is multiplied by the input feature map to perform adaptive feature refinement to obtain the road centerline.

[0034] In some embodiments, based on the image morphology thinning algorithm, the road center line can be extracted from the trajectory peak image. Among them, the image morphology thinning algorithm can include but is not limited to at least one of the following: based on the Hessian matrix line center extraction (Steger) algorithm, grayscale centroid method, Zhang-Suen thinning algorithm, etc. Based on the Hessian matrix line center extraction algorithm, the trajectory imaging image can be regarded as a three-dimensional surface (peak), and the center line (ridge) is determined according to the local curvature to calculate the road center line. Based on the Hessian matrix line center extraction algorithm, a large number of Gaussian convolution operations are performed, and the calculation accuracy is high. The grayscale centroid method has a simple algorithm and high calculation efficiency. The Zhang-Suen thinning algorithm can continuously perform corrosion operations on the binary image pixel array, remove the contour boundary points, and obtain a skeleton connection line with a single pixel width. The thinned graphic skeleton retains most of the features of the original graphic. At least one of the image morphology thinning algorithms can extract at least one road center line. Weighted fusion of at least one road center line can generate the final road center line. Multi-method weighted fusion is used to improve the completeness and accuracy of road centerline extraction.

[0035] In some embodiments, a road centerline can be extracted from a trajectory peak image based on a skeleton extraction model and an image morphology refinement algorithm. For example, a first road centerline is extracted from a trajectory peak image based on a skeleton extraction model. At least one second road centerline is extracted based on at least one item included in the image morphology refinement algorithm. The first road centerline and at least one second road centerline are weightedly fused to generate a road centerline. Based on the skeleton extraction model, a weighted fusion is performed with a traditional image morphology refinement algorithm to generate a road centerline, thereby improving the completeness and accuracy of the road centerline extraction.

[0036] Based on the road centerline, the road boundary line can be automatically generated by a trajectory point density algorithm. In some embodiments, based on the road centerline, the trajectory peak graph is cut across the road surface according to a first preset step length, and the road boundary line is extracted according to a trajectory density layering algorithm.

[0037] There are usually bumps and burrs on the road center line and the road boundary line. In some embodiments, the bumps and burrs on the road center line and the road boundary line are smoothed. For example, the first bumps and burrs at the initial positions of the road center line and the road boundary line are smoothed using the mean method. Among them, the first bumps and burrs can be large bumps and burrs. For another example, the second bumps and burrs in the middle section of the road center line and the road boundary line are smoothed using the SG (Savitzky-Golay) filtering method. Among them, the second bumps and burrs can be small bumps and burrs. For another example, the road center line and the road boundary line are subjected to secondary smoothing. Specifically, the first bumps and burrs at the initial positions of the road center line and the road boundary line are first smoothed using the mean method, and then the second bumps and burrs in the middle section of the road center line and the road boundary line are smoothed using the SG filtering method.

[0038] Step 103, updating the road network data based on the road centerline and the road boundary line.

[0039] In this embodiment, the execution entity may update the road network data based on the road centerline and the road boundary line.

[0040] Road network data can be the integrated data of road-level road network data and lane-level road network data. The main road shape problem scenarios of road network data are as follows:

[0041] First, the lane-level road network data and road-level road network data generated based on special collection are not in the lane group, resulting in the road conditions on the navigation end being rendered outside the lane group.

[0042] Second, the accuracy deviation of road-level network data leads to the misalignment of the lane-level network data generated at the intersection based on the road-level network data and the lane-level network data generated by special collection. Therefore, it is impossible to form an effective intersection surface.

[0043] Third, the accuracy deviation of road-level network data leads to the misalignment of the lane-level network data generated based on the road-level network data before and after the intersection with the lane-level network data generated by special collection. As a result, the navigation end rendering of the straight section has a fold.

[0044] The road network data is modified for scenarios such as the navigation side road outside the lane group surface. Based on the multi-source trajectory data, the full-scenario modification can be implemented, including but not limited to: high-low precision edge roads outside the lane group surface, all high-grade roads or low-grade roads outside the lane group surface, special collection of lane group current changes, and precision correction of road networks without lane-level coverage. Based on the road centerline and road boundary line, the road network data can be automatically updated.

[0045] In some embodiments, the steps for automatically updating the road network data may be as follows:

[0046] First, the road centerline and road boundary line are matched with the road network data to determine the road sections where the road network accuracy deviation is greater than the preset deviation threshold.

[0047] For example, by performing line distance difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined. For another example, by performing angle difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined. For another example, by performing line distance and angle difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined.

[0048] Then, based on the road centerline and the road boundary line, the road sections whose road network accuracy deviation is greater than a preset deviation threshold can be updated.

[0049] For road-level network data, automatic updates can be performed based on the road centerline, thereby effectively solving the accuracy problem of road-level network data.

[0050] Lane-level road network data can be automatically updated based on road center lines and road boundary lines, thereby effectively solving the error problem of lane-level road network data.

[0051] Reconstruct the traditional production model of road network data and break through the problem of large-scale integrated production of road network data. Create an effective data iteration model to continuously improve the quality of the road network and achieve a low-cost, large-scale, high-quality, and efficient data closed loop. After the road network data is updated, the navigation application can be updated in real time, thereby effectively solving the problem of road rendering on the product side and significantly improving the user navigation experience.

[0052] The disclosed embodiments provide a technology for automatic detection of road boundaries and center lines and automatic generation of road networks based on trajectory peak graphs. A low-cost, efficient, and high-precision automatic detection method for changes in road network data based on GPS (Global Positioning System) spatiotemporal trajectories is proposed. In view of the high cost of professional acquisition equipment and slow data updates, an in-depth study is conducted on the automatic update of urban road network topology and road shape changes. Based on low-cost and quickly accessible multi-source trajectory data, the refined application of the road network is integrated to achieve real-time updates of changes in road network topology and road network shape based on multi-source trajectory data, effectively improving the accuracy of road network data, with the advantages of low cost and high efficiency. The navigation application supports users' optimal navigation experience, and at the same time, provides high-precision lane-level data for autonomous driving, which assists in improving the safety of autonomous driving, and has important practical significance.

[0053] Continue to refer Figure 4, which shows the process 400 of another embodiment of the road network data update method according to the present disclosure. The road network data update method includes the following steps:

[0054] Step 401, perform intersection area trajectory retrieval on the crowd-sourced trajectory data to obtain area trajectory data.

[0055] In this embodiment, the execution subject of the road network data update method can perform intersection area trajectory retrieval on the crowd-sourced trajectory data to obtain area trajectory data.

[0056] The execution subject of the road network data update method is usually a server. The server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software or software modules (such as those used to provide distributed services), or as a single software or software module. Specific limitations are not made here.

[0057] In some embodiments, based on the crowd-sourced trajectory data, trajectory data can be retrieved within a preset range of the intersection. Based on the trajectory data, the trajectory points that extend the first preset distance along the road entering the intersection, and / or the trajectory points that extend a preset number or a second preset distance after passing through the intersection can be obtained to obtain area trajectory data.

[0058] Set a certain range (such as 100m×100m) within the intersection to retrieve the trajectory data in the area. Based on the road entering the intersection and extending a certain distance (such as 100m), calculate the straight road chain of the intersection, and obtain the trajectory points with the straight road chain as the smallest unit. At the same time, retain a certain number (such as 30) of subsequent trajectory points after the trajectory passes through the intersection, or the trajectory points within a certain distance range (such as 50m), effectively solving the problem of low trajectory coverage on short road segments.

[0059] Step 402, preprocess the area trajectory data.

[0060] In this embodiment, the above-mentioned execution subject can preprocess the area trajectory data.

[0061] Define the trajectory T = {p}1,…,pi,…,pn (0 < i < n), where pi is the i-th trajectory point. pi = {xi, yi, ti, vi, ai}, xi and yi are the longitude and latitude of pi in the geographic coordinate system, and ti, vi, and ai are the timestamp, linear velocity, and direction angle of pi respectively. The sampling frequency of two consecutive trajectory points is Δt, and the sampling distance is Δd.

[0062] Due to equipment failures and signal errors, trajectory data contains noise and errors, such as sampling anomalies (such as loss of trajectory points) and accuracy anomalies (such as position drift). At the same time, vehicles are stationary or driving slowly due to red lights, congestion, illegal parking, etc., leaving a large number of sampling points in a small area. In addition, crowd-source trajectory data comes from data collected by users under various traffic modes, and there is non-motor vehicle trajectory data. For example, when a navigation application is turned on while riding, the trajectory may even go in the opposite direction on the road. The trajectory fragments caused by the above-mentioned problems are not conducive to the update and identification of road network changes, and such trajectory fragments need to be eliminated.

[0063] In some embodiments, pre-processing may include, but is not limited to, at least one of the following:

[0064] First, the trajectory points of the regional trajectory data are matched to the roads in the road network data.

[0065] Match the vehicle trajectory data to the road network topology. Due to GPS positioning errors, coordinate system conversion errors, map accuracy errors, etc., the vehicle trajectory data may cause the vehicle GPS positioning point to deviate from the vehicle's driving road. In order to determine the exact position of the vehicle in the road network topology, map matching is required.

[0066] For example, the trajectory points of the regional trajectory data are used as the observation sequence, and the roads in the road network data are used as the hidden sequence. The hidden Markov model is used for trajectory matching to calculate the observation probability and transition probability. Based on the observation probability and transition probability, the Viterbi algorithm is used to calculate the hidden sequence under the maximum comprehensive probability as the trajectory point matching result. Based on the trajectory point matching result, the trajectory points of the regional trajectory data are matched to the roads in the road network data.

[0067] Second, the trajectory similarity of the regional trajectory data is calculated, and the trajectories whose trajectory similarity is less than the preset trajectory similarity threshold are eliminated.

[0068] The user's driving trajectory is relatively complex. The starting and ending points are uncertain, the driving direction is uncertain, the trajectories produced by various behaviors before the intersection, and the adjacent road trajectories affect each other, which will seriously affect the road accuracy. Based on Hausdorff distance (HausdorffDistance) and dynamic time warping (Dynamic Time Warping, DTW), the trajectory similarity of regional trajectory data is calculated. The trajectory fitting is more robust, fully utilizes the trajectory line information, and effectively eliminates the interference of invalid and messy trajectories.

[0069] Third, remove the drift trajectory points and / or lane change trajectory points in the regional trajectory data.

[0070] Trajectory data is a kind of point-sample data showing the distribution of road surface characteristics. Considering the spatial distribution characteristics of trajectory data, a density-based clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBSCAN) can be used to finely remove drift trajectory points and / or lane change trajectory points in regional trajectory data.

[0071] Make sections in a sequence perpendicular to the road's direction of travel according to the second preset step length (e.g., 10m). Calculate trajectory section features (e.g., trajectory surface width, trajectory point density), set point density thresholds (e.g., >= 3) and trajectory surface width thresholds (e.g., <= number of lanes * 3.5m). At the same time, based on the density comparison of the high aggregation density of trajectory data on the road surface and the low aggregation density at the road boundary, remove trajectory points drifting on both sides of the road. Calculate the travel direction angle of the trajectory segment within the section, adaptively calculate the direction difference threshold for different road sections, and remove lane change trajectory points.

[0072] Fourthly, the regional trajectory data is segmented based on the trajectory similarity fused with motion inertia constraints.

[0073] The trajectory data can reflect the vehicle's driving behaviors such as going straight, turning left, turning right, and turning around. Based on the angle, the trajectories of different driving behaviors can be segmented to obtain the trajectory of the same driving behavior. Based on the distance, the trajectories of vehicles driving in different positions in the same driving direction can be distinguished.

[0074] It is assumed that all trajectory points of each trajectory are collected in the same driving process (i.e., the driving mode is the same). Factors that express the inertia of vehicle motion may include but are not limited to: direction, position, speed, etc. Due to equipment failure or signal interruption, the position information cannot be obtained, resulting in data loss, making the distance between adjacent points of the trajectory far, or the angle of the local trajectory vector differs greatly from the driving direction of the entire trajectory, presenting a discontinuous driving state, thereby judging whether the vehicle's driving state has changed.

[0075] Based on the steering behavior of the trajectory during continuous driving, the angle and distance of the trajectory points are selected as the trajectory segmentation constraint factors. Set the sampling frequency t and the sampling distance d. When Δd>d (e.g., 20s) or Δd>d (e.g., 40m), the vehicle driving state is considered to have changed and the trajectory is cut into segments.

[0076] Fifth, eliminate the stop points in the regional trajectory data.

[0077] When a vehicle stops or travels at a low speed, a large amount of repeated sampling data (stop points or too close trajectory points) will be generated in a small space. Generally, at the intersection of the road, the density of trajectory points will be too high due to traffic lights or congestion. The trajectory data in the regional trajectory data whose trajectory point density is greater than the preset density threshold is thinned and smoothed. At the same time, for the trajectory points in the regional trajectory data whose timestamp sampling distance is less than the first preset distance threshold and / or whose position sampling distance is less than the second preset distance threshold, the trajectory points except the first trajectory point of repeated sampling are removed.

[0078] Sixth, eliminate non-motor vehicle trajectory data from regional trajectory data.

[0079] Usually, the speed of electric bicycles does not exceed 25km / h, and this speed does not reach the normal traffic flow speed of motor vehicles. Therefore, the characteristic parameters of the trajectories in the regional trajectory data are calculated, and the trajectories whose characteristic parameters in the regional trajectory data are not higher than the preset parameter threshold are eliminated. Among them, the characteristic parameters may include but are not limited to: maximum speed, average speed, minimum speed, trajectory vehicle probability, trajectory confidence, etc.

[0080] In addition, the regional trajectory data can be further filtered based on multi-dimensional feature parameters such as trajectory shape (such as glutinous rice ball trajectory), number of trajectory points (such as <5), number of large-angle trajectory points (such as >3), etc.

[0081] Step 403: Generate a trajectory peak graph based on the crowd-source trajectory data.

[0082] In this embodiment, the execution entity may generate a trajectory peak graph based on crowd-source trajectory data.

[0083] With the development of satellite positioning technology and the popularization of location-aware devices, users generate a large amount of trajectory data with rich dynamic information when using devices for navigation. Crowd-source trajectory data can be collected from navigation users' devices. Due to the large number of devices, low collection cost, fast dynamic update, strong generalization, and high road network coverage, crowd-source trajectory data can contain rich road spatiotemporal information, record vehicle location, movement direction, and status, and can perceive the geometry, topology, and traffic status of the road in real time. Therefore, crowd-source trajectory data can serve as an important data source for rapid generation and perception update of road networks. Crowd-source trajectory data can include road-level trajectory data and lane-level trajectory data.

[0084] Based on the large-scale high-quality multi-source trajectory data accumulated over multiple days, a trajectory peak map can be automatically generated. In some embodiments, based on the spatial distribution information of the multi-source trajectory data, Gaussian blur can be used to highlight the overall distribution of the trajectory, and the trajectory can be projected to generate a trajectory peak map. Among them, the trajectory peak map can distinguish the uplink and downlink trajectories according to the trajectory direction, which helps to improve the road boundary judgment.

[0085] Step 404: Generate a road centerline and a road boundary line based on the trajectory peak graph.

[0086] In this embodiment, the execution entity may generate a road centerline and a road boundary line based on the trajectory peak graph.

[0087] Based on the trajectory peak graph, a road centerline can be automatically generated, wherein the road centerline generation may refer to fitting a road centerline representing the current road or lane from a trajectory cluster having the same turning rule.

[0088] In some embodiments, based on the skeleton extraction model, the road centerline can be extracted from the trajectory peak map. The skeleton extraction model can be an improved UNet network. Figure 2 As shown, the UNet network has a symmetrical U-shaped encoding and decoding structure, which may include an encoder and a decoder. The trajectory peak map is input into the encoder to obtain a feature map of the trajectory peak map. Then, the feature map is input into the decoder to obtain the centerline of the road. The encoder may correspond to the downsampling process of the trajectory peak map, and the decoder may correspond to the upsampling process of the feature map. The encoder may obtain the feature map of the trajectory peak map through a CNN unit. The CNN unit may include but is not limited to: convolutional layer, BN layer, and ReLU layer. The encoder may use a maximum pooling operation to reduce the feature map, thereby extracting higher-level features to obtain scene information. The decoder may use an upsampling operation to restore the original trajectory peak map size. Upsampling may combine the information of each layer of downsampling and the input information of upsampling to restore detail information. As shown Figure 3 As shown in Figure 2, the decoder can introduce CBAM to infer the attention of the feature map along the channel dimension and the spatial dimension in turn to obtain the attention map. Then, the attention map is multiplied by the input feature map to perform adaptive feature refinement to obtain the road centerline.

[0089] In some embodiments, based on the image morphology thinning algorithm, the road center line can be extracted from the trajectory peak map. Among them, the image morphology thinning algorithm can include but is not limited to at least one of the following: based on the Hessian matrix line center extraction algorithm, the grayscale centroid method, the Zhang-Sun thinning algorithm, etc. Based on the Hessian matrix line center extraction algorithm, the trajectory imaging map can be regarded as a three-dimensional surface (peak), and the center line (ridge) is determined according to the local curvature to calculate the road center line. Based on the Hessian matrix line center extraction algorithm, a large number of Gaussian convolution operations are performed, and the calculation accuracy is high. The grayscale centroid method has a simple algorithm and high calculation efficiency. The Zhang-Sun thinning algorithm can continuously perform corrosion operations on the binary image pixel array, remove the contour boundary points, and obtain a skeleton connection line with a single pixel width. The thinned graphic skeleton retains most of the features of the original graphic. At least one of the image morphology thinning algorithms can extract at least one road center line. Weighted fusion of at least one road center line can generate the final road center line. Multi-method weighted fusion can improve the completeness and accuracy of road center line extraction.

[0090] In some embodiments, a road centerline can be extracted from a trajectory peak image based on a skeleton extraction model and an image morphology refinement algorithm. For example, a first road centerline is extracted from a trajectory peak image based on a skeleton extraction model. At least one second road centerline is extracted based on at least one item included in the image morphology refinement algorithm. The first road centerline and at least one second road centerline are weightedly fused to generate a road centerline. Based on the skeleton extraction model, a weighted fusion is performed with a traditional image morphology refinement algorithm to generate a road centerline, thereby improving the completeness and accuracy of the road centerline extraction.

[0091] Based on the road centerline, the road boundary line can be automatically generated by a trajectory point density algorithm. In some embodiments, based on the road centerline, the trajectory peak graph is cut across the road surface according to a first preset step length, and the road boundary line is extracted according to a trajectory density layering algorithm.

[0092] There are usually bumps and burrs on the road center line and the road boundary line. In some embodiments, the bumps and burrs on the road center line and the road boundary line are smoothed. For example, the first bumps and burrs at the initial positions of the road center line and the road boundary line are smoothed using the mean method. Among them, the first bumps and burrs can be large bumps and burrs. For another example, the second bumps and burrs in the middle section of the road center line and the road boundary line are smoothed using the SG (Savitzky-Golay) filtering method. Among them, the second bumps and burrs can be small bumps and burrs. For another example, the road center line and the road boundary line are subjected to secondary smoothing. Specifically, the first bumps and burrs at the initial positions of the road center line and the road boundary line are first smoothed using the mean method, and then the second bumps and burrs in the middle section of the road center line and the road boundary line are smoothed using the SG filtering method.

[0093] Step 405: Update the road network data based on the road centerline and the road boundary line.

[0094] In this embodiment, the execution entity may update the road network data based on the road centerline and the road boundary line.

[0095] Road network data can be the integrated data of road-level road network data and lane-level road network data. The main road shape problem scenarios of road network data are as follows:

[0096] First, the lane-level road network data and road-level road network data generated based on special collection are not in the lane group, resulting in the road conditions on the navigation end being rendered outside the lane group.

[0097] Second, the accuracy deviation of road-level network data leads to the misalignment of the lane-level network data generated at the intersection based on the road-level network data and the lane-level network data generated by special collection. Therefore, it is impossible to form an effective intersection surface.

[0098] Third, the accuracy deviation of road-level network data leads to the misalignment of the lane-level network data generated based on the road-level network data before and after the intersection with the lane-level network data generated by special collection. As a result, the navigation end rendering of the straight section has a fold.

[0099] The road network data is modified for scenarios such as the navigation side road outside the lane group surface. Based on the multi-source trajectory data, the full-scenario modification can be implemented, including but not limited to: high-low precision edge roads outside the lane group surface, all high-grade roads or low-grade roads outside the lane group surface, special collection of lane group current changes, and precision correction of road networks without lane-level coverage. Based on the road centerline and road boundary line, the road network data can be automatically updated.

[0100] In some embodiments, the steps for automatically updating the road network data may be as follows:

[0101] First, the road centerline and road boundary line are matched with the road network data to determine the road sections where the road network accuracy deviation is greater than the preset deviation threshold.

[0102] For example, by performing line distance difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined. For another example, by performing angle difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined. For another example, by performing line distance and angle difference calculation on the road centerline and road boundary line and the road network data, the road section where the road network accuracy deviation is greater than the preset deviation threshold can be determined.

[0103] Then, based on the road centerline and the road boundary line, the road sections whose road network accuracy deviation is greater than a preset deviation threshold can be updated.

[0104] For road-level network data, automatic updates can be performed based on the road centerline, thereby effectively solving the accuracy problem of road-level network data.

[0105] Lane-level road network data can be automatically updated based on road center lines and road boundary lines, thereby effectively solving the error problem of lane-level road network data.

[0106] Reconstruct the traditional production model of road network data and break through the problem of large-scale integrated production of road network data. Create an effective data iteration model to continuously improve the quality of the road network and achieve a low-cost, large-scale, high-quality, and efficient data closed loop. After the road network data is updated, the navigation application can be updated in real time, thereby effectively solving the problem of road rendering on the product side and significantly improving the user navigation experience.

[0107] from Figure 4 It can be seen that Figure 1 Compared with the corresponding embodiment, the road network data updating method in this embodiment adds a preprocessing step of multi-source trajectory data. Therefore, the scheme described in this embodiment can screen out high-quality trajectory data, which is conducive to the update and identification of road network changes.

[0108] Further references Figure 5 , which shows a process 500 of an embodiment of a lane-level road network data updating method according to the present disclosure. The lane-level road network data updating method comprises the following steps:

[0109] Step 501: Generate a lane-level trajectory peak map based on lane-level multi-source trajectory data.

[0110] In this embodiment, the execution body of the lane-level road network data updating method can generate a lane-level trajectory peak graph based on the lane-level multi-source trajectory data.

[0111] Lane-level multi-source trajectory data can be collected from the navigation user's device, including trajectory data generated when the user performs lane-level navigation.

[0112] Based on the large-scale high-quality lane-level multi-source trajectory data accumulated over multiple days, a lane-level trajectory peak map can be automatically generated. Among them, the lane-level trajectory peak map can be a multi-color trajectory peak map, which can distinguish the trajectory direction according to different colors. For example, the lane in the up direction can be green, and the lane in the down direction can be red.

[0113] Step 502: Generate a lane-level road centerline and a lane-level road boundary line based on the lane-level trajectory peak map.

[0114] In this embodiment, the above-mentioned execution entity can generate a lane-level road center line and a lane-level road boundary line based on the lane-level trajectory peak map.

[0115] Based on the lane-level trajectory peak graph, the lane-level road centerline can be automatically generated. The lane-level road centerline generation can refer to fitting a road centerline representing the current lane from a trajectory cluster with the same turning rule.

[0116] Based on the lane-level road centerline, the lane-level road boundary line can be automatically generated through the trajectory point density algorithm.

[0117] Step 503: Update the lane-level road network data based on the lane-level road centerline and the lane-level road boundary line.

[0118] In this embodiment, the above-mentioned execution entity can update the lane-level road network data based on the lane-level road center line and the lane-level road boundary line.

[0119] In some embodiments, the lane-level road network data automatic update steps may be as follows:

[0120] First, the lane-level road centerline and lane-level road boundary line are matched with the lane-level road network data to determine the road sections where the road network accuracy deviation is greater than the preset deviation threshold.

[0121] Then, based on the lane-level road centerline and the lane-level road boundary line, the road sections whose road network accuracy deviation is greater than a preset deviation threshold can be updated.

[0122] Lane-level road network data can be automatically updated based on the lane-level road centerline and lane-level road boundary line, thereby effectively solving the error problem of lane-level road network data.

[0123] The disclosed embodiment provides a lane-level road network update method, which is based on low-cost and quickly available lane-level multi-source trajectory data, integrates road network refinement applications, and realizes real-time updates of changes in road network topology and road network shape based on lane-level multi-source trajectory data, effectively improving the accuracy of road network data, and has the advantages of low cost and high efficiency. The navigation application supports the user's best navigation experience, and at the same time, provides high-precision lane-level data for autonomous driving, assists in improving the safety of autonomous driving, and has important practical significance.

[0124] Further references Figure 6 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a road network data updating device. Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0125] like Figure 6 As shown, the road network data updating device 600 of this embodiment may include: a first generating module 601, a second generating module 602 and an updating module 603. The first generating module 601 is configured to generate a trajectory peak graph based on multi-source trajectory data; the second generating module 602 is configured to generate a road centerline and a road boundary line based on the trajectory peak graph; the updating module 603 is configured to update the road network data based on the road centerline and the road boundary line.

[0126] In this embodiment, in the road network data updating device 600, the specific processing of the first generating module 601, the second generating module 602 and the updating module 603 and the technical effects thereof can be referred to in Figure 1 The relevant descriptions of steps 101 - 103 in the corresponding embodiment are not repeated here.

[0127] In some optional implementations of this embodiment, the first generation module 601 is further configured to: based on the spatial distribution information of the crowd-source trajectory data, use Gaussian blur to highlight the overall distribution of the trajectory, project the trajectory to generate a trajectory peak graph, wherein the trajectory peak graph distinguishes the upstream and downstream trajectories according to the trajectory direction.

[0128] In some optional implementations of the present embodiment, the second generation module 602 includes: a first extraction submodule, configured to extract the road centerline from the trajectory peak image based on a skeleton extraction model and / or an image morphological refinement algorithm; a second extraction submodule, configured to cut the trajectory peak image according to a first preset step length to cross-section the road surface based on the road centerline, and extract the road boundary line according to a trajectory density layering algorithm.

[0129] In some optional implementations of this embodiment, the skeleton extraction model includes an encoder and a decoder; and the first extraction submodule includes: an encoding unit, configured to input the trajectory peak map into the encoder to obtain a feature map of the trajectory peak map; a decoding unit, configured to input the feature map into the decoder to obtain the road centerline.

[0130] In some optional implementations of this embodiment, the decoder includes a feedforward convolutional neural network attention module; and the decoding unit is further configured to: input the feature map into the feedforward convolutional neural network attention module, infer the attention of the feature map along the channel dimension and the spatial dimension in sequence to obtain an attention map; multiply the attention map by the feature map to obtain the centerline of the road.

[0131] In some optional implementations of this embodiment, the image morphological thinning algorithm includes at least one of the following: a Hessian matrix-based line center extraction algorithm, a grayscale centroid method, and a Zhang-Sun thinning algorithm.

[0132] In some optional implementations of this embodiment, the first extraction submodule is further configured to: extract a first road centerline from the trajectory peak image based on a skeleton extraction model; extract at least one second road centerline based on at least one item included in the image morphological refinement algorithm; perform weighted fusion of the first road centerline and at least one second road centerline to generate a road centerline.

[0133] In some optional implementations of the present embodiment, the update module 603 includes: a first matching submodule, configured to match the road centerline and the road boundary line with the road network data to determine the road sections whose road network accuracy deviation is greater than the preset deviation threshold; an update submodule, configured to update the road sections whose road network accuracy deviation is greater than the preset deviation threshold based on the road centerline and the road boundary line.

[0134] In some optional implementations of this embodiment, the first matching submodule is further configured to: calculate the road center line and road boundary line based on line distance and / or angle difference with the road network data to determine the road sections where the road network accuracy deviation is greater than a preset deviation threshold.

[0135] In some optional implementations of this embodiment, the road network data updating device 600 further includes: a smoothing module configured to smooth the protrusions and burrs of the road center line and the road boundary line.

[0136] In some optional implementations of this embodiment, the smoothing module is further configured to: use the mean method to smooth the first bumps and burrs at the initial positions of the road centerline and the road boundary line; and / or use the SG filtering method to smooth the second bumps and burrs in the middle section of the road centerline and the road boundary line.

[0137] In some optional implementations of this embodiment, the road network data updating device 600 further includes: a retrieval module configured to perform intersection area trajectory retrieval on the multi-source trajectory data to obtain area trajectory data; and a preprocessing module configured to preprocess the area trajectory data.

[0138] In some optional implementations of this embodiment, the retrieval module is further configured to: retrieve trajectory data within a preset range of the intersection based on crowd-source trajectory data; based on the trajectory data, extend the trajectory points entering the intersection by a first preset distance, and / or extend the trajectory points after passing through the intersection by a preset number or a second preset distance, to obtain regional trajectory data.

[0139] In some optional implementations of this embodiment, the preprocessing module includes at least one of the following: a second matching submodule, configured to match trajectory points of regional trajectory data to roads in road network data; a first elimination submodule, configured to calculate trajectory similarity of regional trajectory data, and eliminate trajectories whose trajectory similarity is less than a preset trajectory similarity threshold; a second elimination submodule, configured to eliminate drift trajectory points and / or lane change trajectory points in regional trajectory data; a segmentation submodule, configured to perform trajectory segmentation on regional trajectory data based on motion inertia constraint fusion trajectory similarity; a third elimination submodule, configured to eliminate stop points in regional trajectory data; and a fourth elimination submodule, configured to eliminate non-motor vehicle trajectory data in regional trajectory data.

[0140] In some optional implementations of this embodiment, the second matching submodule is further configured to: take the trajectory points of the regional trajectory data as the observation sequence, take the roads in the road network data as the hidden sequence, use the hidden Markov model to perform trajectory matching, and calculate the observation probability and the transition probability; based on the observation probability and the transition probability, use the Viterbi algorithm to calculate the hidden sequence under the maximum comprehensive probability as the trajectory point matching result; based on the trajectory point matching result, match the trajectory points of the regional trajectory data to the roads in the road network data.

[0141] In some optional implementations of this embodiment, the first elimination submodule is further configured to: calculate the trajectory similarity of the regional trajectory data based on the Haushof distance and dynamic time planning.

[0142] In some optional implementations of this embodiment, the second elimination submodule includes: a elimination unit configured to adopt a density-based clustering algorithm to eliminate drift trajectory points and / or lane change trajectory points in the regional trajectory data.

[0143] In some optional implementations of this embodiment, the elimination unit is further configured to: section the regional trajectory data in an order perpendicular to the road travel direction according to a second preset step size, and eliminate drift trajectory points and / or lane change trajectory points in the regional trajectory data based on the characteristics of the section.

[0144] In some optional implementations of this embodiment, the segmentation submodule is further configured to: use the angle and distance of the trajectory point as trajectory segmentation constraint factors to perform trajectory segmentation on the regional trajectory data.

[0145] In some optional implementations of this embodiment, the third elimination submodule is further configured to: perform a thinning and smoothing operation on the trajectory data in the regional trajectory data whose trajectory point density is greater than a preset density threshold; and / or for the trajectory points in the regional trajectory data whose timestamp sampling distance is less than a first preset distance threshold and / or whose position sampling distance is less than a second preset distance threshold, eliminate the trajectory points except the first sampled trajectory point.

[0146] In some optional implementations of this embodiment, the fourth elimination submodule is further configured to: calculate characteristic parameters of the trajectory in the regional trajectory data, wherein the characteristic parameters include at least one of the following: maximum speed, average speed, minimum speed, trajectory vehicle probability, and trajectory confidence; and eliminate trajectories in the regional trajectory data whose characteristic parameters are not higher than a preset parameter threshold.

[0147] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0148] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0149] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0150] like Figure 7As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0151] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0152] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as the road network data update method. For example, in some embodiments, the road network data update method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the road network data update method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the road network data update method in any other appropriate manner (e.g., by means of firmware).

[0153] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0154] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0155] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0156] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0157] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0158] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0159] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions provided by this disclosure can be achieved, and this document does not limit this.

[0160] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for updating road network data, comprising: Generate trajectory peak graph based on crowd-source trajectory data; Based on the trajectory peak graph, generating a road centerline and a road boundary line; The road network data is updated based on the road centerline and the road boundary line.

2. The method according to claim 1, wherein: The generating of the trajectory peak graph based on the crowd-source trajectory data includes: According to the spatial distribution information of the crowd-source trajectory data, Gaussian blur is used to highlight the overall distribution of the trajectory, and the trajectory is projected to generate the trajectory peak graph, wherein the trajectory peak graph distinguishes the uplink and downlink trajectories according to the trajectory direction.

3. The method according to claim 1, wherein: The step of generating a road centerline and a road boundary line based on the trajectory peak graph includes: Extracting the road centerline from the trajectory peak image based on a skeleton extraction model and / or an image morphology thinning algorithm; Based on the road centerline, the trajectory peak graph is cut by cross-cutting the road surface according to a first preset step length, and the road boundary line is extracted according to a trajectory density layering algorithm.

4. The method according to claim 3, wherein: The skeleton extraction model includes an encoder and a decoder; and The step of extracting the road centerline from the trajectory peak image based on a skeleton extraction model and / or an image morphology thinning algorithm includes: Inputting the trajectory peak graph into the encoder to obtain a feature graph of the trajectory peak graph; The feature map is input into the decoder to obtain the road centerline.

5. The method according to claim 4, wherein: The decoder includes a feed-forward convolutional neural network attention module; as well as The step of inputting the feature map into the decoder to obtain the road centerline comprises: Inputting the feature map into the feedforward convolutional neural network attention module, and sequentially inferring the attention of the feature map along the channel dimension and the spatial dimension to obtain an attention map; The attention map is multiplied by the feature map to obtain the road centerline.

6. The method according to claim 5, wherein: The image morphology thinning algorithm includes at least one of the following: a Hessian matrix-based line center extraction algorithm, a grayscale centroid method, and a Zhang-Sun thinning algorithm.

7. The method according to claim 6, wherein: The step of extracting the road centerline from the trajectory peak image based on a skeleton extraction model and / or an image morphology thinning algorithm includes: Extracting a first road centerline from the trajectory peak graph based on the skeleton extraction model; Extracting at least one second road centerline based on at least one item included in the image morphological thinning algorithm; The first road centerline and the at least one second road centerline are weightedly merged to generate the road centerline.

8. The method according to claim 1, wherein: The updating of the road network data based on the road centerline and the road boundary line includes: Matching the road centerline and the road boundary line with the road network data to determine a road section where the road network accuracy deviation is greater than a preset deviation threshold; Based on the road centerline and the road boundary line, the road sections whose road network accuracy deviation is greater than a preset deviation threshold are updated.

9. The method according to claim 8, wherein: The matching of the road centerline and the road boundary line with the road network data to determine the road section whose road network accuracy deviation is greater than a preset deviation threshold comprises: The road center line and the road boundary line are calculated based on line distance and / or angle difference with the road network data to determine the road section where the road network accuracy deviation is greater than a preset deviation threshold.

10. The method according to any one of claims 1 to 9, wherein: The method further comprises: The protrusions and burrs of the road center line and the road boundary line are smoothed.

11. The method according to claim 10, wherein: The step of smoothing the protrusions and burrs on the road center line and the road boundary line includes: Smoothing the first protrusions and burrs at the initial positions of the road center line and the road boundary line by using a mean value method; and / or The SG filtering method is used to smooth the second protrusions and burrs in the middle section of the road center line and the road boundary line.

12. The method according to any one of claims 1 to 9, wherein: The method further comprises: Performing intersection area trajectory retrieval on the multi-source trajectory data to obtain area trajectory data; The regional trajectory data is preprocessed.

13. The method according to claim 12, wherein: The performing intersection area trajectory retrieval on the multi-source trajectory data to obtain the area trajectory data includes: Based on the multi-source trajectory data, trajectory data is retrieved within a preset range of the intersection; Based on the trajectory data, the trajectory points entering the road from the intersection are extended by a first preset distance, and / or the trajectory points after passing through the intersection are extended by a preset number or a second preset distance to obtain the regional trajectory data.

14. The method according to claim 12, wherein: The preprocessing of the regional trajectory data includes at least one of the following: Matching the trajectory points of the regional trajectory data to the roads of the road network data; Calculating the trajectory similarity of the trajectory data in the area, and eliminating trajectories whose trajectory similarity is less than a preset trajectory similarity threshold; Eliminating drift trajectory points and / or lane change trajectory points in the regional trajectory data; Based on the motion inertial constraint fusion trajectory similarity, the trajectory data of the area is segmented; Eliminating the stop points in the trajectory data of the area; The non-motor vehicle trajectory data in the regional trajectory data is eliminated.

15. The method according to claim 14, wherein: The step of matching the trajectory points of the regional trajectory data to the roads of the road network data comprises: Taking the trajectory points of the regional trajectory data as an observation sequence and the roads in the road network data as a hidden sequence, a hidden Markov model is used for trajectory matching to calculate the observation probability and the transition probability; Based on the observation probability and the transition probability, a hidden sequence under the maximum comprehensive probability is calculated using the Viterbi algorithm as a trajectory point matching result; Based on the trajectory point matching result, the trajectory points of the regional trajectory data are matched to the roads of the road network data.

16. The method according to claim 14, wherein: The calculating the trajectory similarity of the regional trajectory data includes: Based on Haushof distance and dynamic time planning, trajectory similarity of the regional trajectory data is calculated.

17. The method according to claim 14, wherein: The removing of drifting trajectory points and / or lane changing trajectory points in the regional trajectory data comprises: A density-based clustering algorithm is used to remove drift trajectory points and / or lane change trajectory points in the regional trajectory data.

18. The method according to claim 17, wherein: The method of using a density-based clustering algorithm to remove drifting trajectory points and / or lane-changing trajectory points in the regional trajectory data includes: The regional trajectory data is sectioned according to a second preset step length in an order perpendicular to the road travel direction, and based on the characteristics of the section, the drift trajectory points and / or lane change trajectory points in the regional trajectory data are eliminated.

19. The method according to claim 14, wherein: The step of performing trajectory segmentation on the regional trajectory data based on the fusion trajectory similarity of the motion inertia constraint includes: The angle and distance of the trajectory points are used as trajectory segmentation constraint factors to perform trajectory segmentation on the regional trajectory data.

20. The method according to claim 14, wherein: The removing of the stationary points in the regional trajectory data comprises: performing a thinning and smoothing operation on the trajectory data in the regional trajectory data whose trajectory point density is greater than a preset density threshold; and / or For trajectory points in the regional trajectory data whose timestamp sampling distance is less than a first preset distance threshold and / or whose position sampling distance is less than a second preset distance threshold, the trajectory points except the first sampled trajectory point are removed.

21. The method according to claim 14, wherein: The step of removing the non-motor vehicle trajectory data from the regional trajectory data includes: Calculating characteristic parameters of the trajectory in the regional trajectory data, wherein the characteristic parameters include at least one of the following: maximum speed, average speed, minimum speed, trajectory vehicle probability, and trajectory confidence; Trajectories whose characteristic parameters in the regional trajectory data are not higher than a preset parameter threshold are eliminated.

22. A road network data updating device, comprising: A first generating module is configured to generate a trajectory peak graph based on crowd-source trajectory data; A second generating module is configured to generate a road center line and a road boundary line based on the trajectory peak graph; The updating module is configured to update the road network data based on the road center line and the road boundary line.

23. The device according to claim 22, wherein: The first generation module is further configured to: According to the spatial distribution information of the crowd-source trajectory data, Gaussian blur is used to highlight the overall distribution of the trajectory, and the trajectory is projected to generate the trajectory peak graph, wherein the trajectory peak graph distinguishes the uplink and downlink trajectories according to the trajectory direction.

24. The device according to claim 22, wherein: The second generation module comprises: A first extraction submodule is configured to extract the road centerline from the trajectory peak image based on a skeleton extraction model and / or an image morphology thinning algorithm; The second extraction submodule is configured to cut the trajectory peak image based on the road centerline and the road surface according to a first preset step length, and extract the road boundary line according to a trajectory density layering algorithm.

25. The device according to claim 24, wherein: The skeleton extraction model includes an encoder and a decoder; and The first extraction submodule comprises: an encoding unit, configured to input the trajectory peak graph into the encoder to obtain a feature graph of the trajectory peak graph; The decoding unit is configured to input the feature map into the decoder to obtain the road centerline.

26. The device according to claim 25, wherein The decoder includes a feed-forward convolutional neural network attention module; and The decoding unit is further configured to: Inputting the feature map into the feedforward convolutional neural network attention module, and sequentially inferring the attention of the feature map along the channel dimension and the spatial dimension to obtain an attention map; The attention map is multiplied by the feature map to obtain the road centerline.

27. The device according to claim 26, wherein: The image morphology thinning algorithm includes at least one of the following: a Hessian matrix-based line center extraction algorithm, a grayscale centroid method, and a Zhang-Sun thinning algorithm.

28. The device according to claim 27, wherein The first extraction submodule is further configured to: Extracting a first road centerline from the trajectory peak graph based on the skeleton extraction model; Extracting at least one second road centerline based on at least one item included in the image morphological thinning algorithm; The first road centerline and the at least one second road centerline are weightedly merged to generate the road centerline.

29. The device according to claim 22, wherein: The update module includes: A first matching submodule is configured to match the road centerline and the road boundary line with the road network data to determine a road section where a road network accuracy deviation is greater than a preset deviation threshold; The updating submodule is configured to update the road sections whose road network accuracy deviation is greater than a preset deviation threshold based on the road centerline and the road boundary line.

30. The device according to claim 29, wherein: The first matching submodule is further configured to: The road center line and the road boundary line are calculated based on line distance and / or angle difference with the road network data to determine the road section where the road network accuracy deviation is greater than a preset deviation threshold.

31. The device according to any one of claims 22 to 30, wherein: The device also includes: The smoothing module is configured to smooth the protrusions and burrs of the road center line and the road boundary line.

32. The device according to claim 31, wherein The smoothing module is further configured to: Smoothing the first protrusions and burrs at the initial positions of the road center line and the road boundary line by using a mean value method; and / or The SG filtering method is used to smooth the second protrusions and burrs in the middle section of the road center line and the road boundary line.

33. The device according to any one of claims 22 to 30, wherein: The device also includes: A retrieval module is configured to perform intersection area trajectory retrieval on the multi-source trajectory data to obtain area trajectory data; The preprocessing module is configured to preprocess the regional trajectory data.

34. The device according to claim 33, wherein: The retrieval module is further configured to: Based on the multi-source trajectory data, trajectory data is retrieved within a preset range of the intersection; Based on the trajectory data, the trajectory points entering the road from the intersection are extended by a first preset distance, and / or the trajectory points after passing through the intersection are extended by a preset number or a second preset distance to obtain the regional trajectory data.

35. The device according to claim 33, wherein The preprocessing module includes at least one of the following: A second matching submodule is configured to match the trajectory points of the regional trajectory data to the roads of the road network data; A first elimination submodule is configured to calculate the trajectory similarity of the regional trajectory data and eliminate trajectories whose trajectory similarity is less than a preset trajectory similarity threshold; A second elimination submodule is configured to eliminate drift trajectory points and / or lane change trajectory points in the regional trajectory data; A segmentation submodule is configured to perform trajectory segmentation on the regional trajectory data based on motion inertial constraint fusion trajectory similarity; A third elimination submodule is configured to eliminate the stationary points in the regional trajectory data; The fourth elimination submodule is configured to eliminate the non-motor vehicle trajectory data in the regional trajectory data.

36. The device according to claim 35, wherein The second matching submodule is further configured to: Taking the trajectory points of the regional trajectory data as an observation sequence and the roads in the road network data as a hidden sequence, a hidden Markov model is used for trajectory matching to calculate the observation probability and the transition probability; Based on the observation probability and the transition probability, a hidden sequence under the maximum comprehensive probability is calculated using the Viterbi algorithm as a trajectory point matching result; Based on the trajectory point matching result, the trajectory points of the regional trajectory data are matched to the roads of the road network data.

37. The device according to claim 35, wherein: The first rejection submodule is further configured to: Based on Haushof distance and dynamic time planning, trajectory similarity of the regional trajectory data is calculated.

38. The apparatus of claim 35, wherein: The second elimination submodule comprises: The elimination unit is configured to adopt a density-based clustering algorithm to eliminate drift trajectory points and / or lane change trajectory points in the regional trajectory data.

39. The device according to claim 38, wherein The culling unit is further configured to: The regional trajectory data is sectioned according to a second preset step length in an order perpendicular to the road travel direction, and based on the characteristics of the section, the drift trajectory points and / or lane change trajectory points in the regional trajectory data are eliminated.

40. The apparatus of claim 35, wherein: The segmentation submodule is further configured to: The angle and distance of the trajectory points are used as trajectory segmentation constraint factors to perform trajectory segmentation on the regional trajectory data.

41. The apparatus of claim 35, wherein: The third rejection submodule is further configured to: performing a thinning and smoothing operation on the trajectory data in the regional trajectory data whose trajectory point density is greater than a preset density threshold; and / or For trajectory points in the regional trajectory data whose timestamp sampling distance is less than a first preset distance threshold and / or whose position sampling distance is less than a second preset distance threshold, the trajectory points except the first sampled trajectory point are removed.

42. The apparatus of claim 35, wherein: The fourth rejection submodule is further configured to: Calculating characteristic parameters of the trajectory in the regional trajectory data, wherein the characteristic parameters include at least one of the following: maximum speed, average speed, minimum speed, trajectory vehicle probability, and trajectory confidence; Trajectories whose characteristic parameters in the regional trajectory data are not higher than a preset parameter threshold are eliminated.

43. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 21.

44. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the method of any one of claims 1-21.

45. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 21.

Images