Road network update method, device, storage medium and electronic device

Abstract

Embodiments of the present application disclose a road network update method, device, storage medium, and electronic device. The method includes: determining a target area of road network information to be updated and a target driving trajectory of a specified vehicle within the target area; generating a road-trajectory image according to an initial road network corresponding to the target area and the target driving trajectory; inputting the road-trajectory image into a trained neural network model to output a target image, where the target image includes at least a target road that does not overlap with the initial road network; and updating the initial road network based on the target image. In this solution, the driving trajectory is combined with the neural network model to predict missing road data, which can improve the accuracy and efficiency of road network update.

Description

Technical Field

[0001] This application relates to the technical field of electronic devices, and in particular, to a road network update method, apparatus, storage medium, and electronic device. Background Art

[0002] There are various methods for updating digital maps, such as manual collection methods, remote sensing image recognition methods, etc. Among them, the manual collection method updates the road network by professional personnel and vehicles equipped with surveying instruments for actual road surveys; the remote sensing image recognition method requires professional personnel to continuously collect a sufficient number of remote sensing images to obtain a large number of available map images for recognition. However, these methods all face the problems of long update cycles and the need for a large amount of human and material resources. Summary of the Invention

[0003] Embodiments of this application provide a road network update method, apparatus, storage medium, and electronic device, which can improve the accuracy and update efficiency of road network updates.

[0004] In a first aspect, embodiments of this application provide a road network update method, including:

[0005] Determine a target area of the road network information to be updated and a target driving trajectory of a specified vehicle within the target area;

[0006] Generate a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory;

[0007] Input the road-trajectory image into a trained neural network model to output a target image, where the target image includes at least a target road, and the target road does not overlap with the initial road network;

[0008] Update the initial road network based on the target image.

[0009] In a second aspect, embodiments of this application provide a road network update apparatus, including:

[0010] A determination unit, configured to determine a target area of the road network information to be updated and a target driving trajectory of a specified vehicle within the target area;

[0011] A generation unit, configured to generate a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory;

[0012] A processing unit, configured to input the road-trajectory image into a trained neural network model to output a target image, where the target image includes at least a target road, and the target road does not overlap with the initial road network;

[0013] An update unit for updating the initial road network based on the target image.

[0014] In one embodiment, the generating unit is configured to:

[0015] Convert the initial road network and the target driving trajectory into a specified image format;

[0016] Perform data overlay on the converted initial network information and the target driving trajectory to obtain a road-trajectory image.

[0017] In one embodiment, the determining unit is configured to:

[0018] Obtain the candidate driving trajectories of the specified vehicle within the target area;

[0019] Cluster the candidate driving trajectories to obtain a plurality of trajectory clusters;

[0020] Extract the centerlines for different trajectory clusters to obtain the target driving trajectory.

[0021] In one embodiment, the update unit is configured to:

[0022] Obtain the geographical location information corresponding to the pixel units in the road-trajectory image;

[0023] Assign the geographical location information to the target image to determine the target geographical location information corresponding to the target road;

[0024] Map the target road to the initial road network according to the target geographical location information to update the initial road network.

[0025] In one embodiment, the device further includes:

[0026] A first acquisition unit for acquiring a sample initial road network corresponding to a sample area and a sample driving trajectory of a preset vehicle within the sample area before determining the target area of the road network information to be updated and the target driving trajectory of the specified vehicle within the target area, and generating a sample road-trajectory image according to the sample initial road network and the sample driving trajectory;

[0027] A second acquisition unit for acquiring the missing road information corresponding to the sample initial road network and generating a missing road image according to the missing road information;

[0028] A construction unit for constructing a training sample according to the sample road-trajectory image and the missing road image;

[0029] A training unit for training a preset neural network model based on the training sample to obtain a trained neural network model.

[0030] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are adapted to be loaded by a processor to execute the above-mentioned road network update method.

[0031] In a fourth aspect, an embodiment of the present application further provides an electronic device, including a processor and a memory, the processor is electrically connected to the memory, the memory is used to store instructions and data, and the processor is used to execute the above-mentioned road network update method.

[0032] In the embodiment of the present application, according to the initial road network corresponding to the target area of the road network information to be updated and the target driving trajectory within the target area, a road-trajectory image is generated and input into a trained neural network model, and then the initial road network is updated based on the target image output by the model. In this solution, the driving trajectory is combined with the neural network model to predict the missing road data, which can improve the accuracy and efficiency of road network update. Description of the Drawings

[0033] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0034] Figure 1 It is a flowchart of a road network update method provided by an embodiment of the present application.

[0035] Figure 2 It is another flowchart of a road network update method provided by an embodiment of the present application.

[0036] Figure 3 It is a structural diagram of a road network update device provided by an embodiment of the present application.

[0037] Figure 4 It is a structural diagram of an electronic device provided by an embodiment of the present application.

[0038] Figure 5 It is another structural diagram of an electronic device provided by an embodiment of the present application. Detailed Embodiments

[0039] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0040] The embodiments of the present application provide a road network update method, device, storage medium, and electronic device. The following will be described in detail respectively.

[0041] In one embodiment, a road network update method is provided, which is applied to electronic devices such as smartphones, tablets, and laptops. Refer to Figure 1 , the specific process of this road network update method can be as follows:

[0042] 101. Determine the target area of the road network information to be updated and the target driving trajectory of the specified vehicle within the target area.

[0043] Among them, the target area is a new area where road network missing mining needs to be done. The specified vehicle can be selected by the product manufacturer itself, and the driving trajectory can be obtained through a unified platform. For example, the specified vehicle can be a vehicle under a certain freight platform or passenger platform that has traveled within the target area.

[0044] In specific implementation, the driving trajectory data of all orders of the vehicle can be obtained through a timed task. Since there may be a lot of noise in the driving trajectory data, it is necessary to improve the quality of the driving trajectory data and extract the driving trajectory data that can restore the real road. That is, in some embodiments, the step of determining the target driving trajectory of the specified vehicle within the target area may include the following operations:

[0045] Obtain the candidate driving trajectories of the specified vehicle within the target area;

[0046] Cluster the driving trajectories to obtain multiple trajectory clusters;

[0047] Extract the center line for different trajectory clusters to obtain the target driving trajectory.

[0048] Specifically, the driving trajectories can be clustered by similarity calculation, and the center line is extracted for different driving trajectory clusters as the key trajectory to improve the trajectory quality.

[0049] When calculating the trajectory similarity, in order to distinguish the driving trajectories that conform to different road forms, the Frechet distance and the direction difference between the start and end points can be used to calculate the similarity between different trajectories. The calculation formula is as follows:

[0050] TSM = M Frechet+W directio *ζ

[0051] Among them, M Frechet represents the Frechet distance between two trajectories, W direction represents the direction difference between the starting and ending points of two trajectories, and ζ is a constant. The larger the value obtained by this formula, the lower the similarity; the smaller the value, the higher the similarity.

[0052] In addition, the determination formula for the angle difference is as follows:

[0053]

[0054] Among them, ω represents the angle difference between two trajectories. T ra represents trajectory a, and T rb represents trajectory b. If the direction difference between the starting and ending points of two trajectories is greater than the angle threshold γ, the direction difference between the two trajectories is 1; otherwise, it is 0.

[0055] Then, based on the similarity measurement results of the similarity matrix, the DBSCAN clustering algorithm is used to divide the trajectories into different trajectory clusters. The DBSCAN method has two important parameters, and the values of minSample (the minimum number of samples) and eps can be appropriately adjusted.

[0056] When extracting the center line, k candidate reference trajectories in the same cluster can be first selected using the Frechet distance, and the Frechet distance between each candidate and the remaining trajectories in the cluster is calculated respectively. The candidate trajectory with the minimum sum of distances is regarded as the reference trajectory; then, using the Force-attraction method, the candidate trajectory and the remaining trajectories in the same cluster are used to iteratively adjust the positions of the points until the attractive and repulsive forces of any point on the trajectory reach equilibrium to obtain a new position. The attractive force calculation formula is as follows:

[0057]

[0058] The repulsive force calculation formula is as follows:

[0059] F2(p j ) = s(x - p j )

[0060] Among them, p j represents any randomly selected point on the trajectory, p k represents a trajectory point in any same trajectory cluster, d(p i , p k ) represents the shortest distance from point p i to point p k , θ represents the angle between point p i and point p kThe directional difference, M and σ are parameters for determining gravity (with values of 1 and 10 respectively), s is the repulsive force coefficient (with a value of 0.005), and x represents point p i The new position.

[0061] In practical applications, before improving the trajectory quality, data preprocessing can also be performed on the trajectory using methods such as segmentation, denoising, and compression. Specifically as follows:

[0062] Trajectory segmentation: For the situation where a large amount of intermediate road information is missing due to signal loss resulting in missing sampling points, considering the normal sampling interval (3 - 5s) and the distance interval between adjacent trajectory points (20m), set the thresholds for the time interval and distance interval for the trajectory, and truncate between points with an overly large time interval or distance interval;

[0063] Trajectory denoising: For the situation where some trajectory lengths are extremely short or the number of trajectory points is extremely small due to trajectory segmentation, set the trajectory length threshold and the number of trajectory point threshold to eliminate them; for the position fluctuations of trajectory points caused by gps signal drift and the self - intersection of trajectories in a very small area due to congestion and human factors, eliminate them through the threshold of the directional difference between adjacent front and rear trajectory points;

[0064] Trajectory compression: Since the vehicle generates a large number of trajectory points in a local area due to stopping at traffic lights or traffic congestion, the Douglas - Peucker algorithm can be used to simplify the trajectory into a set of key points while retaining the shape characteristics of the original trajectory. (The overall idea is: for any trajectory tr, its starting point p s and the ending point p e are connected into a line segment. The farthest point p s between the starting point p e and the ending point p se with a length D k greater than the set distance threshold of 5m is regarded as a key point and retained. Then, connect the starting point (or ending point) and the key point p k into a line segment, and repeat the operation until the starting point (ending point) overlaps with a certain key point. Iteratively execute the key point extraction process on the remaining points on the trajectory t r Finally, obtain a new compressed trajectory formed by the key points.

[0065] 102. Generate a road - trajectory image according to the initial road network corresponding to the target area and the target driving trajectory.

[0066] Among them, the initial road network refers to the known road information in the current target area. In some embodiments, when generating a road - trajectory image according to the initial road network corresponding to the target area and the target driving trajectory, the following operations can specifically be included:

[0067] Convert the initial road network and the target driving trajectory into a specified image format;

[0068] Overlay the initial network information and the target driving trajectory after format conversion to obtain a road-trajectory image.

[0069] Specifically, according to the specified map extent shp file (i.e., the target area), obtain the driving trajectory data (after quality improvement) and road network data within the map extent. Crop the map extent shp file into multiple (e.g., 24) sub-boxes, and then use each sub-box to cut the driving trajectory and road network shp files to obtain multiple road data of size 1024*1024 and convert them into json format. Then, rasterize and vectorize the driving trajectory data and road network data in json format for the corresponding sub-images into tif format and perform data overlay, and save them on a tif image to obtain a road-trajectory image.

[0070] 103. Input the road-trajectory image into the trained neural network model, and output a target image, where the target image includes at least a target road, and the target road does not overlap with the initial road network.

[0071] In this embodiment, the target road is the missing road corresponding to the road network information in the target area. The trained neural network model can be an image segmentation model, which can replace the process of manual data annotation to infer the missing road network in a new area and output the missing road information.

[0072] In this embodiment, the model needs to be pre-trained. Specifically, the problem of obtaining the missing road can be regarded as a binary classification semantic segmentation task. Based on the image segmentation model unet, a better image segmentation model can be produced through model training to quickly and efficiently obtain the missing road information. That is, before determining the target area of the road network information to be updated and the target driving trajectory of the specified vehicle in the target area, the following operations may also be included:

[0073] Obtain the sample initial road network corresponding to the sample area and the sample driving trajectory of the preset vehicle in the sample area, and generate a sample road-trajectory image according to the sample initial road network and the sample driving trajectory;

[0074] Obtain the missing road information corresponding to the sample initial road network, and generate a missing road image according to the missing road information;

[0075] Construct a training sample according to the sample road-trajectory image and the missing road image;

[0076] Train the preset neural network model based on the training sample to obtain the trained neural network model.

[0077] First is the acquisition of the training sample set, which mainly includes two parts of data. One part is the data image (tif) of the driving trajectory and the road overlay, that is, the sample road-trajectory image; the other part is the missing road data image (tif). The former part of the data is obtained by converting the data format, and the latter part of the data is obtained through the image annotation platform.

[0078] Specifically, according to the specified map extent shp file, the driver trajectory data (after quality improvement) and road network data within the map extent can be obtained. The map extent shp file is cropped into multiple (such as 24) sub-boxes, and then each sub-box is used to cut the driver trajectory and road network shp files to obtain multiple road data with a size of 1024*1024 and convert them into json format.

[0079] When obtaining the missing road image, the driver trajectory data and road network data of the corresponding sub-map can be input into the missing road data annotation platform. Through manual recognition and annotation, the missing road json format is obtained and then converted into tif format. Among them, the conversion method is as follows: Read the xmin, xmax, ymin, ymax of the map shp, and calculate the resolution and average resolution values in the x / y directions: x_res / y_res and average_res (there will be a slight difference between x_res and y_res. When defining the size of a single pixel, average_res is used); after the tif attributes are defined, determine whether there is a road at each pixel position. If there is a road, assign the corresponding pixel value (the road network data is 4, the driving trajectory is 1, the missing road is -1, otherwise write 0).

[0080] When obtaining the sample road-trajectory image, the json format of the driver trajectory data and road network data of the corresponding sub-map is rasterized and vectorized into tif format (the same conversion method as in step 2) and the data is overlaid and saved on a single tif image (the pixel value of 5 indicates that both roads exist).

[0081] Then, the two types of data samples obtained are used as the data input of the preset neural network model to train the model.

[0082] To meet the pixel standard of the network input, the pixel values of the image (i.e., the sample road-trajectory image) and label (i.e., the missing road image) in the dataset can be changed. The background pixel value in the image is 100, the single road network pixel value is 120, the single trajectory pixel value is 200, and the pixel value shared by the road network and the trajectory is 170. The background class pixel value in the label is 0, and the rest is 255.

[0083] The data in the obtained training samples is divided according to the ratio of the training set to the test set of 4:1.

[0084] Due to the small amount of data in the training samples, during the model training process, data augmentation parameters (rotation, translation, shearing, flipping, resize) can be set to expand the training sample data.

[0085] The training parameters can be set according to the server GPU configuration and the size of the images in the training sample set. Set batch_size to 5, epoch to 500, and the initial learning rate to 10e-4. After 10 epochs when the loss no longer decreases, the learning rate is reduced by 0.1 times. To prevent overfitting, when the loss no longer decreases after 20 epochs, the training process ends prematurely.

[0086] During model training, the training images and their corresponding labels are input into the network, and the keras deep learning framework is used for training. The input images are convolved to different degrees to extract features at different levels, then upsampled to restore the dimensions, and skip-connected with the previous layer features, so that the feature maps contain both deep semantic feature information and shallow semantic feature information. The upsampling is implemented through transposed convolution. Finally, the original image size is restored, and two types of images are output, for the background and the missing roads. The output missing roads are compared with the label to calculate the loss, and the adam optimizer is used to optimize the learning rate to update the network weights, reducing the loss, and finally obtaining an optimized model.

[0087] Finally, the trained neural network model is used to infer the test images, and the inferred results are compared with the labels corresponding to the test images to calculate the iou (Intersection over Union) and miou (Mean Intersection over Union) to verify the accuracy of the model prediction.

[0088] 104. Update the initial road network based on the target image.

[0089] In this embodiment, the image results inferred by the image segmentation model can be vectorized to obtain the longitude and latitude information of the missing roads to update the road network. That is, in one implementation, when updating the initial road network based on the target image, the following operations can be specifically included:

[0090] Obtain the geographical location information corresponding to the pixel units in the road-trajectory image;

[0091] Assign the geographical location information to the target image to determine the target geographical location information corresponding to the target road;

[0092] Map the target road to the initial road network according to the target geographical location information to update the initial road network.

[0093] Specifically, the missing road images (tif) obtained by model training can be used to splice the results of multiple sub-images of each map sheet, and the longitude and latitude information of the input image (i.e., the road-trajectory image) can be assigned to the output result image. For example, the spliced PNG image can be read, each pixel can be traversed, and according to the RGB values, the pixels at the black background positions can be set to 0, and the rest of the roads to 1. At the same time, the upper left corner coordinates, the scales in the x and y directions are written, and a tif image with longitude and latitude information is exported.

[0094] Then, the center line of the vector image can be extracted using relevant components in GIS (Geographic Information System) and saved in the shp format. For example, the ArcScan module in ArcMap can be used to automatically extract the road center line (some parameters in the vector settings need to be adjusted).

[0095] In addition, according to actual needs, the obtained shp format of the missing roads can be converted into a specified format for saving, such as the json format.

[0096] In some embodiments, since the quality of the road network obtained by backtracking is poor, in order to make the obtained missing roads more convenient for subsequent data warehousing operations, the quality of the missing roads can also be improved through operations such as road closed-loop clipping, road thinning, and road and road network intersection clipping.

[0097] When performing road closed-loop clipping, it can be first determined whether the start and end points of the obtained missing road are the same. If they are the same, it is determined as a closed loop. Then, the number of points of the missing road is calculated and divided into three equal parts, and then the road is clipped into three roads.

[0098] When performing road thinning, the Douglas-Peucker algorithm can be used to simplify the trajectory into a set of key points while retaining the shape characteristics of the original trajectory (this step can refer to the algorithm explanation in the trajectory compression part).

[0099] When performing road and road network intersection clipping, the road network data of the specified map sheet can be first converted into the json format, and then all link objects with the LineString attribute are spliced together to form an object with the MultiLineString attribute; then, each missing road in the map sheet is respectively judged for intersection or inclusion with an object with the MultiLineString attribute. If there is an intersection or inclusion, the different parts of the missing road are taken out and saved.

[0100] As can be seen from the above, the road network update method provided in this embodiment generates a road-trajectory image based on the initial road network corresponding to the target area of the road network information to be updated and the target driving trajectory within the target area, inputs it into the trained neural network model, and then updates the initial road network based on the target image output by the model. In this solution, the driving trajectory is combined with the neural network model to predict missing road data, which can improve the accuracy and efficiency of road network update.

[0101] Reference Figure 2 , in another embodiment of the present application, a method for supplementing missing road networks based on image segmentation is also provided. By combining existing image recognition algorithms and trajectory data and making improvements and optimizations on this basis, the accuracy and coverage rate of road network supplementation can be greatly improved, and the problem of missing road shape information can be solved. This solution mainly includes five major links: trajectory data preprocessing and quality improvement, dataset generation, image segmentation method, image information backtracking, and missing road optimization. Specifically as follows:

[0102] (I) Trajectory data preprocessing and quality improvement

[0103] The trajectory data of all the driver's orders every day can be obtained through a scheduled task. Since there is a lot of noise in the trajectory, it is necessary to use methods of segmentation, denoising, and compression to preprocess the trajectory data, then cluster the trajectory through similarity calculation, and extract the center line as the key trajectory for different trajectory clusters to improve the trajectory quality.

[0104] (II) Dataset generation

[0105] The dataset is divided into two parts, the training set and the test set. The training set is used for training the image segmentation model and mainly includes two parts of data. One part is the data image (tif) of the superposition of the trajectory and the road, and the other part is the data image (tif) of the missing road; the former part is obtained through data format conversion, and the latter part is obtained through an image annotation platform. The test set is used for mining road network missing in new areas after the image segmentation model is finalized, and it is mainly the data image (tif) of the superposition of the trajectory and the road.

[0106] (III) Image segmentation method

[0107] The problem of obtaining missing roads is regarded as a semantic segmentation task of binary classification, and improvements and optimizations are made based on the image segmentation model unet, and finally the required missing roads are obtained through training.

[0108] (IV) Image information backtracking

[0109] Stitch the missing road images (tif) obtained from model training for the results of n sub - images of each map sheet, assign the longitude and latitude information of the input image to the output result image, then extract the road centerlines to obtain road information in shp format, and finally convert it to json format.

[0110] (5) Optimization of missing roads

[0111] Since the quality of the road network obtained by backtracking is poor, the quality of the missing roads can be improved through operations such as circular cropping, road thinning, and cropping by intersection with the road network.

[0112] The method for supplementing the missing road network provided in the embodiments of the present application obtains road data based on trajectory data, greatly reducing the labor and time costs required for manual collection and improving the efficiency of obtaining the missing road network. In the stage of updating roads, the method of image segmentation is adopted instead of the traditional map - matching algorithm, without any parameter tuning, improving the accuracy and coverage rate of the generated missing roads.

[0113] In another embodiment of the present application, a road network updating device is also provided. The road network updating device can be integrated into an electronic device in the form of software or hardware. The electronic device can specifically include devices such as mobile phones, tablet computers, and laptop computers. As Figure 3 shown, the road network updating device 300 may include: a determination unit 301, a generation unit 302, a processing unit 303, and an updating unit 304, where:

[0114] The determination unit 301 is used to determine the target area of the road network information to be updated and the target driving trajectory of a specified vehicle within the target area;

[0115] The generation unit 302 is used to generate a road - trajectory image according to the initial road network corresponding to the target area and the target driving trajectory;

[0116] The processing unit 303 is used to input the road - trajectory image into a trained neural network model and output a target image, where the target image includes at least a target road that does not overlap with the initial road network;

[0117] The updating unit 304 is used to update the initial road network based on the target image.

[0118] In an implementation manner, the generation unit 302 may be used to:

[0119] Convert the initial road network and the target driving trajectory into a specified image format;

[0120] Perform data superposition on the initial network information and the target driving trajectory after conversion to obtain a road - trajectory image.

[0121] In one embodiment, the determining unit 301 may be configured to:

[0122] Obtain candidate driving trajectories of the specified vehicle within the target area;

[0123] Cluster the candidate driving trajectories to obtain multiple trajectory clusters;

[0124] Extract centerlines for different trajectory clusters to obtain the target driving trajectory.

[0125] In one embodiment, the updating unit 304 may be configured to:

[0126] Obtain the geographical location information corresponding to the pixel units in the road-trajectory image;

[0127] Assign the geographical location information to the target image to determine the target geographical location information corresponding to the target road;

[0128] Map the target road to the initial road network according to the target geographical location information to update the initial road network.

[0129] In one embodiment, the road network updating device 300 may further include:

[0130] A first obtaining unit, configured to obtain a sample initial road network corresponding to a sample area and sample driving trajectories of a preset vehicle within the sample area before determining the target area of the road network information to be updated and the target driving trajectory of the specified vehicle within the target area, and generate a sample road-trajectory image according to the sample initial road network and the sample driving trajectories;

[0131] A second obtaining unit, configured to obtain missing road information corresponding to the sample initial road network and generate a missing road image according to the missing road information;

[0132] A constructing unit, configured to construct a training sample according to the sample road-trajectory image and the missing road image;

[0133] A training unit, configured to train a preset neural network model based on the training sample to obtain a trained neural network model.

[0134] As can be seen from the above, the road network updating device provided in the embodiments of the present application can generate a road-trajectory image according to the initial road network corresponding to the target area of the road network information to be updated and the target driving trajectory within the target area, input it into the trained neural network model, and then update the initial road network based on the target image output by the model. In this solution, driving trajectories are used in combination with a neural network model to predict missing road data, which can improve the accuracy and efficiency of road network updating.

[0135] In another embodiment of the present application, an electronic device is further provided. The electronic device may be a smart terminal such as a smart phone or a tablet computer. As Figure 4 shown, the electronic device 400 includes a processor 401 and a memory 402. Among them, the processor 401 is electrically connected to the memory 402.

[0136] The processor 401 is the control center of the electronic device 400. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or loading applications stored in the memory 402, and calling data stored in the memory 402, it executes various functions of the electronic device and processes data, thereby monitoring the entire electronic device.

[0137] In this embodiment, the processor 401 in the electronic device 400 will load the instructions corresponding to the processes of one or more applications into the memory 402 according to the following steps, and the processor 401 will run the applications stored in the memory 402 to implement various functions:

[0138] Determine the target area of the road network information to be updated, and the target driving trajectory of the specified vehicle in the target area;

[0139] Generate a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory;

[0140] Input the road-trajectory image into the trained neural network model to output a target image, where the target image includes at least a target road, and the target road does not overlap with the initial road network;

[0141] Update the initial road network based on the target image.

[0142] In an implementation manner, when generating a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory, the processor 401 specifically performs the following operations:

[0143] Convert the initial road network and the target driving trajectory into a specified image format;

[0144] Perform data overlay on the initial network information and the target driving trajectory after converting the format to obtain a road-trajectory image.

[0145] In an implementation manner, when determining the target driving trajectory of the specified vehicle in the target area, the processor 401 may perform the following operations:

[0146] Obtain the candidate driving trajectories of the specified vehicle in the target area;

[0147] Cluster the candidate driving trajectories to obtain multiple trajectory clusters;

[0148] Extract the center line for different trajectory clusters to obtain the target driving trajectory of the object.

[0149] In one embodiment, when updating the initial road network based on the target image, the processor 401 may perform the following operations:

[0150] Obtain the geographical location information corresponding to the pixel units in the road-trajectory image;

[0151] Assign the geographical location information to the target image to determine the target geographical location information corresponding to the target road;

[0152] Map the target road to the initial road network according to the target geographical location information to update the initial road network.

[0153] In one embodiment, before determining the target area of the road network information to be updated and the target driving trajectory of the specified vehicle in the target area, the processor 401 may perform the following operations:

[0154] Obtain the sample initial road network corresponding to the sample area and the sample driving trajectory of the preset vehicle in the sample area, and generate a sample road-trajectory image according to the sample initial road network and the sample driving trajectory;

[0155] Obtain the missing road information corresponding to the sample initial road network, and generate a missing road image according to the missing road information;

[0156] Construct a training sample according to the sample road-trajectory image and the missing road image;

[0157] Train the preset neural network model based on the training sample to obtain a trained neural network model.

[0158] The memory 402 can be used to store applications and data. The applications stored in the memory 402 contain instructions that can be executed in the processor. The applications can form various functional modules. The processor 401 executes various functional applications and road network updates by running the applications stored in the memory 402.

[0159] In some embodiments, as Figure 5 shown, the electronic device 400 further includes: a display screen 403, a control circuit 404, a radio frequency circuit 405, an input unit 406, a sensor 408, and a power supply 409. Among them, the processor 401 is electrically connected to the display screen 403, the control circuit 404, the radio frequency circuit 405, the input unit 406, the camera 407, the sensor 408, and the power supply 409 respectively.

[0160] The display screen 403 can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device. These graphical user interfaces can be composed of images, texts, icons, videos, and any combination thereof.

[0161] The control circuit 404 is electrically connected to the display screen 403 and is used to control the display screen 403 to display information.

[0162] The radio frequency circuit 405 is used to receive and transmit radio frequency signals to establish wireless communication with other electronic devices or the electronic device itself through wireless communication, and to receive and transmit signals between the server and other electronic devices.

[0163] The input unit 406 can be used to receive input digital, character information, or user characteristic information (such as fingerprints), and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function controls. Among them, the input unit 406 can include a fingerprint recognition module.

[0164] The camera 407 can be used to collect image information. Among them, the camera can be a single camera with one lens, or can have two or more lenses.

[0165] The sensor 408 is used to collect external environmental information. The sensor 408 can include an ambient light sensor, an acceleration sensor, a light sensor, a motion sensor, and other sensors.

[0166] The power supply 409 is used to supply power to each component of the electronic device 400. In some embodiments, the power supply 409 can be logically connected to the processor 401 through a power management system, so as to implement functions such as management of charging, discharging, and power consumption management through the power management system.

[0167] Although Figure 5 not shown in the figure, the electronic device 400 may further include a speaker, a Bluetooth module, etc., which will not be elaborated here.

[0168] As can be seen from the above, the electronic device provided in the embodiment of the present application can generate a road-trajectory image according to the initial road network corresponding to the target area of the road network information to be updated and the target driving trajectory within the target area, input it into the trained neural network model, and then update the initial road network based on the target image output by the model. In this solution, the driving trajectory is combined with the neural network model to predict missing road data, which can improve the accuracy and efficiency of road network update.

[0169] In some embodiments, a computer-readable storage medium is further provided. Multiple instructions are stored in the storage medium, and the instructions are suitable for being loaded by a processor to execute any of the above road network update methods.

[0170] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which can include: read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, etc.

[0171] The above has introduced in detail the road network update method, device, storage medium and electronic device provided by the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those skilled in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A road network update method, characterized in that, including: determining a target area of road network information to be updated and a target driving trajectory of a specified vehicle within the target area; generating a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory; inputting the road-trajectory image into a trained neural network model to output a target image, where the target image at least includes a target road that does not overlap with the initial road network, and the trained neural network model is an image segmentation model; updating the initial road network based on the target image; before determining the target area of the road network information to be updated and the target driving trajectory of the specified vehicle within the target area, further including: obtaining a sample initial road network corresponding to a sample area and a sample driving trajectory of a preset vehicle within the sample area, and generating a sample road-trajectory image according to the sample initial road network and the sample driving trajectory; obtaining missing road information corresponding to the sample initial road network and generating a missing road image according to the missing road information; constructing a training sample according to the sample road-trajectory image and the missing road image; training a preset neural network model based on the training sample to obtain a trained neural network model; determining the target driving trajectory of the specified vehicle within the target area includes: obtaining candidate driving trajectories of the specified vehicle within the target area; clustering the candidate driving trajectories through similarity calculation to obtain a plurality of trajectory clusters; extracting centerlines for different trajectory clusters to obtain the target driving trajectory.

2. The road network update method according to claim 1, wherein The generating a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory includes: converting the initial road network and the target driving trajectory into a specified image format; performing data superposition on the converted initial network information and the target driving trajectory to obtain a road-trajectory image.

3. The road network update method according to claim 1, wherein The updating the initial road network based on the target image includes: obtaining geographical location information corresponding to pixel units in the road-trajectory image; assigning the geographical location information to the target image to determine target geographical location information corresponding to the target road; mapping the target road to the initial road network according to the target geographical location information to update the initial road network.

4. A road network update device, characterized in that, including: a determining unit for determining a target area of road network information to be updated and a target driving trajectory of a specified vehicle within the target area; determining the target driving trajectory of the specified vehicle within the target area includes: obtaining candidate driving trajectories of the specified vehicle within the target area; clustering the candidate driving trajectories through similarity calculation to obtain a plurality of trajectory clusters; extracting centerlines for different trajectory clusters to obtain the target driving trajectory; a generating unit for generating a road-trajectory image according to the initial road network corresponding to the target area and the target driving trajectory; A processing unit, configured to input the road-trajectory image into a trained neural network model and output a target image, where the target image at least includes a target road that does not overlap with the initial road network, and the trained neural network model is an image segmentation model; An updating unit, configured to update the initial road network based on the target image; Before determining the target area of the road network information to be updated and the target driving trajectory of a specified vehicle in the target area, it further includes: Obtaining a sample initial road network corresponding to a sample area and a sample driving trajectory of a preset vehicle in the sample area, and generating a sample road-trajectory image according to the sample initial road network and the sample driving trajectory; Obtaining missing road information corresponding to the sample initial road network, and generating a missing road image according to the missing road information; Constructing a training sample according to the sample road-trajectory image and the missing road image; Training a preset neural network model based on the training sample to obtain a trained neural network model.

5. The road network update device according to claim 4, characterized in that, The generating unit is configured to: Convert the initial road network and the target driving trajectory into a specified image format; Perform data superposition on the converted initial network information and the target driving trajectory to obtain a road-trajectory image.

6. The road network updating device according to claim 5, wherein The updating unit is configured to: Obtain the geographical location information corresponding to the pixel unit in the road-trajectory image; Assign the geographical location information to the target image to determine the target geographical location information corresponding to the target road; Map the target road to the initial road network according to the target geographical location information to update the initial road network.

7. A computer-readable storage medium, characterized in that, Multiple instructions are stored in the storage medium, and the instructions are suitable for being loaded by a processor to execute the road network updating method according to any one of claims 1-3.

8. An electronic device, characterized in that, It includes a processor and a memory, the processor is electrically connected to the memory, and the memory is used to store instructions and data; the processor is used to execute the road network updating method according to any one of claims 1-3.

Images