Elevation-information processing method and apparatus, and electronic device and storage medium

Abstract

The present application relates to the technical field of maps, and in particular relates to an elevation-information processing method and apparatus, and an electronic device and a storage medium, which can be applied in fields such as maps, traffic, and autonomous driving. The method comprises: performing sampling on lane data of roads in a target area, so as to obtain a plurality of sampling points; determining the plurality of sampling points and a plurality of elevation values of each of the plurality of sampling points from a plurality of point cloud identification results which are obtained by identifying a plurality of point clouds of the target area, and using the plurality of elevation values of each sampling point as candidate elevation values of the sampling point, so as to form a set of candidate elevation values of the sampling point; performing fitting on the plurality of sampling points to obtain a fitting result, and on the basis of the fitting result, determining one or more target sampling points among the plurality of sampling points; and respectively executing the following operations for a set of candidate elevation values of each of the one or more target sampling points: determining confidence levels of candidate elevation values in the set of candidate elevation values of the target sampling point, and determining a valid elevation value of the target sampling point on the basis of the confidence levels of the candidate elevation values.

Description

Method, device, electronic device and storage medium for processing elevation information

[0001] This application claims priority to the Chinese patent application with application number 202311212439.6 filed on September 20, 2023, and invention name “A method, device, electronic device and storage medium for processing elevation information”. Technical Field

[0002] The present application relates to the field of map technology, and in particular to a method, device, electronic device and storage medium for processing elevation information. Background Art

[0003] Processing elevation information is crucial for high-precision maps. One related automated elevation processing method uses a trajectory-matching road network approach. This method first matches trajectories to roads using a trajectory adsorption algorithm. Then, based on the elevation information in the trajectory data, it automatically assigns elevation values ​​to road lanes and other data.

[0004] However, current trajectory-matching road network methods rely primarily on vehicle trajectories, which contain limited information. Furthermore, the installation locations of Global Positioning Systems (GPS) on different vehicles vary, leading to errors when using GPS to estimate elevation differences from the ground, affecting the accuracy of GPS-derived elevation information. Consequently, the accuracy of automated elevation processing using trajectory-matching road network algorithms is currently affected by multiple factors, including road network accuracy, GPS positioning accuracy, and trajectory length, making it difficult to improve the upper limit of accuracy.

[0005] Summary of the Invention

[0006] Embodiments of the present application provide a method, device, electronic device, and storage medium for processing elevation information to improve the accuracy of elevation information processing.

[0007] The present invention provides a method for processing elevation information, including:

[0008] determining, from a plurality of point cloud recognition results obtained by recognizing a plurality of point clouds of the target area, the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points, and using the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point to form a candidate elevation value set for the sampling point;

[0009] Fitting the multiple sampling points to obtain fitting results, and determining one or more target sampling points among the multiple sampling points based on the fitting results;

[0010] For each candidate elevation value set of the one or more target sampling points, the following operations are performed: determining a confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining a valid elevation value of the target sampling point based on the confidence level of each candidate elevation value.

[0011] An embodiment of the present application provides a device for processing elevation information, comprising:

[0012] A sampling unit, used to sample lane data of roads in a target area to obtain a plurality of sampling points;

[0013] a fitting unit configured to determine, from a plurality of point cloud recognition results obtained by recognizing a plurality of point clouds of the target area, the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points; use the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point to form a candidate elevation value set of the sampling point; perform fitting on the plurality of sampling points to obtain fitting results; and determine, based on the fitting results, one or more target sampling points among the plurality of sampling points;

[0014] The fusion unit is configured to perform the following operations on a candidate elevation value set of each target sampling point in the one or more target sampling points: determining a confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining a valid elevation value of the target sampling point based on the confidence level of each candidate elevation value.

[0015] An electronic device provided in an embodiment of the present application includes a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of any one of the above-mentioned elevation information processing methods.

[0016] An embodiment of the present application provides a computer-readable storage medium, which includes a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of any one of the above-mentioned elevation information processing methods.

[0017] An embodiment of the present application provides a computer program product, which includes a computer program, and the computer program is stored in a computer-readable storage medium; when a processor of an electronic device reads the computer program from the computer-readable storage medium, the processor executes the computer program, so that the electronic device performs the steps of any one of the above-mentioned elevation information processing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] FIG1 is a schematic diagram of an example of an application scenario in an embodiment of the present application;

[0020] FIG2 is a schematic diagram of an elevation data processing flow in a data production link in an embodiment of the present application;

[0021] FIG3 is a flowchart of an implementation method for processing elevation information in an embodiment of the present application;

[0022] FIG4 is a schematic diagram of a lane line recognition result in a loading target area in an embodiment of the present application;

[0023] FIG5 is a schematic diagram showing a comparison of the effects of a RANSAC fitting algorithm and a linear fitting algorithm in an embodiment of the present application;

[0024] FIG6 is a schematic diagram of a plane fitting process in an embodiment of the present application;

[0025] FIG7 is a schematic diagram of a plane and a standard normal vector in an embodiment of the present application;

[0026] FIG8 is a flow chart of a method for determining the confidence level of a candidate elevation value in an embodiment of the present application;

[0027] FIG9 is a schematic diagram of an interpolation and filling process in an embodiment of the present application;

[0028] FIG10 is an optimization flowchart based on a graph propagation algorithm in an embodiment of the present application;

[0029] FIG11 is a schematic diagram of a road network diagram in an embodiment of the present application;

[0030] FIG12 is a schematic diagram of a candidate set of elevation categories of a road and a corresponding candidate graph in an embodiment of the present application;

[0031] FIG13 is a schematic diagram of a graph propagation optimization process in an embodiment of the present application;

[0032] FIG14 is a flowchart of an implementation method of a void fitting method in an embodiment of the present application;

[0033] FIG15 is a schematic diagram of a hole fitting effect at an intersection according to an embodiment of the present application;

[0034] FIG16 is a schematic diagram of the structure of a device for processing elevation information according to an embodiment of the present application;

[0035] FIG17 is a schematic diagram of a hardware structure of an electronic device to which an embodiment of the present application is applied;

[0036] FIG18 is a schematic diagram of a hardware structure of another electronic device to which an embodiment of the present application is applied. DETAILED DESCRIPTION

[0037] To make the purpose, technical solutions, and advantages of the embodiments of the present application more clear, the technical solutions of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the technical solutions of the present application, but not all of them. Based on the embodiments described in the present application documents, ordinary technicians in this field can derive other embodiments without making any creative efforts.

[0038] The following is an introduction to some concepts involved in the embodiments of this application.

[0039] Elevation: refers to the distance from a certain point along the plumb line to the absolute base surface, which is called absolute elevation, or simply elevation. The absolute base surface is generally based on the characteristic sea surface at a certain coastal location, and the elevation of this characteristic water surface is 0.000 meters. The distance from a certain point along the plumb line to a certain assumed leveling base surface is called the assumed elevation. In the embodiment of the present application, the elevation information refers to the elevation value of a point. In the embodiment of the present application, there are three types of elevation values, namely candidate elevation values, valid elevation values, and target elevation values. Candidate elevation values ​​refer to elevation values ​​obtained through point cloud recognition. Valid elevation values ​​refer to elevation values ​​obtained by weighted fusion of candidate elevation values. Target elevation value refers to the elevation value finally determined based on the optimization of the graph propagation algorithm.

[0040] Point cloud: also known as laser point cloud, refers to a collection of laser scanning points obtained by laser scanning the surface of an object in three-dimensional space. The points in the point cloud can correspond to points on the surface of an object in three-dimensional space. Therefore, point clouds can be used to represent objects in three-dimensional space. In business scenarios related to high-precision maps, point clouds can be collected for road space, and the point cloud portion belonging to the road surface can be identified from the collected point cloud. The point cloud recognition result in the embodiment of the present application refers to the recognition result of the point cloud collected from the road space in the target area, including the latitude and longitude information, elevation values, category information (such as lane line category, guardrail category), etc. of the points in the point cloud.

[0041] Topological relationships refer to the connectivity relationships (also known as access relationships) between roads. For example, if Road A and Road B are accessible, they are connected. In intelligent or autonomous driving scenarios, vehicles can move from one road / lane to another based on the connectivity relationships between roads / lanes, enabling intelligent or autonomous driving based on high-precision maps.

[0042] Random Sampling Consensus (RANSAC): This algorithm calculates the mathematical model parameters of the data based on a set of sample data that contains abnormal data to obtain valid sample data. The basic assumption of the RANSAC algorithm is that the sample contains correct data (inliers, data that can be described by the model, also known as inliers) and abnormal data (outliers, data that deviates far from the normal range and cannot fit the mathematical model, also known as outliers), that is, the data set contains noise. This abnormal data may be caused by incorrect measurements, incorrect assumptions, incorrect calculations, etc. At the same time, RANSAC also assumes that, given a set of correct data, there is a method to calculate the model parameters that conform to this data.

[0043] The method for processing elevation information in the embodiment of the present application can be applied to the mapping field, the transportation field, the autonomous driving field, etc.

[0044] For autonomous driving, changes in geographic elevation data are crucial, helping the vehicle better control its longitudinal and lateral directions. For example, autonomous driving applications require determining slope information and driving strategies based on the elevation differences between adjacent features.

[0045] However, current automated processing methods for elevation information in the mapping field primarily rely on trajectory matching to road networks, using vehicle trajectory points to match the road network. This method calculates the distance between the vehicle trajectory points and roads in the road network, and then identifies the roads that match the vehicle trajectory points based on these distances. However, because vehicle trajectory points are collected using GPS positioning technology, their positions are subject to certain fluctuations due to GPS positioning accuracy limitations. Therefore, existing methods for matching road networks are prone to mismatching.

[0046] Therefore, the accuracy of the current trajectory matching road network method is affected by multiple factors such as road network accuracy, GPS positioning accuracy, and trajectory length, and the upper limit of accuracy is difficult to break through.

[0047] In view of this, embodiments of the present application provide a method, device, electronic device, and storage medium for processing elevation information. The elevation information processing method in this application is implemented based on point cloud recognition results. Multiple sampling points are determined from the point cloud recognition results, and candidate elevation values ​​for the multiple sampling points are obtained. Plane fitting is performed on the multiple sampling points to effectively remove noise points and obtain useful target sampling points. Based on this, these candidate elevation values ​​for the same target sampling point are fused based on the confidence level of the candidate elevation values ​​corresponding to the target sampling points, and the valid elevation values ​​are optimized based on the topological relationship between roads to obtain a processed target elevation value. This processing is based on point cloud recognition results. Compared to current trajectory matching road network methods that primarily rely on trajectories, this application directly evaluates the elevation accuracy of road lane data based on point cloud recognition results. Point cloud recognition results are unaffected by factors such as GPS installation location and contain richer and more accurate information than trajectory data. Furthermore, through methods such as resampling, plane fitting, and graph propagation, road lane data and other data are effectively and automatically processed for elevation, significantly improving accuracy.

[0048] On this basis, based on the point cloud recognition results, automatically assigning elevation data to the road lanes in the map can effectively improve the quality of the elevation data in the current map, and then improve the authenticity of the real environment simulated in the autonomous driving simulation link, so as to provide objects in need (such as driving objects) with a simulation environment that is more consistent with the real environment.

[0049] The following describes some embodiments of the present application in conjunction with the accompanying drawings. The embodiments described herein are only used to illustrate and explain the present application and are not intended to limit the present application. In addition, the embodiments and features in the embodiments of the present application may be combined with each other if there is no conflict.

[0050] As shown in Figure 1, which is a schematic diagram of an application scenario of an embodiment of the present application, the application scenario diagram includes two terminal devices 110 and a server 120.

[0051] In an embodiment of the present application, the terminal device 110 includes but is not limited to mobile phones, tablet computers, laptop computers, desktop computers, e-book readers, intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, aircraft and other devices. Clients related to map and elevation information processing can be installed on the terminal device. The client can be software (such as a browser, map software, etc.), or it can be a web page, applet, etc. The server 120 is a background server corresponding to the software or web page, applet, etc., or it is a server specifically used for elevation information processing, which is not specifically limited in this application. The server 120 can be an independent physical server, or it can be a server cluster or distributed system composed of multiple physical servers, or it can be a cloud server that provides cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDN), and basic cloud computing services such as big data and artificial intelligence platforms.

[0052] The elevation information processing method in each embodiment of the present application can be performed by an electronic device. The electronic device can be a terminal device 110 or a server 120. That is, the method can be performed independently by the terminal device 110 or the server 120, or jointly by the terminal device 110 and the server 120. For example, when performed independently by the server 120, the server 120 may first obtain multiple point clouds of a target area acquired at different times, identify these point clouds to obtain multiple point cloud recognition results, and sample lane data of roads within the target area to obtain multiple sampling points. The server 120 then determines multiple sampling points from each of the obtained point cloud recognition results and obtains a set of candidate elevation values ​​for each of the multiple sampling points. Based on the point cloud recognition results, the server 120 performs fitting on the multiple sampling points to obtain fitting results, and determines a target sampling point from the multiple sampling points based on the fitting results. Furthermore, the server 120 performs weighted fusion on each candidate elevation value in the candidate elevation value set based on the confidence level of each candidate elevation value in the candidate elevation value set for each target sampling point to obtain a valid elevation value for each target sampling point. In addition, the server 120 may also optimize the effective elevation value of each target sampling point based on the topological relationship between the roads in the target area to obtain the target elevation value of each target sampling point.

[0053] On the basis of the above, the server 120 can improve the quality of the elevation data in the map based on the obtained target elevation value, thereby improving the authenticity of the simulation of the real environment in the autonomous driving simulation link, so as to provide a simulation environment that is more consistent with the real environment to objects in need.

[0054] As shown in Figure 2, it is a schematic diagram of the elevation data processing flow of a data production link in an embodiment of the present application. Among them, the data production link refers to delivering map data to demand parties such as car manufacturers.

[0055] During the data production phase, an elevation data processing process can be added, such as the elevation batch processing portion indicated by the arrow in Figure 2, to automatically process road lane data and other data to meet customer requirements for elevation accuracy. Therefore, the elevation information processing method in the embodiments of this application can be applied to the aforementioned data production phase to improve the accuracy of elevation information in map data.

[0056] In an exemplary embodiment, the terminal device 110 and the server 120 may communicate with each other via a communication network.

[0057] In an exemplary embodiment, the communication network is a wired network or a wireless network.

[0058] FIG1 is only an example. In fact, the number of terminal devices and servers is not limited and is not specifically limited in the embodiments of the present application.

[0059] In the embodiments of the present application, when there are multiple servers, the multiple servers can be combined into a blockchain, and the servers are nodes on the blockchain. For example, in the method for processing elevation information disclosed in the embodiments of the present application, the relevant processed data involved can be stored on the blockchain, such as sampling points, point cloud recognition results, candidate elevation values, confidence levels, feature evaluation values, effective elevation values, target elevation values, etc.

[0060] In addition, the embodiments of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving and other scenarios.

[0061] In various embodiments of this application, user information (such as location information when the user uses a map) and other related data may be involved. When these embodiments are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0062] In addition, in various embodiments of the present application, the collection, use and processing of relevant data such as elevation information processing (such as the collection of point clouds within the target area) also comply with relevant laws, regulations and standards of relevant countries and regions.

[0063] The following describes a method for processing elevation information provided by an exemplary embodiment of the present application in combination with the application scenarios described above and with reference to the accompanying drawings.

[0064] 3 is a flowchart of an implementation method for processing elevation information provided by an embodiment of the present application. Taking the server as the execution subject as an example, the specific implementation process of the method is as follows: steps S31 to S35.

[0065] S31: Sampling lane data of roads within the target area to obtain multiple sampling points.

[0066] The target area can be any location in the real world that contains a road. In the embodiment of the present application, a road can also refer to a road section.

[0067] In an embodiment of the present application, the road lane data (such as lane lines) in the target area can be equidistantly resampled to obtain multiple sampling points. For example, when performing equidistant resampling, the distance between each two adjacent sampling points on the lane line is 2 meters, and a sampling point matrix S[m,n,2] can be obtained. Among them, m represents the number of samples, n represents the number of lanes, and 2 represents the longitude and latitude coordinates, that is, (x,y). Lane data is an element in the map and belongs to a type of map data. Lane data can refer to lane line data, for example. A road may have multiple lane lines, corresponding to multiple lanes.

[0068] In addition to the equidistant resampling method listed above, other methods can also be used to sample lane data on roads within the target area, such as sampling according to random distances, etc., which are not specifically limited in this article.

[0069] S32: Determine, from a plurality of point cloud recognition results obtained by recognizing a plurality of point clouds in the target area, the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points, each elevation value corresponding to a point cloud recognition result, and use the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point to form a candidate elevation value set for the sampling point.

[0070] The sampling points in the lane data are projected onto the lane segments in the point cloud recognition results, and the points obtained are considered to be the points corresponding to the sampling points in the point cloud recognition results, representing the sampling points. In an embodiment of the present application, point cloud collection can be performed on the target area at different times, and multiple point clouds of the collected target area can be identified. For example, multiple point cloud recognitions are performed to obtain point cloud recognition results of point clouds collected at different times. For example, the first point cloud collection is performed on the target area in Y1, and the point cloud recognition result 1 of the target area is obtained by identifying the point cloud obtained by the first point cloud collection. The second point cloud collection is performed on the target area in Y2, and the point cloud recognition result 2 of the target area is obtained by identifying the point cloud obtained by the second point cloud collection. The third point cloud collection is performed on the target area in Y3, and the point cloud recognition result 3 of the target area is obtained by identifying the point cloud obtained by the third point cloud collection.

[0071] The point cloud recognition result in the embodiment of the present application may include the latitude and longitude information and elevation value of the point. The elevation value obtained by point cloud recognition can be recorded as candidate elevation information, also known as candidate elevation value.

[0072] As shown in Figure 4, it is a schematic diagram of the lane line recognition result in a target area in an embodiment of the present application. Figure 4 shows that after performing point cloud recognition on the target area, the lane line in the target area can be identified.

[0073] In addition, point cloud recognition can not only identify lane lines, but also further identify other types of points in the target area. For example, in the embodiment of the present application, the target area includes at least a road, and there is at least one lane and corresponding lane lines on the road. In addition, the target area may also include guardrails, curbstones, ground arrows, text, scenes, etc. Accordingly, the sampling points in the target area include but are not limited to some or all of the following: lane line points, guardrail points, curbstone points, ground arrow points, text points, scene points, etc.

[0074] According to an embodiment of the present application, for each sampling point, a candidate elevation value corresponding to the sampling point can be obtained from each point cloud recognition result. For the same sampling point, multiple candidate elevation values ​​can be obtained from multiple point cloud recognition results, and these candidate elevation values ​​can constitute the candidate elevation value set for the sampling point.

[0075] For example, for sampling point P1, candidate elevation value z11 for sampling point P1 can be extracted based on point cloud recognition result 1, candidate elevation value z12 for sampling point P1 can be extracted based on point cloud recognition result 2, and candidate elevation value z13 for sampling point P1 can be extracted based on point cloud recognition result 3. The candidate elevation value set corresponding to sampling point P1 is composed of z11, z12, and z13.

[0076] For example, for sampling point P2, we can extract candidate elevation value z21 based on point cloud recognition result 1, extract candidate elevation value z22 based on point cloud recognition result 2, and extract candidate elevation value z23 based on point cloud recognition result 3. The candidate elevation value set corresponding to sampling point P2 is composed of z21, z22, and z23.

[0077] Similarly, the candidate elevation value set of each sampling point in the sampling point matrix can be obtained.

[0078] S33 , performing fitting on the multiple sampling points to obtain fitting results, and determining one or more target sampling points among the multiple sampling points based on the fitting results.

[0079] Based on the above point cloud recognition results, a plane fitting can be performed on the sampling points in the sampling point matrix to obtain a fitting plane. The plane fitting can be performed on the sampling points in the sampling point matrix based on the elevation values ​​of the sampling points.

[0080] In the embodiment of the present application, there are many ways to fit the sampling points, such as a random sample consensus (RANSAC) fitting algorithm, a linear fitting algorithm, and the like.

[0081] The following example uses the RANSAC algorithm in machine learning to perform plane fitting on the elevation values ​​of point cloud recognition results. The RANSAC algorithm is an iterative method for estimating mathematical model parameters using observed data points, in which the mathematical model can be estimated by repeatedly selecting a set of data points, and iterating until a mathematical model that is considered to be relatively good is estimated. Data points include inliers and outliers. Outliers have no value for model estimation, so this method can also be called an outlier detection method. Inliers refer to the correct sampling points, inlier points, or correct samples in the sample. Outliers refer to noise points in the sample (such as points with large measurement errors, points with inaccurate data values, etc.), which are often called outliers or noise points.

[0082] In general, if a model is directly used to fit a set of observation data containing many noise points, the model will definitely be affected by the noise points, and the larger the proportion of noise points, the less accurate the fitted model will be. RANSAC can be used to solve the outlier problem (noise) in the sample and can handle up to 50% of the outlier cases.

[0083] Refer to Figure 5, which is a schematic diagram comparing the effects of a RANSAC fitting algorithm and a linear fitting algorithm in an embodiment of the present application. Result 2 represents the fitting result obtained using the RANSAC fitting algorithm, and Result 1 represents the fitting result obtained using the linear fitting algorithm. As can be seen from Figure 5, when performing line fitting using the RANSAC fitting algorithm, outliers can essentially be excluded and fitting can be performed based on inliers; however, when performing line fitting using the linear fitting algorithm, the above effect cannot be achieved.

[0084] Similarly, when performing surface fitting (plane fitting) based on the RANSAC fitting algorithm, noise points can also be effectively excluded.

[0085] Based on the above ideas, in the embodiment of the present application, a plane fitting is performed on the multiple sampling points included in the sampling point matrix based on the above-obtained point cloud recognition results using the RANSAC fitting algorithm to obtain a fitting plane. During this fitting process, noise points (i.e., outliers) among the multiple sampling points can be effectively eliminated, and the remaining sampling points are the target sampling points in the embodiment of the present application.

[0086] Refer to Figure 6, which is a schematic diagram of a plane fitting process in accordance with an embodiment of the present application. As shown in Figure 6, sampling points can first be collected using the equidistant resampling method listed above to perform spatial analysis of the target area. Based on this, a plane fitting can be performed on the sampling points based on the RANSAC algorithm and the point cloud recognition results to obtain a fitted plane. The gray plane in Figure 6 is an example of a fitted plane in accordance with an embodiment of the present application.

[0087] In the embodiment of the present application, when obtaining the target sampling points by plane fitting, an exemplary implementation method is:

[0088] According to multiple elevation values ​​of each sampling point among the multiple sampling points, plane fitting is performed on the multiple sampling points to obtain a fitting plane corresponding to the target area.

[0089] For example, a plane fitting may be performed on multiple sampling points based on a RANSAC algorithm to obtain a fitting plane.

[0090] In the embodiment of the present application, after obtaining the fitted plane using the RANSAC algorithm, the fitting quality of the fitted plane still needs to be determined. If the plane is tilted due to insufficient sampling points (large elevation deviations between sampling points cause the fitted plane to be too tilted, and the angle with the standard normal vector is less than a certain angle threshold), it is necessary to degenerate from surface fitting to line fitting.

[0091] An exemplary method for analyzing whether lateral deviation occurs is:

[0092] The angle between the fitted plane and the standard normal vector is compared with a first preset angle threshold, where the standard normal vector is a normal vector perpendicular to the horizontal ground.

[0093] If the angle between the fitting plane and the standard normal vector is not less than a first preset angle threshold, the fitting plane is considered to be free of lateral deviation and the fitting quality meets the standard. Therefore, the sampling point on the fitting plane is used as the target sampling point. Sampling points outside the fitting plane can be considered as noise points to be excluded.

[0094] On the contrary, if the angle between the fitting plane and the standard normal vector is less than the first preset angle threshold, it is considered that the fitting plane has a plane deviation, and the fitting quality of the fitting plane does not meet the standard. It is necessary to degenerate from surface fitting to line fitting, and then determine the target sampling point based on the line fitting result.

[0095] Based on the above implementation, noise points can be effectively removed from the sampling, and subsequent processing can be performed based on the remaining target sampling points, effectively improving the elevation accuracy of the determined sampling points. Furthermore, the above implementation also includes a process for determining the fitting quality of the fitted plane, effectively avoiding the impact of insufficient samples and low accuracy.

[0096] As shown in Figure 7, it is a schematic diagram of a plane and a standard normal vector in an embodiment of the present application. Assume that plane A is the fitting plane and vector B is the standard normal vector. When analyzing whether plane A has deflected, the angle α between plane A and vector B can be analyzed. This angle α can also be understood as the angle between the projection of vector B on plane A and vector B. Vector B shown in Figure 7 is parallel to the z-axis, which means that the standard normal vector in the embodiment of the present application is perpendicular to the horizontal ground.

[0097] Obviously, when plane A is horizontal, vector B is perpendicular to plane A, and the angle α between vector B and plane A is 90°. When plane A is not horizontal, the angle α between vector B and plane A will be less than 90°.

[0098] In the embodiment of the present application, the first preset angle threshold can be flexibly set based on experience or needs. The value of the first preset angle threshold is a positive number and is less than 90°. For example, if the first preset angle threshold is set to 10°, it means that when the angle between vector B and plane A is less than 10°, the fitting quality of plane A is poor and the plane deviates laterally.

[0099] When the fitting plane deviates laterally, the above-mentioned surface fitting needs to be degenerated into line fitting, and then the line fitting result needs to be further analyzed to see whether it deviates laterally.

[0100] An example implementation is as follows:

[0101] If the angle between the fitting plane and the standard normal vector is less than a first preset angle threshold (indicating that the plane has deviated laterally), the two-dimensional latitude and longitude information (e.g., latitude and longitude coordinates) of each sampling point is converted into a linear reference distance. Furthermore, a line fitting is performed on the multiple sampling points based on their respective linear reference distances to obtain a fitted line result for the target area. The fitted line result includes at least one fitted line.

[0102] The linear reference distance represents the distance between a sampling point and the starting point of the lane line to which the sampling point belongs. A lane line is a section of a lane line within the target area, and therefore, the starting point of the lane line is also within the target area. For example, a sampling point is the i-th sampling point on a lane line on a road within the target area. If the sampling points to be fitted are sampled at a sampling distance of two meters, the linear reference distance corresponding to the i-th sampling point is 2i. When performing line fitting, the abscissa of the i-th sampling point is 2i, and the ordinate remains the elevation value.

[0103] In the embodiment of the present application, when performing line fitting on the sampling points, the RANSAC algorithm listed above can also be used to effectively eliminate noise points.

[0104] Similar to the surface fitting described above, after obtaining the fitted line result through the RANSAC algorithm, the fitting quality of the result still needs to be judged. If the fitted line deviates due to insufficient sampling points (for example, the elevation deviation between sampling points is large, resulting in a large inclination angle of the fitted curve, and the angle with the standard normal vector is less than a certain angle threshold), linear interpolation is required to smooth the interpolation.

[0105] For example, after obtaining the fitted straight line results, a method similar to the above-mentioned analysis of whether the fitted plane has deflected is used to further analyze the fitted straight lines that have deflected and the fitted straight lines that have not deflected in the fitted straight line results.

[0106] For example, if the angle between a certain fitted straight line and the standard normal vector is smaller than the second preset angle threshold, it indicates that the fitted straight line is deviated; otherwise, it indicates that the fitted straight line is not deviated.

[0107] Similar to the above-mentioned method for determining plane deviation, when line C is horizontal, vector B is perpendicular to line C, and the angle between vector B and line C is 90°. However, when line C is not horizontal, the angle between vector B and line C may be less than 90°. For example, if the second preset angle threshold is 10°, a deviation of the line fit indicates poor quality.

[0108] In this embodiment of the present application, the unbiased fitted line can be recorded as the target fitted line. That is, the target fitted line is the fitted line whose angle with the standard normal vector is not less than the second preset angle threshold. Furthermore, the sampling points on the target fitted line in the fitted line result are used as target sampling points.

[0109] Generally, even if a fitted line deviates laterally, the deviation won't persist. There will typically be a target fitted line before and after the line (for example, along the longitudinal direction of the road) that hasn't deviated. Therefore, linear interpolation can be used to flatten such deviated lines. This is explained in detail below and will not be repeated here.

[0110] Based on the above implementation, in the event of plane deviation, by degenerating surface fitting to line fitting, in the line fitting scenario, problems caused by insufficient samples can be effectively solved, thereby improving the accuracy of elevation processing.

[0111] S34: For each candidate elevation value set of the one or more target sampling points, the following operations are performed: determining a confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining a valid elevation value of the target sampling point based on the confidence level of each candidate elevation value.

[0112] Based on the above S32, the sampling points in the sampling point matrix can be screened to effectively filter out noise points and obtain target sampling points. Furthermore, for each target sampling point, each target sampling point corresponds to a set of candidate elevation values. Each candidate elevation value set contains at least one candidate elevation value, and each candidate elevation value is associated with a confidence level. This confidence level can be determined based on the point cloud recognition results and reference map data to which the candidate elevation value belongs.

[0113] In the embodiment of the present application, for the same target sampling point, these candidate elevation values ​​can be weightedly fused based on the confidence of each candidate elevation value in the candidate elevation value set corresponding to the target sampling point to obtain a valid elevation value corresponding to the target sampling point.

[0114] In all embodiments of the present application, the confidence level of each candidate elevation value in the candidate elevation value set corresponding to the target sampling point can be determined by the method shown in Figure 8. Referring to Figure 8, it is a flow chart of a method for determining the confidence level of a candidate elevation value in an embodiment of the present application, including the following steps S81 to S84.

[0115] S81: Determine a difference weight for the candidate elevation value based on a difference between a point cloud lane segment formed by a target sampling point and a next adjacent target sampling point in the point cloud recognition result associated with the candidate elevation value and a map lane segment formed by corresponding points in the reference map data. The candidate elevation value is a candidate elevation value for the target sampling point.

[0116] The point cloud recognition result associated with a candidate elevation value specifically refers to the point cloud recognition result from which the candidate elevation value was extracted. For example, in the example above, taking sampling point P1 as the target sampling point, if the candidate elevation value z11 of that sampling point is extracted from point cloud recognition result 1, then z11 is associated with point cloud recognition result 1. Similarly, the candidate elevation value z12 of that sampling point is associated with point cloud recognition result 2, and the candidate elevation value z13 of that sampling point is associated with point cloud recognition result 3.

[0117] In this embodiment of the present application, different candidate elevation values ​​for the same target sampling point are associated with different point cloud recognition results. However, for different target sampling points, their candidate elevation values ​​can be associated with the same point cloud recognition result. For example, the candidate elevation value z11 for target sampling point P1 and the candidate elevation value z21 for target sampling point P2 are both associated with point cloud recognition result 1.

[0118] Furthermore, in the embodiments of the present application, for a target sampling point, its next adjacent target sampling point may refer to a point along a specified direction (e.g., the direction of the lane line, which can be set during sampling). For example, for target sampling point P1, its next adjacent target sampling point refers to the next target sampling point located after P1 along the lane line where target sampling point P1 is located.

[0119] For example, there are 10 sampling points on a lane line, denoted as P1 to P10 along a specified direction. After removing noise points P3 and P5, the remaining 8 sampling points are the target sampling points. For P1, its next adjacent target sampling point is P2. For P2, its next adjacent target sampling point is P4. For P4, its next adjacent target sampling point is P6. For P6, its next adjacent target sampling point is P7. And so on.

[0120] The difference weight includes at least one of a distance weight and an angle weight.

[0121] Taking the example of a difference weight including both distance weight and angle weight, an exemplary implementation of S81 includes the following steps S811 to S812:

[0122] S811: Determine a distance weight for the candidate elevation value based on a lateral distance between the point cloud lane line segment and the map lane line segment. The lateral distance can be understood as the distance between a sampling point on the map lane line segment and the point cloud lane line segment.

[0123] The horizontal direction refers to the direction perpendicular to the lane line's extension, and the vertical direction refers to the lane line's extension direction. Therefore, the horizontal distance represents the distance between the point cloud lane line and the map lane line in the direction perpendicular to the lane line's extension direction.

[0124] In the embodiment of the present application, an exemplary implementation of S811 is as follows:

[0125] The distance weight of the candidate elevation value is determined according to the ratio of the square of the lateral distance to the square of the third preset model parameter.

[0126] In the embodiments of the present application, the third preset model parameter is a preset distance parameter. This distance parameter can be flexibly set based on experience, actual needs, etc., and is generally a positive number. For example, the third preset model parameter value is 20, in centimeters, which represents 20 cm. Alternatively, the third preset model parameter can also be any other distance value, which will not be detailed here.

[0127] For example, one calculation formula for distance weight is as follows: dis =exp(-d ** 2 / 2 / σ1 ** 2) (Formula 1);

[0128] Among them, w dis represents the distance weight, d is the lateral distance, σ1 is the third preset model parameter, d**2 represents the square of d, that is, the square of the lateral distance, and σ1**2 represents the square of σ1, that is, the square of the third preset model parameter.

[0129] The distance weight calculation method listed in the above formula 1 is just a simple example. In addition, any method of calculating the distance weight based on the ratio of the square of the horizontal distance to the square of the third preset model parameter is applicable to the embodiment of the present application and will not be repeated here.

[0130] In the above embodiment, combined with the distance difference between the segment of point cloud lane line and the segment of map lane line, the lateral difference between the segment of point cloud lane line and the segment of map lane line can be effectively analyzed, and the distance weight calculated based on this can be used as one of the confidence calculation parameters to evaluate the accuracy of the candidate elevation value.

[0131] S812: Determine an angle weight of the candidate elevation value based on a longitudinal angle between the segment of the point cloud lane line and the segment of the map lane line.

[0132] The longitudinal angle represents the angle between the segment of point cloud lane line and the segment of map lane line in the direction in which the lane line extends.

[0133] In the embodiment of the present application, an exemplary implementation of S812 is as follows:

[0134] The angle weight of the candidate elevation value is determined according to the ratio of the square of the longitudinal angle to the square of the fourth preset model parameter.

[0135] In the embodiment of the present application, the fourth preset model parameter is a preset angle parameter. The angle parameter can be flexibly set based on experience, actual needs, etc., and is generally a positive number. For example, the fourth preset model parameter has a value of 15, in degrees (°), which represents 15°. Alternatively, the fourth preset model parameter can also be any other angle, which will not be detailed here.

[0136] For example, one calculation formula for angle weight is as follows: angle =exp(-α ** 2 / 2 / σ2 ** 2) (Formula 2);

[0137] Among them, w angle represents the angle weight, α is the longitudinal angle, σ2 is the fourth preset model parameter, α**2 represents the square of α, that is, the square of the longitudinal angle, and σ2**2 represents the square of σ2, that is, the square of the fourth preset model parameter.

[0138] The calculation method of the angle weight listed in the above formula 2 is just a simple example. In addition, any method of calculating the angle weight based on the ratio of the square of the longitudinal angle to the square of the fourth preset model parameter is applicable to the embodiment of the present application and will not be repeated here.

[0139] In the above embodiment, combined with the angular difference between the segment of point cloud lane line and the segment of map lane line, the longitudinal difference between the segment of point cloud lane line and the segment of map lane line can be effectively analyzed, and the angle weight calculated based on this can be used as one of the confidence calculation parameters to evaluate the accuracy of the candidate elevation value.

[0140] S82: Determine an acquisition time attenuation coefficient of the candidate elevation value according to the acquisition time of the point cloud corresponding to the point cloud recognition result.

[0141] In the embodiment of the present application, the acquisition time of the point cloud recognition result can be the time described in S32, that is, the time when the point cloud recognition (including point cloud acquisition and point cloud recognition) of the target area is performed to obtain the point cloud recognition result, or it can refer to the point cloud acquisition time. Different point cloud recognition results correspond to different acquisition times. Assuming that the time unit of the acquisition time is year, the acquisition time of the point cloud recognition result 1 listed above is Y1 year, the acquisition time of the point cloud recognition result 2 is Y2 year, and the acquisition time of the point cloud recognition result 3 is Y3 year.

[0142] The time unit can be further refined, such as specific to month, day, hour, minute, second, etc. The above-mentioned year as the unit is just a simple example. Other time units are also applicable to the embodiments of this application and will not be repeated here.

[0143] When determining the acquisition time attenuation coefficient corresponding to a candidate elevation value based on the acquisition time of each point cloud recognition result, the attenuation coefficient can be determined based on the difference between the acquisition time of the point cloud recognition result associated with the candidate elevation value and the acquisition time of each point cloud recognition result. An example implementation is as follows:

[0144] An acquisition time attenuation coefficient is determined based on the difference between the acquisition time of the point cloud recognition result associated with the candidate elevation value and the maximum acquisition time, wherein the maximum acquisition time is the maximum value among the acquisition times of the point cloud recognition results.

[0145] For example, taking the three point cloud recognition results mentioned above as an example, the corresponding acquisition times are Y1, Y2, and Y3 respectively. Assuming that the maximum value of these three acquisition times is Y3, it means that the maximum acquisition time is Y3.

[0146] Assuming that the acquisition time of the point cloud recognition result 1 associated with a candidate elevation value z11 is Y1, the acquisition time attenuation coefficient corresponding to z11 is determined according to the difference between Y1 and Y3.

[0147] For example, the acquisition time attenuation coefficient of the candidate elevation value can be determined based on the difference between the acquisition time of the point cloud recognition result associated with the candidate elevation value and the maximum acquisition time, multiplied by the fifth preset model parameter. In the embodiment of the present application, the fifth preset model parameter is a preset coefficient parameter that can be flexibly set based on experience, actual needs, etc. For example, the fifth preset model parameter takes a value of 1e-4. Alternatively, the fifth preset model parameter can also be any other coefficient, which will not be detailed here.

[0148] For example, one calculation formula for the acquisition time attenuation coefficient is as follows: τ1 = exp(β*(t-max(t))) (Formula 3);

[0149] Among them, τ1 is the acquisition time attenuation coefficient, t is the acquisition time of the point cloud recognition result associated with a candidate elevation value, max(t) is the maximum acquisition time, and β is the fifth preset model parameter.

[0150] The calculation method of the acquisition time attenuation coefficient listed in the above formula 3 is only a simple example. In addition, any method of calculating the acquisition time attenuation coefficient based on the difference between the acquisition time of the point cloud recognition result associated with a candidate elevation value and the maximum acquisition time is applicable to the embodiments of the present application and will not be detailed here.

[0151] In the above embodiment, the acquisition time attenuation coefficient is analyzed based on the acquisition time of the point cloud, and the confidence is calculated based on this coefficient, which is equivalent to considering the impact of the acquisition time of the point cloud recognition result, and can further improve the evaluation accuracy of the candidate elevation value.

[0152] S83: Determine a preset category reduction coefficient corresponding to the identified category according to the category of the target sampling point identified in the point cloud recognition result associated with the candidate elevation value.

[0153] In the embodiment of the present application, different sampling points may be of different categories, such as lane lines, guardrails, curbstones, ground arrows, text, scenes, etc. listed above.

[0154] Each category may correspond to a preset coefficient, i.e., a preset category weight reduction coefficient. Taking the coefficient as an example, where the value is 0 to 1, different coefficients may be set for different categories, or the same coefficient may be set for some categories. This article does not make any specific restrictions.

[0155] For example, the coefficient corresponding to the guardrail category is 0.5, and the coefficient corresponding to other categories is 1. Of course, other preset coefficients can also be used, which will not be detailed here.

[0156] In this embodiment of the present application, the target sampling point can be identified through point cloud recognition to determine the category of the point identified during the current recognition process. Each category corresponds to a preset category downweighting coefficient, such as τ2. If the target sampling point P1 is determined to belong to the guardrail category based on point cloud recognition result 1 associated with the target sampling point P1, then τ2 = 0.5. If P1 is not a guardrail, then τ2 = 1.

[0157] In the above embodiment, the confidence level is calculated based on the category of point cloud recognition, which is equivalent to considering the impact of the classification accuracy of the point cloud recognition result, and can further improve the evaluation accuracy of the candidate elevation values.

[0158] S84: Determine the confidence level corresponding to the candidate elevation value according to the difference weight, the acquisition time attenuation coefficient, and the preset category weight reduction coefficient.

[0159] In the embodiment of the present application, based on the above S81 to S83, a difference weight (including distance weight w) of a candidate elevation value can be determined according to each point cloud recognition result and reference map data. dis and angle weight w angle ), the acquisition time attenuation coefficient τ1 and the preset category reduction coefficient τ2, and then, based on these parameters, the confidence level corresponding to the candidate elevation value can be determined.

[0160] In the embodiment of the present application, an exemplary implementation of S84 includes the following steps S841 to S842:

[0161] S841: Determine the sum of the product of the distance weight and the first preset model parameter and the product of the angle weight and the second preset model parameter, and record them as target parameters.

[0162] In an embodiment of the present application, the first preset model parameter and the second preset model parameter are a preset weight parameter, which can be flexibly set based on experience, actual needs, etc., and is generally a positive number with a value of 0 to 1, and the sum of the first preset model parameter and the second preset model parameter is 1.

[0163] For example, the first preset model parameter is denoted as c1, and the second preset model parameter is denoted as c2, where c1 + c2 = 1. The values ​​of c1 and c2 can be flexibly set according to needs. For example, if distance weight is given more importance, the first preset model parameter is set to be larger than the second preset model parameter, such as c1 = 0.7 and c2 = 0.3. For another example, if angle weight is given more importance, the second preset model parameter is set to be larger than the first preset model parameter, such as c1 = 0.4 and c2 = 0.6. For another example, the first preset model parameter and the second preset model parameter can be set to be the same, that is, c1 = c2 = 0.5, and so on.

[0164] The values ​​of the above-mentioned first preset model parameter and second preset model parameter are just simple examples. Any preset method is applicable to the embodiments of the present application and will not be described in detail here.

[0165] In the embodiment of the present application, the target parameter is: c1*w dis +c2*w angle .

[0166] S842: The product of the acquisition time attenuation coefficient, the preset category weight reduction coefficient, and the target parameter is used as the confidence of the candidate elevation value.

[0167] For example, one calculation formula for confidence is as follows: p =τ1*τ2*(c1*w dis +c2*w angle ) (Formula 4);

[0168] Among them, w p That is, it represents the confidence level, τ1 is the acquisition time attenuation coefficient, τ2 is the preset category weight reduction coefficient, (c1*w dis +c2*w angle ) is the target parameter.

[0169] The confidence calculation method listed in the above formula 4 is just a simple example. In addition, other methods of calculating confidence using the acquisition time attenuation coefficient, the preset category reduction coefficient and the difference weight are also applicable to the embodiments of the present application and will not be described in detail here.

[0170] In the above embodiment, the confidence of the candidate elevation value is comprehensively evaluated from multiple different dimensions by integrating the distance weight, angle weight, acquisition time screening coefficient and preset category deweighting coefficient, which can ensure that the confidence of the calculated candidate elevation value is more credible.

[0171] When performing plane fitting on multiple sampling points, if no plane deviation occurs, the effective elevation value of the target sampling point can be calculated based on the following method, and then the effective elevation value can be optimized through the graph propagation algorithm.

[0172] The following first explains how to calculate the effective elevation value:

[0173] In the embodiment of the present application, for each target sampling point, after obtaining the candidate elevation values ​​of the target sampling point and the confidence level of each candidate elevation value, the effective elevation value of the target sampling point can be determined by weighted fusion.

[0174] Based on this, an exemplary implementation of step S34 includes the following steps S331 to S332 (not shown in FIG3 ):

[0175] S331: Based on the reference map data and the recognition results of each point cloud, respectively determine the confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point.

[0176] The method of calculating the confidence of each candidate elevation value based on the reference map data and the point cloud recognition results is shown in Figure 7 above. For details, please refer to the above embodiment, and the repeated parts will not be repeated.

[0177] S332: Based on the confidence level of each candidate elevation value, perform weighted fusion on each candidate elevation value to obtain a valid elevation value of the target sampling point.

[0178] In the embodiment of the present application, each target sampling point may correspond to multiple candidate elevation values. Through weighted fusion, the multiple candidate elevation values ​​corresponding to each target sampling point can be fused to obtain a valid elevation value, i.e., the valid elevation value in the embodiment of the present application. The valid elevation value is a unique value.

[0179] For example, by weighted fusion of the above confidences, an effective elevation value z can be obtained. A calculation formula for an effective elevation value is as follows: Z = ∑(W pi *z i ) / ∑W pi(Formula 5);

[0180] Among them, for a target sampling point, z i Indicates the i-th candidate elevation value of the target sampling point, W pi z i confidence level.

[0181] In S322, an exemplary implementation may be: taking the ratio of the sum of the products of the candidate elevation values ​​of the target sampling point and the corresponding confidence levels to the sum of the confidence levels corresponding to the candidate elevation values ​​as the effective elevation value of the target sampling point.

[0182] In the above implementation, the candidate elevation values ​​are weighted and fused together using confidence levels to determine the effective elevation value for the target sampling point. This approach processes candidate elevation values ​​collected at different times, determining the elevation value based on point cloud recognition results. It is independent of GPS and unaffected by factors such as GPS positioning and road network accuracy, helping to improve the accuracy of elevation information processing.

[0183] In the embodiment of the present application, after obtaining the effective elevation value of the target sampling point, the effective elevation value of the target sampling point can be further optimized using a graph propagation algorithm. This part will be explained in detail in the following step S35 and will not be repeated here.

[0184] In addition, if there is local point cloud data missing in the lane (that is, there is no point cloud recognition result in the local area), but there are point cloud recognition results before and after it, linear interpolation can be performed on the local missing area to compensate for the local point cloud data missing and smooth the result.

[0185] An example implementation is as follows:

[0186] If the number of target sampling points on any lane in the target area is less than a preset number threshold, a first target interpolation point on that lane is determined. The first target interpolation point can be one or more. Then, for each first target interpolation point, the following operations are performed:

[0187] Based on the effective elevation values ​​of the target sampling points on both sides of the first target interpolation point along the longitudinal direction of the lane, linear interpolation is performed on the first target interpolation point to obtain a first interpolation elevation value as the effective elevation value of the first target interpolation point.

[0188] The preset quantity threshold can be flexibly set according to actual needs, and can also refer to the settings such as the sampling distance during the above sampling. This article does not make specific restrictions.

[0189] If the number of actual target sampling points on one or more lanes in the target area is less than the preset threshold due to the lack of point cloud data, the first target interpolation point can be selected on the lane based on the sampling distance set above, that is, the point where the effective elevation value is to be calculated through linear interpolation.

[0190] Furthermore, for each first target interpolation point, there should be target sampling points with valid elevation values ​​on both sides of the first target interpolation point (i.e., target sampling points before and after the first target interpolation point along the longitudinal direction of the road where the lane is located). Therefore, the effective elevation value of the first target interpolation point can be calculated based on the effective elevation values ​​of the target sampling points on both sides of the first target interpolation point to achieve linear interpolation.

[0191] In the embodiment of the present application, an exemplary calculation formula for the effective elevation value of the first target interpolation point is as follows: z=(z j (mi)+z i (jm)) / (ji) (Formula 6);

[0192] Where m is the index number of the point to be interpolated, i (left side of m) and j (right side of m) are the index numbers of the target sampling points on both sides of the point to be interpolated. In the embodiment of the present application, the index number represents the sampling point number of the point on the road.

[0193] That is, the effective elevation value of the first target interpolation point is determined by combining the effective elevation value of the target sampling point and the index numbers of the to-be-interpolated point and the target sampling point.

[0194] Refer to Figure 9, which is a schematic diagram of an interpolation and filling process in an embodiment of the present application. After obtaining the fitting plane, interpolation and filling can also be performed based on the linear interpolation method listed above. The interpolation and filling result is shown in Figure 9.

[0195] Based on the above, the effective elevation values ​​of each first target interpolation point can be optimized based on the topological relationship between roads within the target area to obtain the target elevation value of each first target interpolation point. The specific optimization process can be referred to the relevant part of Figure 10 below, and the repeated parts will not be repeated here.

[0196] For example, when the surface fitting degenerates to the line fitting, the non-target fitting line that deviates sideways can also be flattened by linear interpolation. An example implementation is as follows:

[0197] First, for a non-target fitting line in the fitting line result, a second target interpolation point on the non-target fitting line is determined.

[0198] Specifically, according to the sampling distance set above, etc., a second target interpolation point, ie, a point whose effective elevation value is to be calculated through linear interpolation, may be selected on the non-target fitting line.

[0199] Furthermore, for each second target interpolation point, the following operations are performed: based on the effective elevation values ​​of the target sampling points on both sides of the second target interpolation point, the second target interpolation point is linearly interpolated to obtain a second interpolation elevation value as the effective elevation value of the second target interpolation point.

[0200] The target sampling points on both sides of the second target interpolation point specifically refer to the target sampling points on the target fitting curve on both sides of the non-target fitting line. In other words, linear interpolation can be performed on the part with side deviation based on the valid fitting results without side deviation.

[0201] In the embodiment of the present application, the method for calculating the effective elevation value of the second target interpolation point can be the same as the method for calculating the effective elevation value of the first target interpolation point listed above. For reference, the above formula 6 can also be used, and the repeated parts will not be repeated here.

[0202] On the basis of the above, the effective elevation value of each second target interpolation point can also be optimized based on the topological relationship between the roads in the target area to obtain the target elevation value of each second target interpolation point. The specific optimization process can be referred to the relevant part of Figure 10 below, and the repeated parts will not be repeated.

[0203] In the above embodiment, for the fitted straight line that is deviated laterally, the impact can be reduced based on linear interpolation or other methods to improve the accuracy of elevation information processing.

[0204] S35: Based on the topological relationship between the roads in the target area, the effective elevation value of each target sampling point is optimized to obtain the target elevation value of each target sampling point.

[0205] In the embodiment of the present application, the topological relationship between roads refers to the connectivity relationship between roads.

[0206] In the embodiment of the present application, the target sampling points, first target interpolation points or second target interpolation points listed above can be collectively referred to as target points. For any target point. As mentioned above, each target point obtains a valid elevation value, but it does not mean that the elevation value at this time is correct. This is because in a multi-layer road scene, the upper road has better road paving conditions and lighting, so there are more and better point cloud recognition results, while the lower layer will have a large number of missed recognitions. When resampling according to the above method, the sample bias causes RANSAC to tend to fit to the layer with more sampling points, which may cause the obtained effective elevation value to be abnormally wrong. Therefore, it is necessary to distinguish between multiple layers and whether a specific road belongs to the upper layer or the lower layer. To solve this problem, in the embodiment of the present application, the candidate elevation values ​​are clustered, and then based on the clustering results, combined with the topological relationship between the roads in the target area, the effective elevation value of the target point is optimized. That is, the elevation value is optimized through the graph propagation algorithm to obtain the target elevation value of the target point.

[0207] Refer to Figure 10, which is a flowchart of an optimization process based on a graph propagation algorithm in an embodiment of the present application. A schematic diagram of an exemplary method for optimizing elevation values ​​based on a graph propagation algorithm includes the following steps S101 to S104:

[0208] S101: The roads in the target area are used as multiple nodes in the road network diagram. Each node corresponds to at least one elevation category and a feature evaluation value corresponding to each elevation category. The elevation category is obtained by clustering the target sampling points on the road corresponding to the node, and represents the elevation value of the road. The feature evaluation value is: the sum of the confidences of the candidate elevation values ​​of each target sampling point on the road belonging to the corresponding elevation category. In step S101, it can also include: clustering the target sampling points on the road corresponding to each node in the multiple nodes to obtain at least one elevation category for each node, the at least one elevation category representing the elevation value of the corresponding road; determining the sum of the confidences of the candidate elevation values ​​of the target sampling points on the corresponding road belonging to the corresponding elevation category in the at least one elevation category, and using the sum of the confidences as the feature evaluation value of the corresponding elevation category.

[0209] In this embodiment, a graph propagation algorithm is introduced to optimize the overall effect of distinguishing between upper and lower layers on roads. In this graph propagation algorithm, roads are used as nodes, and nodes have a feature score (a score representing the weight of the node) and a label. The label can be an elevation category in this article, and the feature score is the feature evaluation value corresponding to the elevation category.

[0210] As shown in Figure 11, it is a schematic diagram of a road network diagram in an embodiment of the present application. If the actual road network in the target area contains three roads, namely L1, L2 and L3 in the actual road network. Then the road network diagram generated based on the actual road network contains three nodes, such as nodes L1, L2 and L3 in the road network diagram in Figure 11. There are also edges between the nodes corresponding to the connected roads. As shown in Figure 11, nodes L1 and L2 are connected, and nodes L2 and L3 are connected, which means that roads L1 and L2 are connected, and roads L2 and L3 are connected.

[0211] The actual road network and road network diagram in Figure 11 are just simple examples. In reality, the road network in the real world may be more complex, and the corresponding road network diagram will also be more complex, so we will not go into details here.

[0212] In an embodiment of the present application, a graph propagation algorithm can be used to optimize the effective elevation value. The roads in the target area are used as nodes in the road network graph, that is, each road in the target area can be used as a node. Since a road can be a single-layer road or a road in a multi-layer road, each node corresponds to at least one elevation category. For example, a node belonging to a single-layer road corresponds to only one elevation category, and a node belonging to a multi-layer road may correspond to multiple elevation categories. Furthermore, each elevation category corresponding to a node corresponds to a feature evaluation value, which is the sum of the confidence levels of the candidate elevation values ​​of each target sampling point on the road corresponding to the node that belongs to the corresponding elevation category. The roads in the target area can be obtained from reference map data and / or determined based on point cloud recognition results.

[0213] For example, for a certain road (node), there are three elevation categories corresponding to the road: Class A (5 meters), Class B (10 meters), and Class C (15 meters). Class A corresponds to 5 target sampling points on the road, Class B corresponds to 10 target sampling points on the road, and Class C corresponds to 10 target sampling points on the road. Assuming that the sum of the confidence levels of the 5 target sampling points corresponding to Class A is 10, the feature evaluation value corresponding to Class A is 10, that is, the feature score is 10. Similarly, the feature evaluation value corresponding to Class B is 11, that is, the score is 11; the feature evaluation value corresponding to Class C is 9, that is, the score is 9.

[0214] In this case, the scores for categories A, B, and C are similar, making it difficult to determine the correct elevation value for the road. In this case, a graph propagation algorithm can be used for optimization.

[0215] Before optimization, the target sampling points on the roads in the target area are first clustered, and at least one elevation category corresponding to each road (each node) is determined through clustering. An example implementation is as follows:

[0216] First, based on the candidate elevation value sets for each target sampling point, the candidate elevation values ​​are clustered to obtain at least one cluster, wherein the difference between the elevation values ​​of cluster centers of different clusters is greater than a preset elevation threshold. For example, based on the candidate elevation value sets for each target sampling point, the candidate elevation values ​​of the one or more target sampling points are clustered to obtain at least one cluster. Within the at least one cluster, the difference between the elevation values ​​of cluster centers of different clusters is greater than a preset elevation threshold. The preset elevation threshold is a specified height threshold that can be flexibly set based on actual conditions or needs and is generally a positive number, such as 4.5 meters.

[0217] In the embodiment of the present application, there are many ways to cluster the candidate elevation values, such as K-means (KMEANS), fuzzy C-means clustering (FCM), etc., which are not specifically limited in this article.

[0218] Take the KMEANS clustering method as an example. The candidate elevation values ​​in the candidate elevation value set of each target sampling point can be clustered based on the KMEANS clustering method. For example, first, from the candidate elevation value set corresponding to a target sampling point, k candidate elevation values ​​are randomly selected as the k cluster centers of k clusters. Then, based on the distance between each candidate elevation value in the candidate elevation value set and these k cluster centers, it is divided into the cluster where the cluster center closest to it is located. Then, the cluster center of each cluster is updated, for example, the average value of all candidate elevation values ​​in each cluster is taken as the new cluster center of the cluster. Repeat the above process until a certain termination condition is met, for example, the change between the updated cluster center and the cluster center before the update is less than a threshold, or the number of iterations is reached. In an embodiment of the present application, 4 clusters can be determined first. The clusters can then be merged so that the number of clusters begins to decrease until at least one cluster is obtained, and the difference between the z values ​​(candidate elevation values) of the cluster centers of the at least one cluster is greater than a preset elevation threshold. The obtained at least one cluster is a classification of candidate elevation values ​​of the road, ie, an elevation category in this article.

[0219] Then, for each node, the following operations are performed: the elevation value of the cluster center corresponding to each target sampling point associated with the node is determined, and the determined elevation value is used as the elevation value represented by the elevation category corresponding to the node.

[0220] In this embodiment, roads are considered nodes. For a road, the number of clusters obtained after clustering the target sampling points on the road represents the number of elevation categories for the road, and therefore the number of layers to which the road may belong. Each elevation category represents a layer, and the elevation value represented by the elevation category represents the elevation value of that layer.

[0221] For example, a road listed above corresponds to three elevation categories: Class A, Class B, and Class C, which means that the road may be a three-layer road.

[0222] Please refer to FIG12 , which is a schematic diagram of a candidate set of elevation categories of a road and a corresponding candidate graph in an embodiment of the present application.

[0223] Among them, Figure 12 shows that the road L1 may be a two-layer road, corresponding to two elevation categories, and the elevation values ​​represented are Z 11 and Z 12 ; Road L2 may be a three-layer road, corresponding to three elevation categories, represented by elevation values ​​of Z 21 、Z 22 and Z 23 ; Road L3 may be a three-layer road, corresponding to three elevation categories, represented by elevation values ​​of Z 31 、Z 32 and Z 33 The candidate graph generated based on the above elevation categories is shown in FIG12 , which describes various elevation values ​​of the road and the corresponding possible connection relationships.

[0224] In the above implementation, clustering can be used to determine the number of layers corresponding to each road. Furthermore, the topological relationship between roads can be combined to verify the effective elevation value of the target point previously determined based on the point cloud recognition results. If there are any problems, further corrections can be made. The specific implementation is described in S102 to S103 below:

[0225] S102: Select a target node from the multiple nodes according to the feature evaluation value of each node.

[0226] In the embodiment of the present application, when selecting a target node, priority may be given to nodes corresponding to roads with small hierarchical ambiguity, usually single-layer roads.

[0227] Each node may have one or more scores (due to stratification), and each elevation category after KMEANS clustering corresponds to a score. When selecting a target node based on the score corresponding to each node, at least one of the following selection methods may be used:

[0228] Selection method 1: select a node with only one feature evaluation value among the multiple nodes as the target node.

[0229] That is, a node with only one score is selected, and the road corresponding to the node with only one score is a single-layer road.

[0230] Selection method 2: Among the multiple nodes, a node whose ratio of a corresponding feature evaluation value to other feature evaluation values ​​is greater than a preset ratio threshold is selected as the target node.

[0231] In the embodiment of the present application, the preset ratio threshold can be flexibly set according to actual conditions or needs. The value is generally a positive number greater than 1, such as 5, 10, 11, 15, etc.

[0232] For example, if a node corresponds to multiple scores, but among these scores, the ratio of the score of a certain elevation category to the scores of other elevation categories is greater than a preset ratio threshold, then the node can be used as the target node.

[0233] For example, a node whose score of a certain elevation category is more than ten times higher than that of other elevation categories is selected as the target node. The road layering corresponding to this type of node has less ambiguity, and the upper and lower layers can basically be determined.

[0234] In the embodiment of the present application, one target node or multiple target nodes may be selected. When selecting a target node, either the first selection method or the second selection method may be used; or the two selection methods may be combined for selection. When selecting a target node by combining the two selection methods, selection may be performed based on selection method one first. If the target node has not been screened out based on selection method one, or if insufficient target nodes have been screened out, selection may be performed based on selection method two, and so on.

[0235] S103: Starting from the target node, along the topological relationship between multiple nodes in the road network graph, the target elevation category of the target node is propagated to the surrounding areas of the target node, and the target elevation categories of other nodes in the road network graph are obtained according to the propagation results.

[0236] S104: Compare the target elevation category of the node associated with the target point with the effective elevation value of the target point, and obtain the target elevation value of the target point according to the comparison result.

[0237] The following example uses the method 1 of selecting a node corresponding to a single-layer road as the target node. After selecting the target node, starting from the target node, the elevation category of the target node is diffused to the surrounding area along the topological relationship between the nodes in the road network diagram. The road corresponding to the target node is labeled good, and propagated based on the connectivity relationship (topological relationship) between the roads. For a certain road, the elevation category corresponding to the road propagated from the good label can be used as a reference when the current road layer is difficult to distinguish, and is recorded as the target elevation category of the node corresponding to the current road. In this way, the advantage of the global perspective of the road topological relationship in the target area is utilized to improve the overall accuracy of the elevation.

[0238] For example, the node corresponding to the three-layer road listed above is L5, and the selected target node is L4 (single-layer). When propagating based on the connectivity relationship between roads, L4 and L5 are connected, so the elevation category of L4 can be propagated to L5. Based on this, from the three elevation categories corresponding to L5 (Class A - 5 meters, Class B - 10 meters, Class C - 15 meters), the elevation category with the smallest difference from L4 is selected as the target elevation category of L5. For example, if L4 is 10 meters, then L5 can be determined to be 10 meters.

[0239] Similarly, when selecting the target node based on the second selection method, the principle is similar and the repeated parts will not be repeated.

[0240] In an embodiment of the present application, when obtaining a target elevation value of a target point according to the comparison result, an exemplary implementation method is as follows:

[0241] If the comparison result is within the preset difference range, the effective elevation value of the target point is used as the target elevation value of the target point; if the comparison result is outside the preset difference range, the elevation value represented by the target elevation category of the node associated with the target point is used as the target elevation value of the target point.

[0242] For example, for a target sampling point, the calculated effective elevation value is 10 meters, and the elevation value represented by the target elevation category determined by graph propagation is also 10 meters. It can be determined that the target elevation value of the target sampling point is 10 meters.

[0243] For a target sampling point, the calculated effective elevation value is 10 meters, and the elevation value represented by the target elevation category determined by graph propagation is also 15 meters. It can be determined that the target elevation value of the target sampling point is 15 meters.

[0244] In the above implementation, the graph propagation stage can be understood as a post-processing stage, used to verify the accuracy of the previously calculated effective elevation values. If the previously calculated effective elevation values ​​are found to be abnormally erroneous, they can be corrected using the graph propagation algorithm, further improving the accuracy of the elevation information processing.

[0245] Figure 13 is a schematic diagram of a graph propagation optimization process in an embodiment of the present application. In this embodiment, the effective elevation values ​​of each target point obtained above can be verified using the graph propagation methods listed above. If any problems are found, they can be promptly corrected based on graph propagation to achieve smooth road surface and correct elevation.

[0246] This embodiment of the application selects various point cloud recognition results after equidistant resampling of road lane data, requiring high road network accuracy. In some scenarios where road network drawing accuracy is insufficient, the problem of insufficient automated processing accuracy still exists. To eliminate the reliance on road network accuracy and correctness, it is possible to consider automatically constructing a road surface model directly from the point cloud recognition results and matching them with the current road lane data, thereby improving the accuracy of elevation data when processing complex road surfaces.

[0247] The above describes the processing of road elevation values. In addition, the target area may also include intersections. In scenarios where fitting is impossible, such as intersections (where roads and lanes are not related) or insufficient samples, the hole fitting method can be used to automatically process the road lane data.

[0248] Referring to FIG. 14 , which is a flowchart of an implementation of a void fitting method in an embodiment of the present application, the method includes the following steps S141 to S144 :

[0249] S141: Aggregate the roads and lane lines in the intersection into parts to be fitted according to the topological relationship;

[0250] S142: Expand the sampling points of the part to be fitted according to the recognition results of each point cloud;

[0251] S143: fitting the intersection plane corresponding to the intersection according to the effective elevation values ​​of the intersection edge points in the portion to be fitted and the effective elevation values ​​of the expanded sampling points;

[0252] S144: Predicting an interior point in the intersection, and smoothing the intersection edge in the intersection plane according to the prediction result, where the interior point is at least one of a road point and a lane line point in the intersection.

[0253] For intersections, in general, a plane is fitted based on the edge points around the intersection, filling the empty hole and flattening it onto a plane. Then, the elevation values ​​of all elements within the intersection (whether roads or lanes) are copied from this plane.

[0254] As shown in Figure 14, the roads and lanes in the intersection are first aggregated according to the topological relationship to form parts to be fitted (referred to as carto). Then the elevation values ​​of the intersection edge points in carto are fitted (i.e., the effective elevation values ​​described above). On this basis, based on the above-obtained point cloud recognition results, some sampling points (such as digital elevation model (DEM) elevation data or trajectory points) are further expanded. After fitting the plane based on these intersection edge points and the expanded sampling points, the elevation values ​​of the interior points (road points and lane line points in the intersection) are predicted based on the RANSAC algorithm. Then, the fitting results are optimized based on the prediction results, and the intersection edges in the intersection plane are further smoothed.

[0255] Refer to Figure 15, which is a schematic diagram of the intersection hole fitting effect in an embodiment of the present application. The black dots and lines represent lane boundaries, and the gray dots and lines represent the centerlines of the lane boundaries. The lane centerline is a centerline determined based on the left and right lane boundaries. In the embodiment of the present application, the lane centerline can be used in scenarios such as autonomous driving. By reference to the lane centerline, the vehicle can be kept in the exact center of the lane.

[0256] In the embodiment of the present application, intersections are a relatively special scenario because the roads and lane boundaries at the intersections are not bound to each other. The graph propagation algorithm introduced above does not need to process these intersections, and these intersections will be transmitted through.

[0257] In the above implementation, the hole fitting-based approach can handle special scenarios such as intersections, further enriching and improving the processing of elevation information.

[0258] In summary, the automated road lane elevation processing method based on point cloud recognition results proposed in this application can effectively improve the quality of elevation data in current high-precision maps and meet customer requirements for autonomous driving simulation environments. Furthermore, in daily data production operations, it greatly alleviates the difficulty and low efficiency of drawing geometry in 3D models, improving the efficiency and accuracy of various data production operations.

[0259] Based on the same inventive concept, the embodiment of the present application further provides a device for processing elevation information. As shown in FIG16 , it is a schematic structural diagram of the device 1600 for processing elevation information, which may include:

[0260] The sampling unit 1601 is used to sample lane data of the road in the target area to obtain multiple sampling points;

[0261] The fitting unit 1602 determines the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points from a plurality of point cloud recognition results obtained by recognizing a plurality of point clouds in the target area, uses the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point, and forms a candidate elevation value set for the sampling point; performs fitting on the plurality of sampling points to obtain fitting results, and determines one or more target sampling points from the plurality of sampling points based on the fitting results;

[0262] The fusion unit 1603 is configured to perform the following operations for each candidate elevation value set of the one or more target sampling points: determining a confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining a valid elevation value of the target sampling point based on the confidence level of each candidate elevation value.

[0263] The processing device 1600 may further include an optimization unit 1604 configured to optimize the effective elevation value of each target sampling point based on the topological relationship between roads in the target area to obtain a target elevation value of each target sampling point.

[0264] For example, the fusion unit 1603 is specifically configured to:

[0265] Determining, based on the reference map data and each point cloud recognition result in the plurality of point cloud recognition results, a confidence level of each candidate elevation value in the candidate elevation value set of the target sampling point;

[0266] Based on the confidence level of each candidate elevation value, each candidate elevation value is weightedly fused to obtain the effective elevation value of the target sampling point.

[0267] For example, the device further comprises:

[0268] A first interpolation unit 1605 is configured to determine a first target interpolation point on a lane within a target area if the number of target sampling points on a lane within the target area among the one or more target sampling points is less than a preset number threshold;

[0269] Based on the effective elevation values ​​of target sampling points on both sides of the first target interpolation point among the target sampling points on the lane, linear interpolation is performed on the first target interpolation point to obtain a first interpolation elevation value as the effective elevation value of the first target interpolation point.

[0270] The optimization unit 1604 is further configured to:

[0271] Based on the topological relationship between the roads in the target area, the effective elevation value of each first target interpolation point is optimized to obtain the target elevation value of each first target interpolation point.

[0272] For example, the fitting unit 1602 is specifically configured to:

[0273] performing plane fitting on the plurality of sampling points according to the plurality of elevation values ​​of each sampling point in the plurality of sampling points to obtain a fitting plane corresponding to the target area;

[0274] If the angle between the fitting plane and the standard normal vector is not less than a first preset angle threshold, the sampling point in the fitting plane among the multiple sampling points is used as the target sampling point; the standard normal vector is a normal vector perpendicular to the horizontal ground.

[0275] For example, the fitting unit 1602 is further configured to:

[0276] If the angle between the fitting plane and the standard normal vector is less than a first preset angle threshold, the latitude and longitude information of each sampling point is converted into a linear reference distance, where the linear reference distance represents the distance between the sampling point and the starting point of the lane line to which the sampling point belongs; the standard normal vector is a normal vector perpendicular to the horizontal ground; and the lane line is a lane line within the target area;

[0277] According to the linear reference distance of multiple sampling points, line fitting is performed on multiple sampling points to obtain the fitting straight line result of the target area;

[0278] The sampling points on the target fitting line in the fitting line result among the multiple sampling points are used as target sampling points. The target fitting line is a fitting line whose angle with the standard normal vector is not less than a second preset angle threshold.

[0279] For example, the device further includes:

[0280] The second interpolation unit 1606 is configured to determine a second target interpolation point on a non-target fitted line in the fitted line result;

[0281] Determining target sampling points on a target fitting line adjacent to the non-target fitting line along the longitudinal direction of the road in the target area on both sides of the second target interpolation point;

[0282] Based on the effective elevation values ​​of the target sampling points on both sides of the second target interpolation point, linear interpolation is performed on the second target interpolation point to obtain a second interpolation elevation value as the effective elevation value of the second target interpolation point.

[0283] The optimization unit 1604 is further configured to:

[0284] Based on the topological relationship between the roads in the target area, the effective elevation value of each second target interpolation point is optimized to obtain the target elevation value of each second target interpolation point.

[0285] For example, the optimization unit 1604 is specifically configured to:

[0286] Determining a target point having a valid elevation value, the target point comprising at least one of a target sampling point, a first target interpolation point, and a second target interpolation point;

[0287] The roads in the target area are used as nodes in the road network graph;

[0288] By clustering target sampling points on the road corresponding to each node among the plurality of nodes, at least one elevation category of each node is obtained, wherein the at least one elevation category represents an elevation value of the corresponding road;

[0289] Determining a sum of confidences of candidate elevation values ​​of target sampling points on the corresponding road belonging to a corresponding elevation category in the at least one elevation category, and using the sum of confidences as a feature evaluation value of the corresponding elevation category;

[0290] Select a target node from multiple nodes based on the feature evaluation value of each node;

[0291] Starting from a target node, along the topological relationships between multiple nodes in the road network graph, the target elevation category of the target node is propagated to the surrounding areas of the target node, and the target elevation categories of other nodes in the road network graph are obtained according to the propagation results;

[0292] The target elevation category of the node associated with the target point is compared with the effective elevation value of the target point, and the target elevation value of the target point is obtained according to the comparison result.

[0293] For example, the optimization unit 1604 is specifically configured to select a target node by at least one of the following methods:

[0294] A node having only one characteristic evaluation value among the plurality of nodes is used as a target node;

[0295] A node among the multiple nodes, the node for which the ratio of a corresponding feature evaluation value to other feature evaluation values ​​is greater than a preset ratio threshold, is used as a target node.

[0296] For example, the optimization unit 1604 is further configured to obtain the elevation category corresponding to each node in the following manner:

[0297] Clustering the candidate elevation values ​​of one or more target sampling points based on the candidate elevation value set of each target sampling point to obtain at least one cluster, wherein a difference between elevation values ​​of cluster centers of different clusters in the at least one cluster is greater than a preset elevation threshold;

[0298] For each node, the following operations are performed: determine the elevation value of the cluster center corresponding to each target sampling point associated with the node, and use the determined elevation value as the elevation value represented by the elevation category corresponding to the node.

[0299] For example, the optimization unit 1604 is specifically configured to:

[0300] Compare the obtained elevation value corresponding to the target elevation category with the effective elevation value of a corresponding target point in the fitting plane;

[0301] If the comparison result is within the preset difference range, the effective elevation value of the target point is used as the target elevation value of the target point;

[0302] If the comparison result is outside the preset difference range, the elevation value represented by the target elevation category of the node associated with the target point is used as the target elevation value of the target point.

[0303] For example, the device further includes:

[0304] The weight calculation unit 1607 is used to determine the confidence level of each candidate elevation value by:

[0305] determining a difference weight for the candidate elevation value based on a difference between a segment of a point cloud lane line formed by the target sampling point and the next adjacent target sampling point in the point cloud recognition result associated with the candidate elevation value and a segment of a map lane line formed by corresponding points in the reference map data;

[0306] Determining an acquisition time attenuation coefficient of the candidate elevation value according to an acquisition time of a point cloud recognition result associated with the candidate elevation value;

[0307] Determining a preset category reduction coefficient corresponding to the identified category based on the category of the target sampling point identified in the point cloud recognition result associated with the candidate elevation value;

[0308] The confidence level of the candidate elevation value is determined based on the difference weight, the acquisition time attenuation coefficient, and the preset category de-weighting coefficient.

[0309] Illustratively, the difference weight includes at least one of a distance weight and an angle weight;

[0310] The weight calculation unit 1607 is specifically configured to perform at least one of the following operations:

[0311] Determining a distance weight of the candidate elevation value based on a lateral distance between the segment of the point cloud lane line and the segment of the map lane line;

[0312] An angle weight of the candidate elevation value is determined according to a longitudinal angle between the segment of point cloud lane line and the segment of map lane line.

[0313] For example, the difference weight includes a distance weight and an angle weight, and the weight calculation unit 1607 is specifically configured to:

[0314] Determine the sum of the product of the distance weight and the first preset model parameter and the product of the angle weight and the second preset model parameter, and record it as the target parameter;

[0315] The product of the acquisition time attenuation coefficient, the preset category weight reduction coefficient and the target parameter is used as the confidence level corresponding to the candidate elevation value.

[0316] For example, the weight calculation unit 1607 is specifically configured to:

[0317] Determining a distance weight of the candidate elevation value according to a ratio of the square of the lateral distance to the square of a third preset model parameter;

[0318] And, the angle weight of the candidate elevation value is determined according to the ratio of the square of the longitudinal angle to the square of the fourth preset model parameter.

[0319] For example, the weight calculation unit 1607 is specifically configured to:

[0320] An acquisition time attenuation coefficient is determined according to a difference between an acquisition time of a point cloud recognition result associated with the candidate elevation value and a maximum acquisition time, wherein the maximum acquisition time is a maximum value among acquisition times of multiple point cloud recognition results.

[0321] For example, if the target area includes an intersection, the device further includes:

[0322] The intersection processing unit 1608 is used to aggregate the roads and lane lines in the intersection into a part to be fitted according to the topological relationship; sample points of the part to be fitted based on the multiple point cloud recognition results; fit the intersection plane corresponding to the intersection based on the effective elevation values ​​of the intersection edge points in the part to be fitted and the effective elevation values ​​of the expanded sampling points; predict the inner points in the intersection, and smooth the intersection edge in the intersection plane based on the prediction results, where the inner points are at least one of the road points and lane line points in the intersection.

[0323] Because the elevation information processing device in this application is implemented based on point cloud recognition results, plane fitting can be performed on the point cloud recognition results to effectively remove noise points and obtain useful target sampling points. Candidate elevation values ​​for the sampling points can also be obtained from the point cloud recognition results. Based on this, these candidate elevation values ​​for the same target sampling point can be fused based on the confidence level of the candidate elevation values ​​corresponding to the target sampling point, and the effective elevation values ​​can be optimized based on the topological relationship between roads to obtain the processed target elevation value. This processing is based on point cloud recognition results. Compared to current trajectory matching road network methods that use trajectory as the primary basis, this application directly evaluates the elevation accuracy of road lane data based on point cloud recognition results. Point cloud recognition results are not affected by factors such as GPS installation location, and the information they contain is richer and more accurate than trajectory data. Furthermore, through methods such as resampling, plane fitting, and optimization based on topological relationships, road lane data can be effectively and automatically processed for elevation, effectively improving accuracy.

[0324] On this basis, automatic elevation processing of road lanes based on point cloud recognition results can effectively improve the quality of elevation data in the current map, thereby improving the authenticity of the real environment simulated in the autonomous driving simulation link, so as to provide objects in need (such as driving objects) with a simulation environment that is more consistent with the real environment.

[0325] For the convenience of description, the above parts are divided into modules (or units) according to their functions and described separately. Of course, when implementing this application, the functions of each module (or unit) can be implemented in the same or multiple software or hardware.

[0326] After introducing the method and apparatus for processing elevation information according to an exemplary embodiment of the present application, an electronic device according to another exemplary embodiment of the present application will be introduced.

[0327] Those skilled in the art will appreciate that various aspects of the present application can be implemented as systems, methods, or program products. Therefore, various aspects of the present application can be specifically implemented in the following forms: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation that combines hardware and software aspects, which may be collectively referred to herein as a "circuit," "module," or "system."

[0328] Based on the same inventive concept as the above-mentioned method embodiment, an electronic device is also provided in an embodiment of the present application. In one embodiment, the electronic device may be a server, such as server 120 shown in FIG1 . In this embodiment, the structure of the electronic device may be as shown in FIG17 , including a memory 1701, a communication module 1703, and one or more processors 1702.

[0329] Memory 1701 is used to store computer programs executed by processor 1702. Memory 1701 may include a program storage area and a data storage area. The program storage area may store an operating system and programs required for running instant messaging functions, while the data storage area may store various instant messaging messages and operating instruction sets.

[0330] Memory 1701 may be a volatile memory, such as random-access memory (RAM); a non-volatile memory, such as read-only memory, flash memory, a hard disk drive (HDD), or a solid-state drive (SSD); or any other medium capable of carrying or storing a desired computer program in the form of instructions or data structures and accessible by a computer, but is not limited thereto. Memory 1701 may be a combination of the aforementioned memories.

[0331] The processor 1702 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 1702 is configured to implement the above-mentioned method for processing elevation information when calling the computer program stored in the memory 1701 .

[0332] The communication module 1703 is used to communicate with terminal devices and other servers.

[0333] The specific connection medium between the memory 1701, communication module 1703, and processor 1702 is not limited in the embodiments of the present application. In Figure 17, the memory 1701 and processor 1702 are connected via bus 1704. Bus 1704 is depicted as a bold line in Figure 17. The connection methods between other components are merely schematic and are not intended to be limiting. Bus 1704 can be divided into an address bus, a data bus, a control bus, and the like. For ease of description, Figure 17 depicts only one bold line, but does not represent only one bus or one type of bus.

[0334] The memory 1701 stores a computer storage medium, which stores computer executable instructions for implementing the elevation information processing method of the embodiment of the present application. The processor 1702 is used to execute the elevation information processing method, as shown in FIG3 .

[0335] In another embodiment, the electronic device may be another electronic device, such as the terminal device 110 shown in FIG1 . In this embodiment, the structure of the electronic device may be as shown in FIG18 , including: a communication component 1810, a memory 1820, a display unit 1830, a camera 1840, a sensor 1850, an audio circuit 1860, a Bluetooth module 1870, a processor 1880, and other components.

[0336] The communication component 1810 is used to communicate with the server. In some embodiments, it may include a wireless fidelity (WiFi) module. The WiFi module is a short-range wireless transmission technology. Electronic devices can help users send and receive information through the WiFi module.

[0337] The memory 1820 can be used to store software programs and data. The processor 1880 executes various functions and data processing of the terminal device 110 by running the software programs or data stored in the memory 1820. The memory 1820 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. The memory 1820 stores an operating system that enables the terminal device 110 to operate. In the present application, the memory 1820 can store the operating system and various application programs, and may also store a computer program that executes the elevation information processing method of the embodiment of the present application.

[0338] The display unit 1830 can also be used to display information input by the user or provided to the user, as well as a graphical user interface (GUI) for displaying various menus of the terminal device 110. Specifically, the display unit 1830 may include a display screen 1832 disposed on the front of the terminal device 110. The display screen 1832 may be configured in the form of a liquid crystal display, a light-emitting diode, etc. The display unit 1830 can be used to display the relevant interface of the map software in the embodiments of the present application, etc.

[0339] The display unit 1830 can also be used to receive input digital or character information and generate signal input related to the user settings and function control of the terminal device 110. Specifically, the display unit 1830 may include a touch screen 1831 arranged on the front of the terminal device 110, which can collect user touch operations on or near it, such as clicking a button, dragging a scroll box, etc.

[0340] The touch screen 1831 can be covered on the display screen 1832, or the touch screen 1831 and the display screen 1832 can be integrated to realize the input and output functions of the terminal device 110. The integrated display screen can be simply called a touch screen. In this application, the display unit 1830 can display applications and corresponding operation steps.

[0341] Camera 1840 can be used to capture still images, and users can publish images captured by camera 1840 through an application. There can be one or more cameras 1840. The object generates an optical image through the lens and projects it onto the photosensitive element. The photosensitive element can be a charge coupled device (CCD) or a complementary metal-oxide-semiconductor (CMOS) phototransistor. The photosensitive element converts the optical signal into an electrical signal, which is then transmitted to processor 1880 for conversion into a digital image signal.

[0342] The terminal device may further include at least one sensor 1850, such as an acceleration sensor 1851, a distance sensor 1852, a fingerprint sensor 1853, and a temperature sensor 1854. The terminal device may also be configured with other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, an infrared sensor, a light sensor, and a motion sensor.

[0343] The audio circuit 1860, speaker 1861, and microphone 1862 provide an audio interface between the user and the terminal device 110. The audio circuit 1860 can convert the received audio data into an electrical signal and transmit it to the speaker 1861, which converts it into a sound signal for output. The terminal device 110 may also be equipped with a volume button for adjusting the volume of the sound signal. On the other hand, the microphone 1862 converts the collected sound signal into an electrical signal, which is received by the audio circuit 1860 and converted into audio data. The audio data is then output to the communication component 1810 for transmission to, for example, another terminal device 110, or the audio data is output to the memory 1820 for further processing.

[0344] The Bluetooth module 1870 is used to exchange information with other Bluetooth devices having a Bluetooth module through the Bluetooth protocol. For example, the terminal device can establish a Bluetooth connection with a wearable electronic device (such as a smart watch) that also has a Bluetooth module through the Bluetooth module 1870 to exchange data.

[0345] The processor 1880 is the control center of the terminal device. It uses various interfaces and lines to connect the various parts of the entire terminal. By running or executing software programs stored in the memory 1820 and calling data stored in the memory 1820, it performs various functions of the terminal device and processes data. In some embodiments, the processor 1880 may include one or more processing units; the processor 1880 may also integrate an application processor and a baseband processor, wherein the application processor can process an operating system, a user interface, and applications, etc., and the baseband processor can process wireless communications. It is understandable that the above-mentioned baseband processor may not be integrated into the processor 1880. In the present application, the processor 1880 can run an operating system, applications, user interface display and touch response, as well as the method for processing elevation information in the embodiment of the present application. In addition, the processor 1880 is coupled to the display unit 1830.

[0346] In some possible implementations, various aspects of the elevation information processing method provided in the present application can also be implemented in the form of a program product, which includes a computer program. When the program product is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of the elevation information processing method according to various exemplary embodiments of the present application described above in this specification. For example, the electronic device can execute the steps shown in Figure 3.

[0347] The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable 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 thereof.

[0348] The program product of the embodiment of the present application may be a portable compact disc read-only memory (CD-ROM) and include a computer program, and can be run on an electronic device. However, the program product of the present application is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with a command execution system, apparatus, or device.

[0349] A readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries a readable computer program. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with a command execution system, apparatus, or device.

[0350] The computer program embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0351] The computer program for performing the operations of the present application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The computer program can be executed entirely on the user electronic device, partially on the user electronic device, as a separate software package, partially on the user electronic device and partially on a remote electronic device, or entirely on a remote electronic device or server. In cases involving remote electronic devices, the remote electronic device can be connected to the user electronic device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external electronic device (for example, using an Internet service provider to connect through the Internet).

[0352] It should be noted that although several units or subunits of the device are mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, depending on the embodiment of the application, the features and functions of two or more units described above can be embodied in a single unit. Conversely, the features and functions of a single unit described above can be further divided and embodied by multiple units.

[0353] Furthermore, although the operations of the method of the present application are described in a particular order in the accompanying drawings, this does not require or imply that the operations must be performed in this particular order, or that all illustrated operations must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0354] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain a computer-usable computer program.

[0355] The present application is described with reference to the flow chart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present application. It should be understood that each flow process and / or box in the flow chart and / or block diagram and the combination of the flow process and / or box in the flow chart and / or block diagram can be realized by computer program commands. These computer program commands can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processing machine or other programmable data processing device to produce a machine, so that the command executed by the processor of the computer or other programmable data processing device produces a device for realizing the function specified in one flow chart flow or multiple flows and / or one box or multiple boxes of the block diagram.

[0356] These computer program commands may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the commands stored in the computer-readable memory produce a manufactured product including a command device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0357] These computer program commands can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the commands executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0358] Although some embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the described embodiments and all changes and modifications that fall within the scope of this application.

Claims

1. A method for processing elevation information, characterized in that: The method comprises: Sampling lane data of roads in a target area to obtain multiple sampling points; Determine the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points from a plurality of point cloud recognition results obtained by recognizing a plurality of point clouds of the target area, and use the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point to form a candidate elevation value set of the sampling point; Fitting the multiple sampling points to obtain fitting results, and determining one or more target sampling points among the multiple sampling points according to the fitting results; For the candidate elevation value set of each target sampling point in the one or more target sampling points, the following operations are performed respectively: determining the confidence of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining the valid elevation value of the target sampling point according to the confidence of each candidate elevation value.

2. The method according to claim 1, characterized in that The step of determining the confidence of each candidate elevation value in the candidate elevation value set of the target sampling point, and determining the effective elevation value of the target sampling point according to the confidence of each candidate elevation value, includes: Determine the confidence of each candidate elevation value in the candidate elevation value set of the target sampling point based on the reference map data and each point cloud recognition result in the plurality of point cloud recognition results; Based on the confidence of each candidate elevation value, the candidate elevation values ​​are weightedly fused to obtain the effective elevation value of the target sampling point.

3. The method according to claim 1 or 2, characterized in that The method further comprises: If the number of target sampling points on a lane in the target area among the one or more target sampling points is less than a preset number threshold, determining a first target interpolation point on the lane; Based on the effective elevation values ​​of the target sampling points on both sides of the first target interpolation point among the target sampling points on the lane, linear interpolation is performed on the first target interpolation point to obtain a first interpolation elevation value as the effective elevation value of the first target interpolation point.

4. The method according to any one of claims 1 to 3, characterized in that The fitting of the multiple sampling points to obtain fitting results, and determining one or more target sampling points among the multiple sampling points according to the fitting results, includes: performing plane fitting on the multiple sampling points according to multiple elevation values ​​of each sampling point among the multiple sampling points to obtain a fitting plane corresponding to the target area; If the angle between the fitting plane and the standard normal vector is not less than a first preset angle threshold, a sampling point on the fitting plane among the multiple sampling points is used as a target sampling point, wherein the standard normal vector is a normal vector perpendicular to the horizontal ground.

5. The method according to claim 4, characterized in that The method further comprises: If the angle between the fitting plane and the standard normal vector is less than the first preset angle threshold, converting the longitude and latitude information of each sampling point in the multiple point cloud recognition results into a linear reference distance, where the linear reference distance represents the distance between the sampling point and the starting point of the lane line to which the sampling point belongs, and the lane line is a section of the lane line in the target area; Performing line fitting on the plurality of sampling points according to the linear reference distances of the plurality of sampling points to obtain a fitting straight line result of the target area; A sampling point on a target fitting line in the fitting line result among the multiple sampling points is used as a target sampling point, and the target fitting line is a fitting line whose angle with the standard normal vector is not less than a second preset angle threshold.

6. The method according to claim 5, characterized in that The method further comprises: For a non-target fitted straight line in the fitted straight line result, determining a second target interpolation point on the non-target fitted straight line; Determine target sampling points on a target fitting straight line adjacent to the non-target fitting straight line along the longitudinal direction of the road in the target area on both sides of the second target interpolation point; Based on the effective elevation values ​​of the target sampling points on both sides of the second target interpolation point, linear interpolation is performed on the second target interpolation point to obtain a second interpolation elevation value as the effective elevation value of the second target interpolation point.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: Determine a target point having a valid elevation value, wherein the target point includes at least one of a target sampling point, a first target interpolation point, and a second target interpolation point; Based on the topological relationship between the roads in the target area, the effective elevation value of the target point is optimized to obtain the target elevation value of the target point, including: Using roads in the target area as multiple nodes in a road network graph; By clustering target sampling points on the road corresponding to each of the multiple nodes, at least one elevation category of each node is obtained, wherein the at least one elevation category represents an elevation value of the corresponding road; Determine the sum of confidences of candidate elevation values ​​of target sampling points belonging to a corresponding elevation category in the at least one elevation category on the corresponding road, and use the sum of confidences as a feature evaluation value of the corresponding elevation category; Selecting a target node from the plurality of nodes according to a characteristic evaluation value of each node; Starting from the target node, along the topological relationship between multiple nodes in the road network graph, the target elevation category of the target node is propagated to the surrounding of the target node, and the target elevation categories of other nodes in the road network graph are obtained according to the propagation result; The target elevation category of the node associated with the target point is compared with the effective elevation value of the target point, and the target elevation value of the target point is obtained according to the comparison result.

8. The method according to claim 7, characterized in that The selecting a target node from the plurality of nodes according to the feature evaluation value of each node includes at least one of the following methods: Among the multiple nodes, a node having only one characteristic evaluation value is used as a target node; A node among the multiple nodes, the node corresponding to which the ratio of a corresponding feature evaluation value to other feature evaluation values ​​is greater than a preset ratio threshold, is taken as a target node.

9. The method according to claim 7 or 8, characterized in that At least one elevation category corresponding to each node is obtained as follows: Based on the candidate elevation value sets of the target sampling points, clustering the candidate elevation values ​​of the one or more target sampling points to obtain at least one cluster, wherein in the at least one cluster, the difference between the elevation values ​​of cluster centers of different clusters is greater than a preset elevation threshold; For each node, the following operations are performed respectively: the elevation value of the cluster center corresponding to each target sampling point associated with the node is determined, and the determined elevation value is used as the elevation value represented by the elevation category corresponding to the node.

10. The method according to any one of claims 7 to 9, characterized in that: Obtaining a target elevation value of the target point according to the comparison result includes: If the comparison result is within a preset difference range, the effective elevation value of the target point is used as the target elevation value of the target point; If the comparison result is outside the preset difference range, the elevation value represented by the target elevation category of the node associated with the target point is used as the target elevation value of the target point.

11. The method according to any one of claims 1 to 10, characterized in that: Determining the confidence of each candidate elevation value in the candidate elevation value set of the target sampling point includes: Determine a difference weight of the candidate elevation value according to a difference between a segment of point cloud lane line formed by the target sampling point and the next adjacent target sampling point in the point cloud recognition result associated with the candidate elevation value and a segment of map lane line formed by corresponding points in the reference map data; Determining an acquisition time attenuation coefficient of the candidate elevation value according to the acquisition time of the point cloud recognition result associated with the candidate elevation value; According to the category of the target sampling point identified in the point cloud recognition result associated with the candidate elevation value, determining a preset category downweighting coefficient corresponding to the identified category; The confidence level of the candidate elevation value is determined according to the difference weight, the acquisition time attenuation coefficient and the preset category reduction coefficient.

12. The method according to claim 11, characterized in that The difference weight includes at least one of a distance weight and an angle weight; Then, determining the difference weight of the candidate elevation value includes at least one of the following operations: Determining a distance weight of the candidate elevation value according to a lateral distance between the segment of the point cloud lane line and the segment of the map lane line; The angle weight of the candidate elevation value is determined according to the longitudinal angle between the segment of point cloud lane line and the segment of map lane line.

13. The method according to claim 12, characterized in that The difference weight includes a distance weight and an angle weight, and the confidence of the candidate elevation value is determined according to the difference weight, the acquisition time attenuation coefficient and the preset category reduction coefficient, including: Determine the sum of the product of the distance weight and the first preset model parameter and the product of the angle weight and the second preset model parameter, and record it as the target parameter; The product of the acquisition time attenuation coefficient, the preset category weight reduction coefficient and the target parameter is used as the confidence of the candidate elevation value.

14. The method according to claim 12, characterized in that The determining the distance weight of the candidate elevation value according to the lateral distance between the segment of the point cloud lane line and the segment of the map lane line comprises: Determining the distance weight of the candidate elevation value according to the ratio of the square of the lateral distance to the square of the third preset model parameter; And, determining the angle weight of the candidate elevation value according to the longitudinal angle between the segment of the point cloud lane line and the segment of the map lane line includes: The angle weight of the candidate elevation value is determined according to the ratio of the square of the longitudinal angle to the square of the fourth preset model parameter.

15. The method according to any one of claims 11 to 14, characterized in that The step of determining the acquisition time attenuation coefficient of the candidate elevation value according to the acquisition time of the point cloud recognition result associated with the candidate elevation value includes: The acquisition time attenuation coefficient is determined according to the difference between the acquisition time of the point cloud recognition result associated with the candidate elevation value and the maximum acquisition time, wherein the maximum acquisition time is the maximum value of the acquisition times of the multiple point cloud recognition results.

16. The method according to any one of claims 1 to 15, characterized in that: If the target area includes an intersection, the method further includes: Aggregating the roads and lane lines in the intersection into a part to be fitted according to a topological relationship; Expanding sampling points of the part to be fitted according to the multiple point cloud recognition results; According to the effective elevation values ​​of the edge points of the intersection in the part to be fitted and the effective elevation values ​​of the expanded sampling points, the intersection plane of the intersection is obtained by fitting; Predicting an inner point in the intersection, and smoothing an edge of the intersection in the intersection plane according to the prediction result, wherein the inner point is at least one of a road point and a lane line point in the intersection.

17. A device for processing elevation information, characterized in that: include: A sampling unit, used for sampling lane data of roads in a target area to obtain a plurality of sampling points; A fitting unit, configured to determine the plurality of sampling points and a plurality of elevation values ​​of each of the plurality of sampling points from a plurality of point cloud recognition results obtained by recognizing the plurality of point clouds of the target area, and use the plurality of elevation values ​​of each sampling point as candidate elevation values ​​of the sampling point to form a candidate elevation value set of the sampling point; Fitting the multiple sampling points to obtain fitting results, and determining one or more target sampling points among the multiple sampling points according to the fitting results; The fusion unit is used to perform the following operations for the candidate elevation value set of each target sampling point in the one or more target sampling points: determine the confidence of each candidate elevation value in the candidate elevation value set of the target sampling point, and determine the effective elevation value of the target sampling point according to the confidence of each candidate elevation value.

18. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of any one of the methods of claims 1 to 16.

19. A computer-readable storage medium, characterized in that: It comprises a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of any method described in claims 1 to 16.

Images