Map data filtering method and device, electronic equipment and storage medium

Abstract

The invention provides a map data filtering method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a driving track of a vehicle in a current driving process, obtaining vector data associated with the driving track from a vector map, and obtaining true value data associated with the driving track from a high-precision true value map; map feature information corresponding to the driving track is constructed through the vector data and the truth value data, and the map feature information comprises a track surface, a semantic object set and a truth value set; obtaining a region of interest corresponding to the vehicle, and determining a target semantic object matched with the region of interest in the semantic object set; and filtering the truth value set by using the target semantic object to obtain a filtered truth value set, and marking the filtered truth value set based on the trajectory surface to generate target map data. The problems that in the prior art, map redundant data cannot be completely removed, and key map semantics and attributes cannot be reserved are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle maps, and particularly relates to a method, device, electronic device and storage medium for filtering map data. Background Art

[0002] High-precision maps are important auxiliary tools for autonomous vehicles, which contain rich road information, semantics, driving trajectories and other data. During the production of maps, it is necessary to collect original visual, positioning and other data through vehicles, and obtain vectorized geometric information after processing. These vectorized map data contain a large amount of information such as road semantics, driving trajectories, and ground dirty data. Before using these data to make maps, detailed data analysis and filtering are required to eliminate a large amount of invalid or redundant data and obtain key available data.

[0003] In the prior art, map data filtering methods have been proposed from different perspectives, but only a single filtering requirement is concerned, and invalid or redundant data cannot be completely eliminated, resulting in a decrease in the usability of map data. In addition, the existing methods cannot retain key map semantics and attributes, which will have an adverse impact on the subsequent analysis and use of high-precision maps. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method, device, electronic device and storage medium for filtering map data to solve the problem that the prior art cannot completely eliminate redundant data in map data and cannot retain key map semantics and attributes.

[0005] In a first aspect, an embodiment of the present invention provides a method for filtering map data, the method comprising:

[0006] Obtain the driving trajectory during the current driving process of the vehicle, and obtain the vector data associated with the driving trajectory from the vector map and the true value data associated with the driving trajectory from the high-precision ground truth map;

[0007] Construct map feature information corresponding to the driving trajectory by using the vector data and the true value data, wherein the map feature information includes: a trajectory surface, a semantic object set, and a true value set;

[0008] Obtain the region of interest corresponding to the vehicle, and determine the target semantic object in the semantic object set that matches the region of interest;

[0009] Filter the true value set by using the target semantic object to obtain a filtered true value set, and mark the filtered true value set based on the trajectory surface to generate target map data.

[0010] Further, constructing the map feature information corresponding to the driving trajectory by using the vector data and the ground truth data includes:

[0011] Analyze the vector data to obtain trajectory data and object data of multiple semantic objects;

[0012] Perform longitude and latitude coordinate conversion on the trajectory data, the object data, and the ground truth data;

[0013] Generate a trajectory surface by using the trajectory data after coordinate conversion;

[0014] Generate a ground truth set by using the ground truth data after coordinate conversion;

[0015] Extract the contour information of the semantic objects based on the object data after coordinate conversion, and fuse the contour information of multiple semantic objects to obtain the semantic object set;

[0016] Construct map feature information based on the trajectory surface, the ground truth set, and the semantic object set.

[0017] Further, generating a trajectory surface by using the trajectory data after coordinate conversion includes:

[0018] Obtain the first longitude and latitude information from the trajectory data after coordinate conversion;

[0019] Generate trajectory points based on the first longitude and latitude information, and fit the trajectory points into an initial curve;

[0020] Project the initial curve onto a preset coordinate system, and expand a preset distance in a specified direction within the preset coordinate system to generate the trajectory surface.

[0021] Further, generating a ground truth set by using the ground truth data after coordinate conversion includes:

[0022] Obtain the second longitude and latitude information from the ground truth data after coordinate conversion;

[0023] Generate ground truth points based on the second longitude and latitude information, and calculate the Euclidean distance of the ground truth points;

[0024] Calculate the buffer of the ground truth points by using the Euclidean distance, and merge the buffers corresponding to each ground truth point to obtain the ground truth set.

[0025] The method provided by the embodiments of the present application helps to form a more complete and continuous data representation on the map by generating ground truth points and calculating their buffers, thereby providing more comprehensive information; by merging the buffers, a ground truth set can be obtained, which helps to reduce the data volume and improve the data processing efficiency.

[0026] Further, filtering the truth value set by using the target semantic object to obtain a filtered truth value set includes:

[0027] Construct a semantic space tree by using the target semantic object;

[0028] Traverse the semantic space tree to obtain the associated semantic objects of each trajectory point in the trajectory plane;

[0029] Update the semantic object set based on the associated semantic objects to obtain an updated semantic object set;

[0030] Filter the truth value data in the truth value set by using the updated semantic object set to obtain a filtered truth value set.

[0031] Further, updating the semantic object set based on the associated semantic objects to obtain an updated semantic object set includes:

[0032] Index the spatial point set of the associated semantic object in the trajectory plane based on the spatial position of the associated semantic object, where the spatial point set includes: a first spatial point, a second spatial point, a third spatial point, and a fourth spatial point. The first spatial point is the spatial point closest to the associated semantic object, the second spatial point is located in the first direction of the first spatial point, the third spatial point is located in the second direction of the first spatial point, and the fourth spatial point is the spatial point closest to the first spatial point. The second direction is opposite to the first direction;

[0033] Calculate the first horizontal distance from the first spatial point to the second spatial point, the second horizontal distance from the first spatial point to the third spatial point, and the third horizontal distance from the first spatial point to the fourth spatial point;

[0034] Calculate the semantic visual distance by using the first horizontal distance, the second horizontal distance, and the third horizontal distance;

[0035] Mark the associated semantic object according to the semantic visual distance to obtain an updated semantic object set.

[0036] Further, after generating the corresponding target map data, the method further includes:

[0037] Obtain a preset attribute filtering rule;

[0038] Eliminate the attribute abnormal data in the target map data based on the attribute filtering rule.

[0039] The method provided by the embodiments of the present application can clear the errors, inaccuracies or abnormal information in the map data by filtering out the data with abnormal attributes, improving the quality and accuracy of the data; through the filtering rules, the data related to a specific task or application can be screened out, thereby reducing the complexity and redundancy of the map data.

[0040] In a second aspect, an embodiment of the present invention provides a filtering device for map data, the device includes:

[0041] A first acquisition module, configured to acquire the driving trajectory during the current driving of the vehicle, and acquire the vector data associated with the driving trajectory from the vector map and the ground truth data associated with the driving trajectory from the high-precision ground truth map;

[0042] A construction module, configured to construct map feature information corresponding to the driving trajectory by using the vector data and the ground truth data, wherein the map feature information includes: a trajectory surface, a semantic object set, and a ground truth set;

[0043] A second acquisition module, configured to acquire the region of interest corresponding to the vehicle, and determine the target semantic object in the semantic object set that matches the region of interest;

[0044] A filtering module, configured to filter the ground truth set by using the target semantic object to obtain a filtered ground truth set, and mark the filtered ground truth set based on the trajectory surface to generate target map data.

[0045] In a third aspect, an embodiment of the present invention provides an electronic device, including: a memory and a processor, which are communicatively connected to each other, the memory stores computer instructions, and the processor executes the computer instructions to execute the method according to the first aspect or any corresponding embodiment thereof.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to execute the method according to the first aspect or any corresponding embodiment thereof.

[0047] The embodiments of the present application have the following advantages:

[0048] The method provided by the embodiments of this application provides a data basis for filtering map data by obtaining the current driving trajectory of the vehicle and acquiring the associated vector data in the vector map and the associated ground truth data in the high-precision ground truth map. By obtaining the real-time driving trajectory of the vehicle, the map data is filtered and corrected, removing errors and deviations, and improving the accuracy of the map data. By combining the vector data and the ground truth data, map feature information such as trajectory surfaces and semantic object sets can be extracted more accurately, which can more accurately reflect the actual situation of the vehicle's driving and enhance the readability and comprehensibility of the map data. By constructing map feature information such as trajectory surfaces and semantic object sets, the map data can be made more intuitive and can provide richer map information. By obtaining the region of interest of the vehicle, filtering can be performed according to the actual driving range of the vehicle, avoiding filtering out key information related to the vehicle's driving, and improving the pertinence and efficiency of filtering. By determining the target semantic object that matches the region of interest of the vehicle, personalized map data can be provided according to the driving needs and points of interest of different vehicles. By filtering the ground truth set using the target semantic object and marking the filtered ground truth set based on the trajectory surface, target map data can be generated, improving the accuracy and reliability of the map data.

[0049] The method provided by the embodiments of this application can obtain trajectory data and object data of multiple semantic objects by parsing the vector data. After these data are converted through longitude and latitude coordinates, they can more accurately reflect the physical features and semantic objects on the map, thereby extracting effective map information. By using the trajectory data and ground truth data after coordinate conversion, accurate trajectory surfaces and ground truth sets can be generated. These data can be used for further analysis of map features and semantic objects, improving the accuracy and reliability of the map data. By extracting the contour information of semantic objects based on the object data after coordinate conversion and fusing the contour information of multiple semantic objects, a semantic object set can be obtained. This process can remove redundant data, retain important map features and semantic objects, and improve the simplicity and readability of the map data. By constructing map feature information through trajectory surfaces, ground truth sets, and semantic object sets, multiple data sources can be integrated to provide more comprehensive and accurate map information.

[0050] The method provided by the embodiment of the present application can ensure the accuracy of longitude and latitude conversion, reduce the error of map data, and improve the precision of map data by obtaining the first longitude and latitude information from the trajectory data after coordinate conversion; generating trajectory points through the first longitude and latitude information can accurately identify points on the map and ensure the accuracy and integrity of map data; fitting the trajectory points into an initial curve can more intuitively display the trajectory features on the map and achieve the filtering of redundant and irregular trajectory points; by projecting the initial curve onto a preset coordinate system and expanding a preset distance in a specified direction within the preset coordinate system to generate a trajectory surface, the integration of multiple trajectory points is realized, a continuous and complete trajectory surface is generated, more comprehensive map data is provided, and the subsequent filtering process is facilitated.

[0051] The method provided by the embodiment of the present application realizes the process of data filtering based on specific requirements by constructing a semantic space tree, which can meet the requirements in different application scenarios. By associating the update strategy of semantic objects, unnecessary computing and storage consumption can be reduced to a certain extent. Updating the semantic object set can quickly filter the data in the true value set and improve the data processing efficiency. By using the updated semantic object set for data filtering, the relationship between the filtered true value data and the original true value data can be more intuitively understood, providing more accurate, efficient, and interpretable data support for the subsequent filtering process.

[0052] The method provided by the embodiment of the present application can more efficiently organize and query map data by indexing the spatial point set in the trajectory surface based on the spatial positions of associated semantic objects. This indexing method helps to quickly locate and access specific semantic objects; by calculating the distance between spatial points (such as the first horizontal distance from the first spatial point to the second spatial point), not only a quantitative measure of the spatial relationship is provided, but also it helps to understand and analyze map data; by using the horizontal distance to calculate the semantic visual distance, more accurate marking and classification of associated semantic objects can be performed. By marking associated semantic objects with the semantic visual distance, the relationship and relative position between different semantic objects can be more clearly reflected, and the updated semantic object set more accurately reflects the actual situation on the map. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 It is a schematic flowchart of a method for filtering map data according to some embodiments of the present invention;

[0055] Figure 2 is a schematic flow chart of generating a trajectory surface and a set of semantic objects according to some embodiments of the present invention;

[0056] Figure 3 is a schematic flow chart of generating a set of true values according to some embodiments of the present invention;

[0057] Figure 4 is a schematic flow chart of filtering map data according to some embodiments of the present invention;

[0058] Figure 5 is a structural block diagram of a filtering device for a map data according to an embodiment of the present invention;

[0059] Figure 6 is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed implementation manners

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0061] According to an embodiment of the present invention, there is provided a method, a device, an electronic device, and a storage medium for filtering map data. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0062] In this embodiment, a method for filtering map data is provided. Figure 1 is a flowchart of a method for filtering map data according to an embodiment of the present invention. As Figure 1 shown, the process includes the following steps:

[0063] Step S11, obtain the driving trajectory during the current driving of the vehicle, and obtain the vector data associated with the driving trajectory from the vector map and the true value data associated with the driving trajectory from the high-precision true value map.

[0064] It should be noted that a vector map is generated when a paper map is converted into a dot matrix digital image through opto - electrical conversion by a computer graphics and image system, followed by image processing and curve vectorization, or directly through manual tracking digitization. The generated vector map data file can be used for display, modification, annotation, roaming, calculation, management, and printing by a geographic information system. There is a corresponding relationship between the vector map and the ordinary paper map, and it is generated by converting raster data into vector data. In the schematic diagram of the vectorized map, geometric shapes such as lines and polygons are used to represent geographical features. The high - precision ground truth map, also known as the high - definition map (HD Map), is a map specifically for autonomous driving. The high - precision ground truth map can provide navigation information at the road level and lane - level navigation information. The schematic diagram of the high - precision map may display geographical information in the form of pixel points.

[0065] In the embodiment of the present application, the driving trajectory during the current driving is obtained based on the driving information within the vehicle's historical time period. Here, the driving information includes the position changes and GPS signals generated during the vehicle's driving process. Based on the driving trajectory during the current driving, it can be matched in the vector map and the high - precision ground truth map, and the ground truth data associated with the driving trajectory can be obtained.

[0066] Specifically, the specific process of matching the driving trajectory with the vector data in the vector map includes: matching the associated vector file from the vectorized map according to the driving trajectory during the vehicle's current driving. Here, the vector file is compressed and stored in the vectorized map in the protobuf format. After obtaining the vector file associated with the driving trajectory, the vector file is decompressed to obtain the vector data corresponding to the driving trajectory in the vector map.

[0067] Specifically, the specific process of matching the driving trajectory with the ground truth data in the high - precision ground truth map includes: matching the associated pixel points from the vectorized map according to the driving trajectory during the vehicle's current driving; extracting the data of the lane boundaries and road boundaries based on the associated pixel points; generating the corresponding ground truth data based on the extracted lane boundary and road boundary data.

[0068] The method provided by the embodiment of the present application provides a data basis for filtering map data by obtaining the vehicle's current driving trajectory and the associated vector data in the vector map and the associated ground truth data in the high - precision ground truth map according to the current driving trajectory. By obtaining the vehicle's real - time driving trajectory, the map data is filtered and corrected, removing errors and deviations, and improving the accuracy of the map data.

[0069] Step S12: Construct map feature information corresponding to the driving trajectory using vector data and ground truth data. The map feature information includes: trajectory surface, semantic object set, and ground truth set.

[0070] In the embodiment of the present application, the specific process of step S12 includes: parsing the vector data to obtain trajectory data and object data of multiple semantic objects; performing longitude and latitude coordinate conversion on the trajectory data, object data, and ground truth data; obtaining first longitude and latitude information from the trajectory data after coordinate conversion; generating trajectory points based on the first longitude and latitude information and fitting the trajectory points into an initial curve; projecting the initial curve onto a preset coordinate system and expanding a preset distance in a specified direction within the preset coordinate system to generate a trajectory surface; obtaining second longitude and latitude information from the ground truth data after coordinate conversion; generating ground truth points based on the second longitude and latitude information and calculating the Euclidean distance of the ground truth points; calculating the buffer of the ground truth points using the Euclidean distance, merging the buffers corresponding to each ground truth point to obtain a ground truth set; extracting contour information of semantic objects based on the object data after coordinate conversion and fusing the contour information of multiple semantic objects to obtain a semantic object set; constructing map feature information based on the trajectory surface, ground truth set, and semantic object set.

[0071] In the embodiment of the present application, constructing map feature information corresponding to the driving trajectory using vector data and ground truth data includes the following steps A1 - A6:

[0072] Step A1: Parse the vector data to obtain trajectory data and object data of multiple semantic objects.

[0073] In the embodiment of the present application, the vector data contains trajectory data during vehicle driving and object data of multiple semantic objects. The object information of semantic objects can include information such as the positions and shapes of various geographical objects such as roads, buildings, and waters. In vector data, it is often necessary to determine whether a point is within a surface and whether there is a direct correspondence between elements, that is, whether there is a spatial position relationship between vector points, lines, and surfaces. Using the inclusion analysis and data processing methods of vector data and GIS information, the inclusion analysis of existing vector data can be realized, and then the purpose of parsing vector data to obtain trajectory data and object data of multiple semantic objects can be achieved. Since the vector data is serialized and stored according to the Protobuf standard, the specific process of vector data parsing can include: deserializing and parsing the object data and trajectory data of multiple semantic objects in the vector map according to the standard protocol for the vector data; performing coordinate transcoding on the parsed object data and trajectory data to obtain a series of trajectory data with attributes and object data of multiple semantic objects.

[0074] Step A2: Perform longitude and latitude coordinate conversion on the trajectory data, object data, and ground truth data.

[0075] In the embodiments of the present application, the coordinate information of the data points stored in the trajectory data and object data obtained after vector data parsing, and the true value data obtained through the high-precision true value map is the map code (Map Code). The map code is converted into map longitude and latitude information through bit conversion and hexadecimal AND / OR operations. The specific process may include:

[0076] Conversion from Map Code to Bit Array: Map Code is usually a string or number representing a certain coordinate or path on the map. It is necessary to convert Map Code into binary form (bit array). This can be achieved by converting each digit (if it is a number) or each character (if it is a string) of Map Code into binary.

[0077] Processing of Bit Array: Determine the required coordinate accuracy. For example, longitude and latitude usually require six decimal places, but it can also be other precisions. According to the required precision, determine the length of the binary array.

[0078] Hexadecimal AND / OR Operations: Hexadecimal notation is often used to represent binary data; use the OR operation to combine different hexadecimal values. For example, if you have two hexadecimal values "A" and "B", the result of their OR operation will be "C"; use the AND operation to check specific bits. For example, the AND operation of "A" and "C" will return "A" because all bits of "C" are 1, and only the lowest two bits match "A".

[0079] Conversion to Longitude and Latitude: According to the format and encoding method of Map Code, perform specific conversion steps to extract longitude and latitude information from the bitarray and hexadecimal values; longitude and latitude are usually expressed in decimal degrees, ranging from -180 to 180 (longitude) and -90 to 90 (latitude).

[0080] Verification and Adjustment: During the conversion process, it is necessary to verify the accuracy of the data and make adjustments to ensure the correct coordinate representation; test different encoding and decoding methods to ensure integrity and accuracy.

[0081] Step A3: Generate a trajectory surface using the trajectory data after coordinate conversion.

[0082] In the embodiments of the present application, the specific process of step A3 includes: obtaining the first longitude and latitude information from the trajectory data after coordinate conversion; generating trajectory points based on the first longitude and latitude information, and fitting the trajectory points into an initial curve; projecting the initial curve onto a preset coordinate system, and expanding a preset distance in a specified direction within the preset coordinate system to generate a trajectory surface.

[0083] In the embodiment of the present application, the trajectory data after coordinate conversion contains all the longitude and latitude information associated with the trajectory data, that is, the first longitude and latitude information. The first longitude and latitude information is obtained from the trajectory data, and a series of discrete trajectory points can be generated based on the first longitude and latitude information; based on the generated trajectory points, the future driving trajectory can be predicted for trends to obtain corresponding prediction results; the discrete trajectory points and the prediction results are fitted into an initial curve that conforms to the trajectory traveling direction using a curve fitting method, where the curve fitting method can include linear regression, polynomial regression, etc.; after completing the initial curve fitting, since the fitted curve is a two-dimensional geometry with altitude information, it is necessary to project the initial curve into a preset coordinate system, where the preset coordinate system can be a geographic spherical coordinate system; the trajectory points are generated according to the driving route, and in the preset coordinate system, a preset distance can be defined according to the extreme value of the 6m width of the common lanes in the high-precision map, and the width is expanded by 6m in the specified direction to generate a buffer area for the initial curve, that is, a trajectory surface.

[0084] Step A4: Generate a truth set using the truth data after coordinate conversion.

[0085] In the embodiment of the present application, the specific process of step A4 includes: obtaining the second longitude and latitude information from the truth data after coordinate conversion; generating truth points based on the second longitude and latitude information, and calculating the Euclidean distance of the truth points; using the Euclidean distance to calculate the buffer area of the truth points, and merging the buffer areas corresponding to each truth point to obtain a truth set.

[0086] In the embodiment of the present application, the truth data after coordinate conversion contains all the longitude and latitude information associated with the truth data, that is, the second longitude and latitude information. The second longitude and latitude information is obtained from the truth data, and a series of truth points can be generated based on the second longitude and latitude information; based on the generated truth points, the Euclidean distance of each truth point in the same preset coordinate system as the vector data can be calculated, that is, the 6m Euclidean distance of each truth point in the spherical coordinate system. The Euclidean distance is a common method for measuring the straight-line distance between two points and can be calculated according to the point coordinates and the distance formula in the spherical coordinate system; using the calculated Euclidean distance, a boundary line is generated around each truth point, and the boundary line is extended outward by a certain distance to form a buffer surface; the buffer pads of each truth point are merged, overlapping parts are removed, and smoothing processing is performed to obtain a truth set in a certain area.

[0087] Step A5: Extract the contour information of the semantic object based on the object data after coordinate conversion, and fuse the contour information of multiple semantic objects to obtain a semantic object set.

[0088] In the embodiments of the present application, the specific process of generating the semantic object set may include: First, classify the object data according to the actual semantic attributes, where the actual semantic attributes may include points, lines, and planes, which can form road surface markings geometry, lane lines, road edges, roadside signs, traffic lights, etc., and extract the color, object subtype, and recognition confidence of the semantic object and assign them to the corresponding contour information; Second, establish a vector geometry library to store the contour information of the semantic object, and generate corresponding multiple buffers based on the contour information of the semantic object; Merge and smooth the buffers to obtain the semantic object set.

[0089] It should be noted that generating corresponding multiple buffers based on the contour information of the semantic object specifically includes: For linear semantic objects (such as lane lines), project them in the spherical coordinate system, calculate the 6m Euclidean distance of each semantic object, and perform buffer calculation, where the buffer calculation is to consider the possible width and irregularity of the linear object; For planar semantic objects (such as road edges), obtain the border points on its periphery and expand 6m outward based on the border points to calculate the buffer; For point semantic objects (such as traffic lights), expand a 6m buffer circular area centered on this point.

[0090] Step A6, construct map feature information based on the trajectory surface, the ground truth set, and the semantic object set.

[0091] In the embodiments of the present application, construct map feature information based on the trajectory surface, the ground truth set, and the semantic object set obtained from the foregoing steps A3 - A5.

[0092] The method provided by the embodiments of the present application can more accurately extract map feature information, such as the trajectory surface, the semantic object set, etc., by combining vector data and ground truth data, and can more accurately reflect the actual situation of vehicle driving, enhancing the readability and comprehensibility of map data; By constructing map feature information such as the trajectory surface and the semantic object set, the map data can be made more intuitive and can provide richer map information.

[0093] Step S13, obtain the region of interest corresponding to the vehicle, and determine the target semantic object in the semantic object set that matches the region of interest.

[0094] In the embodiments of the present application, the region of interest is a preference range or a specific scenario area marked by the user in the map feature information based on their own preferences. The vehicle system sequentially identifies and records the feature points marked by the user following the marking operation input by the user. After the user's marking is completed, a polygon is generated based on the feature points preferred by the user as the region of interest.

[0095] In the embodiment of the present application, for each semantic object in the semantic object set, the spatial relationship algorithm or geometric calculation method can be used to determine whether each semantic object intersects with the region of interest one by one, and the corresponding judgment result is obtained: if the semantic object intersects with the region of interest, it is considered that the semantic object is within the region of interest; if the semantic object does not intersect with the region of interest, it is considered that the semantic object is not within the region of interest; according to the result of the intersection judgment between the semantic object and the region of interest, the semantic object set is processed: if the semantic object is within the region of interest, the semantic object is retained in the original data; if the semantic object is not within the region of interest, the semantic object is removed from the original data; the semantic objects retained in the processed semantic object set are used as target semantic objects.

[0096] The method provided by the embodiment of the present application can filter according to the actual driving range of the vehicle by obtaining the region of interest of the vehicle, avoid filtering out key information related to the vehicle driving, and improve the pertinence and efficiency of filtering; by determining the target semantic object matching the region of interest of the vehicle, personalized map data can be provided according to the driving requirements and points of interest of different vehicles.

[0097] Step S14, filter the truth value set by using the target semantic object to obtain the filtered truth value set, and mark the filtered truth value set based on the trajectory plane to generate the target map data.

[0098] In the embodiment of the present application, a semantic space tree is constructed by using the target semantic object; the semantic space tree is traversed to obtain the associated semantic objects of each trajectory point in the trajectory plane; the spatial point set of the associated semantic object is indexed in the trajectory plane based on the spatial position of the associated semantic object; the first head-up distance, the second head-up distance, and the third head-up distance are calculated based on the spatial point set; the semantic visual distance is calculated by using the first head-up distance, the first head-up distance, the second head-up distance, and the third head-up distance; the associated semantic objects are marked according to the semantic visual distance to obtain the updated semantic object set; the truth value data in the truth value set is filtered by using the updated semantic object set to obtain the filtered truth value set; the filtered truth value set is marked based on the trajectory plane to generate the target map data.

[0099] In the embodiment of the present application, marking the filtered truth value set based on the trajectory plane to generate the target map data specifically includes: when there is an intersection between the truth value data in the truth value set and the trajectory data in the trajectory plane, assign the traj_mark attribute to 1 and mark that the truth value data is within the visible range of 6m of the driving trajectory; when there is no intersection between the truth value data in the truth value set and the trajectory data in the trajectory plane, assign the traj_mark attribute to 0 and mark that the truth value data is not within the visible range of 6m of the driving trajectory. After the assignment and marking process of the truth value set is completed, the target map data is generated.

[0100] In the embodiments of the present application, the true value set is filtered by using the target semantic object to obtain the filtered true value set, including the following steps B1 - B4:

[0101] Step B1, construct a semantic space tree by using the target semantic object.

[0102] In the embodiments of the present application, the spatial tree (R - tree) is a tree - like data structure used for storing spatial data. The target semantic objects retained in the processed semantic object set are used as keys through their minimum bounding rectangles to construct a spatial index, and the target semantic objects are written into the spatial tree based on the spatial index to obtain the semantic space tree.

[0103] Step B2, traverse the semantic space tree to obtain the associated semantic objects of each trajectory point in the trajectory plane.

[0104] In the embodiments of the present application, the trajectory points in the trajectory plane are used to traverse multiple semantic objects in the semantic space tree to find the semantic objects matched by each trajectory point, that is, to find the intersection of multiple trajectory points in the trajectory plane and multiple semantic objects in the semantic space tree, and the semantic objects in the intersection are the associated semantic objects of each trajectory point.

[0105] Step B3, update the semantic object set based on the associated semantic objects to obtain the updated semantic object set.

[0106] In the embodiments of the present application, the process of updating the semantic object set includes: indexing the spatial point set of the associated semantic object in the trajectory plane based on the spatial position of the associated semantic object, where the spatial point set includes: the first spatial point, the second spatial point, the third spatial point, and the fourth spatial point; calculating the first horizontal distance from the first spatial point to the second spatial point, the second horizontal distance from the first spatial point to the third spatial point, and the third horizontal distance from the first spatial point to the fourth spatial point; calculating the semantic visual distance by using the first horizontal distance, the second horizontal distance, and the third horizontal distance; marking the associated semantic objects according to the semantic visual distance to obtain the updated semantic object set.

[0107] In the embodiments of the present application, the specific process of step B3 includes:

[0108] Step B31, index the spatial point set of the associated semantic object in the trajectory plane based on the spatial position of the associated semantic object, where the spatial point set includes: the first spatial point, the second spatial point, the third spatial point, and the fourth spatial point. The first spatial point is the spatial point closest to the associated semantic object, the second spatial point is located in the first azimuth of the first spatial point, the third spatial point is located in the second azimuth of the first spatial point, the fourth spatial point is the spatial point closest to the first spatial point, and the second azimuth is opposite to the first azimuth.

[0109] In an embodiment of the present application, a first spatial point, i.e., a trajectory point p that is the closest to the spatial position of the associated semantic object in the trajectory plane; a second spatial point, i.e., a trajectory point p1 that is in front of the trajectory point p in the trajectory plane; a third trajectory point, i.e., a trajectory point p2 that is behind the trajectory point p in the trajectory plane; a fourth trajectory point, i.e., a trajectory point p3 that is the closest to the trajectory point p in the trajectory plane, where the trajectory points p1 and p2 are respectively located in opposite directions of the trajectory point p.

[0110] Step B32, calculate a first horizontal distance from the first spatial point to the second spatial point, a second horizontal distance from the first spatial point to the third spatial point, and a third horizontal distance from the first spatial point to the fourth spatial point.

[0111] In an embodiment of the present application, calculating the first horizontal distance from the first spatial point to the second spatial point means calculating the distance d1 between p and p1; calculating the second horizontal distance from the first spatial point to the third spatial point means calculating the distance d2 between p and p2; calculating the third horizontal distance from the first spatial point to the fourth spatial point means calculating the distance d3 between p and p3. The calculation formula for the horizontal distance is as follows:

[0112]

[0113] where d is the horizontal distance, Δh is the height difference between two spatial points, and r is the radius of the earth, i.e., r = 6371 km.

[0114] Step B33, calculate the semantic visible distance using the first horizontal distance, the second horizontal distance, and the third horizontal distance.

[0115] In an embodiment of the present application, calculating the semantic visible distance using the first horizontal distance, the second horizontal distance, and the third horizontal distance means calculating the semantic visible distance L using d1, d2, and d3. Among them, calculating the semantic visible distance can be done by taking the driving front view angle of 83.743633 degrees and calculating the ratio of the horizontal distance to the cosine of the front view angle. The calculation formula for the semantic visible distance is as follows:

[0116]

[0117] where L is the semantic visible distance, d is the horizontal distance, is the driving front view angle, i.e.,

[0118] Step B34, mark the associated semantic object according to the semantic visible distance to obtain an updated semantic object set.

[0119] In the embodiment of the present application, the strategy of marking the associated semantic objects according to the semantic visibility is: when the semantic visibility of the semantic object is less than 6m, the traj_mark attribute is assigned to 1, and it is marked that there is a track within the semantic 6m visibility range; when the semantic visibility of the semantic object is greater than or equal to 6m, the traj_mark attribute is assigned to 0, and it is marked that there is no track within the semantic 6m visibility range. After the assignment and marking are completed, the updated semantic object set is obtained.

[0120] Step B4: Filter the truth value data in the truth value set using the updated semantic object set to obtain a filtered truth value set.

[0121] In an embodiment of the present application, the updated semantic object set is used to filter in the truth value set, and the data matching the updated semantic objects is retained in the truth value set, while the data not matching the updated semantic objects is eliminated. In this way, the intersection of the updated semantic object set and the truth value set is obtained, which is the filtered truth value set.

[0122] The method provided in the embodiment of the present application can improve the accuracy and reliability of map data by filtering a truth value set using a target semantic object and marking the filtered truth value set based on a trajectory surface to generate target map data.

[0123] In the embodiment of the present application, after generating the corresponding target map data, the method further includes the following steps C1-C2:

[0124] Step C1, obtaining preset attribute filtering rules.

[0125] In an embodiment of the present application, in addition to feature points and longitude and latitude information, the target map data also includes attributes such as color, confidence, and object subtype. After the above filtering process is completed, it is necessary to filter the target map data after processing in the previous step according to the pre-set attribute filtering rules, wherein the attribute filtering rules specifically include: filtering the vector data in the target map data according to the color value, that is, eliminating data other than the common road edge line color, wherein the common road edge line color is often white and yellow; filtering the target map data according to the confidence, that is, for data with too low confidence, it is eliminated as dirty data or unreliable data; filtering the target map data according to the object subtype, that is, according to the subtype, the edge line (single solid line / double solid line / single dashed line, etc.), arrow (left turn / right turn / straight go / U-turn, etc.), sign (speed bump / stop line / zebra crossing) and other data are screened in the target map data, and other unnecessary data are eliminated.

[0126] Step C2: Eliminate abnormal attribute data in the target map data based on the attribute filtering rules.

[0127] In the embodiments of the present application, color anomaly data other than the common road edge colors, untrusted data with too low confidence, and shape anomaly data are removed from the target map data based on the attribute filtering rules.

[0128] The method provided by the embodiments of the present application can clear the errors, inaccuracies or anomaly information in the map data by filtering out the data with attribute anomalies, improving the quality and accuracy of the data; and can screen out the data related to specific tasks or applications through the filtering rules, thereby reducing the complexity and redundancy of the map data.

[0129] Figure 2 is a flowchart of generating a trajectory surface and a semantic object set according to an embodiment of the present invention, as Figure 2 shown. Steps 3-8 are the process of generating the trajectory surface, and steps 9-14 are the process of generating the semantic object set. The method includes:

[0130] Step 1, parse the vector data to obtain the trajectory data and the object data of multiple semantic objects;

[0131] Step 2, perform longitude and latitude coordinate conversion on the trajectory data and the object data;

[0132] Step 3, obtain the first longitude and latitude information of the trajectory data;

[0133] Step 4, generate trajectory points based on the first longitude and latitude information;

[0134] Step 5, fit the trajectory points into an initial curve;

[0135] Step 6, project the initial curve onto a preset coordinate system;

[0136] Step 7, expand a preset distance in a specified direction;

[0137] Step 8, generate a trajectory surface;

[0138] Step 9, classify the object data according to the actual semantic attributes;

[0139] Step 10, extract the attributes of the semantic objects and assign them to the corresponding contour information;

[0140] Step 11, establish a vector geometry library to store the contour information of the semantic objects

[0141] Step 12, generate multiple buffers corresponding to multiple semantic objects;

[0142] Step 13, perform merging and smoothing processing on the buffers;

[0143] Step 14, generate a semantic object set.

[0144] Figure 3is a flowchart for generating a ground truth set according to an embodiment of the present invention. As Figure 3 shown, the method includes:

[0145] Step 15: Convert the longitude and latitude coordinates of the ground truth data;

[0146] Step 16: Obtain the second longitude and latitude information of the ground truth data and generate ground truth points;

[0147] Step 17: Calculate the Euclidean distance of the ground truth points;

[0148] Step 18: Calculate the buffer of each ground truth point;

[0149] Step 19: Merge and smooth the buffer;

[0150] Step 20: Generate a ground truth set.

[0151] Figure 4 is a flowchart for filtering map data according to an embodiment of the present invention. As Figure 4 shown, the method includes:

[0152] Step 21: Obtain the semantic object set related to the region of interest in the region of interest determined by the vehicle;

[0153] Step 22: Target semantic objects that match;

[0154] Step 23: Construct a semantic space tree using the target semantic objects;

[0155] Step 24: Traverse the semantic space tree to obtain the associated semantic objects of each trajectory point in the trajectory plane;

[0156] Step 25: Index the first spatial point, second spatial point, third spatial point, and fourth spatial point of the associated semantic object;

[0157] Step 26: Calculate the first eye-level distance, second eye-level distance, and third eye-level distance;

[0158] Step 27: Calculate the semantic visibility distance;

[0159] Step 28: Mark the associated semantic objects to obtain an updated semantic object set;

[0160] Step 29: Filter the ground truth data in the ground truth set to obtain a filtered ground truth set;

[0161] Step 30: Mark the filtered ground truth set based on the trajectory plane;

[0162] Step 31: Generate target map data.

[0163] This embodiment provides a map data filtering device. AsFigure 5 As shown in

[0164] The first acquisition module 51 is configured to acquire the driving trajectory during the current driving of the vehicle, and acquire the vector data associated with the driving trajectory from the vector map and the ground truth data associated with the driving trajectory from the high-precision ground truth map;

[0165] The construction module 52 is configured to construct map feature information corresponding to the driving trajectory by using the vector data and the ground truth data, wherein the map feature information includes: a trajectory surface, a semantic object set, and a ground truth set;

[0166] The second acquisition module 53 is configured to acquire the region of interest corresponding to the vehicle, and determine the target semantic object in the semantic object set that matches the region of interest;

[0167] The filtering module 54 is configured to filter the ground truth set by using the target semantic object to obtain a filtered ground truth set, and mark the filtered ground truth set based on the trajectory surface to generate target map data.

[0168] In an embodiment of the present application, the device further includes: a rejection module, configured to acquire a preset attribute filtering rule; and reject the attribute abnormal data in the target map data based on the attribute filtering rule.

[0169] In an embodiment of the present application, the construction module 52 is configured to parse the vector data to obtain trajectory data and object data of multiple semantic objects; perform longitude and latitude coordinate conversion on the trajectory data, the object data, and the ground truth data; generate a trajectory surface by using the trajectory data after coordinate conversion; generate a ground truth set by using the ground truth data after coordinate conversion; extract the contour information of the semantic object based on the object data after coordinate conversion, and fuse the contour information of the multiple semantic objects to obtain a semantic object set; construct map feature information based on the trajectory surface, the ground truth set, and the semantic object set.

[0170] In an embodiment of the present application, the construction module 52 is configured to obtain first longitude and latitude information from the trajectory data after coordinate conversion; generate trajectory points based on the first longitude and latitude information, and fit the trajectory points into an initial curve; project the initial curve onto a preset coordinate system, and expand a preset distance in a specified direction within the preset coordinate system to generate a trajectory surface.

[0171] In an embodiment of the present application, the construction module 52 is configured to obtain second longitude and latitude information from the ground truth data after coordinate conversion; generate ground truth points based on the second longitude and latitude information, and calculate the Euclidean distance of the ground truth points; use the Euclidean distance to calculate the buffer zone of the ground truth points, and merge the buffer zones corresponding to each ground truth point to obtain a ground truth set.

[0172] In an embodiment of the present application, a filtering module 54 is configured to construct a semantic space tree by using target semantic objects; traverse the semantic space tree to obtain associated semantic objects of each trajectory point in a trajectory plane; update a semantic object set based on the associated semantic objects to obtain an updated semantic object set; and filter true value data in a true value set by using the updated semantic object set to obtain a filtered true value set.

[0173] In an embodiment of the present application, the filtering module 54 is configured to index a set of spatial points of an associated semantic object in a trajectory plane based on the spatial position of the associated semantic object, where the set of spatial points includes: a first spatial point, a second spatial point, a third spatial point, and a fourth spatial point. The first spatial point is the spatial point closest to the associated semantic object. The second spatial point is located in a first direction of the first spatial point. The third spatial point is located in a second direction of the first spatial point. The fourth spatial point is the spatial point closest to the first spatial point, and the second direction is opposite to the first direction; calculate a first horizontal viewing distance from the first spatial point to the second spatial point, a second horizontal viewing distance from the first spatial point to the third spatial point, and a third horizontal viewing distance from the first spatial point to the fourth spatial point; calculate a semantic viewing distance by using the first horizontal viewing distance, the second horizontal viewing distance, and the third horizontal viewing distance; and mark the associated semantic object according to the semantic viewing distance to obtain an updated semantic object set.

[0174] Please refer to Figure 6 , Figure 6 which is a schematic structural diagram of an electronic device provided by an optional embodiment of the present invention. As Figure 6 shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other through different buses and can be installed on a common main board or installed in other ways as needed. The processor can process instructions executed within the electronic device, including instructions stored in the memory or on the memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (such as a server array, a set of blade servers, or a multi-processor system).

[0175] The processor 10 can be a central processing unit, a network processor, or a combination thereof. Among them, the processor 10 can further include a hardware chip. The above hardware chip can be an application-specific integrated circuit, a programmable logic device, or a combination thereof. The above programmable logic device can be a complex programmable logic device, a field programmable gate array, a general array logic, or any combination thereof.

[0176] Among them, the memory 20 stores instructions that can be executed by at least one processor 10, so that the at least one processor 10 executes the methods shown in the above embodiments.

[0177] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of an electronic device presented by a kind of landing page of a small program, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories can be connected to the electronic device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0178] The memory 20 may include a volatile memory, for example, a random access memory; the memory may also include a non-volatile memory, for example, a flash memory, a hard disk, or a solid-state drive; the memory 20 may further include a combination of the above types of memories.

[0179] The electronic device further includes a communication interface 30 for the electronic device to communicate with other devices or communication networks.

[0180] The embodiments of the present invention further provide a computer-readable storage medium. The methods according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code originally stored in a remote storage medium or a non-temporary machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the methods described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium may further include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0181] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for filtering map data, characterized in that, Including: Obtain the driving trajectory during the current driving of the vehicle, obtain the vector data associated with the driving trajectory from the vector map, and obtain the ground truth data associated with the driving trajectory from the high-precision ground truth map; Construct map feature information corresponding to the driving trajectory by using the vector data and the ground truth data, wherein the map feature information includes: a trajectory surface, a set of semantic objects, and a set of ground truths; Obtain the region of interest corresponding to the vehicle, and determine the target semantic object in the set of semantic objects that matches the region of interest; Filter the set of ground truths by using the target semantic object to obtain a filtered set of ground truths, and mark the filtered set of ground truths based on the trajectory surface to generate target map data.

2. The method according to claim 1, wherein The constructing the map feature information corresponding to the driving trajectory by using the vector data and the ground truth data includes: Parse the vector data to obtain trajectory data and object data of multiple semantic objects; Perform longitude and latitude coordinate conversion on the trajectory data, the object data, and the ground truth data; Generate a trajectory surface by using the trajectory data after coordinate conversion; Generate a set of ground truths by using the ground truth data after coordinate conversion; Extract contour information of the semantic objects based on the object data after coordinate conversion, and fuse the contour information of multiple semantic objects to obtain the set of semantic objects; Construct map feature information based on the trajectory surface, the set of ground truths, and the set of semantic objects.

3. The method according to claim 2, wherein The generating the trajectory surface by using the trajectory data after coordinate conversion includes: Obtain first longitude and latitude information from the trajectory data after coordinate conversion; Generate trajectory points based on the first longitude and latitude information, and fit the trajectory points into an initial curve; Project the initial curve onto a preset coordinate system, and expand a preset distance in a specified direction within the preset coordinate system to generate the trajectory surface.

4. The method according to claim 2, wherein The generating the set of ground truths by using the ground truth data after coordinate conversion includes: Obtain second longitude and latitude information from the ground truth data after coordinate conversion; Generate ground truth points based on the second longitude and latitude information, and calculate the Euclidean distance of the ground truth points; Calculate the buffer of the ground truth points by using the Euclidean distance, and merge the buffers corresponding to each ground truth point to obtain the set of ground truths.

5. The method according to claim 1, wherein The filtering the set of ground truths by using the target semantic object to obtain a filtered set of ground truths includes: Construct a semantic space tree by using the target semantic object; Traverse the semantic space tree to obtain the associated semantic objects of each trajectory point in the trajectory surface; Update the set of semantic objects based on the associated semantic objects to obtain an updated set of semantic objects; Filter the ground truth data in the set of ground truths by using the updated set of semantic objects to obtain a filtered set of ground truths.

6. The method according to claim 5, wherein The updating the set of semantic objects based on the associated semantic objects to obtain an updated set of semantic objects includes: Indexing a set of spatial points of the associated semantic object in the trajectory surface based on the spatial position of the associated semantic object, wherein the set of spatial points includes: a first spatial point, a second spatial point, a third spatial point, and a fourth spatial point, the first spatial point is a spatial point closest to the associated semantic object, the second spatial point is located at a first orientation of the first spatial point, the third spatial point is located at a second orientation of the first spatial point, the fourth spatial point is a spatial point closest to the first spatial point, and the second orientation is opposite to the first orientation; Calculate a first horizontal distance from the first spatial point to the second spatial point, a second horizontal distance from the first spatial point to the third spatial point, and a third horizontal distance from the first spatial point to the fourth spatial point; Calculating semantic visible distance using the first horizontal distance, the second horizontal distance, and the third horizontal distance; The associated semantic objects are marked according to the semantic visibility to obtain an updated semantic object set.

7. The method according to claim 1, characterized in that, After generating the corresponding target map data, the method further includes: Get preset attribute filtering rules; Abnormal attribute data in the target map data is eliminated based on the attribute filtering rule.

8. A filtering device for map data, characterized in that, The device comprises: A first acquisition module is used to acquire a driving trajectory of the vehicle during the current driving process, and to acquire vector data associated with the driving trajectory from a vector map and to acquire true value data associated with the driving trajectory from a high-precision true value map; A construction module, used to construct map feature information corresponding to the driving track using the vector data and the true value data, wherein the map feature information includes: a track surface, a semantic object set and a true value set; A second acquisition module is used to acquire the region of interest corresponding to the vehicle and determine a target semantic object in the semantic object set that matches the region of interest; A filtering module is used to filter the truth value set using the target semantic object to obtain a filtered truth value set, and mark the filtered truth value set based on the trajectory surface to generate target map data.

9. An electronic device, characterized in that, include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images