Map fusion methods, devices, electronic devices and storage media

Abstract

This application provides a map fusion method, apparatus, electronic device, and storage medium. The map fusion method includes: acquiring a first map element vector and pose of a target road, and a second map element vector of the target road; correcting the pose represented by a graph based on a pre-generated ground model of the target road, the first map element vector, and the second map element vector; and fusing the first map element vector and the second map element vector of the target road according to the corrected pose. In the solution of this application embodiment, since the ground model of the target road better reflects the actual ground condition, correcting the pose based on the ground model, the first map element vector, and the second map element vector can improve the accuracy of pose correction, thereby further improving the accuracy of map fusion based on the corrected pose.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a map fusion method, apparatus, electronic device, and storage medium. Background Technology

[0002] High-precision maps are widely used in scenarios with high requirements for map information. They possess high accuracy and detailed defined attributes, reaching decimeter-level precision, and include various traffic elements such as road network data, lane network data, lane lines, and traffic sign data. Based on this, high-precision maps can provide safety assurance in various scenarios, such as autonomous driving. They are unaffected by inclement weather and can provide vehicles with prior knowledge, improving driving safety.

[0003] Currently, high-precision maps can be produced based on map data collected by professional surveying vehicles. However, the high cost and limited number of these vehicles result in slow map updates. Therefore, updating high-precision maps using map data collected by low-cost crowdsourced data collection equipment has become an effective way to improve the update speed; this is known as crowdsourced updating.

[0004] In their research on crowdsourced updates, the inventors of this application discovered that it is necessary to integrate road map data (hereinafter referred to as crowdsourced data) collected through crowdsourcing with existing high-precision maps of the same road. This integration ensures the accuracy of the updated map while also updating the existing high-precision map. Therefore, the integration of crowdsourced data with existing high-precision maps is a crucial step in high-precision map updates, and continuously improving and optimizing the integration technology is a problem that those skilled in the art need to address. Summary of the Invention

[0005] In view of this, embodiments of this application provide a map fusion method, apparatus, electronic device, and storage medium to solve or alleviate the above-mentioned problems.

[0006] According to a first aspect of the embodiments of this application, a map fusion method is provided, comprising: acquiring a first map element vector and pose of a target road, and a second map element vector of the target road; correcting the pose represented by a graph based on a pre-generated ground model of the target road, the first map element vector, and the second map element vector; and fusing the first map element vector and the second map element vector of the target road according to the corrected pose.

[0007] According to a second aspect of the embodiments of this application, a map fusion apparatus is provided, comprising: an acquisition module, configured to acquire a first map element vector and pose of a target road, and a second map element vector of the target road; a correction module, configured to correct the pose represented by a graph based on a pre-generated ground model of the target road, the first map element vector, and the second map element vector; and a fusion module, configured to fuse the first map element vector and the second map element vector of the target road according to the corrected pose.

[0008] According to a third aspect of the present application, an electronic device is provided, the electronic device comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, the executable instruction causing the processor to perform an operation corresponding to the map integration method described in the first aspect.

[0009] According to a fourth aspect of the embodiments of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the map integration method as described in the first aspect.

[0010] In the scheme of this application embodiment, since the ground model of the target road can better reflect the actual ground condition, the pose can be corrected based on the ground model, the first map element vector and the second map element vector, which can improve the accuracy of the pose correction. Furthermore, the first map element vector and the second map element vector of the target road can be fused according to the corrected pose, which can also ensure the accuracy of the fused map element vector. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0012] Figure 1 This is a schematic diagram of a map fusion scenario;

[0013] Figure 2A This is a schematic flowchart illustrating a map fusion method according to an embodiment of this application;

[0014] Figure 2B A schematic diagram of a ground model as an example of a map fusion method according to another embodiment of this application;

[0015] Figure 3AA schematic diagram of vector processing for an example of a map fusion method according to another embodiment of this application;

[0016] Figure 3B A schematic diagram of vector processing for another example of a map fusion method according to another embodiment of this application;

[0017] Figure 4A This is a schematic flowchart illustrating a map fusion method according to another embodiment of this application;

[0018] Figure 4B This is a schematic diagram of the pose graph before correction, which is an example of a map fusion method according to another embodiment of this application.

[0019] Figure 4C This is a schematic diagram of a pose graph correction for another example of a map fusion method according to another embodiment of this application;

[0020] Figure 5 This is a schematic block diagram of a map fusion apparatus according to another embodiment of this application;

[0021] Figure 6 The hardware structure of an electronic device is shown in another embodiment of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions in the embodiments of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art should fall within the protection scope of the embodiments of this application.

[0023] The specific implementation of the embodiments of this application will be further described below with reference to the accompanying drawings.

[0024] Figure 1 This is a schematic diagram of a map fusion scenario. (Example) Figure 1 As shown, vehicles 1, 2, and 3 are crowdsourced vehicles, also known as socialized data collection vehicles. These vehicles are equipped with corresponding data collection devices to collect relevant map data. For example, the three vehicles travel along their respective planned routes on the road, collecting relevant road environment data, such as lane line information and road sign information, while traveling. It should be understood that the three vehicles mentioned above are merely illustrative, and there could be more or fewer vehicles. The aforementioned travel routes are also illustrative; in a single data collection process, for the same travel route, the crowdsourced vehicles can perform one data collection (one trip) or multiple data collections (multiple trips). Figure 1As shown, when the real-world environment changes (e.g., new lane lines, altered lane lines, new road signs, or altered road signs), collecting relevant road environment data through crowdsourced vehicles helps to quickly update map data.

[0025] The crowdsourced vehicle illustrated in this example can be a "pre-installed crowdsourcing" method, where the vehicle is equipped with relatively professional data collection equipment to collect high-precision, high-cost, and small-volume data; or it can be a "post-installed crowdsourcing" method, where the vehicle is equipped with third-party smart rearview mirrors, dashcams, and other data collection equipment to collect low-precision, low-cost, and large-volume data.

[0026] In crowdsourced map data collection using crowdsourced vehicles, it's common for the same or different crowdsourced vehicles to collect data multiple times for the same road. Each collected road environment data point generates a single crowdsourced map element vector for that road; multiple collections result in multiple vectors. However, a single crowdsourced map element vector is generally sufficient to represent a single road. Therefore, multiple crowdsourced map element vectors for the same road need to be pre-fused. This fusion process involves three steps: localization, vector reconstruction, and vector map fusion. In the localization step, existing absolute positioning information can be used. For example, existing high-precision maps of the road can be used to calibrate the positioning of each crowdsourced map element, obtaining precise absolute positioning information. In the vector reconstruction step, single-trip vector maps corresponding to each crowdsourced map can be generated. For instance, after localization calibration using a high-precision map in the localization step, the resulting absolute positioning information is output to the reconstruction step to generate single-trip vector maps for each crowdsourced map. In the vector map fusion step, the single-pass vector maps corresponding to each crowdsourced map can be merged to obtain an updated map. For example, the single-pass vector maps corresponding to each crowdsourced map can be merged according to the recursive least squares algorithm.

[0027] It should be noted that while the above methods can achieve map updates, the collected map data itself may deviate significantly from the actual data, resulting in maps with relatively poor accuracy. Therefore, this application provides a map fusion scheme to improve the accuracy of maps obtained through fusion.

[0028] Figure 2A This is a schematic flowchart illustrating a map fusion method according to one embodiment of this application. The method of this embodiment can be executed by any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, tablets, etc.), and PCs. Optionally, the method of this embodiment is applicable to cloud servers such as private clouds, public clouds, or dedicated clouds, or other forms of backend data processing devices. Figure 2A The map fusion method shown includes:

[0029] 210: Obtain the first map element vector and pose of the target road, and the second map element vector of the target road.

[0030] It should be understood that the first map element vector and the second map element vector in this application are map element vectors of the same target road. In one feasible approach, the first map can be a crowdsourced map, and other maps can be updated based on the first map. The first map has a first map element vector and a pose. When the first map is a crowdsourced map, each crowdsourced map has a corresponding map element vector and a pose. In the embodiments of this application, the map to be updated based on the first map is referred to as the second map. In one feasible approach, the second map can be an existing high-precision map (also called a base map). This high-precision map can be the original high-precision map, i.e., a map that has not been updated, or it can be a map that has been updated once or more (two or more times).

[0031] Furthermore, the first map in this embodiment can be a single map or multiple maps. When the first map is implemented as a crowdsourced map, the crowdsourced map can be a single-trip vector map or a multi-trip vector map. During map updates, the first map and the second map can be merged, such as merging the crowdsourced map with a high-precision map each time to generate a new map.

[0032] In this embodiment of the application, both the first map and the second map have map element vectors. The map element vectors can be vectors that indicate any map element, such as lane line vectors, street sign vectors of light poles, etc. This embodiment of the application does not limit this.

[0033] Furthermore, during the map data collection process using crowdsourced vehicles, the image acquisition devices mounted on these vehicles possess corresponding poses, i.e., the position and orientation of the image acquisition devices, such as their spatial location and orientation. Typically, this pose can be represented by the rotation and translation data of the image acquisition devices. In this embodiment, the pose can be recorded when acquiring the image of the first map element vector, thus the acquired first map element vector corresponds to specific pose data. Furthermore, a pose graph can be generated based on the pose data. With the aid of pose data or pose graphs, more accurate map data can be obtained.

[0034] It should be noted that, in the specific implementation of this step, the execution of obtaining the first map element vector and pose of the target road, and obtaining the second map element vector of the target road, can be performed in any order or in parallel.

[0035] For example, you can first obtain the second map element vector, and then obtain the first map element vector that matches the second map element vector. Alternatively, you can first obtain the first map element vector, and then obtain the second map element vector. For example, you can select the second map element vector that matches the first map element vector from among multiple first map element vectors in the second map.

[0036] 220: Based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector, the pose represented by the graph is corrected.

[0037] A terrain model, also known as a digital terrain model, is generated from a series of data representing the spatial distribution of terrain features. A terrain model can be represented as a plane or surface in a world coordinate system, for example... Figure 2B This is a schematic diagram of an example ground model in one embodiment of this application. Because in real-world scenarios, perfectly flat roads do not exist. Figure 2B The curved surface shown illustrates one type of non-flat road, represented by vectors A, B, and C, which represent a section of the road's surface. It should be noted that... Figure 2B The height or degree of curvature data shown are exemplary and should not be construed as limiting the scope of protection of this application. Furthermore, Figure 2B The projections of vectors A, B, and C onto the XOY plane in the world coordinate system can be different from their projections onto the ground model. The projections of vectors A, B, and C onto the ground model vary depending on factors such as road slope. Using a ground model avoids map scale inconsistencies caused by inaccurate height scaling and inaccurate roll angles.

[0038] When correcting pose based on the ground model, the first map element vector, and the second map element vector, taking a high-precision map as an example, if the second map element vector adopts the lane line vector of the high-precision map or the lane line vector in the map updated based on the lane line of the high-precision map, then: on the one hand, because the ground model data and the high-precision map data are both relatively accurate, the aforementioned projection vector and the aforementioned lane line vector are more closely matched with the ground model; on the other hand, the aforementioned projection vector is also more closely matched with the real physical scene, thus making the pose correction effect better.

[0039] It should be understood that the data for the ground model can be acquired externally, for example, collected using ground mapping equipment or equipment with ground mapping capabilities, and uploaded to a server via a network such as the Internet. In another example, the data for the ground model can be generated based on a high-precision map.

[0040] It should also be understood that pose represented by a graph can be called a pose graph, which can represent the position vectors and attitude matrices of poses across multiple frames. Corrections to poses represented by a graph can be made using methods such as constructing pose constraints. For example, based on a ground model of the target road, the first map element vector and the second map element vector can be matched; when the matching degree meets a threshold, the pose is corrected. The aforementioned pose constraints can be constraints on multi-frame poses (multi-frame camera poses or multi-frame acquisition device poses). Pose constraints can be non-linear constraints and can include at least one of distance constraints (e.g., point-to-line distance constraints), relative pose constraints, and absolute pose constraints. Pose graph correction is a backend optimization method for the pose graph. Since the pose is associated with the corresponding first map element vector, correcting the pose can achieve the correction of the first map element vector, improving the accuracy of the first map element vector.

[0041] 230: Based on the corrected pose, fuse the first map element vector and the second map element vector of the target road.

[0042] Taking a first map as a crowdsourced map and a second map as a high-precision map as an example, multiple crowdsourced map element vectors can be obtained based on the corrected pose map. These multiple crowdsourced map element vectors are then fused with the corresponding map element vectors in the high-precision map using methods such as weighted fusion. Alternatively, in another example, multiple crowdsourced map element vectors that do not correspond to vectors in the high-precision map can be obtained based on the corrected pose map (e.g., no corresponding crowdsourced map element exists, or there are new features in the map). These multiple crowdsourced map element vectors are added to the high-precision map for integration. In the specific map fusion process, a new map can be generated using at least one of the following methods: horizontal and vertical calibration, grid division, and vector fusion. When multiple crowdsourced maps exist, each crowdsourced map can be fused with the high-precision map, or some crowdsourced maps can be fused first, and then the fused crowdsourced map can be fused with the high-precision map. This generates a new map.

[0043] In the scheme of this application embodiment, since the ground model of the target road can better reflect the actual ground condition, the pose can be corrected based on the ground model, the first map element vector and the second map element vector, which can improve the accuracy of the pose correction. Furthermore, the first map element vector and the second map element vector of the target road can be fused according to the corrected pose, which can also ensure the accuracy of the fused map element vector.

[0044] In other words, in the embodiments of this application, vectors can be associated with poses, and vectors can be corrected during the pose graph correction process, thereby improving the absolute accuracy of the fused vectors.

[0045] In one implementation of this application, the pose diagram corresponding to the first map can be corrected by the pose constraint between the first map element vector and the second map element vector. In some cases, the second map element vector may include high-precision map element vectors and non-high-precision map element vectors, where the non-high-precision map element vectors can be corresponding map element vectors in a map updated based on the high-precision map element vectors. Therefore, correcting the pose diagram corresponding to the first map by the pose constraint between the first and second map element vectors can include: performing a first correction on the pose diagram corresponding to the first map, such as a crowdsourcing map, by using a first pose constraint between the first map element vector and the high-precision map element vectors; and performing a second correction on the pose diagram corresponding to the first map, such as a crowdsourcing map, by using a second pose constraint between the first map element vector and the non-high-precision map element vectors. Optionally, when the angle and / or distance between the high-precision map element vector and the non-high-precision map element vector is greater than a preset threshold, the result of the second correction is excluded, and the corrected pose diagram is determined based on the result of the first correction. Alternatively, when the angle and / or distance between the high-precision map element vector and the non-high-precision map element vector is less than a preset threshold, a corrected pose map is determined based on the results of the first and second corrections. For example, a weighted vector is obtained by weighting the high-precision map element vector and the non-high-precision map element vector, and the corrected pose map is determined based on the weighted vector. The preset threshold can be appropriately set by those skilled in the art according to actual needs, and this embodiment does not impose any limitations on it.

[0046] In another implementation of the application embodiment, when the first map element vector is a crowdsourced map element vector and the second map element vector is a high-precision map element vector, the pose represented by graph is corrected based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector, including: determining the projection vector of the crowdsourced map element vector of the target road onto the ground model of the target road; and correcting the pose represented by graph based on the pose constraints between the projection vector and the high-precision map element vector.

[0047] In this example, since the projection vector reflects the matching relationship with the high-precision map element vector better than the crowdsourced map element vector, and the projection vector has a smaller data computation amount, the pose represented by the graph is corrected based on the pose constraints between the projection vector and the high-precision map element vector, which improves the data processing efficiency while ensuring the accuracy of the pose correction.

[0048] Alternatively, in the solution of this application embodiment, the high-precision map can be placed on the server, eliminating the need for real-time transmission of the high-precision map, and it is only used when the map is fused, thus saving data processing costs and data transmission costs.

[0049] The following is an illustrative description of the process of correcting the pose represented by graphs based on the pose constraints between the crowdsourced map element vectors used to determine the target road and the ground model of the target road.

[0050] As a first example, the pose constraint includes point-line distance constraint. Then, the step of correcting the pose represented by graph based on the pose constraint between the projection vector and the high-precision map element vector includes: correcting the pose represented by graph by minimizing the value of the expression of the point-line distance constraint between the projection vector and the high-precision map element vector, so as to eliminate the matching error between the crowdsourced map element vector and the high-precision map element vector.

[0051] Specifically, the expression for the point-line distance constraint includes the sum of the squares of multiple point-line distances between the projection vector and the high-precision map element vector. Based on this, the process of minimizing the value of the expression for the point-line distance constraint between the projection vector and the high-precision map element vector includes minimizing the value of the sum of the squares of multiple point-line distances.

[0052] As a concrete example, the expression for the point-to-line distance constraint can be represented as follows:

[0053]

[0054] in,

[0055] in, p represents the projection of each point on the crowdsourced map element vector onto the ground model; p is a point on the crowdsourced map element vector. This represents the displacement in the pose associated with the map element vector; For any two points on the vector of the matched high-precision map element. The ground model is {p r ,d r}; where p r The plane passes through the point d. r The direction of the plane normal; The pose to be corrected; r i (z d ,χ) represents the point-to-line distance constraint.

[0056] As a second example, the pose constraint includes a linear expression consisting of the point-to-line distance constraint between the projection vector and the high-precision map element vector, and the absolute pose constraint of the high-precision map element vector and its multi-frame pose. Based on this, the correction of the pose represented by graphs based on the pose constraint between the projection vector and the high-precision map element vector includes: correcting the pose represented by graphs by minimizing the value of the linear expression. This eliminates the matching error between the crowdsourced map element vector and the high-precision map element vector.

[0057] As a concrete example, this linear expression can be represented as follows:

[0058]

[0059] in, and in, p represents the projection of each point on the crowdsourced map element vector onto the ground model; p is a point on the crowdsourced map element vector. This represents the displacement in the pose associated with the map element vector; For any two points on the vector of the matched high-precision map element. The ground model is {p r ,d r}; where p r The plane passes through the point d. r The direction of the plane normal; The pose to be corrected; r i (z d ,χ) represents the point-to-line distance constraint; r i (z p ,χ) represents the absolute pose constraint.

[0060] As a third example, the pose constraint includes a linear expression consisting of point-to-line distance constraints between the projected vector and the high-precision map element vector, and relative pose constraints between the multi-frame poses of the high-precision map element vector. Based on the pose constraints between the projected vector and the high-precision map element vector, the pose represented by the graph is corrected by minimizing the value of the linear expression. This eliminates the matching error between the crowdsourced map element vector and the high-precision map element vector.

[0061] Specifically, in this linear expression, the point-to-line distance constraint has a first weight and the relative pose constraint has a second weight. The method further includes adjusting the first and second weights based on the constraint state of the point-to-line distance constraint. By adjusting the weights, areas with insufficient constraints are corrected, making the pose correction result more accurate.

[0062] Specifically, based on the constraint state of the point-to-line distance constraint, the first and second weights are adjusted, including: in the target area of ​​the target road, if the crowdsourced map lane line vectors are distributed on one side of the pose, the constraint state of the point-to-line distance constraint is determined to be insufficient, and the first weight is decreased and / or the second weight is increased. The magnitude or step size of the decrease or increase can be appropriately set by those skilled in the art according to the actual situation, as long as it can quickly and accurately correct areas of insufficient constraint. By determining the distance constraint as insufficient through the distribution of the crowdsourced map lane line vectors on one side of their corresponding multi-frame poses, a simplified constraint state determination is achieved, improving the data processing efficiency of pose or pose graph correction.

[0063] It should be understood that insufficient constraint can be determined by the distribution of the pose. For lane lines, the constraint is relatively good when the pose associated with the lane line vector has constraints on both sides, i.e., lane line vector constraints exist on both the left and right sides of the pose. By constraining the distance between the lane line vectors on both sides, a relatively accurate pose estimate can be obtained. Generally, when the pose has constraints on only one side, the constraint is insufficient.

[0064] For example, in pose graph correction, the residual can be reduced by increasing the covariance of the lane line distance constraint. Alternatively, the residual can be increased by decreasing the covariance of the relative pose constraint.

[0065] As a concrete example, the linear expression in this correction method can be represented as follows:

[0066]

[0067] in, and

[0068] in, p represents the projection of each point on the crowdsourced map element vector onto the ground model; p is a point on the crowdsourced map element vector. This represents the displacement in the pose associated with the map element vector; For any two points on the vector of the matched high-precision map element. The ground model is {p r ,d r}; where p r The plane passes through the point d. r The direction of the plane normal; The pose to be corrected; r i (z d ,χ) represents the point-to-line distance constraint; r i (z r ,χ) represents the relative pose constraint.

[0069] Where, r i (z d The weight corresponding to χ) is the first weight; r i (z d The weight corresponding to χ) is the second weight.

[0070] Furthermore, in another implementation of this application, the pose graph corresponding to the crowdsourcing map can be modified by configuring a first weight for the point-to-line distance constraint and a second weight for the relative pose constraint. For example, a weighted representation of the pose of the crowdsourcing map across multiple frames can be constructed, and the value of the pose across multiple frames when the weighted representation reaches an extreme value can be determined. The weighted representation includes a weighted sum of the point-to-line distance constraint and the relative pose constraint. The point-to-line distance constraint is the sum of the squares of the distance vectors between the target point in the first map element vector and the second map element vector in the multiple frames of pose, and the relative pose constraint is the sum of the squares of the relative pose vectors of adjacent poses in the multiple frames of pose.

[0071] Because the multi-frame poses used in constructing crowdsourced maps are weighted representations of pose constraints, it facilitates the rapid acquisition of corrected multi-frame poses. Furthermore, the use of the aforementioned concise and efficient mathematical representations for relative pose constraints and point-to-line distance constraints improves data processing efficiency for pose or pose graph correction while maintaining accuracy.

[0072] As a fourth example, the pose constraint includes a linear expression consisting of the point-line distance constraint between the projection vector and the high-precision map element vector, the absolute pose constraint between the high-precision map element vector and its multi-frame pose, and the relative pose constraint between the multi-frame poses. Based on this, the correction of the pose represented by graphs based on the pose constraint between the projection vector and the high-precision map element vector includes: correcting the pose represented by graphs by minimizing the value of the linear expression. This eliminates the matching error between the crowdsourced map element vector and the high-precision map element vector.

[0073] As a concrete example, this linear expression can be represented as follows:

[0074]

[0075] in, and and

[0076] in, p represents the projection of each point on the crowdsourced map element vector onto the ground model; p is a point on the crowdsourced map element vector. This represents the displacement in the pose associated with the map element vector; For any two points on the vector of the matched high-precision map element. The ground model is {pr ,d r}; where p r The plane passes through the point d. r The direction of the plane normal; The pose to be corrected; r i (z d ,χ) represents the point-to-line distance constraint; r i (z p ,χ) represents the absolute pose constraint; r i (z r ,χ) represents the relative pose constraint.

[0077] Furthermore, when pose constraints include point-to-line distance constraints, absolute pose constraints, and relative pose constraints, each constraint can be configured with a corresponding weight. For example, the weighted representation of pose constraints across multiple frames in a crowdsourced map can include a weighted representation of at least two of relative pose constraints, absolute pose constraints, and point-to-line distance constraints. Preferably, the weighted representation of pose constraints across multiple frames in a crowdsourced map can be a weighted representation of relative pose constraints, absolute pose constraints, and point-to-line distance constraints. For example, point-to-line distance constraints can be configured with a first weight, relative pose constraints can be configured with a second weight, and absolute pose constraints can be configured with a third weight. When adjusting the weights, the weight of one constraint can be increased while the weights of the other two are decreased, or the weights of two constraints can be increased while the weight of the other one is decreased. Then, the pose or pose graph is corrected based on the adjusted weights of each constraint. In one example, the point-to-line distance constraint can be a weighted representation of the distances from multiple points of multiple map element vectors corresponding to multiple frames of poses to the same second map element vector (as an example, a high-precision map element vector). Alternatively, the point-to-line distance constraint can be a weighted representation of the distances from the same point of the same second map element vector (as an example, a high-precision map element vector) to multiple map element vectors corresponding to multiple frames of poses.

[0078] The pose correction was effectively achieved through the methods described in the above examples.

[0079] Since each first map corresponds to a pose to be corrected, in order to improve the efficiency of pose or pose map correction, in one implementation of this application, obtaining the first map element vector of the target road includes: excluding crowdsourced map element vectors with distorted reconstruction shapes from multiple crowdsourced map element vectors of the target road to obtain the first map element vector.

[0080] Since reconstructed map element vectors with distorted shapes significantly impact pose constraints in pose graph correction, excluding these vectors when acquiring the first map element vector of the target road improves data processing efficiency before pose graph correction. In other words, it improves the efficiency of pose graph correction preprocessing, which in turn improves the efficiency of subsequent pose or pose graph correction.

[0081] It should be understood that the above determination of reconstructed shape distortion can be based on the relationship between the corresponding crowdsourced map element vector and the second map element vector. The second map element vector can be a high-precision map element vector or a map element vector previously updated based on the high-precision map element vector.

[0082] Specifically, lane line vectors in crowdsourced map element vectors are typically obtained through point cloud fitting. During the fitting process, due to uneven distribution of the point cloud or unstable pose, point clouds may overlap, causing significant curvature changes in the generated lane line vectors. Therefore, by discriminating the distribution of the reconstructed vectors, excessively curved vectors can be identified. If such vectors are used in subsequent pose or pose map corrections, they will result in incorrect pose correction results.

[0083] It should also be understood that various pose graph correction metrics can be used to identify and process map element vectors with distorted reconstructed shapes. For example, various pose graph correction metrics can be used for error vector removal. In the first example, the degree of curvature or bending of some or all of the map element vectors can be used to determine whether the map element vectors have distorted reconstructed shapes. In the second example, the curvature or radius of curvature of a specific point in the map element vectors can also be used to determine whether the map element vectors have distorted reconstructed shapes. In the third example, straight line fitting can be used to determine whether the variance of the distance between at least one point in the first map element vector and the second map element vector exceeds a threshold. If it exceeds the threshold, it is determined to be a case of distorted reconstructed shapes. In the fourth example, the variance of the distance between at least one point in the second map element vector and the first map element vector can also be determined whether it exceeds a threshold. If it exceeds the threshold, it is determined to be a case of distorted reconstructed shapes. In the fifth example, the variance of the distance between at least one point in the first map element vector and at least one point in the second map element vector can also be determined whether it exceeds a threshold. If it exceeds the threshold, it is determined to be a case of distorted reconstructed shapes. In the sixth example, fitted lines for multiple first map element vectors, such as multiple crowdsourced map element vectors, can be calculated. If the variance of the distances from multiple points on a first map element vector, such as a crowdsourced map element vector, to its fitted line is greater than a threshold, the crowdsourced map element vector is determined to have reconstructed shape distortion and is excluded. Since the variance of the distances from multiple points in a crowdsourced map element vector to its fitted line can indicate the degree of reconstructed shape distortion, and the computational cost of processing the aforementioned variances is relatively small, the data processing efficiency before pose graph correction is improved. That is, the efficiency of pose graph correction preprocessing is improved, thereby improving the efficiency of pose graph correction. In addition, the aforementioned fitted lines can be implemented in any way, such as least-two-way fitting. It should be noted that the thresholds in the above various methods can be appropriately set by those skilled in the art according to the actual situation, and the thresholds in the various methods can be the same or different.

[0084] In another implementation aimed at improving the efficiency of pose or pose graph correction, obtaining the first map element vector of the target road includes: excluding crowdsourced map element vectors from multiple crowdsourced map element vectors of the target road whose included angle with the second map element vector exceeds a threshold, thereby obtaining the first map element vector that matches the second map element vector.

[0085] Since crowdsourced map element vectors whose angle with the second map element vector exceeds a threshold have a significant impact on pose constraints in pose graph correction, excluding these vectors improves data processing efficiency before pose graph correction. In other words, it improves the efficiency of pose graph correction preprocessing, which in turn improves the efficiency of pose or pose graph correction.

[0086] It should be understood that the angles between multiple second map element vectors and multiple crowdsourced map element vectors can be determined in any way, and mismatched vector removal processing can be performed in any way. For example, in one example, multiple slopes of multiple crowdsourced map element vectors can be calculated, and the angles between these multiple slopes and the slopes of the corresponding second map element vectors can be calculated to obtain multiple angles. The matching relationship indicated by the angles that do not meet the threshold conditions can be removed. In addition, map element vectors with reconstructed shape distortion can be excluded to obtain the aforementioned multiple crowdsourced map element vectors. Multiple crowdsourced map element vectors can also be excluded if they have not undergone shape distortion reconstruction processing. In other words, for pose graph correction preprocessing, error vector removal processing and / or mismatched vector removal processing can be performed, and when both error vector removal processing and mismatched vector removal processing are performed, the embodiments of this application do not limit the execution order of the two.

[0087] It should be understood that in a specific implementation, the matching of lane line vectors in the crowdsourced map with high-precision map vectors may result in incorrect matching. This incorrect matching will cause the heading angle of the pose to be incorrect, which in turn will cause the poses of the surrounding neighboring poses to change incorrectly through the transmission of relative poses. By determining the angle of the main direction of the lane line vector matching, the incorrect matching can be removed.

[0088] The following section, in conjunction with the accompanying drawings, explains the processing of the reconstructed shape-distorted map element vectors.

[0089] Figure 3A This is a schematic diagram of vector processing as an example of a map fusion method according to another embodiment of this application. Figure 3A As shown, in one example, the reconstructed shape of vector A can be determined to be undistorted, while the reconstructed shapes of vectors B and C can be determined to be distorted. In another example, the reconstructed shapes of vectors A and B can be determined to be undistorted, while the reconstructed shape of vector C can be determined to be distorted. In other examples, vectors A, B, and C can all be determined to be either distorted or undistorted. In such cases, vectors determined to be distorted can be excluded. It should be understood that the above implementation and illustrations are merely exemplary, and different processing methods or vector representation methods may be used in other examples.

[0090] Figure 3B This is a schematic diagram of vector processing for another example of a map fusion method according to another embodiment of this application. Figure 3BAs shown, in one example, both angles θ1 and θ2 exceed the angle threshold, so vector A can be determined as a match, while vectors B and C are determined as unmatched. In another example, angle θ1 does not exceed the angle threshold, but angle θ2 exceeds the angle threshold, so vectors A and B can be determined as a match, while the reconstructed shape of vector C is determined as unmatched. In other examples, vectors A, B, and C can all be determined as either a match or an unmatched vector. Therefore, vectors determined as unmatched can be excluded. It should be understood that the above implementation and illustrations are merely exemplary; different processing methods or vector presentation methods may be used in other examples.

[0091] Figure 4A This is a schematic flowchart illustrating a map fusion method according to another embodiment of this application. Figure 4A This is a specific implementation of the map fusion method. However, it should be understood that some steps in the following process can be appropriately added, reduced, or changed to implement the method of this embodiment.

[0092] In step 401, the pose graph of the crowdsourced map is determined based on GPS information, IMU information, and crowdsourced map data.

[0093] In step 402, based on the pose graph of the crowdsourced map, multiple initial lane line vectors are obtained through point cloud fitting, and then the process proceeds to step 403 and / or step 404.

[0094] In step 403, the multiple initial lane line vectors are subjected to first vector preprocessing to exclude overly curved vectors, resulting in multiple lane line vectors, which are then processed in step 404.

[0095] In step 404, a second vector preprocessing is performed on multiple lane line vectors to exclude vectors that have angular mismatches with the lane line vectors in the high-precision map, resulting in multiple target lane line vectors, and then proceeding to step 405.

[0096] In step 405, a distance constraint state determination is performed on the pose constraints of the pose graph of the crowdsourced map. If the distance constraint is insufficient, proceed to step 406; if the distance constraint is satisfied, proceed to step 407.

[0097] In step 406, the pose graph of the crowdsourced map is modified using pose constraints, and then the process proceeds to step 408.

[0098] In step 407, the weights of the relative pose constraint and the distance constraint in the pose constraint are adjusted, and then the process proceeds to step 408.

[0099] In step 408, the crowdsourced map and the high-precision map are fused based on the corrected pose graph. It should be understood that the above implementation and illustrations are merely exemplary; different processing methods or vector rendering methods may be used in other examples.

[0100] Figure 4B This is a schematic diagram of the pose graph before correction, representing an example of a map fusion method according to another embodiment of this application. As shown in the figure, before pose graph correction, the dashed crowdsourced lane line vectors (first map element vectors) and the solid lane line vectors (second map element vectors) do not overlap or intersect. Therefore, fusing the crowdsourced map including the first map element vector with the high-precision map including the second map element vector results in a large error. Although poses 1, 2, 3, 4, and 5 in the figure are labeled on the illustrated vehicle, it should be understood that poses 1, 2, 3, 4, and 5 can be for specific acquisition devices, such as cameras or cameras installed on the vehicle. It should be understood that the above implementation and illustrations are merely exemplary, and different processing methods or vector presentation methods can be used in other examples. It should also be understood that the dashed crowdsourced lane line vectors shown in the figure are merely an example; there can be multiple crowdsourced lane line vectors, and the correction effect is more accurate when pose graph correction is performed based on multiple crowdsourced lane line vectors. In addition, multiple crowdsourced lane line vectors can be all or part of the crowdsourced lane line vectors shown (e.g., after removing mismatched or incorrect vectors).

[0101] Figure 4CThis is a schematic diagram of the pose graph after correction, representing another example of the map fusion method according to another embodiment of this application. As shown in the figure, after pose graph correction, although the dashed crowdsourced lane line vector (first map element vector) and the solid lane line vector (second map element vector) do not overlap, their correspondence is better in terms of distance constraints, etc., due to pose constraints. Therefore, the corrected pose graph is more accurate. It should be understood that in this example, only one vector is used as an example, and other vectors are not shown; and schematic diagrams of relative pose constraints and absolute pose constraints are not shown. However, those skilled in the art will understand that for the pose constraints of the embodiments of this application, different types of constraints can play a role in correcting the pose graph. In a preferred embodiment, different types of constraints are weighted to obtain a weighted representation. Based on this weighted representation, the best correction or optimization effect can be achieved in a local area (region of interest) or globally. Furthermore, although poses 1, 2, 3, 4, and 5 in the figure are labeled on the illustrated vehicle, it should be understood that poses 1, 2, 3, 4, and 5 can be applied to specific data acquisition devices, such as cameras or sensors mounted on the vehicle. It should be understood that the above implementation and illustrations are merely exemplary; different processing methods or vector representation methods may be used in other examples. It should also be understood that the dashed crowdsourced lane line vectors shown in the figure are only one example; there can be multiple crowdsourced lane line vectors, and the correction effect is more accurate when pose graph correction is performed based on multiple crowdsourced lane line vectors. Moreover, the multiple crowdsourced lane line vectors can be all or part of the shown crowdsourced lane line vectors (e.g., after removing mismatched or erroneous vectors).

[0102] Figure 5 This is a schematic block diagram of a map fusion apparatus according to another embodiment of this application. Figure 5 The map fusion device can be applied to any suitable electronic device with data processing capabilities, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs. Figure 5 The device includes:

[0103] The acquisition module 510 is used to acquire the first map element vector and pose of the target road, and the second map element vector of the target road.

[0104] The correction module 520 is used to correct the pose represented by graphs based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector.

[0105] The fusion module 530 is used to fuse the first map element vector and the second map element vector of the target road according to the corrected pose.

[0106] In the solution of this application embodiment, since the ground model of the target road can better reflect the actual ground condition, the pose can be corrected based on the ground model, the first map element vector and the second map element vector, which can improve the accuracy of pose correction, thereby further improving the accuracy of map fusion based on the corrected pose.

[0107] In another implementation of this application, the first map element vector is a crowdsourced map element vector, the second map element vector is a high-precision map element vector, and the correction module is specifically used to: determine the projection vector of the crowdsourced map element vector of the target road onto the ground model of the target road; and correct the pose represented by the graph based on the pose constraints between the projection vector and the high-precision map element vector.

[0108] In another implementation of this application, the pose constraint includes a point-to-line distance constraint. Specifically, the correction module is used to: correct the pose represented by the graph by minimizing the value of the expression for the point-to-line distance constraint between the projection vector and the high-precision map element vector, so as to eliminate the matching error between the crowdsourced map element vector and the high-precision map element vector.

[0109] In another implementation of this application, the expression for the point-line distance constraint includes the sum of squares of multiple point-line distances between the projection vector and the high-precision map element vector, and the correction module is specifically used to minimize the value of the sum of squares of multiple point-line distances.

[0110] In another implementation of this application, the pose constraint includes a linear expression consisting of the point-line distance constraint between the projection vector and the high-precision map element vector, and the absolute pose constraint of the high-precision map element vector and its multi-frame pose. Specifically, the correction module is used to correct the pose represented by the graph by minimizing the value of the linear expression.

[0111] In another implementation of this application, the pose constraint includes a linear expression consisting of the point-line distance constraint between the projection vector and the high-precision map element vector and the relative pose constraint between the multi-frame poses of the high-precision map element vector. Specifically, the correction module is used to correct the pose represented by the graph by minimizing the value of the linear expression.

[0112] In another implementation of this application, in the linear expression, the point-to-line distance constraint has a first weight and the relative pose constraint has a second weight. The device further includes an adjustment module that adjusts the first weight and the second weight based on the constraint state of the point-to-line distance constraint.

[0113] In another implementation of this application, the adjustment module is specifically used to: in the target area of ​​the target road, if the crowdsourced map lane line vector is distributed on one side of the pose, determine that the constraint state of the point-line distance constraint is insufficient, and reduce the first weight and / or increase the second weight.

[0114] In another implementation of this application, the pose constraint includes a linear expression consisting of the point-line distance constraint between the projection vector and the high-precision map element vector, the absolute pose constraint between the high-precision map element vector and its multi-frame pose, and the relative pose constraint between the multi-frame pose. Specifically, the correction module is used to correct the pose represented by the graph by minimizing the value of the linear expression.

[0115] In another implementation of this application, the acquisition module is specifically used to: exclude crowdsourced map element vectors with distorted reconstruction shapes from multiple crowdsourced map element vectors of the target road to obtain a first map element vector.

[0116] In another implementation of this application, the acquisition module is specifically used to: calculate the fitted straight line of each of the multiple crowdsourced map element vectors; if the variance of the distance from multiple points on a crowdsourced map element vector to its fitted straight line is greater than a threshold, then the crowdsourced map element vector is determined to be a reconstructed shape distortion, and the crowdsourced map element vector is excluded.

[0117] In another implementation of this application, the acquisition module is specifically used to: exclude crowdsourced map element vectors from multiple crowdsourced map element vectors of the target road whose included angle with the second map element vector exceeds a threshold, and obtain a first map element vector that matches the second map element vector.

[0118] The apparatus of this embodiment is used to implement the corresponding methods in the foregoing method embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here. Furthermore, the functional implementation of each module in the apparatus of this embodiment can be referred to the description of the corresponding part in the foregoing method embodiments, which will also not be repeated here.

[0119] Figure 6 The hardware structure of the electronic device according to another embodiment of this application, such as Figure 6 As shown, the hardware structure of the electronic device may include: a processor 601, a communication interface 602, a storage medium 603, and a communication bus 604.

[0120] The processor 601, communication interface 602, and storage medium 603 communicate with each other through communication bus 604.

[0121] Optionally, the communication interface 602 can be an interface of a communication module.

[0122] Specifically, the processor 601 can be configured to: acquire a first map element vector and pose of the target road, and a second map element vector of the target road; correct the pose represented by graph based on a pre-generated ground model of the target road, the first map element vector, and the second map element vector; and fuse the first map element vector and the second map element vector of the target road according to the corrected pose.

[0123] The processor 601 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor.

[0124] The storage medium 603 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0125] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a storage medium, the computer program containing program code configured to perform the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by a central processing unit (CPU), it performs the functions defined in the methods of this application. It should be noted that the storage medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access storage media (RAM), read-only storage media (ROM), erasable programmable read-only storage media (EPROM or flash memory), optical fibers, portable compact disk read-only storage media (CD-ROM), optical storage media, magnetic storage media, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any storage medium other than a computer-readable storage medium that can send, propagate, or transmit a program configured for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the storage medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0126] Computer program code configured to perform the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as "C" or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions configured to perform a specified logical function. Specific sequences are present in the above specific embodiments, but these sequences are merely exemplary; in actual implementations, these steps may be fewer, more, or executed in a different order. That is, in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0128] The modules described in the embodiments of this application can be implemented in software or hardware. The names of these modules do not necessarily limit the module itself.

[0129] In another aspect, embodiments of this application also provide a storage medium having a computer program stored thereon that, when executed by a processor, implements the method as described in the above embodiments.

[0130] In another aspect, this application also provides a storage medium, which may be included in the apparatus described in the above embodiments; or it may exist independently and not assembled into the apparatus. The storage medium carries one or more programs that, when executed by the apparatus, cause the apparatus to: acquire a first map element vector and pose of a target road, and a second map element vector of the target road; correct the pose represented by a graph based on a pre-generated ground model of the target road, the first map element vector, and the second map element vector; and fuse the first map element vector and the second map element vector of the target road according to the corrected pose.

[0131] The terms "first," "second," "first," or "second" used in the various embodiments of this application may modify various components regardless of their order and / or importance, but these terms do not limit the corresponding components. The above terms are configured only for the purpose of distinguishing an element from other elements. For example, "first user equipment" and "second user equipment" refer to different user equipments, although both are user equipment. For example, without departing from the scope of this application, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element.

[0132] When a component (e.g., a first component) is referred to as being "(operably or communicatively) coupled" or "(operably or communicatively) coupled to" or "connected to" another component (e.g., a second component), it should be understood that the first component is directly connected to the second component or that the first component is indirectly connected to the second component via yet another component (e.g., a third component). Conversely, it can be understood that when a component (e.g., a first component) is referred to as being "directly connected" or "directly coupled" to another component (the second component), no component (e.g., a third component) is inserted between the two.

[0133] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A map fusion method, comprising: Obtain a first map element vector and pose of the target road, and a second map element vector of the target road, wherein the first map element vector is a crowdsourced map element vector, and the second map element vector includes a high-precision map element vector; Based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector, the pose represented by the graph is corrected. The pose represented by the graph is called the pose graph, which represents the position vector and attitude matrix of the pose of multiple frames. Based on the corrected pose, the first map element vector and the second map element vector of the target road are fused.

2. The method according to claim 1, wherein, The step of correcting the pose represented by a graph based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector includes: Determine the projection vector of the crowdsourced map element vector of the target road onto the ground model of the target road; Based on the pose constraints between the projection vector and the high-precision map element vector, the pose represented by the graph is corrected.

3. The method according to claim 2, wherein, The pose constraints include point-line distance constraints, wherein the correction of the pose represented by the graph based on the pose constraints between the projection vector and the high-precision map element vectors includes: By minimizing the value of the expression for the point-to-line distance constraint between the projection vector and the high-precision map element vector, the pose represented by the graph is corrected to eliminate the matching error between the crowdsourced map element vector and the high-precision map element vector.

4. The method according to claim 3, wherein, The expression for the point-line distance constraint includes the sum of the squares of multiple point-line distances between the projected vector and the high-precision map element vector, wherein, The expression for minimizing the point-to-line distance constraint between the projection vector and the high-precision map element vector includes: This minimizes the sum of the squares of the distances between the points and lines.

5. The method according to claim 2, wherein, The pose constraint includes a linear expression consisting of the point-to-line distance constraint between the projection vector and the high-precision map element vector, and the absolute pose constraint between the high-precision map element vector and its multi-frame pose. The step of correcting the pose represented by the graph based on the pose constraint between the projection vector and the high-precision map element vector includes: The pose represented by the graph is corrected by minimizing the value of the linear expression.

6. The method according to claim 2, wherein, The pose constraint includes a linear expression composed of the point-to-line distance constraint between the projection vector and the high-precision map element vector, and the relative pose constraint between the multi-frame poses of the high-precision map element vector. The step of correcting the pose represented by the graph based on the pose constraint between the projection vector and the high-precision map element vector includes: The pose represented by the graph is corrected by minimizing the value of the linear expression.

7. The method according to claim 6, wherein, In the linear expression, the point-to-line distance constraint has a first weight and the relative pose constraint has a second weight. The method further includes: Based on the constraint state of the point-line distance constraint, adjust the first weight and the second weight.

8. The method according to claim 7, wherein, The adjustment of the first weight and the second weight based on the constraint state of the point-line distance constraint includes: If the crowdsourced map lane line vector is distributed on one side of the pose in the target area of ​​the target road, the constraint state of the point-line distance constraint is determined to be insufficient, and the first weight is reduced and / or the second weight is increased.

9. The method according to claim 2, wherein, The pose constraint includes a linear expression consisting of the point-line distance constraint between the projection vector and the high-precision map element vector, the absolute pose constraint between the high-precision map element vector and its multi-frame pose, and the relative pose constraint between the multi-frame poses. The step of correcting the pose represented by the graph based on the pose constraint between the projection vector and the high-precision map element vector includes: The pose represented by the graph is corrected by minimizing the value of the linear expression.

10. The method according to claim 1, wherein, The acquisition of the first map element vector of the target road includes: From the multiple crowdsourced map element vectors of the target road, exclude the crowdsourced map element vectors that are reconstructed with distorted shapes to obtain the first map element vector.

11. The method according to claim 10, wherein, The step of excluding crowdsourced map element vectors with distorted reconstructed shapes from multiple crowdsourced map element vectors of the target road includes: Calculate the fitted straight line for each of the multiple crowdsourced map element vectors; If the variance of the distances from multiple points on a crowdsourced map element vector to its fitted line is greater than a threshold, then the crowdsourced map element vector is determined to have distorted reconstruction shape and is excluded.

12. The method according to claim 1, wherein, The acquisition of the first map element vector of the target road includes: From the multiple crowdsourced map element vectors of the target road, exclude the crowdsourced map element vectors whose angle with the second map element vector exceeds a threshold, and obtain the first map element vector that matches the second map element vector.

13. A map fusion device, comprising: The acquisition module is used to acquire a first map element vector and pose of the target road, and a second map element vector of the target road, wherein the first map element vector is a crowdsourced map element vector, and the second map element vector includes a high-precision map element vector; The correction module is used to correct the pose represented by the graph based on the pre-generated ground model of the target road, the first map element vector, and the second map element vector. The pose represented by the graph is called the pose graph, which represents the position vector and attitude matrix of the pose of multiple frames. The fusion module is used to fuse the first map element vector and the second map element vector of the target road according to the corrected pose.

14. An electronic device, the device comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the method as described in any one of claims 1-12.

15. A storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as claimed in any one of claims 1-12.

Images