Parking map generation method and device, electronic equipment and storage medium

Abstract

The invention relates to the field of map generation, in particular to a parking map generation method and device, electronic equipment and a storage medium. According to the method, firstly, the initial reference of map construction can be accurately determined by obtaining the mapping starting point of the initial map which is being constructed; and then the target local map closest to the mapping starting point is searched, existing map resources around can be efficiently utilized, repeated drawing of similar areas is avoided, and storage space and computing resources are effectively saved. And then the laser point cloud of the initial map is converted to a target local map and feature matching is carried out to obtain an accurate target conversion relationship, so that the connection accuracy between different maps is ensured. And finally, the target map is spliced based on the target map, so that the dispersed maps are integrated, a route can be clearly judged according to the integrated target map during subsequent automatic parking, interference of multiple maps is avoided, and the problem that the route is difficult to judge is solved.

Description

Technical Field

[0001] The present invention relates to the field of map generation, and in particular to a method, device, electronic device and storage medium for generating a parking map. Background Art

[0002] Memory parking is an automatic driving function in the parking lot area. It can assist users to automatically cruise into a parking space according to the memorized route, or automatically park out of a parking space and then cruise to the exit point. Since memory parking is essentially a low-speed automatic driving control process, a certain design operation range must be set during the process to ensure that memory mapping and driving control are feasible. To achieve memory parking, it is often necessary to generate a parking map first. The map covers parking routes and parking spaces. Based on the generated parking map, the vehicle can drive along the corresponding parking route and park in the parking space to achieve automatic memory parking.

[0003] However, in the current memory parking algorithm, a separate parking map is created for each parking route. If there are multiple routes that are close in distance and partially overlap, the system will store a number of maps corresponding to the number of routes. This not only makes it difficult for the system to determine which route to use during automatic parking, but also causes a waste of storage space. Summary of the invention

[0004] In view of this, embodiments of the present invention provide a method, device, electronic device and storage medium for generating a parking map to solve the problem that in the current memory parking algorithm, a separate parking map is created for each parking route. When there are routes that are close in distance and partially overlap, it will make it difficult for the system to determine the route during automatic parking and waste storage space.

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

[0006] Get the starting point of the initial map currently being constructed;

[0007] Acquire a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point;

[0008] Converting the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and performing feature matching between the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map;

[0009] The initial map and the target local map are spliced ​​based on the target conversion relationship to obtain a target map.

[0010] Furthermore, the obtaining of the local map of the target adjacent to the mapping starting point includes:

[0011] The search area is constructed by taking the mapping starting point as the center and the preset distance as the radius;

[0012] determining whether a local map exists in the search area;

[0013] If the local map exists in the search area, the local map closest to the map building starting point is used as the target local map.

[0014] Furthermore, converting the laser point cloud corresponding to the initial map to the target local map to obtain the converted laser point cloud includes:

[0015] Acquire first global pose data collected at the mapping starting point during the mapping process;

[0016] Acquire second global pose data of the target local map;

[0017] Based on the first global pose data and the second global pose data, constructing a pose conversion relationship between the initial map and the local map;

[0018] Based on the posture conversion relationship, the laser point cloud of the initial map is converted to the target local map to obtain a converted laser point cloud.

[0019] Furthermore, the method of performing feature matching using the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map includes:

[0020] Acquire a map feature range corresponding to the target local map, and use the laser point cloud within the map feature range in the converted laser point cloud as the target laser point cloud;

[0021] Matching the target laser point cloud with a feature vector in the map feature range to obtain a first matching degree in a spatial dimension and a second matching degree in a posture dimension between the target laser point cloud and the feature vector;

[0022] Using the first matching degree and the second matching degree to update the posture conversion relationship, to obtain an updated posture conversion relationship;

[0023] The target laser point cloud is transformed using the updated posture transformation relationship until the error between the transformed target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range, thereby obtaining the target transformation relationship.

[0024] Further, the updating of the posture conversion relationship by using the first matching degree and the second matching degree to obtain an updated posture conversion relationship includes:

[0025] Using the first matching degree to adjust the translation parameters in the posture transformation relationship to obtain updated translation parameters;

[0026] adjusting the rotation parameters in the posture transformation relationship using the second matching degree to obtain updated rotation parameters;

[0027] An updated pose transformation relationship is constructed based on the updated translation parameters and the updated rotation parameters.

[0028] Further, the target laser point cloud is converted using the updated posture conversion relationship until the error between the converted target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range, and the target conversion relationship is obtained, including:

[0029] The target laser point cloud is transformed using the updated posture transformation relationship to obtain a transformed target laser point cloud;

[0030] Matching the converted target laser point cloud with the feature vector in the map feature range to obtain a third matching degree in the spatial dimension and a fourth matching degree in the posture dimension between the target laser point cloud and the feature vector;

[0031] Determine an error between the converted target laser point cloud and the target local map based on the third matching degree and the fourth matching degree;

[0032] If the error reaches a preset range, the updated posture transformation relationship is used as the target transformation relationship.

[0033] Furthermore, after the target map is obtained by splicing the initial map and the target local map based on the target conversion relationship, the method further includes:

[0034] Obtain the user's map usage preferences or map usage needs;

[0035] Acquire a functional module corresponding to the map usage preference or the map usage requirement;

[0036] Add the functional module to the target map.

[0037] In a second aspect, an embodiment of the present invention provides a device for generating a parking map, the device comprising:

[0038] The first acquisition module is used to obtain the starting point of the initial map currently being constructed;

[0039] A second acquisition module is used to acquire a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point;

[0040] A conversion module, used to convert the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and use the converted laser point cloud to perform feature matching with the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map;

[0041] A splicing module is used to splice the initial map and the target local map based on the target conversion relationship to obtain a target map.

[0042] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions being stored in the memory, and the processor executing the method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0043] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method of the first aspect or any corresponding embodiment thereof.

[0044] This application first obtains the starting point of the initial map being constructed, and can accurately determine the starting reference for map construction. Then, the local map of the target that is closest to the starting point of the map is found, which can efficiently utilize the existing map resources in the surrounding area, avoid repeated drawing of similar areas, and effectively save storage space and computing resources. The laser point cloud of the initial map is then converted to the target local map and feature matching is performed to obtain an accurate target conversion relationship, thereby ensuring the accuracy of the connection between different maps. Finally, the target map is spliced ​​out based on this, so that the scattered maps can be integrated. During subsequent automatic parking, the route can be clearly determined based on the integrated target map, and it is no longer disturbed by multiple maps, solving the problem of difficulty in determining the route. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 is a flow chart of a method for generating a vehicle map according to some embodiments of the present invention;

[0047] Figure 2 is a flowchart of another method for generating a vehicle map according to some embodiments of the present invention;

[0048] Figure 3 is a structural block diagram of a vehicle map generating device according to an embodiment of the present invention;

[0049] Figure 4 It is a schematic diagram of the hardware structure of the electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0051] According to an embodiment of the present invention, a method, apparatus, electronic device and storage medium for generating a parking map are provided. 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 a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0052] In this embodiment, a method for generating a parking map is provided. Figure 1 is a flow chart of a method for generating a parking map according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0053] Step S101, obtaining the starting point of the initial map currently being constructed.

[0054] In the embodiment of the present application, obtaining a local map of a target adjacent to a mapping starting point includes:

[0055] Step A1: Use the starting point of the map as the center and the preset distance as the radius to construct the search area.

[0056] Specifically, in the application scenario of memory parking, when the vehicle is ready to park and start building a parking map, the starting point of this map construction will be determined first. This starting point is usually the initial location where the vehicle enters the parking area and begins to record parking-related environmental information (for example, the location of the vehicle when it just enters the area near a parking space in the parking lot, and its coordinate information can be obtained through the vehicle positioning system). Then, based on the pre-set distance value (this distance is determined based on comprehensive factors such as the actual parking scene, parking lot size, and system design requirements, for example, it is set to 50 meters), with the starting point of the map construction as the center and this preset distance as the radius, a circular search area is delineated on a two-dimensional plane (corresponding to the plane layout of the parking lot) through relevant geometric calculations and spatial definition methods, so as to clarify the approximate range of subsequent searches for existing local maps.

[0057] Step A2: Determine whether there is a local map in the search area.

[0058] Specifically, the vehicle's memory parking system will interact and communicate with the relevant modules that store map data, and traverse the existing map data records within the constructed search area (the map data can be an environmental information map of a local area of ​​the parking lot that has been pre-collected and stored by previous parking operations or other methods, and contains information such as parking space locations, channel directions, and obstacle distribution in the form of point clouds or other forms). The system determines whether the geographic coordinate range corresponding to these map data intersects with the current search area, or determines whether there is a local map that meets the requirements within the search area through feature matching and other methods (such as checking whether key landmark features in the map fall within the search area), so as to know whether there are available stored map resources.

[0059] Step A3: If there is a local map in the search area, the local map closest to the map building starting point is used as the target local map.

[0060] Specifically, if it is found that there are indeed local maps in the search area, the system will further calculate the distance between each local map and the current map starting point. The calculation method of this distance can be determined based on the spatial distance between the center coordinates of the map or the coordinates of the key feature points and the coordinates of the starting point of the map (for example, the distance between two points on a two-dimensional plane is calculated using the Euclidean distance formula). After performing such distance calculations on all local maps in the search area, by comparing their distances to the starting point of the map, the local map closest to the starting point of the map is screened out and selected as the target local map. Subsequently, further operations such as coordinate conversion and map merging can be performed around this target local map to better utilize existing map resources, avoid repeated map construction, and solve the problems of route determination and storage space waste mentioned earlier.

[0061] Step S102, obtaining a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point.

[0062] In the embodiment of the present application, converting the laser point cloud corresponding to the initial map to the target local map to obtain the converted laser point cloud includes the following steps B1-B4:

[0063] Step B1, obtaining the first global pose data collected at the starting point of the map construction process.

[0064] Specifically, when the vehicle starts to build a parking map, the starting point of the map is the position of the vehicle when the parking operation starts (its coordinates can be obtained through the on-board high-precision positioning system, such as the integration of GPS, Beidou and other satellite positioning and the on-board inertial navigation system to determine its position coordinates in the parking lot).

[0065] At the same time, the vehicle's various sensors (such as laser radars and cameras installed around the vehicle body) will monitor the vehicle's own posture information in real time. The posture is usually described by Euler angles (including roll angle, pitch angle, and yaw angle). For example, the laser radar can obtain the vehicle's posture by scanning the relative angle changes of the surrounding environment combined with its own calibration information. At the starting point of the map, the system will integrate the acquired vehicle position coordinate information (longitude, latitude, and altitude. In the plane scene of the parking lot, the altitude information is relatively fixed, and the plane coordinates play a key role) and the Euler angle information of the posture, and record and save it as the first global posture data. This set of data fully reflects the position and posture state of the vehicle in the entire parking lot space at the starting point of the map.

[0066] Step B2, obtaining second global pose data of the target local map.

[0067] Specifically, when the target local map is initially constructed or stored, its corresponding global position information is also recorded. Its position information is also determined by corresponding positioning means (for example, the coordinate position record of the vehicle when constructing the local map, or the coordinate conversion of key landmarks in the map in the overall coordinate system of the parking lot) to clarify its coordinate situation (latitude, longitude and altitude information) in the parking lot environment.

[0068] As for posture, it is based on the map orientation, angle and other information determined when constructing the local map (for example, the angle between the channel in the map and the overall coordinate axis of the parking lot, etc., and the corresponding Euler angle is obtained by analyzing the map features and combining the angle measurement to represent the posture). By integrating these position and posture-related data, it constitutes the second global posture data of the target local map, which represents the specific state of the local map in the entire parking lot space.

[0069] Step B3: constructing a posture conversion relationship between the initial map and the local map based on the first global posture data and the second global posture data.

[0070] Specifically, after obtaining the first global pose data of the starting point of the map and the second global pose data of the target local map, the system needs to construct a pose conversion relationship between the two through mathematical calculations and coordinate transformation algorithms.

[0071] From the perspective of position, it is necessary to calculate the translation parameters between the coordinate system of the starting point of mapping (the initial map coordinate system, a local coordinate system built with the starting point of mapping as the relative origin) and the target local map coordinate system. For example, the difference in longitude and latitude and the difference in plane coordinates between the two can be used to determine the distance required for translation in the x, y, and z directions (mainly the x and y directions in the parking lot plane scene) to make the coordinate position of the initial map correspond to the coordinate position of the target local map.

[0072] In terms of posture, the difference between the Euler angles (roll, pitch, yaw) of the two is used to determine what kind of rotation operation is needed to make the posture of the initial map consistent with that of the target local map. By combining the calculations of these translation and rotation related parameters, and using mathematical tools such as coordinate transformation matrices (for example, constructing a homogeneous coordinate transformation matrix containing translation and rotation matrices), the posture transformation relationship between the initial map and the local map is constructed. This relationship can accurately describe how to transform the elements in the initial map to the coordinate system of the target local map so that subsequent operations can be performed in the same spatial reference system.

[0073] Step B4, based on the posture conversion relationship, the laser point cloud of the initial map is converted to the target local map to obtain a converted laser point cloud.

[0074] Specifically, during the memory parking process, the surrounding environment information collected by the vehicle using sensors such as lidar is presented in the form of laser point clouds. These laser point clouds initially record their coordinate position and posture based on the initial map coordinate system where the mapping starting point is located.

[0075] According to the constructed posture transformation relationship, for each laser point cloud in the initial map, through the corresponding coordinate transformation calculation (i.e., the coordinate vector of the laser point cloud is multiplied by the transformation matrix corresponding to the posture transformation relationship and other mathematical operations), it is translated in space according to the calculated translation parameters, and its posture is rotated according to the rotation parameters, so as to transform these laser point clouds from the coordinate system of the initial map to the coordinate system of the target local map. After performing such a conversion operation on all laser point clouds, the converted laser point clouds are obtained.

[0076] Step S103, converting the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and performing feature matching between the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map.

[0077] In the embodiment of the present application, feature matching is performed using the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map, including the following steps C1-C4:

[0078] Step C1, obtaining a map feature range corresponding to the target local map, and taking the laser point cloud within the map feature range in the converted laser point cloud as the target laser point cloud.

[0079] Specifically, for the target local map (which contains relevant information such as parking space outlines, channel boundaries, and fixed obstacles in the parking lot in the form of point clouds, etc.), we must first analyze the feature range formed by the various geographical elements inside it. For example, by extracting key features such as the marked parking space polygon area in the map (defined by the parking space boundaries composed of multiple point clouds), the starting and ending points and width range of the channel (the point clouds on both sides of the boundary determine its range), and other fixed obstacles (such as columns, etc., the point cloud distribution around them determines the space range they occupy). Using relevant geometric algorithms or traversal methods based on map data structures, the coordinate range of each feature in space is clarified, and the map feature range corresponding to the entire target local map is determined in combination. This range defines which spatial areas belong to the key parts of the map that have practical significance.

[0080] Taking parking spaces as an example, by extracting the coordinates of the four corner points of the parking space and the coordinate information of the boundary point cloud, we can determine a rectangular or polygonal range to represent the feature range of the parking space in the map. After such a range definition is made for all similar key features, the complete map feature range is obtained.

[0081] Next, for the laser point cloud obtained after the coordinate conversion (these laser point clouds were originally collected based on the initial map and have now been converted to the target local map coordinate system), determine whether the coordinate position of each laser point cloud falls within the previously determined map feature range. Screening can be performed by simple coordinate range comparison (for example, determining whether the x and y coordinates of the laser point cloud are between the minimum x coordinate and the maximum x coordinate of a feature, and between the minimum y coordinate and the maximum y coordinate) or more complex point-polygon relationship judgment algorithms (if the feature range is a polygon, determine whether the point is inside the polygon).

[0082] The laser point clouds whose coordinate positions are within the range of map features are selected. These selected laser point clouds are directly related to the key features of the target local map. They are used as target laser point clouds, and subsequent matching and other operations are carried out around these target laser point clouds. This can eliminate interference from laser point clouds in blank areas or irrelevant areas of the map, allowing operations to focus more on valid data.

[0083] Step C2, matching the target laser point cloud with the feature vector in the map feature range, and obtaining a first matching degree in the spatial dimension and a second matching degree in the posture dimension between the target laser point cloud and the feature vector.

[0084] Specifically, first, for the target laser point cloud (i.e., the laser point cloud that is within the target local map feature range after screening), an extraction method based on local geometric features can be used. For example, for each laser point cloud point, calculate the normal vector of the point cloud in a certain neighborhood around it, and determine the direction of the normal vector of the point by analyzing the distribution of the point cloud in the neighborhood. The normal vector can reflect the local geometric shape and orientation information of the point cloud. At the same time, calculate the curvature information in the neighborhood of the point cloud point. The curvature can reflect the degree of curvature of the point cloud surface. For example, in a parking lot scene, the curvature of the point cloud at the boundary of the parking space will be different from the curvature of the point cloud on the channel. Combine local geometric features such as normal vectors and curvature to form a feature vector that can characterize the laser point cloud point. Perform this operation on all points in the target laser point cloud, and finally obtain a set of feature vectors to describe the characteristics of the entire target laser point cloud.

[0085] For each feature in the target local map feature range (such as parking spaces, passages, fixed obstacles, etc.), its corresponding feature vector also needs to be extracted. Taking the parking space as an example, first determine the point cloud data of the parking space boundary, and then calculate the normal vector, curvature and other features of the parking space boundary point cloud according to the above-mentioned local geometric feature extraction method for laser point cloud, and combine them into a feature vector to describe the geometric properties of the parking space feature in space.

[0086] The similarity between the feature vector of the target laser point cloud and the feature vectors of each feature in the map feature range is calculated in the spatial dimension (i.e., x, y, and z dimensions). One of the commonly used calculation methods is the Euclidean distance. For a feature vector of the target laser point cloud and a feature vector of a feature in the map (such as a parking space), the similarity is measured by calculating their Euclidean distance in the feature vector space.

[0087] Since the target laser point cloud will match multiple features in the map, it is necessary to comprehensively consider these matching situations to obtain the overall first matching degree. The weighted average method can be used to assign different weights to each feature according to its importance in the map. For example, in a parking scene, the parking space feature is relatively more important and can be assigned a higher weight, while the channel feature has a slightly lower weight.

[0088] The similarity values ​​obtained by matching each feature vector are multiplied by the corresponding weights and then summed, and then divided by the total weights to obtain a comprehensive similarity quantization value, which is the first matching degree between the target laser point cloud and the feature vector in the spatial dimension. It can reflect the overall degree of fit between the target laser point cloud in terms of spatial position, shape, etc. and each feature in the map feature range. The higher the first matching degree, the better the match in the spatial dimension.

[0089] In the attitude dimension (i.e., roll, pitch, and yaw dimensions), we must first obtain elements that can reflect attitude information from the extracted feature vectors. For example, for the feature vector of the target laser point cloud, the distribution of the point cloud normal vectors it contains can indirectly reflect the attitude angle information of the point cloud as a whole. By analyzing the angular relationship between the normal vectors of multiple point cloud points in space and combining certain mathematical transformations, we can obtain the approximate attitude angles of the point cloud as a whole in the roll, pitch, and yaw directions.

[0090] Similarly, for features in the map feature range (such as channels, parking spaces, etc.), the corresponding attitude angle information is parsed from its feature vector, and then the difference in attitude angle between the target laser point cloud and the map feature is calculated. For example, in the yaw direction, the angle difference between the two is calculated. The smaller the angle difference, the more similar the yaw attitude is; similar angle difference calculations are performed for roll and pitch.

[0091] In addition to angle difference calculation, a similarity calculation method based on posture feature vector can also be used, similar to the cosine similarity in the spatial dimension, to construct a vector representation specifically for posture features (for example, roll, pitch, and yaw angle values ​​are combined into a vector), and then calculate the similarity between them to measure the matching of the posture dimension.

[0092] Similar to calculating the first matching degree in the spatial dimension, in the attitude dimension, comprehensive statistics are performed on the matching of the target laser point cloud and the map features at various angles of roll, pitch, and yaw (reflected by angle difference or attitude feature vector similarity, etc.). Similarly, weighted averaging and other methods can be used to assign corresponding weights according to the importance of different attitude angles to the overall attitude matching (for example, in some scenarios, the yaw angle is more critical and the weight is relatively larger).

[0093] After weighted summation of the matching conditions of each attitude angle, a comprehensive quantitative value is obtained, namely the second matching degree between the target laser point cloud and the feature vector in the attitude dimension, which reflects the overall consistency level between the attitude of the target laser point cloud and the map feature attitude. The higher the second matching degree, the better the matching effect in the attitude dimension.

[0094] Step C3, updating the posture conversion relationship using the first matching degree and the second matching degree to obtain an updated posture conversion relationship.

[0095] Specifically, the posture transformation relationship is updated using the first matching degree and the second matching degree to obtain an updated posture transformation relationship, including: adjusting the translation parameters in the posture transformation relationship using the first matching degree to obtain updated translation parameters; adjusting the rotation parameters in the posture transformation relationship using the second matching degree to obtain updated rotation parameters; constructing an updated posture transformation relationship based on the updated translation parameters and the updated rotation parameters.

[0096] First, the first matching degree reflects the degree of fit between the target laser point cloud and the map feature vector in the spatial dimension (x, y, z dimension). When the first matching degree is low, it means that there is a deviation in the spatial position, and analysis is needed to determine the specific error in each dimension. For example, if the first matching degree calculated by methods such as Euclidean distance or cosine similarity is lower than the preset reasonable threshold, it means that the target laser point cloud is not accurately aligned with the map feature in spatial position.

[0097] The error can be further clarified by comparing the coordinates of the target laser point cloud with the coordinate difference of the corresponding position of the map feature. For example, in the x direction, the difference between the average x coordinate of each point in the target laser point cloud and the corresponding position in the map feature (such as the coordinate of the center of the parking space in the x direction) is calculated. If the difference is positive, it means that the target laser point cloud is biased to the right in the x direction as a whole, otherwise it is biased to the left; similarly, the same coordinate difference analysis is performed on the y and z directions to obtain the specific position deviation in each spatial dimension.

[0098] According to the analyzed spatial dimension error, a corresponding translation parameter adjustment strategy is formulated. Usually, an adjustment method based on the error ratio can be adopted, for example, an adjustment coefficient is set (the coefficient can be set according to actual conditions and experience, and is generally between 0 and 1, such as 0.1, 0.2, etc.).

[0099] Taking the x direction as an example, if the coordinate deviation of the target laser point cloud in the x direction relative to the map feature is calculated to be Δx (unit: meter), and the x-direction translation parameter in the current posture transformation relationship is Tx (unit: meter), then the updated x-direction translation parameter Tx' can be calculated according to the following formula: Tx'=Tx+k×Δx (where k is the adjustment coefficient). In this way, according to the direction (positive or negative) and size of the deviation, the translation parameter is reasonably adjusted by adjusting the coefficient, so that the target laser point cloud moves in the x direction in a direction closer to the correct position of the map feature. For the y and z directions, a similar method is also used. According to the coordinate deviation in each direction, combined with the corresponding adjustment coefficient, the updated y-direction translation parameter Ty' and z-direction translation parameter Tz' are calculated respectively, so as to achieve targeted adjustment of each translation parameter based on the spatial dimension error.

[0100] Through the above calculations, the updated translation parameters Tx', Ty' and Tz' are obtained, completing the update of the translation parameters in the pose transformation relationship. After the update, the new translation parameters need to be verified for rationality to avoid unreasonable adjustment results (such as the translation parameters causing the target laser point cloud to exceed the reasonable range of the map, etc.).

[0101] The second matching degree reflects the matching degree between the target laser point cloud and the map feature vector in the attitude dimension (roll, pitch, yaw dimension). When the second matching degree does not meet the requirements, it means that there is a deviation in the attitude, and it is necessary to analyze the specific error situation in each attitude angle.

[0102] It is determined by the angle difference obtained when calculating the attitude dimension matching previously. For example, in the yaw angle (yaw) direction, the difference Δyaw (unit: degree) between the overall yaw angle of the target laser point cloud and the map feature (such as the yaw angle corresponding to the correct direction of the channel) is calculated. If Δyaw is not 0, it indicates that there is a deviation in the yaw attitude. A positive value indicates that the target laser point cloud has rotated a certain angle clockwise relative to the map feature, and a negative value indicates that it has rotated a corresponding angle counterclockwise. The same is true for the roll angle (roll) and pitch angle (pitch). The specific angle differences of Δroll and Δpitch are calculated respectively to clarify the specific deviations at each attitude angle.

[0103] According to the analyzed attitude dimension error, formulate the adjustment strategy of rotation parameters. It is common to use rotation matrix for adjustment. Taking yaw angle (yaw) as an example, assuming that the rotation parameter corresponding to yaw in the current posture transformation relationship is Ryaw (in radians, the angle value can be converted to radian value according to the conversion relationship between angle and radian), according to the angle deviation Δyaw (converted to radians), construct a rotation matrix Rz (Δyaw) rotating around the z axis. Then, through matrix multiplication and other operations, the current rotation parameter is combined with this rotation matrix to update the rotation parameter Ryaw' in the yaw direction, so that the target laser point cloud can rotate in the yaw attitude in a direction closer to the correct attitude of the map feature. For the roll angle (roll) and pitch angle (pitch), the rotation matrices around the x-axis and y-axis (such as Rx (Δroll) and Ry (Δpitch)) are constructed respectively. According to the similar matrix operation method, combined with the current rotation parameters, the rotation parameters Rroll' in the Roll direction and Rpitch' in the pitch direction are updated to achieve targeted adjustment of each rotation parameter.

[0104] Step C4, using the updated posture transformation relationship to transform the target laser point cloud, until the error between the transformed target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range, and the target transformation relationship is obtained.

[0105] Specifically, the target laser point cloud is transformed using the updated posture transformation relationship until the error between the transformed target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range to obtain the target transformation relationship, including: using the updated posture transformation relationship to transform the target laser point cloud to obtain the transformed target laser point cloud; matching the transformed target laser point cloud with the feature vector in the map feature range to obtain a third matching degree in the spatial dimension and a fourth matching degree in the posture dimension between the target laser point cloud and the feature vector; determining the error between the transformed target laser point cloud and the target local map based on the third matching degree and the fourth matching degree; if the error reaches the preset range, using the updated posture transformation relationship as the target transformation relationship.

[0106] First, the position of each point in the target laser point cloud should be moved according to the translation parameters involved in the updated pose transformation relationship. The pose transformation relationship will clearly give the distance values ​​that need to be translated in the three directions of x, y, and z. For any point in the target laser point cloud, check its original coordinate value in the x direction, and then move the point in the x direction by the corresponding distance according to the translation parameter in the x direction in the pose transformation relationship to obtain a new x coordinate. For example, if the original x coordinate of a point is 5 meters, and the x-direction translation parameter in the pose transformation relationship requires a translation of 2 meters to the right, then the new x coordinate of this point becomes 7 meters. The same reasoning applies to the y and z directions, that is, the positions of the points are moved according to the translation parameters corresponding to their respective directions, so that each point is finally translated in space. The overall position of the target laser point cloud in space changes in a direction that is more compatible with the target local map.

[0107] Next, we will proceed to the posture rotation part. The posture conversion relationship contains rotation parameter information about the three posture angles of roll, pitch, and yaw. First, from the perspective of yaw, according to the corresponding rotation parameters, we determine that the target laser point cloud needs to be rotated around a certain axis (usually an axis perpendicular to the ground, similar to the axis around which a car turns) by a certain angle. For example, if the rotation parameters require the target laser point cloud to rotate 30 degrees clockwise, then the entire target laser point cloud is rotated according to this requirement, so that its orientation on the plane changes and it is more in line with the direction of the corresponding features in the target local map. Then, for the roll angle (roll) and pitch angle (pitch), according to their respective corresponding rotation parameters, they are rotated around the appropriate axis (roll corresponds to the axis of the vehicle rolling left and right, and pitch corresponds to the axis of the vehicle raising and lowering its head) at corresponding angles, so that the overall posture of the target laser point cloud is adjusted in three-dimensional space to the state expected by the posture conversion relationship. Through such coordinate translation and attitude rotation, all points in the target laser point cloud are transformed, thereby obtaining the transformed target laser point cloud.

[0108] For the converted target laser point cloud, it is necessary to extract the feature vector that can represent its characteristics. This can be extracted by analyzing the local situation around each point in the point cloud, such as checking how other points around the point are distributed, whether they are relatively concentrated or scattered, so as to determine some information that can describe the characteristics of the point, such as the degree of curvature of the local area where the point is located, the general direction, etc. Combining this information forms the feature vector element corresponding to this point. After all points in the target laser point cloud are processed in this way, a complete set of feature vectors that can characterize the target laser point cloud is obtained. At the same time, for each feature contained in the target local map feature range, such as parking spaces and channels in the map, similar ideas are used to extract their feature vectors, that is, to analyze the local characteristics of the point cloud corresponding to these features and summarize the corresponding feature vectors, so that both sides have a set of feature vectors for comparison.

[0109] In the spatial dimension, the feature vectors of the converted target laser point cloud are compared and matched with the feature vectors of each feature in the feature range of the target local map. The specific approach is to check the feature vectors of the target laser point cloud and a feature in the map (such as the feature vector corresponding to a parking space) one by one to see how similar they are in spatial description. One way is to see whether the spatial position and shape they reflect are close. If the spatial shape, position and other features represented by the two feature vectors are very similar, then the distance between them will be relatively small, and it can be considered that the degree of match in the spatial dimension is high; conversely, if the difference is large, the distance is large and the degree of match is low. After such a comparison of the target laser point cloud and all relevant features in the map, different weights are assigned to each comparison result according to the importance of these features in the map. For example, the parking space feature is more important, the weight is higher, and the channel feature weight is slightly lower. Finally, all comparison results are combined according to the weights to calculate a total value. This value is the third matching degree between the target laser point cloud and the feature vector in the spatial dimension. It can overall reflect the degree of fit between the converted target laser point cloud in terms of spatial position, shape, etc. and the various features in the target local map feature range.

[0110] In terms of attitude dimension, find out the part that can reflect attitude information from the feature vector just extracted, such as analyzing from the feature vector which direction the target laser point cloud is tilted, how much it is tilted to the left or right, or how much it is pitched up or down, etc., and similarly obtain the situation of the target laser point cloud in the attitude angles of roll, pitch, and yaw. Similarly, perform such attitude analysis on the feature vectors of each feature in the feature range of the local map of the target, and then compare the situation of the attitude angles on both sides to see if they are consistent or close. For example, if the yaw angle of the target laser point cloud is very close to the yaw angle of a channel feature in the map, it means that the attitude of yaw is well matched. After such comparison of the three attitude angles of roll, pitch, and yaw, weights are assigned according to the importance of each attitude angle in the overall attitude matching, and these comparison results are combined to calculate a value. This value is the fourth matching degree between the target laser point cloud and the feature vector in the attitude dimension, which can reflect the consistency level of the converted target laser point cloud in attitude and the local map features of the target.

[0111] The third matching degree reflects the matching situation in the spatial dimension. Generally speaking, the higher the matching degree means the smaller the error in the spatial dimension. The spatial dimension error can be measured in this way. If you have known what kind of spatial position deviation range different third matching degrees correspond to through a lot of tests or experience accumulation, you can find the corresponding spatial position error range according to the currently calculated third matching degree to determine the size of the spatial dimension error. Or in a more intuitive way, looking back at the calculation of the third matching degree before, it is calculated by comparing the similarity of the target laser point cloud and the map feature vector in spatial features. Then these values ​​used to measure the similarity (such as the value indicating the distance) can be comprehensively counted, such as finding their average distance. This average distance value can roughly reflect the degree of deviation between the target laser point cloud and the target local map in the spatial dimension, that is, the spatial dimension error.

[0112] The fourth matching degree reflects the matching of the attitude dimension. Similar to the spatial dimension, the error size of the attitude dimension can also be inferred based on the matching degree. If there is a pre-set correspondence between the fourth matching degree and the attitude angle error, then the corresponding attitude angle error range is found based on the current fourth matching degree. If there is no such correspondence, look at the attitude angle comparison involved in the previous calculation of the fourth matching degree, such as the angle difference between the target laser point cloud and the map feature in the roll angle, and the difference in pitch and yaw angles. Combine these angle differences and calculate a value in a suitable way (such as finding the absolute value of the angle difference or other statistical methods). This value can represent the degree of deviation between the target laser point cloud and the target local map in the attitude dimension, that is, the attitude dimension error.

[0113] After determining the spatial dimension error and the attitude dimension error, combine them to get an indicator that can fully reflect the overall error between the converted target laser point cloud and the target local map. According to the importance of the spatial dimension and the attitude dimension in the entire matching process, you can give them a certain weight. For example, if you think that the spatial dimension and the attitude dimension are equally important, then give them a weight of 0.5. Then multiply the spatial dimension error by its corresponding weight, and then multiply the attitude dimension error by its corresponding weight. Finally, add these two results together. The resulting value is the comprehensive error. This comprehensive error can reflect the overall deviation between the converted target laser point cloud and the target local map.

[0114] The preset error range is set according to the specific needs of the actual use of the map and the requirements for map accuracy. For example, if this map is used in a high-precision autonomous driving scenario, the map accuracy requirements are very high, requiring the spatial position error to be controlled at the centimeter level, and the attitude angle error to be controlled within a few tenths of a degree, so that the autonomous driving vehicle can accurately drive and park according to the map. If it is used in an ordinary map navigation scenario, the accuracy requirements are not so high, the spatial position error is allowed to be at the decimeter level, and the attitude angle error is within a few degrees to meet the use requirements. These preset error range values ​​are generally determined after a large number of actual tests, combined with existing relevant standards in the industry, and analysis of actual application scenarios.

[0115] Compare the calculated comprehensive error with the preset error range. If the spatial position error in the comprehensive error is smaller than the preset spatial position error range, and the attitude angle error is also smaller than the preset attitude angle error range, it means that the matching degree between the converted target laser point cloud and the target local map has met the expected requirements. At this time, the currently updated posture transformation relationship is regarded as the target transformation relationship.

[0116] This application obtains the characteristic range of the target local map and filters out the target laser point cloud. This step can focus on effective data, remove irrelevant interference, and make subsequent processing more targeted. Then, the target laser point cloud is matched with the feature vector, and the first matching degree of the spatial dimension and the second matching degree of the posture dimension are obtained respectively. Based on this, the posture conversion relationship is updated, that is, precise adjustments are made from the two aspects of translation and rotation parameters to construct an updated posture conversion relationship. In this way, the posture conversion between maps can be more in line with the actual situation and the accuracy of map fusion can be improved. And further use the updated posture conversion relationship to repeatedly convert and match the target laser point cloud again, determine the third and fourth matching degrees to measure the error, until the error reaches the preset range to obtain the target conversion relationship. The target conversion relationship finally obtained can highly accurately reflect the correct correspondence between different maps in space and posture, greatly improving the accuracy of map splicing and fusion, ensuring the high-quality integration of map data, and providing a reliable and accurate map foundation for subsequent application scenarios such as memory parking.

[0117] Step S104, the initial map and the target local map are spliced ​​based on the target conversion relationship to obtain the target map.

[0118] In an embodiment of the present application, all elements in the initial map (such as various geographical features, objects, etc. represented by point cloud data) are moved according to the translation parameters in the target transformation relationship. For example, if the target transformation relationship indicates that the initial map as a whole needs to be translated to the right by a certain distance in the x direction, then each point cloud point in the initial map, each coordinate point representing an object, etc., are moved along the x-axis in the corresponding direction by the specified distance. Similarly, for the y and z directions (if a three-dimensional space is involved), the corresponding coordinates are also moved according to the corresponding translation parameters, so that the initial map is initially aligned with the target local map in space, just like moving two puzzle pieces to roughly corresponding positions on a plane.

[0119] Next, the initial map is adjusted in attitude based on the rotation parameters in the target transformation relationship. Take the yaw angle as an example. If the target transformation relationship requires the initial map to rotate clockwise around a vertical axis (similar to the axis around which a car turns) by a certain angle, then all elements in the initial map are rotated according to this requirement to change their orientation so that they are more in line with the attitude requirements for splicing with the target local map. Similar operations are performed for roll and pitch angles, which are rotated around the corresponding axes to the corresponding angles so that the attitude angles of the initial map in three-dimensional space can match those of the target local map, just like adjusting the angles of the two puzzle pieces to fit together.

[0120] After coordinate translation and attitude rotation, although the initial map as a whole has been aligned with the target local map in the general direction, it is still necessary to further carefully match the key features in the map to ensure the accuracy of the stitching. For example, the parking space boundaries, channel directions and other features in the initial map should be compared with the corresponding parking spaces, channels and other features in the target local map to see whether their positions, shapes, angles, etc. can accurately correspond. If it is found that some features still have slight deviations, it may be necessary to re-check the target conversion relationship or fine-tune the local area so that these key features can perfectly overlap, just like carefully comparing each pattern when doing a puzzle to ensure a perfect fit.

[0121] In addition to key features, we also need to pay attention to some details in the map, such as some small obstacles and special signs that may exist in the map. These details should also be matched and integrated so that the initial map and the target local map can be naturally and smoothly connected in all aspects. For example, if there is a small street lamp pole sign in the initial map, it is necessary to ensure that it is consistent with the position of the street lamp pole and the surrounding environment in the target local map after splicing, to avoid abrupt or unreasonable situations, and to make the entire spliced ​​map look as a complete and unified whole as much as possible.

[0122] After it is determined that the initial map and the target local map are accurately aligned in coordinates, posture, and various features and details, data fusion operations can be performed. All data in the initial map and the target local map (including point cloud data, attribute information, etc.) are merged together according to the corresponding positions and rules to form a new data set. For example, the point cloud data in the initial map is added to the spatial area corresponding to the target local map, and the relevant information such as geographic attributes and object categories carried by each of them are integrated, so that these data together constitute a richer and more complete map data set, just like fusing two pieces of puzzle pieces into a complete puzzle picture.

[0123] Finally, based on the merged data, a complete and directly usable target map is constructed through the corresponding map generation algorithm or processing flow. This target map not only contains the geographic information originally covered by the initial map and the target local map, but also can accurately reflect the actual geographical environment of the entire area after precise splicing, providing a reliable map foundation for subsequent application scenarios such as path planning, navigation, geographic analysis, etc.

[0124] This application first obtains the starting point of the initial map being constructed, and can accurately determine the starting reference for map construction. Then, the local map of the target that is closest to the starting point of the map is found, which can efficiently utilize the existing map resources in the surrounding area, avoid repeated drawing of similar areas, and effectively save storage space and computing resources. The laser point cloud of the initial map is then converted to the target local map and feature matching is performed to obtain an accurate target conversion relationship, thereby ensuring the accuracy of the connection between different maps. Finally, the target map is spliced ​​out based on this, so that the scattered maps can be integrated. During subsequent automatic parking, the route can be clearly determined based on the integrated target map, and it is no longer disturbed by multiple maps, solving the problem of difficulty in determining the route.

[0125] In the embodiment of the present application, after the initial map and the target local map are spliced ​​based on the target conversion relationship to obtain the target map, Figure 2 As shown, the method also includes:

[0126] Step S201, obtaining the user's map usage preference or map usage requirement.

[0127] In an embodiment of the present application, the system will present a special interactive interface to the user, which can be a settings page on a mobile application, or a specific functional area on a web page, etc. The interface will guide users to express their map usage preferences or needs in a variety of ways. For example, in the form of a list of options, users are asked to check their favorite map display style, such as simple style, detailed style, or preferred map color scheme, such as light mode for easy viewing during the day, dark mode for night use, etc.; some text input boxes will also be set up to facilitate users to directly enter their specific usage needs, such as requiring the map to accurately display the real-time flow of people in a certain type of place (such as parks, hospitals, etc.), or hoping that the map has the function of displaying historical change information in a specific area, etc.

[0128] In addition to the information that users actively fill in or select, the system will also conduct an in-depth analysis of the user's past map usage records. Check the map areas that users have frequently viewed, the map functions that they frequently use (for example, whether they always use the navigation function to go to certain types of places, such as restaurants and shopping malls), and the content that they stay on the map for a long time. By analyzing these usage habits, we can summarize the user's potential map usage preferences. For example, if we find that the user has viewed hiking routes in the mountains many times, we can infer that the user has a preference for map functions and information related to outdoor hiking.

[0129] At the same time, the system will also collect feedback from other channels to further clarify the user's map usage preferences or needs. For example, the suggestions for improving map functions left by users in the evaluation section of the application, or the relevant feedback on map usage experience in the questionnaire surveys participated in, etc., by integrating these scattered information and conducting comprehensive analysis, the user's map usage preferences or map usage needs can be obtained more comprehensively and accurately.

[0130] Step S202: obtaining a functional module corresponding to a map usage preference or a map usage requirement.

[0131] In the embodiment of the present application, a function module database is pre-built, which covers a variety of map-related function modules, and each function module has a detailed description and classification. For example, according to the function type, it is divided into navigation function modules (including navigation functions for different travel modes such as driving navigation, walking navigation, and cycling navigation), information display function modules (such as displaying real-time traffic conditions, recommendations for surrounding popular attractions, air quality information, etc.), map editing function modules (supporting users to mark locations, draw routes, etc.), etc. Each function module has corresponding code implementation and related interfaces to facilitate subsequent addition to the target map.

[0132] Based on the user's map usage preferences or needs, the system will search and match in the functional module database. If the user indicates that he has a preference for outdoor adventure maps, the system will filter out corresponding functional modules such as displaying contour lines, marking water sources, and providing outdoor camping site recommendations; if the user's needs are to facilitate the search for parking information in the city, the system will locate relevant functional modules such as real-time parking space query and parking fee standard display. This matching process is achieved by extracting keywords from user needs and comparing text with functional module descriptions, ensuring that the found functional modules can accurately meet user expectations.

[0133] After determining the preliminary matching function modules, it is necessary to further verify whether these function modules are applicable to the current target map. Factors to be considered include the coverage of the target map (for example, some function modules are only applicable to maps of specific urban areas, and cannot be added if the target map does not cover this area), the scale accuracy of the map (for example, some high-precision measurement function modules require the map to have a sufficiently fine scale to support, and if the target map accuracy does not meet the requirements, it is not suitable) and the type of map (for example, there are differences in the function modules applicable to indoor maps and outdoor maps). After such applicability verification, the accurate list of function modules that can be added to the target map is finally determined.

[0134] Step S203, adding the functional module to the target map.

[0135] In the embodiment of the present application, first, for each functional module to be added, ensure that its related code files, configuration files, and dependent resources (such as pictures, data files, etc.) are all in an available state. Check whether the interface of the functional module is compatible with the system architecture of the target map. If there is incompatibility, some adaptation and adjustment work is required, such as modifying interface parameters, updating data transmission format, etc., so that it can interact and integrate with the target map smoothly.

[0136] According to the development framework and architecture design of the target map, the prepared functional modules are added one by one. For example, if the target map is developed based on the framework of the web map, the corresponding code of the functional module is embedded into the script file of the web page through the corresponding front-end and back-end code integration methods, and appropriate space is reserved for the display and operation of the functional module on the page layout, such as adding navigation bar buttons, information display windows and other elements; if it is a mobile map application, the code of the functional module is integrated into the code engineering of the application, and according to the mobile interface design specifications, the operation interface of the functional module is reasonably laid out on the mobile phone screen for user convenience.

[0137] After adding the functional modules, the entire target map should be fully tested to see whether the newly added functional modules can operate normally and whether they can work together with the original functions and data of the target map without conflicts or abnormalities. For example, when testing the added real-time traffic conditions functional module, check whether the map can accurately display the traffic congestion of each road section, and whether the display and data update of the functional module are normal when switching the map zoom level or switching different areas. If problems are found during the test, such as functional module freezing, display abnormalities, or errors in interaction with other functions, the corresponding code and configuration need to be adjusted and optimized until the functional module can function stably and accurately in the target map to meet the user's map usage preferences and needs.

[0138] This application first obtains the user's map usage preferences or needs, which is like building a cornerstone for subsequent personalized services. Whether it is a user who is keen on outdoor adventures and desires the map to present detailed terrain, water sources and other information, or an urban commuter who urgently needs real-time traffic congestion and parking space guidance, it can be accurately captured. Then, based on the information obtained, the corresponding functional modules are matched to quickly and accurately screen out tools that meet the needs, such as locking in an accurate cycling navigation route planning module for cycling enthusiasts, and fishing out a module for introducing historical sites in specific areas and recommending tour routes for history and culture enthusiasts. Finally, the selected functional modules are added to the target map. In this way, different users can have a set of functions that fit their own usage scenarios, which improves the convenience of use and deeply meets the needs of personalized navigation, exploration, information query, etc. in multiple scenarios.

[0139] In this embodiment, a parking map generation device is also provided, which is used to implement the above embodiments and preferred implementations, and will not be repeated here. As used below, the term "module" may be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also and is conceived.

[0140] This embodiment provides a parking map generation device, such as Figure 3 As shown, including:

[0141] The first acquisition module 301 is used to acquire the map construction starting point of the initial map currently being constructed;

[0142] The second acquisition module 302 is used to acquire a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point;

[0143] The conversion module 303 is used to convert the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and use the converted laser point cloud to perform feature matching with the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map;

[0144] The splicing module 304 is used to splice the initial map and the target local map based on the target conversion relationship to obtain the target map.

[0145] In the embodiment of the present application, the second acquisition module 302 is used to construct a search area with the mapping starting point as the center and a preset distance as the radius; determine whether there is a local map in the search area; if there is a local map in the search area, the local map closest to the mapping starting point is used as the target local map.

[0146] In the embodiment of the present application, the conversion module 303 is used to obtain the first global pose data collected at the starting point of the map construction process; obtain the second global pose data of the target local map; based on the first global pose data and the second global pose data, construct a pose conversion relationship between the initial map and the local map; based on the pose conversion relationship, convert the laser point cloud of the initial map to the target local map to obtain the converted laser point cloud.

[0147] In the embodiment of the present application, the conversion module 303 is used to obtain the map feature range corresponding to the target local map, and use the laser point cloud within the map feature range in the converted laser point cloud as the target laser point cloud; match the target laser point cloud with the feature vector in the map feature range to obtain a first matching degree in the spatial dimension and a second matching degree in the posture dimension between the target laser point cloud and the feature vector; update the posture transformation relationship using the first matching degree and the second matching degree to obtain an updated posture transformation relationship; transform the target laser point cloud using the updated posture transformation relationship until the error between the converted target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range to obtain the target transformation relationship.

[0148] In an embodiment of the present application, the conversion module 303 is used to adjust the translation parameters in the posture conversion relationship using the first matching degree to obtain updated translation parameters; adjust the rotation parameters in the posture conversion relationship using the second matching degree to obtain updated rotation parameters; and construct an updated posture conversion relationship based on the updated translation parameters and the updated rotation parameters.

[0149] In the embodiment of the present application, the conversion module 303 is used to convert the target laser point cloud using the updated posture conversion relationship to obtain the converted target laser point cloud; match the converted target laser point cloud with the feature vector in the map feature range to obtain a third matching degree between the target laser point cloud and the feature vector in the spatial dimension and a fourth matching degree in the posture dimension; determine the error between the converted target laser point cloud and the target local map based on the third matching degree and the fourth matching degree; if the error reaches a preset range, use the updated posture conversion relationship as the target conversion relationship.

[0150] In an embodiment of the present application, the device also includes: a processing module for obtaining a user's map usage preference or map usage requirement; obtaining a functional module corresponding to the map usage preference or map usage requirement; and adding the functional module to a target map.

[0151] See also Figure 4 , Figure 4 is a schematic diagram of the structure of an electronic device provided by an optional embodiment of the present invention, such as Figure 4As shown, the electronic device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the electronic device, including instructions stored in or on the memory to display graphical information of the 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 (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0152] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0153] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0154] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created by the use of an electronic device based on the presentation of a small program landing page, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0155] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0156] The electronic device further comprises a communication interface 30 for the electronic device to communicate with other devices or a communication network.

[0157] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method 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 storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also 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 a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0158] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for generating a parking map, characterized in that: The method comprises: Get the starting point of the initial map currently being constructed; Acquire a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point; Converting the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and performing feature matching between the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map; The initial map and the target local map are spliced ​​based on the target conversion relationship to obtain a target map.

2. The method according to claim 1, characterized in that The obtaining of the local map of the target adjacent to the mapping starting point includes: The search area is constructed by taking the mapping starting point as the center and the preset distance as the radius; determining whether a local map exists in the search area; If the local map exists in the search area, the local map closest to the map building starting point is used as the target local map.

3. The method according to claim 1, characterized in that The step of converting the laser point cloud corresponding to the initial map to the target local map to obtain the converted laser point cloud includes: Acquire first global pose data collected at the mapping starting point during the mapping process; Acquire second global pose data of the target local map; Based on the first global pose data and the second global pose data, constructing a pose conversion relationship between the initial map and the local map; Based on the posture conversion relationship, the laser point cloud of the initial map is converted to the target local map to obtain a converted laser point cloud.

4. The method according to claim 3, characterized in that The method of performing feature matching using the converted laser point cloud and the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map includes: Acquire a map feature range corresponding to the target local map, and use the laser point cloud within the map feature range in the converted laser point cloud as the target laser point cloud; Matching the target laser point cloud with a feature vector in the map feature range to obtain a first matching degree in a spatial dimension and a second matching degree in a posture dimension between the target laser point cloud and the feature vector; Using the first matching degree and the second matching degree to update the posture conversion relationship, to obtain an updated posture conversion relationship; The target laser point cloud is transformed using the updated posture transformation relationship until the error between the transformed target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range, thereby obtaining the target transformation relationship.

5. The method according to claim 4, characterized in that The updating of the posture conversion relationship by using the first matching degree and the second matching degree to obtain an updated posture conversion relationship includes: Using the first matching degree to adjust the translation parameters in the posture transformation relationship to obtain updated translation parameters; adjusting the rotation parameters in the posture transformation relationship using the second matching degree to obtain updated rotation parameters; An updated pose transformation relationship is constructed based on the updated translation parameters and the updated rotation parameters.

6. The method according to claim 4, characterized in that The target laser point cloud is converted by using the updated posture conversion relationship until the error between the converted target laser point cloud and the feature vector in the spatial dimension and the posture dimension reaches a preset range, and the target conversion relationship is obtained, including: The target laser point cloud is transformed using the updated posture transformation relationship to obtain a transformed target laser point cloud; Matching the converted target laser point cloud with the feature vector in the map feature range to obtain a third matching degree in the spatial dimension and a fourth matching degree in the posture dimension between the target laser point cloud and the feature vector; Determine an error between the converted target laser point cloud and the target local map based on the third matching degree and the fourth matching degree; If the error reaches a preset range, the updated posture transformation relationship is used as the target transformation relationship.

7. The method according to claim 1, characterized in that After the initial map and the target local map are spliced ​​based on the target conversion relationship to obtain the target map, the method further includes: Obtain the user's map usage preferences or map usage needs; Acquire a functional module corresponding to the map usage preference or the map usage requirement; Add the functional module to the target map.

8. A parking map generation device, characterized in that: The device comprises: The first acquisition module is used to obtain the starting point of the initial map currently being constructed; A second acquisition module is used to acquire a target local map adjacent to the mapping starting point, wherein the target local map is a local map with the shortest distance from the mapping starting point; A conversion module, used to convert the laser point cloud corresponding to the initial map to the target local map to obtain a converted laser point cloud, and use the converted laser point cloud to perform feature matching with the point cloud of the target local map to obtain a target conversion relationship between the initial map and the local map; A splicing module is used to splice the initial map and the target local map based on the target conversion relationship to obtain a target map.

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