Map construction method and device

Abstract

The invention discloses a map construction method and device, and relates to the technical field of computers. According to one specific embodiment, the method comprises the steps that data collected by a plurality of vehicles are processed, track data corresponding to each vehicle are generated, and the data comprise point cloud data; determining an absolute position of each point cloud point in the point cloud data based on a positioning system, and performing multi-track registration on track data corresponding to each vehicle based on the absolute position to generate an initial map track; and generating a positioning track based on the absolute position of each point cloud point and the signal intensity of each point cloud point, and performing track alignment on the positioning track and the initial map track to construct a map. According to the embodiment, the data acquisition cost can be saved, and the expansion speed of the unmanned distribution station can be improved; and multi-track registration is performed on the track data of the multiple vehicles based on the absolute position of each point, so that the data splicing effect is improved, and the map precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for constructing a map. Background Art

[0002] High-precision maps are essential elements for highly automated driving. In order to promote autonomous driving areas, the scope of high-precision maps needs to be continuously expanded. For example, in the scenario of unmanned delivery of terminal logistics, in the process of expanding the scope and number of unmanned delivery sites of terminal logistics, it is necessary to quickly build high-precision maps of unmanned delivery sites. The production of existing high-precision maps is usually based on the data collection of the autonomous driving range of terminal logistics unmanned delivery vehicles by dedicated data collection vehicles, and then based on SLAM (Simultaneous Localization and Mapping) technology to stitch the laser radar point cloud data, process the stitched point cloud data, and then produce high-precision maps. At the same time, there are also data collection based on unmanned delivery vehicles, using GPS (Global Positioning System), inertial navigation, wheel speed meter and laser radar, relying on SLAM technology to stitch point clouds, and then produce high-precision maps.

[0003] In the process of implementing the present invention, the inventors found that the data collection mode based on ordinary vehicles such as unmanned delivery vehicles has a slow data collection speed, poor collection quality, and poor splicing effect, which limits the map accuracy; although the data collection and splicing effects of special data collection vehicles are better than those of unmanned delivery vehicles, special data collection vehicles are of higher value, limited in number, and have a high transportation cost-effectiveness, which limits the expansion speed of unmanned delivery sites. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and device for map construction, which can use multiple ordinary vehicles such as unmanned delivery vehicles to collect data at the same time, without the need to use special data collection vehicles for data collection, thereby saving data collection costs and increasing the expansion speed of unmanned delivery sites; based on the absolute position of each point, the trajectory data of multiple vehicles are multi-track aligned, thereby improving the data splicing effect and improving the map accuracy.

[0005] To achieve the above objective, according to one aspect of an embodiment of the present invention, a method for constructing a map is provided, comprising: Processing the data collected from the multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data; Determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position to generate an initial map trajectory; A positioning trajectory is generated based on the absolute position of each point cloud point and the signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct a map.

[0006] Optionally, before performing multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position, the method further includes: using a radar semantic segmentation model to obtain a semantic label for each point cloud point in the point cloud data; constructing a physical model according to the semantic labels of each point cloud point; constructing an optimization equation according to the physical model, and optimizing the trajectory data corresponding to each vehicle based on the optimization equation.

[0007] Optionally, the entity model includes an object model and a plane model; before constructing the optimization equation according to the entity model, it also includes: based on the plane model merging strategy, merging two plane models whose distance in the normal direction of the plane model is less than a first threshold and whose normal angle is less than a second threshold into one plane model.

[0008] Optionally, multi-track registration is performed on the trajectory data corresponding to each vehicle based on the absolute position, including: for two trajectory data corresponding to any two vehicles, judging whether the two trajectory data need to be multi-track registered according to the absolute position; when it is determined that multi-track registration is needed, selecting a specified number of key frames on the two trajectory data, and obtaining the absolute positions of the specified number of key frames; based on the absolute positions of the specified number of key frames, using a point cloud registration method to perform multi-track registration on the two trajectory data.

[0009] Optionally, based on the absolute positions of the specified number of key frames, a point cloud registration method is used to perform multi-track registration on the two trajectory data, including: based on the absolute positions of the specified number of key frames, performing multiple multi-track registrations on the two trajectory data through point cloud registration; performing accuracy evaluation based on the registration results of the multiple multi-track registrations, and optimizing the registration results based on the evaluation results.

[0010] Optionally, accuracy evaluation is performed based on the registration results of multiple multi-track registrations, including: obtaining the relative posture transformation matrix of the two track data according to the registration results of each multi-track registration in the registration results of the multi-track registration; based on the initial relative posture transformation matrix between the two track data, calculating the error value between the initial relative posture transformation matrix and the relative posture transformation matrix to obtain the error value corresponding to the registration result of the multi-track registration; judging the accuracy of the registration results of the two track data according to the error value corresponding to the registration results of the multiple multi-track registrations.

[0011] Optionally, judging the accuracy of the registration results of the two track data according to the error values ​​corresponding to the registration results of the multiple multi-track registrations, includes: obtaining a sample set consisting of the error values ​​corresponding to the registration results of the multiple multi-track registrations; obtaining a change in the error value by counting the covariance states of six degrees of freedom of translation and rotation corresponding to the sample set; and judging the accuracy of the registration results of the two track data according to the change in the error value.

[0012] Optionally, the positioning trajectory and the initial map trajectory are aligned to construct a map, including: performing trajectory least squares estimation on the positioning trajectory and the initial map trajectory to obtain an offset value of the positioning trajectory and the initial map trajectory; using the offset value to align the positioning trajectory and the initial map trajectory to obtain the final absolute position of the initial map trajectory to construct a map.

[0013] According to another aspect of an embodiment of the present invention, there is provided a map construction device, comprising: A trajectory generation module, used to process the data collected by multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data; A track registration module, used to determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the track data corresponding to each vehicle based on the absolute position to generate an initial map track; The trajectory alignment module is used to generate a positioning trajectory based on the absolute position of each point cloud point and the signal strength of each point cloud point, and to align the positioning trajectory with the initial map trajectory to construct a map.

[0014] According to another aspect of an embodiment of the present invention, there is provided an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the map construction method provided by an embodiment of the present invention.

[0015] According to another aspect of the embodiments of the present invention, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the map construction method provided by the embodiments of the present invention is implemented.

[0016] According to another aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the map construction method provided by the embodiment of the present invention is implemented.

[0017] An embodiment of the above invention has the following advantages or beneficial effects: by processing the data collected from multiple vehicles respectively, the trajectory data corresponding to each vehicle is generated, and the data includes point cloud data; the absolute position of each point cloud point in the point cloud data is determined based on the positioning system, and the trajectory data corresponding to each vehicle is multi-track aligned based on the absolute position to generate the initial map trajectory; the positioning trajectory is generated based on the absolute position of each point cloud point and the signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct the map. The technical solution can be based on multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, and after the data collection is completed, synchronous data processing is performed to generate the trajectory data of each vehicle, and then based on the trajectory data of multiple vehicles, multi-track alignment is performed to complete the multi-vehicle data splicing, and finally, the absolute position of each point cloud point determined by the positioning system is combined to align the trajectory to construct the map. The present invention uses multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, and there is no need to use a dedicated data collection vehicle to collect data, so that the data collection cost can be saved and the expansion speed of unmanned delivery sites can be improved; based on the absolute position of each point, the trajectory data of multiple vehicles are multi-track aligned, which improves the data splicing effect and the map accuracy.

[0018] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention. Figure 1 is a schematic diagram of main steps of a method for constructing a map according to an embodiment of the present invention; Figure 2 is a schematic diagram of the implementation principle of a map construction system according to an embodiment of the present invention; Figure 3 is a schematic diagram of main modules of a map construction device according to an embodiment of the present invention; Figure 4 is an exemplary system architecture diagram to which embodiments of the present invention may be applied; Figure 5 It is a schematic diagram of the structure of a computer system of a terminal device or a server suitable for implementing an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0021] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage and other aspects of user personal information involved in the technical solution disclosed in the present invention are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security and national security.

[0022] In order to solve the problems existing in the prior art, the present invention proposes a map construction method, which can collect data based on multiple unmanned delivery vehicles at the same time, perform synchronous data processing based on SLAM technology after data collection is completed to generate trajectory data for each vehicle, and then perform multi-track alignment based on the trajectory data of multiple vehicles to complete multi-vehicle data splicing, and finally align the trajectories in combination with the absolute positions of each point cloud point determined by the positioning system to construct a map. The present invention solves the splicing of multi-vehicle data, and can also perform splicing quality assessment, perform absolute position estimation based on the positioning results of point cloud data collected by multiple vehicles, and improve relative positioning accuracy based on local cluster adjustment. Among them, the positioning system in the present invention is, for example, the Beidou satellite navigation system, the global positioning system, and the like.

[0023] Figure 1 FIG. 1 is a schematic diagram of the main steps of the method for constructing a map according to an embodiment of the present invention. Figure 1 As shown, in the embodiment of the present invention, the map construction method mainly includes the following steps S101 to S103.

[0024] Step S101: Process the data collected by multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data. The trajectory data includes the position of each point constituting the trajectory and the posture information of the radar or camera at each point. In an embodiment of the present invention, ordinary vehicles such as unmanned delivery vehicles can be used for data collection, and there is no need to use special data collection vehicles for data collection, thereby saving data collection costs and increasing the expansion speed of unmanned delivery sites. In an embodiment of the present invention, in addition to point cloud data, the data collected by the vehicle may also include image data of the surrounding environment, and location information at the corresponding time.

[0025] According to an embodiment of the present invention, when processing data collected by multiple vehicles, specifically, the data collected by each vehicle can be processed separately and simultaneously in a multi-threaded manner to generate trajectory data corresponding to each vehicle. Wherein, when generating trajectory data, a time sequence window can be dynamically maintained, wherein the number of key frames included in the window is N, and the splicing result of these N frames of data is used as a local reconstructed map. This local reconstruction result has better freshness than the high-precision map produced offline. Then fill in new frames in time sequence and discard old frames, use this local reconstructed map to estimate the posture of the radar or camera of the N+1th frame, and then insert the N+1th frame as a new frame into the local reconstructed map, and discard the oldest frame of data in the window. Specifically, the data of the inertial navigation and wheel speed meter can be used to integrate the change between consecutive frames for motion compensation, and at the same time as the initial value of the inter-frame posture to generate the trajectory data corresponding to the vehicle.

[0026] According to one embodiment of the present invention, after the trajectory data corresponding to each vehicle is generated, before the trajectory data corresponding to each vehicle is multi-track registered based on the absolute position in step S102, the trajectory data corresponding to each vehicle may be optimized. Specifically, the optimization process is: using a radar semantic segmentation model to obtain the semantic label of each point cloud point in the point cloud data; constructing a physical model according to the semantic labels of each point cloud point; constructing an optimization equation according to the physical model, and optimizing the trajectory data corresponding to each vehicle based on the optimization equation. The physical model includes an object model and a plane model. In addition, before constructing the optimization equation according to the physical model, it also includes: based on a plane model merging strategy, two plane models whose distance in the normal direction of the plane model is less than a first threshold and whose normal angle is less than a second threshold are merged into one plane model.

[0027] According to an embodiment of the present invention, the trajectory data corresponding to each vehicle can be optimized based on the cost equation of the constructed local window bundle adjustment optimization. Bundle adjustment, also known as BA optimization, is a technology for extracting the optimal 3D model in reconstruction. Its purpose is to adjust the observation posture and the spatial position of the feature point so that the light rays (bundles of light rays) reflected from each feature point can converge to the observation optical center after adjustment. This process involves optimizing the spatial coordinates of landmark points under multiple poses and postures.

[0028] Specifically, after generating the trajectory data corresponding to each vehicle, the present invention uses the Lidar Semanticsegmentation model (a prior art) to obtain the semantic label of each point cloud point in the point cloud frame emitted by the radar, and constructs an entity model according to the semantic label of each point cloud point. For example, if the label is a rod-shaped object (tree trunk, telephone pole, fire hydrant, etc.), a columnar object model (Model ( , , r), the cylindrical object model contains any point on the center line of the column , the centerline unit direction vector , and column radius information r); if the label is a building (such as a building, a sentry box, a temporary building), a plane model (Model ( , ), the model contains any point on the plane , and the unit normal vector of the plane ). According to the above columnar object model and plane model, by constructing the least squares optimization problem about the columnar object model and the plane model, the cost equation of the cluster optimization adjustment is constructed as follows: Among them, the optimization variables include the corresponding rotation information R, translation information t and all corresponding plane models ( , ) and columnar object model parameters ( , , r); represents all point cloud points and their labels in the current frame that are labeled as rods and are used to construct the columnar object model, and M is the number of points labeled as rods in the current frame; represents all point cloud points and their labels in the current frame that are labeled as buildings and used to construct the plane model. N is the number of points labeled as buildings in the current frame. k represents the features for constructing a columnar object model or a plane model. O represents the number of features for constructing a columnar object model. P represents the number of features for constructing a plane model.

[0029] The first part of the above formula This describes the distance formula from a point to a columnar object, which is also the cost function of the columnar object model; Part II It describes the distance formula from a point to a plane, which is also the cost function of the plane model.

[0030] In addition, in order to maintain the scale of the number of models, the present invention uses a plane model merging strategy, that is, when the normal vector angle of adjacent plane models is less than a certain threshold, the two planes are considered to belong to the same building edge, and the two plane models will be merged. When the distance between the positions of the two plane models in the normal direction is less than a certain threshold, the two plane models are considered to be adjacent plane models. The normal vector angle of the two plane models is calculated according to the following formula: ; Among them, n 1 and n 2 are the normal vectors of the two plane models.

[0031] The normal vector n of the merged plane model is calculated according to the following formula: ; Among them, n 1 and n 2 are the normal vectors of the two plane models, and p and q represent the number of observation points selected by the two plane models.

[0032] Step S102: Determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position to generate an initial map trajectory. In an embodiment of the present invention, the positioning system is, for example, the Beidou satellite navigation system (hereinafter referred to as the Beidou system), and the Beidou system and real-time dynamic measurement technology are used as the initial value of the absolute position.

[0033] According to one embodiment of the present invention, performing multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position may specifically include: for two trajectory data corresponding to any two vehicles, judging whether the two trajectory data need to be registered with multiple tracks according to the absolute position; if it is determined that multi-track registration is needed, selecting a specified number of key frames on the two trajectory data, and obtaining the absolute positions of the specified number of key frames; based on the absolute positions of the specified number of key frames, performing multi-track registration on the two trajectory data using a point cloud registration method.

[0034] According to the absolute position of each point cloud point in the point cloud data, the absolute position of the trajectory data corresponding to each vehicle can be obtained, and then, it can be determined whether the trajectory data needs to be aligned according to the absolute position of the vehicle's trajectory data. Specifically, in one embodiment, for two trajectory data corresponding to any two vehicles, samples are taken from the two trajectory data respectively by sampling, and the distance between the same point on the two trajectory data obtained by sampling is calculated. When the distance is less than a set threshold, it is determined that the two trajectory data need to be aligned.

[0035] Then, key frames are selected on the two trajectories, and the point cloud registration method is used to obtain the registration results of the two trajectories. The point cloud registration method can use the ICP (Iterative Closest Point, ICP algorithm is based on data registration method, using the nearest point search method to solve an algorithm based on free-form surfaces) algorithm or NDT (Normal Distributions Transform, an algorithm for point cloud registration, the core idea of ​​which is to discretize the space into squares or cubes, and then divide the point cloud into different grids, and describe the local characteristics of the point cloud, such as shape and direction, by constructing a normal distribution model of the point cloud in each grid) algorithm. In the algorithm practice, the range-based enumeration search method can also be considered according to the error range of the absolute position. Since the point cloud registration method is a conventional method, it will not be repeated here. Here, when selecting the key frame, specifically, the point cloud data frames with a specified number of points before and after the previous sampling point can be used as key frames.

[0036] According to another embodiment of the present invention, based on the absolute positions of the specified number of key frames, the two track data are multi-track registered using a point cloud registration method, which may specifically include: based on the absolute positions of the specified number of key frames, the two track data are multi-track registered multiple times by point cloud registration; accuracy evaluation is performed based on the registration results of the multiple multi-track registrations, and the registration results are optimized based on the evaluation results. That is to say, when performing multi-track registration, multiple registrations may be performed, and then the accuracy of the registration is evaluated based on the registration results of the multiple registrations, so as to select the registration result with the best evaluation result as the final registration result, so as to optimize the registration result.

[0037] According to an embodiment of the present invention, performing accuracy evaluation based on the registration results of multiple multi-track registrations may specifically include: obtaining a relative posture transformation matrix of the two track data according to the registration result of each multi-track registration in the registration results of the multi-track registration; calculating an error value between the initial relative posture transformation matrix and the relative posture transformation matrix based on an initial relative posture transformation matrix between the two track data, to obtain an error value corresponding to the registration result of the multi-track registration; and judging the accuracy of the registration results of the two track data according to the error values ​​corresponding to the registration results of the multiple multi-track registrations.

[0038] In an embodiment of the present invention, judging the accuracy of the registration results of the two track data according to the error values ​​corresponding to the registration results of the multiple multi-track registrations may specifically include: obtaining a sample set consisting of the error values ​​corresponding to the registration results of the multiple multi-track registrations; obtaining a change in the error value by counting the covariance states of the six degrees of freedom of translation and rotation corresponding to the sample set; and judging the accuracy of the registration results of the two track data according to the change in the error value.

[0039] After obtaining the registration result of the two trajectories, that is, the relative pose transformation matrix After (the kth transformation from the gi trajectory to the gj trajectory), since there is a natural relative pose transformation matrix between the two trajectories , where O is the origin of the radar or camera coordinate system. Therefore, the error values ​​of the two pose transformation matrices can be obtained. .

[0040] There will be multiple relative pose changes between the two trajectories, and the The accuracy of the current two trajectory registrations is judged by the changes in . Specifically, the present invention judges the accuracy of trajectory registration by counting the covariance states of the six degrees of freedom of translation and rotation, wherein the covariances of the six degrees of freedom of translation and rotation can be calculated separately, and different thresholds are used to make accuracy judgments. Among them, the relative posture transformation is a matrix that represents six degrees of freedom, namely, translations in three directions (x, y, z) and rotations in three directions (roll, pitch, yaw), and the covariance statistics of these six degrees of freedom can be performed separately.

[0041] Through step S102, multi-track registration of the track data corresponding to each vehicle can be performed to generate an initial map track.

[0042] Step S103: Generate a positioning track based on the absolute position of each point cloud point and the signal strength of each point cloud point, and align the positioning track with the initial map track to build a map. After the initial map track is generated in step S102, the position information corresponding to the data collected by all vehicles and the signal strength of the point cloud data will be counted, and the collected point cloud data will be screened according to the signal strength, and the point cloud points at the position with high signal strength (exceeding the set threshold) will be used to form the positioning track. After that, the track alignment is performed based on the positioning track and the initial map track to build a map.

[0043] According to an embodiment of the present invention, aligning the positioning trajectory and the initial map trajectory to construct a map may specifically include: performing trajectory least squares estimation on the positioning trajectory and the initial map trajectory to obtain the offset value of the positioning trajectory and the initial map trajectory; aligning the positioning trajectory and the initial map trajectory using the offset value to obtain the final absolute position of the initial map trajectory to construct a map. Specifically, the least squares method constructs the error value through the corresponding posture points in the trajectory, and minimizes the error of the corresponding points in the two trajectories by optimizing the transformation state (R, t) of the overall trajectory. Wherein, R is the rotation information and t is the translation information.

[0044] Figure 2 FIG. 1 is a schematic diagram showing the implementation principle of a map construction system according to an embodiment of the present invention. Figure 2 As shown, in one embodiment of the present invention, taking data collection by three vehicles as an example, the map construction system first uses a plurality of trajectory generation modules in a multi-threaded manner to simultaneously process the data collected by the plurality of vehicles and generate trajectory data corresponding to each vehicle; then, based on the BeiDou system and real-time kinematic measurement technology (RTK), multi-track alignment and the accuracy evaluation of the multi-track alignment results are performed to generate an initial map trajectory; finally, the absolute position accuracy problem is processed in the joint optimization module, a positioning trajectory is generated based on the absolute position and signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct a map.

[0045] Figure 3 Schematic diagram of the main modules of the map construction device according to an embodiment of the present invention. Figure 3 As shown, in the embodiment of the present invention, the map construction device 300 mainly includes a trajectory generation module 301 , a trajectory registration module 302 and a trajectory alignment module 303 .

[0046] A trajectory generation module 301 is used to process the data collected by multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data; A track registration module 302 is used to determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the track data corresponding to each vehicle based on the absolute position to generate an initial map track; The trajectory alignment module 303 is used to generate a positioning trajectory based on the absolute position of each point cloud point and the signal strength of each point cloud point, and align the positioning trajectory with the initial map trajectory to construct a map.

[0047] According to one embodiment of the present invention, it also includes a trajectory data optimization module (not shown in the figure), which is used to: before performing multi-trajectory registration on the trajectory data corresponding to each vehicle based on the absolute position, obtain the semantic label of each point cloud point in the point cloud data using a radar semantic segmentation model; construct a physical model according to the semantic label of each point cloud point; construct an optimization equation according to the physical model, and optimize the trajectory data corresponding to each vehicle based on the optimization equation.

[0048] According to another embodiment of the present invention, the entity model includes an object model and a plane model; before constructing the optimization equation according to the entity model, the trajectory data optimization module (not shown in the figure) is also used to: based on the plane model merging strategy, merge two plane models whose distance in the normal direction of the plane model is less than a first threshold and whose normal angle is less than a second threshold into one plane model.

[0049] According to another embodiment of the present invention, the trajectory registration module 302 can also be used to: for two trajectory data corresponding to any two vehicles, determine whether the two trajectory data need to be multi-track registered according to the absolute position; if it is determined that multi-track registration is needed, select a specified number of key frames on the two trajectory data, and obtain the absolute positions of the specified number of key frames; based on the absolute positions of the specified number of key frames, use a point cloud registration method to perform multi-track registration on the two trajectory data.

[0050] According to another embodiment of the present invention, when the trajectory registration module 302 performs multi-track registration on the two trajectory data using a point cloud registration method based on the absolute positions of the specified number of key frames, it can be specifically used to: perform multiple multi-track registrations on the two trajectory data through point cloud registration based on the absolute positions of the specified number of key frames; perform accuracy evaluation based on the registration results of the multiple multi-track registrations, and optimize the registration results based on the evaluation results.

[0051] According to another embodiment of the present invention, when the trajectory registration module 302 performs accuracy evaluation based on the registration results of multiple multi-trajectory registrations, it can be specifically used to: for each registration result of the multi-trajectory registration in the registration results of the multi-trajectory registration, obtain the relative posture transformation matrix of the two trajectory data according to the registration result of the multi-trajectory registration; based on the initial relative posture transformation matrix between the two trajectory data, calculate the error value between the initial relative posture transformation matrix and the relative posture transformation matrix to obtain the error value corresponding to the registration result of the multi-trajectory registration; and judge the accuracy of the registration results of the two trajectory data according to the error value corresponding to the registration results of the multiple multi-trajectory registrations.

[0052] According to another embodiment of the present invention, when the trajectory registration module 302 judges the accuracy of the registration results of the two trajectory data according to the error values ​​corresponding to the registration results of the multiple multi-trajectory registrations, it can be specifically used to: obtain a sample set consisting of error values ​​corresponding to the registration results of the multiple multi-trajectory registrations; obtain a change in the error value by counting the covariance states of the six degrees of freedom of translation and rotation corresponding to the sample set; and judge the accuracy of the registration results of the two trajectory data according to the change in the error value.

[0053] According to another embodiment of the present invention, the trajectory alignment module 303 can also be used to: perform trajectory least squares estimation on the positioning trajectory and the initial map trajectory to obtain an offset value of the positioning trajectory and the initial map trajectory; use the offset value to align the positioning trajectory and the initial map trajectory to obtain the final absolute position of the initial map trajectory to construct a map.

[0054] According to the technical solution of the embodiment of the present invention, by processing the data collected by multiple vehicles respectively, the trajectory data corresponding to each vehicle is generated, and the data includes point cloud data; the absolute position of each point cloud point in the point cloud data is determined based on the positioning system, and the trajectory data corresponding to each vehicle is multi-track aligned based on the absolute position to generate the initial map trajectory; the positioning trajectory is generated based on the absolute position of each point cloud point and the signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct the technical solution of the map, which can be based on multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, after the data collection is completed, synchronized data processing is performed to generate the trajectory data of each vehicle, and then based on the trajectory data of multiple vehicles, multi-track alignment is performed to complete the multi-vehicle data splicing, and finally, the absolute position of each point cloud point determined by the positioning system is combined to align the trajectory to construct the map. The present invention uses multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, without the need to use a dedicated data collection vehicle to collect data, thereby saving data collection costs and improving the expansion speed of unmanned delivery sites; based on the absolute position of each point, the trajectory data of multiple vehicles are multi-track aligned, which improves the data splicing effect and improves the map accuracy.

[0055] Figure 4 An exemplary system architecture 400 is shown to which the method or apparatus for building a map according to an embodiment of the present invention may be applied.

[0056] like Figure 4As shown, system architecture 400 may include terminal devices 401, 402, 403, network 404 and server 405. Network 404 is used to provide a medium for communication links between terminal devices 401, 402, 403 and server 405. Network 404 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0057] Users can use terminal devices 401, 402, 403 to interact with server 405 through network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, 403, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0058] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0059] Server 405 may be a server that provides various services, such as a background management server that provides support for websites browsed by users using terminal devices 401, 402, and 403 (for example only). The background management server may process the data collected from multiple vehicles for the received map building request and other data, and generate trajectory data corresponding to each vehicle, the data including point cloud data; determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position to generate an initial map trajectory; generate a positioning trajectory based on the absolute position of each point cloud point and the signal strength of each point cloud point, align the positioning trajectory with the initial map trajectory to construct a map, and feed back the processing result (such as the constructed map - for example only) to the terminal device.

[0060] It should be noted that the map construction method provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the map construction device is generally set in the server 405 .

[0061] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0062] Reference below Figure 5 , which shows a schematic diagram of the structure of a computer system 500 of a terminal device or a server suitable for implementing an embodiment of the present invention. Figure 5The terminal device or server shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0063] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage part 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the system 500 are also stored. The CPU 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0064] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, a modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 510 as needed, so that a computer program read therefrom is installed into the storage section 508 as needed.

[0065] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-mentioned functions defined in the system of the present invention are executed.

[0066] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0067] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0068] The units or modules involved in the embodiments of the present invention may be implemented in software or hardware. The units or modules described may also be arranged in a processor. For example, they may be described as: a processor including a trajectory generation module, a trajectory registration module, and a trajectory alignment module. The names of these units or modules do not, in some cases, constitute limitations on the units or modules themselves. For example, the trajectory generation module may also be described as "a module for processing data collected from multiple vehicles respectively to generate trajectory data corresponding to each vehicle".

[0069] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: processing the data collected from multiple vehicles respectively to generate trajectory data corresponding to each vehicle, the data including point cloud data; determining the absolute position of each point cloud point in the point cloud data based on the positioning system, and performing multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position to generate an initial map trajectory; generating a positioning trajectory based on the absolute position of each point cloud point and the signal strength of each point cloud point, and aligning the positioning trajectory with the initial map trajectory to construct a map.

[0070] According to the technical solution of the embodiment of the present invention, by processing the data collected by multiple vehicles respectively, the trajectory data corresponding to each vehicle is generated, and the data includes point cloud data; the absolute position of each point cloud point in the point cloud data is determined based on the positioning system, and the trajectory data corresponding to each vehicle is multi-track aligned based on the absolute position to generate the initial map trajectory; the positioning trajectory is generated based on the absolute position of each point cloud point and the signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct the technical solution of the map, which can be based on multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, after the data collection is completed, synchronized data processing is performed to generate the trajectory data of each vehicle, and then based on the trajectory data of multiple vehicles, multi-track alignment is performed to complete the multi-vehicle data splicing, and finally, the absolute position of each point cloud point determined by the positioning system is combined to align the trajectory to construct the map. The present invention uses multiple unmanned delivery vehicles and other ordinary vehicles to collect data at the same time, without the need to use a dedicated data collection vehicle to collect data, thereby saving data collection costs and improving the expansion speed of unmanned delivery sites; based on the absolute position of each point, the trajectory data of multiple vehicles are multi-track aligned, which improves the data splicing effect and improves the map accuracy.

[0071] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for map construction, characterized in that: include: Processing the data collected from the multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data; Determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the trajectory data corresponding to each vehicle based on the absolute position to generate an initial map trajectory; A positioning trajectory is generated based on the absolute position of each point cloud point and the signal strength of each point cloud point, and the positioning trajectory and the initial map trajectory are aligned to construct a map.

2. The method according to claim 1, characterized in that Before performing multi-track registration on the track data corresponding to each vehicle based on the absolute position, the method further includes: Using a radar semantic segmentation model to obtain a semantic label for each point cloud point in the point cloud data; Building a physical model according to the semantic labels of each point cloud point; An optimization equation is constructed according to the entity model, and the trajectory data corresponding to each vehicle is optimized based on the optimization equation.

3. The method according to claim 2, characterized in that The entity model includes an object model and a plane model; before constructing the optimization equation according to the entity model, it also includes: Based on the plane model merging strategy, two plane models whose distance in the normal direction of the plane models is less than a first threshold and whose normal angle is less than a second threshold are merged into one plane model.

4. The method according to claim 1, characterized in that: Performing multi-track registration on the track data corresponding to each vehicle based on the absolute position includes: For two track data corresponding to any two vehicles, judging whether the two track data need to be multi-track aligned according to the absolute positions; When it is determined that multi-track registration is required, selecting a specified number of key frames on the two track data, and obtaining absolute positions of the specified number of key frames; Based on the absolute positions of the specified number of key frames, a point cloud registration method is used to perform multi-track registration on the two track data.

5. The method according to claim 4, characterized in that Based on the absolute positions of the specified number of key frames, a point cloud registration method is used to perform multi-track registration on the two track data, including: Based on the absolute positions of the specified number of key frames, performing multiple multi-track registrations on the two track data by point cloud registration; The accuracy of the registration results of multiple multi-track registrations is evaluated, and the registration results are optimized based on the evaluation results.

6. The method according to claim 5, characterized in that The accuracy is evaluated based on the registration results of multiple multi-track registrations, including: For each registration result of the multi-track registration, a relative position transformation matrix of the two track data is obtained according to the registration result of the multi-track registration; Based on the initial relative posture transformation matrix between the two trajectory data, calculating the error value between the initial relative posture transformation matrix and the relative posture transformation matrix, and obtaining the error value corresponding to the registration result of the multi-trajectory registration; The accuracy of the registration results of the two track data is judged according to the error values ​​corresponding to the registration results of the multiple multi-track registrations.

7. The method according to claim 6, characterized in that Judging the accuracy of the registration results of the two track data according to the error values ​​corresponding to the registration results of the multiple multi-track registrations, including: Acquire a sample set consisting of error values ​​corresponding to the registration results of the multiple multi-track registrations; By counting the covariance states of the six degrees of freedom of translation and rotation corresponding to the sample set, the change of the error value is obtained; The accuracy of the registration result of the two trajectory data is judged according to the change of the error value.

8. The method according to claim 1, characterized in that Aligning the positioning trajectory with the initial map trajectory to construct a map includes: Performing trajectory least squares estimation on the positioning trajectory and the initial map trajectory to obtain an offset value of the positioning trajectory and the initial map trajectory; The positioning track and the initial map track are aligned using the offset value to obtain the final absolute position of the initial map track to construct a map.

9. A map construction device, characterized in that: include: A trajectory generation module, used to process the data collected by multiple vehicles respectively to generate trajectory data corresponding to each vehicle, wherein the data includes point cloud data; A track registration module, used to determine the absolute position of each point cloud point in the point cloud data based on the positioning system, and perform multi-track registration on the track data corresponding to each vehicle based on the absolute position to generate an initial map track; The trajectory alignment module is used to generate a positioning trajectory based on the absolute position of each point cloud point and the signal strength of each point cloud point, and to align the positioning trajectory with the initial map trajectory to construct a map.

10. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 8.

11. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images