A method, system, device, and storage medium for constructing a global point cloud map.

Abstract

This invention discloses a method, system, device, and storage medium for constructing a global point cloud map. The method includes: acquiring several differential GPS data points of a vehicle at different times and parsing several global pose information points at different times from them; acquiring several frames of point cloud data around the vehicle at different times and parsing several laser odometry information points at different times from them; fusing the several global pose information points and several laser odometry information points based on a factor graph optimization algorithm to obtain multiple global odometry information points of the vehicle at different times; and constructing a global point cloud map based on the multiple global odometry information points and their associated multi-frame point cloud data points at different times. This invention utilizes a factor graph optimization algorithm to fuse point cloud data and differential GPS data, resulting in a more accurate and reliable global point cloud map.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method, system, device, and storage medium for constructing a global point cloud map. Background Technology

[0002] With the continuous development and advancement of science and technology, computer technology, modern sensing technology, and artificial intelligence technology have been gradually applied to the automotive field. By implementing perception, planning, and control functions on vehicles, autonomous driving can be achieved. Among these, generating a global point cloud map of the vehicle's driving environment is an important aspect of autonomous driving technology.

[0003] The current proposed method for constructing global point cloud maps uses GPS sensors to assign pose information to synchronously acquired point cloud data, and uses the location information contained therein to stitch the point cloud together to form the corresponding point cloud map. However, the problem with this construction method is that it cannot take into account the transformation between continuous point cloud data. When there is an error in the pose information provided by the GPS sensor, it may cause problems such as overlap and ghosting in the stitched point cloud map. Summary of the Invention

[0004] This invention provides a global point cloud map construction method, system, device, and storage medium to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0005] Firstly, a method for constructing a global point cloud map is provided, the method comprising:

[0006] Step S110: Obtain several differential GPS data of the vehicle at different times, and parse out several global pose information at different times from them;

[0007] Step S120: Obtain several frames of point cloud data around the vehicle at different times, and parse out several laser odometry information at different times.

[0008] Step S130: Based on the factor graph optimization algorithm, fuse the several global pose information and the several laser odometry information to obtain multiple global odometry information of the vehicle at different times;

[0009] Step S140: Construct a global point cloud map based on the multiple global odometer information and the associated multi-frame point cloud data at different times.

[0010] Furthermore, each differential GPS data includes position data, attitude data, and GPS week time. The position data includes longitude, latitude, and altitude values, and the attitude data includes roll angle, pitch angle, and heading angle.

[0011] Further, in step S110, the corresponding global pose information is parsed from each differential GPS data, including:

[0012] The GPS week time in the differential GPS data is converted to Greenwich Mean Time;

[0013] The longitude and latitude values ​​in the differential GPS data are converted to the UTM coordinate system for representation, and the corresponding global pose information is generated by combining the altitude and attitude data in the differential GPS data.

[0014] Furthermore, in step S120, each laser odometry information is obtained by parsing two adjacent frames of point cloud data. The specific implementation process includes:

[0015] Obtain the first frame of point cloud data at time T and the second frame of point cloud data at time T+1, where T is a positive integer;

[0016] Based on the LeGO-LOAM algorithm, feature matching is performed on the first frame point cloud data and the second frame point cloud data to obtain pose transformation information;

[0017] The laser odometry information at time T is added to the pose transformation information to obtain the laser odometry information at time T+1.

[0018] Wherein, when T=1, the laser odometry information at time T is the pose data represented by the point cloud data of the first frame.

[0019] Furthermore, the implementation process of step S130 includes:

[0020] The global pose information and the laser odometry information are time-aligned, and the first global pose information with the alignment relationship is used as the prior factor.

[0021] From all the remaining global pose information that have alignment relationships, select multiple global pose information that meet the predetermined selection rules, and then construct the corresponding multiple GPS factors.

[0022] Using all laser odometry information with alignment relationships, construct all corresponding odometry factors and all their associated pose node factor pairs;

[0023] By connecting the prior factors, the multiple GPS factors, all odometry factors, and all pose node factors in a time synchronization manner, a factor graph is obtained.

[0024] The factor graph is nonlinearly optimized to obtain the global odometry information associated with each pose node factor.

[0025] Furthermore, the predetermined selection rule includes at least the following: the time interval between the two global pose information corresponding to two adjacent GPS factors falls within a preset time range.

[0026] Furthermore, the implementation process of step S140 includes:

[0027] A single frame of point cloud data for a vehicle at any given time contains multiple point cloud data points. Each point cloud data point is projected onto a global coordinate system using global odometer information at the same time point. Each frame of point cloud data is then processed in the manner described above to construct a global point cloud map.

[0028] Secondly, a global point cloud map construction system is provided, the system comprising:

[0029] The first parsing module is used to acquire several differential GPS data of the vehicle at different times and parse several global pose information at different times from them.

[0030] The second parsing module is used to acquire several frames of point cloud data around the vehicle at different times, and to parse out several laser odometry information at different times.

[0031] The fusion module is used to fuse the several global pose information and the several laser odometry information based on the factor graph optimization algorithm to obtain multiple global odometry information of the vehicle at different times.

[0032] The construction module is used to construct a global point cloud map based on the multiple global odometer information and the associated multi-frame point cloud data at different times.

[0033] Thirdly, a computer device is provided, comprising:

[0034] At least one memory;

[0035] At least one processor;

[0036] The memory stores a computer program, and the processor executes the computer program as described in the first aspect of the global point cloud map construction method.

[0037] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the global point cloud map construction method as described in the first aspect.

[0038] The present invention has at least the following beneficial effects: the pose transformation between consecutive point clouds can be effectively determined by using several frames of point cloud data, the global position information can be effectively determined by using several differential GPS data, and the fusion of point cloud data and differential GPS data can be achieved by further using the factor graph optimization algorithm, which can make the final global point cloud map more accurate and reliable. Attached Figure Description

[0039] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0040] Figure 1 This is a flowchart illustrating a global point cloud map construction method according to an embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram of the composition of a global point cloud map construction system according to an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the hardware structure of the computer device in an embodiment of this disclosure. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0045] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a global point cloud map construction method provided by an embodiment of the present invention. The method includes the following steps:

[0046] Step S110: Obtain several differential GPS data of the vehicle at different times, and parse out several global pose information at different times from them;

[0047] Step S120: Obtain several frames of point cloud data around the vehicle at different times, and parse out several laser odometry information at different times.

[0048] Step S130: Based on the factor graph optimization algorithm, fuse the several global pose information and the several laser odometry information to obtain multiple global odometry information of the vehicle at different times;

[0049] Step S140: Construct a global point cloud map based on the multiple global odometer information and the associated multi-frame point cloud data at different times.

[0050] In this embodiment of the invention, the differential GPS data of the vehicle at any given time includes GPS cycle time, attitude data, and position data, wherein the attitude data includes yaw, pitch, and roll, and the position data includes altitude, latitude, and longitude.

[0051] In this embodiment of the invention, step S110 actually describes the parsing of the global pose information covered by the differential GPS data of the vehicle at any given time. The specific implementation process includes the following:

[0052] Step S111: Obtain differential GPS data collected by the GPS inertial navigation device installed on the vehicle at any time, and convert the GPS week time contained therein into Greenwich Mean Time.

[0053] Step S112: Convert the latitude and longitude values ​​contained in the differential GPS data to the UTM (Universal Transverse Mercator Grid System) coordinate system for representation, and then combine the altitude and attitude data contained in the differential GPS data to generate the corresponding global pose information (X,Y,Z,Roll,Pitc,Yaw), where Z is the altitude value.

[0054] It should be noted that the timestamp corresponding to each frame of point cloud data proposed in step S120 above is assigned by the lidar, which uses Greenwich Mean Time (GMT) and sets the zero point to 00:00 on January 1, 1970; while the timestamp corresponding to each differential GPS data proposed in step S110 above is assigned by the GPS inertial navigation system, which uses GPS week time and sets the zero point to 00:00 on January 6, 1980. By implementing step S111 above, this embodiment of the invention can meet the time alignment processing of these two types of data in subsequent operations.

[0055] It should be noted that step S112 above uses an existing latitude and longitude to UTM conversion algorithm, and the corresponding calculation process is as follows:

[0056]

[0057] In the formula, v is the deformation of the UTM projection length, and e is the eccentricity. Here, A represents the latitude value contained in the differential GPS data, λ represents the longitude interval, λ0 represents the longitude value of the central meridian, and λ / 6 must be a rounded-down value, and s represents the latitude value. The associated meridian radius, E is the eastward distance, which is the X value in the global pose information mentioned above, E0 is the offset, k0 is the central meridian scale factor, α is the equatorial radius, and T is a user-defined parameter, derived from the formula. In this representation, C is a user-defined parameter, derived from the formula. The representation is as follows: N is the northward distance, which is the Y value in the global pose information mentioned above. When the projection point is located in the Northern Hemisphere, N0 is 0 km, and when the projection point is located in the Southern Hemisphere, N0 is 10000 km.

[0058] In this embodiment of the invention, step S120 actually describes the parsing of two frames of point cloud data around the vehicle at adjacent times to extract the laser odometry information it covers. The specific implementation process includes the following:

[0059] Step S121: Obtain the first frame of point cloud data and the second frame of point cloud data. The first frame of point cloud data is a frame of point cloud data around the vehicle obtained at time T, and the second frame of point cloud data is a frame of point cloud data around the vehicle obtained at time T+1, where T is a positive integer.

[0060] Step S122: Perform feature matching on the first frame point cloud data and the second frame point cloud data based on the LeGO-LOAM (Lightweight and Ground-Optimized LidarOdometry and Mapping on Variable Terrain) algorithm to obtain pose transformation information;

[0061] Step S123: The pose transformation information is accumulated with the laser odometry information at time T obtained earlier to obtain the laser odometry information at time T+1.

[0062] In step S121 above, the external parameters of the lidar and GPS inertial navigation equipment installed on the vehicle are first calibrated to determine the position transformation relationship between the lidar and the GPS inertial navigation equipment. Then, based on the position transformation relationship, a frame of point cloud data acquired by the lidar at time T is subjected to overall translation and rotation processing to obtain the first frame of point cloud data. And based on the position transformation relationship, a frame of point cloud data acquired by the lidar at time T+1 is subjected to overall translation and rotation processing to obtain the second frame of point cloud data. By performing the above point cloud transformation processing, this embodiment of the invention can ensure that the origin position of each frame of point cloud data acquired by the lidar is the same as the origin position of each differential GPS data acquired by the GPS inertial navigation equipment.

[0063] In step S122 above, the implementation process of using the LeGO-LOAM algorithm includes the following steps:

[0064] (1) The first frame of point cloud data is segmented using the laser radar beam scanning towards the ground to obtain the first ground point cloud. The remaining point cloud is then clustered to obtain several first point cloud clusters. After filtering out the smaller number of first point cloud clusters, all the remaining first point cloud clusters and the first ground point cloud are labeled and bound.

[0065] (2) For the first distance image formed by the first ground point cloud, the first distance image is horizontally divided into multiple equal first sub-images; for each first sub-image, the curvature value corresponding to each row of point cloud is calculated; when the curvature value does not exceed the predetermined threshold, the row of point cloud is classified as surface point feature and attached with the label bound to the first ground point cloud; or when the curvature value exceeds the predetermined threshold, the row of point cloud is classified as edge point feature and attached with the label bound to the first ground point cloud.

[0066] Similarly, for the first distance image formed by each first point cloud cluster, surface point features and edge point features are extracted according to the image segmentation and curvature comparison methods described above;

[0067] (3) The second frame point cloud data is segmented using the laser radar beam scanning towards the ground to obtain the second ground point cloud. The remaining point cloud is then clustered to obtain several second point cloud clusters. After filtering out the smaller number of second point cloud clusters, all the remaining second point cloud clusters and the second ground point cloud are labeled and bound.

[0068] (4) For the second distance image formed by the second ground point cloud, the second distance image is horizontally divided into multiple equal second sub-images; for each second sub-image, the curvature value corresponding to each row of point cloud is calculated. When the curvature value does not exceed the predetermined threshold, the row of point cloud is classified as a surface point feature and the label bound to the second ground point cloud is attached. Or when the curvature value exceeds the predetermined threshold, the row of point cloud is classified as an edge point feature and the label bound to the second ground point cloud is attached.

[0069] Similarly, for the second distance image formed by each second point cloud cluster, surface point features and edge point features are extracted according to the image segmentation and curvature comparison methods described above;

[0070] (5) Using the LM nonlinear optimization method (i.e., the Levenberg-Marquardt method), the two types of features extracted from the first frame point cloud data in steps (1) to (2) above and the two types of features extracted from the second frame point cloud data in steps (3) to (4) above are matched according to the same type of label. All face point features with matching relationship are subjected to pose transformation to obtain the height change value, roll angle change value and pitch angle change value. All edge point features with matching relationship are subjected to pose transformation to obtain the lateral horizontal displacement change value, longitudinal horizontal displacement change value and heading angle change value. Then, the final pose transformation information can be obtained based on the six change values ​​obtained above.

[0071] It should be noted that in step S123 above, when T=1, the laser odometry information at time T is actually the pose data covered by the first frame point cloud data, wherein the pose data includes attitude data and position data.

[0072] In this embodiment of the invention, the specific implementation process of step S130 includes the following:

[0073] Step S131: Time-align the several global pose information obtained through step S110 and the several laser odometry information obtained through step S120, and obtain the first global pose information that has an alignment relationship and is the earliest in time, and then input it into the existing GTSAM (Georgia Tech Smoothing and Mapping library) library to generate prior factor output.

[0074] Step S132: Using pre-defined selection rules, filter all remaining global pose information that has an alignment relationship (i.e., not including the global pose information obtained through step S131 above) to obtain multiple global pose information, and then input them into the GTSAM library to generate multiple corresponding GPS factor outputs.

[0075] Step S133: Obtain all laser odometry information with alignment relationships, and then input it into the GTSAM library to generate all corresponding odometry factors; since each laser odometry information is calculated by combining two frames of point cloud data at adjacent times, input the pose data covered by each frame of point cloud data into the GTSAM library to generate pose node factor output, and then obtain a pose node factor pair connected to each odometry factor;

[0076] Step S134: Connect the prior factors obtained in step S131, the multiple GPS factors obtained in step S132, and all odometry factors and pose node factors obtained in step S133 in a time synchronization manner to obtain a factor graph.

[0077] Step S135: Directly call the iSAM2 (Incremental Smoothing and Mapping) algorithm from the GTSAM library to perform nonlinear optimization processing on the factor map, thereby obtaining the global odometry information associated with each pose node factor.

[0078] In step S131 above, the purpose of determining the prior factor is to provide a fixed initial global pose for the entire global odometry system, thereby ensuring that each global odometry information obtained through the subsequent step S135 is a unique solution.

[0079] In step S132 above, the pre-defined selection rules include: first, the time interval between the global pose information currently defined as a GPS factor and the previous global pose information defined as a GPS factor should be within a preset time range, the preset time range being greater than or equal to 5 seconds, and no other GPS factor associated with a time value falling within the time interval associated with the above two GPS factors; second, the geometric distance error between the global pose information currently defined as a GPS factor and the global odometry pose data obtained at the same moment after performing nonlinear optimization should be within a preset distance range, the preset distance range being greater than or equal to 3 meters.

[0080] In step S133 above, it is necessary to directly merge two pose node factors that appear at the same time into one pose node factor for representation. For example, the odometry factor at time T+1 is connected to the pose node factor at time T and the pose node factor at time T+1. The odometry factor at time T+2 is connected to the pose node factor at time T+1 and the pose node factor at time T+2. At this time, it is necessary to directly merge the two pose node factors that appear at time T+1 so that only one pose node factor at time T+1 appears in the final factor graph.

[0081] In step S134 above, the prior factors obtained in step S131 need to be connected with the pose node factors at the earliest time (obtained from all pose node factor pairs obtained in step S133 above). Then, each GPS factor obtained in step S132 is connected with a pose node factor at the same time (also obtained from all pose node factor pairs obtained in step S133 above). All of the above connections are constrained and created by the GTSAM library.

[0082] In this embodiment of the invention, the specific implementation process of step S140 includes the following:

[0083] Step S141: Using the multiple timestamp information corresponding to the multiple global odometer information, obtain multiple frames of point cloud data that are the same as the multiple timestamp information from the multiple frames of point cloud data obtained through step S120 above;

[0084] Step S142: Since each frame of point cloud data actually contains multiple point cloud data, firstly obtain a global odometer information that is the same as the timestamp information of the point cloud data of that frame, and then multiply each point cloud data with the global odometer information respectively, so that the point cloud data of that frame can be represented in the global coordinate system through projection transformation.

[0085] Step S143: Process the multi-frame point cloud data according to the implementation method of step S142 above, and then construct a global point cloud map.

[0086] In this embodiment of the invention, the pose transformation between consecutive point clouds can be effectively determined using several frames of point cloud data, and the global position information can be effectively determined using several differential GPS data. Furthermore, the fusion of point cloud data and differential GPS data is achieved by using a factor graph optimization algorithm, which makes the final global point cloud map more accurate and reliable.

[0087] Please refer to Figure 2 , Figure 2This is a schematic diagram illustrating the composition of a global point cloud map construction system provided in an embodiment of the present invention. The system includes:

[0088] The first parsing module 210 is used to acquire several differential GPS data of the vehicle at different times and parse several global pose information at different times from them.

[0089] The second parsing module 220 is used to acquire several frames of point cloud data around the vehicle at different times, and to parse several laser odometry information at different times. Each laser odometry information needs to be calculated by combining two frames of point cloud data around the vehicle at adjacent times.

[0090] The fusion module 230 is used to fuse the plurality of laser odometry information and the plurality of global pose information using a factor graph optimization algorithm, thereby obtaining multiple global odometry information of the vehicle at different times.

[0091] The construction module 240 is used to construct a global point cloud map by combining the multiple global odometer information and the associated multi-frame point cloud data at different times.

[0092] The content of the above method embodiments is applicable to this system embodiment. The functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are the same as those in the above method embodiments. Therefore, they will not be repeated here.

[0093] Furthermore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the global point cloud map construction method described in the above embodiments. The computer-readable storage medium includes, but is not limited to, any type of disk (including floppy disks, hard disks, optical disks, CD-ROMs, and magneto-optical disks), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic cards, or optical cards. In other words, the storage device includes any medium on which a device (e.g., a computer, mobile phone, etc.) stores or transmits information in a readable form, and can be a read-only memory, a disk, or an optical disk, etc.

[0094] also, Figure 3This is a schematic diagram of the hardware structure of a computer device provided in an embodiment of the present invention. The computer device includes components such as a processor 320, a memory 330, an input unit 340, and a display unit 350. Those skilled in the art will understand that... Figure 3 The illustrated device structure is not intended to limit all devices and may include more or fewer components than shown, or combine certain components. The memory 330 can be used to store the computer program 310 and various functional modules. The processor 320 runs the computer program 310 stored in the memory 330, thereby performing various functional applications and data processing of the device. The memory can be internal memory or external memory, or include both internal and external memory. Internal memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, or random access memory. External memory may include hard disks, floppy disks, ZIP disks, USB flash drives, magnetic tapes, etc. The memory 330 disclosed in the embodiments of this invention includes, but is not limited to, these types of memory. The memory 330 disclosed in the embodiments of this invention is only an example and not a limitation.

[0095] Input unit 340 is used to receive signal input and user-input keywords. Input unit 340 may include a touch panel and other input devices. The touch panel can collect user touch operations on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel) and drive the corresponding connection device according to a pre-set program; other input devices may include, but are not limited to, one or more of physical keyboards, function keys (such as play control buttons, power buttons, etc.), trackballs, mice, joysticks, etc. Display unit 350 can be used to display user-input information or information provided to the user, as well as various menus of the terminal device. Display unit 350 may be in the form of a liquid crystal display, organic light-emitting diode, etc. Processor 320 is the control center of the terminal device, connecting various parts of the entire device through various interfaces and lines, performing various functions and processing data by running or executing software programs and / or modules stored in memory 330, and calling data stored in memory.

[0096] As one embodiment, the computer device includes at least one processor 320, at least one memory 330, and at least one computer program 310, wherein the at least one computer program 310 is stored in the at least one memory 330 and configured to be executed by the at least one processor 320, and the at least one computer program 310 is configured to perform the global point cloud map construction method in the above embodiment.

[0097] Although the description of this application has been quite detailed and particularly focused on several of the described embodiments, it is not intended to limit itself to any of these details or embodiments or any particular embodiment. Rather, it should be considered as effectively covering the intended scope of this application by referring to the appended claims and taking into account the prior art, which provides for a broad possible interpretation of these claims. Furthermore, the foregoing description of this application with respect to embodiments foreseeable by the inventors is intended to provide a useful description, and non-substantial modifications to this application that have not yet been foreseen may still represent equivalent modifications.

Claims

1. A method for constructing a global point cloud map, characterized in that, The method includes: Step S110: Obtain several differential GPS data of the vehicle at different times, and parse out several global pose information at different times from them; Step S120: Obtain several frames of point cloud data around the vehicle at different times, and parse out several laser odometry information at different times. Step S130: Based on the factor graph optimization algorithm, fuse the several global pose information and the several laser odometry information to obtain multiple global odometry information of the vehicle at different times; Step S140: Construct a global point cloud map based on the multiple global odometer information and the associated multi-frame point cloud data at different times; In step S120, each laser odometry information is obtained by parsing two adjacent frames of point cloud data. The specific implementation process includes: Obtain the first frame of point cloud data at time T and the second frame of point cloud data at time T+1, where T is a positive integer; Based on the LeGO-LOAM algorithm, feature matching is performed on the first frame point cloud data and the second frame point cloud data to obtain pose transformation information; The laser odometry information at time T is added to the pose transformation information to obtain the laser odometry information at time T+1. Wherein, when T=1, the laser odometry information at time T is the pose data represented by the point cloud data of the first frame; The implementation process of step S130 includes: The global pose information and the laser odometry information are time-aligned, and the first global pose information with the alignment relationship is used as the prior factor. From all the remaining global pose information that have an alignment relationship, select multiple global pose information that meet the predetermined selection rules, and then construct multiple corresponding GPS factors; wherein, the predetermined selection rules include at least the time interval between the two global pose information corresponding to two adjacent GPS factors falling within a preset time range. Using all laser odometry information with alignment relationships, construct all corresponding odometry factors and all their associated pose node factor pairs; By connecting the prior factors, the multiple GPS factors, all odometry factors, and all pose node factors in a time synchronization manner, a factor graph is obtained. The factor graph is nonlinearly optimized to obtain the global odometry information associated with each pose node factor.

2. The global point cloud map construction method according to claim 1, characterized in that, Each differential GPS data set includes position data, attitude data, and GPS week time. The position data includes longitude, latitude, and altitude values, and the attitude data includes roll, pitch, and heading angles.

3. The global point cloud map construction method according to claim 2, characterized in that, In step S110, the corresponding global pose information is parsed from each differential GPS data, including: The GPS week time in the differential GPS data is converted to Greenwich Mean Time; The longitude and latitude values ​​in the differential GPS data are converted to the UTM coordinate system for representation, and the corresponding global pose information is generated by combining the altitude and attitude data in the differential GPS data.

4. The global point cloud map construction method according to claim 1, characterized in that, The implementation process of step S140 includes: A single frame of point cloud data for a vehicle at any given time contains multiple point cloud data points. Each point cloud data point is projected onto a global coordinate system using global odometer information at the same time point. Each frame of point cloud data is then processed in the manner described above to construct a global point cloud map.

5. A global point cloud map construction system, characterized in that, The system includes: The first parsing module is used to acquire several differential GPS data of the vehicle at different times and parse several global pose information at different times from them. The second parsing module is used to acquire several frames of point cloud data around the vehicle at different times, and to parse out several laser odometry information at different times. The fusion module is used to fuse the several global pose information and the several laser odometry information based on the factor graph optimization algorithm to obtain multiple global odometry information of the vehicle at different times. The construction module is used to construct a global point cloud map based on the multiple global odometer information and the associated multi-frame point cloud data at different times. In the second parsing module, each laser odometry information is obtained by parsing two adjacent frames of point cloud data. The specific implementation process includes: Obtain the first frame of point cloud data at time T and the second frame of point cloud data at time T+1, where T is a positive integer; Based on the LeGO-LOAM algorithm, feature matching is performed on the first frame point cloud data and the second frame point cloud data to obtain pose transformation information; The laser odometry information at time T is added to the pose transformation information to obtain the laser odometry information at time T+1. Wherein, when T=1, the laser odometry information at time T is the pose data represented by the point cloud data of the first frame; The implementation process of the fusion module includes: The global pose information and the laser odometry information are time-aligned, and the first global pose information with the alignment relationship is used as the prior factor. From all the remaining global pose information that have an alignment relationship, select multiple global pose information that meet the predetermined selection rules, and then construct multiple corresponding GPS factors; wherein, the predetermined selection rules include at least the time interval between the two global pose information corresponding to two adjacent GPS factors falling within a preset time range. Using all laser odometry information with alignment relationships, construct all corresponding odometry factors and all their associated pose node factor pairs; By connecting the prior factors, the multiple GPS factors, all odometry factors, and all pose node factors in a time synchronization manner, a factor graph is obtained. The factor graph is nonlinearly optimized to obtain the global odometry information associated with each pose node factor.

6. A computer device, characterized in that, include: At least one memory; At least one processor; The memory stores a computer program, and the processor executes the computer program to implement the global point cloud map construction method as described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the global point cloud map construction method as described in any one of claims 1 to 4.

Images