Method, device and system for collaboratively building a point cloud map

Abstract

The present disclosure provides a method, device and system for cooperatively constructing a point cloud map, and relates to the field of point cloud map construction. The method comprises: acquiring point cloud data of a corresponding road section collected by a plurality of vehicles when moving on different road sections, different vehicles corresponding to different road sections in a preset route, and any adjacent first road section and second road section being provided with an overlapping area; determining a transformation matrix of point cloud data of the second road section transformed to point cloud data of the first road section by using point cloud data of the overlapping area; transforming the point cloud data of the second road section by using the transformation matrix; and splicing the point cloud data of the first road section and the transformed point cloud data of the second road section to construct a point cloud map of the preset route. Thus, the efficiency of constructing a point cloud map of the entire route is improved, and the method can be applied to a business scenario of online construction of a point cloud map.

Description

Technical Field

[0001] This disclosure relates to the field of point cloud map construction, and in particular to a method, device and system for collaboratively constructing point cloud maps. Background Technology

[0002] The point cloud maps used in the field of autonomous driving are currently constructed by specialized map data collection vehicles equipped with expensive, high-precision map data collection equipment, such as high-precision combined inertial navigation equipment and high-precision multi-line LiDAR equipment, to collect data on a preset route. The backend equipment then processes the collected data offline to build the point cloud map. Summary of the Invention

[0003] This disclosure proposes a scheme for multiple vehicles to collaboratively construct a point cloud map. Different vehicles collect point cloud data from different road segments along a preset route. Using the point cloud data from overlapping areas of adjacent road segments, a transformation matrix is ​​determined for the point cloud data of those two segments. Based on the transformed point cloud data, the point cloud data from different road segments are stitched together to construct a point cloud map of the entire route. This improves the efficiency of constructing a point cloud map of the entire route and is applicable to business scenarios involving online point cloud map construction. Furthermore, since each vehicle collects point cloud data for one road segment, it is not necessary to collect point cloud data for the entire route, thus reducing the accuracy requirements of the map data acquisition equipment carried by each vehicle.

[0004] Some embodiments of this disclosure propose a method for collaboratively constructing point cloud maps, including:

[0005] The point cloud data of the corresponding road segments are collected when multiple vehicles move on different road segments. Different vehicles correspond to different road segments in the preset route. Any adjacent first road segment and second road segment have overlapping areas.

[0006] Using the point cloud data of the overlapping region, determine the transformation matrix from the point cloud data of the second road segment to the point cloud data of the first road segment;

[0007] The point cloud data of the second road segment is transformed using the transformation matrix;

[0008] The point cloud data of the first road segment and the transformed point cloud data of the second road segment are stitched together to construct the point cloud map of the preset route.

[0009] In some embodiments, the point cloud data of each road segment includes: each single-frame point cloud of the road segment and corresponding label data for each single-frame point cloud; wherein, determining the transformation matrix of the point cloud data of the second road segment to the point cloud data of the first road segment includes: determining the transformation matrix of the point cloud data of the second road segment to the point cloud data of the first road segment by registering the single-frame point clouds of the overlapping area in the first road segment and the single-frame point clouds of the overlapping area in the second road segment with the same label data.

[0010] In some embodiments, determining the transformation matrix from the point cloud data of the second road segment to the point cloud data of the first road segment includes:

[0011] The single-frame point cloud of the overlapping area in the first road segment and the single-frame point cloud of the overlapping area in the second road segment with the same labeled data are used as the target point cloud and the source point cloud, respectively, to obtain the matching point pair between the source point cloud and the target point cloud;

[0012] Construct rotation and translation matrices based on matching point pairs;

[0013] The source point cloud is transformed using rotation and translation matrices;

[0014] Calculate the error between the transformed source point cloud and the target point cloud;

[0015] If the error is greater than the preset error, continue to iteratively execute the above steps. If the error is not greater than the preset error, stop iterating and determine the rotation and translation matrix corresponding to the error as the transformation matrix for transforming the point cloud data of the second segment to the point cloud data of the first segment.

[0016] In some embodiments, the labeling data corresponding to each single-frame point cloud of each road segment includes at least one of the following: the current global position information and heading information of the vehicle in the road segment when collecting the single-frame point cloud of the road segment; and the environmental image captured by the vehicle in the road segment when collecting the single-frame point cloud of the road segment.

[0017] In some embodiments, when the similarity of the feature vectors of the environmental images corresponding to different single-frame point clouds is greater than a preset value, they are determined to be different single-frame point clouds with the same labeled data; when the global position information and heading information of the vehicles corresponding to different single-frame point clouds are the same, they are determined to be different single-frame point clouds with the same labeled data.

[0018] In some embodiments, the point cloud data of the second road segment includes: point cloud data of each single frame point cloud of the second road segment and local pose information corresponding to each single frame point cloud. The point cloud data of each single frame point cloud of the second road segment includes: position information of each point constituting the single frame point cloud in a local coordinate system with the vehicle of the second road segment as the origin. The local pose information corresponding to the single frame point cloud of the second road segment is: the relative pose information of the vehicle of the second road segment when acquiring the single frame point cloud of the second road segment relative to the initial pose when acquiring the initial single frame point cloud of the second road segment. Transforming the point cloud data of the second road segment using the transformation matrix includes: multiplying the local pose information corresponding to each single frame point cloud of the second road segment by the transformation matrix, and then multiplying it by the point cloud data of the single frame point cloud.

[0019] In some embodiments, the method further includes: clustering each point in each road segment based on the point cloud data of each road segment; and removing the clustered point sets that match the shape of a preset moving object from the point cloud data of each road segment.

[0020] In some embodiments, the point cloud data for each road segment is point cloud data after removing a set of points of a preset moving object. The removal operation includes the vehicles of each road segment clustering each point of the road segment based on the point cloud data of the road segment, and removing the set of points obtained from the clusters that match the shape of the preset moving object from the point cloud data of the road segment.

[0021] In some embodiments, constructing the point cloud map of the preset route includes: constructing the point cloud map of the preset route using LeGO-LOAM based on the point cloud data of the first road segment and the transformed point cloud data of the second road segment.

[0022] In some embodiments, it also includes:

[0023] Obtain point cloud data of the third road segment that has changed in the preset route collected by any vehicle. The third road segment has an overlapping area with the fourth road segment that has not changed in the preset route.

[0024] Using the point cloud data of the overlapping region, determine the transformation matrix from the point cloud data of the third segment to the point cloud data of the fourth segment;

[0025] The point cloud data of the third road segment is transformed using the transformation matrix;

[0026] The point cloud data of the fourth route segment and the transformed point cloud data of the third route segment are stitched together to construct the point cloud map of the preset route.

[0027] Some embodiments of this disclosure propose a system for collaboratively constructing point cloud maps, including:

[0028] Multiple vehicles are configured to collect point cloud data of a corresponding road segment as they move on that segment, and send the collected point cloud data to a device that collaboratively builds a point cloud map. Different vehicles correspond to different road segments in a preset route.

[0029] The device for collaboratively constructing point cloud maps is configured to perform the method for collaboratively constructing point cloud maps as described in any of the embodiments.

[0030] In some embodiments, the device that collaboratively constructs the point cloud map is one of a plurality of vehicles.

[0031] Some embodiments of this disclosure propose an apparatus for collaboratively constructing point cloud maps, including:

[0032] The communication device is configured to acquire point cloud data of the corresponding road segments collected by multiple vehicles moving on different road segments.

[0033] Memory; and

[0034] A processor coupled to the memory is configured to execute the method for collaboratively constructing a point cloud map as described in any embodiment, based on instructions stored in the memory.

[0035] In some embodiments, the device for collaboratively constructing a point cloud map is one of a plurality of vehicles; the device for collaboratively constructing a point cloud map further includes:

[0036] The point cloud data acquisition device is configured to acquire individual frame point clouds of the corresponding road segment as it moves along the corresponding road segment.

[0037] The pose detection device is configured to detect the relative pose information of the current pose of each single frame point cloud relative to the initial pose of the initial single frame point cloud.

[0038] The tag data acquisition device is configured to acquire the corresponding tag data of each single frame point cloud while each single frame point cloud is being acquired.

[0039] In some embodiments, the tag data acquisition device includes: a global positioning and navigation device; or, it includes: a camera device.

[0040] Some embodiments of this disclosure provide a non-transitory computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method for collaboratively constructing a point cloud map as described in any embodiment. Attached Figure Description

[0041] The accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. This disclosure can be more clearly understood from the following detailed description with reference to the accompanying drawings.

[0042] Obviously, the accompanying drawings described below are merely some embodiments of this disclosure. Those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0043] Figure 1 A flowchart illustrating a method for collaboratively constructing point cloud maps according to some embodiments of this disclosure is shown.

[0044] Figure 2 This illustration shows an application scenario of collaboratively constructing point cloud maps according to some embodiments of the present disclosure.

[0045] Figure 3A flowchart illustrating a method for collaboratively constructing point cloud maps according to other embodiments of this disclosure is shown.

[0046] Figure 4 A schematic diagram of a system for collaboratively constructing point cloud maps according to some embodiments of this disclosure is shown.

[0047] Figure 5 A schematic diagram of an apparatus for collaboratively constructing point cloud maps according to some embodiments of the present disclosure is shown. Detailed Implementation

[0048] The technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0049] Unless otherwise specified, the terms "first," "second," etc., in this disclosure are used to refer to different objects and are not used to indicate size or sequence.

[0050] Figure 1 A flowchart illustrating a method for collaboratively constructing point cloud maps according to some embodiments of this disclosure is shown.

[0051] like Figure 1 As shown, the method of this embodiment includes steps 110-140.

[0052] In step 110, the device that collaboratively constructs the point cloud map acquires point cloud data of the corresponding road segments collected by multiple vehicles moving on different road segments. Different vehicles correspond to different road segments in the preset route, and any adjacent first road segment and second road segment have overlapping areas.

[0053] The device that collaboratively constructs a point cloud map can be one of multiple vehicles or other devices. The vehicle collecting the point cloud data could be, for example, an autonomous vehicle. Autonomous vehicles can collect point cloud data using their onboard sensors, eliminating the need for additional sensors.

[0054] The point cloud data collected for each road segment includes: each single-frame point cloud of that road segment and the corresponding label data for each single-frame point cloud. The single-frame point cloud of that road segment includes: the point cloud data of the single-frame point cloud and the corresponding local pose information of the single-frame point cloud.

[0055] The point cloud data (or first point cloud data) of a single frame for each road segment includes: the position information of each point constituting the single frame point cloud in a local coordinate system with the vehicle of that road segment as the origin. The first point cloud data can be acquired, for example, by a point cloud data acquisition device, such as a lidar device. The corresponding local pose information of the single frame point cloud for each road segment is: the relative pose information of the vehicle in that road segment when acquiring each single frame point cloud of that road segment, relative to its initial pose when acquiring the initial single frame point cloud of that road segment. The corresponding local pose information of the single frame point cloud can be acquired, for example, by a pose detection device, such as an inertial navigation device.

[0056] The second point cloud data of a single-frame point cloud can be determined based on the first point cloud data and the corresponding local pose information of the single-frame point cloud. The second point cloud data for each road segment's single-frame point cloud includes the position information of each point constituting the single-frame point cloud in a local coordinate system with the initial pose of the vehicle in that road segment as the origin. For example, the second point cloud data is obtained by multiplying the corresponding local pose matrix of the single-frame point cloud by the first point cloud data. The formula is as follows:

[0057]

[0058] in, Let R represent the local pose matrix. 3×3 P represents a 3x3 attitude matrix. 3×1 This represents a 3x1 matrix of positions, where (x, y, z) represents the position information of a point in the first point cloud data, and (x′, y′, z′) represents the position information of a point in the second point cloud data.

[0059] Based on the first point cloud data and the corresponding local pose information of each single frame point cloud of a road segment, or based on the second point cloud data of each single frame point cloud of a road segment, a local point cloud map of the road segment can be constructed.

[0060] The labeling data for each road segment's single-frame point cloud includes, for example, at least one of the following: the vehicle's current global position and heading information when acquiring the single-frame point cloud for that road segment, or an environmental image captured by the vehicle while acquiring the single-frame point cloud for that road segment. The labeling data is acquired by a labeling data acquisition device. For example, the global position and heading information can be acquired by a GPS device mounted on the vehicle, and the environmental images can be acquired by a camera device.

[0061] In some embodiments, after acquiring point cloud data for each road segment, the device collaboratively constructing the point cloud map clusters the points within each road segment based on the point cloud data for each road segment. For example, it uses the second point cloud data of each single-frame point cloud of each road segment (i.e., the position information of each point constituting the single-frame point cloud in a local coordinate system with the initial pose of the vehicle in that road segment as the origin) to cluster the points within each road segment. Then, it removes the clustered point sets from the point cloud data of each road segment that match the shape of a preset moving object. Thus, noise generated by moving objects such as pedestrians and vehicles is removed from the point cloud data.

[0062] The point cloud data sent by each vehicle to the device collaboratively building the point cloud map for each road segment can be point cloud data after removing a preset set of points representing moving objects. This removal operation involves each vehicle clustering the points of that road segment based on the point cloud data. For example, using the second point cloud data of each single-frame point cloud of that road segment (i.e., the position information of each point constituting that single-frame point cloud in a local coordinate system with the initial pose of the vehicle in that road segment as the origin), the points of that road segment are clustered. Then, the point sets obtained from the clustering that match the shape of the preset moving objects are removed from the point cloud data of that road segment. Thus, noise generated by moving objects such as pedestrians and vehicles is removed from the point cloud data, and the amount of information transmitted between the vehicles and the device collaboratively building the point cloud map can be reduced.

[0063] In step 120, the device that collaboratively constructs the point cloud map uses the point cloud data of the overlapping area to determine the transformation matrix from the point cloud data of the second road segment to the point cloud data of the first road segment.

[0064] By registering single-frame point clouds of overlapping regions in the first segment and single-frame point clouds of overlapping regions in the second segment with the same labeled data, the transformation matrix for transforming the point cloud data of the second segment to the point cloud data of the first segment is determined.

[0065] For example, the Iterative Closest Point (ICP) algorithm can be used for registration between different point clouds. Specifically, single-frame point clouds of overlapping regions in the first segment and the second segment with the same labeled data are used as the target point cloud and the source point cloud, respectively, to obtain matching point pairs between the source and target point clouds; a rotation and translation matrix is ​​constructed based on the matching point pairs; the source point cloud is transformed using the rotation and translation matrix; the error between the transformed source and target point clouds is calculated; if the error is greater than a preset error, the above steps are iteratively executed; if the error is not greater than the preset error, the iteration stops, and the rotation and translation matrix corresponding to the error is determined as the transformation matrix for transforming the point cloud data of the second segment to the point cloud data of the first segment.

[0066] Different single-frame point clouds are classified as having the same labeled data if the global position and heading information of the vehicles corresponding to them are identical. Alternatively, they are classified as having the same labeled data if the similarity of the feature vectors of the environmental images corresponding to the different single-frame point clouds is greater than a preset value. Or, they are classified as having the same labeled data if the global position and heading information of the vehicles corresponding to the different single-frame point clouds are identical, and the similarity of the feature vectors of the environmental images corresponding to the different single-frame point clouds is greater than a preset value.

[0067] In step 130, the device that collaboratively constructs the point cloud map uses the transformation matrix to transform the point cloud data of the second road segment.

[0068] The transformation method includes, for example, multiplying the local pose information of each single-frame point cloud of the second road segment by the transformation matrix, and then multiplying it by the point cloud data of that single-frame point cloud (first point cloud data). This transforms the point cloud data of the second road segment into the coordinate system of the first road segment.

[0069]

[0070] in, The expression represents the local pose information of a single frame point cloud. (x,y,z) represents the position information of a point in the first point cloud data of a single frame point cloud. (x′′,y′′,z′′) represents the position information of a point in the transformed point cloud data of a single frame point cloud. N is an intermediate product of the operation and has no practical meaning. This represents the transformation matrix.

[0071] In step 140, the device that collaboratively constructs the point cloud map stitches together the point cloud data of the first road segment and the transformed point cloud data of the second road segment to construct the point cloud map of the preset route.

[0072] As mentioned above, the point cloud data of the second road segment has been transformed into the coordinate system of the first road segment. Therefore, by stitching together the second point cloud data of each single frame point cloud of the first road segment and the transformed point cloud data of each single frame point cloud of the second road segment, the point cloud maps of the first and second road segments can be constructed.

[0073] If the preset route is divided into more than two segments, the method for constructing the point cloud map of other adjacent segments is the same as the method for constructing the point cloud map of the first and second segments, and will not be repeated here.

[0074] For any first and second segment of a preset route, using the point cloud data of the first segment and the transformed point cloud data of the second segment, the LeGO-LOAM tool can be used to construct a point cloud map of the preset route. In other words, by inputting the point cloud data of each segment into the LeGO-LOAM tool, a point cloud map of the entire route can be constructed.

[0075] The aforementioned scheme for collaborative point cloud map construction by multiple vehicles involves different vehicles collecting point cloud data from different segments of a preset route. By utilizing the point cloud data from overlapping areas of adjacent segments, a transformation matrix is ​​determined for the point cloud data of those adjacent segments. Based on the transformed point cloud data, the point cloud data from different segments are stitched together to construct a point cloud map of the entire route. This improves the efficiency of constructing a point cloud map for the entire route and is applicable to business scenarios involving online point cloud map construction. Furthermore, since each vehicle collects point cloud data for one segment, eliminating the need to collect point cloud data for the entire route, the accuracy requirements for the map data acquisition equipment on each vehicle are reduced.

[0076] Figure 2 This illustration shows an application scenario of collaboratively constructing point cloud maps according to some embodiments of the present disclosure.

[0077] Four vehicles working together Figure 2 The diagram shows a point cloud map of a rectangular route. Each vehicle is responsible for collecting point cloud data for one segment of the route. Adjacent segments have overlapping areas. One of the four vehicles acts as the collaborative point cloud map construction device, referred to as the master vehicle. The other vehicles are designated as vehicle 1, vehicle 2, and vehicle 3. The four vehicles move in the same direction along their respective assigned segments, simultaneously collecting point cloud data for their respective segments. Vehicles 1, 2, and 3 transmit their collected point cloud data to the master vehicle. The master vehicle then uses the point cloud data collected by vehicles 1, 2, and 3, along with its own collected data, to construct the point cloud map. Figure 1 The method shown in the embodiment can construct a point cloud map of the entire rectangular route.

[0078] For example, if the road segment of the main vehicle is adjacent to the road segment of vehicle 1, the point cloud data of vehicle 1's road segment is transformed to the coordinates of the main vehicle's road segment. By stitching the point cloud data of the main vehicle's road segment with the transformed point cloud data of vehicle 1's road segment, a point cloud map of both the main vehicle's road segment and vehicle 1's road segment can be constructed. Similarly, if the road segment of the main vehicle is adjacent to the road segment of vehicle 3, the point cloud data of vehicle 3's road segment is transformed to the coordinates of the main vehicle's road segment. By stitching the point cloud data of the main vehicle's road segment with the transformed point cloud data of vehicle 3's road segment, a point cloud map of both the main vehicle's road segment and vehicle 3's road segment can be constructed. Since the point cloud data of the main vehicle's route segment and the route segment of vehicle 1 have been unified to the same coordinate system, the main vehicle's route segment and the route segment of vehicle 1 are treated as a merged route segment. This merged route segment is adjacent to the route segment of vehicle 2. The point cloud data of vehicle 2's route segment is transformed to the coordinate system of the merged route segment (i.e., the main vehicle's route segment). By stitching the point cloud data of the merged route segment and the transformed point cloud data of vehicle 2's route segment, a point cloud map of the merged route segment and vehicle 2's route segment can be constructed. Following a similar process, the point cloud map of the entire rectangular route is finally constructed.

[0079] Figure 3 A flowchart illustrating a method for collaboratively constructing point cloud maps according to other embodiments of this disclosure is shown.

[0080] like Figure 3 As shown, the method of this embodiment is in Figure 1 Based on the illustrated embodiment, the method further includes steps 350-380.

[0081] In step 350, assuming a certain segment (third segment) of the preset route changes, any vehicle collects point cloud data of the changed third segment and sends it to the device that collaboratively constructs the point cloud map. The device that collaboratively constructs the point cloud map obtains the point cloud data of the changed third segment of the preset route collected by any vehicle. The third segment overlaps with a segment of the preset route that has not changed (fourth segment).

[0082] In step 360, the device that collaboratively constructs the point cloud map uses the point cloud data of the overlapping area to determine the transformation matrix from the point cloud data of the third segment to the point cloud data of the fourth segment.

[0083] The method for determining the transformation matrix from the point cloud data of the third segment to the point cloud data of the fourth segment is the same as the method for determining the transformation matrix from the point cloud data of the second segment to the point cloud data of the first segment. For details, please refer to the above.

[0084] In step 370, the device that collaboratively constructs the point cloud map uses the transformation matrix to transform the point cloud data of the third road segment.

[0085] The method for transforming the point cloud data of the third road segment using a transformation matrix is ​​the same as the method for transforming the point cloud data of the second road segment using a transformation matrix. For details, please refer to the above.

[0086] In step 380, the device collaboratively constructing the point cloud map stitches together the point cloud data of the fourth road segment and the transformed point cloud data of the third road segment to construct the point cloud map of the preset route. The stitching method is detailed above.

[0087] Therefore, when a local point cloud map changes, only the point cloud map of that local road segment needs to be reconstructed, improving the update efficiency of the point cloud map for the entire route.

[0088] Figure 4 A schematic diagram of a system for collaboratively constructing point cloud maps according to some embodiments of this disclosure is shown.

[0089] like Figure 4 As shown, the system in this embodiment includes: multiple vehicles 410 and a device 420 for collaboratively constructing a point cloud map. The device for collaboratively constructing the point cloud map can be, for example, one of the multiple vehicles, referred to as the master vehicle, or other devices.

[0090] Multiple vehicles 410 are configured to collect point cloud data of the corresponding road segment while moving on the corresponding road segment, and send the collected point cloud data to the device that collaboratively builds the point cloud map. Different vehicles correspond to different road segments in the preset route.

[0091] The device 420 for collaboratively constructing point cloud maps is configured to perform the method for collaboratively constructing point cloud maps according to any embodiment.

[0092] For example, a device that collaboratively constructs a point cloud map acquires point cloud data of corresponding road segments collected by multiple vehicles moving on different road segments. Different vehicles correspond to different road segments in a preset route, and any adjacent first and second road segments have overlapping areas. Using the point cloud data of the overlapping areas, a transformation matrix is ​​determined to transform the point cloud data of the second road segment to the point cloud data of the first road segment. The transformation matrix is ​​used to transform the point cloud data of the second road segment. The point cloud data of the first road segment and the transformed point cloud data of the second road segment are then stitched together to construct a point cloud map of the preset route.

[0093] For example, a device that collaboratively constructs a point cloud map acquires point cloud data of a third road segment that has changed in a preset route collected by any vehicle. The third road segment overlaps with a fourth road segment that has not changed in the preset route. Using the point cloud data of the overlapping area, a transformation matrix is ​​determined to transform the point cloud data of the third road segment into the point cloud data of the fourth road segment. The point cloud data of the third road segment is transformed using the transformation matrix. The point cloud data of the fourth road segment and the transformed point cloud data of the third road segment are then stitched together to construct a point cloud map of the preset route.

[0094] Figure 5 A schematic diagram of an apparatus for collaboratively constructing point cloud maps according to some embodiments of the present disclosure is shown.

[0095] like Figure 5 As shown, the device 420 for collaboratively constructing point cloud maps in this embodiment includes:

[0096] The communication device 421 is configured to acquire point cloud data of the corresponding road segments collected by multiple vehicles moving on different road segments.

[0097] Memory 422; and

[0098] A processor 423 coupled to the memory is configured to execute a method for collaboratively constructing a point cloud map according to any embodiment, based on instructions stored in the memory, as detailed above and will not be repeated here.

[0099] The communication device 421 may be, for example, a wireless communication device, such as a wireless local area network communication device, or a mobile network communication device, such as a fifth-generation mobile network communication device, a fourth-generation mobile network communication device, etc.

[0100] The memory 422 may include, for example, system memory, fixed non-volatile storage media, etc. The system memory stores, for example, the operating system, application programs, boot loader, and other programs.

[0101] The device 420 for collaboratively building a point cloud map can be, for example, one of multiple vehicles. In this case, the device 420 for collaboratively building a point cloud map also includes:

[0102] The point cloud data acquisition device 424 is configured to acquire each single frame of point cloud of the corresponding road segment when moving on the corresponding road segment, such as a lidar device. The point cloud data acquired by the lidar device includes not only the position information of the point, but also information such as the intensity of the reflected signal.

[0103] Pose detection device 425, configured to detect relative pose information of the current pose of each single-frame point cloud relative to the initial pose of the initial single-frame point cloud, such as an inertial navigation device; and

[0104] The tag data acquisition device 426 is configured to acquire the corresponding tag data of each single frame point cloud while each single frame point cloud is acquired, such as a global positioning and navigation device, a camera device, etc.

[0105] The aforementioned devices 421-426 can be connected via bus 427, for example.

[0106] This disclosure also proposes a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for collaboratively constructing a point cloud map according to any embodiment.

[0107] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more non-transitory computer-readable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer program code.

[0108] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0110] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0111] The above description is only a preferred embodiment of this disclosure and is not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the protection scope of this disclosure.

Claims

1. A method for collaboratively constructing point cloud maps, characterized in that, include: The system acquires point cloud data of various road segments collected by multiple vehicles moving on different road segments. Different vehicles correspond to different road segments in a preset route. Any adjacent first and second road segments have overlapping areas. The point cloud data of each road segment includes: each single frame point cloud of the road segment and corresponding label data for each single frame point cloud. The corresponding label data for each single frame point cloud of each road segment includes: environmental images captured by the vehicles on the road segment when collecting the single frame point cloud of the road segment. Point sets obtained by clustering that match the shape of a preset moving object in the point cloud data of each road segment are removed. The moving objects include pedestrians and moving vehicles. Using point cloud data from overlapping regions, a transformation matrix is ​​determined to transform the point cloud data of the second road segment to the point cloud data of the first road segment. This includes: when the similarity of the feature vectors of the environmental images corresponding to different single-frame point clouds is greater than a preset value, they are determined to be different single-frame point clouds with the same labeled data; or, when the global position information and heading information of the vehicles corresponding to different single-frame point clouds are the same, and the similarity of the feature vectors of the environmental images corresponding to different single-frame point clouds is greater than a preset value, they are determined to be different single-frame point clouds with the same labeled data. By registering the single-frame point clouds of overlapping regions in the first road segment and the single-frame point clouds of overlapping regions in the second road segment with the same labeled data, the transformation matrix is ​​determined to transform the point cloud data of the second road segment to the point cloud data of the first road segment. The point cloud data of the second road segment is transformed using the transformation matrix; The point cloud data of the first road segment and the transformed point cloud data of the second road segment are stitched together to construct the point cloud map of the preset route.

2. The method according to claim 1, characterized in that, Determine the transformation matrix from the point cloud data of the second segment to the point cloud data of the first segment, including: The single-frame point cloud of the overlapping area in the first road segment and the single-frame point cloud of the overlapping area in the second road segment with the same labeled data are used as the target point cloud and the source point cloud, respectively, to obtain the matching point pair between the source point cloud and the target point cloud; Construct rotation and translation matrices based on matching point pairs; The source point cloud is transformed using rotation and translation matrices; Calculate the error between the transformed source point cloud and the target point cloud; If the error is greater than the preset error, continue to iteratively execute the above steps. If the error is not greater than the preset error, stop iterating and determine the rotation and translation matrix corresponding to the error as the transformation matrix for transforming the point cloud data of the second segment to the point cloud data of the first segment.

3. The method according to claim 1, characterized in that, The corresponding labeled data for each single-frame point cloud of each road segment also includes: the current global position information and heading information of the vehicle in the road segment when the single-frame point cloud of the road segment is collected.

4. The method according to claim 1, characterized in that, The point cloud data of the second road segment includes: point cloud data of each single frame point cloud of the second road segment and local pose information corresponding to each single frame point cloud. The point cloud data of the single frame point cloud of the second road segment includes: position information of each point constituting the single frame point cloud in a local coordinate system with the vehicle of the second road segment as the origin. The local pose information of the single frame point cloud of the second road segment is: the relative pose information of the vehicle of the second road segment when collecting the single frame point cloud of the second road segment relative to the initial pose when collecting the initial single frame point cloud of the second road segment. Transforming the point cloud data of the second road segment using the transformation matrix includes: multiplying the local pose information of each single frame point cloud of the second road segment by the transformation matrix, and then multiplying it by the point cloud data of the single frame point cloud.

5. The method according to claim 1, characterized in that, The removal includes: Based on the point cloud data of each road segment, cluster the points of each road segment; From the point cloud data of each road segment, remove the clusters that match the shape of the preset moving object to obtain the point set.

6. The method according to claim 1, characterized in that, The removal process involves clustering the points of each road segment based on the point cloud data of the road segment, and removing the clustered point sets that match the shape of a preset moving object from the point cloud data of the road segment.

7. The method according to claim 1, characterized in that, Constructing the point cloud map of the preset route includes: Based on the point cloud data of the first route segment and the transformed point cloud data of the second route segment, a point cloud map of the preset route is constructed using LeGO-LOAM.

8. The method according to claim 1, characterized in that, Also includes: Obtain point cloud data of the third road segment that has changed in the preset route collected by any vehicle. The third road segment has an overlapping area with the fourth road segment that has not changed in the preset route. Using the point cloud data of the overlapping region, determine the transformation matrix from the point cloud data of the third segment to the point cloud data of the fourth segment; The point cloud data of the third road segment is transformed using the transformation matrix; The point cloud data of the fourth route segment and the transformed point cloud data of the third route segment are stitched together to construct the point cloud map of the preset route.

9. A system for collaboratively constructing point cloud maps, characterized in that, include: Multiple vehicles are configured to collect point cloud data of a corresponding road segment as they move on that segment, and send the collected point cloud data to a device that collaboratively builds a point cloud map. Different vehicles correspond to different road segments in a preset route. A device for collaboratively constructing point cloud maps is configured to perform the method for collaboratively constructing point cloud maps as described in any one of claims 1-8.

10. The system according to claim 9, characterized in that, The device that collaboratively constructs the point cloud map is one of multiple vehicles.

11. A device for collaboratively constructing point cloud maps, characterized in that, include: The communication device is configured to acquire point cloud data of the corresponding road segments collected by multiple vehicles moving on different road segments. Memory; as well as A processor coupled to the memory, the processor being configured to execute the method for collaboratively constructing a point cloud map as described in any one of claims 1-8, based on instructions stored in the memory.

12. The device according to claim 11, characterized in that, The device that collaboratively constructs the point cloud map is one of multiple vehicles; The devices for collaboratively constructing point cloud maps also include: The point cloud data acquisition device is configured to acquire individual frame point clouds of the corresponding road segment as it moves along the corresponding road segment. The pose detection device is configured to detect the relative pose information of the current pose of each single frame point cloud relative to the initial pose of the initial single frame point cloud. The tag data acquisition device is configured to acquire the corresponding tag data of each single frame point cloud while each single frame point cloud is being acquired.

13. The device according to claim 12, characterized in that, The tag data acquisition device includes: a global positioning and navigation device; or, a camera device.

14. A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for collaboratively constructing a point cloud map as described in any one of claims 1-8.

Images