Point cloud labeling method and device

Abstract

The invention provides a point cloud labeling method and device, and relates to the field of data processing. According to the invention, a plurality of nodes are created in the point cloud to mark the lane lines. When the nodes are marked, the plane positions of the nodes are preliminarily determined through the first marking operation of marking the visual angle, then the coordinates of the nodes in the height direction are determined through the second marking operation of adjusting the visual angle, and therefore all the nodes capable of reflecting the lane line are determined in the three-dimensional space. Therefore, the marking of the lane lines can be realized through additionally created nodes, and the marking difficulty is reduced. In addition, the nodes are presented through multiple visual angles, the distribution conditions of the point clouds in different coordinate systems can be presented more visually, and therefore the node labeling precision is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and specifically to a point cloud annotation method and device. Background Art

[0002] Point cloud data (hereinafter referred to as point cloud) can generally refer to a data set of points in a specific coordinate system. For example, a point cloud can be represented as the three-dimensional position information of a point in three-dimensional space and its related information (such as the color information of the point). Point cloud is widely used in various fields.

[0003] In the field of autonomous driving and related fields, point clouds can be collected by a mobile measurement system (MMS) or other similar devices and used to characterize the scanning results of the road surface. Therefore, data processing such as lane line detection and obstacle recognition can be performed based on the point cloud, so as to select and execute appropriate driving operations to achieve autonomous driving. The above process generally requires the lane lines in the point cloud to be marked in advance, so as to achieve lane line detection and / or lane division. Therefore, how to accurately mark lane lines in point clouds is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the invention

[0004] In view of this, an embodiment of the present application provides a point cloud annotation method and device, which preliminarily determines the plane position of the node through a first annotation operation of the annotation perspective, and then determines the coordinates of the node in the height direction through a second annotation operation of adjusting the perspective, thereby achieving accurate annotation of the lane line to solve the above-mentioned technical problems.

[0005] According to one aspect of the present disclosure, the present disclosure provides a point cloud annotation method, the method comprising: displaying a point cloud to be annotated through an annotation perspective along a reference direction, wherein the point cloud to be annotated includes a point cloud representing a lane line; determining a target node and its first position information based on a first annotation operation on the point cloud to be annotated, wherein the lane line is annotated by a plurality of nodes including the target node; displaying the target node through an adjustment perspective perpendicular to the reference direction; determining the second position information of the target node under the adjustment perspective based on a second annotation operation on the target node and the first position information, wherein the second position information at least includes coordinate information of the target node in the reference direction.

[0006] According to one aspect of the present disclosure, the present disclosure also provides a point cloud annotation device, which includes: one or more processors, and a memory for storing programs, wherein the program includes instructions, and when the instructions are executed by the processor, the processor executes the aforementioned method.

[0007] A point cloud annotation method and device are provided based on the embodiments of the present application. The present application annotates lane lines by creating multiple nodes in a point cloud. When annotating nodes, the plane position of the node is preliminarily determined by a first annotation operation of the annotation perspective, and then the coordinates of the node in the height direction are determined by a second annotation operation of adjusting the perspective, thereby determining each node that can reflect the lane line in three-dimensional space. Therefore, the annotation of lane lines in the present application can be achieved through additionally created nodes, reducing the difficulty of annotation. In addition, the present application presents nodes through multiple perspectives, which can more intuitively present the distribution of point clouds in different coordinate systems, thereby improving the annotation accuracy of nodes. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 This is an application scenario diagram of a point cloud annotation scenario provided in some embodiments of the present application.

[0010] Figure 2 This is an exemplary flowchart of the point cloud annotation method provided in some embodiments of the present application.

[0011] Figure 3 This is a schematic diagram of the point cloud annotation interface provided by some embodiments of the present application.

[0012] Figure 4 It is an exemplary flowchart of a second location information determination method provided in some embodiments of the present application.

[0013] Figure 5 It is an exemplary flow chart of a method for determining initial coordinate information provided in some embodiments of the present application.

[0014] Figure 6 This is an exemplary device module diagram of a point cloud annotation device provided in some embodiments of the present application.

[0015] Figure 7 This is an example diagram of a point cloud annotation interface provided in some embodiments of the present application. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0017] Application Overview

[0018] In autonomous driving and related technologies, lane marking is generally achieved by selecting point clouds corresponding to lane lines from point clouds to be marked by a marking staff. The marking staff can select point clouds corresponding to lane lines through a point cloud marking interface to mark lane lines.

[0019] Furthermore, considering that each point in the point cloud contains three-dimensional coordinate information (such as three-dimensional coordinate information in the world coordinate system), it is generally necessary to project the point cloud at a certain perspective during annotation. The point cloud is first converted from the point cloud coordinate system to the device coordinate system, and then converted to the screen coordinate system, and then displayed to the annotator for viewing. Then, based on the annotation operation of the annotator at this perspective, the point cloud corresponding to the lane line is determined. At the same time, since the annotation operation is performed on the two-dimensional image obtained by the point cloud projection, after the annotator inputs the annotation operation through the display device (such as creating a node), the coordinates corresponding to the annotation operation (such as the node) also need to be converted from the screen coordinate system to the device coordinate system, and then converted to the point cloud coordinate system to determine the target point cloud.

[0020] Specifically, the screen coordinate system refers to the coordinates based on which the user clicks or touches on the screen. The screen coordinate system is a two-dimensional coordinate system. For example, when the screen is displayed in full screen, the upper left corner of the screen is used as the origin, and the horizontal X-axis coordinate and the vertical Y-axis coordinate represent the position of the point on the screen; or when the screen is displayed in a window, the upper left corner of the window is used as the origin. The device coordinate system is a left-handed three-dimensional coordinate system with the origin at the center of the screen. Compared with the screen coordinate system, it has one more dimension (such as the Z axis, which usually points to the inside of the screen). When the screen coordinate system is converted to this coordinate system, the X and Y axis coordinates are usually converted according to certain rules, and then the Z axis coordinate takes a preset value. For example, in full-screen display, assuming that the canvas (i.e., screen) of the screen coordinate system is h high and w wide, and there is a point (x, y) in the screen coordinate system, and the coordinates of this point in the device coordinate system are (x1, y1, z1), then x1 = (x / w)*2-1, y1 = -(y / h)*2+1, where the value ranges of x1, y1, and z1 are all [-1, 1]. The point cloud coordinate system is also a three-dimensional coordinate system, which can be understood as the coordinate system used when the point cloud is collected and stored, usually the world coordinate system. When the device coordinate system needs to be converted to this coordinate system, it can be converted through inverse projection.

[0021] In the aforementioned technology, the point cloud to be annotated is generally projected along the Z axis / height direction of the world coordinate system for the annotators to view and select. In this process, it is easy for the annotators to make mistakes when selecting the point cloud. For example, when obstacles (such as roadblocks, roadside green belts, etc.) block the lane line, because the height information is eliminated by the projection, the point cloud selected by the annotator may be the point cloud of the obstacle above the lane line, thus affecting the subsequent recognition of the lane line and the division of the lane.

[0022] In response to the above problems, an embodiment of the present application provides a point cloud annotation method and device. The present application annotates lane lines by creating multiple nodes in a point cloud. When annotating nodes, the plane position of the node is preliminarily determined by a first annotation operation of annotating the perspective, and then the coordinates of the node in the height direction are determined by a second annotation operation of adjusting the perspective, thereby determining each node that can reflect the lane line in three-dimensional space. Therefore, the annotation of lane lines in the present application can be achieved through additionally created nodes, reducing the difficulty of annotation. In addition, the present application presents nodes through multiple perspectives, which can more intuitively present the distribution of point clouds in different coordinate systems, thereby improving the annotation accuracy of nodes. Various non-limiting embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0023] Exemplary application scenarios

[0024] Figure 1 This is an application scenario diagram of a point cloud annotation scenario provided in some embodiments of the present application.

[0025] like Figure 1 As shown, the point cloud annotation scene 100 may include a user terminal 110 , an annotator 120 , a point cloud annotation interface 130 , a processor 140 , and a memory 150 .

[0026] In the above-mentioned point cloud annotation scenario 100, the processor 140 and the memory 150 can be associated with the aforementioned user terminal 110, and the user terminal 110 is used to present the point cloud annotation interface 130, so that the annotator 120 can perform lane line annotation in the point cloud annotation interface 130 through interactive operations on the user terminal 110.

[0027] During the lane line marking process, the processor 140 can process the point cloud data so that the point cloud marking interface 130 can display the point cloud to be marked through the marking perspective along the reference direction. The processor 140 can respond to the first marking operation of the marking personnel and determine the target node among the multiple nodes for marking the lane line and the first position information of the target node based on the first marking operation on the point cloud to be marked. The processor 140 can process the point cloud to be marked so that the point cloud marking interface 130 displays the target node through an adjusted perspective perpendicular to the reference direction. The processor 140 can respond to the second marking operation of the marking personnel and determine the second position information of the target node based on the second marking operation on the target node and the first position information under the adjusted perspective. For more information about the above marking process, please refer to Figure 2 and its related description.

[0028] The user terminal 110 may refer to one or more terminal devices or software used by the annotator 120 for point cloud annotation. The user terminal 110 may include a display part and an input part. The display part may be used to display the point cloud annotation interface 130, and the input part may be operated by the annotator 120 to input the annotation operation to annotate the lane line. In some embodiments, the user terminal 110 may be one of the mobile devices 110-1, tablet computers 110-2, laptop computers 110-3, desktop computers 110-4, or any combination thereof, which have input and / or output functions.

[0029] The annotator 120 may perform an annotation operation based on the point cloud annotation interface 130 , so that the processor 140 executes a corresponding instruction in response to the annotation operation.

[0030] It should be noted that the limitation of the annotation operation (such as the subsequent first annotation operation and the second annotation operation) in this application does not mean to limit the actions of the annotation personnel 120, but to provide interactive controls / interaction logic corresponding to the annotation operation to carry the corresponding interactive operations. For example, the aforementioned first annotation operation can be a node creation interactive operation, that is, the processor 140 can create a node at a corresponding position based on a specific interactive action (such as a mouse click), thereby inputting the first annotation operation. For another example, the aforementioned second annotation operation can be a node position adjustment operation, that is, the processor 140 can create a corresponding control based on the node, and the annotation personnel 120 can drag the control to adjust the position of the node, thereby inputting the second annotation operation.

[0031] The point cloud annotation interface 130 may refer to an interactive interface for performing point cloud annotation. The point cloud annotation interface 130 may be presented in the user terminal 110, and carry the annotation operation of the annotator 120 through the control, and then feed back the annotation operation to the processor 140, so that the processor 140 determines the point cloud corresponding to the lane line in the point cloud to be annotated based on the annotation operation of the annotator 120.

[0032] The processor 140 can process data and / or information obtained from other devices or system components. The processor 140 can be integrated into the corresponding device as needed to process the point cloud data so that the user terminal 110 presents the aforementioned point cloud annotation interface 130. For example, when the aforementioned point cloud annotation is a local task, the processor 140 can be a computing device in the user terminal 110. For another example. When the aforementioned point cloud annotation is a cloud task / remote task, the processor 140 can be a computing device in a server that provides related services.

[0033] In some embodiments, the processor 140 may include one or more sub-processing devices (e.g., a single-core processing device or a multi-core multi-core processing device). As an example only, the processor 140 may include a central processing unit (CPU), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), etc., or any combination thereof.

[0034] The memory 150 may be used to store data and / or instructions. The memory 150 may include one or more storage components, each of which may be an independent device or a part of another device. In some embodiments, the memory 150 may include a random access memory (RAM), a read-only memory (ROM), a mass storage device, or the like, or any combination thereof. Exemplarily, the mass storage device may include a magnetic disk, an optical disk, a solid-state disk, or the like.

[0035] In some embodiments, the aforementioned processor 140 and the memory 150 may construct a point cloud annotation device. In the point cloud annotation device, the memory 150 may store the relevant data of the present application (such as point cloud, relevant parameters of the annotation operation, etc.), and the processor 140 may execute preset program instructions based on calling the relevant data from the memory 150, so that the point cloud annotation device forms multiple functional modules, thereby performing one or more functions described in the present application.

[0036] In some embodiments, the aforementioned point cloud annotation device can be integrated into the user terminal 110 and form a point cloud annotation interface 130 for annotation, so as to determine the lane lines in the point cloud based on the interactive operation with the annotation user 120. For more information about the point cloud annotation device, please refer to Figure 6 and its related description.

[0037] In some embodiments, the aforementioned memory 150 may be implemented based on a computer-readable storage medium storing a program, wherein the instructions, when executed by one or more processors of a computing device (such as the processor 140), cause the computing device to perform the point cloud annotation method / function described in this application.

[0038] Exemplary point cloud annotation method

[0039] Figure 2 is an exemplary flow chart of a point cloud annotation method provided in some embodiments of the present application. Figure 2 The process P200 shown can be executed by the point cloud annotation device 600, and each step of P200 can be executed by the internal functional module of the point cloud annotation device 600.

[0040] like Figure 2 As shown, P200 may include the following steps:

[0041] S210 , displaying the point cloud to be annotated through the annotation viewing angle along the reference direction. In some embodiments, S210 may be executed by the display module 610 .

[0042] The annotation perspective can refer to the presentation perspective of the annotated point cloud, and the reference direction can refer to the presentation direction of the annotation perspective. The annotation perspective is generally a bird's-eye view, and the reference direction is generally a bird's-eye view. The point cloud to be annotated is a collection of points that need to be annotated by the annotator. The point cloud to be annotated can be formed based on the point cloud collected from the road surface and include a point cloud representing the lane line, so that the annotator can mark the lane line.

[0043] S220 : Determine a target node and its first position information based on a first labeling operation on the point cloud to be labeled. In some embodiments, S220 may be performed by the node determination module 620 .

[0044] Among them, the aforementioned first annotation operation may be an operation in which the annotator triggers a node creation instruction in the annotation perspective, that is, the annotator may trigger an instruction to cause the node determination module 620 to create a series of nodes to represent the lane line (the lane line may be annotated by multiple nodes). The target node may refer to the node formed by the current first annotation operation, that is, the node currently being operated. The first position information may refer to the initial position information of the target node determined based on the relevant parameters of the first annotation operation (such as the coordinate information of the position pointed to by the cursor, or the input coordinate information).

[0045] The initial position information refers to the position information in the device coordinate system, point cloud coordinate system or other preset coordinate system. Unless otherwise specified, the position information refers to the position information in the preset coordinate system.

[0046] S230 , displaying the target node by adjusting the viewing angle perpendicular to the reference direction. In some embodiments, S230 may be executed by the display module 610 .

[0047] The adjusted perspective may refer to a point cloud presentation perspective that is perpendicular to the reference direction / annotation perspective, and based on the adjusted perspective, the coordinate information of the target node in the reference direction may be presented.

[0048] S240 , under the adjusted viewing angle, based on the second marking operation on the target node and the first position information, determine the second position information of the target node. In some embodiments, S240 may be performed by the node adjustment module 630 .

[0049] The second annotation operation may refer to an adjustment operation on the position information of the target node. The second position information may refer to the position information of the target node adjusted based on the second annotation operation. Because the second annotation operation is performed under the adjustment perspective, the second position information may at least include the coordinate information of the target node in the reference direction.

[0050] In some embodiments, the aforementioned S210 can be executed in response to a point cloud annotation request from an annotator. That is, the processor can call the corresponding point cloud to be annotated from the memory in response to the point cloud annotation request from the annotator, and display the point cloud to be annotated in the point cloud annotation interface based on the annotation perspective. Among them, the annotation perspective and the reference direction are generally preset in the relevant interface, and the annotator can select a suitable interface in the point cloud annotation interface for annotation (generally a bird's-eye view). The point cloud to be annotated can be obtained through MMS, LiDAR or other laser scanning equipment, and this application does not limit this.

[0051] In some embodiments, in order to further improve the display effect of the point cloud to be annotated, the reference plane reflecting the road surface can be determined first, and then the reference direction perpendicular to the reference plane can be determined. Considering that the point cloud to be annotated is generally formed based on laser scanning of the road surface, the road surface point cloud reflecting the road surface can be determined from the point cloud to be annotated, and then the reference plane is fitted based on the road surface point cloud, so as to determine the direction perpendicular to the reference plane as the reference direction. Among them, the road surface point cloud can be determined based on the relevant method of the reference plane of the point cloud data in the relevant technology. For example, the road surface point cloud (or the reference plane of the point cloud) can be determined based on a set of continuous points with the lowest height, and the specific method will not be repeated here.

[0052] As a result, the coordinates of the road surface point cloud in the reference direction are basically constant, making it easier to determine relevant parameters of the road surface (such as the threshold range of the lane line point cloud). At the same time, the point cloud to be annotated will not be misaligned due to the inclination of the reference direction (especially the higher point cloud), thereby improving the display effect. Among them, the misalignment phenomenon can refer to the fact that when the reference direction is not perpendicular to the road surface, the horizontal coordinate information of the point cloud with a certain height is different from the actual coordinate information when it is displayed along the annotation perspective of the reference direction.

[0053] In some embodiments, the aforementioned S220 may be executed in response to a first annotation operation. That is, the processor may execute based on relevant parameters of the interactive operation in response to the interactive operation of the annotator in the point cloud annotation interface. Among them, the first annotation operation can be understood as a target node creation operation within the annotation perspective, and the relevant controls of the point cloud annotation interface can collect the relevant position information of the interactive operation, thereby creating a target node at the corresponding position. Among them, the first position information of the target node can be determined based on the position corresponding to the first annotation operation.

[0054] For example, the first annotation operation can be defined as a trigger operation containing position information (such as clicking the left or right mouse button, or clicking a specific button), and the annotator can create a target node at the corresponding position (such as the cursor position) by executing the first annotation operation. At this time, the processor can map the cursor position in the point cloud annotation interface to the space where the point cloud to be annotated is located to determine the first position information of the target node.

[0055] Considering that the annotation perspective displays the target node to be annotated along the reference direction, the first position information determined under the annotation perspective generally does not have the coordinate information of the reference direction, and only includes the coordinate information on the reference plane (such as the coordinate information in the longitudinal direction of the plane and the coordinate information in the transverse direction of the plane, or the coordinate information of the X-axis and Y-axis in the device coordinate system). In some embodiments, the processor can also estimate the coordinate information of the target node in the reference direction based on the point cloud related to the target node. For related content, please refer to Figure 5 and its related description.

[0056] In some embodiments, the aforementioned S230 can be implemented by adjusting the display viewing angle in the point cloud annotation interface, that is, presenting the point cloud to be annotated and the target node by adjusting the viewing angle.

[0057] To ensure that the coordinate information of the target node in the reference direction can be determined in the subsequent steps, the coordinate information of the target node in the reference direction can at least be adjusted in the annotation operation of adjusting the viewing angle. Thus, in some embodiments, the position information of any object (such as a node, point cloud, etc.) in the annotation viewing angle can include the coordinate information in the longitudinal direction of the plane and the coordinate information in the transverse direction of the plane, and the position information of any object in adjusting the viewing angle includes at least the coordinate information in the reference direction. Among them, the reference plane formed by the longitudinal direction of the plane and the transverse direction of the plane (such as the aforementioned road surface) is perpendicular to the reference direction.

[0058] In some embodiments, the adjustment perspective may include at least a first adjustment perspective and a second adjustment perspective, wherein the position information of the first adjustment perspective includes coordinate information in a reference direction and coordinate information in a horizontal direction of the plane, and the position information of the second adjustment perspective includes coordinate information in a reference direction and coordinate information in a vertical direction of the plane. The reference direction may be recorded as the z-axis, the horizontal direction of the plane may be recorded as the x-axis, and the vertical direction of the plane may be recorded as the y-axis, that is, the aforementioned annotation perspective may resolve the coordinates of any object in the (x, y) direction, the first adjustment perspective may resolve the coordinates of any object in the (x, z) direction, and the second adjustment perspective may resolve the coordinates of any object in the (y, z) direction, thereby realizing multi-angle resolution of the target node.

[0059] In some embodiments, the aforementioned plane direction can be constructed based on the lane line direction to facilitate the analysis of the road. The horizontal plane direction is along the extension direction of the lane line in the reference plane, and the vertical plane direction is perpendicular to the aforementioned horizontal plane direction in the reference plane. As an alternative example, the plane direction can also be a preset value when collecting point cloud data to ensure the uniformity of the coordinates of the annotation information.

[0060] In some embodiments, the specific contents of marking the viewing angle and adjusting the viewing angle can be found in Figure 3 , Figure 6 And the related descriptions are not repeated here.

[0061] In some embodiments, the aforementioned S240 can be executed in response to a second annotation operation. That is, the processor can respond to the interactive operation of the annotator in the point cloud annotation interface and execute based on the relevant parameters of the interactive operation. Among them, the second annotation operation can be understood as the target node adjustment operation within the adjustment viewing angle, and the relevant controls of the point cloud annotation interface can collect the relevant position information in the screen coordinate system corresponding to the interactive operation, so as to further determine the coordinate information of the target node in the reference direction (i.e., the z-axis coordinate in the device coordinate system), thereby combining with the first position information to determine the second position information.

[0062] In some embodiments, the annotator can perform an adjustment operation such as dragging on the target node in the adjustment interface, thereby transmitting the second annotation operation to the processor. The processor can adjust the target node based on the second annotation operation and determine the coordinate information of the target node in the reference direction. The processor can also determine the second position information based on the first position information and the coordinate information of the target node in the reference direction.

[0063] In some embodiments, the target node can be presented in the form of a specific control in the adjustment interface, so as to facilitate the annotator to perform the second annotation operation. For more information about the control, see Figure 4 and its related description.

[0064] It should be noted that the target node in this application is a specific reference to the node corresponding to the current operation. When marking lane lines, multiple nodes are required to be marked, and the relevant steps of the target node (such as S220 to S240) can be repeated and the target node can be rotated, so as to determine multiple nodes to complete the marking of the lane lines.

[0065] Based on the above point cloud annotation method, when annotating nodes, the plane position of the node is preliminarily determined through the first annotation operation of the annotation perspective, and then the coordinates of the node in the height direction are determined through the second annotation operation of adjusting the perspective, so as to determine each node that can reflect the lane line in three-dimensional space. Therefore, the annotation of lane lines in this application can be achieved through additionally created nodes, without the need to select the corresponding point cloud, which reduces the difficulty of annotation. In addition, this application presents nodes through multiple perspectives, which can more intuitively present the distribution of point clouds in different coordinate systems, thereby improving the annotation accuracy of nodes.

[0066] In some embodiments, based on the above steps, multiple nodes of the lane line can be determined, and annotation information reflecting the attributes of the lane line can be generated to further describe the information of the lane line. Figure 2 As shown, P200 may also include the following steps:

[0067] S250: Determine lane line labeling information based on multiple nodes. The labeling information may refer to information reflecting lane line related attributes. In some embodiments, S250 may be performed by the lane line labeling module 640.

[0068] In some embodiments, to ensure that multiple nodes can truly reflect the lane lines, the processor in the aforementioned S250 can map multiple nodes to the image data based on the mapping relationship with the image data to determine whether the annotation is accurate. At this time, the processor can first determine the image data corresponding to the point cloud to be annotated. Then, the image data is mapped based on multiple nodes to determine a mapping image containing multiple nodes. Finally, based on the mapping image and the third annotation operation, the annotation information of the lane line is determined. Among them, the third annotation operation is at least used to adjust multiple nodes to determine that multiple nodes can reflect the lane line.

[0069] For example, if there is no problem with the lane marking, the third marking operation may include a determination operation (such as a touch operation to determine the space) to determine that the lane marking is completed. If there is a problem with the lane marking, the third marking operation may include a node selection operation to determine the node that needs to be adjusted as the target node, and return to the aforementioned steps (such as S230, S240) to readjust until the lane marking is completed.

[0070] It should be noted that when the point cloud to be annotated is collected, the relevant collection settings (such as the aforementioned MMS) can synchronously collect image data, so that the point cloud to be annotated corresponds to the image data, thereby constructing a mapping relationship between the point cloud to be annotated (and its related spatial coordinate system) and the image data. The aforementioned mapping processing can be performed based on this mapping relationship. For example, the aforementioned mapping relationship can be characterized as a mapping relationship between each point in the point cloud to be annotated and each pixel in the image data, thereby, the node can be mapped to the image data based on the point cloud related to the node. For another example, the aforementioned mapping relationship can be characterized as a conversion relationship from the coordinate system of the point cloud to be annotated to the pixel of the image data, and the node can be directly mapped to the image data based on the conversion relationship.

[0071] In some embodiments, in order to facilitate the labeling personnel to determine the labeling of the lane line based on the mapped nodes, the labeling line of the lane line can be determined based on the nodes and mapped. That is, the processor can determine the labeling line of the lane line based on each target node of the lane line. Then, based on the position information in the multiple nodes, the multiple nodes and the labeling lines are mapped in the image data to determine a mapping image containing multiple nodes. Among them, the labeling line includes multiple nodes of the lane line, which can refer to a vector line segment formed by connecting multiple nodes in sequence. The mapping method of the labeling line is similar to that of the node, which will not be repeated here.

[0072] In some embodiments, the lane line annotation information may also include other relevant information of the lane line, such as the road direction of the lane line, whether the lane line can be changed, whether the lane line is a tidal lane, etc. Thus, the processor may also determine the attribute information of the lane line based on the fourth annotation operation of the lane line in response to the fourth annotation operation of the annotator, wherein the fourth annotation operation is used to input the attribute information.

[0073] Based on the lane line marking information determined by the above method, in actual application, the lane line marking information can be used to divide lanes in the navigation map to determine the lane information of the lane, so as to determine the driving instructions to be executed in combination with the driving destination, thereby realizing automatic driving. For example, the city roads can be scanned in advance, and the lane line marking information can be determined based on the above scheme. When the vehicle is driving, the collected information can be compared with the pre-marked information to determine the marking information and lane information of the current driving lane.

[0074] Example point cloud annotation interface

[0075] To further illustrate the display and interaction in the point cloud annotation interface, the present application may also provide an exemplary point cloud annotation interface. Figure 3 This is a schematic diagram of the point cloud annotation interface provided by some embodiments of the present application.

[0076] like Figure 3 As shown, the point cloud annotation interface 300 may include an annotation perspective 300A, a partial magnified view 300B of the annotation perspective, a first adjustment perspective 300C, and a second adjustment perspective 300D. That is, the interface where the annotation perspective is located also includes a partial magnified view of the target node and an adjustment perspective, so as to avoid changing the interface when the perspective is switched, thereby facilitating the operation of the annotator.

[0077] like Figure 3 As shown, the reference coordinate axes of the annotation perspective 300A may be the x-axis and the y-axis (i.e. the aforementioned plane horizontal and plane vertical directions), and the annotation perspective 300A includes a plurality of point clouds 310 to be annotated. The point clouds 310 to be annotated may present different display styles according to the actual content of the point clouds.

[0078] In the annotation perspective 300A, the point cloud 310 to be annotated can be of any shape, such as a circular point cloud, a pentagonal point cloud, and a star point cloud. The circular point cloud can reflect the point cloud of the road surface, the pentagonal point cloud can reflect the point cloud of the lane line, and the star point cloud can reflect the point cloud of roadside obstacles (such as green belts).

[0079] It should be noted that the above display styles are generally determined based on relevant parameters of the point cloud (such as actual color, height information, etc.), and are only used to guide the annotator to make reasonable annotations. In practice, different types of point clouds are generally presented in different colors. For details, please refer to Figure 6 In addition, different display styles of point clouds are generally only used to guide labelers in labeling. In actual labeling, there may be situations where the point clouds corresponding to lane lines do not need to be labeled (such as the point clouds corresponding to turn signs).

[0080] In some embodiments, in order to determine the display style of the point cloud to be annotated, the processor may first determine the point cloud to be annotated and its location information when loading the point cloud to be annotated into the view (such as the annotation view 300A). Then, the display style of the point cloud to be annotated is determined based on the location information of the point cloud to be annotated. Finally, the point cloud to be annotated is displayed through the annotation perspective based on the location information and display style of the point cloud to be annotated. Among them, the display style can be adjusted according to the actual situation, and generally includes the aforementioned three categories: road surface, lane line and obstacle.

[0081] As an exemplary illustration only, the point cloud may be classified according to its location information and color information to determine the display style. Among them, the point cloud slightly higher than the road surface and in yellow / white may be the lane line point cloud.

[0082] As shown in the annotation perspective 300A, the processor can create multiple nodes for each lane line in the annotation perspective 300A in response to the first annotation operation and related annotation operations of the annotator, thereby implementing the annotation of the lane line. For example, for the lane line 320 (represented by a rectangle filled with a grid) in the annotation perspective 300A, the processor can create multiple nodes 330 (represented by a square shape) in response to a series of annotation operations of the annotator based on the lane line 320, and reflect the lane line 320 by connecting each node 330.

[0083] As shown in the annotation perspective 300A, the processor may record the node created by the annotator's current first annotation operation as a target node 340 (indicated by a square filled with a dot pattern) and display it in the annotation perspective 300A. It should be noted that the special display of the fill pattern of the target node 340 in the annotation perspective 300A is only an example, and in practice, how to display the target node 340 may also be determined as needed. Figure 6 Another display style is also provided, and please refer to the relevant drawings and text descriptions.

[0084] Based on the aforementioned target node 340, the target node 340 and its surrounding point clouds can be displayed in other perspectives. For example, the target node 340 and its surrounding point clouds can be displayed in the local magnified image 300B, the first adjusted perspective 300C, and the second adjusted perspective 300D. Among them, the surrounding point cloud of the target node 340 may include points within a preset area with the target node 340 as the center.

[0085] In some embodiments, in order to facilitate the annotation personnel to observe the coordinates of the target node 340, other perspectives can respectively reflect different coordinate systems. Among them, the local magnified image 300B can reflect the coordinates of the target node 340 under the x-axis and y-axis, the first adjustment perspective 300C can reflect the coordinates of the target node 340 under the x-axis and z-axis, and the second adjustment perspective 300D can reflect the coordinates of the target node 340 under the y-axis and z-axis.

[0086] In some embodiments, the target node in other perspectives can be presented through the node control 350. Thus, the annotator can adjust the position of the node control in the perspective to determine the appropriate position of the target node, thereby determining the second position information. For more information about determining the second position information based on the node control, see Figure 4 and its related description.

[0087] In some embodiments, when the target node / node control is presented in a perspective involving a reference direction, it can be displayed in a preset position. For example, the target node can be fixedly displayed at the center of the perspective. In some embodiments, when the target node is displayed in an adjusted perspective involving a reference direction, the initial position of the target node can also be generated based on the point cloud related to the target node, so as to be displayed. For more information about the initial position (initial coordinate information), please refer to Figure 5 and its related description.

[0088] Exemplary adjustment method based on node control

[0089] Figure 4 is an exemplary flow chart of a method for determining second location information provided in some embodiments of the present application. Figure 4 The process P400 shown may be executed by the display module 610 .

[0090] like Figure 4 As shown, P400 may include the following steps:

[0091] S410: Based on the first position information, create a node control corresponding to the target node while adjusting the viewing angle.

[0092] The node control may be a control that can interact with the annotator. The node control is at least bound to the coordinate information of the target node in the reference direction, so that the annotator can adjust the coordinate information of the target node in the reference direction based on the node control.

[0093] Combined with the above Figure 3 , the node controls under each viewing angle can be bound to the coordinates of the target node under the corresponding viewing angle. So that when the annotator controls the node controls 350 under a certain viewing angle, the node controls of each viewing angle can be linked. For example, the node controls 350 in the partial enlarged view 300B can be bound based on the x-axis coordinates and y-axis coordinates of the target node 340, the node controls 350 in the first adjustment viewing angle 300C can be bound based on the x-axis coordinates and z-axis coordinates of the target node 340, and the node controls 350 in the second adjustment viewing angle 300S can be bound based on the y-axis coordinates and z-axis coordinates of the target node 340.

[0094] Based on the above node control, the second annotation operation includes a control adjustment operation for adjusting the viewing angle of the node control. To determine the second position information, P400 may also include the following steps:

[0095] S420, based on the control adjustment operation, determine the coordinate information of the target node in the reference direction.

[0096] Among them, the annotator adjusts the coordinate information of the node control in the reference direction based on the control adjustment operation, thereby determining the coordinate information of the target node in the reference direction.

[0097] S430: Determine second position information based on the coordinate information of the target node in the reference direction and the first position information.

[0098] The aforementioned second position information can be determined based on the position of the adjusted node control, that is, the coordinate information of the node control in each direction after the adjustment can be used as the second position information of the target node. Considering that the aforementioned control adjustment operation is based on the first position information to adjust at least the coordinate information of the target node in the reference direction, the above corresponding process can also be understood as determining the second position information based on the coordinate information of the target node in the reference direction and the first position information.

[0099] Therefore, based on the above node controls, the second annotation operation of the annotator is simplified. The annotator can refer to the point cloud around the target node to determine the adjustment position when annotating; in addition, the adjustment of the node control under a certain perspective can automatically realize the adjustment of the node control under other perspectives based on the linkage of the node control under different perspectives, thereby facilitating the annotator to determine the second position information.

[0100] Exemplary method for determining initial coordinate information

[0101] Figure 5 is an exemplary flow chart of a method for determining initial coordinate information provided in some embodiments of the present application. Figure 5 The process P500 shown may be executed by the display module 610 .

[0102] like Figure 5 As shown, P500 may include the following steps:

[0103] S510: Based on the first position information, determine a target point cloud that matches the target node from the point cloud to be labeled.

[0104] The target point cloud includes the coordinate information of the target node in the reference direction / points related to the coordinate information in the point cloud to be annotated. For example, the target point cloud may include points within a certain range of the location of the target node.

[0105] S520: Determine initial coordinate information of the target node in the reference direction based on the coordinate information of the target point cloud in the reference direction.

[0106] The initial coordinate information may reflect an estimated value of the coordinate information of the target node in the reference direction based on the target point cloud.

[0107] S530: Based on the initial coordinate information, the target node is displayed by adjusting the viewing angle.

[0108] In some embodiments, the aforementioned S510 may be determined by screening the point cloud under the annotation perspective using the first position information as a screening condition. For example, points whose distances meet preset requirements may be directly screened from the annotation perspective based on the first position information to form a target point cloud.

[0109] In some embodiments, the aforementioned S520 may be performed based on a statistical algorithm such as a mean, a median, etc. That is, when executing the aforementioned S520, the processor may perform statistical processing (such as calculating an average value, a weighted average value, etc.) on the coordinate information of the target point cloud in the reference direction, and then determine the initial coordinate information based on the processing result of the statistical processing.

[0110] In some embodiments, the aforementioned S530 can directly display the target node based on the initial coordinate information. For example, the initial coordinate information can be combined with the first position information to determine the specific position of the target node in the adjustment perspective. Exemplarily, for the first adjustment perspective, the z-axis coordinate of the target node can be determined based on the aforementioned S510 and S520, and then combined with the x-axis coordinate in the first position information, the target node can be presented in the first adjustment perspective.

[0111] In some embodiments, the aforementioned S510 can also be determined based on multiple screenings of the point cloud to be annotated, so as to improve the accuracy of the initial coordinate information, reduce the adjustment of the target node by the second annotation operation, and thus improve the work efficiency of the annotators. At this time, the adjacent point cloud can be determined from the point cloud to be annotated based on the first position information. Then, based on the coordinate information of the adjacent point cloud in the reference direction, the target point cloud is determined from the adjacent point cloud. Among them, the adjacent point cloud includes a point cloud whose position information meets the proximity requirement of the target node. For example, the proximity requirement may be that the distance between the point cloud and the target node in the xy plane is less than a preset threshold. For another example, the proximity requirement may be that after the point cloud to be annotated is clustered with the target node as the center, the adjacent node is a point cloud within the point cloud clustering cluster.

[0112] In some embodiments, considering that the adjacent point cloud is determined only based on the first position information, it may contain point clouds other than the lane line point cloud (such as point clouds corresponding to obstacles covering the lane line), thereby interfering with the determination of the initial coordinate information. Therefore, the adjacent point cloud can be secondary screened based on the coordinate information in the reference direction. At this time, the adjacent point cloud can be screened based on the threshold range of the coordinate information of the adjacent point cloud in the reference direction and the coordinate information of the lane line in the reference direction. The target point cloud is determined based on the processing result of the screening process. Among them, the threshold range can be a range that can reflect the lane line.

[0113] In some embodiments, the aforementioned threshold range can be determined based on the coordinate information of the road surface where the lane line is located in the reference direction. For example, the height range of the lane line relative to the road surface can be estimated according to the lane line painting standard to determine the threshold range.

[0114] Based on the above method for determining the initial coordinate information, a relatively accurate reference direction coordinate information can be estimated as much as possible before the annotator performs the second annotation operation, thereby reducing the adjustment of the target node by the annotator and improving the annotation efficiency.

[0115] Exemplary point cloud annotation device

[0116] Figure 6 This is an exemplary device module diagram of a point cloud annotation device provided in some embodiments of the present application.

[0117] In some embodiments, Figure 6 The point cloud annotation device 600 shown can be integrated in a terminal (such as the user terminal 110). In this case, the display screen of the user terminal can serve as the display module 610 of the point cloud annotation device 600 to present an interactive interface (such as the point cloud annotation interface 130), so that relevant users (such as the annotation personnel 120) can annotate multiple nodes in the point cloud to be annotated based on the interactive operation of the interactive interface to mark the lane lines, thereby executing one or more functions described in the present application.

[0118] like Figure 6 As shown, in order to implement the functions described in this application, the point cloud annotation device 600 may include multiple functional modules such as a display module 610, a node determination module 620, and a node adjustment module 630.

[0119] The display module 610 can be used to display the point cloud to be annotated through the annotation perspective along the reference direction, and to display the target node through the adjustment perspective perpendicular to the reference direction. The point cloud to be annotated includes a point cloud representing a lane line. For more information about displaying the point cloud to be annotated, please refer to the aforementioned S210, S230 and related descriptions.

[0120] Specifically, the display module 610 can be implemented by an independent display device or a display device integrated with other functions; it can be a display device with a physical body or a display device without a physical body. The display device can be implemented in different forms and sizes according to actual needs, and this application does not limit this, but these different implementations should not be considered to exceed the scope of this application.

[0121] The node determination module 620 can be used to determine the target node and its first position information based on the first annotation operation on the point cloud to be annotated, wherein the lane line is annotated by multiple nodes including the target node. For more information about the first annotation operation, please refer to the aforementioned S220 and its related description.

[0122] The node adjustment module 630 can be used to determine the second position information of the target node based on the second annotation operation on the target node and the first position information under the adjustment perspective, wherein the second position information at least includes the coordinate information of the target node in the reference direction. For more information about the second annotation operation, please refer to the aforementioned S240 and its related description.

[0123] In some embodiments, Figure 6 As shown, the point cloud annotation device 600 may further include a lane line annotation module 640. The lane line annotation module 640 may be used to determine the annotation information of the lane line based on the plurality of nodes. For more information on determining the annotation information of the lane line, please refer to the aforementioned S250 and its related description.

[0124] Point cloud annotation interface example

[0125] Figure 7 This is an example diagram of a point cloud annotation interface provided by some embodiments of the present application. Figure 7 With the aforementioned Figure 3 The first and second adjustment perspectives are consistent and both include a marked perspective, a local magnified view of the marked perspective, a first adjustment perspective, and a second adjustment perspective.

[0126] exist Figure 7 In the point cloud annotation interface shown, the interface can display different types of point clouds. Figure 7 It can display point clouds that reflect ground conditions, point clouds that may reflect lane lines (and related signs), and point clouds that may reflect obstacle conditions.

[0127] In some embodiments, the aforementioned different types of point clouds may be presented in different styles in the point cloud annotation interface, so that annotators can construct nodes based on different types of point clouds to annotate lane lines.

[0128] like Figure 7 As shown, in Figure 7In the color interactive interface shown, the nodes can be represented by red cubes, and the target nodes can be represented by nodes with coordinate axes. The blue points can be point clouds that may reflect the ground conditions, the green points can be point clouds that may reflect the lane lines (and related signs), and the red points can be point clouds that may reflect the obstacles. Figure 7 The related text in can be the attribute information added for the lane line. Figure 7 When the interactive interface shown is converted into a black and white image, various point clouds can be presented in different grayscales, so that the labeling personnel can distinguish the various point clouds and label the lane lines based on the point clouds of the lane lines.

[0129] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, and will not be described in detail here.

[0130] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0131] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0132] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0133] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0134] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0135] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program check codes.

[0136] It should be noted that, in the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0137] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A point cloud annotation method, characterized in that: The method comprises: Displaying a point cloud to be annotated through an annotation perspective along a reference direction, wherein the point cloud to be annotated includes a point cloud representing a lane line; Based on a first labeling operation on the point cloud to be labeled, determining a target node and first position information thereof, wherein the lane line is labeled by a plurality of nodes including the target node; Displaying the target node by adjusting the viewing angle perpendicular to the reference direction; Under the adjusted viewing angle, second position information of the target node is determined based on a second marking operation on the target node and the first position information, wherein the second position information at least includes coordinate information of the target node in the reference direction.

2. The point cloud annotation method according to claim 1, characterized in that: The determining, based on the second marking operation on the target node and the first position information, second position information of the target node includes: Adjust the target node based on the second marking operation, and determine coordinate information of the target node in the reference direction; The second position information is determined based on the first position information and the coordinate information of the target node in the reference direction.

3. The point cloud annotation method according to claim 1, characterized in that: The displaying the target node by adjusting the viewing angle perpendicular to the reference direction includes: Based on the first position information, a node control corresponding to the target node is created in the adjusted viewing angle; wherein the node control is at least bound to the coordinate information of the target node in the reference direction.

4. The point cloud annotation method according to claim 3, characterized in that: The second annotation operation includes a control adjustment operation on the node control in the adjustment perspective, and the determining the second position information of the target node based on the second annotation operation on the target node and the first position information includes: Based on the control adjustment operation, determining the coordinate information of the target node in the reference direction; The second position information is determined based on the coordinate information of the target node in the reference direction and the first position information.

5. The point cloud annotation method according to claim 1, characterized in that: The displaying the target node by adjusting the viewing angle perpendicular to the reference direction includes: Based on the first position information, determining a target point cloud matching the target node from the point cloud to be annotated; Based on the coordinate information of the target point cloud in the reference direction, determining the initial coordinate information of the target node in the reference direction; Based on the initial coordinate information, the target node is displayed by adjusting the viewing angle.

6. The point cloud annotation method according to claim 5, characterized in that: The step of determining a target point cloud matching the target node from the point cloud to be annotated based on the first position information includes: Based on the first position information, determine a neighboring point cloud from the point cloud to be annotated, wherein the neighboring point cloud includes points whose position information meets the proximity requirement of the target node; The target point cloud is determined from the neighboring point clouds based on the coordinate information of the neighboring point clouds in the reference direction.

7. The point cloud annotation method according to claim 6, characterized in that: The determining the target point cloud from the adjacent point cloud based on the coordinate information of the adjacent point cloud in the reference direction includes: Performing screening processing on the adjacent point cloud based on the coordinate information of the adjacent point cloud in the reference direction and the threshold range of the coordinate information of the lane line in the reference direction; The target point cloud is determined based on a processing result of the screening process.

8. The point cloud annotation method according to claim 7, characterized in that: The determining the target point cloud from the adjacent point cloud based on the coordinate information of the adjacent point cloud in the reference direction further includes: The threshold range is determined based on coordinate information of the road surface where the lane line is located in the reference direction.

9. The point cloud annotation method according to claim 5, characterized in that: The determining, based on the coordinate information of the target point cloud in the reference direction, the initial coordinate information of the target node in the reference direction includes: Performing statistical processing on the coordinate information of the target point cloud in the reference direction; The initial coordinate information is determined based on a processing result of the statistical processing.

10. The point cloud annotation method according to claim 1, characterized in that: The method further includes: determining labeling information of the lane line based on the multiple nodes.

11. The point cloud annotation method according to claim 10, characterized in that: The determining, based on the multiple nodes, the labeling information of the lane line includes: Determining image data corresponding to the point cloud to be annotated; Performing mapping processing on the image data based on the multiple nodes to determine a mapping image including the multiple nodes; Based on the mapping image and a third labeling operation, labeling information of the lane line is determined, wherein the third labeling operation is at least used to adjust the plurality of nodes.

12. The point cloud annotation method according to claim 11, characterized in that: The performing mapping processing on the image data based on the multiple nodes to determine a mapping image including the multiple nodes includes: Determine a marking line of the lane line based on each target node of the lane line, wherein the marking line includes the multiple nodes of the lane line; Based on the position information of the plurality of nodes, the plurality of nodes and the annotation lines are mapped in the image data to determine a mapping image including the plurality of nodes.

13. The point cloud annotation method according to claim 1, characterized in that: The position information of the annotation viewing angle includes coordinate information in the longitudinal direction of the plane and coordinate information in the transverse direction of the plane, and a reference plane formed by the longitudinal direction and the transverse direction of the plane is perpendicular to the reference direction; the position information of the adjustment viewing angle at least includes coordinate information in the reference direction.

14. The point cloud annotation method according to claim 13, characterized in that: The adjustment viewing angle includes a first adjustment viewing angle and a second adjustment viewing angle, wherein the position information of the first adjustment viewing angle includes coordinate information in the reference direction and coordinate information in the horizontal direction of the plane, and the position information of the second adjustment viewing angle includes coordinate information in the reference direction and coordinate information in the vertical direction of the plane.

15. The point cloud annotation method according to claim 14, characterized in that: The plane is transversely along the extension direction of the lane line in the reference plane, and the plane is longitudinally perpendicular to the first adjustment viewing angle in the reference plane.

16. The point cloud annotation method according to claim 1, characterized in that: The step of displaying the point cloud to be annotated through the annotation viewing angle along the reference direction includes: Determine the point cloud to be annotated and its position information; Determining a display style of the point cloud to be annotated based on the position information of the point cloud to be annotated; The point cloud to be annotated is displayed through the annotation viewing angle based on the position information of the point cloud to be annotated and the display style.

17. The point cloud annotation method according to claim 1, characterized in that: The interface where the annotated perspective is located also includes a local magnified image of the target node and the adjusted perspective.

18. The point cloud annotation method according to claim 1, characterized in that: The method further comprises: Attribute information of the lane line is determined based on a fourth marking operation on the lane line, wherein the fourth marking operation is used to input the attribute information, and the marking information of the lane line is used to divide the lane in the navigation map to determine the lane information of the lane.

19. A point cloud annotation device, characterized in that: The device comprises: A display module, configured to display the point cloud to be annotated through an annotation viewing angle along a reference direction, and to display the target node through an adjustment viewing angle perpendicular to the reference direction; wherein the point cloud to be annotated includes a point cloud representing a lane line; A node determination module, configured to determine a target node and first position information thereof based on a first labeling operation on the point cloud to be labeled, wherein the lane line is labeled by a plurality of nodes including the target node; A node adjustment module is used to determine the second position information of the target node under the adjustment viewing angle based on a second marking operation on the target node and the first position information, wherein the second position information at least includes the coordinate information of the target node in the reference direction.

20. The point cloud annotation device according to claim 19, characterized in that: The device also includes a lane line marking module, which is used to determine the marking information of the lane line based on the multiple nodes.

Images