Data processing method, electronic device, and computer storage medium

Abstract

The application provides a data processing method, an electronic device and a computer storage medium. The data processing method comprises: acquiring a plurality of source poses of a collection device during collection of a source point cloud, the source poses comprising positions and directions of the collection device in a world coordinate system at different time points; acquiring a plurality of source ground truth points in the source point cloud, the source ground truth points being labeled for a target object, the source ground truth points comprising a timestamp and a first three-dimensional coordinate in the world coordinate system; calculating three-dimensional coordinates of the source ground truth points in a collection device coordinate system according to a target time point indicated by the timestamp and a source pose of the collection device at the target time point; acquiring a plurality of new poses of the collection device under a test point cloud; converting the three-dimensional coordinates of the source ground truth points in the collection device coordinate system to second three-dimensional coordinates in the world coordinate system corresponding to the test point cloud according to the new pose of the collection device at the target time point; and calculating feature points of the target object in the test point cloud based on the second three-dimensional coordinates. The method can reduce the labeling cost.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of high-precision map, and particularly relate to a data processing method, an electronic device and a computer storage medium. BACKGROUND

[0002] In the process of map production and positioning of an autonomous vehicle, point cloud matching needs to be performed. For example, a plurality of sets of point cloud data are obtained by performing data collection on a target object at least twice by using a collection device. Since the pose of the collection device is different in different collection times, the three-dimensional coordinates of the same target object calculated from the point cloud data collected in different collection times in a world coordinate system are different, that is, there is a deviation between the point cloud data, and the point cloud data needs to be aligned to eliminate the deviation.

[0003] In order to verify the matching accuracy of the aligned point cloud data, some feature points (such as corner points or end points of the target object) need to be selected from the point cloud data, and the matching accuracy is determined according to the distance deviation between the feature points. However, since the point cloud data is a sampling of the real environment, there is a gap between the points of the point cloud data, and it is very easy to miss the feature points in the sampling process due to interference, occlusion and other reasons, or some feature points are invisible, which makes it difficult to determine the feature points, and manual labeling is required on the point cloud data. However, since the data iteration speed is fast, labeling is required every time, which results in high labor cost and waste of a large amount of manpower and material resources. SUMMARY

[0004] In view of this, embodiments of the present application provide a data processing scheme to at least partially solve the above problems.

[0005] According to a first aspect of embodiments of the present application, a data processing method is provided, comprising: acquiring a plurality of source poses of a collection device during collection of source point cloud, the source poses including positions and directions of the collection device in a world coordinate system at different time instants; acquiring a plurality of source ground truth points labeled on a target object in the source point cloud, the source ground truth points including a time stamp and a first three-dimensional coordinate in the world coordinate system; calculating a three-dimensional coordinate of the source ground truth point in a collection device coordinate system according to a target time instant represented by the time stamp and a source pose of the collection device at the target time instant; acquiring a plurality of new poses of the collection device under test point cloud; converting the three-dimensional coordinate of the source ground truth point in the collection device coordinate system into a second three-dimensional coordinate in the world coordinate system corresponding to the test point cloud according to the new pose of the collection device at the target time instant; and calculating a feature point of the target object in the test point cloud based on the second three-dimensional coordinate.

[0006] According to a second aspect of the embodiments of this application, a data processing method is provided, comprising: displaying a source point cloud of a target object in a display interface; receiving a selection operation on a plurality of shape points in the source point cloud; determining source ground truth points of the target object based on the selected plurality of shape points; and determining feature points of the target object in the test point cloud using the method of the first aspect described above based on the source ground truth points.

[0007] According to a third aspect of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; the memory is used to store at least one executable instruction, wherein the executable instruction causes the processor to perform an operation corresponding to the data processing method described in the first or second aspect.

[0008] According to a fourth aspect of the embodiments of this application, a computer storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the data processing method as described in the first or second aspect.

[0009] According to a fifth aspect of the embodiments of this application, a computer program is provided that, when executed by a processor, implements the method as described in the first or second aspect.

[0010] According to the data processing scheme provided in the embodiments of this application, for the test point cloud transformed from the source point cloud, the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device are determined based on the timestamp and source pose of the source ground truth point corresponding to the source point cloud. Based on the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device and the new pose of the transformed test point cloud, the second three-dimensional coordinates of the source ground truth point corresponding to the test point cloud in the world coordinate system are determined. In this way, it is not necessary to label the ground truth point under the test point cloud, and the source ground truth point in the source point cloud can be reused, thereby reducing costs and time. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.

[0012] Figure 1A This is a flowchart of the steps of a data processing method according to Embodiment 1 of this application;

[0013] Figure 1B for Figure 1A A schematic diagram of the annotation of the source truth points of a long line in the embodiment shown;

[0014] Figure 1C Figure 1 shows a schematic diagram of a short marking line in an embodiment of the present application; Figure 1A Figure 2 shows a schematic diagram of a source true value point of a short marking line in an embodiment of the present application;

[0015] Figure 1D Figure 3 shows a schematic diagram of a road surface with an arrow direction sign in an embodiment of the present application; Figure 1A Figure 4 shows a schematic diagram of a source true value point of a road surface with an arrow direction sign in an embodiment of the present application;

[0016] Figure 1E Figure 5 shows a schematic diagram of a sign in an embodiment of the present application; Figure 1A Figure 6 shows a schematic diagram of a source true value point of a sign in an embodiment of the present application;

[0017] Figure 1F Figure 7 shows a schematic diagram of a projection fitting in an embodiment of the present application; Figure 1A Figure 8 shows a schematic diagram of a projection fitting in an embodiment of the present application;

[0018] Figure 1G Figure 9 shows a schematic diagram of a sub-step of step S106 according to Embodiment One of the present application;

[0019] Figure 1H Figure 10 shows a schematic diagram of a sub-step of step S112 according to Embodiment One of the present application;

[0020] Figure 2 Figure 11 shows a step flow chart of a data processing method according to Embodiment Two of the present application;

[0021] Figure 3 Figure 12 shows a structural block diagram of a cloud server according to Embodiment Three of the present application;

[0022] Figure 4 Figure 13 shows a structural schematic diagram of an electronic device according to Embodiment Four of the present application. DETAILED DESCRIPTION

[0023] In order to make the personnel in the art better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by the person skilled in the art should belong to the scope of protection of the embodiments of the present application.

[0024] The specific implementation of the embodiments of the present application will be further described below in combination with the drawings of the embodiments of the present application.

[0025] Embodiment One

[0026] Referring to Figure 11, a step flow chart of a data processing method according to Embodiment One of the present application is shown. Figure 1A

[0027] ​For the convenience of clearly describing the method of the embodiment, before the implementation of the method is described in detail, the application scenario is exemplarily described as follows:

[0028] The method can be applied to scenarios such as map production and positioning in the automatic driving process. Taking map production as an example, in order to obtain map data, a vehicle (such as a high-precision map vehicle equipped with a laser radar) carrying a collection device can be driven along a road, and the target objects (such as trees, signboards, and markings of the road) on both sides of the road are scanned by the collection device during driving, so as to obtain point cloud data. The points of the point cloud data correspond to a certain point on the target object. Based on the three-dimensional coordinates P l of the points in the point cloud data in the coordinate system of the collection device, the pose T wi of the collection device in the world coordinate system, and the parameters T il of the collection device, the three-dimensional coordinates P w of the target object in the world coordinate system, that is, the geographical position thereof, can be determined, thereby realizing the production of the map. Based on the point cloud data, the three-dimensional coordinates of the target object in the world coordinate system can be determined, and the map can be made. Specifically, the three-dimensional coordinates P w may be expressed as: P w =T wi ·T il ·P l .

[0029] Since the real environment changes, in order to ensure that the map can be updated in time when the environment changes, the road can be driven once every period of time, so as to collect newer point cloud data. However, when the road is driven, it cannot be ensured that the trajectory of the vehicle is the same every time, that is, the pose of the collection device can be different, so that even if the position of the target object does not change, the three-dimensional coordinates P w of the same target object in the world coordinate system determined by the point cloud data collected at different times can be different, and thus the point cloud data collected at different times needs to be aligned.

[0030] The alignment manner is, for example, that the point cloud data A is taken as the target point cloud, and the point cloud data B is taken as the source point cloud. The point cloud data B is converted to obtain new point cloud data C (denoted as test point cloud), and the distance between the same feature points (that is, the same point on the same target object, which can be referred to as the same name point) in the point cloud data C and the point cloud data A is calculated to determine the matching accuracy. It should be noted that the feature points are, for example, the corner points or end points of the target object, but are not limited thereto, and can also be the center axis or edge line of the target object.

[0031] From the above alignment process, in order to calculate the distance between the feature points, the feature points need to be determined from the point cloud data C and the point cloud data A respectively, but due to the lack of point cloud density, occlusion causes the feature points to be missing or the feature points themselves are invisible, etc. Directly obtain the feature points. This requires manual labeling of some points in the point cloud data (because the accuracy of manually labeled points is high, so it can be called ground truth points), based on these manually labeled ground truth points, the matching accuracy can be directly determined, or the feature points can be calculated according to the ground truth points, and then the matching accuracy can be calculated according to the feature points, but no matter which way it is, there is a problem of frequent iteration of point cloud data and re-labeling of ground truth points, which causes a large labor burden under the condition of frequent iteration of data.

[0032] In order to solve this problem, an embodiment of the present application provides a method, by which the reuse of ground truth points can be realized, thereby reducing the cost and time of labeling ground truth points on point cloud data.

[0033] In the embodiment, the method comprises the following steps:

[0034] Step S102: Obtain a plurality of source poses of the acquisition device during acquisition of the source point cloud.

[0035] The source pose includes the position and direction of the acquisition device in the world coordinate system at different times. A plurality of source poses can constitute a source trajectory of the acquisition device. Generally, the source pose of the acquisition device can be detected based on the IMU unit (inertial measurement unit) on the vehicle carrying the acquisition device.

[0036] Step S104: Obtain a plurality of source ground truth points labeled on the target object in the source point cloud.

[0037] In the embodiment, in order to facilitate labeling, the source ground truth points are visible shape points (denoted as SP) in the target object or points calculated based on the visible shape points. The source ground truth points include a timestamp and a first three-dimensional coordinate in the world coordinate system. The source point cloud can be a point cloud obtained by scanning the target object by a laser radar.

[0038] The number and position of the source ground truth points labeled on the target object of different geometric shapes can be different. For example, the target object includes at least one of a first type of road marking, a second type of road marking, a sign, a road surface arrow direction mark, and a round rod body, the length of the first type of road marking is greater than a set value, and the length of the second type of road marking is less than or equal to the set value. The set value can be the maximum value of the second road marking. For example, 3m, 5m, etc.

[0039] For example, as shown in FIG. 1, the target object is a road surface arrow direction mark, and the source ground truth points are the visible shape points of the target object. Figure 1BAs shown, the target object is the first type of road marking (also called long marking, such as long solid line on the road, stop line, etc.), which has a rectangular geometric shape and a large ratio of length to width. For such a target object, the source ground truth points of the target object include at least two pairs of points, and the two points in a pair are located on different sides of the center axis of the target object. Of course, in order to obtain more accurate feature points (i.e., points on the center axis) subsequently, multiple pairs of points can be labeled along the direction of the center axis. The timestamp of the source ground truth point can be the time when the point is collected.

[0040] As shown, Figure 1C the target object is the second type of road marking (also called short marking, such as dashed line or zebra crossing on the road, etc.), which also has a rectangular geometric shape, but has a smaller ratio of length to width than the first type of road marking. For such a target object, the 6-point method can be used to label the source ground truth points, that is, the source ground truth points include at least two points on each side of the target object, and a top point and a bottom point in the direction of the center axis of the target object. The timestamp of the source ground truth point can be the time when the point is collected.

[0041] An example of a labeling method is, for example, along the direction of the center axis of the target object, 2 source ground truth points are labeled on the left long edge to determine the direction of the left edge, 2 source ground truth points are labeled on the right long edge to determine the direction of the right edge, 1 source ground truth point is selected on the bottom short edge, which does not need to be on the center axis (in fact, it is likely that there is no point on the center axis), but needs to select a point that is the bottom of the edge that can be recognized by the naked eye, and 1 source ground truth point is selected on the top short edge, which also does not need to be on the center axis, but needs to select a point that is the top of the edge that can be recognized by the naked eye.

[0042] As shown, Figure 1D and 1E the target object is the sign or the road direction sign with an arrow, when the target object is a sign, it has a rectangular or rounded rectangular geometric shape, and the source ground truth points include at least two points on each side of the target object. The timestamp of the source ground truth point can be the time when the point is collected.

[0043] An example of a labeling method is, for example, the 8-point method, that is, facing the front of the sign, starting from the left edge, selecting 2 source ground truth points on each edge in a counterclockwise order.

[0044] When the target object is a road direction sign with an arrow, its geometric shape is complex and can be a single arrow or a combination of multiple arrows, and the source ground truth points include at least two points on each side of the target object. The timestamp of the source ground truth point can be the time when the point is collected.

[0045] An example of the labeling manner is, for example, starting from the bottom short side, selecting 2 ground truth points on each edge in a counterclockwise order.

[0046] When the target object is a circular rod, the geometric shape thereof is a cylinder, and the ground truth points include a center point of a top end section of the target object and a center point of a bottom end section of the target object. Optionally, to further completely represent the cylinder, a radius of the top end section and a radius of the bottom end section can also be included.

[0047] It should be noted that, since the ground truth points of the circular rod are not points actually existing in the source point cloud, but points calculated based on points in the source point cloud, the time stamp of the ground truth points can be the corresponding average time of the point cloud contained by the circular rod.

[0048] A feasible process for determining the ground truth points is, for example, as follows.

[0049] Process A1: determining a main axis direction of the circular rod according to a point cloud contained by the circular rod.

[0050] The point cloud contained by the circular rod can be a point cloud selected manually and denoised, so as to remove interference points and improve accuracy. By performing principal component analysis on the point cloud contained by the circular rod, the main axis direction thereof can be determined.

[0051] Process B1: selecting a first end point set and a second end point set from the point cloud contained by the circular rod according to the main axis direction.

[0052] A feasible way of obtaining the first end point set and the second end point set can be as follows: slicing the point cloud contained by the circular rod at a first distance from an apex along a direction perpendicular to the main axis direction to obtain a top slicing cylinder segment, and points located in the top slicing cylinder segment belong to the first end point set. Similarly, slicing the point cloud contained by the circular rod at a second distance from a lowest point along a direction perpendicular to the main axis direction to obtain a bottom slicing cylinder segment, and points located in the bottom slicing cylinder segment belong to the second end point set.

[0053] The first distance and the second distance can be determined as needed, and the two can be the same or different, which is not limited. For example, the first distance and the second distance are the same, and are one fifth of the distance from the apex to the lowest point. Since the more points contained by the first end point set and the second end point set, the greater the calculation load, the first distance and the second distance can also be determined according to the computing power.

[0054] Process C1: projecting points in the first end point set onto a top end surface of the circular rod along the main axis direction, and projecting points in the second end point set onto a bottom end surface of the circular rod along the main axis direction.

[0055] The top end surface can be an end surface passing through the top point and perpendicular to the main axis direction. The shape and size of the top end section of the circular rod body can be obtained by projecting the points in the first end point set along the main axis direction onto the top end surface. In this way, the problem that the point cloud is sparse and the top end section cannot be accurately determined can be solved.

[0056] Similarly, the bottom end section of the circular rod body can be obtained by projecting the points in the second end point set along the main axis direction onto the bottom end surface of the circular rod body.

[0057] Process D1: determining the ground truth point of the top end of the circular rod body according to the projection result of the first end surface.

[0058] The projection result of the first end surface can fit the top end section of the circular rod body, as shown in FIG. 2. Figure 1F The center point of the center of the top end section can be determined as the ground truth point of the top end.

[0059] Process E1: determining the ground truth point of the bottom end of the circular rod body according to the projection result of the second end surface.

[0060] The projection result of the second end surface can fit the bottom end section of the circular rod body, and the center point of the center of the bottom end section can be determined as the ground truth point of the bottom end.

[0061] In summary, regardless of the geometric shape of the target object, the ground truth point in the embodiment can be determined by labeling the visible points in the source point cloud, thereby reducing the labeling difficulty and improving the labeling efficiency.

[0062] Step S106: calculating the three-dimensional coordinates of the ground truth point in the coordinate system of the acquisition device according to the target time indicated by the time stamp and the source pose of the acquisition device at the target time.

[0063] Since the ground truth point not only includes its three-dimensional coordinates in the world coordinate system, but also includes the time stamp, the source pose of the acquisition device when the source point cloud is collected can be determined based on the time stamp, and then the three-dimensional pose of the source point cloud in the coordinate system of the acquisition device can be inversely calculated. The inverse calculation principle is as follows:

[0064] The first three-dimensional coordinates P w can be expressed as: P w =T wi ·T il ·P l . Wherein, P w is the first three-dimensional coordinates. T wi is the source pose of the acquisition device at the i-th time. T il is the intrinsic parameter of the acquisition device and the IMU unit (inertial measurement unit). P lThis can be the three-dimensional coordinates of the source truth point in the coordinate system of the acquisition device. Once the inherent parameters of the acquisition device (such as installation position and angle) are determined, they will not change, and the three-dimensional coordinates of the source truth point in the coordinate system of the acquisition device also cannot be changed after acquisition. Therefore, given P... w T wi 、 and T il In certain cases, the three-dimensional coordinates can be calculated by inverse solution.

[0065] The three-dimensional coordinates calculated by inverse kinematics can be denoted as P. i It can be represented as P i =T wi -1 ·P w .

[0066] In one example, such as Figure 1G As shown, step S106 can be implemented through the following sub-steps:

[0067] Sub-step S1061: Obtain the source pose of the acquisition device at the target time.

[0068] Because the acquisition device acquires point cloud data at a higher frequency than the IMU unit outputs source poses, the source point cloud at some moments does not have a directly corresponding source pose. To solve this problem, sub-step S1061 can be implemented through the following process:

[0069] Process A1: Determine whether the time corresponding to the multiple source poses includes the target time.

[0070] If the time corresponding to the source pose includes the target time, the source pose corresponding to the target time can be obtained directly. Alternatively, if it does not include the target time, process B1 is executed.

[0071] Process B1: Obtain the preceding and following times adjacent to the target time from the times corresponding to the multiple source poses.

[0072] For example, if the target time is time t, the previous time could be time t-1, and the later time could be time t+1.

[0073] Process C1: Obtain the source pose at the previous time step and the source pose at the later time step.

[0074] Process D1: Interpolate the source pose at the previous time step and the source pose at the later time step, and determine the source pose of the acquisition device at the target time step based on the interpolation result.

[0075] For example, linear interpolation is performed on the source pose at the earlier time step and the source pose at the later time step to determine the source pose at the target time step.

[0076] Sub-step S1062: determining the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device according to the first three-dimensional coordinates of the source ground truth point at the target time and the source pose of the acquisition device at the target time.

[0077] The first three-dimensional coordinates are denoted as P w The source pose is denoted as T wi The three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device can be expressed as: P i i = T wi -1 · P w .

[0078] Step S108: obtaining a plurality of new poses of the acquisition device under the test point cloud.

[0079] The new poses of the acquisition device are used to map the source point cloud to the test point cloud. For example, when aligning the source point cloud with the target point cloud, the source point cloud is mapped to the test point cloud by changing the source pose to the new pose. In this embodiment, the new poses correspond to the source poses one by one, and the plurality of new poses can constitute a new trajectory of the acquisition device. In the process of aligning the source point cloud with the target point cloud, the error in alignment can be reduced by constantly changing the new pose.

[0080] Step S110: converting the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device into second three-dimensional coordinates in the world coordinate system corresponding to the test point cloud according to the new pose of the acquisition device at the target time.

[0081] When solving the source ground truth point, the new pose of the source ground truth point at the target time needs to be obtained. Since the sampling frequency of the source point cloud is higher than that of the IMU unit, a part of the source point cloud at a time does not have a corresponding new pose.

[0082] To solve this problem, the new pose corresponding to the target time can be obtained in the following way: determining whether the time corresponding to the plurality of new poses contains the target time. If the time corresponding to the new pose contains the target time, the corresponding new pose can be directly obtained.

[0083] Alternatively, if the target time is not contained, the time before and after the target time adjacent to the target time is obtained from the time corresponding to the plurality of new poses. The new pose of the time before and the new pose of the time after are obtained. Linear interpolation is performed on the new pose of the time before and the new pose of the time after, and the new pose of the acquisition device at the target time is determined according to the interpolation result.

[0084] Based on the obtained new pose, the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device can be expressed as: P w = T wi · P​i , the correct solution is to calculate the second three-dimensional coordinates P' of the source ground truth point in the world coordinate system corresponding to the test point cloud w .

[0085] Step S112: based on the second three-dimensional coordinates, calculate the feature points of the target object in the test point cloud.

[0086] Different ways can be used to calculate the feature points in the test point cloud for target objects of different geometric shapes. In an example, as shown in FIG. 12, step S112 can include the following sub-steps: Figure 1H

[0087] Step S1121: determine the geometric shape corresponding to the target object.

[0088] For example, the first type of road marking and the second type of road marking can be rectangular. The sign can be rectangular, rounded rectangular, or circular, etc. The round rod body can be cylindrical. The road surface arrow direction mark can be an arrow or a complex geometric body fused by multiple arrows, etc.

[0089] Step S1122: according to the geometric shape of the target object and the second three-dimensional coordinates of the source ground truth point, calculate the three-dimensional coordinates of the feature points of the target object in the test point cloud in the world coordinate system.

[0090] When the target object is the first type of road marking, the three-dimensional coordinates of at least two feature points on the center axis of the target object in the world coordinate system are calculated according to the second three-dimensional coordinates of at least two pairs of source ground truth points. In this way, the center axis can be determined through at least two feature points. The center axis can evaluate the matching accuracy in the remaining 2 directions other than the direction of the center axis.

[0091] When the target object is the second type of road marking, the corresponding edge is calculated according to the second three-dimensional coordinates of the source ground truth points on each edge, the center axis of the second type of road marking is calculated according to each edge, the top end point on the center axis is obtained by projecting the second three-dimensional coordinates of the top point onto the center axis. The bottom end point on the center axis is obtained by projecting the second three-dimensional coordinates of the bottom point onto the center axis. The center axis can be determined through the top end point and the bottom end point, and the end points of the center axis are determined, so the matching accuracy can be determined in 3 directions.

[0092] When the target object is a sign or a road surface arrow direction mark, the edge is calculated according to the second three-dimensional coordinates of the source ground truth points on the edge, and the three-dimensional coordinates of the intersection point of the adjacent two edges in the world coordinate system are the three-dimensional coordinates of the corner point.

[0093] If the target object is a round rod body, the second three-dimensional coordinates of the source ground truth point are taken as the three-dimensional coordinates of the feature point in the world coordinate system. ​

[0094] In determining the matching accuracy of the test point cloud and the target point cloud, the three-dimensional coordinates of the feature points in the test point cloud in the world coordinate system and the three-dimensional coordinates of the same feature points in the target point cloud in the world coordinate system are calculated to obtain the distance between the two feature points. If the distance is less than a distance threshold (which can be determined as needed), it indicates that the matching accuracy of the test point cloud meets the requirements, otherwise it indicates that it does not meet the requirements.

[0095] By the labeling method of the embodiment, the problem that the actual feature points of some target objects are not directly visible is solved. The shape points that can be significantly distinguished are used as source ground truth points, and the three-dimensional coordinates of the accurate feature points can be calculated based on the source ground truth points. For the problem that the test point cloud is different after each conversion of the source point cloud, the source ground truth points are recalculated based on the time stamp added to the shape points and the source pose and new pose of the acquisition device, so that the ground truth points can be labeled once and used permanently.

[0096] The method only needs to select easily distinguishable shape points as source ground truth points, and then automatically calculates accurate semantic feature points based on the source ground truth points (even if these feature points are not visible in the source laser point cloud), thereby ensuring the reliability of the evaluation index. It is also applicable to low-cost lidar with low point density, greatly expanding the scope of application of the method.

[0097] By adding a time stamp to the source ground truth points and combining the source pose and the new pose, the three-dimensional coordinates of the source ground truth points in the world coordinate system relative to the test point cloud can be calculated. The three-dimensional coordinates of the newly calculated source ground truth points match the test point cloud, so that the ground truth can be labeled once and used permanently, greatly saving time and labor costs.

[0098] This way solves the problem that the existing automatic selection of the same name point is affected by the sampling density of the laser point cloud, and the selected same name point is not the actual same name point, leading to inaccurate evaluation. It also solves the problem that in some cases, the actual feature points (corner points, end points, etc.) may have data missing, leading to the inability to directly select, and the laser point cloud of the rod-shaped object can only be sampled on the outer surface, making it impossible to directly label the feature line (such as the axis); most landmark lines have a certain width, and directly selecting the central axis has a large error.

[0099] In summary, the source ground truth points of the method can select easily distinguishable shape points, which can automatically calculate accurate semantic feature points based on the shape points, thereby ensuring the reliability of the evaluation index and greatly expanding the scope of application of the method. Since the source ground truth points have a time stamp, the source ground truth points of the test point cloud can be determined based on the time stamp, the source pose and the new pose after the source point cloud is converted, thereby realizing ground truth reuse and greatly saving time and labor costs.

[0100] Embodiment Two

[0101] Referring to Figure 2 FIG. 2 shows a step flow chart of the data processing method according to the embodiment two of the present application.

[0102] The data processing method can be configured in a display terminal, and the method comprises the following steps:

[0103] Step S202: displaying a source point cloud of a target object in a display interface.

[0104] The display interface can be an interface of the display terminal, and the source point cloud of the target object is displayed in the display interface to facilitate a user to mark source ground truth points therefrom.

[0105] Step S204: obtaining a plurality of shape points in the source point cloud as the source ground truth points.

[0106] In an example, the user can select some shape points in the displayed source point cloud as the source ground truth points. For example, two points located on different sides of a center axis of a first type of road marking can be selected as a pair of points, and at least two pairs of points are selected from the source point cloud as the source ground truth points.

[0107] For a second type of road marking, two points can be selected from each of two side edges as the source ground truth points, and a top point and a bottom point can be selected as the source ground truth points.

[0108] For a signboard or a direction sign with an arrow on a road surface, at least two points can be selected from each edge as the source ground truth points.

[0109] For a circular rod, interference points can be manually removed, and then the point cloud of the circular rod is analyzed by an algorithm to determine a center point of a top cross section and a center point of a bottom cross section of the circular rod as the source ground truth points.

[0110] An automatically generated mode can comprise the following steps: determining a main axis direction of the circular rod according to the point cloud contained in the circular rod; selecting a first end point set and a second end point set from the point cloud contained in the circular rod according to the main axis direction; projecting points in the first end point set onto a top end surface of the circular rod along the main axis direction, and projecting points in the second end point set onto a bottom end surface of the circular rod along the main axis direction; determining a source ground truth point of a top end of the circular rod according to a projection result of the first end surface; and determining a source ground truth point of a bottom end of the circular rod according to a projection result of the second end surface.

[0111] The point cloud of the circular rod can be determined by manual screening and denoising. The main axis direction can be determined by principal component analysis. The first end point cloud and the second end point set can be determined by manually selecting a slice position, or the distance of the slice can be directly set.

[0112] The timestamp of the source ground truth point can be an average time corresponding to the point cloud contained in the round stick.

[0113] Step S206: Based on the source ground truth point, the feature point of the target object in the test point cloud is determined using the method described above.

[0114] For the source ground truth point, the three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device can be inversely calculated according to the first three-dimensional coordinates of the source ground truth point in the world coordinate system and the source pose of the corresponding acquisition device, and then the second three-dimensional coordinates of the source ground truth point in the world coordinate system corresponding to the test point cloud can be calculated according to the new pose of the acquisition device and the inversely calculated three-dimensional coordinates of the source ground truth point in the coordinate system of the acquisition device.

[0115] According to the second three-dimensional coordinates of the source ground truth point and the geometric shape of the target object, the three-dimensional coordinates of the feature point of the target object in the test point cloud in the world coordinate system can be calculated.

[0116] The data processing method of the embodiment only needs to point and select a shape point that is easy to distinguish as a source ground truth point, and accurate feature points can be automatically generated by the algorithm, thereby reducing the difficulty of labeling the feature points. This method is also applicable to low-cost lidar with low point density, greatly expanding the scope of application.

[0117] Embodiment three

[0118] Referring to Figure 3 , a structural schematic diagram of a cloud server according to Embodiment Three of the present application is shown.

[0119] The cloud server 302 can be connected with the client terminal, and the cloud server is used to receive the source ground truth points labeled on the source point cloud transmitted by the client terminal, and execute the method described above according to the source ground truth points, so as to reuse the source ground truth points labeled on the source point cloud to the test point cloud after converting the source point cloud to the test point cloud, thereby not needing to label the ground truth points on the test point cloud again, so as to reduce the workload of labeling.

[0120] Embodiment four

[0121] Referring to Figure 4 , a structural schematic diagram of an electronic device according to Embodiment Four of the present application is shown, and the specific implementation of the electronic device is not limited in the specific embodiments of the present application.

[0122] As Figure 4 shown, the electronic device can include a processor 402, a communications interface 404, a memory 406, and a communications bus 408.

[0123] wherein:

[0124] The processor 402, the communication interface 404, and the memory 406 communicate with each other through the communication bus 408.

[0125] The communication interface 404 is configured to communicate with other electronic devices or servers.

[0126] The processor 402 is configured to execute the program 410, and specifically can execute the related steps in the foregoing data processing method embodiments.

[0127] Specifically, the program 410 can include program codes including computer operation instructions.

[0128] The processor 402 can be a processor CPU, or an application specific integrated circuit ASIC, or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the smart device can be processors of the same type, such as one or more CPUs; or can be processors of different types, such as one or more CPUs and one or more ASICs.

[0129] The memory 406 is configured to store the program 410. The memory 406 can include a high-speed RAM memory, and can also include a non-volatile memory such as at least one disk memory.

[0130] The program 410 can be specifically used for causing the processor 402 to perform the operations corresponding to the foregoing method.

[0131] The specific implementation of each step in the program 410 can refer to the corresponding description in the corresponding steps and units in the foregoing data processing method embodiments, and will not be described here. Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the devices and modules described above can refer to the corresponding process description in the foregoing method embodiments, and will not be described here.

[0132] It should be noted that, according to the needs of implementation, each component / step described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or part of the operations of the components / steps can be combined into a new component / step, to achieve the purpose of the embodiments of the present application.

[0133] The methods according to the embodiments of the present application described above can be implemented in hardware, firmware, or software, or a combination of them, and can be stored in a recording medium such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk, or downloaded from a network and stored in a local recording medium, so that the methods described herein can be processed by such software using a general-purpose computer, a special-purpose processor, or programmable or dedicated hardware such as an ASIC or FPGA. It can be understood that the computer, processor, microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the data processing methods described herein. Furthermore, when a general-purpose computer accesses code for implementing the data processing methods shown herein, the execution of the code converts the general-purpose computer into a special-purpose computer for executing the data processing methods shown herein.

[0134] Those skilled in the art can understand that the units and method steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of the present application.

[0135] The above embodiments are only used to illustrate but not limit the embodiments of the present application, and a person of ordinary skill in the art can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application, therefore all equivalent technical solutions belong to the scope of the embodiments of the present application, and the patent protection scope of the embodiments of the present application should be defined by the claims.

Claims

1. A data processing method, comprising: obtaining a plurality of source poses of a collection device during collection of a source point cloud, the source poses comprising positions and orientations of the collection device in a world coordinate system at different time instants; obtaining a plurality of source ground truth points in the source point cloud, the source ground truth points being labeled for a target object, the source ground truth points comprising time stamps and first three-dimensional coordinates in the world coordinate system, the source ground truth points being shape points visible in the target object or points calculated based on the shape points visible in the target object; calculating three-dimensional coordinates of the source ground truth points in a coordinate system of the collection device according to a target time instant indicated by the time stamps and a source pose of the collection device at the target time instant; obtaining a plurality of new poses of the collection device under a test point cloud, the test point cloud being obtained by transforming the source point cloud; transforming the three-dimensional coordinates of the source ground truth points in the coordinate system of the collection device into second three-dimensional coordinates in the world coordinate system corresponding to the test point cloud according to the new pose of the collection device at the target time instant; calculating feature points of the target object in the test point cloud based on the second three-dimensional coordinates.

2. The method of claim 1, wherein, The calculating of the three-dimensional coordinates of the source ground truth points in the coordinate system of the collection device according to the target time instant indicated by the time stamps and the source pose of the collection device at the target time instant comprises: obtaining the source pose of the collection device at the target time instant; determining the three-dimensional coordinates of the source ground truth points in the coordinate system of the collection device according to the first three-dimensional coordinates of the source ground truth points at the target time instant and the source pose of the collection device at the target time instant.

3. The method of claim 2, wherein, The obtaining of the source pose of the collection device at the target time instant comprises: determining whether the target time instant is included in time instants corresponding to the plurality of source poses; if the target time instant is not included, obtaining a preceding time instant and a following time instant adjacent to the target time instant from the time instants corresponding to the plurality of source poses; obtaining the source pose at the preceding time instant and the source pose at the following time instant; interpolating the source pose at the preceding time instant and the source pose at the following time instant, and determining the source pose of the collection device at the target time instant according to an interpolation result.

4. The method of claim 1, wherein, The calculating of the feature points of the target object in the test point cloud based on the second three-dimensional coordinates comprises: determining a geometric shape corresponding to the target object; calculating three-dimensional coordinates of the feature points of the target object in the test point cloud in the world coordinate system according to the geometric shape of the target object and the second three-dimensional coordinates of the source ground truth points at the target time instant.

5. The method of claim 1, wherein, The target object comprises at least one of a first type of road marking line, a second type of road marking line, a sign, a direction sign with an arrow on a road surface, and a round pole body, the first type of road marking line having a length greater than a set value, and the second type of road marking line having a length less than or equal to the set value.

6. The method of claim 5, wherein, if the target object is the first type of road marking line, the source ground truth points of the target object comprise at least two pairs of points, and two points in a pair of points are located on different sides of a center axis of the target object; and / or, if the target object is a second type of road marking, the source ground truth points include at least two points on each side of the target object, and a top point and a bottom point in the direction of the center axis of the target object; and / or, if the target object is a signboard or a road surface arrow direction sign, the source ground truth points include at least two points on each side of the target object; and / or, if the target object is a round rod, the source ground truth points include a center point of a top end section of the target object and a center point of a bottom end section of the target object.

7. A data processing method, comprising: displaying a source point cloud of a target object in a display interface; receiving a selection operation on a plurality of shape points in the source point cloud; determining source ground truth points of the target object according to the selected plurality of shape points; determining feature points of the target object in the test point cloud based on the source ground truth points, using the method of any one of claims 1-6.

8. The method of claim 7, wherein, The determining of the source ground truth points of the target object according to the selected plurality of shape points comprises: if the target object is a round rod, determining a main axis direction of the round rod according to a point cloud contained by the round rod; selecting a first end point set and a second end point set from the point cloud contained by the round rod according to the main axis direction; projecting points in the first end point set onto a top end surface of the round rod along the main axis direction, and projecting points in the second end point set onto a bottom end surface of the round rod along the main axis direction; determining source ground truth points of a top end of the round rod according to the projection of the points in the first end point set onto the top end surface of the round rod along the main axis direction; determining source ground truth points of a bottom end of the round rod according to the projection of the points in the second end point set onto the bottom end surface of the round rod along the main axis direction.

9. The method of claim 8, wherein, The determining of the source ground truth points of the target object according to the selected plurality of shape points further comprises: determining an average time corresponding to the point cloud contained by the round rod as a time stamp of the source ground truth points of the top end and the bottom end of the round rod.

10. The method of claim 8 or 9, wherein, The determining of the source ground truth points of the top end of the round rod according to the projection of the points in the first end point set onto the top end surface of the round rod along the main axis direction comprises: fitting a top end section of the round rod according to the projection of the points in the first end point set onto the top end surface of the round rod along the main axis direction, and taking a center point of the top end section as the source ground truth points of the top end.

11. The method of claim 8 or 9, wherein, The determining of the source ground truth points of the bottom end of the round rod according to the projection of the points in the second end point set onto the bottom end surface of the round rod along the main axis direction comprises: fitting a bottom end section of the round rod according to the projection of the points in the second end point set onto the bottom end surface of the round rod along the main axis direction, and taking a center point of the bottom end section as the source ground truth points of the bottom end.

12. An electronic device comprising: a processor, a memory, a communication interface, and a communication bus, the processor, the memory, and the communication interface being in communication with each other through the communication bus; The memory is configured to store at least one executable instruction, and the executable instruction causes the processor to perform operations corresponding to the data processing method in any one of claims 1-6 or any one of claims 7-11. 13.A computer storage medium having stored thereon a computer program, which, when executed by a processor, implements the data processing method in any one of claims 1-6 or any one of claims 7-11.

14. A computer program product comprising a computer program, wherein, The program, when executed by the processor, implements the data processing method in any one of claims 1-6 or any one of claims 7-11.

Images