Method for creating digital maps for a vehicle or robot

Abstract

The invention relates to a method (100) for creating a digital map for controlling a vehicle or a robot. Furthermore, the invention relates to a device, a computer program, and a machine-readable storage medium.

Description

[0001] The invention relates to a method for creating digital maps for a vehicle or a robot. Furthermore, the invention comprises a device, a computer program, and a machine-readable storage medium. State of the art

[0002] Methods for creating digital maps are known from the state of the art.

[0003] Digital maps can be created using measurement data collected during test drives, for example, using a vehicle's sensors. The completeness and precision of such digital maps can be increased by conducting multiple test drives or by having multiple sensors record measurement data during a single test drive. The measurement data obtained in this way may differ in part due to external circumstances, for example, even if they describe a virtually identical environmental area. Disclosure of the invention

[0004] It is therefore an object of the invention to provide an improved method for creating digital maps, which provides an improved digital map.

[0005] This object is achieved by a method according to the independent patent claim. Advantageous embodiments are specified in the dependent claims.

[0006] According to a first aspect of the invention, a method for creating a digital map for a vehicle or a robot is proposed. The method comprises the following steps: - Using at least a first point cloud sequence and a second point cloud sequence, each point cloud sequence comprising at least a first point cloud and a second point cloud, a point cloud comprising measurement points with measurement data over an area surrounding a location position, the first point cloud and the second point cloud at least partially overlapping in a first overlap area, - Determining a first point cloud pair, each comprising a point cloud from the first point cloud sequence and a point cloud from the second point cloud sequence, wherein the point clouds of the first point cloud pair each represent an at least partially identical first area of ​​the environment, wherein first corresponding elements are determined for the measurement points of the point clouds of the first point cloud pair, - Determining a second point cloud pair, each comprising a point cloud from the first point cloud sequence and a point cloud from the second point cloud sequence, wherein the point clouds of the second point cloud pair represent an at least partially identical second area of ​​the environment, wherein second corresponding elements are determined for the measurement points of the point clouds of the second point cloud pair, - Determining first further corresponding elements for the measuring points of the first overlap area from the point clouds of the first and the second point cloud sequence, - Calculating a transformation for the point clouds of the first and second point cloud sequence based on the first, second and first further corresponding elements, - Creating a digital map with measurement points and measurement data based on the transformed point clouds of the first and second point cloud sequence using several location positions.

[0007] The map can be used to control a function, in particular a driving function of a vehicle or robot. The technical advantage can be achieved of providing an improved method for creating a digital map with measurement points and measurement data. This is achieved by using a large number of specific corresponding elements to calculate the transformation of the first point cloud sequence and the second point cloud sequence. The individual point clouds of the point cloud sequence have a first overlap area for this purpose. To estimate the transformation, in addition to the corresponding elements of a point cloud pair from the point cloud sequences, corresponding elements from the overlap area are used. By using the corresponding elements from the overlap area, it is possible to reduce a potential error in the calculation for the transformation of the point cloud sequences.Thus, an improvement in accuracy and an improvement in error reduction can be provided for the creation of a digital map. Furthermore, the proposed method demonstrates an efficient approach for creating a digital map for a vehicle and / or a robot from different point cloud sequences.

[0008] In an advantageous embodiment, the measurement points of the measurement data for the point clouds are recorded using a sensor. The sensor records the measurement points at periodic intervals and preferably at a constant intensity.

[0009] This provides the technical advantage that the point clouds describing an area of ​​the environment always describe a virtually constant size. The periodic spacing also ensures that the overlapping areas of the point clouds also remain constant during recording. This allows for improved comparison or assignment of point cloud pairs from the different point cloud sequences.

[0010] In a further embodiment, the point cloud sequences are recorded by a vehicle and / or a robot during one or more journeys.

[0011] In a further embodiment, point correspondences and / or patch correspondences and / or line correspondences are used as corresponding elements.

[0012] This offers the advantage that the method can be performed independently of the corresponding elements. Either only point correspondences, only patch correspondences, or only line correspondences can be used. On the other hand, it is also conceivable that different correspondences can be combined. For example, point correspondences and patch correspondences can be combined.

[0013] In a further embodiment, the method is carried out with a point cloud registration, wherein the point cloud registration is carried out correspondence-based with an FCGF method.

[0014] This offers the technical advantage that the method can be implemented particularly robustly and efficiently. Thus, it also enables efficient processing of large volumes of point cloud sequences or point clouds. Furthermore, the embodiment offers high accuracy for point cloud alignment, even in complex environments.

[0015] In an advantageous embodiment, the calculation of the transformation for the point clouds of the point cloud sequences is based on a RANSAC algorithm.

[0016] This offers the technical advantage that the method can be used in a particularly versatile manner and with reduced effort. The algorithm is efficient because it only requires a limited number of iterations. This allows it to process extremely large amounts of data. The method can also be used efficiently for processing a large number of point cloud sequences.

[0017] In a further embodiment, each point cloud sequence has at least a third point cloud sequence, wherein the first and the third point clouds at least partially overlap in a second overlap region, wherein a third point cloud pair is determined. The third point cloud pair each has a point cloud from the first point cloud sequence and a point cloud from the second point cloud sequence, wherein the point clouds of the third point cloud pair represent an at least partially identical third region of the environment, wherein third corresponding elements are determined for the measurement points of the point clouds of the third point cloud pair, wherein second further corresponding elements are determined for the measurement points of the second overlap region of the first and the second point cloud sequence.The third corresponding elements and the second further corresponding elements are additionally taken into account for the calculation of the transformation for the point clouds of the first point cloud sequence and the second point cloud sequence.

[0018] This offers the technical advantage of further increasing the accuracy of the method. By using the third point cloud pair and the second additional corresponding elements, the accuracy of the method can be further increased.

[0019] In a further embodiment, a second transformation is determined for the point clouds of the first point cloud sequence and the second point cloud sequence based on the first additional corresponding elements and the second additional corresponding elements. The second transformation is used to assess the quality of the first transformation.

[0020] This provides the technical advantage of providing additional security for the calculation of the first transformation by comparing it with the calculation of the second transformation. This makes the method more secure and robust than conventional methods.

[0021] According to a second aspect of the invention, a device is proposed which is configured to carry out all steps of the method according to the first aspect.

[0022] According to a third aspect, a computer program is provided, comprising the instructions which, when the computer program is executed by a computer, for example by the device according to the second aspect, cause the computer to carry out a method according to the first aspect.

[0023] According to a fourth aspect, a machine-readable storage medium is provided on which the computer program according to the third aspect is stored.

[0024] The invention is explained in more detail below using exemplary figures and embodiments. Herein: Fig. 1 a flowchart of a method according to a first aspect; Fig. 2 a device according to a second aspect; Fig. 3 a machine-readable storage medium according to a fourth aspect; Fig. 4 a first point cloud pair and a possible problem with an alignment of point clouds; Fig. 5 a first and a second point cloud sequence for the Fig. 4 shown point cloud pair and a solution of the problem shown; Fig. 6 another first point cloud sequence and another second point cloud sequence; Fig. 7 an exemplary determination of corresponding elements for the point cloud sequences from Fig. 6; Fig. 8 shows an example of a selection of the corresponding elements for the first point cloud pair from Fig. 7.

[0025] Fig. 1 shows a flowchart of a method 100 according to a first aspect.

[0026] The method 100 is used to create a digital map for a vehicle or robot. The method 100 comprises the following steps: In a first step 110, at least a first point cloud sequence and a second point cloud sequence are used. Each of the point cloud sequences comprises at least a first point cloud and a second point cloud. The point clouds comprise measurement points with measurement data over an area of ​​an environment for a location. The first point cloud and the second point cloud overlap at least partially in a first overlap area. The point cloud sequences preferably originate from one or more measurement runs and are created by sensors.

[0027] In a second step 120, a first point cloud pair is determined, each comprising one point cloud from the first point cloud sequence and one point cloud from the second point cloud sequence. The point clouds of the first point cloud pair each describe an at least partially identical first area of ​​the environment for a location. First corresponding elements are determined for the measurement points of the point clouds of the first point cloud pair. Point correspondences can be used as corresponding elements.

[0028] In a third step 130, a second point cloud pair is determined, each comprising one point cloud from the first point cloud sequence and one point cloud from the second point cloud sequence. The point clouds of the second point cloud pair represent an at least partially identical second area of ​​the environment for a location. Second corresponding elements are determined for the measurement points of the point clouds of the second point cloud pair. The corresponding elements for the point cloud pairs can be determined, for example, using a method such as FCGF.

[0029] In a fourth step 140, first further corresponding elements for the measuring points of the first overlap area are determined from the point clouds of the first and the second point cloud sequence.

[0030] In a fifth step 150, a transformation for the point clouds of the first and second point cloud sequence is calculated on the basis of the first, the second and the first further corresponding elements.

[0031] In a sixth step 160, a digital map with measurement points and measured values ​​is created on the basis of the transformed point clouds of the first and second point cloud sequence using a plurality of location positions.

[0032] The locations are located on a track or road. The measurement points of the measurement data for the point clouds can be recorded, for example, using a sensor. A sensor as defined in this description is, for example, one of the following sensors: radar sensor, lidar sensor, image sensor, in particular an image sensor of a video camera, ultrasonic sensor, infrared sensor, or magnetic field sensor.

[0033] Thus, a point cloud can be generated or created, for example, by using lidar data and / or radar data and / or video camera data and / or infrared data and / or magnetic field data and / or ultrasound data.

[0034] The point cloud sequences can be recorded by a vehicle and / or a robot at different locations during one or more journeys. For this purpose, the vehicle and / or robot is equipped with one of the previously described sensors. Consequently, the point cloud sequences and their point clouds preferably contain measurement points and measurement data from the sensor measurements across a vehicle or robot environment.

[0035] The measurement points and measurement data can be, for example, road markings. It is conceivable that the measurement points and measurement data include, for example, longitudinal markings, area markings, boundary markings, transverse markings, or parking markings.

[0036] The measurement points of the measurement data for the point clouds acquired with the sensor are preferably recorded at periodic intervals and at constant intensity.

[0037] The digital map created by the method 100 described above can be used, for example, for the navigation of a vehicle or a robot. It is thus conceivable that the map created by this method 100 could be used for vehicle navigation in semi-autonomous or fully autonomous driving mode of a vehicle or robot.

[0038] Fig. 2 shows a device 1 according to a second aspect.

[0039] The device 1 is configured to carry out all steps of the previously described method 100, even in advantageous embodiments. The device 1 can, in particular, be a computer. To carry out the method 100, at least a first point cloud sequence 10 and a second point cloud sequence 11 are transmitted to the device 1. The point cloud sequences 10, 11 can be stored, for example, in a cloud or on a data storage device of the device 1. The point cloud sequences 10, 11 are recorded, for example, during one or more journeys of a vehicle 4 with a sensor 6. The digital map 5 created by the device 1 and the method 100 can then be transmitted to a vehicle 4. The transmissions can be wireless, for example, via WLAN. The vehicle 4 can then use the generated digital map 4, for example, to control a driving function in a partially or fully autonomous driving mode.

[0040] Fig. 3 shows a machine-readable storage medium 3 according to a fourth aspect.

[0041] A computer program 2 according to the third aspect is stored on the machine-readable storage medium 3. The computer program 2 comprises instructions which, when executed by a computer, cause the computer to carry out a method according to the independent claim. The machine-readable storage medium 3 can, for example, be read by the device 1 from Fig. 2 can be read.

[0042] Fig. 4 schematically shows a first point cloud pair 20 and a possible problem in an alignment of point clouds.

[0043] A first point cloud pair 20 is shown. The first point cloud pair 20 has a first point cloud 12 from a first trip 40. The first point cloud 12 from the first trip 40 has measurement points 15 over a first area 55. The measurement points 15 for the first area 55 can, for example, describe vehicle surroundings from the first trip 40. Thus, the first point cloud 12 from the first trip 40 has measurement points 15 which can, for example, be represented as lines and can represent road markings. In this example, the measurement points 15 are three road markings which, by way of example, run approximately parallel and are arranged at approximately the same distance from one another. Furthermore, the first point cloud pair 20 has a first point cloud 12 from a second trip 41, which describes the first area 55.The first point cloud 12 from the second trip 41 also has measurement points 15 that describe properties of the first area 55. During the second trip 41, two lane markings were recorded in the first point cloud 12. The two lane markings run parallel to each other, for example. During the second trip 41, the third lane marking from the first trip 40 was not recorded. This can be the case, for example, if the external environments during the second trip 41 differ from the first trip 40. For example, a vehicle may have obscured the third lane marking during the second trip 41. The measurement points 15 can, for example, have been recorded by a sensor, for example a lidar radar.

[0044] In order to convert the first two point clouds 12 of the first point cloud pair 20 into a suitable transformation, the first two point clouds 12 from the first run 40 and the second run 41 are compared with each other using correspondence-based alignments. Such correspondence-based alignments can, for example, be performed using point correspondences and / or patch correspondences and / or line correspondences.

[0045] Furthermore, a possible first alignment 45 and a possible second alignment 46 for the first point cloud pair 20 are shown as examples. Based on the recorded measurement points 15 from the first run 40 and the second run 41, both alignments 45 and 46 appear plausible. Thus, it cannot be ruled out that a possibly incorrect alignment 45 or 46 is included in the calculation of a transformation for the digital map.

[0046] This article explains a potential problem for aligning measurement points from point clouds from two different measurement runs. As already described, two point clouds can have different measurement points 15 despite describing the same first area 55. The following figures show how this problem can be resolved.

[0047] Fig. 5 shows a first and a second point cloud sequence 10, 11 for the Fig. 4 shown point cloud pair 20.

[0048] A first point cloud sequence 10 from a first journey 40 has a first point cloud 12 and a second point cloud 13. By way of example, the first point cloud sequence 10 has a third point cloud 14. The point clouds 12, 13, 14 from the first point cloud sequence 10 have measurement points 15. The measurement points 15 can, for example, have been recorded by a sensor during a first journey 40. The measurement points 15 can, for example, be lane markings, and in this case, three lane markings have been recorded as measurement points 15. The lane markings or the measurement points 15 can, by way of example, describe a vehicle environment. The first point cloud 12 and the second point cloud 13 have a first overlap area 50. The first overlap area 50 has, by way of example, information from the measurement points 15 or the lane markings of the first point cloud 12 and the second point cloud 13.This is achieved, for example, by recording the measurement points 15 during the first run 40 by a sensor at periodic intervals and preferably at a constant intensity. The sensor can, for example, be configured to record data every 10 meters for an area within a radius of 12 meters. The first point cloud 12 and the third point cloud 14 further comprise, by way of example, a second overlap area 51. The second overlap area 51 thus contains information about the measurement data from the first point cloud 12 and the third point cloud 14.

[0049] Furthermore, a second point cloud sequence 11 from a second trip 41 is shown. The second point cloud sequence 11 from the second trip 41 can again be recorded using a sensor at periodic intervals and preferably at constant intensity. The second point cloud sequence 11 has a first point cloud 12 and a second point cloud 13. By way of example, the second point cloud sequence 11 has a third point cloud 14. The point clouds 12, 13, and 14 of the second point cloud sequence 11 again contain measurement points 15. The measurement points 15 are again lane markings and can describe a vehicle's surroundings from the second trip 41.In this exemplary embodiment, a first lane marking was not recorded in the first point cloud 12 of the second point cloud sequence 11. One possible reason for this could be, for example, a parked car that prevented the first lane markings from being recorded in the first point cloud 12 during the second trip 41. The second lane and the third lane were also recorded continuously in the second point cloud sequence 11. The second point cloud sequence 11 also has a first overlap area 50 and a second overlap area 51.

[0050] The respective first point clouds 12 from the point cloud sequences 10, 11 describe an approximately identical first region 55. These two first point clouds 12 from the two point cloud sequences 10, 11 again form the first point cloud pair 20. The respective second point clouds 13 and third point clouds 14 from the first point cloud sequence 10 and the second point cloud sequence 11 describe an at least partially identical second region 56 and an at least partially identical third region 57. By using the two overlapping regions 50, 51, the previously identified problem for the alignment of the first point cloud pair 20 can now be resolved. Since the different point cloud sequences 10, 11 contain information about the preceding and subsequent point clouds of the first point cloud pair 20, the positioning of the measurement points 15 for the first point cloud pair 20 can be determined.The right image shows that the second alignment 46 with the measurement points 15 can place the lane markings in the correct position. This is achieved by a correspondence-based comparison of the first point clouds 12 and the use of corresponding elements from the two overlapping areas 50, 51.

[0051] A more detailed exemplary description of how to solve this alignment is given in the following Fig. 6 to 8 are explained in more detail.

[0052] Fig. 6 shows another first point cloud sequence 10 and another second point cloud sequence 11.

[0053] The first point cloud sequence 10 and the second point cloud sequence 11 can be used for the procedure from Fig. 1 can be used to create a digital map. The first and second point cloud sequences 10, 11 have a first point cloud 12 and a second point cloud 13. By way of example, they also have a third point cloud 14. The point clouds 12, 13, 14 from the first point cloud sequence 10 and the second point cloud sequence 11 have measurement points 15 with measurement data about an area of ​​an environment. The first point cloud sequence 10 and the second point cloud sequence 11 can, for example, have been recorded during a first trip 40 and during a second trip 41. By way of example, the point cloud sequences 10, 11 were recorded during two different trips 40 and 41 by the same sensor. The sensor recorded measurement points 15, which describe lane markings, by way of example. The point clouds 12, 13, 14 from the point cloud sequences 10, 11 were recorded with the sensor at periodic intervals with preferably constant intensity.The intensity and periodic intervals for recording the point cloud sequences 10, 11 were adjusted so that the point clouds overlap in a first overlap area 50 and a second overlap area 51. Thus, the first point clouds 12 overlap with the second point clouds 13 in a first overlap area 50, and the first point clouds 12 overlap with a third point cloud 14 in a second overlap area 51.

[0054] Fig. Figure 7 shows an exemplary determination of corresponding elements 30, 31, 32, 35, 36 for the point cloud sequences 10, 11 from Fig. 6.

[0055] From the two point cloud sequences 10, 11 determined during the first run 40 and the second run 41, point cloud pairs are determined for the first point cloud 12, the second point cloud 13, and the third point cloud 14. A first point cloud pair 20 comprises the first point cloud 12 from the first point cloud sequence 10 and the first point cloud 12 from the second point cloud sequence 11. The first two point clouds 12 describe an approximately identical first area 55.

[0056] A second point cloud pair 21 comprises a second point cloud 13 from the first point cloud sequence 10 and a second point cloud 13 from the second point cloud sequence 11. The respective second point clouds 13 describe an approximately identical second region 56 of the environment.

[0057] A third point cloud pair 22 comprises a third point cloud 14 from the first point cloud sequence 10 and a third point cloud 14 from the second point cloud sequence 11. The two third point clouds 14 describe an approximately identical third region 57.

[0058] For scan matching, also called point cloud registration, the point cloud pairs 20, 21, 22 from the point cloud sequences 10, 11 are matched based on correspondence. For this purpose, first corresponding elements 30 are determined for the first point cloud pair 20. The first corresponding elements 30 can be point correspondences, for example. It is also conceivable for the first corresponding elements 30 to be patch correspondences and / or line correspondences. Second corresponding elements 31 are determined in the second point cloud pair 21. The second corresponding elements 31 are also point correspondences, for example. Third corresponding elements 32 can also be determined for the third point cloud pair 22. The third corresponding elements 32 can also be point correspondences. Furthermore, first further corresponding elements 35 are determined in the second point cloud pair 21.Second additional corresponding elements 36 can also be determined from the third point cloud pair 22. The further corresponding elements 35, 36 can also be point correspondences.

[0059] The determination of the previously described corresponding elements 30, 31, 32, 35, 36 or the scan matching or the point cloud registration can be determined using a method such as FCGF.

[0060] The following is the Fig. The point cloud registration shown in Figure 7 using an FCGF method is briefly described again using general formulas.

[0061] For the point cloud registration of two point cloud sequences X = (X 1 , X 2 , ... ), Y = (Y 1 ,Y 2 , ...) are calculated for each point cloud pair X i , Y i first point correspondences Ci={(xki,yki)},xki∈Xi,yki∈Yi determined by a method such as FCGF. Furthermore, it is assumed that the overlapping areas within the point cloud sequences are known, and this can be done with xki∈Xi+1 for xki∈Xi and yki∈Yi+1 be determined.

[0062] Subsequently, the point correspondences C i After that, a set of Ai={(xki−1,yki−1)|xki−1∈Xi,yki−1∈Yi}∪{(xki+1,yki+1)|xki+1∈Xi,yki+1∈Yi} Then the set C i ∪ A i used.

[0063] The result of this calculation is calculated using Fig. 8 described.

[0064] Fig. 8 shows an example result for the corresponding elements of the first point cloud pair 20 from Fig. 7.

[0065] The first point clouds 12 from the first run 40 and the second run 41 are compared using the first corresponding elements 30, second corresponding elements 31, and third corresponding elements 32. The first point cloud 12 describes the first area 55. To calculate the transformation, only the second corresponding elements 31 are used, which were identified as the first additional corresponding elements 35 by the previously described calculation. This also applies to the third corresponding elements 32 and the resulting second additional corresponding elements 36. Using the first corresponding elements 30 and the first additional corresponding elements 35 and the second additional corresponding elements 36, an affine transformation for the first point cloud pair 20 can be calculated.

[0066] The transformation calculation is based, for example, on a RANSAC algorithm. The calculated transformation can then be used to create a digital map using the measurement data from the point cloud sequences.

[0067] This digital map can then be used for the navigation of a vehicle and / or a robot, even for semi-autonomous or fully autonomous operation.

[0068] To increase the safety requirements for the transformation calculation, a second transformation calculation can also be performed based solely on the first additional corresponding elements 35 and the second additional corresponding elements 36. The second transformation calculation can then be compared with the first transformation calculation. Depending on whether the transformation calculations differ from one another or are identical, the method can also be used to perform an additional safety assessment for the transformation of the point clouds from the point cloud sequences.

[0069] Although the invention has been described above with reference to specific embodiments, a person skilled in the art can also implement embodiments not disclosed or only partially disclosed without deviating from the essence of the invention.

Claims

[1] Method (100) for creating a digital map for controlling a vehicle or a robot, comprising the following steps: - Using (110) at least one first point cloud sequence (10) and one second point cloud sequence (11), wherein each point cloud sequence (10, 11) comprises at least one first point cloud (12) and one second point cloud (13), wherein a point cloud comprises measurement points (15) with measurement data over an area of ​​an environment of a location position, wherein the first point cloud (12) and the second point cloud (13) overlap at least partially in a first overlap area (50), - determining (120) a first point cloud pair (20) each comprising a point cloud from the first point cloud sequence (10) and a point cloud from the second point cloud sequence (11), wherein the point clouds of the first point cloud pair (20) each represent an at least partially identical first region (55) of the environment, wherein first corresponding elements (30) are determined for the measurement points (15) of the point clouds of the first point cloud pair (20), - determining (130) a second point cloud pair (21) with one point cloud from the first point cloud sequence (10) and one point cloud from the second point cloud sequence (11), wherein the point clouds of the second point cloud pair (21) represent an at least partially identical second region (56) of the environment, wherein second corresponding elements (31) are determined for the measuring points (15) of the point clouds of the second point cloud pair (21), - determining (140) first further corresponding elements (35) for the measuring points (15) of the first overlap area (50) from the point clouds of the first (10) and the second point cloud sequence (11), - calculating (150) a transformation for the point clouds of the first (10) and second point cloud sequence (11) on the basis of the first (30), the second (31) and the first further corresponding elements (35), - Creating (160) a digital map with measuring points and measured values ​​based on the transformed point clouds of the first (10) and second point cloud sequence (11) using several locations. [2] Method (100) according to claim 1, wherein the measuring points (15) of the measuring data for the point clouds are detected by a sensor, wherein the detection of the measuring points by the sensor takes place at periodic intervals and preferably at constant intensity. [3] Method (100) according to claim 1 or 2, wherein the point cloud sequences (10, 11) are recorded by a vehicle and / or a robot during one or more journeys. [4] Method (100) according to one of the preceding claims, wherein point correspondences and / or patch correspondences and / or line correspondences are used as corresponding elements. [5] Method (100) according to one of the preceding claims, wherein the method (100) is carried out with a point cloud registration, wherein the point cloud registration is carried out correspondence-based with an FCGF method. [6] Method (100) according to one of the preceding claims, wherein the calculation of the transformation for the point clouds of the point cloud sequences (10, 11) is based on a RANSAC algorithm. [7] Method (100) according to one of the preceding claims, wherein each point cloud sequence (10, 11) has at least a third point cloud (14), wherein the first (12) and the third point cloud (14) at least partially overlap in a second overlap region (51), wherein a third point cloud pair (22) is determined, wherein the third point cloud pair (22) each has a point cloud of the first (10) and the second point cloud sequence (11), wherein the point clouds of the third point cloud pair (22) represent an at least partially identical third region (57) of the environment, wherein third corresponding elements (32) are determined for the measurement points (15) of the point clouds of the third point cloud pair (22), wherein second further corresponding elements (36) are determined for the measurement points (15) of the second overlap region (51) of the first (10) and the second point cloud sequence (11),wherein the third (32) and the second further corresponding elements (36) are additionally taken into account for the calculation of the transformation for the point clouds of the first (10) and the second point cloud sequence (11)., [8] Method according to claim 7, wherein a second transformation for the point clouds of the first (10) and the second point cloud sequence (11) is determined on the basis of the first further (35) and the second further corresponding elements (36), wherein the second transformation is used for an assessment of the quality of the first transformation. [9] Device (1) which is arranged to carry out all steps of the method (100) according to one of the preceding claims. [10] Computer program (2) comprising the instructions which, when executed by a computer, cause the computer to carry out a method (100) according to one of claims 1 to 8. [11] Machine-readable storage medium (3) on which the computer program (2) according to claim 10 is stored.

Images