Method for the 3D measurement of an environment and scanning assembly

Abstract

A method for the 3D measurement of an environment, for example in an architectural building survey, and a scanning assembly having a 3D laser scanner, said scanning assembly being designed to carry out a method of this type, are disclosed. Proceeding from a performed scan (data capture A), the next best scanner view having the greatest potential both to expand, in the peripheral regions, the already-captured environment geometry and to complete the already-captured environment geometry in itself is proposed by means of spatial analysis (preparation B, analysis C), for example by ray tracing of the voxel cloud (ray tracing 1). The results from the preparation B are superposed and a scan capture is simulated. The position proposals must take into account the physical setup conditions of the scanner measurement setup, e.g. the accessibility, the setup height and a minimum and maximum distance from the previous view. The analysis C of the view candidates is carried out preferably again by means of ray tracing (ray tracing 2).

Description

[0001] Method for 3D measurement of an environment and scanning arrangement

[0002] Description

[0003] The invention relates to a method for measuring an environment and a scanning arrangement operated according to such a method.

[0004] The 3D measurement of objects using laser scanners is of considerable importance in practice. For complex or difficult-to-access objects, multiple laser scans are always taken consecutively from different locations and saved in a common project folder. The scans must then be converted into a common, higher-level coordinate system. This process is referred to as "registration." Applicant's patent EP 3 056 923 B1 describes a scanning system in which this registration is performed in the field using a handheld device, with another scan being performed using a laser scanner in parallel with the registration.

[0005] The challenge with such solutions is positioning the laser scanner (2D or 3D) in the field in such a way that the entire area to be surveyed is actually captured. When scanning a larger or more complex area, considerable experience is required to select suitable positions for the scanner. Nevertheless, it often happens that the entire area is not surveyed comprehensively, requiring extensive follow-up surveying to fill in any gaps.

[0006] WO 2022 / 074083 A1, also by the applicant, discloses a mobile scanning arrangement that can be operated in a SLAM mode. The scanning arrangement is mounted on a carrier vehicle, for example, a SKID or an AGV, or positioned on a type of backpack carried by a person moving in the environment to be measured. This mobile scanning arrangement has a computing unit that is connected to the scanning device, for example, a 2D or 3D scanner. The scanning arrangement is further configured with a location detection device, wherein the computing unit is designed to evaluate the data acquired by the scanning device, operated, for example, in an MMS-SLAM rotation mode, and the location detection device, and to determine a trajectory of the scanning device.The computing unit is also designed to determine at least one position for performing a static scan if the data quality of the trajectory or the resulting scan is insufficient. Such a mobile scanning arrangement is very flexible and can be used in environments where a carrier vehicle or a person can move freely. However, this concept reaches its limits in environments where free movement is not possible.

[0007] In such cases, the first scanning process with the recording of static scans from several viewpoints is preferred.

[0008] Mobile autonomous robots use systems in which a scanning device generates a map of the environment while the robot is moving, enabling the robot to orient itself within the environment. However, such concepts do not involve the precise measurement of the environment or a measurement object, but rather simply the measurement of the environment in which the autonomous robot is to be moved.

[0009] In contrast, the invention is based on the object of creating a method for measuring an environment and a scanning arrangement operable according to such a method, which enables the precise detection of even complex environments. This object is achieved with regard to the method by the combination of features of patent claim 1 and with regard to the scanning arrangement by the features of the independent patent claim 14.

[0010] Advantageous further developments of the invention are the subject of the subclaims.

[0011] The method according to the invention for 2D or 3D measurement of an environment is carried out according to the following steps, whereby the sequence of the method steps is variable.

[0012] In a first step, a scanning arrangement is positioned at a location in the environment to be measured in a conventional manner. By appropriately controlling a scanning device (hereinafter referred to as a scanner), the environment is scanned. Then, through appropriate analysis, as described, for example, in the aforementioned EP 3 056 923 B1, a spatial image of the visible surface of the environment is created. This image is articulated, for example, as a 3D point cloud and provides a scaled model of the environment.

[0013] This scan data is then added to any existing inventory data (which will be discussed later) and correctly aligned to a common reference system. The scan is then registered accordingly in this project-specific coordinate system.

[0014] In a further process step, the registered scan data are converted into a 3D database and a 3D data structure cloud is created from this database.

[0015] Up to this point, data acquisition is no different from that of conventional scanning systems. In a subsequent preparatory step, these 3D data structures are analyzed as described below, and candidates for possible scanner positions are generated in parallel. The 3D data structure cloud is classified into 3D data structures that have measurement coordinates assigned to them and those that do not. The latter 3D data structures thus represent hidden or invisible areas of the measured environment.

[0016] In a further process step, candidate viewpoints are determined from the 3D data structures without measurement coordinates according to predetermined boundary conditions and scan images of the determined candidate viewpoints are then simulated, preferably by ray tracing.

[0017] The results of this simulation are then analyzed and a new concrete viewpoint is selected according to predefined criteria.

[0018] The scanner is then moved to this new static position, and another scan is initiated. This scan is then evaluated in the manner described above, with the process being repeated until the environment is completely captured and all 3D data structures are filled with measurement coordinates.

[0019] The inventive concept thus enables the measurement of an environment, for example during a building survey, without prior knowledge of this environment. Using spatial analysis, the best next scanner position (best next view) is suggested, which has the greatest potential to both expand the already recorded environmental geometry in the peripheral areas and complete it in itself. As explained in more detail below, these position suggestions must take into account the physical setup conditions of the scanner measurement setup, e.g., accessibility, setup height, and minimum distance. In a preferred embodiment of the invention, the 3D data structure cloud is implemented as a voxel cloud.

[0020] The simulation is carried out, as already indicated above, by ray tracing, i.e. by virtually scanning the voxel CIoud with radial rays from a fictitious nodal point of the original scan.

[0021] In the above-described classification of the 3D data structure cloud (VoxelCloud), the data structures (voxels) without measurement coordinates are differentiated into parts that, with reference to a nodal point, lie behind and / or in front of a 3D data structure (voxel) filled with measurement coordinates.

[0022] The computational effort can be further reduced if ground coordinates are extracted from the 3D data structure cloud (VoxelCloud) before the station coordinates are determined, whereby these extracted ground coordinates can be further reduced according to certain boundary conditions relating to the ground or the location.

[0023] Such constraints require, for example, that discontinuities on the ground are avoided when selecting the new station. Furthermore, to improve measurement accuracy, a minimum and maximum distance between the station and a previous station is specified, and a minimum clearance above a ground point is required to account for the height of the scanning device.

[0024] When analyzing the viewpoint candidates for selecting the new viewpoint from the aforementioned extracted ground coordinates, criteria such as the number of 3D data structures, in particular voxels, without measurement coordinates in the field of view of the viewpoint candidate or the point density or point resolution in the field of view assigned when creating the 3D data structure (voxels) can preferably be taken into account.

[0025] It is particularly preferred if the number of 3D data structures, in particular the voxels, without measurement coordinates in the field of view behind a 3D data structure (voxel) filled with measurement coordinates are taken into account during the analysis.

[0026] In this analysis, hole edges and hole edge areas of the 3D data structures, especially the voxels, are also taken into account to improve measurement accuracy.

[0027] As explained above, the viewpoint candidates can be analyzed using ray tracing according to the criteria mentioned. The viewpoint candidates extracted in this way then serve as the basis for selecting a new viewpoint.

[0028] In the scanning arrangement according to the invention, it is provided that the data processing for determining the new position in the field is carried out, for example, by means of a handheld / tablet that is in data connection with the scanning device or a computer integrated into the scanning device.

[0029] Preferred embodiments of the invention are explained in more detail below with reference to schematic drawings. They show:

[0030] Figure 1 shows a basic structure of an embodiment of a scanning arrangement according to the invention;

[0031] Figure 2 shows a workflow of an embodiment of the inventive

[0032] Method for 3D measurement of an environment and Figure 3 shows a detailed representation of the workflow according to Figure 2.

[0033] Figure 1 shows the basic structure of a scanning arrangement 1 according to the invention. This has a scanning device, for example a 3D laser scanner 2, as offered by the patent applicant under the trademark Z+F-Imager®. This laser scanner 2 has a rotating measuring head 4, which rotates about a horizontal axis (view according to Figure 1 ) and via which a laser beam is directed onto the environment to be measured. A transmitting and receiving unit of the laser scanner 2 is arranged in a housing 6, which can be pivoted at least 180° about a vertical pivot axis 8, so that the environment can be scanned almost completely by pivoting the housing 6 about the pivot axis 8 while the measuring head 4 rotates. The laser scanner 2 is positioned in the environment via a tripod 10 or the like - this area of ​​the environment is referred to below as the viewpoint 12.

[0034] The housing 6 can contain a memory for the scan data and an evaluation unit, via which the acquired scan data can be evaluated. This evaluation, in particular the registration / registration, can also be carried out using a tablet 14 or handheld device shown in Figure 1, which is preferably in contactless data connection with the laser scanner 2. This data connection can be established, for example, via Bluetooth or a connection regulated according to the WiFi standard. To determine the scanner position and the scanner orientation, the laser scanner 2 can be designed with an integrated navigation system, which, for example, enables a GNSS-independent determination of the absolute position of the laser scanner 2 or at least a relative position to a known location in the environment (in the field).

[0035] This tablet 14 is used for a preferably targetless registration of the scan, which is facilitated by the fact that the position of the laser scanner 2 and its orientation, either relative to a previously known location or as an absolute position, are known. After the complete survey of the surroundings described below, a field book containing the viewpoints 12 and the scanner orientation 16 is stored in the tablet 14. The results can then be transmitted, as indicated in Figure 1, to a central server 18 after a complete survey, for example, via a WLAN connection.

[0036] Figure 2 shows the basic process steps of the inventive method for determining the "next best view." The inventive workflow can be divided into the main steps of data acquisition A, preparation B, and analysis C. Data acquisition A is generally carried out according to the usual methodology. In the illustrated embodiment, the scanning arrangement 1 explained with reference to Figure 1 is used with a laser scanner 2—hereinafter referred to as the scanner. In principle, however, other scanning arrangements for 2D or 3D measurement can also be used.

[0037] As indicated in the workflow, the scanner 2 according to the invention is brought into the environment by the person performing the measurement process and positioned at a preselected location 12 via the tripod 10. The scanner 2 is then controlled in a conventional manner, so that a first recording is created with the capture of a spatial image of the visible surface of the environment articulated as a point cloud.

[0038] The resulting scan, or more precisely the scan data representing a scaled model of the captured environment, is then registered in the next process step, whereby this scan data is possibly added to inventory data captured in a previous step and is correctly aligned to a common, higher-level reference Z-coordinate system.

[0039] Up to this point, the method according to the invention corresponds to the conventional scanning process. After this registration, the scan data is converted into a database, preparing the data structure for a holistic spatial analysis, thus enabling subsequent spatial point retrieval more efficiently.

[0040] In the next step, a 3D data structure cloud, in this case a voxel cloud, is created from this database. This creates a cube structure with defined edge lengths, whose components (voxels) contain numerical properties, such as point density. Accordingly, this step reduces the data to an overall point cloud with resolution information.

[0041] The voxel cluster resulting from the data reduction is then further processed according to the invention in two parallel process steps, with properties being assigned to the components (voxels) as described below. In parallel, candidates for possible scanner positions are generated by selection based on predefined boundary conditions.

[0042] As shown in preparation B on the left, the method according to the invention initially performs an initial ray tracing, whereby the voxels of the VoxelColour are classified with attributes. The voxels are virtually scanned and classified using radial rays from the nodal point of the original scan. According to the invention, this classification is achieved by distinguishing between at least three groups: a) "seen" - these are voxels filled with measurement coordinates, which therefore do not require any additions; b) "empty" - this refers to voxels without measurement coordinates that are located upstream of a "seen" voxel on the virtual scanning beam; c) "hidden" - this grouping represents voxels without measurement coordinates, but which are located directly downstream of a "seen" voxel on the virtual scanning beam. The voxels classified in this way are then further processed in the manner explained in more detail below.

[0043] In parallel with the classification of voxels into "seen," "empty," and "occluded," the floor surfaces are extracted, as shown in Preparation B on the right in Figure 2. This results from the requirement that a scanner viewpoint is usually physically only possible on one floor surface, so this is determined using a search algorithm and marked as a voxel group (containing the viewpoint candidates) in the voxel cloud.

[0044] The extracted ground surfaces are then reduced to voxel groups representing the candidate viewpoints using boundary conditions. Such boundary conditions can vary depending on the scanning task. For example, it can be specified that the voxel groups must not contain any discontinuities on the ground, thus avoiding viewpoints located in a stepped area or other unevenness. Furthermore, voxel groups are considered that lie within an area defined by a minimum and a maximum distance with reference to completed viewpoints (from previously performed scans). Furthermore, a certain minimum free space above the respective ground point may be required so that the installation height and a minimum measurement distance of scanner 2 are taken into account.

[0045] As already mentioned, these boundary conditions can be extended depending on the scanning task.

[0046] The viewpoint candidates generated in this way are then arranged in a regular grid and analyzed in the sequence labeled Analysis C. The results from Preparation B are overlaid, and a scan is again simulated to determine the "next best viewpoint." The viewpoint candidates determined from Preparation B are again analyzed using ray tracing (Ray Tracing 2), as shown in Figure 2, according to, for example, two criteria. These criteria are, of course, variable and expandable. In the specific embodiment, the number of visible holes is taken into account. These are voxels from Preparation B that are classified as "hidden" and within the field of view of the viewpoint candidates. These voxels assigned to the "visible holes" preferably contain information about both the hole edges and the hole edge regions.Another criterion is the point density - this considers the point density assigned to the voxels when the VoxelCioud was created.

[0047] This analysis is performed for each candidate viewpoint and saved. The results of the simulation described above (ray tracing 2) are then compared and listed in a weighted order, with the viewpoint with the highest score being the new viewpoint sought (next-best view).

[0048] Scanner 2 is then moved to the new position determined in this way, and a new scan is started. This process is repeated iteratively until the environment is completely captured and only voxels associated with measurement coordinates remain.

[0049] Figure 3 shows the previously described ray tracing 1 of preparation B in a more detailed representation.

[0050] As explained above, in ray tracing 1 performed in preparation step B, the components of the voxel cluster are first classified according to the "occluded" attribute. This means that voxels without measurement coordinates that are located directly downstream of a "seen" voxel on the ray tracing scanning beam ("holes") are classified. After this sub-step, the "normal" voxels ("seen") and the "occluded" voxels are extracted. In part 2 of ray tracing 1 according to preparation B, the voxels classified as "empty," i.e., the "empty" voxels seen, are then captured and saved.

[0051] This is followed by the second ray tracing step (Raytracing 2) in Analysis C, which was explained in detail using Figure 2.

[0052] As indicated in Figure 3, the voxels determined by ray tracing 1 can also be taken into account when extracting the ground surfaces.

[0053] Disclosed are a method for 3D measuring an environment and a scanning arrangement designed to carry out such a method, wherein, based on a scan performed, the next best scanner viewpoint (next best view) is proposed by means of spatial analysis.

[0054] List of reference symbols:

[0055] 1 Scanning arrangement

[0056] 2 scanners

[0057] 4 measuring head

[0058] 6 housings

[0059] 8 swivel axis

[0060] 10 Tripod

[0061] 12 Point of view

[0062] 14 Tablet / Handheld

[0063] 16 Scanner orientation

[0064] 18 servers

[0065] A Data recording

[0066] B Preparation

[0067] C Analysis

Claims

Patent claims 1 . Procedure for measuring an environment with the following steps: - Positioning a scanning array at a location in the environment; - Scanning the environment using the scanning arrangement and creating a spatial image of the visible surfaces of the environment; - Registration / alignment of the captured scan data to a common reference system; - Converting the scan data into a 3D database and creating a 3D Data structure cloud; - Classification of the 3D data structure cloud into 3D data structures filled with and without measurement coordinates; - Determining new viewpoint candidates from the 3D data structures without Measurement coordinates according to given boundary conditions; - Simulating scans of the identified candidate viewpoints; - Analysis of the simulation and selection of a new viewpoint; - Move the scanner to the new location and - Repeat the above steps until the environment is completely covered.

2. The method according to claim 1, wherein the 3D data structure cloud is a VoxelCloud.

3. Method according to claim 1 or 2, wherein the simulation is carried out by ray tracing, in particular a virtual scanning of the 3D data structure cloud, in particular the voxel cloud, with radial rays from a nodal point of the original scan.

4. The method according to claim 3, wherein the 3D data structures without measurement coordinates contain portions that lie behind or in front of a 3D data structure filled with measurement coordinates with respect to a nodal point.

5. Method according to one of the preceding claims, wherein ground coordinates are extracted from the 3D data structure cloud, in particular the VoxelCloud, before the position candidates are determined.

6. Method according to claim 5, wherein the extracted ground coordinates are reduced according to boundary conditions relating to the ground and the location.

7. Method according to claim 6, wherein the boundary conditions include: - Avoiding discontinuities on the ground; - Maintaining a minimum and maximum distance between the position and previous positions; - Maintain a minimum space above a ground point.

8. Method according to one of the preceding claims, wherein the analysis of the location data for the selection of the new viewpoint is carried out from the extracted ground coordinates, in particular according to one of the following criteria: - Number of 3D data structures, in particular voxels, without measurement coordinates in the field of view of the viewpoint candidate; - in the creation of the 3D data structure cloud, especially the VoxelCloud, the 3D data structure, in particular the voxel, assigned point density or Point resolution in the field of view.

9. Method according to claim 8, wherein the number of 3D data structures, in particular voxels, without measurement coordinates in the field of view behind a 3D data structure filled with measurement coordinates is taken into account in the analysis.

10. Method according to claim 9, wherein hole edges and hole edge regions of the 3D data structures, in particular the voxels, are also taken into account in the analysis.

11. Method according to one of claims 8 to 10, wherein the viewpoint candidates are analyzed by means of ray tracing.

12. Method according to one of claims 8 to 11, wherein the viewpoint candidates resulting from the analysis are listed and a selection of the new viewpoint is made from this list.

13. Method according to one of the preceding claims, wherein the data processing for determining the new position in the environment is carried out by means of a tablet / handheld (14) having a data connection with the scanning arrangement or a computer integrated into the scanning arrangement.

14. Scanning arrangement, in particular 3D scanning device, which is designed to be operated according to a method according to one of the preceding claims 1 to 13.

Images