Method and arrangement for measuring the progress of a construction

Description

[0001] Contemporary buildings are increasingly being created using digital building models, particularly so-called BIM models (Building Information Modeling). Such BIM models typically contain a detailed 3D model of the building along with semantic information describing the individual structural elements, such as walls, ceilings, floors, windows, doors, stairs, or pillars, and their relationships to one another.

[0002] The representation of a building in a digital building model usually reflects the planned state of the building. The building should then be constructed in such a way that its structural elements are as congruent as possible with their representation in the building model. In reality, however, deviations of varying magnitude often occur, which should be identified and monitored whenever possible. Furthermore, it is generally desirable to track the current construction progress of the building during the construction phase and compare it with the plans.

[0003] To record deviations from the plan or the construction progress, 3D scanners, such as lidar or laser scanners, are often used. These scanners are used to scan the structure regularly, often several times a day. A particular problem, however, is that 3D scanners generate large quantities of discrete spatial sample points, so-called point clouds of objects, which, although roughly consistent over repeated scans, are very difficult to compare or reproduce in detail. This is due, among other things, to the fact that the sample points from different scans are usually gridded in different ways. In addition, differences in lighting conditions lead to more or less noticeable differences in color and brightness between different scans. This makes scanner-based detailed comparisons of objects scanned at different times, and thus the accurate recording and documentation of construction progress, considerably more difficult.

[0004] The publication "Point-to-point Comparison Method for Automated Scan-vs-BIM Deviation Detection" by Jingdao Chen and Yong K. Cho in the Proceedings of the 17th International Conference on Computing in Civil and Building Engineering, Tampere, Finland, 2018, describes a method for comparing point clouds with existing BIM models. However, detailed recording and documentation of construction progress typically requires considerable effort.

[0005] Stefanie Zollmann et al., "Augmented Reality for Construction Site Monitoring and Documentation," Proceedings of the IEEE, IEEE New York, US, Vol. 102, No. 2, February 1, 2014, pages 137–154, ISSN: 0018-9219, DOI: 10.1109 / JPROC.2013.2294314, disclose the monitoring of construction progress via uniformly recorded 3D reconstruction using aerial photographs and BIM data. Specific BIM structures can be identified semi-automatically. Visualization is achieved using a fragment shader.

[0006] Alexander Braun et al., "Automated Progress Monitoring Based on Photogrammetric Point Clouds and Precedence Relationship Graphs," 35th International Symposium on Automation and Robotics in Construction (ISARC 2018), June 18, 2015, pages 1–7, ISSN: 2413-5844, DOI: 10.22260 / ISARC2015 / 0034, https: / / www.pf.bgu.tum.de / pub / 2015 / braun_co_stilla_isarc15_pap.pdf, discloses image-based monitoring of construction progress under multiple occlusions based on a precedence relationship graph. One variant uses GPS-based aerial photographs.

[0007] It is an object of the present invention to provide a method and an arrangement by means of which construction progress can be determined and documented more efficiently.

[0008] This object is achieved by a method having the features of patent claim 1, by an arrangement having the features of patent claim 12, by a computer program product having the features of patent claim 13 and by a computer-readable storage medium having the features of patent claim 14.

[0009] To measure the construction progress of a building, a digital building model of the structure is imported. Such a structure can be, in particular, a public or private building, a residential building, an office building, a factory building, a power plant, a tunnel, a bridge, a road, a shaft, or any other structural construction. Construction progress can also refer to progress in the reconstruction or demolition of the building. Using the building model, the spatial arrangement of a building element surface within the building is determined for each building element. Furthermore, the building is scanned using a 3D scanner, generating a large number of spatial scanning points. At least some of the scanning points are then spatially assigned to corresponding building element surfaces.According to the invention, a building element-specific depth map is generated for each building element area based on the scanning points specifically assigned to that building element area. To determine construction progress, the generated depth maps are then compared with previously generated depth maps.

[0010] To carry out the method according to the invention, an arrangement for measuring construction progress in a building, a computer program product and a computer-readable, preferably non-volatile storage medium are provided.

[0011] The method according to the invention, the arrangement according to the invention and the computer program product according to the invention can be carried out in particular by means of one or more computers, one or more processors, application-specific integrated circuits (ASICs), digital signal processors (DSPs), a cloud infrastructure and / or so-called "field programmable gate arrays" (FPGAs).

[0012] A key advantage of the invention is that, thanks to the generation and comparison of depth maps specific to building elements, a complex and often inaccurate comparison of point clouds is no longer necessary. Furthermore, the amount of data to be documented can generally be significantly reduced – not least due to the often improved compressibility and comparability of depth maps. Furthermore, many process steps can be performed particularly efficiently using conventional graphics processors.

[0013] Advantageous embodiments and further developments of the invention are specified in the dependent claims.

[0014] According to an advantageous development of the invention, a deviation of a depth map currently generated for a building element surface from a depth map previously generated for this building element surface can be determined. Depending on the determined deviation, construction progress information specific to this building element surface can then be output. In particular, a difference between the depth maps can be determined as a deviation, e.g. in the form of a difference image. A large deviation or difference can indicate significant construction progress. In particular, it can be provided to compare each deviation with a predetermined threshold value and, if the threshold value is exceeded, to output corresponding construction progress information. The previously generated depth map can, in particular, also be generated from the building model.In this case, construction progress can be determined relative to a planned building state.

[0015] Advantageously, a depth map currently generated for a building element surface can be saved in the form of a deviation from a depth map previously generated for this building element surface. The deviation can be determined as described above. Since such deviations often have significantly smaller values ​​than the involved depth maps, memory requirements can be significantly reduced in many cases. In particular, for small deviations, difference images of the depth maps can be saved instead of the current depth maps, and for larger deviations, the current depth maps themselves can be saved. The deviations can be determined particularly efficiently using a fragment shader, in particular by a graphics processor.

[0016] According to an advantageous embodiment of the invention, the generated depth maps can be versioned individually for each building element surface based on the determined deviations. This allows for particularly efficient versioning, since little or no data needs to be stored for building element surfaces that do not change or change only slightly.

[0017] According to an advantageous implementation variant of the invention, one or more building element surfaces can be determined based on discontinuities in the course of a surface normal of a building element and / or based on discontinuities in the course of a material property, e.g., a color, a reflectivity, and / or a translucency of a building element. In many cases, a course of surface normals and / or material properties can be taken from the digital building model. In this way, different building element surfaces can be delimited from one another and / or a building element can be divided into different building element surfaces.

[0018] According to a further advantageous implementation variant of the invention, the digital building model can comprise a building graph in which building elements are stored as nodes and relationships between building elements as edges. This allows one or more building element surfaces to be determined for a respective building element and stored as additional nodes assigned to the respective building element. Furthermore, a depth map generated for a respective building element surface can be stored in association with the node of this building element surface. An association of a building element surface to an associated building element and / or an association of a depth map to an associated building element surface can each be represented by an edge of the building graph. In particular, building element surfaces and / or depth maps can be added to the digital building model as a further hierarchy level.In this way, the determined building element areas and / or depth maps can be managed in a particularly efficient manner.

[0019] According to a further advantageous embodiment of the invention, the concealment and / or visibility of a respective building element surface can be checked. The assignment of scanning points to a respective building element surface can then take place depending on its visibility and / or concealment. In particular, a respective building element surface can be discarded, not saved and / or marked if it is detected as concealed or invisible. Furthermore, building elements concealed by cladding, internal building elements or building elements located in other rooms, or their building element surfaces, can be marked as concealed or invisible. If necessary, a respective building element surface can be divided into concealed or visible parts. For example, a wall bordering several rooms can be divided into the wall surfaces visible in each individual room.The check for occlusion or visibility can be done, for example, using the IFCRelSpaceBoundary attribute of a BIM model.

[0020] Furthermore, the position of the 3D scanner relative to the structure can be read. The check for occlusion and / or visibility of a particular building element surface can then be performed based on the read position. In particular, if the position is known, a visibility cone for the 3D scanner can be determined, bounded by known objects in the space. Using such a visibility cone, the scanning points can be used to efficiently check whether or where additional, unknown objects are located between the 3D scanner and a particular building element.

[0021] Furthermore, depending on the occlusion and / or visibility of a respective building element surface and on a check of whether the measured sampling points spatially correspond to the respective building element surface, the respective building element surface can be marked as missing. A building element surface can be marked as missing in particular if it is identified as visible or not occluded and no sampling points are assigned to it.

[0022] According to a further advantageous embodiment of the invention, the depth maps can be generated using a vertex shader, in particular a graphics processor. Such vertex shaders, especially those of graphics processors, are generally specifically designed for such calculations and allow for particularly efficient generation of depth maps.

[0023] An embodiment of the invention is explained in more detail below with reference to the drawings, each of which illustrates schematically: Figure 1: an extension of a building model to include building element surfaces, Figure 2: an arrangement according to the invention when measuring construction progress, Figure 3: the arrangement according to the invention during a second measurement and Figure 4: a comparison of sampling points of different measurements.

[0024] Figure 1 illustrates an extension of a digital building model BM of a building B by building element surfaces F1, ..., FN to an extended building model BME. Building B can be a public or private building, a residential building, an office building, a factory building, a power plant, a tunnel, a bridge, a road, or another structural structure.

[0025] The digital building model BM specifies a planned state of building B and can preferably be a so-called BIM model (BIM: Building Information Modeling) or another CAD model of building B. The building model BM describes a multitude of building elements of building B, such as walls, ceilings, floors, windows, doors, stairs, and / or other specific construction elements, in machine-readable form using a multitude of data sets specifying a respective building element. The building model BM also specifies a multitude of relationships between building elements using corresponding data sets. For example, a respective relationship specification can be used to assign a door or window to the wall into which the respective door or window is to be installed according to the plan.

[0026] In the present exemplary embodiment, the building model BM is stored in a database DB and, in particular, comprises a building graph. In this building graph, building elements of building B are represented by nodes, and relationships between the building elements are represented by edges.

[0027] In the present exemplary embodiment, the building graph comprises the following building elements: building B itself, several floors S1 and S2 of building B, as well as a wall W and a pillar PF on floor S1. Floors S1 and S2 are assigned to building B via edges represented by arrows, just as wall W and pillar PF are assigned to floor S1. In addition to building elements B, S1, S2, PF, and W, the planned building B comprises many other building elements, which are also stored in the building graph but are not shown for reasons of clarity.

[0028] The building model BM is retrieved from the database DB by a computer PC and expanded into an extended building model BME. The computer PC has one or more processors (CPU) for executing method steps of the inventive method, as well as one or more memories (MEM) for storing data to be processed. Furthermore, the computer PC includes a graphics card (GPU) with one or more graphics processors. The graphics card (GPU) includes a plurality of vertex shaders and fragment shaders for executing method steps of the inventive method.

[0029] To extend the building model BM, the PC determines a spatial arrangement of surfaces or surface elements of each building element of the building model BM. The determined surfaces or surface elements, along with their spatial arrangement, are assigned to the respective building element in the building graph as its building element surfaces via edges.

[0030] Figure 1 illustrates, by way of example, how visible or exposed surfaces or surface elements of the pillar PF are determined using the structural model BM and assigned to it as structural element surfaces F1, ..., FN in the structural graph. By adding the structural element surfaces F1, ..., FN and, if necessary, other structural element surfaces not shown, the digital structural model BM is expanded into the extended digital structural model BME.

[0031] Criteria for determining surfaces or surface elements of building elements can be discontinuities in the surface normal of a given building element or discontinuities in a material property, such as brightness or color. For building elements with simple geometries, such as cuboid or cylindrical building elements, a relatively simple decomposition into building element surfaces can be carried out using the BM building model. For more complex curved surfaces, a lower level of detail can be used alternatively or additionally, or a minimal, cuboid-shaped bounding box can be determined around the building element.

[0032] By assigning building element surfaces to a respective building element, data about the spatial arrangement of visible building element surfaces can be managed naturally in the extended building model BME. Building element surfaces are preferably defined as a further hierarchy level in the extended building model BME. If necessary, individual building element surfaces can be further subdivided in the extended building model BME, for example, to specify partial visibility or partial concealment.

[0033] Figure 2shows a schematic representation of an arrangement according to the invention with a computer PC and a 3D scanner S3D coupled thereto for measuring construction progress on a building B. The construction progress can also relate to progress in demolition, conversion, or gutting. Where identical or corresponding reference symbols are used in the figures, these reference symbols designate identical or corresponding entities that can be configured, or implemented as described in the respective figure.

[0034] In Figure 2 Structural elements BE1, BE2, BE3, and BE4 of structure B are shown as examples and symbols. For the structural elements BE1, ..., BE4, their structural element areas F1, ..., F4 were determined as described above and assigned to the respective structural element BE1, BE2, BE3, or BE4 in the extended structural model BME. The structural element areas F1, ..., F4 are shown in Figure 2highlighted by thickened lines. The extended building model BME is stored on the PC.

[0035] To measure the construction progress, the structure B or its visible part is scanned by the 3D scanner S3D, for example a lidar or a laser scanner, whereby a plurality of spatial sampling points P are generated. The sampling points P are in Figure 2 illustrated by dotted lines.

[0036] The sample points P are transmitted from the 3D scanner S3D to the computer PC. The computer PC assigns the sample points P to spatially corresponding building element surfaces, here F1, ..., F4. For this purpose, a check is carried out for each sample point P to determine which of the building element surfaces, here F1, ..., F4, it spatially corresponds to. Alternatively or additionally, for each building element surface F1, ..., F4, it can be checked to determine which sample points P spatially correspond to it.

[0037] A spatial correspondence can be determined, in particular, by the spatial coordinates of a respective scanning point P and a respective building element surface F1, ..., F4 matching within a specified tolerance range, possibly after a suitable shift and / or rotation of a reference system. To support such a check or correlation, images from the 3D scanner S3D can be registered using optical markers. Furthermore, a spatial position of the 3D scanner S3D can be recorded and taken into account during the check or correlation.

[0038] The assignment of sample points P to building element surfaces F1, ..., F4 can be repeated until either all or almost all sample points P have been assigned to a building element surface, or until sample points P remain that cannot be clearly assigned. The latter sample points P can be discarded for further processing, or an attempt could be made to detect, measure, or identify objects not intended for the building model BM using the non-assignable sample points P.

[0039] Based on the sampling points P that can be assigned to a respective building element surface F1, ... or F4, a building element-specific depth map is generated according to the invention for the relevant building element surface F1, ... or F4. The texture carries color information of the relevant sampling points, while the depth map carries depth information of these sampling points. Such a depth map is often also referred to as a displacement map. Depth maps are conventionally used by graphics processors when rendering 3D objects. The depth maps are projected as images onto the 3D objects to be rendered in order to ensure the most realistic representation of the 3D objects possible. Since many graphics processors are specifically designed for such tasks, the generation of the depth maps can be carried out particularly efficiently by the graphics processor GPU of the PC.

[0040] The generation of depth maps is described below using the Figure 3 and 4 described in more detail. Figure 3 shows the already in Figure 2 The arrangement shown in a second measurement is shown schematically. Compared to the Figure 2 In the state of structure B shown, a protruding structure VA, such as wall cladding, was installed on the structural element surface F3 during construction. The protruding structure VA partially conceals the structural element surface F3.

[0041] The protruding structure VA is scanned by the 3D scanner S3D, generating the scanning points P'. Due to the protruding structure VA, the scanning points P' are offset in the direction of the 3D scanner S3D, unlike the other scanning points P assigned to the building element surface F3. Due to the spatial proximity of the scanning points P' to the building element surface F3, these scanning points P' are nevertheless assigned to this building element surface F3 as spatially corresponding.

[0042] The generation of a depth map for a respective building element area is explained in more detail below using the example of the building element area F3.

[0043] First, a so-called vertex shader of the GPU is used to calculate the distances of all sample points P and P' assigned to the building element surface F3 from the building element surface F3 and store them in association with this building element surface F3. Sample points that do not belong to the building element surface F3 can be removed via a clipping stage. For this purpose, a tolerance range can be provided to define the depth of a clipping level.

[0044] Based on the distances between points that remain within the tolerance range, even flat objects on the building element surface F3 can be represented, even if no separate building element exists in the building model BM, such as sockets, switches, or cable ducts. To ensure that building objects can be represented with sufficient resolution, a corresponding resolution can be specified in advance for each depth map; for example, the size of the building element surface divided by twice the resolution of the 3D scanner S3D.

[0045] Using a rendering pipeline of the GPU graphics card, the sample points P and P' assigned to the building element surface F3 are interpolated onto this building element surface F3. This applies to both depth information and color information of the sample points P and P'. The depth information is stored as a depth map or displacement map, while the color information is saved as a texture. Each building element surface F1, ..., F4 can thus be assigned two images, namely a texture and a depth map with the same resolution. These images can be stored efficiently, especially if there is a uniform distribution of the sample points P and P'. In the extreme case, if all sample points P and P' for a building element surface are the same distance from this building element surface, the corresponding depth map can be encoded with a single value.

[0046] Figure 4illustrates a comparison of sampling points of different measurements using the example of the building element area F3.

[0047] The upper part of Figure 4 illustrates the context of Figure 2 described first measurement, while the lower part of Figure 4 in connection with Figure 3 The second measurement described illustrates this.

[0048] During the first measurement, the building element surface F3 was not yet partially covered by the protruding structure VA, so that the 3D scanner S3D could sense the building element surface F3 and generate a set of scanning points P that was almost adjacent to the building element surface F3. Figure 4 The distance shown between the sampling points P and the building element surface F3 is greatly exaggerated for reasons of clarity.

[0049] Based on the building element surface F3 and the sample points P, a first texture TX1 and a first depth map DM1 are generated for the building element surface F3. If the sample points P have a nearly constant distance to the building element surface F3 within the measurement accuracy, the depth map DM1 can be encoded or compressed very efficiently. The same applies to the texture TX1 if the building element surface F3 has a uniform or nearly uniform color.

[0050] During the second measurement, as already mentioned above, the building element surface F3 is partially obscured by the protruding structure VA. As a result, the sample points P' originating from the protruding structure VA are offset forward. Based on the building element surface F3 and the sample points P and P', a second texture TX2 and a second depth map DM2 are generated, as described above. Due to a uniformly greater distance of the sample points P' from the building element surface F3, a higher distance or a lower depth is encoded in a lower, hatched part of the second depth map DM2 than in the upper, non-hatched part.

[0051] Initial versions of a respective depth map, here DM1, or texture, here TX1, can be saved in a version management system as a base version in a compressed format for each building element area. A respective initial version can also be created for different building element areas at different times. This means that building objects that will only be created in later construction phases may still be missing.

[0052] Later versions of a respective depth map, here DM2, or texture, here TX2, can advantageously be stored relative to a respective earlier version. In the present case, the first textures, here TX1, and first depth maps, here DM1, are stored as such. The second textures, here TX2, and depth maps, here DM2, are advantageously not stored as such, but only in the form of a difference to the first textures TX1 and depth maps DM1. In the present case, a difference image ΔTX is formed between the respective second texture TX2 and the respective first texture TX1, as well as a difference image ΔDM between the respective second depth map DM2 and the respective first depth map DM1.In many cases, the textures and depth maps of different measurements differ only slightly—especially for building element surfaces that have not been altered—and can therefore be very efficiently encoded or compressed in the form of difference images, here ΔTX and ΔDM. Furthermore, construction progress information can be derived particularly easily from a difference image.

[0053] In particular, textures and depth maps, or their difference images, can be saved instead of the sample points, which generally results in significant storage space savings. The data is stored in association with the relevant building element surface.

[0054] When assessing construction progress, it is often necessary to determine whether a building element surface without assigned scanning points actually does not exist or no longer exists, or whether it was simply not recorded due to occlusion, for example. In such cases, the graphics pipeline of the GPU graphics card can be used to determine which surfaces can be viewed by the S3D 3D scanner based on a recorded position of the S3D 3D scanner, for example, through so-called z-culling.

[0055] In particular, with a known recording position and for known objects, a visibility cone limited by the latter can be calculated. Based on the sampling points and the visibility cone, it is then possible to efficiently check whether and where other unknown objects are located between the S3D 3D scanner and a given building element surface. If no sampling points can be assigned to building elements that were assessed as visible in this way, it can often be assumed that these building elements are not yet or no longer present in the current life cycle of the building.

Claims

1. Computer-implemented method for measuring progress in the construction of a building (B), wherein a) a digital building model (BM) of the building (B) is read in, b) a spatial arrangement of a building element surface (F1,...,F4) in the building (B) is determined on the basis of the building model (BM) for a respective building element (BE1,...,BE4), c) the building (B) is scanned by means of a 3D scanner (S3D), wherein a multiplicity of spatial scanning points (P, P') are generated, d) at least some of the scanning points (P, P') are assigned to spatially corresponding building element surfaces (F1,...,F4), characterized in that e) a building element surface-specific depth map (DM2) is generated for a respective building element surface (F1,...,F4) on the basis of the scanning points (P, P') specifically assigned to this building element surface, and f) the generated depth maps (DM2) are compared with previously generated depth maps (DM1) in order to determine construction progress.

2. Method according to Claim 1, characterized in that a deviation (ΔTX, ΔDM) of a depth map (DM2) currently generated for a building element surface from a depth map (DM1) previously generated for this building element surface is determined, and in that, depending on the deviation (ΔTX, ΔDM) determined, construction progress information specific to this building element surface is output.

3. Method according to either of the preceding claims, characterized in that a depth map (DM2) currently generated for a building element surface is stored in the form of a deviation (ΔTX, ΔDM) from a depth map (DM1) previously generated for this building element surface.

4. Method according to Claim 3, characterized in that the deviation (ΔTX, ΔDM) is determined by means of a fragment shader, in particular by a graphics processing unit (GPU).

5. Method according to Claim 3 or 4, characterized in that the generated depth maps (DM1, DM2) are versioned individually for a respective building element surface (F1,...,F4) on the basis on the deviations (ΔTX, ΔDM) determined.

6. Method according to any of the preceding claims, characterized in that one or more building element surfaces (F1,...,F4) are determined on the basis of discontinuities in the profile of a surface normal of a building element (BE1,...,BE4) and / or on the basis of discontinuities in the profile of a material property of a building element (BE1,...,BE4).

7. Method according to any of the preceding claims, characterized in that the digital building model (BM) comprises a building graph in which building elements are stored as nodes (B, S1, S2, PF, W) and relationships between building elements are stored as edges, in that one or more building element surfaces are determined for a respective building element and are stored as additional nodes (F1,...,FN) assigned to the respective building element, and in that a depth map generated for a respective building element surface is stored in a manner assigned to the node of this building element surface.

8. Method according to any of the preceding claims, characterized in that occlusion and / or visibility of a respective building element surface (F1,...,F4) are / is checked, and in that the assignment of scanning points (P, P') to a respective building element surface (F1,...,F4) is effected depending on the visibility and / or occlusion thereof.

9. Method according to Claim 8, characterized in that a position of the 3D scanner (S3D) relative to the building (B) is read in, and in that the checking for occlusion and / or visibility of a respective building element surface (F1,...,F4) is effected depending on the position read in.

10. Method according to Claim 8 or 9, characterized in that depending on occlusion and / or visibility of a respective building element surface (F1,...,F4) and depending on a check as to whether measured scanning points (P, P') spatially correspond to the respective building element surface (F1,...,F4), the respective building element surface (F1,...,F4) is marked as missing.

11. Method according to any of the preceding claims, characterized in that the generation of the depth maps (DM1, DM2) is effected by means of a vertex shader, in particular by a graphics processing unit (GPU).

12. Arrangement for measuring progress in the construction of a building (B), configured for carrying out a method according to any of the preceding claims.

13. Computer program product, configured for carrying out a method according to any of Claims 1 to 11.

14. Computer-readable storage medium having a computer program product according to Claim 13.