Fast feature recognition and mesh generation in structural design.

Abstract

The method identifies structural features of a structure by obtaining a global point cloud representation of the structure and obtaining target point clouds representing target structural features, where the target point cloud is a subset of the global point cloud representation of the structure. The global structural information and the target point cloud are provided to a global registration process, where the global registration process generates a globally registered representation of the structure, where the global structural information is derived from the global point cloud representation of the structure. The globally registered representation of the structure and the target point cloud are provided to a local registration process, where the local registration process generates one or more matching point clouds, where each of the one or more matching point clouds is a subset of the global point cloud representation of the structure. Mesh generation rules are then applied to each matching point cloud to generate a corresponding volume for the matching point cloud, where the corresponding volume has a mesh spacing, size, and orientation appropriate for the location of the respective matching point cloud.

Description

[Technical Field]

[0001] Copyright A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of this patent document or patent disclosure, as it appears in the Patent and Trademark Office file or patent records, but otherwise reserves any and all copyrights whatsoever.

[0002] Cross-Reference to Related Applications This application claims priority from U.S. Provisional Patent Application No. 63 / 399,351, filed August 19, 2022, which is incorporated herein by reference in its entirety.

[0003] The present invention relates to computer engineering, and more particularly to techniques for recognizing features within computer-designed structures. [Background technology]

[0004] Computer engineering analysis processes typically apply algorithms to a digital representation of a product. For example, in computational fluid dynamics, a mesh is generated that represents the shape of the product and forms the basis for the flow characteristics around the product. As is known in the art, a mesh is a discrete representation of a subsection of a continuous geometric space.

[0005] The mesh regions of a design are generally not uniform in size, but are smaller and of higher density in areas where features of interest with higher complexity exist, such as, but not limited to, flow characteristics in aircraft fuselage design.

[0006] A problem with traditional computer design techniques arises from the fact that the specification and identification of meshes in a design remains largely a human endeavor. In some situations, a human must labor over the design to identify all of the locations where features with particular properties reside. Furthermore, some techniques that achieve some level of automation require significant time and effort for a human to complete a user-driven means of characterizing features to be placed in the structure.

[0007] Another problem with conventional computer design techniques is their relatively slow execution speed. The inventors of the subject matter described herein have determined through research and inventiveness that a contributing factor to this problem is related to conventional methods of encoding structural designs (e.g., tessellated data, vector representations provided as output from computer-aided design programs, etc.). [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] Rusu et al., “Fast point Feature Histograms (FPFH) for 3D Registration”, 2009 IEEE International Conference on Robotics and Automation, 2009, pp. 3212-3217, doi: 10.1109 / ROBOT.2009.5152473 [Non-patent document 2] Zhou et al., “Fast Global Registration”, 9906, 10.1007 / 978-3-319-46475-6_47 (2016) [Non-patent document 3] Gelfand et al., “Robust Global Registration”, SGP05: Eurographics Symposium on Geometry Proceedings, The Eurographics Association pp. 197-206 (2005) Summary of the Invention [Problem to be solved by the invention]

[0009] In view of the above, there exists a need in the art to address the above and / or related problems. [Means for solving the problem]

[0010] The words "comprise" and "comprising", when used herein, are to be construed as specifying the presence of stated features, values, steps or components; however, it should be emphasized that the use of these terms does not exclude the presence or addition of one or more other features, values, steps, components or groups thereof.

[0011] Additionally, in some instances (e.g., in the claims and abstract) reference letters may be provided to facilitate identifying various steps and / or elements, however, the use of reference letters is not intended to constrain or suggest that the so-referenced steps and / or elements should be performed or operated in any particular order.

[0012] According to one aspect of the present invention, these and other objects are achieved in a technique (e.g., a method, an apparatus, a non-transitory computer-readable storage medium, a program means) for identifying structural features of a structure. In some, but not necessarily all, preferred embodiments consistent with the present invention, identifying the structural features includes obtaining a global point cloud representation of the structure and obtaining a target point cloud representing structural features of a target, the target point cloud being a subset of the global point cloud representation of the structure. The global structural information and the target point cloud are provided to a global registration process, which generates a globally aligned representation of the structure, the global structural information being derived from the global point cloud representation of the structure. The globally aligned representation of the structure and the target point cloud are provided to a local registration process, which generates one or more matching point clouds, each of the one or more matching point clouds being a subset of the global point cloud representation of the structure.

[0013] In another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, the global structural information is a global point cloud representation of the structure.

[0014] In still other aspects of some, but not necessarily all, embodiments consistent with the present invention, identifying structural features includes generating a set of one or more extracted structural features based on a global point cloud representation of the structure, each of the one or more extracted structural features being a pose-invariant characterization of a local geometry around a point in the global point cloud representation of the structure, and the global structural information being the set of one or more extracted structural features. In other aspects of some, but not necessarily all, of these embodiments, generating a set of one or more extracted structural features based on the global point cloud representation of the structure includes determining a point feature histogram (PFH) based on the global point cloud representation of the structure. In one possible alternative, generating a set of one or more extracted structural features based on the global point cloud representation of the structure includes determining a fast point feature histogram (FPFH) based on the global point cloud representation of the structure.

[0015] In other aspects of some, but not necessarily all, preferred embodiments consistent with the present invention, identifying structural features includes downsampling the global point cloud representation of the structure to generate a downsampled global point cloud representation of the structure, and generating the set of one or more extracted structural features based on the global point cloud representation of the structure includes generating the set of one or more extracted structural features from the downsampled global point cloud representation of the structure.

[0016] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, generating one or more matching point clouds includes identifying a subset of the locally aligned representation of the structure as one of the one or more matching point clouds when a comparison of the subset of the locally aligned representation of the structure with the target structural feature produces a predetermined comparison result.

[0017] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, the predetermined comparison result is a predetermined root mean square error between a subset of the locally aligned representations of the structure and the target structural features.

[0018] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, the global alignment process is a Random Sample Consensus (RANSAC) process.

[0019] In another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, the local registration process includes determining an Iterative Closest Point (ICP) value.

[0020] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, obtaining a global point cloud representation of the structure includes: Obtaining point cloud data by converting a CAD (Computer Aided Design) geometric representation of the structure; Obtaining point cloud data by transforming a discretized representation of the geometry of the structure; Obtaining point cloud data by converting a surface mesh representation of the structure; Obtaining point cloud data by converting a volume mesh representation of that structure; obtaining point cloud data by transforming sensor data collected during flight testing of the structure; Obtaining point cloud data by transforming sensor data collected during physical testing of the structure Contains one or more of the following:

[0021] In yet another aspect of some, but not necessarily all, embodiments consistent with this invention, identifying the structural feature includes obtaining rules describing a volume (three-dimensional data set) associated with the target structural feature and obtaining contextual information about the location of the target structural feature. For each of the one or more matching point clouds, a corresponding volume is generated, the corresponding volume having a mesh spacing, a size, and a pose, the mesh spacing being generated by the rules; and each of the size and pose being generated based on the contextual information about the location of the target structural feature and the contextual information about the location of each of the one or more matching point clouds.

[0022] In another aspect of some, but not necessarily all, embodiments consistent with the present invention, identifying structural features includes obtaining one or more additional target point clouds representing structural features of the target, the one or more additional target point clouds having sizes different from one another and different from the target point cloud; providing the global structural information and each of the one or more additional target point clouds for each of the one or more additional target point clouds to a global registration process, which generates one or more additional globally aligned representations of the structure therefrom; providing the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds for each of the one or more additional target point clouds to a local registration process, which generates one or more additional sets of one or more matching point clouds therefrom.

[0023] In another aspect of some, but not necessarily all, embodiments consistent with the present invention, a computer program product is configured to perform any one or more of the aspects described herein.

[0024] In yet another aspect of some, but not necessarily all, embodiments consistent with the present invention, a non-transitory computer-readable storage medium includes program instructions that, when executed by one or more processors, perform any one or more of the aspects described herein.

[0025] In yet another aspect of some, but not necessarily all, embodiments consistent with the present invention, a system comprises one or more processors configured to perform any one or more of the aspects described herein.

[0026] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, there is provided a structural feature recognition apparatus for use in computer engineering, the structural feature recognition apparatus configured to perform any one or more of the aspects described herein.

[0027] In yet another aspect of some, but not necessarily all, preferred embodiments consistent with the present invention, a computer engineering system is provided that includes a structural feature recognition device for use in computer engineering, the structural feature recognition device configured to perform any one or more of the aspects described herein.

[0028] The objects and advantages of the present invention will be understood from the following detailed description read in conjunction with the drawings. [Brief explanation of the drawings]

[0029] [Figure 1] 1 is a flowchart of the operation of a structural feature identification apparatus consistent with the present invention. [Figure 2] 1 is a flowchart of the operations associated with the step / element of obtaining a point cloud representation of a structure. [Figure 3] 1 is a flowchart of some operations of a structural feature recognition apparatus according to some embodiments consistent with the present invention. [Figure 4]1 is a flowchart of some operations of a structural feature recognition apparatus according to some embodiments consistent with the present invention. [Figure 5] 1 is a point cloud representation of an exemplary aircraft. [Figure 6] FIG. 6 shows a computer-generated image of the nose section of the aircraft of FIG. 5. [Figure 7] 7A, 7B, 7C, 7D, and 7E are diagrams illustrating aspects of feature identification and mesh generation according to embodiments of the present invention. [Figure 8] FIG. 10 illustrates a range of volumes subject to new rules according to some embodiments of the present invention. [Figure 9] FIG. 1 illustrates structural features of a target and nine other vortex sources that are automatically identified by a preferred embodiment of the present invention. [Figure 10] FIG. 10 illustrates a method for automatically copying the original rule and creating copy spaces at other matching locations. [Figure 11] FIG. 1 illustrates a suitable structural feature recognition device configured to perform any and / or all of the operations described and illustrated herein. DETAILED DESCRIPTION OF THE INVENTION

[0030] Detailed Description Various features of the present invention will now be described with reference to the drawings, in which like parts are identified with the same reference characters.

[0031] Various aspects of the invention will now be described in more detail with reference to a number of preferred embodiments. To facilitate an understanding of the invention, many aspects of the invention will be described in terms of sequences of operations to be performed by a computer system or other hardware element capable of executing programmed instructions. It will be recognized that in each embodiment, various operations can be performed by specialized circuitry (e.g., analog and / or discrete logic gates interconnected to perform specialized functions), by one or more processors programmed with an appropriate set of instructions, or by a combination of both. References to "circuitry configured to" perform one or more described operations will be used herein to refer to any such embodiment (i.e., one or more specialized circuits alone, one or more programmed processors, or any combination thereof). Furthermore, the invention may additionally be considered entirely embodied in any form of non-transitory computer-readable carrier, such as a semiconductor memory, magnetic disk, or optical disk, that contains a suitable set of computer instructions that cause a processor to perform the techniques described herein. Accordingly, various aspects of the invention can be embodied in many different forms, and all such forms are contemplated as being within the scope of the invention. For each of the various aspects of the present invention, any of these implementations described above may be referred to herein as "logic circuitry configured to" perform the described operations, or alternatively as "logic circuitry" that performs the described operations.

[0032] One aspect of embodiments of the present invention is the use of point cloud representations of structural designs. Unlike traditional uses of, for example, named elements and geometric typologies, the use of point clouds is more flexible and facilitates the adaptation of recent advances in machine vision technology to a new purpose: solving feature recognition in structural design.

[0033] Another aspect of embodiments of the present invention is feature recognition / identification using geometric similarity recognition and alignment. Unlike prior art, embodiments consistent with the present invention can detect similarities that are not constrained by size or type because the underlying data structure is simply an arbitrary collection of points in 3D (Three Dimensional) space (i.e., a point cloud).

[0034] Yet another aspect of some, but not necessarily all, embodiments of the present invention is the modification of mesh generation rules based on contextual information about the location, and the application of the modified rules to automatically identified structural features.

[0035] These and other aspects will now be described in further detail in the following description.

[0036] Registration algorithms are currently known in the art and are widely used in image processing applications. The goal of registration is generally to transform a collection of separate data sets so that all of these data sets are related to the same coordinate system. This is useful, for example, when it is desirable to overlay separately acquired images (e.g., any combination of photographs, sensor data, computer-generated images, etc.) to produce a single image. One of many possible representative uses of registration can be found in augmented reality technology, where computer-generated graphic images are overlaid on one or more real-world images. Augmented reality technology allows a user to experience computer-generated images as existing in intended locations in the real world.

[0037] Through research and inventiveness, the inventors have recognized that alignment algorithms can be used outside the context of image processing, more specifically for the purpose of locating structural features that match target features of interest within a global structure. However, fast global alignment of sub-scale features within much larger shapes is a difficult challenge, especially when using point cloud representations. Automated machine vision tends to assume similar point counts in the source and target point clouds, and alignment is preceded by a segmentation step (dividing the target into multiple distinct parts, one of which is aligned to the source - e.g., a car moving through a machine vision system based on LIDAR (Laser Intensity Direction Ans Ranging)).

[0038] The inventors of the subject matter described herein recognized that registration functions can be applied to identify the location of features within a structure. However, unlike applications such as machine vision, feature matching poses the unique challenge of matching portions of a continuous target set to a source point set, where there may not be a straightforward way to partition the data set a priori.

[0039] The above-mentioned challenges are addressed in various embodiments of the present invention, such as that shown in Figure 1, which is in one respect a flowchart of the operation of a structural feature identification apparatus consistent with the present invention. In other respects, the blocks shown in Figure 1 may be thought of as representing means 100 (e.g., hardwired or programmable circuitry, or other processing means) for performing the described operations.

[0040] In one aspect of an embodiment consistent with the present invention, the process of recognizing features within an engineering model or structure and applying associated rules to further computational steps (computational processing) is uniquely based on points, rather than any higher-order resolution (such as CAD or images). This is because a point cloud is a comprehensive data representation, allowing information about volumes (rather than surfaces) and measured data values ​​(rather than computational calculations) to be input into a single process without requiring specific data / file formats and without requiring a type of representation that is not suited to the data being represented—for example, the surface tufting method on a Boeing 737 flight test. Thus, one step is to obtain a point cloud representation of the structure (step 101).

[0041] Some, but not necessarily all, embodiments further include converting any other data representations to a point cloud before global alignment to the additional rule library. This differs from geometry-based methods in which rules relate to named and tagged geometric shapes (surfaces) before discretization. This makes the present process ideally suited for computational engineering data processing based on the shape of the product, regardless of representation. Thus, obtaining a point cloud representation of the structure may need to include an additional conversion step. This is illustrated in FIG. 2, which is, in one respect, a flowchart of the operations associated with step / element 101.

[0042] As shown in Figure 2, a workspace can contain data sources from one or more of the following: i. Point Cloud 201 ii. CAD Geometry 203 iii Shape discretization 205 iv. Computational meshing of surfaces and / or volumes 207 v. Flight test sensor data and data obtained through physical experiments209 vi. Sensor data collected during physical testing of the structure (e.g., full-scale or mock-up); vii. Expert knowledge 211 encoded in tables, text or images, or input via a software GUI (Graphical User Interface).

[0043] Valuable information related to computer engineering can be obtained from physical experiments using real-world structures related to computer-designed structures. In such cases, data from the physical experiments can come from any one or more of the following: electronic sensors, transcribed direct measurements, annotated photographs, and other forms of data collection. Examples of electronic sensors include, but are not limited to, electro-optical sensors, pressure gauges, thermal sensors, infrared sensors, accelerometers, acoustic sensors (microphones), and strain gauge sensors.

[0044] In each case, a data point is generated that has the value of interest. Conversion to point cloud format is straightforward because the data points and values ​​do not require additional structure.

[0045] Unless the data originates as a point cloud 201, the remaining formats are first converted to a point cloud representation (step 213) before all the data is collected as a single data set 215.

[0046] 1, a next optional step is to downsample the point cloud dataset 215. This reduces the amount of data that needs to be processed. For ease of explanation, references to a "point cloud dataset" used herein will refer to the original dataset of step 101 if optional downsampling 103 is omitted in a given embodiment, or to the downsampled point cloud dataset if optional downsampling 103 is instead performed.

[0047] The next optional step is feature point extraction (step 105), in which the point cloud dataset is processed to generate data representing one or more extracted features that characterize the local geometry around one or more points in the point cloud dataset, where the extracted features are preferably pose-invariant and have good discriminatory power. Algorithms for performing suitable feature point extraction, such as point feature histograms (PFH) and fast point feature histograms (FPFH), are known in the art, as exemplified by Rusu et al., “Fast point Feature Histograms (FPFH) for 3D Registration,” 2009 IEEE International Conference on Robotics and Automation, 2009, pp. 3212-3217, doi: 10.1109 / ROBOT.2009.5152473. Therefore, a complete description of such algorithms is beyond the scope of this disclosure.

[0048] Next, in step 107, a target point cloud is obtained. The target point cloud is a subset of the global point cloud representation and represents structural features of interest (e.g., to the designer of the structure). The target point cloud can be provided by a user of the inventive technique, or it can be provided by artificial intelligence (AI) that has been trained to automatically identify types of features of interest.

[0049] Because one goal of the design process is to generate appropriate meshes for each structural feature of interest, and these meshes (e.g., in the example presented herein, the meshes defining the vortices generated by the vortex shedding sources) are not represented in the original global point cloud, another step is to obtain rules (step 109) that describe the meshes to be generated for the target point cloud, taking into account contextual information about the location of the target point cloud. Note that the mesh for the target point cloud can be, but need not be, at the location of the target point cloud. To take one non-limiting example, if the target point cloud represents a vortex shedding source, the region requiring mesh refinement is downstream of the vortex shedding source itself. Similar to the data representing that structure, the rules can be initially expressed in any data source.

[0050] Rules for a computational engineering process can be linked to specific features in a point cloud representation of a product and associated data sources, but the rules themselves do not need to be in the point cloud representation. One example is mesh generation for aerodynamics, where parts of an aircraft fuselage require specific processing—usually input by an expert or feedback from downstream processes. After matching point clouds corresponding to additional instances of a target structural feature of interest are identified at different locations on the aircraft fuselage (or, more generally, any type of structure being designed), computational engineering rules are applied at the feature locations to generate an appropriate mesh associated with the identified feature through a transformation in space via an alignment transformation.

[0051] As previously mentioned, registration is a process by which separate data sets having different coordinate systems can be aligned to a common coordinate system, such as in image processing applications. The inventors have recognized that finding registration points can also be useful in computer engineering environments to indicate when a subset of points in a point cloud match points of known features, and for this reason, the next step involves performing a global registration, which seeks to find a match between each target point cloud and the extracted features (or, more generally, with the global point cloud, in embodiments that do not include feature extraction).

[0052] To accomplish this quickly, it is advantageous to first perform a registration of the target point cloud with the global point cloud (step 111), and then fine-tune the initial feature registration by performing a local registration (using a different algorithm) between the target point cloud and the globally registered point cloud. The output of the local registration is a set of one or more matching points corresponding to the target point cloud's structural features of interest. Thus, in a preferred embodiment, the initial registration procedure can be, for example, a random sample consensus (RANSAC) process or similar. Global registration processes are generally known and are not relevant to the embodiments of the present invention described herein (e.g., for aligning images of three-dimensional (3D) shapes to create a single 3D image from separate images). See, for example, Zhou et al., "Fast Global Registration," 9906, 10.1007 / 978-3-319-46475-6_47 (2016) (Non-Patent Document 2).

[0053] The local registration process can be, for example, an iterative closest point (ICP) process or similar. The use of ICP outside the context of the embodiments of the present invention described herein (e.g., for registering images of 3D shapes for the purpose of creating a single 3D image from separate images) is generally known in the art. See, for example, Gelfand et al., "Robust Global Registration," SGP05: Eurographics Symposium on Geometry Proceedings, The Eurographics Association, pp. 197-206 (2005).

[0054] The advantage of performing the global and local registrations in sequence is that the global registration of the target point cloud with the global point cloud aligns the point cloud data so that the points are concentrated around potential features of interest within the structure. This concentration has the effect of generating a subset of points for the local registration to consider, thereby allowing the local registration process to work satisfactorily. Otherwise, the local registration alone would be ineffective.

[0055] To ensure accuracy in this preferred embodiment, a registration goodness-of-fit measure is generated for each matching point identified by the registration process. This can be done, for example, by generating a goodness-of-fit estimate that represents the distance between the target point cloud and each of the automatically identified matching points (step 115). The goodness-of-fit measure can be, for example, a root mean square error (RMSE) value.

[0056] The goodness-of-fit measure of the matching points is compared to a threshold (representing a minimum acceptable level of goodness-of-fit), and if the goodness-of-fit of the identified matching points satisfies the comparison (the "Yes" path out of decision block 117), context information describing the identified matching points and their associated locations is stored or otherwise made available to the designer (step 119).

[0057] Once matching points for the target points within the global point cloud are identified, corresponding meshes are generated for these matching points. Doing so requires that the rules for mesh generation take into account the context of these matching points with respect to the context of where the target points are located (e.g., the pose and surrounding elements of each given point cloud), since the newly generated mesh should fit within the context of the matching point cloud locations. Rules for generating computer-engineered meshes at the identified matching point cloud locations are created by adapting / transforming the rules associated with the target point clouds based on the relationship between the target point cloud locations and the matching point cloud locations (step 121). The transformed rules are then applied to generate a respective mesh for each matching point cloud (step 121).

[0058] A preferred embodiment including the location-dependent rule transformation of step 121 is illustrated in Figure 3, which is in one respect a flowchart of some operations of a structural feature recognition apparatus consistent with the present invention. In other respects, the blocks shown in Figure 3 may be considered to represent means 300 (e.g., hardwired or programmable circuitry, or other processing means) for performing the described operations. Operations involved in this aspect include: - Input of a target point cloud by a user or a downstream process (step 301). - Input of location capture (step 303), which serves as a reference context to compare with the context (location) of the matching point cloud. - Add the locations and rules to the library (step 305). -Compare the library locations with the global model (step 307), which is further explained in relation to FIG. - Transforming rules for matching locations based on a comparison of the library locations with the matching point cloud locations (step 309). - Apply the transformed rules to generate a mesh on the identified matching points (step 311).

[0059] Referring again to decision block 117 in FIG. 1, if the goodness of fit of the identified matching points does not satisfy the comparison with the goodness of fit threshold level (the "No" path out of decision block 117), then the identified matching points are not considered further.

[0060] Some, but not necessarily all, additional aspects of the present invention include multi-scale registration, which is advantageous in use cases where the details of the mesh generation rules to be applied to identified features are not identical at different locations. For example, the same vortex generator shape may require a longer downstream volume of mesh refinement when placed at the leading edge of a wing than when placed near the windshield of an aircraft. Considering this aspect in more detail, it can be seen that using deep learning algorithms for multi-scale registration (matching point clouds and locations) requires automatically generating recognizable patterns of points at multiple length scales determined from a global model. For example, when viewing the surface of a sphere from a very small distance from the surface, the sphere appears flat; when viewing the entire sphere, we see a collection of points equidistant from the center point; and when moving very far away, we see only a single point. For an aircraft, if we consider a vortex source near the nose of the aircraft, we look at a vortex source on a flat surface; or on a cylinder; or just in front of a flat surface (the aircraft's window); or near the front of a long, cylindrical fuselage. All depend on the distance of the observation point relative to the structure (or equivalently, on the size of the point cloud under consideration). For the same vortex source near the rear tail of the aircraft, we look at a vortex source on a flat surface; or on a cylinder; or a large vertically rising surface (the vertical stabilizer) and a large horizontally extending surface (the horizontal stabilizer); or, more precisely, near the rear of a large cylinder, where the hydrodynamic boundary layer is significantly thicker than at the front of the aircraft. Again, what we look at depends on the size of the point cloud under consideration.

[0061] In view of the above, in some, but not necessarily all, aspects of embodiments of the present invention, it is often advantageous in a computational engineering design process to automatically represent a global point cloud at multiple (N) useful length scales with associated library entries as multi-scale segmentations, with each scale acting as a modifier to the base rule library. To illustrate this aspect, reference is made to FIG. 4, which in one respect is a flowchart of the operation of a computational engineering design tool consistent with the present invention. In other respects, the blocks shown in FIG. 4 may be considered to represent means 400 (e.g., hardwired or programmable circuitry, or other processing means) for performing the described operations. Operations included in this aspect include: - Multi-scale definition of target point clouds for a given structural feature (generate target point clouds of N different sizes / scales for a given structural feature) (step 401). The selection of different scales can be input by a user or can be generated automatically by deep learning techniques. - Perform a fast registration (e.g., a global registration followed by a local registration as illustrated by steps 111 and 113 in Figure 1) for each of the N different target point clouds (step 403). Some matching points may be identified in more than one of the N runs of fast registration, others may only be present once, it all depends on what the features are and how similar or dissimilar the locations of the features involved are at different scales. - Multi-scale rule adaptation to generate rules appropriate for each of the N different sized point clouds (step 405). -Deep Learning (Step 407).

[0062] Additional aspects of embodiments consistent with the present invention will now be described with reference to a non-limiting example. Figure 5 is a global point cloud representation of an aircraft 500 (in this case, a Boeing 737). In this example, it is desired to identify ten independent, counter-rotating pairs of vortex generators at the front of the aircraft 500, of which only one pair will be mentioned. The pairs of vortex generators thus function as structural features as described above.

[0063] To identify vortex sources in this design, the first step is to obtain a good quality digital geometric description (such as a CAD) that represents with sufficient accuracy the shape of the object being considered. In this case, the geometry is a CAD representation of aircraft 500.

[0064] To represent the shape as a point cloud, a base mesh is generated by sampling the geometry into an octree (or other suitable discretization). For this particular aircraft, the global point cloud 500 contains approximately one million surface points.

[0065] Figure 6 shows a computer-generated image 600 of the global point cloud 500 of the aircraft. It can be seen that the point cloud is already concentrated around small features or high curvature.

[0066] 7A, 7B, 7C, 7D, and 7E illustrate aspects of feature identification and mesh generation. Referring first to FIG. 7A (for simplicity, the fuselage is sliced ​​at the centerline running from the front to the rear of the aircraft, and only one slice is shown in the figure), the basic rules (feature, size, surface curvature) do not fully dictate the required mesh spacing. To address this issue, one of the vortex generators 701 (the target structural feature) is identified (e.g., by an expert user) in the original geometry. The expert user then adds a set of rules 703 for generating a mesh with the proper spacing given the location of the target feature. Using "sourcing rules," the expert adds more information before the final mesh is generated.

[0067] Figure 7A shows a vortex source 701 and rules 703 for generating a mesh in context (i.e., as the vortex source is located in the nose portion 600 of the aircraft's global point cloud 500). For clarity, Figure 7B is an isolated, zoomed-in illustration of one of the vortex sources 701; Figure 7C is an isolated illustration of the rules 703 for generating a mesh; Figure 7D is an isolated illustration of the rules 703 for generating a mesh, shown in position relative to one of the vortex sources 701; and Figure 7E illustrates a mesh 705 generated from the rules when applied in context at the location of a target vortex source.

[0068] FIG. 8 shows the extent of the volume 801 to which the new rules are imposed when placed within an architectural context.

[0069] Referring now to FIG. 9, in one aspect of an embodiment of the present invention, this technique eliminates the need for an expert to repeat the rule specification process for each of the 10 pairs of vortex sources. Instead, embodiments of the present invention start with a target point cloud 901 (corresponding to a target structural feature of interest) and, given a point cloud representation of the feature, apply the fast feature recognition described above to find the same shape in the remainder of the volume. In particular, FIG. 9 illustrates this aspect, showing the target point cloud 901 (in this example, a vortex source represented by approximately 300 points) and nine other automatically identified point clouds 903 that also correspond to vortex sources.

[0070] The rules exist as a set of points, and these points are used to define the mesh spacing. Once a target set of point cloud matches is found for the original vortex source, the mesh generation rules are migrated to the new location. Figure 10 illustrates this aspect, showing the original volume 1001 generated from the original rule and how the original rule is automatically transformed (i.e., scaled and reoriented) and applied to the matching point cloud 903 to generate clone (duplicate, copy) volumes 1003 at the other nine matching locations.

[0071] In some, but not all, other aspects consistent with the present invention, specified rules for each structural feature (in this example, vortex shedding sources) are used to determine the level of mesh refinement (spacing) in the vicinity. If initial rules are specified a priori, downstream use of the resulting mesh may confirm or negate the need for such spacing. For example, the spacing may be overly conservative (fine). In such cases, gradients in the resulting flow solution are captured by the less fine (i.e., coarser) mesh. The associated rules can then be automatically modified to maintain solution accuracy with better optimized spacing.

[0072] The rules for any given identified feature may depend on the context of the feature, and some, but not necessarily all, embodiments modify the rules based on the context. Context information includes: 1. The purpose of the model. For example, the accuracy required for drag is more stringent than the accuracy required for lift. 2. Surrounding geometry. For example, a vortex source on the leading edge of a wing will require different downstream mesh refinement to maintain adequate solution accuracy compared to a vortex source located near the windshield. This uses a larger region surrounding the target point cloud, sampled with a larger voxel size for the feature points. Thus, the geometric contexts of vortex sources on the leading edge are all similar to each other, but very different from the nose of the aircraft. Each context requires its own rule modification.

[0073] In yet other aspects of some, but not necessarily all, embodiments of the present invention, data sources external to the Computational Fluid Dynamics (CFD) process may also be used to modify the rules. For example, measurements of flows on an actual aircraft (photographs of pressure taps, tufting, oil scars) may establish the need for modified spacing rules once the measured data is made available in point cloud format.

[0074] Additional aspects of some, but not necessarily all, embodiments of the present invention are apparent from the following program code (written in the Python programming language), which shows precisely how to identify pairs of ten independent and reciprocal vortex source sources in front of the global point cloud 500 of the aircraft of Figure 5, given only one of those pairs, which serves as a non-limiting example from which one of ordinary skill in the art will gain further insight into how to make and use embodiments consistent with the present invention.

[0075] JPEG2025530699000002.jpg154166JPEG2025530699000003.jpg199166JPEG2025530699000004.jpg19916 6JPEG2025530699000005.jpg196166JPEG2025530699000006.jpg211166JPEG2025530699000007.jpg19416 6JPEG2025530699000008.jpg201166JPEG2025530699000009.jpg204166JPEG2025530699000010.jpg19716 6JPEG2025530699000011.jpg194166JPEG2025530699000012.jpg212166JPEG2025530699000013.jpg10160

[0076] Additional aspects of embodiments consistent with the present invention will now be described with reference to Figure 11, which illustrates a suitable structural feature recognizer and mesh generator 1101 configured to perform any and / or all of the operations described and illustrated herein. In particular, structural feature recognizer and mesh generator 1101 includes circuitry configured to perform any or any combination of the various functions described herein. Such circuitry may be, for example, entirely hardwired circuitry (e.g., one or more application specific integrated circuits - "ASICs"). However, what is shown in the preferred embodiment of Figure 11 is a programmable circuit, comprising one or more processors 1103, and in some, but not necessarily all, embodiments of the invention, other processing resources capable of performing some of the processing described herein, coupled to one or more memory devices 1105 (e.g., random access memory, magnetic disk drives, optical disk drives, read-only memory, etc.) and an interface 1107 that allows bidirectional communication of data and / or control signals with other components (e.g., input / output devices). An exhaustive list of other possible elements is beyond the scope of this specification.

[0077] The memory device 1105 stores program means 1109 (e.g., a set of processor instructions) that are configured to cause one or more processors 1103 to control other apparatus elements to perform any of the aspects described herein. The memory device 1105 may also store data (not shown) that represent various constant or variable parameters that one or more processors 1103 may require and / or that the processor may require in performing its functions, such as those described by the program means 1109.

[0078] The present invention has been described with reference to specific embodiments. However, those skilled in the art will readily understand that the present invention can be embodied in specific forms other than those of the above-described embodiments by learning and applying aspects of the technology from this disclosure. Therefore, the described embodiments are merely illustrative and should not be considered limiting in any way. The scope of the present invention is further indicated by the appended claims, rather than by the foregoing description, and all modifications and equivalents that fall within the scope of the claims are intended to be embraced therein.

Claims

1. 1. A method for identifying structural features of a structure, comprising: obtaining a global point cloud representation of the structure; obtaining a target point cloud representing structural features of a target, the target point cloud being a subset of the global point cloud representation of the structure; providing global structural information and the target point cloud to a global registration process to generate a globally registered representation of the structure from the global structural information and the target point cloud, wherein the global structural information is derived from the global point cloud representation of the structure; providing the globally aligned representation of the structure and the target point cloud to a local alignment process to generate one or more matching point clouds from the globally aligned representation of the structure and the target point cloud, each of the one or more matching point clouds being a subset of the global point cloud representation of the structure; A method comprising:

2. The method of claim 1 , wherein the global structural information is the global point cloud representation of the structure.

3. generating a set of one or more extracted structural features based on the global point cloud representation of the structure, each of the one or more extracted structural features being a pose-invariant characterization of a local geometry around a point in the global point cloud representation of the structure; The method of claim 1 , wherein the global structural information is a set of the one or more extracted structural features.

4. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, The method of claim 3 , comprising determining a point feature histogram (PFH) based on the global point cloud representation of the structure.

5. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, The method of claim 3 , comprising determining a fast point feature histogram (FPFH) based on a global point cloud representation of the structure.

6. downsampling the global point cloud representation of the structure to generate a downsampled global point cloud representation of the structure; generating the set of one or more extracted structural features based on the global point cloud representation of the structure, The method of claim 3 , comprising generating the set of one or more extracted structural features from the downsampled global point cloud representation of the structure.

7. generating the one or more matching point clouds, 2. The method of claim 1, wherein comparing a subset of the locally aligned representation of the structure with the target structural feature comprises, upon generating a predetermined comparison result, identifying the subset of the locally aligned representation of the structure as one of the one or more clouds of matching points.

8. The method of claim 7 , wherein the predetermined comparison result is a predetermined root mean square error between the subset of the locally aligned representations of the structure and the target structural features.

9. The method of claim 1 , wherein the global alignment process is a Random Sample Consensus (RANSAC) process.

10. The method of claim 1 , wherein the local registration process includes determining an iterative closest point (ICP) value.

11. obtaining the global point cloud representation of the structure, obtaining point cloud data by transforming a CAD geometric representation of said structure; obtaining point cloud data by transforming a discretized representation of the shape of the structure; obtaining point cloud data by transforming a surface mesh representation of said structure; obtaining point cloud data by transforming a volume mesh representation of said structure; obtaining point cloud data by transforming sensor data collected during flight testing of the structure; and obtaining point cloud data by transforming sensor data collected during physical testing of said structure; The method of claim 1 , comprising one or more of:

12. obtaining rules describing a volume associated with the target structural feature; obtaining context information about the location of the target structural feature; generating a corresponding volume for each of the one or more matching point clouds, the corresponding volume having a mesh spacing, a size, and an orientation; the mesh spacing is generated according to the rule, The method of claim 1 , wherein each of the size and pose is generated based on the context information about a location of a structural feature of the target and context information about a location of each of the one or more matching point clouds.

13. acquiring one or more additional target point clouds representing structural features of the target, the one or more additional target point clouds having different sizes from each other and from the target point cloud; for each of the one or more additional target point clouds, providing the global structural information and each of the one or more additional target point clouds to the global registration process to generate one or more additional globally registered representations of the structure from the global structural information and each of the one or more additional target point clouds; for each of the one or more additional target point clouds, providing the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds to the local registration process to generate one or more additional sets of the one or more matching point clouds from the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds; The method of claim 1 , comprising:

14. 1. A non-transitory computer-readable storage medium containing program instructions that, when executed by one or more processors, perform a method for identifying structural features of a structure, the method comprising: The method comprises: obtaining a global point cloud representation of the structure; obtaining a target point cloud representing structural features of a target, the target point cloud being a subset of the global point cloud representation of the structure; providing global structural information and the target point cloud to a global registration process to generate a globally registered representation of the structure from the global structural information and the target point cloud, wherein the global structural information is derived from the global point cloud representation of the structure; providing the globally aligned representation of the structure and the target point cloud to a local alignment process to generate one or more matching point clouds from the globally aligned representation of the structure and the target point cloud, each of the one or more matching point clouds being a subset of the global point cloud representation of the structure; 1. A non-transitory computer-readable storage medium comprising:

15. The non-transitory computer-readable storage medium of claim 14 , wherein the global structural information is the global point cloud representation of the structure.

16. The method comprises: generating a set of one or more extracted structural features based on the global point cloud representation of the structure, each of the one or more extracted structural features being a pose-invariant characterization of a local geometry around a point in the global point cloud representation of the structure; The non-transitory computer-readable storage medium of claim 14 , wherein the global structural information is a set of the one or more extracted structural features.

17. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 17. The non-transitory computer-readable storage medium of claim 16, comprising determining a point feature histogram (PFH) based on the global point cloud representation of the structure.

18. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 17. The non-transitory computer-readable storage medium of claim 16, comprising determining a fast point feature histogram (FPFH) based on the global point cloud representation of the structure.

19. The method comprises: downsampling the global point cloud representation of the structure to generate a downsampled global point cloud representation of the structure; generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 17. The non-transitory computer-readable storage medium of claim 16, further comprising generating the set of one or more extracted structural features from the down-sampled global point cloud representation of the structure.

20. generating the one or more matching point clouds, 15. The non-transitory computer-readable storage medium of claim 14, wherein comparing a subset of the locally aligned representation of the structure with the target structural feature comprises, upon generating a predetermined comparison result, identifying the subset of the locally aligned representation of the structure as one of the one or more clouds of matching points.

21. 21. The non-transitory computer-readable storage medium of claim 20, wherein the predetermined comparison result is a predetermined root-mean-square error between a subset of the locally aligned representations of the structure and the target structural features.

22. 15. The non-transitory computer-readable storage medium of claim 14, wherein the global alignment process is a Random Sample Consensus (RANSAC) process.

23. 15. The non-transitory computer-readable storage medium of claim 14, wherein the local alignment process includes determining an iterative closest point (ICP) value.

24. obtaining the global point cloud representation of the structure, obtaining point cloud data by transforming a CAD geometric representation of said structure; obtaining point cloud data by transforming a discretized representation of the shape of the structure; obtaining point cloud data by transforming a surface mesh representation of said structure; obtaining point cloud data by transforming a volume mesh representation of said structure; obtaining point cloud data by transforming sensor data collected during flight testing of the structure; and obtaining point cloud data by transforming sensor data collected during physical testing of said structure; 15. The non-transitory computer-readable storage medium of claim 14, comprising one or more of:

25. The method comprises: obtaining rules describing a volume associated with the target structural feature; obtaining context information about the location of the target structural feature; generating a corresponding volume for each of the one or more matching point clouds, the corresponding volume having a mesh spacing, a size, and an orientation; the mesh spacing is generated according to the rule, 15. The non-transitory computer-readable storage medium of claim 14, wherein each of the size and pose is generated based on the context information about a location of a structural feature of the target and context information about a location of each of the one or more matching point clouds.

26. The method comprises: acquiring one or more additional target point clouds representing structural features of the target, the one or more additional target point clouds having different sizes from each other and from the target point cloud; for each of the one or more additional target point clouds, providing the global structural information and each of the one or more additional target point clouds to the global registration process to generate one or more additional globally registered representations of the structure from the global structural information and each of the one or more additional target point clouds; for each of the one or more additional target point clouds, providing the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds to the local registration process to generate one or more additional sets of the one or more matching point clouds from the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds; 15. The non-transitory computer-readable storage medium of claim 14, comprising:

27. 1. A system for identifying structural features of a structure, comprising: one or more non-transitory memories storing program instructions; and one or more processors for executing said program instructions, The one or more processors execute the program instructions to: obtaining a global point cloud representation of the structure; obtaining a target point cloud representing structural features of a target, the target point cloud being a subset of the global point cloud representation of the structure; providing global structural information and the target point cloud to a global registration process to generate a globally registered representation of the structure from the global structural information and the target point cloud, wherein the global structural information is derived from the global point cloud representation of the structure; providing the globally aligned representation of the structure and the target point cloud to a local alignment process to generate one or more matching point clouds from the globally aligned representation of the structure and the target point cloud, each of the one or more matching point clouds being a subset of the global point cloud representation of the structure; A system that runs

28. 28. The system of claim 27, wherein the global structural information is the global point cloud representation of the structure.

29. the one or more processors: generating a set of one or more extracted structural features based on the global point cloud representation of the structure, each of the one or more extracted structural features being a pose-invariant characterization of a local geometry around a point in the global point cloud representation of the structure; 28. The system of claim 27, wherein the global structural information is a set of the one or more extracted structural features.

30. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 30. The system of claim 29, comprising determining a point feature histogram (PFH) based on the global point cloud representation of the structure.

31. generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 30. The system of claim 29, comprising determining a fast point feature histogram (FPFH) based on a global point cloud representation of the structure.

32. the one or more processors: configured to further perform the step of downsampling the global point cloud representation of the structure to generate a downsampled global point cloud representation of the structure; generating the set of one or more extracted structural features based on the global point cloud representation of the structure, 30. The system of claim 29, further comprising generating the set of one or more extracted structural features from the downsampled global point cloud representation of the structure.

33. generating the one or more matching point clouds, 28. The system of claim 27, wherein comparing a subset of the locally aligned representation of the structure with the target structural feature comprises, upon generating a predetermined comparison result, identifying the subset of the locally aligned representation of the structure as one of the one or more matching point clouds.

34. 34. The system of claim 33, wherein the predetermined comparison result is a predetermined root mean square error between the subset of the locally aligned representations of the structure and the target structural features.

35. 28. The system of claim 27, wherein the global alignment process is a random sample consensus (RANSAC) process.

36. 28. The system of claim 27, wherein the local registration process includes determining an iterative closest point (ICP) value.

37. obtaining the global point cloud representation of the structure, obtaining point cloud data by transforming a CAD geometric representation of said structure; obtaining point cloud data by transforming a discretized representation of the shape of the structure; obtaining point cloud data by transforming a surface mesh representation of said structure; obtaining point cloud data by transforming a volume mesh representation of said structure; obtaining point cloud data by transforming sensor data collected during flight testing of the structure; and obtaining point cloud data by transforming sensor data collected during physical testing of said structure; 28. The system of claim 27, comprising one or more of:

38. the one or more processors: obtaining rules describing a volume associated with the target structural feature; obtaining context information about the location of the target structural feature; generating a corresponding volume for each of the one or more matching point clouds, the corresponding volume having a mesh spacing, a size, and an orientation; the mesh spacing is generated according to the rule, 28. The system of claim 27, wherein each of the size and pose is generated based on the context information about a location of a structural feature of the target and context information about a location of each of the one or more matching point clouds.

39. the one or more processors: acquiring one or more additional target point clouds representing structural features of the target, the one or more additional target point clouds having different sizes from each other and from the target point cloud; for each of the one or more additional target point clouds, providing the global structural information and each of the one or more additional target point clouds to the global registration process to generate one or more additional globally registered representations of the structure from the global structural information and each of the one or more additional target point clouds; for each of the one or more additional target point clouds, providing the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds to the local registration process to generate one or more additional sets of the one or more matching point clouds from the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds; 28. The system of claim 27, further configured to:

40. 1. A structural feature recognition system for use in computer engineering, the structural feature recognition system being configured to identify structural features of a structure, the system comprising: a circuit configured to obtain a global point cloud representation of the structure; a circuit configured to obtain a target point cloud representing a structural feature of a target, the target point cloud being a subset of the global point cloud representation of the structure; a circuit configured to provide global structural information and the target point cloud to a global registration process to generate a globally registered representation of the structure from the global structural information and the target point cloud, the global structural information being derived from the global point cloud representation of the structure; a circuit configured to provide the globally aligned representation of the structure and the target point cloud to a local alignment process to generate one or more matching point clouds from the globally aligned representation of the structure and the target point cloud, each of the one or more matching point clouds being a subset of the global point cloud representation of the structure; A structural feature recognition device comprising:

41. 41. The structural feature recognition apparatus of claim 40, wherein the global structural information is the global point cloud representation of the structure.

42. a circuit configured to generate a set of one or more extracted structural features based on the global point cloud representation of the structure, each of the one or more extracted structural features being a pose-invariant characterization of a local geometry around a point in the global point cloud representation of the structure; 41. The structural feature recognition apparatus of claim 40, wherein the global structural information is a set of the one or more extracted structural features.

43. a circuit configured to generate the set of one or more extracted structural features based on the global point cloud representation of the structure, 43. The structural feature recognition apparatus of claim 42, comprising circuitry configured to determine a point feature histogram (PFH) based on the global point cloud representation of the structure.

44. a circuit configured to generate the set of one or more extracted structural features based on the global point cloud representation of the structure, 43. The structural feature recognition apparatus of claim 42, comprising circuitry configured to determine a fast point feature histogram (FPFH) based on the global point cloud representation of the structure.

45. the structural feature recognition device comprising circuitry configured to downsample the global point cloud representation of the structure to generate a downsampled global point cloud representation of the structure; a circuit configured to generate the set of one or more extracted structural features based on the global point cloud representation of the structure, 43. The structural feature recognition apparatus of claim 42, comprising circuitry configured to generate the set of one or more extracted structural features from the downsampled global point cloud representation of the structure.

46. a circuit configured to generate the one or more matching point clouds, 41. The structural feature recognition apparatus of claim 40, comprising circuitry configured to identify a subset of the locally aligned representation of the structure as one of the one or more matching point clouds when a comparison of the subset of the locally aligned representation of the structure with the target structural feature produces a predetermined comparison result.

47. 47. The structural feature recognition apparatus of claim 46, wherein the predetermined comparison result is a predetermined root mean square error between the subset of the locally aligned representations of the structure and the target structural feature.

48. 41. The structural feature recognition apparatus of claim 40, wherein the global alignment process is a random sample consensus (RANSAC) process.

49. 41. The structural feature recognition apparatus of claim 40, wherein the local alignment process includes determining an iterative closest point (ICP) value.

50. a circuit configured to obtain the global point cloud representation of the structure, a circuit configured to obtain point cloud data by transforming a CAD geometric representation of the structure; a circuit configured to obtain point cloud data by transforming a discretized representation of the shape of the structure; a circuit configured to obtain point cloud data by transforming a surface mesh representation of the structure; a circuit configured to obtain point cloud data by transforming a volume mesh representation of the structure; a circuit configured to acquire point cloud data by transforming sensor data collected during flight testing of the structure; and a circuit configured to acquire point cloud data by transforming sensor data collected during physical testing of the structure; 41. The structural feature recognition apparatus of claim 40, comprising one or more of:

51. a circuit configured to obtain rules describing a volume associated with a structural feature of the target; a circuit configured to obtain contextual information about a location of the target structural feature; and a circuit configured to generate a corresponding volume for each of the one or more matching point clouds, the corresponding volume having a mesh spacing, a size, and an orientation; the mesh spacing is generated according to the rule, 41. The structural feature recognition apparatus of claim 40, wherein each of the size and pose is generated based on the context information about a location of the target structural feature and context information about a location of each of the one or more matching point clouds.

52. circuitry configured to acquire one or more additional target point clouds representative of structural features of the target, the one or more additional target point clouds having different sizes from each other and from the target point cloud; a circuit configured to generate, for each of the one or more additional target point clouds, one or more additional globally aligned representations of the structure from the global structural information and each of the one or more additional target point clouds by providing the global structural information and each of the one or more additional target point clouds to the global registration process; a circuit configured to generate one or more additional sets of one or more matching points from each of the one or more additional globally aligned representations of the structure and the one or more additional target point clouds by providing the one or more additional globally aligned representations of the structure and each of the one or more additional target point clouds to the local alignment process; 41. The structural feature recognition apparatus of claim 40, comprising:

53. 41. A computer engineered system comprising the structural feature recognition apparatus of claim 40.

54. obtaining the global point cloud representation of the structure, The method of claim 1 , comprising converting a global non-point cloud representation of the structure into the global point cloud representation of the structure.

55. obtaining the global point cloud representation of the structure, 15. The non-transitory computer-readable storage medium of claim 14, comprising converting a global non-point cloud representation of the structure into the global point cloud representation of the structure.

56. obtaining the global point cloud representation of the structure, 28. The system of claim 27, comprising converting a global non-point cloud representation of the structure into the global point cloud representation of the structure.

57. a circuit configured to obtain a global point cloud representation of the structure, 41. The structural feature recognition apparatus of claim 40, comprising circuitry configured to convert a global non-point cloud representation of the structure into the global point cloud representation of the structure.

Images