Computer-implemented method for segmenting measurement data from an object measurement

Abstract

The invention relates to a computer-implemented method of segmenting measurement data from a measurement of an object having at least one material transition, generating a digital object representation comprising the at least one material transition from the measurement data, having a plurality of spatially resolved image information of the object, the method having the steps of: determining measurement data having at least one microstructure with an extension smaller than a predetermined extension; determining at least two homogeneous regions in the measurement data and / or the digital object representation, at least one of which has a microstructure; analyzing local similarities of the plurality of spatially resolved image information; adjusting the extension of the respective homogeneous regions until at least one boundary region of the respective homogeneous regions is arranged at an expected position of the material transition; segmenting the digital object representation based on the adjusted homogeneous regions. An improved computer-implemented method of segmenting measurement data from a measurement of an object having a microstructure is also provided, which provides a correct detection of material transitions from the measurement data of the object.

Description

Technical Field

[0001] This invention relates to a computer-implemented method for segmenting measurement data from object measurements. Background Technology

[0002] To check whether the manufactured object meets the desired set conditions for quality control, the object is measured and compared with those desired set conditions. Measurement can be performed, for example, as a dimensional measurement. Dimensional measurement can be achieved, for example, by scanning different points on the object's surface. Alternatively, computed tomography (CT) can be performed, in which the obtained measurement data is analyzed. In this case, the internal surfaces of the object can also be inspected. The measurement data can then exist, for example, in the form of volumetric data or be converted into volumetric data. To distinguish different object regions in the measurement data, the measurement data is segmented into different regions. This is particularly meaningful, for example, in visualization, reverse engineering, multi-component functional analysis, and material and material property simulation. Furthermore, the measurement data can be preprocessed before performing the method. For example, artifact correction (e.g., metal artifacts), photopolymerization, or geometr-based scattered light correction and data filtering (e.g., Gaussian filtering or median filtering) can be applied to the measurement data.

[0003] However, the segmentation of volumetric data for multi-material measurements has so far been unsatisfactory because the segmentation algorithm needs to be specifically adjusted for each material transition between two specific materials. Therefore, for example, a smaller threshold must be used when analyzing grayscale values ​​to identify material transitions between materials with smaller grayscale values ​​in the measurement data compared to the threshold used to identify transitions between materials with larger grayscale values. Thus, the prospect of obtaining volumetric data segmentation based on an overall threshold is slim. Many algorithms fail to segment the different materials correctly, especially when the measurement data contains artifacts or microstructures. Furthermore, correct segmentation is insufficient to provide accurate measurements at all material transitions, i.e., insufficient to precisely determine the location of material transitions. Summary of the Invention

[0004] Therefore, providing an improved computer-implemented method for segmenting measurement data from measurements of objects with microstructures can be considered the task of the present invention, wherein the method provides correct identification of material transitions from the object measurement data.

[0005] In one aspect of the invention, a computer-implemented method is provided for segmenting measurement data from an object measurement, wherein the object has at least one material transition region, wherein a digital object representation including the at least one material transition region is generated from the measurement data, wherein the digital object representation has multiple spatially resolved image information of the object, wherein the method comprises the steps of: determining measurement data, wherein the measurement data has at least one microstructure having an extension range smaller than a predetermined extension range; determining at least two homogeneous regions in the measurement data and / or the digital object representation, at least one of the at least two homogeneous regions having a microstructure; analyzing the local similarity of the multiple spatially resolved image information; adjusting the extension range of each homogeneous region until at least one boundary region of each homogeneous region is arranged at a desired location of the material transition region; and segmenting the digital object representation based on the adjusted homogeneous regions.

[0006] This invention employs various algorithms to segment objects with microstructures. In this case, the algorithm studies different display patterns of the object's measurement data. By using different algorithms along with their respective advantages and disadvantages, the strengths of the algorithms used can be fully utilized as optimally as possible. For example, an algorithm can be used to first analyze the image information of the image data, where, for example, each image piece of information is compared with locally adjacent image information to determine homogeneous regions. This can be called pre-segmentation. Furthermore, this can advantageously be based on three-dimensional measurement data, but two-dimensional measurement data that can also be logically associated with three-dimensional measurement data can also be used. Thus, similar image information is aggregated into a homogeneous region. At least one homogeneous region is determined in this way. In this case, the algorithm on which the homogeneous region is based may be inaccurate; therefore, the boundary of the homogeneous region does not coincide with the location of the material transition zone that defines the homogeneous region. Another algorithm can be used to analyze the local similarity of the image information. Local similarity analysis can determine regions where the image information is only slightly similar to adjacent image information. These regions can be considered as expected locations of material transition zones. These expected locations can also be obtained, for example, from the theoretical geometry of the object or other displays of the measurement data. The boundary region of the homogeneous zone is then adjusted using another algorithm, for example, by moving its position. In this case, the extent of the homogeneous zone can be changed. The position of the boundary region is adjusted until it includes the expected location of the material transition zone. The shortcomings of each algorithm can therefore be compensated for by employing other algorithms. In this case, the boundary region is understood to refer to the portion of the homogeneous zone used to define the homogeneous zone. In this case, the boundary region can have a predetermined extent of extension within the homogeneous zone.

[0007] In this example, during local similarity display, regions with values ​​exceeding a predetermined threshold for local similarity can be considered material transition zones between different material regions. The region defined by this material transition zone then corresponds entirely to the material that has the largest portion of that region after pre-segmentation. It is also possible that no complete material transition zone is formed between the material regions. This can be achieved, for example, by a "closing" shape operation, in which case the associated material transition zones are coexisting and the small regions between them are removed.

[0008] Therefore, the segmentation of the digital object representation is based on adjusted homogeneous regions between at least two homogeneous regions. Here, the determination of the expected location of the material transition region may involve searching within a small search range at the edges of the homogeneous regions. In this case, pre-segmentation (e.g., coarse segmentation) is performed before the step of segmenting the digital object representation based on adjusted homogeneous regions, which acts as the main segmentation (e.g., finer segmentation), which is arranged together with the steps of analyzing the local similarity of multiple spatially resolved image information and adjusting the extension range of each homogeneous region to at least one boundary region of each homogeneous region, at a expected location of the material transition region. Pre-segmentation may, for example, include the step of determining at least two homogeneous regions in the measurement data and / or digital object representation, wherein at least one of the at least two homogeneous regions has a microstructure. In this case, when local similarity decreases, the material transition region between the homogeneous regions can be determined, for example, in the subsequent main segmentation. Otherwise, the related homogeneous regions merge. In this case, the material transition region may be, for example, a material surface, two adjacent material surfaces, multiple material transition sections separated by narrow material boundaries, or a transition section of the internal structure of a single material. The result of coarse segmentation can be the detection of the extent of a homogeneous region, but it can also be the detection of texture-like areas or microstructures within that homogeneous region. Next, the segmentation step for digital object representation can be performed based on the adjusted homogeneous region. Microstructures within the homogeneous region can be detected better by combining pre-segmentation and main segmentation compared to without such a combination.

[0009] Microstructures can be understood, for example, as textures or structures with an extension range smaller than a predetermined extension range. The size of at least one microstructure is determined using the predetermined extension range. This predetermined extension range can be set by the user or by mapping specifications for the method.

[0010] In this context, a homogeneous region is understood to refer to a region with a uniform material or a uniform mixture of materials. Image information, for example, could be grayscale values ​​obtained from measurement data in computed tomography X-ray photography within the object's dimensional measurement range.

[0011] Furthermore, regions are considered homogeneous when their measurement data or image information lies, for example, between two thresholds (e.g., between an upper threshold and a lower threshold), meaning that local measurement data are similar or have similar values ​​within that region, i.e., when local similarity is high. In one example, the image information of homogeneous regions in a digital object representation could therefore have grayscale values ​​within a narrow range. These regions within the object can have a uniform material or a uniform mixture of materials. Therefore, homogeneous regions are not absolutely uniform but can have fluctuations within tolerance. The thresholds can be predetermined or determined when defining the homogeneous regions. However, the homogeneity of a region does not need to be defined by grayscale values. In another example, regions with fibrous materials but similar fiber orientations can be considered homogeneous, even if the grayscale values ​​themselves are not uniform in this case. However, the pattern defined by the texture derived from the fibers is homogeneous. The material of a region or the entire object can, for example, be a single piece of material; that is, material transitions in material transition areas could, in this example, be transitions between different material structures or between a single piece of material and the background.

[0012] Material transition zones can, for example, be transitions between areas of biomaterials, welds, or regions with different fiber orientations. It is not necessary for this material transition zone to have a distinct material surface. In another example, the material transition zone can approximate or represent a surface in measurements and CAD models.

[0013] Furthermore, the at least one material transition region can be, for example, a multi-material transition region. The term multi-material here does not simply refer to regions of multiple uniform individual materials. The presence of fibers or pores can also clearly indicate regions of separated materials within a single block of material, even when the base material remains constant. Regions with different properties, especially when the material composition is the same or similar, can also be clearly interpreted as separated materials. Therefore, the background of a CT scan (typically the air surrounding the object) can also be considered material in the measurement data.

[0014] In other words, in addition to image information representing the background of the object, the object also contains at least two materials in the measurement data to confirm the material transition (e.g., the surface).

[0015] According to another example, after the step of determining at least two homogeneous regions in the measurement data and / or digital object representation (where at least one of the at least two homogeneous regions has a microstructure), a predetermined material can be assigned to the at least two homogeneous regions.

[0016] Therefore, this method can assign specific materials to these segmented regions based on the expected material transition zone locations and the representation of the segmented object, after pre-segmentation, main segmentation, or subsequent optional accurate surface determination or accurate determination of material transition zone locations, provided that this has not been done beforehand, for example, through pre-registering theoretical geometry (such as a CAD model). If no theoretical geometry exists, this can be performed using different methods. Typical grayscale values ​​of the segmented regions can be compared, for example, with a list of expected materials and their associated grayscale values ​​or absorption coefficients. Alternatively or additionally, a component catalog may exist based on the geometry of the segmented regions, containing geometries that can be compared with the geometry of the regions.

[0017] According to another example, the method may be specified to have the following additional steps: when it is determined in at least one step of the method that the extension range of at least one homogeneous region is less than a predetermined extension range, theoretical image information is determined from said at least one homogeneous region.

[0018] During analysis, a small, new region may be identified that has not yet been assigned to a material. Because of its small size, it is difficult to determine the (perhaps constant) grayscale level to identify the material to which it belongs. This is especially true when the point spread function (PSF) of the recording system is on the same order of magnitude as or larger than the size of the small region. In this case, adjacent material or homogeneous regions adversely affect or distort the grayscale value of the small region. Therefore, in this example, a theoretically correct or undisturbed grayscale value needs to be determined for this small region. This theoretically correct or undisturbed grayscale value is the uniform grayscale value that the material possesses when it is not disturbed by adjacent material. This can be achieved, for example, by utilizing knowledge of the uniform grayscale values, PSF, and / or the geometry of the boundary surfaces of adjacent regions. Based on this knowledge, it can be calculated, for example (perhaps in a model-based manner), what the correct grayscale level of the small region must be to produce the grayscale distribution that appears in the measurement data. In another example, this theoretically correct or undisturbed grayscale value can be obtained using mathematical expansion. Using this theoretically correct or undisturbed grayscale value, it can now be assigned to the material.

[0019] In another example, when it is determined in at least one step of the method that the extension range of at least one homogeneous region is less than the extension range of a predetermined region, a predetermined material may be assigned to the homogeneous region of the at least two homogeneous regions based on the topology of one of the at least two homogeneous regions.

[0020] In this scenario, new regions are assigned materials using topological analysis. Information related to the connecting materials in the environment is employed. When a new region is found, for example, in a material transition zone between two materials, topological analysis may assign this new region to an adjacent region. The topology of the new region and adjacent regions is examined. The examined topology is used to determine whether a connection exists between the new region and the adjacent regions. The portion of the new region that is connected to one of the adjacent regions is assigned to that adjacent region and connected to the corresponding adjacent region. In this case, the entire new region may also be connected to the corresponding adjacent region. Similarly, the examined topology can be used to determine the boundary surfaces between the new region or a portion of the new region and adjacent regions. Adjacent regions with boundary surfaces with the new region or the portion thereof can be excluded from connection to the new region or its portion. Furthermore, when the examined topology determines that no connection with adjacent regions exists, the new region may form a separate homogeneous region. For this purpose, this information is used to check whether the region belongs to one of the materials, for example, based on a theoretically correct or undisturbed grayscale level.

[0021] Alternatively, or additionally, certain areas adjacent to the new area may have been excluded because the new area is necessarily a separate area due to its topological structure.

[0022] In another example, the analysis of local similarity can be based on the process of change of multiple spatially resolved image information and / or the local differences of multiple spatially resolved image information.

[0023] When the image information is, for example, grayscale values, the change process can represent the gradient of the spatially resolved grayscale values. When the homogeneous region is based on texture, local similarity can be determined, for example, by utilizing local differences in the image information. In this case, the gradient display is preferably the absolute value of the local gradient. They indicate the increase in value near the material transition region.

[0024] Furthermore, before determining at least two homogeneous regions of the digital object representation, the method may, for example, include the step of aligning a digital representation of the theoretical geometry with the digital object representation; wherein the determination of at least two homogeneous regions is performed based on the digital representation of the theoretical geometry.

[0025] To this end, for example, the expected location of the material transition zone can be obtained from the theoretical geometry to at least obtain a rough pre-alignment of the measurement data. In this case, the theoretical geometry can be a CAD model of the object. The region of the theoretical geometry or the CAD model can then be assigned to the corresponding region of the measurement data. The computer-implemented method can therefore utilize existing knowledge of the theoretical geometry when determining the location of the material transition. This can be performed as part of pre-segmentation.

[0026] Alternatively or additionally, information about the geometry of the object may be used from measurements taken by other sensors, such as optical methods like ribbon light projection.

[0027] For example, after the step of determining at least two homogeneous regions in the measurement data and / or digital object representation (where at least one of the at least two homogeneous regions has a microstructure), the method may include the step of creating a labeled segment that defines the homogeneous regions by means of spatially resolved labeled values ​​in the measurement data and / or digital object representation, wherein the spatial resolution of the labeled segment is higher than the spatial resolution of the measurement data.

[0028] When it is necessary to segment microstructures (such as thin layers of material between two other materials), detection reliability can be further improved by adapting the homogeneous region to the microstructure after pre-segmentation. In this case, the main segmentation can be performed, for example, by means of a marked domain segment that defines the homogeneous region as a separate material region, for example, which can be defined by a material transition region. The marked domain segment and the measurement data do not need to have the same orientation for this purpose. In order to ensure that the material region can also be adapted to the microstructure, i.e., a small, thin, and / or narrow structure, the resolution of the marked domain segment can be selected to be higher than the resolution of the measurement data to be segmented, i.e., finer.

[0029] A labeled segment assigns a material to a location in the representation of a digital object. For this purpose, different values ​​or ranges of image information (e.g., grayscale values) can be assigned. Thus, a specific range, for example, between two thresholds, can be assigned to different materials under different circumstances. Simultaneously, this assignment defines homogeneous regions. The labeled segment implicitly represents the approximate location of material transition zones. According to another example, distance values ​​of distance segments can be assigned to each labeled segment, where the distance value defines the shortest distance to the nearest boundary of the relevant homogeneous region. This distance segment implicitly represents the location of the surface. The final material transition zones of different materials can be stored with subvoxel accuracy using a single (perhaps unsigned) distance segment. The distance segment here represents or stores the location of the surface. In this case, a single labeled value can also be assigned to multiple distance values ​​and thus, for example, to different homogeneous regions in overlapping areas. Together with the labeled segment, which material transition zone is involved is determined for each region of the surface. This can be shown by the materials shown adjacent to each other in the labeled segment. Since a marked field will usually appear when defining a surface anyway, a distance field is a particularly efficient and feasible way to describe or store that surface.

[0030] In another example, after determining at least two homogeneous regions in the measurement data and / or digital object representation (where at least one of the at least two homogeneous regions has a microstructure), the method may include the following steps: creating a labeled segment that defines the homogeneous region by means of a spatially resolved labeled value in the measurement data and / or digital object representation, wherein, at least in a predetermined region of the object representation, the spatial resolution of the labeled segment is higher than the spatial resolution of the measurement data, and wherein, in the remaining regions of the object representation, the spatial resolution of the labeled segment is at most the same as the spatial resolution of the measurement data.

[0031] Resolution can be varied locally to optimize computation time. A high resolution can be selected for locations where microstructures are estimated. In principle, a low resolution can be maintained in areas where no microstructures are estimated or where microstructures cannot be arranged. In this case, the estimate can be included in mapping specifications, for example, derived from an analysis of at least two homogeneous regions, also known as pre-segmentation, or set by the user. Alternatively, the estimate can be derived from additive manufacturing process data, derived from empirical values, or calculated from manufacturing process simulations.

[0032] In another example, the step of determining at least two homogeneous regions may have the following sub-step: obtaining multiple edges between the at least two homogeneous regions.

[0033] Therefore, material transition regions can be analyzed for multiple edges during pre-segmentation. This analysis can be model-based and can identify regions with multiple edges, thus revealing previously undiscovered regions existing between two outer materials. These material regions can then be considered during pre-segmentation (e.g., within labeled domain segments). In principle, it is also conceivable that even more other material regions within a single material transition region can be identified in this way. Such multi-edge searching can be based on model segmentation, which can identify grayscale transitions where at least two material transitions are represented in a superimposed manner.

[0034] Multiple edges should be understood as edge regions that include at least two different surfaces.

[0035] This method may include, for example, the following steps after digital object representation segmentation: analyzing image information of the determined boundary region to identify erroneous portions of the material transition zone within the determined boundary region; and correcting the erroneous portions of the material transition zone within the determined boundary region.

[0036] Following the segmentation step based on the digital object representation of the adjusted homogeneous region, an analysis of the segmented material transition region can be performed to identify any remaining erroneous segments, i.e., incorrect parts. Here, the analysis is not based on local similarity representation, but on the grayscale values ​​themselves. Because the material boundaries were determined quite accurately at that time, it is guaranteed that the analysis is completed at the correct location. This is particularly advantageous. If erroneous segments are identified within the material transition region, the segmentation is corrected. New material regions are introduced as needed. In this case, model-based methods, such as the multi-edge search described above, can also be used.

[0037] Additionally, the method may include the following steps: acquiring at least one heterogeneous region in the object representation; analyzing the spatially resolved image information of the at least one heterogeneous region to determine the texture within the at least one heterogeneous region; and segmenting the at least one heterogeneous region.

[0038] Especially in additive manufacturing, microstructures, such as textures, are often generated. These microstructures can be better segmented by acquiring at least one heterogeneous region in the object representation and analyzing the spatially resolved image information of that at least one heterogeneous region to determine the texture within it. Alternative methods, different from those described above, can be used to identify the heterogeneous region.

[0039] According to another example, the at least one heterogeneous region may contain measurement data of the powdery region of the object.

[0040] In objects produced using additive manufacturing methods, there are often regions with unmelted powder, which can be mistaken for the fixed geometry of the manufactured object. The combination of microstructure identification and texture analysis steps thus improves the accuracy of measuring the overall geometry of additively manufactured objects.

[0041] The at least one heterogeneous region can be treated as a region with different characteristics from the heterogeneous region in further analysis.

[0042] Regions identified as heterogeneous areas in the measurement data are treated separately in further analysis. In this case, the heterogeneous areas are considered as separate regions with different material properties. These material properties can therefore differ from, for example, those of the corresponding solid material. In particular, this may also mean that the heterogeneous areas are evaluated as not belonging to the geometry of the object. The heterogeneous areas are used as the background of the object because they cannot withstand any load. Further analysis may, for example, be dimensional measurement techniques, defect analysis, or mechanical simulation.

[0043] On the other hand, a computer program product is provided having instructions executable on a computer, which, when executed on the computer, cause the computer to perform the method described above.

[0044] The advantages, functions, and improvements of the computer program product derive from the advantages, functions, and improvements of the methods described above. Therefore, reference is made to the above description. A computer program product may, for example, be a data carrier storing computer program elements having instructions executable by a computer. Alternatively or additionally, a computer program product may also refer to, for example, a permanent or volatile data storage device having such a computer program element, such as flash memory or working memory. However, this does not exclude other types of data storage devices having such a computer program element. Attached Figure Description

[0045] Other features, details, and advantages of the present invention are derived from the statements of the claims and the following description of embodiments in conjunction with the accompanying drawings, wherein:

[0046] Figure 1 A flowchart of the computer-implemented method is shown.

[0047] Figure 2 A flowchart illustrating the sub-steps of an exemplary implementation including the determining step is shown.

[0048] Figure 3 A schematic diagram of the multi-material transition region is shown.

[0049] Figure 4a and Figure 4b The diagram illustrates the material transition before and after the creation of a new region.

[0050] Figures 5a to 5e A schematic diagram illustrating a series of steps in an exemplary implementation of the method is shown. Detailed Implementation

[0051] The computer-implemented method for segmenting measurement data from object measurements is generally indicated by reference numeral 100 in the accompanying drawings. In the following text, it will be referred to by means of... Figure 1 To illustrate the computer implementation method 100.

[0052] Figure 1 A flowchart illustrating one embodiment of a computer-implemented method 100 for segmenting measurement data from an object measurement is shown. In this case, the object has at least one material transition zone. Furthermore, the measurement data includes at least one microstructure having an extension range smaller than a predetermined extension range. This predetermined extension range can, for example, be preset by a user or a mapping specification. Additionally, this predetermined extension range can, for example, be determined by simulating the manufacture of the object.

[0053] In the first step 102, measurement data of the object is determined. In this case, the measurement data can be determined, for example, by means of computed tomography (CT) measurements. However, this does not preclude other methods for determining the measurement data, such as magnetic resonance imaging (MRI). The measurement data is used to generate a digital object representation containing at least one material transition region. This digital object representation includes multiple spatially resolved image information of the object.

[0054] When the measurement data is CT data, it is not necessarily required that each voxel contain only one gray value. Therefore, it can be multimodal data, i.e., data from multiple sensors or from multi-energy CT scans, thus having multiple gray values ​​for each voxel. Furthermore, in method 100, the analysis results based on the initial measurement data can be used as additional spatially resolved gray values, such as the analysis results of fiber orientation or local porosity. Even when no visible spectral colors are displayed, additional information, such as that referred to as color channels, can therefore be interpreted as color voxel data. This additional information can be advantageously used in method 100.

[0055] In optional step 114, the digital representation of the object's theoretical geometry is aligned with the digital object representation derived from the measurement data determined according to step 102. The digital representation of the object's theoretical geometry can, for example, be a CAD representation of the object created prior to its fabrication. The geometry in the CAD model does not necessarily have to be described as a surface or material transition zone. Instead, or additionally, it can be implicitly represented as image stacks, voxel volumes, or distance domain segments. This can be particularly useful in additive manufacturing. Furthermore, this information can be directly or effortlessly transformed into labeled domain segments. However, this does not preclude other representations of the theoretical geometry.

[0056] At least two homogeneous regions are determined in the measurement data and / or the digital representation of the object based on the theoretical geometry. Because the material transition region and the object component or object region with homogeneous material are known in the digital representation of the theoretical geometry, the homogeneous regions in the measurement data or in the digital representation of the object generated from the measurement data can be inferred from the digital representation of the theoretical geometry after alignment in step 114.

[0057] During alignment, i.e., when adapting the shape region of the theoretical geometry to the measurement data, it is important to consider which materials are involved in the grayscale transitions and how they are distributed. From the material distribution, the direction of the material transition can be obtained. This information is usually known in the theoretical geometry and can be easily determined locally based on the measurement data in different situations. This avoids mismatched material transition areas being assigned to each other, which would lead to incorrect alignment.

[0058] The alignment can also be performed using a non-rigid mapping between measurement data and theoretical geometry.

[0059] In another step 104, at least two homogeneous regions are identified in the measurement data and / or digital object representation. For this purpose, image information is analyzed to determine the presence of homogeneous regions, such as regions within a grayscale range or with similar texture. In this case, at least one of the at least two homogeneous regions has a microstructure. Alternatively, all homogeneous regions may have a microstructure.

[0060] In optional step 116, a labeled region can be created for the determined homogeneous region, which is defined by spatially resolved labeled values ​​in the measurement data and / or digital object representation. In this case, the spatial resolution of the labeled region is higher than that of the measurement data. That is, the labeled region can be defined, for example, with sub-voxel accuracy. Furthermore, the resolution can be locally varied to optimize computation time. A larger labeled region resolution can be chosen where microstructures are estimated. In principle, a lower resolution can be maintained in regions where no microstructures are estimated or where it is impossible to establish microstructures.

[0061] In another optional step 118, a labeled segment may be created during optional pre-segmentation, which defines a homogeneous region by means of spatially resolved labeled values ​​in measurement data and / or digital object representations.

[0062] A labeled segment can be combined with a signed or unsigned distance segment. In this case, each labeled value is assigned at least one distance value from the distance segment. The distance value describes the distance to the nearest boundary surface of the homogeneous region. A separate distance segment can be created for each material.

[0063] The interface of the homogeneous region is arranged within the material transition region. In this case, a single marker value can be assigned to multiple distance domain segments, and therefore can be assigned to multiple distance values. That is, the material transition regions for each material within the object can be represented by their respective distance domain segments. Using distance domain segments, the size of the homogeneous region can be recorded with minimal memory usage and computational cost.

[0064] In this situation, existing knowledge can be utilized, for example, to indicate that only a specific volume of a certain material within an object is present within the measurement range. This can be taken into account when creating marked segments, by not assigning larger associated regions to that material. This can reduce segmentation errors.

[0065] Therefore, for example, there might be a spiral of a certain maximum size within the measurement area. If a large area of ​​the material is allocated at a certain location in the measurement volume using this method, it can be determined in this way that the allocation may be incorrect.

[0066] In principle, the alignment or registration of theoretical geometry (e.g., a CAD model) can be performed by adapting the material transition regions from the measurement to the corresponding material transition regions of the theoretical geometry. That is, searching for the poses where they coincide as well as possible. In this case, specific features of the geometry, such as corners and edges, can be explicitly identified to find a suitable correspondence. In this scenario, the user or mapping specifications can choose which materials, material transitions, or components of the theoretical geometry should be considered or excluded. Furthermore, knowing the type of transition in the measurement data can prevent incorrect correspondence assignments and consequently, incorrect registration.

[0067] For example, each registration between measurement data and theoretical geometry can also be performed in a non-rigid manner.

[0068] Furthermore, when creating a labeled segment, measurement data can be searched from a database for known geometrical elements (e.g., spirals). If a geometrical element is identified within the measurement volume, or a similar geometrical element within a predetermined range is identified, the knowledge of that theoretical geometry can be used for further analysis, for example, by assigning appropriate material labels to the grayscale range during pre-segmentation, or by adapting the associated theoretical geometry to that geometrical element. Alternatively or additionally, appropriate evaluation schemes can be automatically retrieved. In another example, objects can be automatically identified or named in the scene tree. Searching for known geometrical elements from the database can also be performed in other steps of method 100.

[0069] In another optional step 112, a specific material can be assigned to the homogeneous region, for example, by performing an analysis of image information associated with the region and present in grayscale values ​​within the initial measurement data. In this case, at least one of the at least two homogeneous regions has a microstructure.

[0070] In step 106, the local similarity of multiple spatially resolved image information is analyzed. In this case, for example, the variation process of the multiple spatially resolved image information can be analyzed. Alternatively or additionally, the local differences of the multiple spatially resolved image information can be analyzed. Local differences can be calculated more quickly and are more robust at multi-material transition zones compared to using the variation process. Based on the local similarity, the expected locations of material transition zones between the various components of the object can be determined. These expected locations of the material transition zones are the locations of the expected boundaries of the homogeneous regions determined in step 104.

[0071] Subsequently, in another step 108, the homogeneous regions are adjusted. To do this, the extent of each homogeneous region is altered so that the boundaries of each homogeneous region are positioned at the intended locations of the material transition regions. The intended locations of the material transition regions thus define the homogeneous regions in the object representation.

[0072] In another step 110, at least two homogeneous regions from the digital object representation are segmented. Step 110, together with steps 106 and 108, is referred to as the main segmentation. In this case, the homogeneous regions determined in the digital object representation are adjusted and demarcated from each other.

[0073] In step 110, information from other sensors may be used. When adjusting the position of the material transition zone, the surface information obtained using these sensors can be used to extend the material transition zone in that direction or to prevent the material transition zone from extending beyond the defined surface.

[0074] Subsequently, when it is determined in at least one step of the method that the extension range of at least one homogeneous region is less than the predetermined region extension range, theoretical image information can be determined based on the at least one homogeneous region in optional step 122. In this case, topological analysis can be performed on the redefined region. In this case, the image information in the new region is analyzed and, for example, checked to confirm, for example, which regions the grayscale values ​​are in or whether a specific texture can be detected. A material can be assigned to the new region through topological analysis. In this case, using information related to the connecting materials in the environment, i.e., using the topology of the region, the material in the new region can be assigned to adjacent regions.

[0075] In another optional step 124, the image information of the determined boundary region can be analyzed to identify erroneous portions of the material transition zone within the determined boundary region. The accurate location of the material transition zone is determined in this case during the segmentation in step 110. The accurate location of the material transition zone ensures that the analysis of the image information in step 124 searches the surface at the correct location. When an erroneous segmentation is determined in step 110, that is, when the erroneous location in step 110 has been determined for a portion of the material transition zone in step 124, there is a high probability that these locations will be found in step 124. For this purpose, for example, a multilateral search method or other model-based methods can be employed.

[0076] In another optional step 126, erroneous portions of the material transition zone of the determined boundary region can be corrected after step 124. For this purpose, for example, if step 124 determines that a separate new region exists within a previously incorrectly segmented material region, a new material region is introduced.

[0077] In another optional step 128, heterogeneous regions can be obtained in the object representation, i.e., in the measurement data. These heterogeneous regions may, for example, appear in the produced object during an additive manufacturing process when, for instance, the powder is not properly or only partially melted during additive manufacturing. In this way, the object may have powder inclusions that form heterogeneous regions within the object in the measurement data. These heterogeneous regions typically form microstructures, which may be formed within homogeneous regions.

[0078] In another optional step 130, spatially resolved image information of at least one heterogeneous region may be analyzed. This operation is performed to determine the texture in the at least one heterogeneous region. In particular, when the heterogeneous region contains unmelted powder, the heterogeneous region may instead have a texture based on powder particles. The texture may also be due to fibers in the material represented by the heterogeneous region.

[0079] In another optional step 132, the at least one heterogeneous region is segmented. In this case, the heterogeneous region is defined from surrounding homogeneous regions or potentially other heterogeneous regions. In this case, a material can be assigned to the heterogeneous region, as in step 110, wherein the assignment is based on image information within the heterogeneous region.

[0080] Figure 2 An optional sub-step 120 of step 104 is shown. In this case, sub-step 120 includes obtaining multiple edges between at least two homogeneous regions.

[0081] Figure 3 An example of a multi-material transition region with multiple edges is shown. In this case, Figure 3 Materials 48, 58, and 56 are shown. In this case, material 48 is arranged between materials 58 and 56 and has a much smaller extension range compared to the other two materials. Material transition region 52 is arranged between materials 48 and 58. Material transition region 50 is arranged between materials 48 and 56. Overall, these two material transition regions 50 and 52 form a multi-material transition region, which is difficult to distinguish using conventional methods. Typically, conventional segmentation methods would detect such a region as only one material transition region. However, using the computer-implemented method 100 of the above invention, the identification of multiple material transition regions that are very close to each other can be performed.

[0082] Figure 4a and Figure 4b The optional step 112, already described above, is explained in detail. A digital representation 10 of the image information, representing a cross-sectional view of the object, is shown here. In this case, the object has sections 12, 14, and 16. Section 12... Figure 4aThe middle section is separated from the partition 14 by the material transition zone 20. The partition 16 is separated from the partitions 12 and 14 by the material transition zone 21.

[0083] The microstructures are arranged in region 23, which is drawn with dashed lines, but they are in... Figure 4a It is indistinguishable in the text and therefore not shown.

[0084] An analysis was performed on region 23 and partitions 12, 14, and 16 by examining the topology. In this example, the analysis showed that region 23 is separated from partitions 12, 14, and 16. Therefore, according to... Figure 4b This forms a new region as partition 25, which is arranged between partitions 12, 14, and 16. This new region here represents a microstructure with an extension range of 27 or 29, both of which are smaller than the predetermined extension range.

[0085] Alternatively, when analysis of the topology indicates that region 23 is connected to one or more of partitions 12, 14, and 16, the new region may be merged with one of partitions 12, 14, or 16. The new region may also be distributed among multiple partitions. In this case, multiple portions of the new region 25 are distributed among partitions 12, 14, and / or 16.

[0086] The following uses Figures 5a to 5e The optional steps 116 and 118 of method 100, as well as several other steps, are described in detail, illustrating the use of labeled fields in relation to method 100. In this case, Figure 5a The illustration schematically shows a digital representation 10 of image information from measurement data of a local region of an object. This schematic digital object representation could, for example, be a cross-sectional view obtained from a computed tomographic X-ray radiograph. In this case, the image information could be grayscale values, for clarity reasons. Figure 5a It is not shown as a grayscale value. Only the transition area where the grayscale value changes significantly is shown as a line.

[0087] The object has partitions 12, 14, 16, and 18, whose image information forms homogeneous regions. Partition 12 is defined relative to partition 14 by material transition region 20. Partition 12 is defined relative to partitions 16 and 18 by material transition region 22. Material transition region 24 is located between partitions 16 and 18. However, transition regions 26, 28, and 30 can also be seen in the digital representation 10 of the image information, but these are not material transition regions and are caused by shadows or other artifacts.

[0088] In this case, traditional algorithms have problems distinguishing transition regions 26, 28, and 30 from material transition regions 20, 22, and 24. Therefore, optional pre-segmentation can be performed first, in which image information is analyzed.

[0089] in this case, Figure 5b The grid used as the marker segment 32 is used as an example. Figure 5a The image information is represented by 10. The marker segment 32 can have any ideal resolution and may be coarser than voxels or pixels, with voxel / pixel accuracy or sub-voxel / sub-pixel accuracy. The marker segment 32 and / or the distance segment have the same structure and resolution as the measurement data in most cases. However, for example, a lower resolution and thus a larger cell can be chosen, or an anisotropic resolution and thus a rectangle instead of a square can be chosen. In addition, the structure can be adjusted, for example, using a tetrahedron instead of a cube. Furthermore, it is not necessarily necessary to use one or more distance segments to represent the material transition region with sub-voxel accuracy. This may only be necessary when or after the location of the material transition region has been determined. Therefore, by processing only the marker segment during segmentation and using the distance segment only when the location of the material transition region has been determined, computation time and storage bits can be saved.

[0090] When image information is, for example, grayscale values, grayscale values ​​below a certain threshold can be assigned to a first material, such as air, which... Figure 5b This is indicated by the symbol "o". Gray values ​​above another threshold are assigned to the second material, which... Figure 5b The grayscale value between the two thresholds is indicated by a "+". This grayscale value can be assigned to a third material, which... Figure 5b In Chinese, it is represented by "x".

[0091] This marker segment can be used in conjunction with the distance segment.

[0092] Furthermore, information about the various parts of an object derived from the theoretical geometry can be used, for example, in a connector with digital codes 1-9 to obtain information about the corresponding material. Therefore, areas of the same material can also be distributed among different parts of the object. In this way, the analysis of measurement data becomes clearer. Ideally, these areas are listed or represented according to a hierarchical structure already defined in the theoretical geometry.

[0093] Similarly, regions of the same material that are separated or unconnected within the marked area can also be automatically assigned to different parts.

[0094] In the next step, according to Figure 5c To determine the representation 34 obtained from the local similarity analysis of the image information. This could be, for example, a gradient representation. Material transition regions 20, 22, and 24 are clearly visible here. Transition regions 26-30 are not visible in this representation. However, unlike the representation 10 of the image information, the various regions of the object cannot be qualitatively distinguished from each other. That is, from the representation based on... Figure 5c The material of the partition cannot be inferred from the view.

[0095] This indicates that 34 is logically associated with the marked field segment 32, as shown in... Figure 5d As illustrated in the example. In this case, it is discernible that the homogeneous region is not defined by material transition zones 20, 22, and 24 in all parts. Therefore, during the main segmentation, the boundaries of the homogeneous region are shifted by remarking the homogeneous region, for example, changing "o" to "x" at arrows 36 and 40, and changing "+" to "x" at arrow 38. The area marked with "o" at arrow 36 or 40 is... Figure 5e The region marked "x" disappears and now belongs to the region marked "o". At arrow 38, the region marked "+" shrinks, and the region marked "o" expands. Similar situations occur at arrows 42, 44, and 46. At arrows 46 and 44, the two previously separated homogeneous regions marked "+" expand together, where the region marked "x" disappears.

[0096] Alternatively or additionally, to create marked regions, areas belonging to a single material can be identified in the digital object representation. These markings are intelligently and automatically extended to the next material transition zone. It is also possible that a material transition zone can be specified by the user and automatically expanded until that transition zone, for example, merges with other material transition zones, so that the user does not need to specify the entire transition zone. Therefore, precise marking is not required. Furthermore, operations such as opening, closing, erosion, and expansion, inversion, Boolean operators, or smoothing tools (such as filters) can be used to process the regions within the marked regions.

[0097] Additionally, areas that, from the user's perspective, contain material transition zones can be highlighted. In this case, anchor points can be set where the processing can be performed as material transition zones and (arguably) as meta-information, or by directly altering image information in the representation of local similarity.

[0098] Alternatively, erroneous material transition zones can be removed or weakened. After processing, the marked segment is recalculated based on this. In this case, a warning can also be issued if a meaningful material transition zone cannot be found at the user-defined location.

[0099] Alternatively, surface-based determination of local data quality can be employed. In this case, each material transition zone can be assigned a quality value representing the precision of the material transition zone.

[0100] Local similarity representations can be calculated from measurement data, especially volume data, using various methods. For example, the Sobel operator, Laplace filter, or Cannibal algorithm can be employed. The choice of algorithm and its parameterization can be performed manually by the user. Therefore, for example, the algorithm that produces the best results when creating the labeled domain segment can be selected based on a preview image. Furthermore, the representation of local similarity can be processed with filters before adjusting the labeled domain segment to obtain the best possible results. One example is using a Gaussian filter to minimize the adverse effects of noise on the results during labeled domain segment adjustment.

[0101] Depending on the algorithm, even after adjusting the labeled domain segments, smaller regions may be incorrectly segmented. To eliminate this, sub-steps can optionally be performed.

[0102] In this case, shape operators, such as open and / or closed, can be applied to each material region to remove some small areas.

[0103] Furthermore, associated regions smaller than the predetermined maximum value can be deleted and assigned to surrounding materials. Alternatively, a different or larger maximum value can be assigned to regions surrounded by two or more other materials, or even the region can be left undeleted while regions surrounded by only one other material are treated with the aforementioned maximum value. In this way, a thin layer of material between the other two materials can be maintained.

[0104] Figure 5e The result of the main segmentation is shown here. Here, the boundaries of the marked domain segments roughly coincide with the material transition regions 20, 22, and 24. These portions, or materials 12, 14, and 16, are thus segmented.

[0105] Based on the adjusted marker domains, material transition zones, which can represent local surfaces, can be calculated with higher precision. Another algorithm specifically designed for this purpose can be employed. In this case, the precise location of the material transition zone is searched within a small circular region, such as a small voxel. This is typically a prerequisite for accurate dimensional measurements based on CT data.

[0106] In principle, different algorithms can be used for this purpose. For example, algorithms that directly operate on the measurement data. These can determine the local location of the surface by means of local or global thresholds, or by searching for the maximum gradient or inflection point of the gray value distribution.

[0107] Alternatively, the precise local location of a material transition region can be determined, for example, in a representation of local similarity or a display of gradients or changes, by adapting a quadratic polynomial to a grayscale distribution. The location of the extrema of this polynomial can then be used as the location of the surface.

[0108] However, the above explanation does not exclude other algorithms.

[0109] From the labeled domain segments and the representation implicitly stored therein, one can deduce perhaps the approximate surface normal direction of the surface in the material transition zone, or the material in the material transition zone. This can be used by several algorithms to obtain accurate results. Alternatively, if available, this can be obtained from theoretical geometry such as a CAD model.

[0110] This is then combined with an algorithm that may require or utilize information about the starting surface to calculate the exact location of the surface.

[0111] In addition, cone-beam artifacts, sampling artifacts, and noise can be reduced before or after creating the labeled domain segment.

[0112] The present invention is not limited to one of the aforementioned embodiments, but can be modified in a variety of ways.

[0113] All features and advantages derived from the claims, description and drawings, including structural details, spatial arrangement and method steps, are essential to the invention, whether individually or in multiple combinations.

Claims

1. A computer-implemented method for segmenting measurement data from an object measurement, wherein, The object has at least one material transition region, wherein a digital object representation comprising the at least one material transition region is generated using the measurement data, wherein the digital object representation has a plurality of spatially resolved image information relating to the object, wherein the method (100) has the following steps: - determining (102) the measurement data, wherein the measurement data has at least one microstructure having an extension which is less than a predetermined extension; - aligning (114) a digital representation of a theoretical geometry with the digital object representation; - determining (104) at least two homogeneous regions in the measurement data and / or in the digital object representation, wherein at least one of the at least two homogeneous regions has a microstructure, wherein the determination (104) of the at least two homogeneous regions is performed on the basis of the digital representation of the theoretical geometry; - analyzing (106) a local similarity of the plurality of spatially resolved image information; - adjusting (108) the extension of the homogeneous regions until at least one boundary region of the homogeneous regions is arranged at an expected position of a material transition region, wherein the expected position of the material transition region between various different components of the object is determined from the local similarity; - segmenting (110) the digital object representation on the basis of the adjusted homogeneous regions.

2. The method of claim 1, wherein, The at least one material transition region is a multi-material transition region.

3. The method according to claim 1 or 2, characterized in that, in After the step of determining (104) the at least two homogeneous regions in the measurement data and / or in the digital object representation, a predetermined material is assigned (112) to the at least two homogeneous regions, at least one of the at least two homogeneous regions having a microstructure.

4. The method of claim 3, wherein, When the extension of at least one of the at least two homogeneous regions is determined to be less than the predetermined extension in at least one of the steps of the method, a predetermined material is assigned (112) to this homogeneous region of the at least two homogeneous regions on the basis of a topological structure of this homogeneous region.

5. The method of claim 1 wherein, The analysis (106) of the local similarity is based on a change process of the plurality of spatially resolved image information and / or on a local difference of the plurality of spatially resolved image information.

6. The method of claim 1, wherein the step of After the step of determining (104) the at least two homogeneous regions in the measurement data and / or in the digital object representation, the method (100) has the following steps, wherein at least one of the at least two homogeneous regions has a microstructure: - creating (116) a marker field which defines the homogeneous regions by means of spatially resolved marker values in the measurement data and / or in the digital object representation; wherein the spatial resolution of the marker field is higher than the spatial resolution of the measurement data.

7. The method of claim 1, wherein the step of After the step of determining (104) the at least two homogeneous regions in the measurement data and / or in the digital object representation, the method (100) has the following steps, wherein at least one of the at least two homogeneous regions has a microstructure: - creating (116) a marker field which defines the homogeneous regions by means of spatially resolved marker values in the measurement data and / or in the digital object representation; wherein the spatial resolution of the marker field is higher than the spatial resolution of the measurement data. - creating (118) a label field segment, which defines the homogeneous region by means of the spatially resolved label values in the measurement data and / or in the digital object representation; wherein, at least in a predetermined area of the object representation, the spatial resolution of the label field segment is higher than the spatial resolution of the measurement data, and wherein, in the remaining area of the object representation, the spatial resolution of the label field segment is at most as high as the spatial resolution of the measurement data.

8. The method of claim 1 wherein, The step of determining (104) at least two homogeneous regions has the following substeps: - obtaining (120) a plurality of edges between the at least two homogeneous regions.

9. The method of claim 1 wherein, The method (100) has the following further steps: - determining (122) theoretical image information from at least one homogeneous region, when it is determined in at least one of the steps of the method that the extension of the at least one homogeneous region is smaller than a predetermined area extension.

10. The method of claim 1 wherein, The method (100) has the following further steps after the step of segmenting (110) the digital object representation: - analyzing (124) the image information relating to the determined boundary region in order to identify erroneous portions of the material transition region in the determined boundary region; - correcting (126) the erroneous portions of the material transition region of the determined boundary region.

11. The method of claim 1 wherein, The method (100) has the following further steps: - obtaining (128) at least one inhomogeneous region in the object representation; - analyzing (130) spatially resolved image information relating to the at least one inhomogeneous region in order to determine a texture within the at least one inhomogeneous region; - segmenting (132) the at least one inhomogeneous region by means of the texture.

12. The method of claim 11, wherein, The at least one inhomogeneous region has measurement data relating to a powdery region of the object.

13. The method of claim 11 or 12, wherein, The at least one inhomogeneous region is treated in further analysis as a region having properties of a material different from the material of the inhomogeneous region.

14. A computer program product having instructions executable on a computer, which, when the instructions are executed on the computer, cause the computer to perform the method according to any one of claims 1 to 13.

Images