Method for determining optimized projection views for sharp object edge imaging of an object to be inspected using robotic computed tomography

Abstract

A method is provided for determining optimized projection views for sharp object edge mapping of an object to be inspected using robotic computed tomography. The following steps are performed: determining a projection view in 3D Radon space, performing a collision check, and determining a consensus set as a subset of the points in 3D Radon space, where each point in 3D Radon space has a predefined weight. If the sum of the weights of the currently determined consensus set is higher than a value of one of the best current consensus sets, the currently determined consensus set is included in the best consensus sets, and the consensus set with the lowest value is removed.Upon reaching the termination criterion, the iterations are stopped, and the highest-scoring projection views are selected. A local check is then performed to determine the remaining pixel intensity in the resulting X-ray projection images from the projection views of the best consensus sets. The consensus set with the highest total score is selected and saved as the projection view to be used.

Description

[0001] The invention relates to the field of non-destructive testing of objects using robotic computed tomography.

[0002] Computed tomography (CT) is a non-destructive testing method used in the automotive industry. High-resolution scans of large objects are not possible with conventional systems due to the limited space available in the X-ray cabinet. Robot-based computed tomography, also known as RCT or robotic computed tomography, enables the acquisition of high-resolution scans of large objects. The RCT method uses two cooperating industrial robots equipped with X-ray hardware as end effectors, allowing for the implementation of multiple projection views and thus trajectories. A projection view describes the positions and orientations of the radiation source and detector for a given scanning geometry. A trajectory describes a set of multiple projection views.The paths between projection views are not part of the trajectory. Due to the six degrees of freedom of the RCT method, it allows for more flexible trajectory shapes than conventional systems, which can significantly improve the resulting data quality in scans of regions with limited accessibility.

[0003] In practice, situations frequently arise where standard trajectories, such as circular or spiral trajectories, cannot be fully executed due to limited accessibility. Consequently, the resulting datasets exhibit significant artifacts. In addition to the potential loss of information due to accessibility limitations, conventional trajectories used in real-world RCT applications neglect material-specific attributes within the study area during manual trajectory planning, which can promote the formation of metal artifacts.

[0004] Optimization methods for RCT trajectories can be divided into task-specific methods (test task known) and task-unspecific methods (test task unknown). Task-unspecific optimization methods attempt to scan a region of interest as completely as possible. Task-specific optimization methods for RCT trajectories attempt to find optimal projection views to best represent a predefined test task. Within the scope of this invention, a task-specific optimization method is described that maps the edges of a test object as sharply as possible.

[0005] Trajectory optimization methods typically refer to a set (a subset) of possible projection views, which is obtained by discretizing a circular, spherical, or spiral path with a constant radius in three-dimensional space. In the case of certain very challenging applications, existing trajectory optimization algorithms exhibit shortcomings. Depending on the degree of granularity in the discretization of the projection views, optimal views within the discrete subset of projection views may not be captured, thus escaping the algorithm's detection. Furthermore, accessibility and collision control are often neglected.

[0006] Therefore, a problem exists that when checking an object using robotic computed tomography, artifacts are generated in the recorded data for large objects due to limited accessibility (i.e., limited projection views) and possible excessive absorption of the X-ray beam, thus preventing optimal inspection of the object.

[0007] A solution to the problem can be found in the publication LINDE, M. [et al.]: Trajectory optimization for few-view robot-based CT: Transitioning from static to object-specific acquisition geometries. In: Tomography of Materials and Structures, Vol. 7, 2025, Article-No. 100058, pp. 1-14. ISSN 2949-673X, published after the filing date.

[0008] Further state of the art can be found in the publication LINDE, M. [et al.]: Selecting feasible trajectories for robot-based x-ray tomography by varying focus detector-distance in space restricted environments. In: Journal of Nondestructive Evaluation, Vol. 43, 2024, No. 2, Article No. 65, 16 pp. ISSN 0195-9298.

[0009] Further state of the art can be found in the publication HATAMIKIA, S. [et al.]: Source-detector trajectory optimization in cone-beam computed tomography: a comprehensive review on today's state-of-the-art. In: Physics in Medicine & Biology, Vol. 67, 2022, Article No. 16, 16TR03, 22 pp. ISSN 1361-6560.

[0010] Further state of the art can be found in documents DE 10 2016 213 403 A1 or DE 10 2010 022 285 A1.

[0011] It is an object of the invention to provide an improved method for determining optimized projection views for a testing task known prior to scanning using robotic computed tomography.

[0012] This task is solved by the features of independent claims. Advantageous embodiments are the subject of dependent claims.

[0013] A method is provided for determining optimized projection views for sharp object edge mapping of an object to be inspected using robotic computed tomography. In the first step, surface data of the object is processed using a 3D Radon transformation, so that for each surface in spatial space a point is created in 3D Radon space. The following steps are repeated until the number of selected projection views corresponds to the number of desired projection views. In the second step, the following substeps are repeated up to a predefined termination criterion and executed using a RANSAC algorithm: Determining a projection view in 3D Radon space by modeling a sphere from an arbitrary selection of data points in 3D Radon space.which intersects the center of the test region (of the object) and which has a diameter range, performing a collision check in which collision-free accessibility of the resulting source and detector poses is verified, determining a consensus set as a subset of the points in 3D Radon space whose distance to the spherical cap within the 3D Radon space is less than a predefined tolerance distance, where each point in 3D Radon space has a predefined weight, and in the case that the sum of the weights of the consensus set just determined has a higher value than a value of one of the currently best consensus sets, the consensus set just determined is included in the best consensus sets and the consensus set that has the lowest value within the currently best consensus sets is removed,Upon reaching the termination criterion, the repetitions are stopped, and the projection views (consensus sets) with the highest scores from the second step are passed to a third step. In the third step, a projection check is performed, i.e., a local inspection of the remaining pixel intensity in the resulting X-ray projection images of the projection views from the best consensus sets from the second step is carried out in the areas that reference sampled planes. The consensus set with the highest total score based on the weighting is selected and saved as the projection view to be used. In a fourth step, the weighting of the points in the radon space belonging to the selected projection view is reduced. Thus, a new weighting is applied.

[0014] This allows optimized projection views to be determined directly, without having to define multiple discrete, predefined reference geometries (projection views).

[0015] One implementation proposes assigning lower weights to the points in the 3D Radon space of already selected projection views in order to determine further optimized projection views. This ensures the novelty of the information.

[0016] In one implementation, the termination criterion is a predetermined number of repetitions or a predetermined time period for executing the procedure. In another implementation, the predetermined number of repetitions is specified by the user. This optimizes the procedure in terms of both time and quality.

[0017] One embodiment provides that the diameter of the spherical domes created in each projection view within the radon room is selected within a predetermined range, ensuring that a specified exposure time is not exceeded and that the entire area of ​​interest is covered by the emitted radiation. This guarantees that the entire object is covered and that the required exposure time is not excessively long.

[0018] In one embodiment, it is provided that, to eliminate metal artifacts, a value is incorporated into the weighting by analyzing the X-ray attenuation, and by projecting each scanned point in the 3D radon space onto a 2D detector using projection matrices that map the point in the 3D radon space onto the 2D detector, whereby the remaining intensity is determined within a predefined range and, if this falls below a predefined value, the point in the 3D radon space receives the value zero and thus does not contribute to the information gain.

[0019] In one embodiment, to ensure accessibility to the source and detector, an accessible space for the radiation source and detector is determined. If the determined projection view lies within this space, it is classified as usable; if it lies outside the space, it is discarded. In another embodiment, if the second substep determines that the detector position is not reachable without collision, the object-detector distance is iteratively reduced until the detector is reachable without collision.

[0020] In one implementation, a trajectory is compiled from several selected projection views, which the robot then follows to inspect the object.

[0021] Furthermore, a computer program product is provided which can be directly loaded into a memory of a programmable computing unit, comprising program code means to execute a method according to one of the preceding claims when the computer program product is executed in the computer.

[0022] Furthermore, a device for inspecting an object using robotic computed tomography is provided, comprising two controllable robots positioned opposite each other over the object to be inspected. One of the robots is equipped with a radiation source for emitting X-rays, and the other robot is equipped with a detector for receiving X-rays that have passed through the object. The robots are controlled in their movement such that they follow a trajectory determined according to the described method. That is to say, they have a control unit configured to execute the method.

[0023] Further features and advantages of the invention will become apparent from the following description of exemplary embodiments of the invention, with reference to the figures in the drawing, which shows details of the invention, and from the claims. The individual features can be implemented individually or in any combination in a variant of the invention.

[0024] Preferred embodiments of the invention are explained in more detail below with reference to the accompanying figures. Fig. Figure 1 shows a cross-sectional image of a cone-shaped X-ray beam in a projection view in the spatial domain in 2D view. Fig. 2 shows a cross-sectional view of one from the in Fig. 1 shown cone-shaped X-ray beam resulting representation in the 3D radon room. Fig. Figure 3 shows a cross-sectional view of a schematic 3D radon room representation for carrying out the method according to an embodiment of the present invention. Fig. Figure 4 shows a schematic flowchart of the method according to one embodiment of the present invention.

[0025] In the following figure descriptions, identical elements or functions are marked with the same reference symbols.

[0026] The following describes an improved method for determining optimized projection views (for determining the optimal positioning of a radiation source and a radiation detector) for an object to be inspected using robotic computed tomography. This method can be used for the non-destructive examination of any object with clearly defined edges, particularly vehicle components and sections of entire vehicles. The advantage of this method is that the quality of the object's reconstruction (especially its edges) can be improved with fewer projections than previously required. The method can be used in both easily accessible and difficult-to-access areas of the object under investigation.As mentioned at the beginning, the procedure uses two cooperating industrial robots equipped with X-ray hardware (one robot with radiation source, the other with detector) as end effectors.

[0027] The proposed trajectory optimization offers the possibility of improving the quality of object reconstruction with fewer projections than previously required. Trajectory optimization refers to the selection of a specific subset of projection views (position of the radiation source, position of the detector, orientation of the radiation source, orientation of the detector) that, compared to other subsets of the same size, result in an improved reconstruction volume with fewer artifacts. The consideration of edges is particularly important in the field of industrial, non-destructive testing.

[0028] A key aspect of the method is that, for optimized reconstruction of an object's structures, X-rays strike the edges of the object under investigation as tangentially as possible. Instead of applying a quality criterion to a discrete input set of projection views, as was previously the case, highly informative projection views are used directly from the object's CAD data. The principle of the 3D Radon transformation is applied to identify the most optimal projection views, which can be freely positioned in space and are not subject to the limitations of pre-discrete acquisition points or fixed system distances.

[0029] A 3D Radon transformation integrates a function f (x, y, z), which describes a (to be verified and therefore known) 3D object, over planes defined by the normal vector η→ They are described and positioned at a distance c from the origin (center of rotation). The 3D Radon transformation is defined as follows: Rf(η→,c)=∫−∞+∞∫−∞+∞∫−∞+∞f(x→)δ(x→Tη→−c)dxdydz.

[0030] Geometrically, the 3D Radon transformation establishes a mapping in which each plane within the spatial domain uniquely corresponds to a point within the 3D Radon space, as shown in Fig. 1 and Fig. 2 shown.

[0031] Within the 3D Radon space, a cone-shaped projection (X-rays are emitted in a cone shape) forms the surface of a spherical cap (in the two-dimensional area circular cap) in the area of ​​interest (ROI) of the imaging.

[0032] The diameter of the sphere determined by the spherical cap corresponds to the distance from the (X-ray) source s to the origin (center) o of the object (also called focus-object distance - FOD). The minimum and maximum permissible FOD are determined by a CT specialist for each specific application.

[0033] The position of the source s of a corresponding cone-ray projection in 3D Radon space is determined by: S→=ctr→sphere‖ctr→sphere‖⋅FOD.

[0034] The direction of the detector position (which must be chosen corresponding to the position of the source s) is determined by the inversion of s→ The distance of the detector from the origin o is defined by the ODD (object-detector distance), which is specified by a ratio value that denotes the ratio of FOD and ODD, ensuring that the area of ​​interest (ROI) is fully displayed on the sensor surface and at high magnification.

[0035] In Fig. Figure 1 shows a spatial domain, which represents a cross-sectional image of a cone-ray projection in the region of interest (ROI). Fig. Figure 2 shows the resulting cross-sectional representation in the radon room. The area of ​​interest (ROI) describes the scan area of ​​the object to be examined. The following are shown in Fig. 1 and Fig. Two surfaces, 01-03, are located within the area to be recorded, and one surface, O4, is located outside of it. Projections in 3D Radon space form spherical caps, which are represented as circular caps in 2D space. Surfaces in the spatial domain are each represented as a point in Radon space and define both the orientation and position of surfaces 01-04. Hereafter, only the term Radon space is used, always referring to 3D Radon space. The figures are shown as 2D views (sectional views) for clarity only.

[0036] The proposed procedure is in Fig. 4 is shown schematically and is described below.

[0037] After the transformation of the object to be tested into the radon space in the first step S1 of the procedure, the goal is now to identify those spherical caps (in Fig. 3 dotted circles within the ROI) are used to identify points that intersect the origin o (center) of the object, meaning the central ray of the cone-shaped X-ray beam geometry always points towards the object's center. X-rays within the cone-shaped X-ray beam geometries correctly scan surfaces when they are tangentially grazed. The goal is to identify as many high-value points in the radon space as possible within an acceptable error distance from the surface of the spherical cap, as described in Fig. 3. An example of a projection view (position of the source s) is shown. The diameter d of the sphere (the circle in the 2D view) is chosen such that: d min < d < d max . d max The distance s of the source from the origin o (center of the object) is chosen such that it is not so great as to require an excessively long exposure time. The minimum distance d minis chosen so that the entire object can still be detected by the radiation (X-ray beam) emitted by the source s, i.e., the ROI is covered.

[0038] In Fig. Figure 3 shows an example of a multitude of points in the radon chamber, which are to be scanned by the X-rays emitted from the source s. As mentioned earlier, it is important that at least one of the cone-shaped X-ray beams strikes the surfaces (points in the radon chamber) as tangentially as possible. To achieve this and to require as few projections as possible, a so-called RANSAC (RANSAC = random sample consensus) algorithm is applied in a second step S2.

[0039] After a 3D Radon transformation has been applied to the known data of the object to be checked in a first step S1 of the procedure, in order to simplify data processing, data points in the 3D Radon space are arbitrarily selected in a first substep S21 of the second step S2 of the procedure. These points are needed to model a sphere (sphere; circle in 2D view) that intersects the origin o (center of the object) and has a diameter d, where: d min < d < d max These data points are used to determine the model parameters (sphere), i.e., to model the sphere that describes the projection with the corresponding source position s, as in Fig. 3 indicated.

[0040] In a second substep, S22, a collision check is performed to verify that the resulting source and detector poses are reachable without collision. The source pose is calculated as described in substep S21. The detector position can be determined using the ratio of the FOD (field of detection) and ODD (objective detection). The detector orientation must always be chosen so that the central beam from the source strikes the detector center orthogonally. Using a tool in the kinematic digital twin of the system, collision-free spaces for the source and detector robots can be calculated. A raycasting algorithm and the odd-even rule of intersections are then used to check whether the resulting projection views lie within these collision-free spaces.If collision control reveals that the detector position is not reachable without collisions, the ODD can be iteratively reduced in size to mark the projection view as reachable. Increasing the ODD is not possible due to the risk of an incomplete representation of the ROI on the detector. Changing the FOD (source robot) is not advisable, as this would alter the analysis of the information quality of the corresponding projection view within the radon chamber.

[0041] In a third substep, S23, a subset of points in the radon space is determined whose distance to the spherical cap (model curve; spherical segment within the ROI) is less than a predefined tolerance distance. This subset is called the consensus set. Each point in the radon space is assigned a value between 0 and 1, which varies depending on how frequently it has been sampled. The more frequently it has been sampled, the lower the value, which describes the novelty content of this point in the radon space. In the first execution of the procedure, i.e., the determination of the first projection view, all points are assigned the value 1. The more frequently they appear in a best projection view, the lower their value, i.e., their weighting, becomes. Thus, the novelty of the information can be ensured, as described below.

[0042] In a fourth substep S24, if the values ​​of the points in the Radon space (i.e., the sum of the weights) in the consensus set are higher than the values ​​of the currently best consensus sets (from previous iterations, if any), the consensus set is included within the current optimal consensus sets, and the least effective consensus set within the current best consensus sets (i.e., the one with the lowest value) is removed.

[0043] Steps S21-S24 are repeated until a predefined termination criterion A is met. The number of repetitions required for steps S21-S24 can be empirically determined by a user. A termination criterion A can also be a predefined time period. Step S2 ends upon reaching termination criterion A by transferring the projection views (consensus sets) with the highest ratings from step S2 to step S3.

[0044] In a third step, S3, a projection check is performed, i.e., a local verification of the remaining pixel intensity in the resulting X-ray projection images of the projection views of the best consensus sets from the second step, S2, in the areas that reference scanned planes. For this purpose, the scanned 3D points in the radon space are projected onto the 2D detector using 3x4 projection matrices, and within a radius around these points, it is ensured that the pixel intensity is greater than 10% of the maximum pixel intensity. If this is not ensured for all pixels within the radius, the weighting of the corresponding point in the radon space for this projection view is set to zero (no added information). The third step, S3, incorporates the X-ray projection-specific evaluation of the added information per scanned point in the radon space for the best projection views (consensus sets) from the second step, S2.The weights for each consensus set are then summed. The consensus set with the highest total is selected and saved as the projection view to be used.

[0045] In a fourth step, S4, a new weighting is applied; that is, the weighting of the points in the Radon space belonging to the selected projection view is reduced, as previously described. Thus, when optimizing the next projection view, these points no longer count as much as points that have not yet been sampled, and the novelty of the information is maintained.

[0046] Steps S2 to S4 are embedded in a loop. The loop iterates until the number of selected projection views equals the number of desired projection views.

[0047] A high-quality projection view exhibits the following characteristics: high information content, avoidance of metal artifacts, novelty of the information, and accessibility. High information content refers to the sum of the weights of the points scanned with a projection view in the radon space.

[0048] The high information content is achieved through the RANSAC algorithm, which allows optimized projection views (spherical caps) to be found in the radon space.

[0049] Furthermore, it is important to avoid metal artifacts. When X-rays pass through high-density metal objects, image quality and data interpretability are degraded by a variety of physical phenomena, primarily due to effects such as photon deficiency and beam hardening. Metal artifacts are characterized by dark and light fringes in the CT volume. Many approaches exist to reduce metal artifacts. The simplest methods filter out projection views with extremely long penetration lengths through the object under investigation. More sophisticated methods use projection-based algorithms to reduce the influence of metal artifacts by replacing metal-induced, corrupted projection views with interpolated data derived from adjacent, unaffected projection views.The proposed method analyzes X-ray attenuation by locally evaluating the contribution of high-information pixel regions to the formation of metal artifacts (in the best consensus sets). In this method, each identified point in the radon space is projected onto a 2D detector using projection matrices. x=P⋅X, where X = (X, Y, Z, 1) T the point in radon space in homogeneous world coordinates, P the 3x4 projection matrix which maps the point in radon space onto the 2D detector and x = (x, y, 1) TThe pixel position is represented in homogeneous detector coordinates. A region with a radius around x is then sketched, and the remaining intensity is determined to ensure it exceeds a predefined minimum intensity threshold, which might be, for example, 10%. If the threshold cannot be met, the point in the radon space is assigned the value 0 for this specific projection view. Summing the values ​​assigned to the points in the radon space from the best projection views yields the optimal projection view for the current iteration.

[0050] To reduce the repeated scanning of identical structures and the resulting projection clustering, all points in the radon space are assigned a weight, as previously described. Initially, the weight for all points in the radon space is 1. If a point in the radon space is scanned a second (or subsequent) time by the projection view (sphere cap), it receives a lower weight. Thus, projection views that include more newly scanned than previously scanned points in the radon space are weighted more heavily. This ensures that the object's areas are scanned more evenly and that the novelty of the information is preserved.

[0051] Furthermore, it is important to ensure the (collision-free) accessibility of the source and detector to the corresponding projection views. This is particularly important in areas of limited accessibility. For this purpose, a space for the source and detector can be determined in a kinematic digital twin of the system, within which projection views can be accessed without collisions. If a position of the source and / or detector lies in an area that is determined to be non-collision-free (i.e., inaccessible), the corresponding projection view is discarded.

[0052] The proposed method enables object-based identification of the optimal positions (sharpest edge imaging) for a source to emit X-rays, allowing for the non-destructive creation of a three-dimensional CT image of an object. The projection views can be directly derived from the data (surface data) of the object under inspection, e.g., from CAD data, without requiring the definition of multiple discrete, predefined acquisition geometries. Furthermore, the proposed method allows for flexible focus-object distances (FOD).

[0053] The described method allows for the identification of projection views with high information content in 3D Radon space using a modified RANSAC algorithm. This enables the determination of arbitrary projection views that are no longer restricted by discretized sets of predefined acquisition geometries. Suppression of metal artifacts is achieved through local attenuation-based quality control of high-information areas within the projection images. Additionally, projection views are checked for accessibility. The algorithm performs particularly well in scenarios where components with distinct edges, as is common with industrial parts, are inspected within areas with highly restricted access.

[0054] The described method is advantageously implemented as a computer program that can execute the steps of the procedure when run on a computer. The computer can be one (or more) control units of one or more robots used to perform the CT scan. Naturally, the robots serving as source and detector are interconnected (i.e., they operate collaboratively) such that if the robot serving as the radiation source is moved, the robot carrying the detector is moved to a position opposite the robot serving as the radiation source (in order to receive the radiation).

[0055] An object that can be inspected using robotic computed tomography can be a vehicle component, in particular a part of the body, or the entire body. Depending on the type of object to be inspected, a suitable trajectory is determined using the described method, specific to the object or application.

Claims

[1] Method for determining optimized projection views for sharp object edge imaging for an object to be inspected by robot computed tomography, wherein - in a first step (S1) surface data of the object are processed using a 3D Radon transformation, so that for each surface in local space a point is created in 3D Radon space, the following steps are repeated until the number of selected projection views equals the number of desired projection views, wherein - in a second step (S2) the following sub-steps are repeated up to a specified termination criterion (A) and executed using a RANSAC algorithm: - (S21) Determining a projection view in the 3D Radon space by modeling a sphere from an arbitrary selection of data points in the 3D Radon space, which intersects the center of the object and has a diameter range, and - (S22) Performing a collision check to verify that the resulting source and detector poses are reachable without collision, - (S23) Determining a consensus set as a subset of the points in 3D Radon space whose distance to the spherical cap within 3D Radon space is less than a given tolerance distance, where each point in 3D Radon space has a given weight, and - (S24) if the sum of the weights of the consensus set just determined has a higher value than one of the currently best consensus sets, the consensus set just determined is included in the best consensus sets and the consensus set with the lowest value within the currently best consensus sets is removed, and - Upon reaching the termination criterion (A), the repetitions are terminated and the projection views with the highest ratings in the second step (S2) are passed to a third step (S3), whereby - in the third step (S3) a local check of a remaining pixel intensity in resulting X-ray projection images of the projection views of the best consensus sets from the second step (S2) is carried out in the areas which reference sampled planes, whereby the consensus set which has the highest sum due to the weighting is selected and saved as the projection view to be used, and - in a fourth step (S4) the weighting of the points in the radon space that belong to the selected projection view is reduced. [2] Method according to claim 1, wherein, to determine further optimized projection views, the points in the 3D Radon space of already selected projection views are given a lower weighting. [3] Method according to claim 1 or 2, wherein the termination criterion (A) is a predetermined number of repetitions or a predetermined time period for carrying out the method. [4] Method according to claim 3, wherein the predetermined number of repetitions is predetermined by a user. [5] Method according to one of the preceding claims, wherein the diameter of the spherical caps formed in the radon room for each projection view is selected within a predetermined range, whereby a predetermined exposure time is not exceeded and whereby the entire area of ​​interest is covered by the emitted radiation. [6] Method according to one of the preceding claims, wherein a value is included in the weighting to eliminate metal artifacts by analyzing the X-ray attenuation, and by projecting each point identified as scanned in the 3D radon space onto a 2D detector using projection matrices which map the point in the 3D radon space onto the 2D detector, wherein the remaining intensity is determined within a predetermined range and, if it falls below a predetermined value, the point in the 3D radon space is assigned the value zero. [7] Method according to one of the preceding claims, wherein in the second substep (S22) an accessible space for the radiation source and the detector is determined to ensure the accessibility of the source and the detector, and if the determined projection view is within the space, this projection view is classified as usable, and if it is outside the space, this projection view is discarded. [8] Method according to one of the preceding claims, wherein in the second substep (S22) it is determined that the detector position cannot be reached without collision, an object-detector distance is iteratively reduced until the detector can be reached without collision. [9] Computer program product which can be loaded directly into a memory of a programmable computing unit, comprising program code means for carrying out a method according to any of the preceding claims when the computer program product is executed in the computer. [10] Device for checking an object by means of robot computed tomography, comprising two controllable robots which are positioned opposite each other over the object to be checked, wherein one of the robots is equipped with a radiation source for emitting X-rays and the other robot with a detector for receiving X-rays which have passed through the object, and additionally comprising a control unit which is configured to carry out a method according to any one of claims 1 to 8.

Images