DE102020130442A1 - Calibration device - Google Patents

Abstract

The invention relates to a calibration device and a method for determining the respective position of components (308, 309, 311) of a computer tomograph (300) to one another, the respective position including at least the respective relative positions of the components (308, 309, 311) to one another , and wherein at least one laser interferometric structure (313-316, 333-336) is arranged in or on the computer tomograph (300) for determining the respective position.

Description

For the exact measurement with a computer tomograph, in particular for the exact geometric measurement of dimensional features on workpieces with a computer tomograph, the exact knowledge of the arrangement of the components radiation source (source), radiation detector (detector) and turntable for receiving and rotating the workpiece to be examined is related on their position (position and angular position) relative to each other. The geometry is described, for example, by so-called SOUV vectors, which are taken into account in the reconstruction of the radiographic images recorded, whereby separate SOUV vectors can also be assigned to each radiographic image, for example if the geometry changes due to recognized drift phenomena during the recording of the multiple radiographic images. The geometry includes, among other things, the image scale (or the magnification) of the image as an important parameter of the measurement, which also results from the ratio FDD (focus to detector distance) to FOD (focus to object distance), i.e. the position of the workpiece ( and thus the axis of rotation of the turntable) between the source and detector, more precisely between the focal point (focus) of the source emitting the measurement radiation and the detector. For the highest possible flexibility, the corresponding devices are equipped with movement axes in order to set different configurations as required, such as different imaging scales, different cone angles of the measurement radiation from the source detected by the detector (distance from source to detector) or the detection of one or more workpiece details in areas. For this purpose, the components are moved towards one another. For exact measurements, these movements must be known exactly in order to know exactly the geometry that is present after the movement has been completed. Therefore, measuring axes are used for this purpose, the movements and movements of which must be measured in relation to one another.

To determine the geometry of a computer tomograph, indirect methods have so far been used as far as possible. To determine the image scale, for example, a calibration object is irradiated and the image on the detector is related to the known dimensions of the calibration object or a displacement of the calibration object on the detector is related to a known displacement of the calibration object. The WO2005119174A1 the applicant describes, for example, a method with a 4-ball standard as a measuring body. The DE102008044437 the applicant describes a similar method that is implemented directly on the workpiece to be measured. Also the DE102014113977 the applicant uses a calibrated reference body to determine the magnification (or image scale). Corresponding methods which can be combined with the following inventive idea are also described in the present application.

In the calibration models available in this way, in particular the angular positions of the components are not fully taken into account. In most cases, only the one angular position present during the calibration is included in the calibration data, without the angles themselves being recorded and locally changing angles being taken into account in the calibration model. The resulting deviations are hardly relevant if the measuring field of the sensor is small. In the case of large detectors, such as image processing systems with large image fields and computer tomographs, however, there are significant deviations. In the context of the present invention, these angular positions, that is to say rotations, of the components are to be taken into account in the calibration model.

Furthermore, various methods for calibrating measuring axes, i.e. for directly determining the deviations of the measuring axes from coordinate measuring machines (also referred to as geometric errors, such as pitch errors, straightness, squareness, etc.) are known. These are devices with tactile or optical sensors in which the workpiece and the sensor are moved relative to one another in two or three spatial directions. Depending on the design, measuring devices are provided for this purpose, which move only the workpiece or only the sensor in two or three spatial directions, or the sensor is moved in one or two directions and the workpiece in the remaining spatial directions. The sensor is usually attached to a so-called quill and the workpiece rests on a measuring table. Both start from a common, fixed base, depending on the arrangement of the measuring axes, fixed or movable.

For the determination of the geometric errors of the aforementioned coordinate measuring machine, see EP1990605 Laser interferometer, which determines the distance between a transmitter arranged on the base or the measuring table and a reflector arranged on the quill in different positions of the measuring axes. The construction is very complex and requires several of the expensive interferometer arrangements.

The DE102007004934 alternatively describes a so-called laser tracker, the laser beam direction of which is variably adjustable in order to enable detection of a reflector, for example arranged on the quill, in different measuring directions and thus in a much more flexible manner. The disadvantage is that the process is very complex, since the Transmitter must be implemented several times to record the complete geometric error.

The EP2390737 describes a corresponding method for determining the geometric errors in which the transmitter of the laser interferometer is arranged on the spindle of a machine tool, and reflectors are detected that emanate from the stationary base.

Another alternative solution is shown by the DE102016113531 . With only one transmitter and one evaluation device, several reflectors are recorded sequentially, in that separate outputs of the transmitter can be controlled by a multiplexer (fiber coupler). Each measuring section has its own reference section integrated for this purpose. The targeting of the respective reflector is carried out with the respective controlled output. Coordinate measuring machines are used with a reflector to be converted or having several reflector elements or several fixed reflectors which are arranged on the quill and arrangement of the outputs on the measuring table or the base. The method is for example in the DE102017103455 described.

An exactly opposite principle describes the WO2015079019 . The transmitter is located on the quill and the reflectors are arranged on the measuring table or the base.

The disclosure of the aforementioned methods and arrangements according to the prior art is to be understood as part of the following inventive idea. However, these methods and devices all have the following limitation in common. Only two components are moved towards each other, namely the workpiece to be measured and the sensor used to detect the workpiece or the quill to which it is attached. To determine the geometry of a computer tomograph, however, the deviations in the movement (pitch, tilt, straightness, squareness, etc.) of three components (source, detector and turntable) relative to one another are always necessary. A particular challenge is that one of the components is arranged on one side and one on the other side of the workpiece (or rotary table). The source emits radiation that the detector receives. In addition to the position of the source and detector in relation to one another and in relation to the workpiece, the angular positions must also be observed, which would lead to imaging errors if ignored. These deviations are contained in the rectangularity in classic coordinate measuring machines, but must be recorded and corrected separately in the case of a computer tomograph. The correct detection of the imaging scale must be carried out for different configurations, in particular for different imaging scales.

A further object of the present invention is therefore to determine the geometry of a computer tomograph (CT geometry) precisely, quickly and inexpensively, in particular for several configurations, i.e. positions of the components source, detector and turntable relative to one another, in particular imaging scales, cone angles of the detector detected measuring radiation of the source (distance source to detector) and / or necessary positions for the area-by-area detection of one or more workpiece details. The measuring axes provided for positioning should advantageously not be measured before each computed tomographic measurement of a workpiece, but for example only once when the computed tomograph is started up or at fixed intervals, for example after a certain time or number of measurements. The devices required for this should then be usable for other devices or other purposes in order to save costs. This should also reduce the set-up time and the calibration time.

To solve this, the invention provides that the determination of the CT geometry of a computer tomograph is carried out by using at least one laser interferometric structure arranged in or on the computer tomograph, the respective positions of the components source, detector and turntable on a base and / or to one another are recorded. The CT geometry is preferably recorded for one or more configurations, in particular imaging scales. The laser interferometric structure is preferably used only when the computer tomograph is put into operation and / or at fixed intervals and can be used for another purpose in the meantime.

The laser interferometric arrangements known from the prior art are used by arranging the two components transmitter and reflector on the base or on the components. Several laser interferometric arrangements can also be used at the same time. There are a total of several possible combinations. A first possibility is to attach the transmitter or reflector to the base and the respective counterpart (transmitter or reflector) one after the other or at the same time to one of the components source, detector, turntable. The second possibility provides that the two components transmitter and reflector are arranged on two of the components source, detector, turntable and two of the three possible combinations (source-detector, source-workpiece, workpiece-detector) are realized. Here, the measurement effort is low overall, but the constructive implementation is more complex.

1. 1. Strahlungsquelle, insbesondere Röntgenstrahlungsquelle (Quelle)
2. 2. Strahlungsdetektor, insbesondere Röntgenstrahlendetektor (Detektor)
3. 3. Drehtisch zur Aufnahme und Drehung des zu untersuchenden Werkstücks,

und wobei die jeweilige Lage zumindest die jeweiligen Relativpositionen der Komponenten zueinander und vorzugsweise die jeweilige Winkellagen der Komponenten zueinander umfasst, die sich dadurch auszeichnet, dass für die Bestimmung der jeweiligen CT-Geometrie zumindest ein laserinterferometrischer Aufbau im bzw. am Computertomografen angeordnet ist, wobei die jeweiligen Lagen der Komponenten in Bezug zu einer feststehenden Basis (Basis) und/oder relativ zueinander erfasst werden.A device according to the invention therefore comprises a calibration device for determining the geometry (SOUV vectors) or several different geometries of a computer tomograph (CT geometry), i.e. the respective position of the components of the computer tomograph with respect to one another, the components being:

1. 1. Radiation source, in particular X-ray source (source)
2. 2. Radiation detector, especially X-ray detector (detector)
3. 3. Turntable for picking up and rotating the workpiece to be examined,

and wherein the respective position comprises at least the respective relative positions of the components to one another and preferably the respective angular positions of the components to one another, which is characterized in that at least one laser interferometric structure is arranged in or on the computer tomograph for determining the respective CT geometry, the respective positions of the components are recorded in relation to a fixed base (base) and / or relative to one another.

In particular, the invention is characterized in that one, two or all three components start from a drive unit that can be displaced in one, two or three spatial directions, preferably start from measuring axes, the drive unit or the respective measuring axis moving the respective component relative to one another causes on the base.

As a rule, it is not necessary for all three components to be displaceable in all three spatial directions. For reasons of redundancy, it is sufficient if a maximum of two of the components can be moved in each spatial direction.

The invention is preferably characterized in that the laser interferometric structure has at least one unit (transmitter) that emits at least one measuring beam, at least one unit (reflector) that reflects or deflects a measuring beam, such as a plane mirror or triple mirror, and an evaluation unit, preferably in the transmitter or on a the device receiving the transmitter at least one reference path for generating interference to be processed by the evaluation unit with the measuring beam is arranged.

According to a proposal to be particularly emphasized, it is provided that at least one reflector is arranged on at least one of the components, preferably each of the components, and / or that at least one reflector extends from the base.

The invention therefore preferably provides that at least one transmitter starts from a fixed base and / or that at least one transmitter starts from at least one of the components.

In particular, the invention is characterized in that a transmitter emanates from a fixed base, the measuring beam of which can be aligned successively to the several reflectors on the several components, the transmitter having means for aligning the measuring beam with which the measuring beam of the movement of the relative thereto moved reflector can be tracked, the transmitter preferably being attachable to the base at different positions.

It is preferably provided that a transmitter emanates from one of the components, its measuring beam being able to be aligned successively to the multiple reflectors on the multiple other components or the base, the transmitter having means for aligning the measurement beam to the respective reflector to be detected, with which the measuring beam of the movement of the respectively detected reflector moved relative to it can be tracked, wherein the transmitter can preferably be attached to the component and / or to different components at different positions.

It should also be emphasized that a transmitter having a plurality of measuring beams emanates from a stationary base or a device receiving the transmitter that is connected to the base, the measuring beams being directed simultaneously or one after the other onto a reflector of one component each.

The invention is preferably characterized in that a transmitter having a plurality of measuring beams emanates from one of the components or from a device receiving the transmitter that is connected to one of the components, the measuring beams each simultaneously or one after the other on a reflector in each case on one of the other components or the base are directed.

According to a suggestion that should be particularly emphasized, it is provided that several reflectors are arranged on at least one component or the base, onto which a measuring beam of the transmitter is directed simultaneously or one after the other in order to detect the relative position of the respective component in several spatial directions.

Another device according to the invention comprises a calibration device for determining calibration data of a coordinate measuring machine, the at least one sensor having more than one measuring dimension, in particular a linear or matrix-shaped detection area comprising sensor, and one or more axes of movement (drive units), in particular measuring axes for the relative movement of the sensor and the measuring table or rotary table receiving the workpiece to be measured, which is characterized by the fact that the calibration device is provided and arranged on or in the coordinate measuring machine, and has a position detection device which is designed to detect the position resulting from the relative movement between the sensor and the workpiece receiving measuring table or turntable, in particular the resulting relative position and / or angular position or angular position to one another.

In particular, the invention is characterized in that the position detection device comprises at least one optical distance sensor, preferably a laser interferometric structure (313-316, 333-336), and / or a camera and / or a computer tomography sensor (computer tomograph) and / or that the position detection device has the computed tomography sensor (computed tomography) contained in the coordinate measuring machine is formed.

The invention also provides a method for determining the geometry (SOUV vectors) or several different geometries of a computer tomograph (CT geometry) using the aforementioned device to be recorded on a base by means of the laser interferometric structure.

In particular, the invention is characterized in that different positions of the components are set by means of the drive units.

It is preferably provided that for several positions between the components set by means of the drive units, i.e. for several different relative positions of the components to one another, in particular for different imaging scales and / or different cone angles, with one or more transmitters or the several measuring beams of the respective transmitter one after the other or simultaneously the distances or changes in distance and / or angular positions or angular position changes between the respective transmitter and the respective detected reflector are determined, and the CT geometry set in each case is determined therefrom.

The image scale changes with the distance between the turntable and the source and the detector. If the turntable is moved towards the source, the image scale increases. A change in the image scale also results from the change in the distance between the detector and the source, with the cone angle also changing in the process. If the detector is positioned away from the source, the result is an enlargement of the image scale and a reduced cone angle. The present cone angle results from the distance between the source and the detector and is independent of the position of the turntable.

It should also be emphasized that a correction for the measurement positions of the components determined with the measurement axes is determined from the determined distances or changes in distance and / or angular positions or angular position changes, preferably in the form of a look up table.

The invention is preferably characterized in that the CT geometry or the several different CT geometries are determined and stored with the computer tomograph before the actual measurement by the positions assumed by the measuring axes or respective combinations of the positions assumed by the several measuring axes certain CT geometry is assigned, and after a change in the position of the components determined by means of the measuring axes, the respectively assigned CT geometry is assigned to the following computed tomography.

Allocation means that the SOUV vectors corresponding to the present CT geometry are used as a basis for the further evaluation, in particular used for the reconstruction of the radiographic images to form volume data.

The invention also provides a method for determining the calibration data of a coordinate measuring device using one of the devices described above, which is characterized in that the positions of the components, in particular relative positions and / or angular positions or angular position to one another, are recorded with the calibration device.

In particular, the invention is characterized in that different positions of the components are set by means of the axes of movement (drive units).

The invention is preferably characterized in that for several positions between the components set by means of the movement axes (drive units), i.e. for several different relative positions of the components to one another, in particular for different imaging scales and / or different cone angles of a computer tomograph, with the Position detection device and / or with one or more transmitters or the multiple measuring beams of the respective transmitter successively or simultaneously the distances or changes in distance and / or angular positions or angular position changes between the detection device and the respective component and / or between the respective transmitter and the respective detected reflector can be determined , and from this the respectively set position of the components and / or the respectively set CT geometry is determined.

According to a suggestion that should be particularly emphasized, it is provided that a correction for the measurement positions of the components determined with the measurement axes is determined from the determined distances or distance changes and / or angular positions or angular position changes, preferably in the form of a look up table.

The invention therefore preferably provides that the CT geometry or the several different CT geometries are determined and stored with the computer tomograph prior to the actual measurement by the positions assumed by the measuring axes or respective combinations of the positions assumed by the several measuring axes certain CT geometry is assigned, and that after a change in position of the components determined by means of the measuring axes, the respectively assigned CT geometry is assigned to the following computer tomography.

In particular, the invention is characterized in that at least first correction data are calculated from the calibration data, which are made available for the correction of measurement data determined during subsequent measurements of workpieces, correction data resulting preferably from the first correction data being calculated with the addition of further correction data, which are made available for the correction of later measurements of workpieces.

It is preferably provided that the first correction data are calculated from the calibration data of the geometric deviations of the movement axes used for moving the components, and that deviations in the position, in particular relative positions and / or angular positions, of unmoved components, for example of the sensor such as Detector and / or the X-ray source and / or the turntable, or detector-specific corrections such as bad pixel corrections, shading corrections or the like can be used.

It should also be emphasized that further correction data are determined by measuring calibrated material measures with the coordinate measuring machine and / or with the computer tomograph.

The invention is preferably characterized in that the application of the further correction data includes the virtual displacement (translation) and / or rotation (rotation) of the detector data.

• - hierarchische Kaskadierung, wobei nacheinander Korrekturdaten für die Positionen und/oder Winkellagen, beispielsweise Rechtwinkligkeitsfehler und/oder Geradheitsfehler von einer oder mehreren Bewegungsachsen, insbesondere Messachsen, und/oder feststehender Komponenten aufgenommen und angewendet werden, und anschließend Korrekturdaten für die Positionen und/oder Winkellagen weiterer Bewegungsachsen, insbesondere Messachsen, und/oder feststehender Komponenten aufgenommen und angewendet werden, oder
• - Optimierung auf Basis der Gesamtkorrekturdaten, wobei die Korrekturdaten für alle Bewegungsachsen, insbesondere Messachsen, und feststehenden Komponenten aufgenommen und dann erst angewendet werden.

According to a suggestion that should be particularly emphasized, any combination of different methods for calculating the correction data is used to determine the correction data, in particular:

• - Hierarchical cascading, with correction data for the positions and / or angular positions, for example squareness errors and / or straightness errors, of one or more axes of movement, in particular measuring axes, and / or fixed components being recorded and applied one after the other, and then correction data for the positions and / or angular positions further axes of movement, in particular measuring axes, and / or fixed components are recorded and used, or
• - Optimization on the basis of the total correction data, the correction data for all movement axes, in particular measuring axes, and fixed components being recorded and only then applied.

With hierarchical cascading, errors that occur during the first calibration are included in later calibrations. In return, these procedures are easier to implement. If the calibration is carried out on the basis of the entire calibration data, a more precise calibration is possible, since there is no error propagation. However, these procedures are usually more complicated.

• - Korrektur der Positionen der Bewegungsachsen, insbesondere Messachsen, durch entsprechendes Positionieren dieser und/oder
• - Korrektur der Position und/oder Winkellage feststehender Komponenten durch Justieren dieser, insbesondere Drehen des Detektors und/oder
• - Rechnerische Korrektur der aus den Korrekturdaten hervorgehenden Lageabweichungen der Komponenten, insbesondere durch
  • ○ Korrektur, insbesondere Verschiebung und/oder Skalierung und/oder Kippen der aufgenommenen Messdaten des Sensors, insbesondere der von einem ein- oder zweidimensionale Sensorfläche aufweisenden Bildsensor wie Zeilen- oder Matrix-Kamera oder Röntgendetektor, vorzugsweise Resampling der Messdaten in der sich aus der Korrektur ergebenden korrigierten Lage, besonders bevorzugt der Durchstrahlungsbilddaten eines Röntgenstrahlendetektor, und/oder
  • ○ Berücksichtigung einer abweichenden CT-Geometrie, wie beispielsweise Abbildungsmaßstab,
    • ■ bei der Rekonstruktion der mit dem Computertomografiesensor (Computertomografen) aufgenommenen Durchstrahlungsbilder, insbesondere Korrektur der den Durchstrahlungsbildern zugeordneten SOUV-Vektoren, und/oder
    • ■ durch Korrektur, insbesondere Verschiebung und/oder Skalierung und/oder Kippen der Durchstrahlungsbilder, und/oder
    • ■ durch Korrektur, insbesondere Verschiebung und/oder Skalierung der aus den Durchstrahlungsbildern rekonstruierten Volumendaten (Voxeldaten).

The invention therefore preferably provides that for the measurement on workpieces, the correction data determined are used according to one or more of the following methods, preferably with these being used alternatively or in cascade (one after the other):

• Correction of the positions of the movement axes, in particular measuring axes, by positioning them accordingly and / or
• - Correction of the position and / or angular position of fixed components by adjusting them, in particular rotating the detector and / or
• - Computational correction of the positional deviations of the components resulting from the correction data, in particular by
  • ○ Correction, in particular shifting and / or scaling and / or tilting of the recorded measurement data of the sensor, in particular of an image sensor having a one- or two-dimensional sensor surface such as line or matrix camera or X-ray detector, preferably resampling of the measurement data resulting from the correction resulting corrected position, particularly preferably the radiographic image data of an X-ray detector, and / or
  • ○ Consideration of a deviating CT geometry, such as the imaging scale,
    • During the reconstruction of the radiographic images recorded with the computed tomography sensor (computer tomograph), in particular correction of the SOUV vectors assigned to the radiographic images, and / or
    • Through correction, in particular shifting and / or scaling and / or tilting of the radiographic images, and / or
    • Through correction, in particular shifting and / or scaling, of the volume data (voxel data) reconstructed from the radiographic images.

The present invention also relates to a computed tomography method in which physical effects such as artifacts when radiating through a workpiece are corrected by means of simulation methods.

Correction methods known from the prior art for measurement data from a computed tomography, which work by means of simulation, are shown in FIG WO2013167616 and DE102013107745 named by the applicant. In this case, the correction data is determined by forming the difference in a simulation taking into account the artifacts and a simulation without taking into account artifacts (generalized physical properties during the irradiation). As input data for the forward projection in the context of the simulation of radiographic image data, data are required which represent the shape or dimensions of the workpiece. For this purpose, the nominal data of the workpiece such as CAD data or STL data (i.e. surface point data in STL format, Standard Triangulation Language) of a measurement of the workpiece that has already taken place (master part measurement) are used in both documents. The calculation and application of the correction takes place optionally or also mixed in the radiographic image data, the volume data reconstructed therefrom, or the surface point data determined from the volume data. This means that the differences in the simulations are formed, for example, in the radiographic image data and are then also applied directly to the radiographic image data of the real measurement. Alternatively, the differences in the radiographic image data can first be reconstructed and then subtracted from the volume data of the measurement. Provision is also made for the differences in the simulations to be formed first in the volume data reconstructed from the simulated radiographic image data and then either to correct the volume data of the measurement or to first determine surface data therefrom and correct the surface data of the measurement. Reference is here made in full to the applicant's publications cited here.

When using the correction methods described in the aforementioned documents, however, it has been found that the nominal data or STL data do not match the actual dimensions of the workpiece (real workpiece shape) sufficiently to allow an exact simulation of the artifacts and thus an exact correction guarantee. Another problem is that, in the case of measurement data with particularly high levels of artifacts, STL data, i.e. a surface point representation, cannot even be determined from the volume data, which means that these are not available for the simulations and the CAD data must be used as input data , or the STL data also deviate too much from the real workpiece shape, which in each case leads to the aforementioned disadvantage.

A further object of the present invention is therefore to avoid or at least reduce deviations that can occur when determining a correction for measurement data of a computed tomography by simulating physical effects such as artifacts, in particular when determining the correction by forming the difference in a simulation Taking into account the artifacts and a simulation without taking these artifacts into account, so that nominal data such as CAD data can be used. If STL data from a measurement of the workpiece is available, it should also be able to be used as input data for the simulations. Expected deviations in the simulation due to the deviations of the STL data from the actual workpiece shape should also be reduced or avoided.

To solve this, the invention provides that the known correction, referred to here as a preliminary correction (Sim artifacts), which results from the difference between a simulation with (Sim-Vol) and a Simulation without (ideal vol) consideration of the physical effects to be corrected such as artifacts, by applying a transformation rule (distortion MAP) to a final correction (real artifact) for the measurement data (real vol). Alternatively, the invention provides for the transformation rule to be applied to the simulation data (Sim-Vol) and (Ideal-Vol) and the difference between the resulting corrected simulation data (Sim-Vol-K) and (Ideal-Vol-K) for the determination to use the final correction (real artifact). However, since it results from this that the application of the transformation rule to the simulation data (Ideal-Vol) leads directly to the corrected measurement data (Koor-Vol) without taking into account the physical effects to be corrected, a correspondingly shortened method is also provided according to the invention.

According to the invention, the transformation rule is the mapping of the simulation data taking into account the physical effects (Sim-Vol) on the measurement data (Real-Vol). With this procedure, the part of the deviations of the simulation data, taking into account the artifacts to the real measured artifact-afflicted measurement data, which results from the deviation of the input data for the simulations to the real workpiece shape, is recorded and used to calculate the simulation data (Sim-Vol) and (Ideal -Vol) or the preliminary correction from the formation of the difference between the two simulation data (Sim-Vol) and (Ideal-Vol) or, in the case of the shortened method, only to adapt the simulation data (Ideal-Vol) accordingly.

The correction is particularly preferably carried out on the basis of the volume data. It is advantageous here that measurement data is also available in the form of volume data if these are so subject to artifacts that surface data (STL data) cannot be calculated or have large deviations from the actual workpiece shape. However, the invention basically also provides for the correction to be implemented on the basis of the radiographic image data or the surface point data or mixed, as explained above in relation to the prior art documents of the applicant.

The following explanations are given by way of example for the case of determining and applying the correction from or on the volume data. The simulation, taking into account artifacts (hereinafter referred to as Sim-Vol), is initially carried out on the basis of input data in the form of nominal data such as CAD data of the workpiece. If available, surface data (STL data) of a measurement of the corresponding workpiece can also be used. The simulation is then also carried out on the identical input data without taking the artifacts into account. The simulations are each carried out in the form of a so-called forward projection, with simulated radiographic image data being calculated and reconstructed into simulated volume data. In the next step, the transformation rule is determined that maps the artifact-affected simulation data in volume data format (Sim-Vol) onto the measured volume data (Real-Vol) reconstructed from the measured radiographic image data.

The transformation rule works on the basis of the gray values of the voxels of the volume data and reallocates the gray values in such a way that the differences between the simulation data and the measurement data with regard to the position shift between the gray value transitions are minimized by means of affine and / or non-affine transformations. These gray value transitions are later representative of the surface data to be determined. In this way, the dimensional deviations between the simulation data and the actually measured data in the area of the volume data are recorded in order to subsequently enable a more precise correction of the measured volume data with this information.

In the case of the correction based on the volume data, the SIM artifact is an artifact volume that is formed by calculation, in particular by forming the difference from the simulation with and the simulation without taking into account the artifacts. Either this artifact volume is adapted to the actual workpiece shape by applying the transformation rule, or the simulation volume data without (Ideal-Vol) consideration of the artifacts and possibly also the simulation volume data with (Sim-Vol) consideration of the artefacts themselves are adjusted by applying the transformation rule to the correct workpiece geometry (to ideal-Vol-K or Sim-Vol-K), which means that the simulation data corrected with the transformation rule (ideal-Vol-K) directly to the corrected measurement volume data (Korr-Vol) or the difference between (Sim-Vol-K) and (Ideal-Vol-K) lead directly to the final correction data (real artifact). The transformation rule results in scaling, rotation, translation, shearing and thus shifting of the gray values in the artifact volume (Sim artifact) or the simulation volume data (Ideal-Vol) and possibly (Sim-Vol) in such a way that the deviations in the edge locations ( Gray value transitions) between simulation data and measured data can be corrected. The mentioned reassignment of the gray values is not necessarily source-free, especially with scaling operations, i.e. the sum of the gray values of a volume can be changed by the transformation rule.

In the event that the corrected simulation data (Ideal-Vol-K) are not to be used directly as corrected measurement volume data (Corr-Vol), in the next step the final correction data (real artifact) are used to correct the volume data of the real measurement (Real -Vol), preferably by subtracting or adding the gray values of the final correction data (real artifact).

From the volume data (Korr-Vol) finally corrected in this way, the surface data, that is to say surface measuring points, are then determined by means of surface determination (also referred to as segmentation), which are preferably used to determine dimensional dimensions on the workpiece. The finally corrected volume data can also be used to determine internal features such as inclusions or voids in the workpiece.

• - Berechnung einer Transformationsvorschrift (Verzeichnungs-MAP), die die Simulationsdaten mit Berücksichtigung der physikalischen Effekte (Sim-Vol) auf die Messdaten (Real-Vol) abbildet, und
• - A) Anwendung der Transformationsvorschrift
  • ◯ auf vorläufige Korrekturdaten (Sim-Artefakt) zur Bestimmung der endgültigen Korrekturdaten (Real-Artefakt), wobei die vorläufigen Korrekturdaten (Sim-Artefakt) gebildet werden durch Differenzbildung zwischen Simulationsdaten mit Berücksichtigung der zu korrigierenden physikalischen Effekte (Sim-Vol) und Simulationsdaten ohne diese Effekte (Ideal-Vol), oder
  • ◯ auf die Simulationsdaten mit Berücksichtigung der zu korrigierenden physikalischen Effekte (Sim-Vol) zur Bestimmung von korrigierten Simulationsdaten (Sim-Vol-K) und auf die Simulationsdaten ohne diese Effekte (Ideal-Vol) zur Bestimmung von korrigierten Simulationsdaten (Ideal-Vol-K), und Bestimmung der endgültigen Korrekturdaten (Real-Artefakt) durch Differenzbildung zwischen (Sim-Vol-K) und (Ideal-Vol-K), und Anwendung der endgültigen Korrekturdaten (Real-Artefakt) auf die Messdaten (Real-Vol) zur Bestimmung der korrigierten Messdaten (Korr-Vol) oder
• - B) Anwendung der Transformationsvorschrift auf die Simulationsdaten ohne Berücksichtigung der zu korrigierenden physikalischen Effekte (Ideal-Vol) zur Bestimmung der korrigierten Messdaten (Korr-Vol).

The invention provides a method for calculating and applying a correction (real artifact) for measurement data (real volume) determined by means of computed tomography, the measurement data preferably being that of the workpiece to be measured in several rotational positions when irradiated by a radiation source (source) X-ray source, emitted measuring radiation, radiographic image data recorded by means of a detector, preferably 2D x-ray detector, and / or volume data reconstructed from the radiographic image data and / or surface point data determined from the volume data, with simulation data initially taking into account the physical effects to be corrected, such as artifacts, for the correction computed tomography (Sim-Vol) and simulation data without these effects (Ideal-Vol) are calculated, with nominal data of the workpiece, preferably CAD data, or STL data of a master part measurement de as input data (CAD / STL) for the simulations s workpiece, which is characterized by the following steps:

• - Calculation of a transformation rule (distortion MAP) which maps the simulation data to the measurement data (Real-Vol), taking into account the physical effects (Sim-Vol), and
• - A) Application of the transformation rule
  • ◯ on preliminary correction data (sim artifact) to determine the final correction data (real artifact), the preliminary correction data (sim artifact) being formed by forming the difference between simulation data taking into account the physical effects to be corrected (Sim-Vol) and simulation data without these effects (ideal-vol), or
  • ◯ on the simulation data taking into account the physical effects to be corrected (Sim-Vol) to determine corrected simulation data (Sim-Vol-K) and on the simulation data without these effects (Ideal-Vol) to determine corrected simulation data (Ideal-Vol-K) K), and determining the final correction data (real artifact) by calculating the difference between (Sim-Vol-K) and (Ideal-Vol-K), and applying the final correction data (real artifact) to the measurement data (Real-Vol) to determine the corrected measurement data (Korr-Vol) or
• - B) Application of the transformation rule to the simulation data without taking into account the physical effects to be corrected (Ideal-Vol) to determine the corrected measurement data (Corr-Vol).

• - Berechnung der Transformationsvorschrift (Verzeichnungs-MAP),
• - Differenzbildung zwischen den Simulationsdaten (Sim-Vol) und (Ideal-Vol),
• - Anwendung der Transformationsvorschrift auf die vorläufigen Korrekturdaten (Sim-Artefakt),
• - Anwendung der Transformationsvorschrift auf die Simulationsdaten (Sim-Vol) und/oder (Ideal-Vol),
• - Differenzbildung zwischen den korrigierten Simulationsdaten (Sim-Vol-K) und (Ideal-V ol- K),
• - Anwendung der endgültigen Korrekturdaten (Real-Artefakt) auf die Messdaten (Real-Vol) zur Bestimmung der korrigierten Messdaten (Korr-Vol)
• - Verwendung der korrigierten Simulationsdaten (Ideal-Vol-K) als korrigierte Messdaten (Korr-Vol).

In particular, the invention is characterized in that all of the steps carried out from the following list take place on the respective radiographic image data or the respective volume data or the respective surface point data:

• - Calculation of the transformation rule (distortion MAP),
• - Difference formation between the simulation data (Sim-Vol) and (Ideal-Vol),
• - Application of the transformation rule to the preliminary correction data (SIM artifact),
• - Application of the transformation rule to the simulation data (Sim-Vol) and / or (Ideal-Vol),
• - Difference between the corrected simulation data (Sim-Vol-K) and (Ideal-V ol-K),
• - Application of the final correction data (real artifact) to the measurement data (Real-Vol) to determine the corrected measurement data (Korr-Vol)
• - Use of the corrected simulation data (Ideal-Vol-K) as corrected measurement data (Korr-Vol).

It is preferably provided that the final correction (real artifact) is applied to the measurement data (real vol) by subtracting or adding the final correction data (real artifact).

It should also be emphasized that the transformation rule assigns the gray values of the volume data of the simulation with artifacts (Sim-Vol), gray values of the volume data of the measurement data (Real-Vol) as a function of location, and preferably changes them, for example in the form of an affine Mapping and / or non-affine mapping and / or a look up table (LUT), with the transformation rule preferably mapping the shift of the gray value progressions assigned to the surface transitions by rearranging the gray values.

In particular, the invention is characterized in that the transformation rule is determined by optimizing a cost function, this cost function consisting of several individual terms, such as equality metrics and / or functions thereof and / or restricting boundary conditions.

The invention is also characterized in that the finally corrected measurement data from computer tomography (Korr-Vol) are used to determine features inside the workpiece, such as inclusions or voids, by evaluating the finally corrected volume data and / or for determining dimensional data Dimensions of the workpiece can be used by evaluating the surface point data.

According to a proposal to be particularly emphasized, it is provided that the method is used in a computer tomograph designed as a coordinate measuring machine, which is designed to determine dimensions on workpieces from the surface point data, that is to say to link surface points to dimensions.

The invention also relates to a method for improving the quality, in particular resolution and / or sharpness and / or contrast, of image data which were recorded using imaging methods.

The quality, in particular the resolution of image data recorded by means of imaging methods, is improved in clinical computed tomography applications by sharpening X-ray images or reconstructed volumes by means of convolutional neural networks (CNN), which are very homogeneous data. In the metrological field, especially in computed tomography for metrological purposes or image processing using optical sensors in coordinate measuring technology (optical coordinate measuring machines), however, very heterogeneous data are available. The corresponding optics have a so-called Point Spread Function (PSF), which, depending on various recording parameters, such as exposure time, lighting or magnification in optics with multiple or adjustable magnifications (or image scales), leads to blurring (blurring) in the image data to lead. The blurring depends not only on the actual recording optics but also on the workpiece to be measured. With a fixed number of pixels of the cameras used for imaging, there is also the challenge that measurements can only be made with either high resolution in a small measuring field or low resolution in a large measuring field.

A further object of the present invention is therefore to improve the quality, in particular the resolution and / or the sharpness and / or the contrast, of image data recorded by means of imaging methods, quickly and easily. In particular, the task is also to transfer the high quality of a sub-area in the image data to other sub-areas of the image data with low quality. Another object of the invention is to reduce the measurement time in a computer tomography and at the same time to achieve high-quality, in particular high-resolution and / or high-contrast results in all areas of the image data that are otherwise only achieved with a longer measurement time or other spectra of the measurement radiation.

In a particular object of the invention, the combination of the respective advantages of measurements with high and low quality is to be realized. In particular, the image data should be corrected for a specific PSF of an optical system of the inspection or coordinate measuring device used and / or for a specific workpiece to be measured.

As a solution, the invention therefore also provides that image data of low quality are converted into image data of high quality, at least in some areas, by means of a correction algorithm. The correction algorithm is based on a learned relationship between the image data of low and high quality (so-called image pairs). To determine the relationship, image data of high and low quality, of the scenes that are at least regionally identical, or of excerpts of these (image pairs), are recorded. For the areas in which high and low resolution data overlap, training is carried out using a machine learning process (AI) such as an artificial neural network, whereby a correction algorithm is learned that is used in real measurements to correct the areas of the image data in which only Image data of low quality are present in order to convert them into image data of high quality. This task is solved by the AI.

In a first embodiment, the invention provides that the imaging method is an image processing method by means of an optical sensor in a coordinate measuring machine is applied. The corresponding optical sensor can be, for example, zoom optics or an optical sensor with two imaging beam paths that cover a small area of the workpiece with high resolution or quality (large imaging scale) and a large area of the workpiece with lower quality ( low imaging scale), for example, through a common front lens, the two areas being centrally aligned with one another, that is to say superimposing one another in a central area, as is the case for example in FIG DE102019118321.9 the applicant is described. In this case, data with high quality, in particular high resolution, are already present in an overall image in the inner area, but only data with low quality, in particular low resolution, in the outer area. In order to transfer the high quality of the data in the inner area to the outer area, a neural network (network) of a machine learning process such as artificial intelligence (AI) is trained with the image data in the inner area, with high-quality and low-quality data being available here. The correction algorithm (network) learned in the process is then transferred to the image data in the outer area in order to improve it.

A second embodiment of the invention relates to a computer tomographic method. Here, radiographic images are recorded in several rotary positions with long exposure times, with high quality data being produced, and again radiographic images with a short picture recording time for each rotary position, with low-quality data being produced. A corresponding neural network is trained with these two image data sets. In the later measurement, the recording with a short recording time, i.e. low quality, is sufficient to generate high quality image data after the application of the neural network. In this way, the measurement time can be reduced.

In a special embodiment of this idea, it is provided that during the real computed tomographic measurement, in at least some rotational positions, image pairs with low and additionally also high quality, i.e. short and long exposure times, are recorded in order to use them additionally for training the neural network. As a result, the network is adapted to the specific workpiece and / or the specific computer tomograph in relation to its specific parameters.

In a further embodiment of the invention, use is provided for a computed tomography method in which the same workpiece is first recorded in two measurements with different exposure. In a first measurement with a small focal point, in particular low power, i.e. lower voltage and / or low current of the x-ray tube, a measurement is carried out with high resolution, with the low power under certain circumstances only allowing partial areas of the workpiece to be irradiated with high contrast. In a second measurement, a large focal spot and high power are used, although the resolution or quality is lower, but areas that are difficult to penetrate are also penetrated with high contrast. For the training of a neural network, the areas in the image data (radiographs) are now used that are irradiated with high contrast in both measurements. With a network trained in this way, it is also possible to correct the areas of the remaining image data for which a high-contrast image was not possible because the corresponding images were recorded too dark or too light.

Training a neural network means optimizing neural parameters (weights and biases) to solve a problem.

A ready-to-use trained network is a neural network whose variables or neural parameters (weights and biases) have been optimized in such a way that the desired optimization (e.g. sharpness calculation of images) of sufficient quality is achieved for any measurement object.

A pre-trained network is a neural network whose variable or neural parameters (weights and biases) have already been roughly optimized for a task (e.g. sharp calculation of images), but this optimization is not specialized enough compared to the ready-to-use trained network therefore cannot be used directly as a correction algorithm.

An untrained neural network is a neural network whose variables or neural parameters (weights and biases) are initialized randomly or without being able to solve a specific task.

To solve this, the invention provides a method for correcting image data of a workpiece recorded with an imaging device, such as an optical sensor, preferably an optical sensor in a coordinate measuring machine, or a radiation-based inspection and / or measuring machine, in particular a computer tomograph, the image data being transmitted through the imaging device can be generated in at least a first quality (low quality), such as first resolution (low resolution) or first sharpness (low sharpness), which is characterized in that, by means of a correction algorithm, the low quality image data is at least partially in Image data of high quality, preferably high resolution (high resolution) and / or high sharpness (high sharpness), are converted, the correction algorithm being based on a learned relationship between image data of low quality and image data of high quality, and the correction algorithm for learning the relationship between image data high and low quality (image pairs) of an identical scene or several identical scenes of at least one section of the workpiece to be measured and / or other workpieces of the same or different shape in relation to the workpiece to be measured are made available.

• - geringerem Abbildungsmaßstab einer Optik und/oder
• - geringerem Abbildungsmaßstab des Computertomografen und/oder
• - Bildaufnahme während der Bewegung der Messachsen des den optischen Sensor beinhaltenden Koordinatenmessgeräts und/oder
• - Bildaufnahmen während der Drehbewegung des Drehtisches des Computertomografen und/oder
• - Bildaufnahmen mit erster Belichtung mit dem Computertomografen, wie erster Spannung und/oder Strom und/oder Leistung und/oder Spektrum und/oder Belichtungszeit je Bild und/oder Anzahl der Bildüberlagerungen je Bild und/oder
• - ersten Messmodus des Computertomografen, wie Kegelstrahltomografie und die Bilddaten hoher Qualität in einer zweiten Betriebsart mit der identischen Vorrichtung oder einer anderen Vorrichtung, mit vorzugsweise ansonsten gleichen Abbildungseigenschaften wie der Vorrichtung, deren Bilddaten zu korrigieren sind, aufgenommen werden, beispielsweise bei im Vergleich zur ersten Betriebsart
• - größerem Abbildungsmaßstab einer Optik und/oder
• - größerem Abbildungsmaßstab des Computertomografen und/oder
• - Bildaufnahme während des Stillstandes der Messachsen des den optischen Sensor beinhaltenden Koordinatenmessgeräts und/oder
• - Bildaufnahmen während des Stillstandes der Drehbewegung des Drehtisches des Computertomografen und/oder
• - Bildaufnahmen mit zweiter Belichtung mit dem Computertomografen, wie zweiter Spannung und/oder Strom und/oder Leistung und/oder Spektrum und/oder Belichtungszeit je Bild und/oder Anzahl der Bildüberlagerungen je Bild und/oder
• - zweitem Messmodus des Computertomografen, wie Helixtomografie.

In particular, the invention is characterized in that the image data of low quality are recorded in a first operating mode of the device, for example when compared to a second operating mode

• - lower image scale of an optics and / or
• - lower image scale of the computer tomograph and / or
• - Image recording during the movement of the measuring axes of the coordinate measuring device containing the optical sensor and / or
• - Image recordings during the rotary movement of the turntable of the computer tomograph and / or
• - Image recordings with the first exposure with the computer tomograph, such as the first voltage and / or current and / or power and / or spectrum and / or exposure time per image and / or number of image overlays per image and / or
• - First measurement mode of the computed tomography, such as cone beam tomography and the high quality image data are recorded in a second operating mode with the identical device or another device, preferably with otherwise the same imaging properties as the device whose image data are to be corrected, for example when compared to the first Operating mode
• - Larger image scale of an optics and / or
• - Larger image scale of the computer tomograph and / or
• - Image recording during the standstill of the measuring axes of the coordinate measuring device containing the optical sensor and / or
• - Image recordings during the standstill of the rotary movement of the turntable of the computer tomograph and / or
• - Image recordings with second exposure with the computer tomograph, such as second voltage and / or current and / or power and / or spectrum and / or exposure time per image and / or number of image overlays per image and / or
• - Second measurement mode of the computer tomograph, such as helix tomography.

It is preferably provided that image data of low quality are recorded in a first operating mode of the device and image data of high quality in a second operating mode of the identical device, in particular with the two operating modes having different exposures and / or spectra with a computer tomograph or with the two operating modes different optical Have magnifications in an optical sensor.

It should also be emphasized that the image data of low and high quality (image pairs) are at least partially superimposed locally, with image data assigned to identical areas being used for learning the relationship.

The invention is also characterized in that the correction algorithm for converting image data of low quality into image data of high quality is generated and implemented in a self-learning manner using a machine learning method (KI), such as an artificial neural network (network), preferably a convolutional neural network (CNN) ), particularly preferably U-Net and / or Residual Neural Network (ResNet) or Generative Adversarial Network (GAN), with low-quality image data being used as input data for training the network and when using the network, and with high-quality image data as target data when training the network or as output data when using the network.

According to a suggestion that should be particularly emphasized, it is provided that the correction algorithm used is an operationally trained network or a pre-trained network with additional training from partial data from a measurement of the workpiece and subsequent application of the network to the low-quality image data.

In particular, the invention is characterized in that, for the training of the network, image data of high quality and image data of low quality (image pairs), in particular low resolution, using the specific optics or other optics with the specific recording parameters that are provided for the actual optics, be used to optimize the correction algorithm on the Point Spread Function (PSF) of the specific optics and / or on the to effect specific recording parameters with regard to blurring in the image data of low quality, preferably with the application of a ready-to-use trained network that has been optimized for a specific measurement task or optics or the use of a pre-trained network with additional training on the image data with the specific Optics.

It is preferably provided that high and low quality image data (image pairs), in particular high and low resolution from measurement of a specific workpiece, are used for training the network, with the optimization of the correction algorithm being effected on the specific measurement of this workpiece, and preferably for this purpose the use of a pre-trained network with additional training on partial data sets of the image data of the workpiece is provided.

• - das Rauschen minimiert, vorzugsweise auf das Rauschen der Bilddaten hoher Qualität minimiert ist und/oder
• - das Messfeld maximiert, vorzugsweise auf die Größe des Messfelds der Bilddaten niedriger Qualität maximiert ist und/oder
• - unter Verwendung erster und zweiter Belichtungen und/oder Spektren die Durchstrahlung des Werksrücks optimiert, vorzugsweise auf kontrastreiche Bilddaten optimiert wird und/oder
• - die Messzeit bei gegebener Verbesserung der Bildqualität minimiert wird.

It should also be emphasized that image data of high and low quality (image pairs), in particular high and low resolution or high and low contrast, are combined to form resulting image data which have the high quality, with.

• the noise is minimized, preferably minimized to the noise of the high quality image data, and / or
• - the measurement field is maximized, preferably maximized to the size of the measurement field of the image data of low quality and / or
• - Using first and second exposures and / or spectra, the irradiation of the back of the work is optimized, preferably optimized for high-contrast image data and / or
• - The measurement time is minimized with a given improvement in the image quality.

The invention is also characterized in that the correction algorithm is applied to any areas of the image data of low quality, in particular areas for which there is no overlap with the image data of high quality.

In particular, the invention is characterized in that the correction of image data of low quality takes place outside the area in which training of the image data of high quality took place, preferably the training took place in a central area of the image data and the correction algorithm in an area outside the central one Area lying area of the image data is applied.

Further details, advantages and features of the invention emerge not only from the claims, the features to be derived from them - individually and / or in combination - but also from the following description of the figures.

• 1 eine Prinzipdarstellung der erfindungsgemäßen Anordnung zur Bestimmung der CT-Geometrie,
• 2 eine Prinzipdarstellung des Ablaufs eines erfindungsgemäßen Korrekturverfahrens für Artefakte bei einer Computertomografie,
• 3 eine Prinzipdarstellung des Ablaufs eines alternativen erfindungsgemäßen Korrekturverfahrens für Artefakte bei einer Computertomografie,
• 4 eine Prinzipdarstellung des Ablaufs eines alternativen verkürzten erfindungsgemäßen Korrekturverfahrens für Artefakte bei einer Computertomografie und
• 5 eine Prinzipdarstellung des Ablaufs eines erfindungsgemäßen Verfahrens zur Verbesserung der Qualität, insbesondere Auflösung von Bilddaten eines bildgebenden Verfahrens.

Show it:

• 1 a schematic diagram of the arrangement according to the invention for determining the CT geometry,
• 2 a schematic representation of the sequence of a correction method according to the invention for artifacts in a computer tomography,
• 3 a schematic representation of the sequence of an alternative correction method according to the invention for artifacts in a computer tomography,
• 4th a schematic representation of the sequence of an alternative shortened correction method according to the invention for artifacts in a computer tomography and
• 5 a basic representation of the sequence of a method according to the invention for improving the quality, in particular the resolution of image data of an imaging method.

1 shows an exemplary arrangement according to the invention of a computer tomograph 300 with a laser interferometric structure (laser interferometer) 301 to determine the CT geometry. From the base 302 go over several measuring axes 303 to 306 The components of the computer tomograph are flexibly connected 300 from, namely the source 308 , the one for picking up and rotating the workpiece to be measured later by means of computer tomography 310 trained turntable 309 and the detector 311 . For recording the positions of the measuring axes 303 to 306 are the multiple reflectors 313 to 316 arranged on these. The reflectors are from the measuring beams 323 to 326 the transmitter 333 to 336 targeted. In the embodiment shown here, the transmitters go 333 to 336 not from the base 302 from, but from the components 308 and 311 or are connected to them via a measuring axis running perpendicular to the respective measuring beam. When the turntable is moving 309 The imaging scale FDD / FOD is set along the spatial direction (Z-axis) designated by Z. Moving the detector 311 or source 308 in the Z direction, the detected cone angle 312 adapt to the measurement radiation. It is sufficient here, detector 311 or source 308 to move. For the determination of the CT geometry in the Z direction, the relative measurement between the detector is therefore essential 311 and source 308 medium of the transmitter 333 and reflector 313 as well as the relative measurement between the detector 311 and turntable 309 medium of the transmitter 335 and reflector 315 sufficient.

In this exemplary embodiment, only movements or positions of the components with respect to one another in the X direction and in the Z direction are determined. According to the invention, provision is also made for corresponding arrangements to be provided for the detection in the Y direction, particularly preferably the movement of the turntable 309 in the Y direction (perpendicular to the plane of the drawing). It is also provided that the transmitter 333 to 336 from the base 302 go out.

This exemplary embodiment describes a first alternative in which the transmitter 333 to 336 operated one after the other by means of multiplexers with the identical beam generation and evaluation unit, which is not shown here. According to the invention, however, it is also provided that there are separate interferometers for each transmitter 333 to 336 to use. According to the invention, an interferometer (not shown here) with a rotatable and pivotable measuring beam on the base is also provided 302 or one of the components 308 , 309 , 311 to be arranged, preferably to be arranged one after the other at several positions, the measuring beam successively on the reflectors 313 to 316 directed and when moving the assigned measuring axis 323 to 326 is tracked (in the sense of a laser tracker).

The transmitters become advantageous 333 to 336 after the CT geometry has been determined or the measuring axes have been calibrated 303 to 316 dismantled. This makes it possible to use the measuring beams 323 to 326 also to run through the areas in which the workpiece is measured during the computed tomographic measurement 310 is arranged.

2 shows an exemplary basic illustration of the sequence of a correction method according to the invention for artifacts in a computer tomography. As input data 500 for simulations 501 , 502 the computed tomographic measurement of a workpiece is based on the nominal data in the form of CAD data or surface data in STL format of the workpiece from, for example, a master part measurement. The simulation 501 takes place without taking artefacts into account 503 and delivers the simulation data Ideal-Vol. The simulation 502 takes into account selected artifacts 503 and delivers the simulation data Sim-Vol. The simulation is the calculation of radiographic image data of the workpiece by means of forward projection, whereby the parameters of the respective computer tomograph such as focal spot size, spectrum and power of the emitted measurement radiation of the radiation source, size, pixel size, number of pixels, pixel spacing of the detector and position of the components radiation source, detector and workpiece, as well as the position of the axis of rotation of the turntable for rotating the workpiece relative to one another, are taken into account. By way of example, the simulated radiographic image data are first reconstructed into volume data, so that ideal-Vol and Sim-Vol are referred to here as volume data. These are then subtracted from one another, that is to say the gray values of the respective spatially corresponding voxels of the volume data are subtracted from one another in order to form the simulated artifact volume that the preliminary correction data Sim artifact 504 forms. From the real measurement of the workpiece with the computer tomograph, the volume data Real-Vol, reconstructed from the exempted radiographic images, are also available. From Real-Vol and the volume data Sim-Vol from the simulation 502 taking into account the artifacts 503 the transformation rule (distortion MAP) 505 determines which sim-vol maps to real-vol. The transformation rule 505 is then sim artifact on the artifact volume 504 (preliminary correction data) applied to the final correction data real artifact 506 to build. The final correction data real artifact 506 Real-Vol are subtracted from the volume data of the measurement, i.e. the gray values of the assigned voxels are subtracted in order to determine the corrected volume data Korr-Vol. These corrected measurement data are evaluated in the volume range in order to determine features inside the workpiece, such as inclusions or voids. Alternatively or additionally, surface point data are determined from the corrected volume data Korr-Vol by means of surface extraction methods or segmentation, from which the dimensional dimensions of the workpiece are determined.

In the 3 an alternative procedure to the 2 shown, the simulated volume data Ideal-Vol and Sim-Vol by applying the transformation rule 505 are converted into the corrected volume data Ideal-Vol-K and Sim-Vol-K, the difference between which is formed to produce the final correction data real artifact 506 to build. It should be noted here that the application of the transformation rule 505 on the simulated volume data Sim-Vol leads to the corrected simulated volume data Sim-Vol-K agreeing with the measurement data Real-Vol. As a result, the final correction data correspond to real artifacts 506 hence the difference between the measurement data Real-Vol and the corrected simulated volume data Sim-Vol-K, which means that the corrected volume data of the measurement Korr-Vol correspond exactly to the corrected volume data Sim-Vol-K. The application of the transformation rule 505 on the simulated volume data Sim-Vol can therefore be omitted and the corrected volume data of the measurement Korr-Vol are transferred directly by applying the transformation rule 505 on the simulated volume data obtained ideal-Vol, as shown in the 4th is shown.

5 shows an exemplary basic illustration of the sequence of a method according to the invention for improving the quality, in particular the resolution of image data of an imaging method. The fuzzy image data that occurs during a real measurement 400 are corrected by using an artificial intelligence or neural network and, if necessary, also combined with other high quality image data to form the high quality image data 401 . For training 402 Artificial intelligence (AI) uses high quality measurement data sets (data set A (sharp)) and low quality data sets (data set A (unsharp)). In addition, a training 403 the AI on a special set of optics parameters 1 respectively. Here, data sets B are in the parameter set 1 recorded when the image is in focus and out of focus. In a further optional additional training of the AI, an optimization for a special measurement takes place by recording the data sets C in sharp and fuzzy images.

The data sets labeled “sharp” or “unsharp” represent examples of image data of high quality (“sharp”) or low quality (“unsharp”). The “optics 1” can be an optical sensor of an imaging method, but also a radiation source-detector combination of a computer tomograph. The “data record A” in 402 can be any image data record recorded on any training object with any imaging sensor, on the basis of which the AI learns to convert fuzzy data into sharp data. The fuzzy data can be generated in a variety of ways. For example, the sharp data can be low-pass filtered. It is also possible to generate the fuzzy data by changing the magnification of the optics. In the case of computed tomography, the fuzzy data can be generated by recording while rotating instead of at a standstill, or by reducing or changing the exposure (reduced exposure time, reduced number of image averages, reduced beam power or changed spectrum of the beam source). The optional additional training 403 now takes into account the properties of the specific sensor “optics 1” used later for the measurement and its measurement parameters “parameter set 1”, such as the present magnification of the optical components used and the pixel size of the camera chip of the camera used.

This training 403 therefore uses the different "data set B". With the optional training 404 a different "data record C" is used, which can be recorded during the actual measurement. This is therefore part of the measurement data that is required for training 404 are available anyway or can be added with little additional effort. A first example of this is the case of the two combined optical beam paths from the DE102019118321.9 of the applicant. In the inner area, image data from two magnifications and thus resolutions (sharp and out of focus) are already available, so that the training can take place continuously parallel to the measurement in order to train a sharp calculation of the outer areas, for which only image data of low magnification and thus resolution (out of focus) are present. A second example of the training 404 is the additional recording of radiographic images of the workpiece to be measured in a computer tomography with changed recording parameters. These can be radiographic images that are exposed for a longer period of time or when the axis of rotation is at a standstill instead of in motion at a reduced number of rotational positions, but also a complete second measurement with changed recording parameters. For areas in the radiographic images that were so sharply mapped instead of previously blurred, training can now take place that is applied to the remaining blurred areas. The workout 404 can also take place on a similar object compared to the workpiece actually to be measured, for example on a master part or other workpieces from the same series.

QUOTES INCLUDED IN THE DESCRIPTION

This list of the documents listed by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent literature cited

• WO 2005119174 A1 [0002]
• DE 102008044437 [0002]
• DE 102014113977 [0002]
• EP 1990605 [0005]
• DE 102007004934 [0006]
• EP 2390737 [0007]
• DE 102016113531 [0008]
• DE 102017103455 [0008]
• WO 2015079019 [0009]
• WO 2013167616 [0047]
• DE 102013107745 [0047]
• DE 102019118321 [0070, 0099]

Claims (46)

Calibration device for determining the geometry (SOUV vectors) or several different geometries of a computer tomograph (300) (CT geometry), i.e. the respective position of the components (308, 309, 311) of the computer tomograph (300) relative to one another, the components (308 , 309, 311) are: 1. Radiation source, in particular X-ray source (source) (308) 2. Radiation detector, in particular X-ray detector (detector) (311) 3. Turntable (309) for receiving and rotating the workpiece (310) to be examined, and wherein the respective position comprises at least the respective relative positions of the components (308, 309, 311) to one another and preferably the respective angular positions of the components (308, 309, 311) to one another, characterized in that at least one laser interferometric Structure (313-316, 333-336) is arranged in or on the computer tomograph (300), the respective positions of the components (308, 309, 311) in relation to g to a fixed base (base) (302) and / or relative to each other. Device according to Claim 1 , characterized in that one, two or all three components (308, 309, 311) start from a drive unit (303-316) which can be displaced in one, two or three spatial directions, preferably start from measuring axes, the drive unit (303-316 ) or the respective measuring axis causes a relative displacement of the respective component (308, 309, 311) with respect to the base (302). Device according to Claim 1 or 2 , characterized in that the laser interferometric structure (313-316, 333-336) has at least one unit (transmitter) (333-336) emitting at least one measuring beam (323-326), at least one reflecting or deflecting a measuring beam (323-326) Unit (reflector) (313-316) such as plane mirror or triple mirror and an evaluation unit, preferably in the transmitter (333-336) or on a device receiving the transmitter (333-336) at least one reference path for the generation of by the evaluation unit processing interference with the measuring beam (323-326) is arranged. Device according to at least one of the preceding Claims 1 to 3 , characterized in that at least one reflector (313-316) is arranged on at least one of the components (308, 309, 311), preferably each of the components (308, 309, 311) and / or that at least one reflector (313 -316) starts from the base (302). Device according to at least one of the preceding Claims 1 to 4th , characterized in that at least one transmitter (333-336) starts from a fixed base (302) and / or that at least one transmitter (333-336) starts from at least one of the components (308, 309, 311). Device according to at least one of the preceding Claims 1 to 5 , characterized in that a transmitter (333-336) extends from a fixed base (302), the measuring beam (323-326) of which can be aligned one after the other to the multiple reflectors (313-316) on the multiple components (308, 309, 311) the transmitter (333-336) having means for aligning the measuring beam (323-326) with which the measuring beam (323-326) can be tracked to the movement of the reflector (313-316) moved relative thereto, the transmitter ( 333-336) is preferably attachable to the base 302 in different positions. Device according to at least one of the preceding Claims 1 to 8th , characterized in that a transmitter (333-336) emanates from one of the components (308, 309, 311), its measuring beam (323-326) successively targeting the multiple reflectors (313-316) on the multiple other components (308 , 309, 311) or the base (302), the transmitter (333-336) having means for aligning the measuring beam (323-326) with the respective reflector (313-316) to be detected, with which the measuring beam ( 323-326) the movement of the respectively detected reflector (313-316) moved relative to it can be tracked, the transmitter (333-336) preferably at different positions on the component (308, 309, 311) and / or on different components ( 308, 309, 311) can be attached. Device according to at least one of the preceding Claims 1 to 7th , characterized in that a transmitter (333-336) having a plurality of measuring beams (323-326) emanates from a fixed base (302) or from a device receiving the transmitter (333-336) which is connected to the base (302) is connected, the measuring beams (323-326) each being directed simultaneously or one after the other onto a reflector (313-316) each of a component (308, 309, 311). Device according to at least one of the preceding Claims 1 to 8th , characterized in that a transmitter (333-336) having a plurality of measuring beams (323-326) emanates from one of the components (308, 309, 311) or from a device receiving the transmitter (333-336) which is connected to one of the components ( 308, 309, 311), the measuring beams (323-326) each being directed simultaneously or one after the other onto a reflector (313-316) of one of the other components (308, 309, 311) or the base (302). Device according to at least one of the preceding Claims 1 to 9 , characterized in that a plurality of reflectors (313-316) are arranged on at least one component (308, 309, 311) or the base (302), onto each of which a measuring beam (323-326) of the transmitter (333-336) is applied simultaneously or is directed one after the other in order to detect the relative position of the respective component (308, 309, 311) in several spatial directions. Calibration device for determining calibration data of a coordinate measuring machine, the at least one sensor having more than one measuring dimension, in particular a linear or matrix-shaped detection area, and one or more axes of movement (drive units), in particular measuring axes for the relative movement of the sensor and the workpiece to be measured, a measuring table or rotary table to one another , characterized in that the calibration device is provided and arranged on or in the coordinate measuring machine, and has a position detection device which is designed to detect the position between the sensor and the workpiece receiving the measuring table or rotary table, in particular the resulting relative position and / or angular position or angular position to one another. Device according to Claim 11 , characterized in that the position detection device comprises at least one optical distance sensor, preferably a laser interferometric structure (313-316, 333-336), and / or a camera and / or a computed tomography sensor (computed tomography) and / or that the position detection device by the Computed tomography sensor (computed tomograph) is formed. Method for determining the geometry (SOUV vectors) or several different geometries of a computer tomograph (CT geometry) using the device according to at least one of the preceding Claims 1 to 12th , characterized in that the positions of the components to one another and / or to a base are detected by means of the laser interferometric structure. Procedure according to Claim 13 , characterized in that different positions of the components are set by means of the drive units. Procedure according to Claim 13 or 14th , characterized in that for several positions between the components set by means of the drive units, i.e. for several different relative positions of the components to one another, in particular for different imaging scales and / or different cone angles, with one or more transmitters or the several measuring beams of the respective transmitter one after the other or simultaneously the distances or changes in distance and / or angular positions or changes in angular position between the respective transmitter and the respective detected reflector are determined, and the CT geometry set in each case is determined therefrom. Method according to at least one of the preceding Claims 13 to 15th , characterized in that a correction for the measurement positions of the components determined with the measurement axes is determined from the determined distances or distance changes and / or angular positions or angular position changes, preferably in the form of a look up table. Method according to at least one of the preceding Claims 13 to 16 , characterized in that the CT geometry or the several different CT geometries are determined and stored with the computer tomograph prior to the actual measurement, in that the positions assumed by the measuring axes or respective combinations of the positions assumed by the several measuring axes are each determined thereby Geometry is assigned, and that after a change in position of the components determined by means of the measuring axes, the respectively assigned CT geometry is assigned to the following computed tomography. Method for determining the calibration data of a coordinate measuring machine using the device according to at least one of the preceding Claims 1 to 12th , characterized in that the positions of the components, in particular relative positions and / or angular positions or angular position to one another, are detected with the calibration device. Method according to at least one of the preceding Claims 13 to 18th , characterized in that different positions of the components are set by means of the axes of movement (drive units). Method according to at least one of the preceding Claims 13 to 19th , characterized in that for several positions set between the components by means of the movement axes (drive units), that is for several different relative positions of the components to one another, in particular for different imaging scales and / or different cone angles of a computer tomograph, with the position detection device and / or with one or more transmitters or the several measuring beams of the respective transmitter successively or simultaneously the distances or changes in distance and / or angular positions or angular position changes between of the detection device and the respective component and / or between the respective transmitter and the respective detected reflector, and from this the respectively set position of the components and / or the respectively set CT geometry is determined. Method according to at least one of the preceding Claims 13 to 20th , characterized in that a correction for the measurement positions of the components determined with the measurement axes is determined from the determined distances or distance changes and / or angular positions or angular position changes, preferably in the form of a look up table. Method according to at least one of the preceding Claims 13 to 21 , characterized in that the CT geometry or the several different CT geometries are determined and stored with the computer tomograph prior to the actual measurement, in that the positions assumed by the measuring axes or respective combinations of the positions assumed by the several measuring axes are each determined thereby Geometry is assigned, and that after a change in position of the components determined by means of the measuring axes, the respectively assigned CT geometry is assigned to the following computed tomography. Method according to at least one of the preceding Claims 13 to 22nd , characterized in that at least first correction data are calculated from the calibration data, which are made available for the correction of measurement data determined in later measurements of workpieces, with correction data resulting from the first correction data with the addition of further correction data, which are preferably calculated for the correction later measurements of workpieces are made available. Method according to at least one of the preceding Claims 13 to 23 , characterized in that the first correction data are calculated from the calibration data of the geometric deviations of the movement axes used for moving the components, and that preferably as further correction data, deviations in the position, in particular relative positions and / or angular positions, of unmoved components, for example the sensor such as Detector and / or the X-ray source and / or the turntable, or detector-specific corrections such as bad pixel corrections, shading corrections or the like can be used. Method according to at least Claim 23 or 24 , characterized in that further correction data are determined by measuring calibrated material measures with the coordinate measuring machine and / or with the computer tomograph. Method according to at least one of the preceding Claims 23 to 25th , characterized in that the application of the further correction data comprises the virtual displacement (translation) and / or rotation (rotation) of the detector data. Method according to at least one of the preceding Claims 13 to 26th , characterized in that any combination of different methods for calculating the correction data is used to determine the correction data, in particular: hierarchical cascading, with correction data for the positions and / or angular positions, for example squareness errors and / or straightness errors of one or more axes of movement, one after the other, in particular measuring axes and / or fixed components are recorded and applied, and then correction data for the positions and / or angular positions of further movement axes, in particular measuring axes, and / or fixed components are recorded and applied, or optimization on the basis of the total correction data, the correction data for all axes of motion, especially measuring axes, and fixed components can be recorded and only then used. Method according to at least one of the preceding Claims 13 to 27 , characterized in that, for the measurement on workpieces, the correction data determined is applied according to one or more of the following methods, preferably these being applied alternatively or cascaded (one after the other): Correction of the positions of the movement axes, in particular measuring axes, by positioning them accordingly and / or - Correction of the position and / or angular position of fixed components by adjusting them, in particular rotating the detector and / or - Computational correction of the positional deviations of the components resulting from the correction data, in particular by ○ correction, in particular shifting and / or scaling and / or Tilting the recorded Measurement data of the sensor, in particular the image sensor having a one- or two-dimensional sensor surface, such as a line or matrix camera or X-ray detector, preferably resampling of the measurement data in the corrected position resulting from the correction, particularly preferably the radiographic image data of an X-ray detector, and / or ○ Consideration of a deviating CT geometry, such as image scale, ■ when reconstructing the radiographic images recorded with the computed tomography sensor (computer tomograph), in particular correcting the SOUV vectors assigned to the radiographic images, and / or ■ through correction, in particular shifting and / or scaling and / or tilting the radiographic images and / or by correcting, in particular shifting and / or scaling, the volume data (voxel data) reconstructed from the radiographic images. Method for calculating and applying a correction (real artifact) for measurement data (real volume) determined by means of computed tomography, the measurement data using the measurement radiation emitted by the workpiece to be measured in several rotational positions when irradiated by a radiation source (source), preferably an X-ray source of a detector, preferably 2D X-ray detector, recorded radiographic image data and / or volume data reconstructed from the radiographic image data and / or surface point data determined from the volume data, for the correction initially simulation data taking into account the physical effects to be corrected, such as artifacts in the computed tomography (Sim-Vol ) and simulation data are calculated without these effects (Ideal-Vol), with nominal data of the workpiece, preferably CAD data, or STL data of a master part measurement of the workpiece being used as input data (CAD / STL) for the simulations, geke Characterized by the following steps: - Calculation of a transformation rule (distortion MAP), which maps the simulation data, taking into account the physical effects (Sim-Vol), on the measurement data (Real-Vol), and - A) Application of the transformation rule o to preliminary correction data (Sim artifact) to determine the final correction data (real artifact), the preliminary correction data (SIM artifact) being formed by forming the difference between simulation data taking into account the physical effects to be corrected (Sim-Vol) and simulation data without these effects (ideal -Vol), or o on the simulation data taking into account the physical effects to be corrected (Sim-Vol) to determine corrected simulation data (Sim-Vol-K) and on the simulation data without these effects (Ideal-Vol) to determine corrected simulation data (Ideal-Vol-K), and determination of the final correction data (real artifact) by means of differences Calculation of the reference between (Sim-Vol-K) and (Ideal-Vol-K), and application of the final correction data (Real-Artifact) to the measurement data (Real-Vol) to determine the corrected measurement data (Korr-Vol) or - B) Application of the transformation rule to the simulation data without taking into account the physical effects to be corrected (Ideal-Vol) to determine the corrected measurement data (Corr-Vol). Procedure according to Claim 29 , characterized in that all steps carried out from the following list are carried out on the respective radiographic image data or the respective volume data or the respective surface point data: - calculation of the transformation rule (distortion MAP), - difference formation between the simulation data (Sim-Vol) and (ideal- Vol), - Application of the transformation rule to the preliminary correction data (Sim-Artifact), - Application of the transformation rule to the simulation data (Sim-Vol) and / or (Ideal-Vol), - Difference formation between the corrected simulation data (Sim-Vol-K ) and (Ideal-V ol-K), - Application of the final correction data (Real-Artifact) to the measurement data (Real-Vol) to determine the corrected measurement data (Korr-Vol) - Use of the corrected simulation data (Ideal-Vol-K ) as corrected measurement data (Corr-Vol). Procedure according to Claim 29 or 30th , characterized in that the final correction (real artifact) is applied to the measurement data (real vol) by subtracting or adding the final correction data (real artifact). Method according to at least one of the preceding Claims 29 to 31 , characterized in that the transformation rule assigns the gray values of the volume data of the simulation with artifacts (Sim-Vol), gray values of the volume data of the measurement data (Real-Vol) as a function of location, and preferably changes them, for example in the form of an affine mapping and / or non-affine mapping and / or a look-up table (LUT), with the transformation rule preferably mapping the shift of the gray value progressions assigned to the surface transitions by rearranging the gray values. Method according to at least one of the preceding Claims 29 to 32 , characterized in that the transformation rule is determined by optimizing a cost function, this cost function from several individual terms, such as equality metrics and / or functions thereof and / or restricting boundary conditions. Method according to at least one of the preceding Claims 29 to 33 , characterized in that the finally corrected measurement data of the computed tomography (Korr-Vol) are used to determine features in the workpiece interior of the workpiece, such as inclusions or voids, in that the finally corrected volume data are evaluated and / or used to determine dimensional dimensions of the workpiece by evaluating the surface point data. Method according to at least one of the preceding Claims 29 to 34 , characterized in that the method is used in a computer tomograph designed as a coordinate measuring machine, which is designed to determine dimensions on workpieces from the surface point data, that is to say to link surface points to dimensions. Method for correcting image data of a workpiece recorded with an imaging device, such as an optical sensor, preferably an optical sensor in a coordinate measuring device, or radiation-based inspection and / or measuring device, in particular computer tomograph, the image data being provided by the imaging device in at least a first quality (low quality), such as first resolution (low resolution) or first sharpness (low sharpness) are generated, characterized in that by means of a correction algorithm the image data of low quality at least in areas in image data of high quality, preferably high resolution (high resolution) and / or high sharpness (high sharpness), are converted, wherein the correction algorithm is based on a learned relationship between image data of low quality and image data of high quality, and wherein the correction algorithm for learning the relationship between higher and lower image data Quality (image pairs) of an identical scene or several identical scenes of at least one section of the workpiece to be measured and / or other workpieces of the same or different shape in relation to the workpiece to be measured are made available. Procedure according to Claim 36 , characterized in that the image data of low quality are recorded in a first operating mode of the device, for example when compared to a second operating mode - lower imaging scale of an optics and / or - lower imaging scale of the computer tomograph and / or - image recording during the movement of the measuring axes The coordinate measuring device containing the optical sensor and / or - image recordings during the rotary movement of the turntable of the computer tomograph and / or - image recordings with the first exposure with the computer tomograph, such as the first voltage and / or current and / or power and / or spectrum and / or exposure time per image and / or number of image overlays per image and / or first measurement mode of the computer tomography, such as cone beam tomography and the high quality image data in a second operating mode with the identical device or another device, preferably with otherwise the same imaging properties as of the device whose image data are to be corrected, for example in the case of - compared to the first operating mode - larger imaging scale of an optical system and / or - larger imaging scale of the computer tomograph and / or - image recording while the measuring axes of the coordinate measuring device containing the optical sensor are at a standstill and / or - image recordings during the standstill of the rotary movement of the turntable of the computer tomograph and / or - image recordings with a second exposure with the computer tomograph, such as second voltage and / or current and / or power and / or spectrum and / or exposure time per image and / or number of Image overlays per image and / or second measurement mode of the computer tomograph, such as helix tomography. Procedure according to Claim 36 or 37 , characterized in that image data of low quality are recorded in a first operating mode of the device and high quality image data in a second operating mode of the identical device, in particular wherein the two operating modes have different exposures and / or spectra with a computer tomograph or wherein the two operating modes have different optical Have magnifications in an optical sensor. Method according to at least one of the preceding Claims 36 to 38 , characterized in that the image data of low and high quality (image pairs) are at least partially superimposed locally, with image data assigned to identical areas being used for learning the relationship. Method according to at least one of the preceding Claims 36 to 39 , characterized in that the correction algorithm for converting image data of low quality into High quality image data is generated and implemented in a self-learning manner using a machine learning process (AI), such as an artificial neural network (network), preferably convolutional neural networks (CNN), particularly preferably U-Net and / or residual neural network (ResNet) or generative Adversarial Network (GAN), where image data of low quality are used as input data for training the network and when using the network, and where image data of high quality are used as target data when training the network or as output data when using the network. Method according to at least one of the preceding Claims 36 to 40 , characterized in that an operationally trained network or a pre-trained network with additional training from partial data of a measurement of the workpiece and subsequent application of the network to the image data of low quality is used as the correction algorithm. Method according to at least one of the preceding Claims 36 to 41 , characterized in that image data of high quality and image data of low quality (image pairs), in particular low resolution, using the specific optics or other optics with the specific recording parameters that are provided for the actual optics are used for training the network to effect the optimization of the correction algorithm on the point spread function (PSF) of the respective specific optics and / or on the specific recording parameters with regard to blurring in the image data of low quality, preferably with the use of a ready-to-use network which is based on a specific Measurement task or optics has been optimized or the use of a pre-trained network with additional training on the image data with the specific optics. Method according to at least one of the preceding Claims 36 to 42 , characterized in that high and low quality image data (image pairs), in particular high and low resolution, from measurement of a specific workpiece are used for training the network, the correction algorithm being optimized for the specific measurement of this workpiece, and preferably for this purpose the use of a pre-trained network with additional training on partial data sets of the image data of the workpiece is provided. Method according to at least one of the preceding Claims 36 to 43 , characterized in that image data of high and low quality (image pairs), in particular high and low resolution or high and low contrast, are combined to form resulting image data which have the high quality, wherein - the noise is minimized, preferably to the noise of the image data higher Quality is minimized and / or - the measuring field is maximized, preferably maximized to the size of the measuring field of the image data of low quality and / or - using first and second exposures and / or spectra the irradiation of the factory back is optimized, preferably optimized for high-contrast image data and / or - the measurement time is minimized with a given improvement in the image quality. Method according to at least one of the preceding Claims 36 to 44 , characterized in that the correction algorithm is applied to any areas of the image data of low quality, in particular areas for which there is no overlap with the image data of high quality. Method according to at least one of the preceding Claims 36 to 45 , characterized in that the correction of image data of low quality takes place outside the area in which a training of the image data of high quality has taken place, preferably the training has taken place in a central area of the image data and the correction algorithm to an area outside of the central area of the Image data is applied.

Images