GB2314621A

GB2314621A - Calibrating camera image fields

Info

Publication number: GB2314621A
Application number: GB9713454A
Authority: GB
Inventors: Ernst Neumann
Original assignee: Jenoptik Jena GmbH; Jenoptik AG
Current assignee: Jenoptik AG
Priority date: 1996-06-25
Filing date: 1997-06-25
Publication date: 1998-01-07
Also published as: FR2750208A1; ITTO970549A1; IT1293383B1; DE19625361A1; ITTO970549A0; ATA70797A; GB9713454D0

Abstract

In a system for measuring the geometry of large objects such a sheets of metal using a plurality of cameras each viewing only a part of the object, the cameras are calibrated so that the relationship between the image fields of the cameras can be determined. To calibrate the cameras, a calibration pattern 33 that is considerably smaller than the large object is accurately positioned in the field of view of each of the cameras in turn using a positioning device 30. The calibration pattern 33 comprises at least one geometric element 34, each geometric element including at least one characteristic point (37, fig 5). The calibration step involves selecting and or moving the calibration pattern so that at least three characteristic points are captured in the object field of each camera. The co-ordinates of the characteristic points of the geometric elements 34 are then transformed into the co-ordinate system of the positioning device 30.

Description

2314621 PROCEDURE AND APPARATUS FOR THE GEOMETRIC MEASUREMENT OF LARGE

OBJECTS WITH A MULTI-CAMERA ARRANGEMENT The invention relates to a procedure and an apparatus for the geometric measurement of large objects, preferably for objects extending mainly in two dimensions such as large metal sheets or car-body parts. Its application is in installations which capture geometric or location- dependent primary image 10 data of objects with the aid of several cameras, and obtain geometric measurement data or location-dependent features through subsequent electronic processing. State-of-the-art technology is familiar with installations and procedures in 15 which multi-camera systems are used for the measurement and capture of large objects. Such a system is described by RAIDINGER et al in "Multi-camera systems for the inspection of large metal sheets" [Mehrkamerasysteme fCJr die Kontrolle von grogen Blechen] ("Metal sheets, Bands, Pipes" [Bleche, Bander, Rohre] 20 9-1993, pp. 36-38, Vogel Publ. & Print. Ltd [Vogel Verlag u. Druck GmbH & Co KG], Wurzburg). The system includes several positioned cameras, located at relevant contour sections, in this case at the corners. In order to be able to determine relevant object contours in relation to a unified co-ordinate system, permanent calibration is carded out through a type of object-screen which has 25 so-called pass-marks for the four cameras' entire possible working range, and which is located, e. g., beneath the object on the object carder. The disadvantage in on-line inspection is that it is necessary to ensure that the object's edges and pass-marks (object screen) do not intersect or obscure each other, for all possible deviations in object position. Other similar 30 solutions provide for suitable calibration patterns, which are adjusted to the object's size and exhibit measurement markings at the relevant contour sections, i.e. the camera positions selected. Besides the considerable size of such a calibration sometimes resulting from this approach, it is also a particular drawback that new calibration patterns always need to be provided 35 and adjusted for different objects, and hence different camera positions.

2 The invention addresses the problem of finding a new method for the geometric measurement of large objects with a multi-camera arrangement allowing calibration, which can be re-adjusted in a straightforward manner to different objects and where the calibration pattern is of simple design.

The invention proposes to achieve this through a procedure for the geometric measurement of large objects with a multi-camera arrangement, in which the number of cameras arranged in space can be freely selected in accordance with the object's size and complexity. These cameras are directed towards characteristic parts of the object, without their individual object fields necessarily covering the entire object; the metric combination of their object fields and hence image fields is achieved through a calibration process which relies on a calibration pattern considerably smaller than the object being measured, and containing geometric elements used to fix points in an object space. This pattern is displaced in defined steps, is assigned a location in a unified co-ordinate system in object space via a sufficiently precise positioning device, and this assigned location is captured and stored. The calibration pattern is positioned in at least such a number of steps that at least three characteristic points, specified by at least one geometric element of the calibration pattern, is captured in each camera's object field within one calibration process. A co-ordinate transformation then takes place, in which the co-ordinates of the geometric elements' characteristic points are converted from the calibration pattern's local co-ordinate system into the positioning device's unified co-ordinate system.

Since the local calibration pattern can take on a number of different forms, there are different variants of the positioning sequence which occur during a calibration process.

When using a small-area calibration pattern containing only one geometric element which also defines one characteristic point, it is desirable to position it a least three times, at different locations, in each camera's object field.

When using a small-area calibration pattern containing only one geometric element which defines at least three characteristic points, it is necessary to ensure that positioning occurs at least once in each camera's object field.

3 Ideally, the calibration pattern's positioning is reduced to a onedimensional movement when the pattern is essentially extended only one-dimensionally, where movement proceeds approximately orthogonally to the direction of the pattern's longitudinal extension, and three characteristic points are positioned in each camera's object field. It is necessary to ensure that at least one geometric element, with no fewer than three characteristic points, is positioned at least once in each camera's object field.

The same effect results from at least a single positioning of at least three geometric points, which each time define only one characteristic point and do not lie on a straight line.

A further variant, in the case of a long calibration pattern, results from positioning two geometric elements, each with one characteristic point, at least twice in each camera's object field.

In the case of two-dimensional measurements, the calibration pattern is appropriately guided, via an x-y positioning device, immediately above and parallel to a flat object-carrier, where the camera points essentially vertically towards the object-carrier.

In the case of three-dimensional measurements, it makes sense to position geometric elements spatially in the object-space and to arrange cameras at least pair-wise as a stereoscopic system. The generation of characteristic points in each camera's object field, arranged in at least two different planes, is best achieved with a calibration pattern containing at least one geometric element, via a (31D-) manipulator acting in three-dimensional space, which presents the calibration pattern to each camera's object field. When using the recommended calibration pattern, which contains a suitable number of characteristic points arranged in at least two planes, it is moved appropriately in two dimensions, in order to present it to the stereoscopic camera system's object fields.

As proposed by the invention, the solution involves an apparatus for the measurement of large objects with a multi-camera arrangement of cameras whose number is chosen according to the object's size and complexity, with the cameras installed on a fitting which is capable of spatial positioning, and pointing towards the object's features of interest, without their individual object 4 fields necessarily covering the entire object, as well as a device to carry the inspected object; and a calibration pattern, which can be introduced along the carrier device, used to calibrate the cameras. The task is achieved by the calibration pattern being considerably smaller than the measured object, and containing at least one geometric element for fixing points in the object space; each geometric element of the calibration pattern containing at least one characteristic point which can be simply interpolated and extracted from the camera images. Each camera can be presented with at least three characteristic points within one calibration process, spanning a sufficiently large area in the camera's object field; the calibration pattern is rigidly coupled to a sufficiently precise positioning device, through which it can be assigned a precise location in the object space along the carrier device. This assigned location is captured and stored. The system also contains a processing unit for co-ordinate transformation, with which the co-ordinates of the geometric elements' characteristic points are converted from the calibration pattern's local co-ordinates into the positioning device's uniform co-ordinate system.

Two-dimensional measurements are designed to use appropriate polygonal shapes as geometric elements. The simplest here would be a triangle whose corners serve as characteristic points. Rectangles, preferably squares, are equally appropriate as geometric elements, where the corners as well as the diagonal intersection are available as characteristic points.

One can further consider closed conic sections as geometric elements for two-dimensional measurements, where centres or foci are available as characteristic points. The preferred characteristic point used would be the centre of a circle.

Useful geometric elements for three-dimensional measurements include objects bounded by polygons, such a pyramids, truncated pyramids, cubes, cuboids etc. Equally appropriate, though with a smaller "yield" of characteristic points, are rotational bodies such as cones, spheres or rotation ellipsoids. Spheres, with their centre used as characteristic point, are the preferred option.

If polygons or objects bounded by straight lines are used as geometric elements, it is sufficient for this purpose to position only one geometric element in each camera's image field within one calibration process.

In the case of shapes or objects bounded by curved lines, and for calibration redundancy purposes, it is useful to position several geometric elements in each camera's object field. Ideally this would happen through rigid linkage of the elements during the calibration process, but can also be achieved through multiple positioning of the same geometric element via the positioning device.

If the calibration pattern in use only contains one geometric element with one characteristic point (such as a circle or sphere), it is useful if the calibration pattern is capable of being translated within each camera's object field at least twice in different directions of the plane or space, using the same scheme. On the other hand, it is best if the calibration pattern does not extend considerably beyond the area of one camera image field, contains several geometric elements and is capable of being translated two-dimensionally from one object field of the camera to another.

A further useful variant provides for a calibration pattern which essentially extends one-dimensionally across one dimension (ideally the shorter one) of the carrier device, and contains several geometric elements arranged regularly, with the calibration pattern being essentially capable of movement only one-dimensionally, perpendicular to its longest dimension. Here it is recommended for the geometric elements to be spaced equidistantly along a straight line. Preferably they would even be spaced along two parallel straight lines, possibly with the two rows of elements being displaced relative to each other (e.g. each element facing a gap).

The invention is based on the idea of calibrating a multi-camera system, used to measure large objects, step-wise with a local calibration pattern. This fundamental idea is suitable for all devices which determine object parameters and location-dependent features, both in two-dimensional and three-dimensional space.

6 The pattern used for calibration is formed of geometric elements, whose position in respect of characteristic points is known with high precision, and which permit the characteristic points to be unambiguously identified in the camera image through uncomplicated extraction and interpolation, even when image quality deteriorates due to blurred edges, for example.

Through sufficiently precise determination of the calibration pattern's spatial location, via the positioning device, the spatially-displaced characteristic points of the geometric elements can be represented in a unified co- ordinate system, and allow metric linking of the camera image fields with each other in a known fashion.

With the procedure described by the invention and the apparatus proposed for it, it is possible to perform photogrammetric measurements of large objects with a calibrated multi-camera arrangement, where the calibration pattern is designed to be simple and small and the multi-camera arrangement can be simply recalibrated as necessary, for different objects.

We further explain the invention in more detail, with the aid of one example describing a possible implementation. The diagrams show:

Fig. 1: A schematic side-view of the apparatus proposed by the invention, for continuous conveyor-belt inspection of essentially plane objects.

Fig. 2: A plan-view of Fig. 1.

Fig. 3: An object with selected object fields of the individual cameras, directed towards the object's distinctive geometric features.

Fig. 4: A design favoured by us for a calibration system, in which the calibration pattern contains only one geometric element.

Fig. 5: An electronically-combined camera image, in which a geometric element which defines exactly one point, was positioned four times in the camera object field.

7 Fig. 6: A calibration pattern with four geometric elements defining up to 20 characteristic points, with conversion of the local coordination system into the positioning device's unified coordinate system. 5 Fig. 7: A calibration pattern of essentially one-dimensional extension, which is positioned at right-angles to its longitudinal extension. Fig. 8: Another design of a one-dimensionally extended calibration 10 pattern, capable of being positioned one-dimensionally in a direction at right angles to that. Fig. 9.. Example of the design of a three-dimensional measuring apparatus based on the invention. 15 Fig. 10: One design of a three- dimensional calibration pattern.

The procedure as described the invention, consists essentially of the following steps:

- Setting up several cameras to point towards selected geometric features of the object; - calibrating the cameras in a unified object-space co-ordinate system, through:

9 Step-wise displacement of a calibration pattern, which is considerably smaller than the objects to be measured and which contains geometric elements for fixing points in object space; 0 Assigning location to the calibration pattern through the capturing and storing of a positioning device's displacement parameters; 8 executing defined displacement steps, whose number and size are so selected that at least three characteristic points, defined through at least one geometric element of the calibration pattern, are positioned in each camera's object field during one calibration process; & performing a co-ordinate transformation, in which the known coordinates of the geometric elements' characteristic points are converted from the calibration pattern's local co-ordinate system into the positioning devices' unified system; determination of geometric features (dimensions, distances, angles, linearity deviations etc.), from the actual camera images of presented objects.

Below, the steps involved in the procedure described by the invention, which mainly relate to execution and design of the calibration process, are explained in the context of the apparatus proposed by the invention, without the procedure being limited to this type of installation or apparatus. Fig. 1 shows schematically the simplest case, involving the measurement of 20 plane surfaces e.g. for testing metal sheets cut to size. A conveyor belt 1 used as a carrier device for the measured objects (here metal sheets) is moved under a multi-camera arrangement 2 which can be freely adjusted. The multicamera arrangement 2 only becomes capable of measurement, in the sense of precise photogrammetric measurement procedures, with the addition of a 25 calibration system 3, since the object fields of the individual cameras in the multi-camera arrangement 2 do not overlap (or only do so in special and exceptional cases). Hence the calibration system is the key to a metric linkage of the camera images with each other, as well as to calibration of the measurements within 30 each individual image field. Fig. 2 shows in plan view the sheet-metal testing system sketched in fig. 1. A metal sheet 4 which is to be tested (being an example for an arbitrary twodimensional object) is transported on conveyor belt 1. The presence of metal sheet 4 in the exposure domain of electronic cameras 16 to 21 is signalled by 35 an arrangement of detectors, and an exposure is triggered.

9 Cameras 16 to 21 can be positioned (e.g. pair-wise) on transverse guides 7, 8, 9 across conveyor belt l's direction of travel, where transverse guides 7 to 9 in turn can be displaced via runners 10 to 15 on longitudinal guides 5 and 6.

This displacement-capability of cameras 16 to 21 serves to select relevant object contour domains. In the case of metal sheet 4, the example shows the corners and a section of the long edge being selected. The number of cameras can also be freely selected in accordance with the task involved. As a rule, for a rectangular metal sheet which is only being tested for conformity of edge dimensions, the four cameras 16, 17, 20 and 21 suffice. Cameras 18 and 19 serve to capture edge straightness and to support rectangularity tests, if any.

Fig. 3 shows, in this context, the arrangement of object fields 22 to 27 of cameras 16 to 21, relative to sheet 4's corners and edges.

A current state-of-the-art calibration system, not shown in fig. 2, would consist of a precisely-dimensioned metal sheet 4 acting in conjunction with an object screen (e.g. pass-marks on the object carrier), which in the case of large metal sheets 4 would result in the need to move unwieldy calibration patterns.

In contrast, the invention goes one step in a new direction, shown schematically in fig. 4.

Calibration system 3, as proposed by the invention, consists of a positioning device 30 with a small-area calibration pattern 33 (i.e. substantially smaller than the measured object's dimensions).

Fig. 4 shows that the positioning device 30 allows, as a rule, twodimensional positioning of calibration pattern 33 for two-dimensional measurements, implemented via longitudinal guides 28 and via transverse guide 29 coupled through runners 31.

Thus, positioning device 30 allows the calibration pattern 33 to be precisely positioned as required in the entire object space covered by cameras 16 to 21, through object fields 22 to 27. Hence, independently of the camera system's format setting, the invention proposes that calibration pattern 33 be positioned step-wise in the object fields 22 to 27 of each camera 16 to 21.

This occurs in the calibration pattern 33 used in fig. 4, which only contains one geometric element 34 (a circle in this example) with only one characteristic point (its centre), by the calibration pattern 33 being positioned at least three, or preferably four times, in each camera object field 22 to 27. This results in the generation, in each (electronically combined) camera image 36, as shown in fig. 5, of circular images, from which each of the centre points 37 to 42 can be interpolated and extracted with the desired accuracy, such that in this specific example, four points not lying on a straight line are available for calibration within camera images 36, and for camera images 36 in relation to each other. Calibration, in this case, proceeds completely via the positioning device and the location assignment in its unified co-ordinate system.

Fig. 6 shows another variant. In this case, calibration pattern 33 consists of geometric elements 34 in the form of quadrangles, preferably squares 41 to 44, which each provide as characteristic points the diagonal intersections 45 and corner points 46. In this case, camera image 36 already shown in fig. 5 would contain the entire calibration pattern 33 (or at least all geometric elements 34 in the form of squares 41 to 44) from fig. 6 within the one photograph.

Even with the imaging of only a single square 41, as long as its dimensions are sufficient, a sufficient quantity of characteristic points is already available, which allow calibration once they are presented once to each camera object field. Hence, the calibration pattern as shown in fig. 6 allows redundant calibration to be achieved.

On the other hand, one may of course only select certain characteristic points from geometric elements 34 (e.g. the diagonal intersections 45 of squares 41 to 44), and use the remaining points (in this case, therefore, the corner points 46) for improved interpolation and extraction of the selected diagonal intersections 45.

Based on this last selection of characteristic points, the calibration process sequence is as follows.

At the start of a camera calibration, e.g. camera 16 in fig. 2, calibration pattern 33 is so positioned in object field 22 of this camera 16, that all squares 41 to

44 are captured simultaneously. The diagonal intersections 45 are defined as the characteristic points of calibration pattern 33, through the location of the geometric elements 34, which in this case are the squares 41 to 44. The diagonal intersections 45 are precisely known in the local co-ordinate system (x', y') of the calibration pattern 33 with its origin 50, so that transformation of co-ordinates x' and y' into the unified co-ordinate system x, y of the positioning device 30 with its origin 51 can proceed in a straightforward manner, by adding the co-ordinate differences Ax and Ay between the co ordinate origins and 51 to the local co-ordinates x' and y' of the diagonal intersections 45.

The coordinate differences Ax and Ay are derived here from the actual position of calibration pattern 33 within the unified co-ordinate system (x, y) of the positioning device 30.

The geometric elements 34 captured by camera 16 allow sufficiently precise extraction of the fixed characteristic points. These points serve to calculate the transformation rule between image and object co-ordinates in a known manner, through which a calibration of the cameras to the co-ordinates is achieved.

Fig. 7 shows an implementation of the calibration procedure described in the invention, in which calibration pattern 33 only needs to be positioned in one dimension. Two guides 28 are provided for this purpose; on them move two runners 31, and between them is rigidly fitted a long itud ina lly- extended, essentially one-dimensional calibration pattern 33. The special feature of this calibration pattern 33 consists in a linear arrangement of geometric elements 34, which must be designed to be so tightly spaced that at least three characteristic points not lying on a straight line can be positioned in image fields 22 to 27 of cameras 16 to 21, in every conceivable position of the latter.

This happens best with geometric elements 34 in the form of circles, as shown in fig. 7, where the distance between circles must be smaller than the size of the camera's object field, so that at least two circles lie simultaneously in one object field. In addition, this optimal variant requires a second positioning per camera object field. Appropriately to this method, the circles are arranged to be equidistant.

When using a series of geometric elements 34 which define at least three characteristic points, as for example triangles, rectangles, squares etc., it suffices if at least one geometric element 34 can be positioned in each object field. This is regularly ensured by again having the spacing smaller than the size of the camera object field. Higher density, or multiple positioning within one camera image field, lead to redundancy and improved accuracy, as already mentioned.

12 Returning now to circles as geometric elements 34, and if one wishes nevertheless to avoid multiple positioning within one camera object field, Fig.

8 provides a suitable solution for the design of a calibration pattern 33. The rows of circles, arranged in parallel, are either exactly aligned, so that each camera is presented with at least one square formed of four circles when the calibration pattern 33 is correspondingly positioned, or they are parallel but displaced relative to each other, so that at least one configuration of circles arranged in a triangle is visible in the camera's object field. Remembering that the position of the geometric elements 34 (here circles) can only be altered each time one-dimensionally and essentially at right angles to the longitudinal extension of calibration pattern 33, the co-ordinates of the characteristic points (here circle centres) are converted into the unified co-ordinate system (x, y) through a simple transformation of the local co-ordinate direction X' + AX = X (with y'= y).

As an extension of the apparatus and sequences described above, fig. 9 shows a device for capturing the spatial position of a body, specifically a car body 64 shown in stylised form, before the technical process of underside treatment. This treatment is to be performed with a robot arm (manipulator 65). Therefore individual work steps must be learned, by guiding the manipulator 65 to the corresponding position. The manipulator 65 moves during this process in a unified co-ordinate system (x, y, z) whose origin is 69.

So that deviations of a car body 64 from a model body used during the learning process do not lead to faults in the finishing process, they must be taken into account (calibrated) in the positioning of manipulator 65 during the finishing process.

In order to make spatial position photographs, cameras 58 to 63 (at least) are arranged pair-wise as stereoscopic systems and directed towards details of interest in the object (car body 64). In the example, two cameras at a time capture reference points 66, 67 or 68 in the body 64. Cameras 58 to 63 are calibrated for manipulator 65's unified co-ordinate system.

In accordance with the procedure described in the patent, calibration proceeds through the manipulator 65 driving a 3D calibration pattern 33 into the relevant object field of the camera pairs (or in the general case, a camera group consisting of several cameras). The position of the reference points 66, 13 67 and 68, for example (in the case of the current body 64) is used to locate the calibration pattern 33. This spatial location of the calibration pattern 33 is known via the setting of manipulator 65, in its own co-ordinate system (x, y, z). In accordance with the invention, transformation of the characteristic points of the geometric elements of calibration pattern 33 into the unified co- ordinate system (x, y, z) is based on the position of calibration pattern 33.

The 3D-calibration pattern 33, introduced above as a "black box" for three dimensional measurements, is shown in a recommended implementation in fig. 10. It consists, ideally, of a plane base-plate 70, to which the geometric elements are attached in the form of spheres 71 on orthogonal rods 72 of different lengths. Spheres 71 define, through their centres, characteristic points in various planes within the local co-ordinate system (x', y', z') of calibration pattern 33 with its origin at 73. Transformation of the local co ordinates x', y', z' proceeds in the same manner as in the two- dimensional measurement rigs described above.

In the event that the measured objects, as e.g. in the case of car body 64's underside, are nearly flat objects with insubstantial projections and recesses compared with their basic shape, it can also be appropriate to move the calibration pattern 33 in two dimensions only (in our case, for example, exclusively with its base-plate 70 along a parallel to the y-z plane). Plane surfaced bodies are suitable as geometric elements 34, again in analogy to the 2D-case, where in order to utilise the corners as characteristic points such bodies should be used, as far as possible, in which the same corners are simultaneously and ambiguously visible to all the cameras in one assigned group (e.g. camera pair) through their stereoscopic layout. For example, right pyramids and truncated pyramids are suitable here. If it is desired to achieve calibration redundancy with rotational bodies or to represent the required set of calibration pattern 33's characteristic points with fewer geometric elements, then (again inanalogy to the 2D-calibration pattern 33) one can utilise the foci or centres of rotation ellipsoids but also the apex, apex foot or centre of the base of right circular cones or truncated cones.

14 The nature of the geometric elements 34, in turn, requires within all camera object fields either multiple positioning (e.g. with spheres, ellipsoids, cones, i.e. with bodies which individually do not provide the required number of characteristic points), or single positioning of the calibration pattern 33 when a geometric element 34 on its own defines more than three characteristic points, or the characteristic points are provided in the required number through several geometric elements 34 rigidly connected through the calibration pattern 33.

If one selects suitable plane surfaces as geometric elements 34 and arranges them in different planes (without the surfaces overlapping in the camera object field), they can be equally used in a 3D-calibration pattern 33 instead of the spheres 71 of fig. 10, by attaching them to rods 72 parallel to the base plate 70. The use of circles or squares is recommended.

Although in principle, selection of the appropriate geometric elements within a calibration pattern 33 is unrestricted, one should decide on one shape for reasons of interpolation and extraction of the characteristic points from the photographs.

Equally, redundant calibration through single positioning via characteristic points rigidly connected within a calibration pattern 33 (regardless of whether this is defined through one or several geometric elements), is preferable to multiple positioning of a single point or two characteristic points within the camera object field.

With the procedure described by the invention and the proposed apparatus, it is possible to calibrate with simple means which involve a specially designed, local calibration pattern 33 and a positioning device with defined and sufficiently precise camera location, a multi-camera layout for demanding photogrammetric measurement tasks, without its object fields overlapping or the calibration pattern having to be simultaneously visible to all cameras.

IS,

Claims

CLAIMS:

1 Procedure for the geometric measurement of large objects with a multi camera arrangement, in which any number of cameras, chosen according to the object's complexity and size, are spatially arranged, where the cameras are directed towards characteristic parts of the object in an object space, without their individual object fields necessarily covering the entire object and the metric linkage of their object fields and hence of their image fields are achieved through a calibration process having the following features:

- A calibration pattern (33), considerably smaller than the measured object and containing geometric elements (34, 41, 42, 43, 44, 71) for fixing points in object space, is displaced in defined steps; - using a sufficiently precise positioning device (30, 65), the calibration pattern (33) is located in a unified co-ordinate system in object space and this location is captured and stored; - the calibration pattern (33) is positioned in at least such a number of steps, that no fewer than three characteristic points (37, 38, 39, 40, 45, 46), specified by at least one geometric element (34, 41, 42, 43, 44, 71) of the calibration pattern (33), are captured within one calibration process in the object field (22, 23, 24, 25, 26, 27) of each camera; and - a co-ordinate transformation takes place, in which the co-ordinates of the characteristic points of geometric elements (34, 41, 42, 43, 44, 71) are converted from the local co-ordinate system of the calibration pattern (33) into the unified co-ordinate system of the positioning device (30, 65).

2. Procedure in accordance with claim 1, with the following features:

The calibration pattern (33), as a local calibration pattern (33) with only one geometric element (34) which defines exactly one characteristic point (37, 38, 39, 40), is positioned in the object field (22, 23, 24. 25, 26, 27) of each camera at least three times at different locations which do not lie on a straight line.

16

3. Procedure in accordance with claim 1, with the following features:

The calibration pattern (33), as a local calibration pattern (33) with only one geometric element (41, 42, 43, 44) which defines at least three characteristic points (45, 46), is positioned at least once in the object field (22, 23, 24, 25, 26, 27) of each camera.

4. Procedure in accordance with claim 1, with the following features:

The calibration pattern (33), as an essentially one-dimensionally extended calibration pattern (33) with several geometric elements (34), is moved essentially one-dimensionally at right angles to the direction of its longitudinal extent, where as a result of the positioning movement at least three characteristic points (37, 38, 39, 40) are generated in the object field (22, 23, 24, 25, 26, 27) of each camera by geometric elements (34).

5. Procedure in accordance with claim 4, with the following features:

At least one geometric element (41, 42, 43, 44) with at least three characteristic points (45, 46) is positioned at least once in the object field (22, 23, 24, 25, 26, 27) of each camera.

6. Procedure in accordance with claim 4, with the following features:

At least three geometric elements (34) with at least one characteristic point each (37, 38, 39 or 40 as relevant) are positioned at least once in the object field (22, 23, 24, 25, 26, 27) of each camera.

7. Procedure in accordance with claim 4, with the following features:

At least two geometric elements (34) with one characteristic point each (37, 38, 39, 40) are positioned at least twice in the object field (22, 23, 24, 25, 26, 27) of each camera.

8. Procedure in accordance with claim 1, with the following features:

In two-dimensional measurements, the calibration pattern (33) is guided directly above and parallel to a plane object carrier (1) via an x-y positioning device (30), with the cameras directed essentially at right angles to the object carrier (1).

1-7

9. Procedure in accordance with claim 1, with the following features:

In three-dimensional measurements, calibration pattern (33) is used to position spatially-arranged geometric elements (71) in object space, and the cameras are installed at least pair-wise as a stereoscopic system.

10. Procedure in accordance with claim 9, with the following features:

Calibration pattern (33), consisting of at least one spatial geometric element (71), is presented via a 3D-manipulator (65) to camera object fields (22, 23, 24, 25, 26, 27).

11. Procedure in accordance with claim 9, with the following features:

Calibration pattern (33), containing characteristic points arranged spatially in at least two planes, is moved, and presented to the stereo camera system, two-dimensionally in a plane in object space.

12. Apparatus for the geometric measurement of large objects with a multi camera arrangement, containing a number of cameras selected in accordance with the object's size and complexity, with the cameras, attached to a fitting and capable of being spatially positioned, directed towards the object's features of interest, without their individual object fields necessarily covering the entire object, as well as c, carrier device for the examined object, where a calibration pattern (33) is capable of being moved along the carrier device for calibrating the geometric distance relationships between the camera image fields, with the following features:

- Calibration pattern (33) is considerably smaller than the measured object and contains at least one geometric element (34, 41, 42, 43, 44, 71) for fixing points in object space; - each geometric element (34, 41, 42, 43, 44, 71) of calibration pattern 30 (33) contains at least one characteristic point (37, 38, 39, 40, 45, 46) which can be simply interpolated and extracted from the camera images, where it is possible to present at least three characteristic points (37, 38, 39, 40, 45, 46), which span a sufficiently large area in the camera's object field (22, 23, 24, 25, 26, 27), to each camera 35 within one calibration process, 1 13 calibration pattern (33) is rigidly coupled to a sufficiently precise positioning device (30, 65), with which calibration pattern (33) can be assigned a precise location in object space and this assigned location is captured and stored; - a processing unit for co-ordinate transformation, with which the characteristic points (37, 38, 39, 40, 45, 46) of geometric elements (34, 41, 42, 43, 44, 71) are converted from the local co-ordinate system of the calibration pattern (33) into the unified co-ordinate system of the positioning device (30, 65).

13. Apparatus in accordance with claim 12, with the following features:

Polygonal shapes are used as geometric elements (34, 41, 42, 43, 44, 71) for two-dimensional measurements.

14. Apparatus in accordance with claim 13, with the following features:

The geometric element (34, 41, 42, 43, 44, 71) is a triangle, with the corners used as characteristic points.

15. Apparatus in accordance with claim 13, with the following features:

The geometric element (34, 41, 42, 43, 44, 71) is a rectangle, preferably a square, with the corners (46) and diagonal intersection (45) being available as characteristic points.

16. Apparatus in accordance with claim 12, with the following features:

The geometric elements (34, 41, 42, 43, 44, 71) for two-dimensional measurements are closed conic sections, with the centres and foci being available as characteristic points.

17. Apparatus in accordance with claim 16, with the following features:

The geometric element (34) is a circle, preferably a square, with the centre (37, 38, 39, 40) used as characteristic point.

j

18. Apparatus in accordance with claim 12, with the following features:

The geometric elements (34, 41, 42, 43, 44, 71) for three-dimensional measurements are bodies bounded by straight lines.

19. Apparatus in accordance with claim 12, with the following features:

The geometric elements (34, 41, 42, 43, 44, 71) for three-dimensional measurements are rotational bodies.

20. Apparatus in accordance with claim 19, with the following features:

The geometric element (34) is a sphere (71).

21. Apparatus in accordance with claim 12, with the following features:

The geometric elements (34, 41, 42, 43, 44, 71) for three-dimensional measurements are plane surfaces, arranged in different planes, preferably parallel to a base-surface of the calibration pattern (33).

22. Apparatus in accordance with one of the claims 13, 14, 15 and 18, with the following features:

Only one geometric element (34, 41, 42, 43, 44, 71) at a time can be positioned in the object field (22, 23, 24, 25, 26, 27) of each camera within one calibration process.

23. Apparatus in accordance with one of the claims 13 to 21, with the following features:

Several geometric elements (34, 41, 42, 43, 44, 71) at a time can be positioned in each camera's object field (22, 23, 24, 25, 26, 27) within one calibration process, in order to generate at least three characteristic points (37, 38, 39, 40, 45, 46) per camera object field (22, 23, 24, 25, 26, 27).

24. Apparatus in accordance with claim 12, with the following features:

Calibration pattern (33) contains only one geometric element (34, 71), which defines only one characteristic point (37, 38, 39, 40), with the calibration pattern (33) being capable of displacement at least twice in different directions within each camera's object field (22, 23, 24, 25, 26, 27).

25. Apparatus in accordance with claim 12, with the following features:

Calibration pattern (33) does not essentially extend beyond the area of one camera image field (36), contains several geometric elements (34, 41, 42, 43, 44, 71) and is capable of two-dimensional displacement from one camera object field (22, 23, 24, 25, 26, 27) to another.

26. Apparatus in accordance with claim 12, with the following features:

Calibration pattern (33) essentially extends one-dimensionally over one direction of the carrier apparatus (1) and contains several geometric elements (34) arranged in a regular layout, where calibration pattern (33) is essentially only capable of positioning one-dimensionally at right angles to its longitudinal extent.

27. Apparatus in accordance with claim 26, with the following features:

The geometric elements (34) are arranged equidistantly along a straight line.

28. Apparatus in accordance with claim 26, with the following features:

The geometric elements (34) are arranged equidistantly along two parallel straight lines, with the rows of geometric elements (34) displaced relative to each other.