DE102005052044A1 - Transparent object measuring device e.g. for defining jet level production mechanism, has first profile line at front surface of transparent object and second profile line at rear surface with radiation provided from jet - Google Patents

Abstract

The device has a first profile line (54) at a front surface (28a, 28c) of a transparent object (24) and a second profile line at a rear surface (28b) with radiation provided from a jet (14). A mechanism (18, 16) receives radiation from a pre-determined line of sight from the profile lines diagonally to the jet level, in order to receive a profile line of the profile lines. A computer (44) evaluates the profile line admission to receive a measurement result for the transparent object. The result is related to the front surface and the rear surface of the transparent object. An independent claim is included for a method, and a computer program.

Description

The present invention employs deals with the measurement of transparent objects, e.g. of bottles and in particular bottled bottoms.

The Production of glass bottles for Food, chemicals or pharmaceuticals require one high production quality and high process reliability. Compliance with the customer specifications by the products is of high importance for the customer and therefore also for the manufacturer. For this reason, were online inspection systems within quality assurance increasingly popular, since they allow the manufacturer perform a 100% test of their products. Due to the speed requirements From 3 to 10 objects per second, contactless high-speed solutions are required.

It There are different possibilities the review of bottle bottoms perform. However, most of them do not allow the desired 100% compliance test the simultaneously desired precision. An optical interferometric pointwise measurement is for a planar Measuring the bottle bottoms too slow. Tomographic procedures for inspection or application enable coded light also not high processing speeds. Normal transmitted light method, the 2-D information collect, allow Although the detection of unwanted inclusions of bubbles in the glass, allow but not the review of others important parameters for their compliance with specifications, such as. the review of Maintaining a minimum floor thickness, a maximum floor thickness, a maximum skewness of the inner bottom, a maximum penetration depth, a maximum puncture depth or a maximum depth of the puncture or thickness, planarity and parallelism specifications.

It That would be why desirable, to have a corresponding measurement method for transparent objects, that makes it possible quickly obtain relatively accurate survey results from transparent objects received referring to the external and inner surface or refer to a front and a back surface of the object, such as e.g. the floor thickness or the other criteria mentioned above.

The The object of the present invention is therefore a method and a device for measuring a transparent object to create that it allows fast, relatively accurate measurements of transparent objects obtained, which refer to both the front and the rear surface.

These The object is achieved by a device according to claim 1 and a method according to claim 16 solved.

A inventive device for measuring a transparent object comprises a beam plane generating device for generating a beam plane intersecting the transparent object a first profile line on a front surface of the transparent object and a second profile line on a rear surface of the transparent object to define at the radiation of the beam plane is scattered. A Receiving device is provided to the at the first and the second profile line scattered radiation from a predetermined Viewing direction at an angle to record the beam plane to a profile line recording the to receive first and second profile line. Furthermore, the device comprises an evaluation device for evaluating the profile line recording, based on the same a survey result for the transparent Object obtained on the front surface and the back surface of the transparent object is related.

A Knowledge on which the present invention is based, consists in this case in that the measurement of transparent objects with respect to their front and back surface for example, in bottle bottoms, because the surfaces often such that they are not just a total reflection or enable a specular reflection but also a diffuse reflection or scattering. by virtue of Transparency is made possible that a residual, not diffusely scattered light, even the rear area can reach, so that by receiving obliquely to the beam plane by means of Triangulation related to the front and back surface Survey variables determined can be.

In order to Fulfills the present invention all the criteria required for quality inspection in Met mass production should be. First the exam takes place or measurement without contact, which facilitates installation and installation in a production line. For one areal Surveying transparent objects is indeed a relative movement between Radiation level and transparent objects necessary, but mostly have to the ones to be measured or tested transparent objects anyway in a conveyor belt of a processing unit be transported to the other, so that there is no additional effort accrues.

According to one embodiment According to the present invention, the beam plane is perpendicular to the direction of relative movement aligned between beam plane and transparent object while the Pixel array is arranged such that a line perpendicular to Relative movement direction and the irradiation direction substantially is parallel to a pixel row of the pixel array. In this way it will enable the height information and the lateral, i. perpen- dicular to the direction of relative movement and to the direction of irradiation defined, location information on separately divide the row and column coordinates within the pixel array. A seeker of the profile lines in the pixel array brightness image Unity can therefore be limited to a column-by-column investigation.

According to one another embodiment The present invention provides an extraction device used for each pixel column to extract a first and a second pixel position who are candidates for Serve places where the first or second profile line on the Pixel array is mapped. This happens, for example, by Find brightness maxima within each column using thresholds. The pixel positions thus found are in metric height values re-scaled to get height images in this way, one per channel or level. The fields can be filtered. moreover allows a manipulation of the fields in such a way that the through the Depth images represented front or back surface substantially perpendicular to a height axis of the elevation images that extends beyond the determination of a height value histogram from the position-corrected height images the altitude values the position corrected height images the front or rear surface can be assigned around the front surface representing To obtain point cloud and a point cloud representing the rear surface. The procedure is simple and thus allows the implementation of the corresponding processing steps in a sufficiently small Time, for example, to carry out the evaluation online can.

preferred embodiments The present invention will be described below with reference to FIGS enclosed drawings closer explained. Show it:

1 a sketch of a device for measuring transparent objects according to an embodiment of the present invention;

2a a sketch to illustrate the measurement setup of 1 in view of its ability to determine height;

2 B a sketch of a variation of the device of 1 in view of the structure for beam plane generation and image pickup to enable the recording from two different directions of view, according to another embodiment of the present invention;

2c a partial sectional view of a bottle bottom;

3a and 3b Gray value images, as they are with the structure according to 2 B were obtained;

4 a graph of an exemplary partial linear rescaling function for rescaling the profile line pixel positions in profile line height values;

5a a graph showing the output of the camera from the construction of 2 B represents, according to an embodiment of the present invention;

5b a graph representing the altitude values as seen from the camera output of 5a obtained by reprecipitation;

6 a flowchart for illustrating the operation of the device according to 2 B from the profile line recording on the evaluation of the same up to the receipt of the desired Vermes sungs results according to an embodiment of the present invention;

7 a flowchart illustrating the steps for noise removal according to an embodiment of the present invention;

8a a graph showing a histogram after that for noise removal 7 is determined and used;

8b a graph showing a histogram that assigns the height values to the front or back surface in 6 is determined and used;

9a and 9b Gray value representations of a height image, as it is in accordance with the structure of 2 B has been obtained before and after the noise removal 7 ;

10 a flowchart illustrating the steps, as for superimposing the height images of the different views in 6 be carried out according to an embodiment of the present invention;

11 a flow chart illustrating the steps as shown in 6 for leveling the height images, according to an embodiment of the present invention;

12 a spatial representation of the point clouds for the bottom bottom and bottom inside, as reflected in the method 6 yield; and

13 Graphs illustrating the results of reproducibility tests, which at an Implementation of a device according to 2 B have been carried out;

Even though in the following preferred embodiments closer to the present invention be noted that the same or functionally identical Elements in the figures are provided with the same reference numerals, and that a repeated description of their function is omitted. It should also be noted that, although the following embodiments with the measurement of bottle bottoms employ, but that the present invention also to the measurement of others transparent objects, such as e.g. Glassware and the like. Furthermore, the present invention is not limited to those shown embodiments limited, as will be discussed after the figure description.

1 shows a device for measuring and subsequent classification of bottle bottoms. The device comprises a laser diode 10 and a cylinder lens disposed in front of it 12 for generating a light fan 14 , Furthermore, the device comprises a pixel array, such as a CCD matrix camera, for example. 16 and an optic, such as a lens, 18 for mapping a relevant section 20 of the fan of light 14 or an inspection volume on the pixel array 16 , A conveyor 22 moves the bottles to be measured 24 in a direction of movement 26 such that the bottle bottoms 28 the image area or the relevant section 20 happen, in such a way that the bottom of the bottle 28 each the light fan 14 traverses in an orientation at the bottom of the bottle 28 except for operational inaccuracies substantially perpendicular to the light fan 14 stands. In other words, the conveyor carries or holds 22 the bottles to be measured 24 in an orientation where a bottle axis 30 essentially parallel to a direction of light incidence or an irradiation direction 32 is. pixel array 16 and lens 18 are arranged to the rele vant section 20 along a line of sight 34 to record obliquely to the plane of the fan of light 14 is and lies in a plane that is through the line of sight 32 and the direction of movement 26 is spanned. Preferred angles will follow below. The pixel array 16 is arranged in the image plane to which the lens 18 the section of interest 20 of the fan of light 14 maps, in such a way that a column direction 36 the pixel 16a of the pixel array 16 extending in a plane passing through the direction of movement 26 and the direction of irradiation 32 is spanned while the row direction 38 of the pixel array 16 is aligned perpendicular to this plane.

A rotary encoder 40 is with the sponsor 22 coupled to the progression or position of the bottles to be measured 24 along the direction of movement 26 to capture and at equidistant intervals, ie, if the conveyor 22 again a predetermined increment has covered, over a line 42 one Trigger signal to the pixel array 16 output to the pixel array 16 with the movement of the conveyor 22 to synchronize. The pixel array 16 synchronizes his pictures with the trigger signals on the line 42 and outputs the output signal representing the respective image recording to a control unit 44 from, for example, a computer on which, inter alia, an evaluation program runs, which will be described in more detail below, and the actual surveying tasks takes over, and other control programs, such as those based on the survey results compliance with certain specifications for the bottom of the bottle 28 and those that cause a rejection of defective bottles based on the result of the check.

After the construction of the device of 1 are described below, first by means of 1 Described the tasks and problems that the device of 1 has to cope with. In particular, it is the task of the device of 1 made of glass bottles 24 to inspect in a fast and contactless manner. More specifically, the bottle bottoms are supposed to 28 be checked. These are characterized by three surfaces, namely the ground contact surface or the footprint 28a , the inner bottom or the inner bottom surface 28b and the puncture or the puncture area 28c , These areas must fulfill certain criteria, as will become clearer in the following. The device takes over from 1 the detection of the actual values, whereas the control device 44 takes over the comparison of the actual values with nominal values and adopts the corresponding measures in the event of non-compliance.

An example of a power request is, for example, 5 bottles per second. This requirement is met by the device of 1 by using a 3D range sensor that is made up of the most competent 10 . 12 . 16 . 18 . 40 and 44 composed and the floors 28 scans while the bottles 24 the light fan or the light plane 14 happen, for example, a resolution of 1 mm in the transport direction 26 and a resolution of 0.10 mm perpendicular thereto can be achieved.

The following will now be based on 1 First of all, the measuring principle roughly explained. The measuring principle for detecting the 3D shapes of the floors 28 is based on laser triangulation or light-sheet or sheet-off-light distance measurement. To illustrate this, shows 2 a side view of the illumination and image receiving part of 1 from a direction perpendicular to the exposure direction 32 and the direction of movement 26 spanned level. As you can see, inscribed in 2a the angle α is the angle between the viewing direction and the optical axis 34 which is out of the lens 18 and the pixel array composing camera 46 and the lot 48 from the optical center of the lens 18 to the light level 14 , The distance of the optical center of the lens 18 from the light plane 14 is labeled B. With b is the image-side distance of the pixel array 16 from the optical center or the main plane or the image-side main plane of the objective 18 designated. The optical axis 34 the camera 46 cuts the light plane 14 in one point 50 around which the relevant section 20 the light plane 14 extends the camera 46 on it or sharp on the pixel array 16 maps. With r is now in 2a a length that is different from the point 50 in the direction of exposure 32 extends. r is also referred to below as the height value. The height values r thus indicate at which point or at which height r the object to be measured r 24 at a certain position in the direction of movement 26 the illumination direction or illumination axis 32 cuts. Similarly, with surface points of the object to be measured 24 Procedures that are located outside the plane, through the axes 32 and 34 is spanned. For them, the height value r is also parallel to the illumination axis 32 determined, and indeed from the line, through the place 50 runs and perpendicular to the axes 32 and 34 stands.

As will be discussed below, the shots are taken by the camera 46 or the individual pictures of the camera 46 formed by respective brightness value arrays corresponding to the image of the region of interest 20 correspond to the light fan 14 is projected. An extraction device discussed in more detail below 52 that either, as in 1 indicated in the control device 44 or, as in 2a indicated in the camera 46 itself, can be arranged, but extracts from the brightness images the information more relevant to the further measurement method by determining column-by-pixel positions as candidates for locations at which the images are irradiated by the light fan 14 on the object to be measured 24 generated laser or profile line 54 through the lens 18 on the pixelarray 16 is shown. This information forms the actually relevant output signal of the camera 46 or the institution 52 where the candidate pixel positions are given, for example, in pixel line numbers, or, as shown in FIG 2a is indicated, in units of pixel lines as an offset to the optical center 56 of the pixel array 16 or to the pixel line passing through the optical center 56 runs, this offset in 2a with s is specified. In 1 is for example the course of the picture 58 the laser line 54 on the pixelarray 16 indicated.

Now that all relevant parameters of the measurement setup have been explained, the explanation of the measurement principle is continued. The light plane 14 passing through the laser 10 is generated, is perpendicular to the inspecting surface 28 , The diffraction projection lens 12 widens to the generation of the light plane 14 or the light fan the light beam 60 the laser 10 orthogonal to the transport direction 26 on what, for example, a light plane 14 of about 0.10 mm thickness. The camera or the sensor 46 who is the scene 20 is an angle which is in a range of 20 ° to 70 °, and more preferably in a range of 30 ° to 60 °, and more preferably about 55 °, or alternatively about 35 °, with respect to the illumination direction 32 inclined to observe a cubic inspection volume of, for example, about 50 mm edge length, which provides a theoretical resolution of 0.08 mm for the height values r determined from triangulation basis.

The profile detection depends on the diffusely reflected part of the incident light, that for the contour lines 58 in the captured image in the pixel array 16 responsible for. Looking closer at the laser-object interaction, such reflection occurs when the laser fan 14 occurs on the object surface. Depending on the opacity of the material of the object to be measured, a part of the laser light penetrates under certain refraction effects and is again diffusely reflected a second time, namely when the laser light leaves the object again. Because of these two reflections, the captured image of the pixel array contains 16 actually two laser line contours, namely one for the top and the front surface 28a . 28c and another for the lower and rear surfaces, respectively 28b , however, in 1 for clarity, only a contour line 58 for the surface or front surface 28c . 28a is shown. The depth information obtained from these two lines can be related to each other, allowing for thickness, parallelism, and planarity measurements, as described below.

When inspecting non-coplanar surfaces with an orthogonal laser projector and a tilted camera assembly, as in the construction of FIG 1 at the surfaces 28c . 28b , the lighting by laser 10 and lens 12 and the camera 46 If this is the case, occlusion effects can occur resulting in a loss of altitude or depth data. This loss can be caused, for example, by unwanted total reflections on the object to be inspected 24 arise. One possible countermeasure against such effects is to capture redundant information from different perspectives. An easy way would be to build the lens 18 and pixel array 16 in 1 symmetrical to the viewing plane 14 to provide again. In 2a this would correspond to a reflection of the camera 46 at the level 14 , This structure would be expensive, so 2 B an alternative solution is provided in which it is possible to record on one and the same pixel array two views, one on one half of the pixel array and the other on the other half. This helps to minimize costs and reduces the overhead of synchronization with a second camera implementation.

As it is in 2 B is shown, the concept of redundancy by means of a camera 46 realized by the fact that the camera 46 a system of three mirrors 62a . 62b and 62c that captures the image from two different directions 34a and 34b allows. In particular shows 2 B only that part of the measuring device which differs from the one in 1 different. The relevant part includes the illumination and image pickup pages. As you can see, the mirrors are 62a - 62c parallel to the light plane 14 arranged. Here is the mirror 62b essentially coplanar with the plane of light 14 arranged such that the light plane 14 close to the mirror 62b passes to the bottle to be measured 24 hold true. The mirror 62a and 62c are substantially axisymmetric to each other and offset to the light plane 14 arranged. The camera 46 is again below the desired inclination to the illumination axis 32 arranged, as well as at 1 the case was. Your optical axis 34 is preferably positioned so that it is on an object to be measured 24 nearest edge of the deflecting mirror 62a meets. In this way, the detectable scene ( 46 ) is divided into two parts, namely a part which receives light, that of the inspection volume 20 over the mirror 62a into the camera 46 and on the other hand light, that of the inspection volume 20 over the mirror 62c and the deflecting mirror 62b into the camera 46 think of something with one and the same camera 46 the recording or recording of the inspection volume 20 from two different directions 34a and 34b allows in the in 2 B also shown symmetrically to the light plane 14 are.

Returning to the explanation of the principle of measurement, let us refer to 2a explains how the through the pixel array 16 obtained brightness or gray value images in connection with the height information r. As it is in 1 can be seen projected the construction of diode 10 and lens 12 a thin laser line 54 on the object 28 , where the camera 18 . 16 respectively. 46 this line 54 viewed from another position, namely at the aforementioned angle of for example 55 ° to the illumination direction 32 , This leaves the laser line 54 image like the contour of a section of the object 24 appear, which is why, as explained later, from the contour line 58 obtained height information is often referred to as a profile or profile line. The 3D shape of the object 24 is then finally reconstructed profile by profile while the same camera 18 . 16 respectively. 54 and the laser setup 10 . 12 happens.

berechnet werden, die durch die Triangulationsgeometrie definiert ist, wie sie in 2a angezeigt ist:

^|Nⁿ → ^|Rⁿ : s(t) → r(t)

wobei B, wie bereits erwähnt, die Baslinie b den Abstand zwischen der optischen Mitte der Kamera 46 und dem Pixelarray bzw. der Sensorebene 16 angibt und α den Betrachtungs winkel bezüglich der Basislinie angibt, wobei ergänzend hierbei auf Ranger, I.M.: Introduction to 3-d range imaging verwiesen wird.As already mentioned, the information of interest are the profiles r (t) ε ^| R ⁿ and not the complete grayscale image I (t) ε ^| N ^{(m, n)} , where t represents the time and m, n indicate a pixel position in row and column. Consequently, the landing positions of the laser plane must 14 from the area to be measured from the gray value image of the pixel array 16 be extracted what the extraction device 52 takes over, and finally leads to the actually interesting output signal, namely the sensor offset values s _i (t) ε ^| N, which indicate a pixel position or a row of pixels within that column for a given column, possibly at the laser line 54 on the pixelarray 16 has been imaged or at the respective pixel column of the contour line 58 is cut. Once these offset values or offsets have been extracted, the corresponding height values r _i (t) representing the height of the surface to be measured 28 as in 2a displayed, defined by an illustration

which is defined by the triangulation geometry, as in 2a is displayed:

^| N ⁿ → ^| R ^n: s (t) → r (t)

where B, as already mentioned, the baseline b the distance between the optical center of the camera 46 and the pixel array or sensor plane 16 and α indicates the viewing angle with respect to the baseline, reference being additionally made to Ranger, IM: Introduction to 3-d range imaging.

As As can be seen, the choice of the baseline B and the triangulation angle α have the greatest influence on the resolution and the accuracy of the measurements, which is why great care in your selection should be applied.

As it emerged from the preceding, the actually relevant information is not in the gray scale or brightness images as they are through the pixel array 16 be generated, but in the latter by the extraction device 52 extracted pixel positions s per column. This is based on 1 explained again. Due to the structure, one column corresponds to the pixel array 16 essentially a line that is in the light plane 14 parallel to the direction of illumination 32 runs. The sequence of pixel columns thus virtually touch the laser line 54 in the transverse direction, ie perpendicular to the through the illumination direction 32 and direction of movement 26 spanned level, from. The light that is on the profile line 54 is diffusely reflected at the location that intersects one of the thus defined virtual lines, within the pixel column corresponding to that line, to a particular pixel within that pixel column through the lens 18 displayed. The through the extraction device 52 output information s indicates this pixel information for this pixel column.

Now, in the present case, the extraction means determines 52 Of course, not just one pixel position per column, because not just the contour line 58 the outer surface 28c respectively. 28a to a contour line 58 on the pixelarray 16 leads, but also on the inside 28b diffuse reflected light. It is for example in 1 assumed that the contour line 58 the image of the laser line 58 on the outside surface 28c . 28a corresponded. Then another contour line (not shown) is above, ie opposite to the direction of the arrow 36 in 1 from, the contour line 58 to be expected from the inside 28b equivalent. The extraction device 52 Therefore, it outputs at least two pixel positions for each pixel column in its output, one for the contour line 58 the front surface and another for the rear surface.

But now arise in the brightness image of the pixel array 16 except at the contour lines 58 other bright spots such as due to the above-mentioned unwanted total reflections on the considered glass bottle. According to the present embodiment, the extraction means extracts 52 Therefore, even more than two, for example, four pixel positions for each column representing candidates for locations where one of the contour lines 58 crosses the respective column. The first pixel position from the four Pi Xelpositions for each column thereby form a first channel or a first level of the output signal of the extraction device 52 , The respective second pixel position of the respective four pixel positions for each column form the second channel or the second level, etc. In this case, the pixel positions of a column in the different channels are preferably arranged in descending or ascending order. One way in which the extraction device 52 For example, the determination of the four pixel positions per pixel column involves the use of individual thresholds. In this case leads the extraction device 52 For each column, follow these steps:
It starts at the pixel position within the current pixel column, the object side to a point in the light plane 14 corresponds closest to the lighting. From there, it checks the brightness values to see if they exceed a predetermined threshold, the first brightness value that exceeds the threshold defines the first pixel position for the first channel. From this line the threshold is updated if necessary, such as reduced. The latter takes into account the fact that under normal circumstances, ie, in the absence of reflections, it is to be expected that the pixel position first found will be the contour line of the outer surface 28c . 28a which, of course, is the brightest, because of the inside 28b Diffused back reflected light only from that on the outer surface 28c . 28a transmitted light is derived. Immediately adjacent pixel lines for which the brightness values in this column exceed the updated threshold are initially ignored because it is assumed that the searched contour lines are spaced on the image side. From the line, since the brightness values in the column fall below the current threshold again, determines the extraction device 52 furthermore, for the individual pixels in the pixel column in the predetermined direction, whether one of the brightness values of these pixels exceeds the current threshold value again and, if this is the case, outputs the corresponding pixel position for the second channel for the corresponding column. Thereafter, the corresponding steps are performed again, namely the threshold value is updated, ie reduced, ignoring the brightness value exceeding the updated threshold value and the pixel position at which the brightness curve in this column exceeds the new threshold value, etc., again.

Of course, in the extraction 52 the throughput and processing speed requirements are met. One option is the IVP Ranger M50, which features an Intel StrongARM processor and a proprietary CMOS image sensor with a resolution of 1536 × 512 pixels with square pixels of 18 μm page length. A full line is even processed there in parallel: each column of this "smart vision sensor" has its own RISC processor, which very quickly searches for laser line impingement points in its assigned column, of course, there are other design possibilities, except those presented here.

As described above, the extractor extracts 52 four offset vectors from each image of the pixel array 16 , namely one offset vector per channel or plane. These offset vectors are referred to below as s ⁽ⁱ⁾ (t), i = 0,..., 3, where i indexes the channel or plane, and s is a vector having one coefficient per column. In other words, the vectors s ⁽ⁱ⁾ (t) can be represented as s (I) (t) = (s (I) 0 (t), ..., s (I) (N-1) (t)) ε | N n i = 0, ..., (l-1) (2) where s _j ⁽ⁱ⁾ (t) thus corresponds to the extracted pixel position from the image at time t in column j for channel i, and where, in the above exemplary construction, l = 4 is the number of channels and n = 1,536 indicates the number of pixel rows or pixel locations in each pixel column.

As already mentioned, from a theoretical point of view, it would be sufficient to extract only two channels or lines to achieve planarity and parallity measurements, but in practice the additional channels or lines can be used for error detection and correction purposes. In the following further description of the operation of the device of 1 It is assumed to be in accordance with the variations of 2 B is constructed, ie, such that two different views on the pixel array 16 be imaged, the two different directions 34a and 34b correspond.

As mentioned above, the sensor or pixel array becomes 16 through the external encoder 40 triggered the progression of the conveyor belt 22 what makes the measurement discussed in more detail below regardless of the different production cycles. It can also, although in

1 not shown, for example, the laser 10 through the encoder 40 be triggered, such as always be activated when the next bottle to be measured again 24 begins through the light plane 14 go through to be switched off again when this bottle is the light plane 14 leaves until the next bottle of light level 14 reached, etc.

Since the output signal of the triangulation sensor consists of lens 18 , Pixel array 16 and extraction device 52 are the laser line impact positions s (t) having non-dimensional sensor offset values s _i (t), mapping of the offset values to metric heights is still necessary, and generally, this can be done by applying Equation 1, yielding the profiles r (t). Due to design-related conditions that affect the mirror configuration, the sensor offset values are not optimally aligned to the image axes, as for example in 5a you can see the output of the extraction device 52 in an exemplary case for 3 channels, Laser Line 1, Laser Line 2 and Laser Line 3, shows. This in turn leads to a sheared contour lines, as in 5b is shown, the recalculated in metric heights offset values 5a indicating that in 5a and 5b along the x-axis the pixel column number is plotted while in 5a along the y-axis the offset values are plotted in arbitrary units and in 5b the metric heights are also plotted in arbitrary units.

The compensation of this perturbation as well as the translation of the sensor offset values into metric heights is therefore accomplished in accordance with an embodiment of the present invention by a calibration method which provides the following figure:

vervollständigt wird. Die Abbildung kann dabei für alle Spalten gleich sein oder für jede Spalte einzeln bestimmt werden, um weitere Lagefehler von beispielsweise den Spiegeln 62a-62c auszugleichen.In this calibration method is in the measuring device of 1 . 2 B a coplanar calibration target, such as a box, within the inspection volume 20 with a suitable resolution, such as a resolution of 0.5 mm, moved up and down. The laser line projected onto the target 54 is as described by the extraction device 52 converted to sensor offset values associated with the height at which the target is positioned. By repeating this process at different positions of the target in the inspection volume 20 is that as a detection volume 20 scanned and thus obtained a calibration grid, that for all columns together or for each column individually assigns the particular pixel positions the set calibration levels. Based on this support grid, missing values for pixel positions to which no calibration height has been assigned can be assigned interpolated values obtained, for example, by linear interpolation from the sample points, thereby reducing the image

is completed. The image can be the same for all columns or be determined individually for each column to further misalignment of example, the mirrors 62a - 62c compensate.

For example, a height-adjustable table with micrometer screw is used for the calibration process. The target in the above example, for example, serves a glass plate similar material as the bottles to be tested 24 , ie with a similar refractive index, and possibly with the same thickness as the test objects, ie with a similar thickness as the bottle bottoms 28 , Then, for example, M measurements can be performed at different heights H with, for example, a stepwise of dH = 1 mm. N × B height profiles are obtained for each measurement, where N is the number of channels, in the present exemplary case 4, and B is the block size for determining the profile of the profiles, ie the number of shots taken by the trigger signal on the management 42 be triggered corresponds. At the same time, the mm value of a micrometer screw is determined with maximum accuracy. By averaging, this yields M value pairs, namely a value pair m with m = 1... M from the respective acquired mm value ρ _{G, m} and the associated position value from the N channels or offset value ρ ~ _{G, m} . For the offset values p ~ between these interpolation point offset values, the mapping into the metric heights ρ can then be carried out piecewise in a linear manner, as shown in FIG 4 and is illustrated by the following forms:

The functioning of the device is now described below 1 in the variation 2 B during the measurement of the bottles 24 or the bottle bottoms 28 described. Before reference is made in detail to the individual steps, the procedure is explained in advance roughly. The process starts at the output of the extraction device 52 , From the calibrated profiles, the height information from opposite views of the bottle 24 have a height image is generated. The two views are aligned with each other to recover data that was partially lost during profile recording went. Then, the bottom contact surface of the bottle is extracted, whereupon an alignment correction is performed, the positional tolerances in the positioning of the bottles 24 compensated. Finally, a histogram is calculated on the basis of which an identification of the bottle surfaces is carried out, after which the parameters are calculated which make it possible to decide whether to accept or reject the test object.

wandelt ein Fließkommabild in eine (vorzeichenlose) Ganzzahldarstellung um.Before starting with the detailed presentation of the procedure, the notation used should also be explained. Namely, a height image with m rows, n columns and p levels or channels as A ε ^| F ^{(m, n, p)} , where ^| F ^{| | |} R has floating point height values. The i-th plane or i-th channel is denoted by A ⁽ⁱ⁾ , where the value of the pixel at (x, y) is denoted by A ⁽ⁱ⁾ (x, y). The illustration

converts a floating-point image into an (unsigned) integer representation.

The measurement of a test object 24 begins with data collection 102 that is, the shots of the transparent object 24 , As already mentioned above, at step 102 the laser 10 through the encoder 40 switched on immediately before the test object 24 the light plane 14 reached. While the bottle 24 the plane crosses, controls the encoder 40 the pixel array 16 such that the same takes pictures at times when the bottle 24 an integer multiple of an incremental distance in the direction of movement 26 has covered. This results in r recordings of four channels each, for which the extraction device 52 provides four offset vectors s ⁽⁰⁾ (t), ..., s ⁽³⁾ (t) per shot or image at the respective time t. These vectors can be considered as the lines of a new generated image, with s ⁽ⁱ⁾ (t _p ) associated with p = 0 to r of the i-th plane. By juxtaposing these successively obtained vectors, one thus obtains offset value images S ⁽⁰⁾ ,..., S ⁽³⁾ , namely a pro-channel or plane. This will be in the step 104 in each case one height image R is obtained by rescaling the offset values of the profile lines contained in these offset values into metric heights, such as, for example, using the illustration already mentioned above

The i-th plane R ⁽ⁱ⁾ contains the profiles r ⁽ⁱ⁾ (t _j ) obtained from the r acquisitions at the times t _j . With a suitable choice of trigger triggering by the rotary encoder 40 or with a suitable choice of Weglängeninkrements the encoder 40 For example, R consists of r = 50 lines containing the undistorted metric height values.

auf die Höhenbilder R angewendet wird. Dabei wird ein apriori-Wissen über das Inspektionsvolumen 20 ausgenutzt, nach welchem die Höhenwerte nur in einem bestimmten Höhenbereich liegen können. Insbesondere wird in dem Schritt 106a jeder Höhenwert aus R gemäß der folgenden Gleichung modifiziert, woraus sich das verbesserte Bild R erhalten wird.

wobei r_lo den niedrigsten und r_hi den höchsten Höhenwert bezeichnen soll, der aus den Abmessungen des Inspektionsvolumens 20 bekannt bzw. zu erwarten ist.In a subsequent step 106 Then, a noise removal is performed on each height image R ⁽ⁱ⁾ . The noise removal is in 7 shown in more detail. To 7 will be in a first step 106a performed an efficient removal of invalid height data or restricted all height data to an allowable height range by a simple clipping transformation C

is applied to the height images R. This is an apriori knowledge about the inspection volume 20 exploited, according to which the altitude values can only be in a certain altitude range. In particular, in the step 106a each height value of R is modified according to the following equation, from which the improved image R is obtained.

where r _lo is to denote the lowest and r _hi the highest altitude value, from the dimensions of the inspection volume 20 known or expected.

Unfortunately, the irregular surfaces of the specimens produce 24 unwanted total reflections that the interfere with collected altitude data. This effect is almost always observable when the bottle 24 straight into the light plane 14 invades or leaves, and sometimes in between. Fortunately, this type of disturbance is characterized by only occasional outliers, whereas the elevation values actually coming from the areas to be examined generally have similar levels. This justifies the subsequent sub-step 106b in which the improved images R - are again subjected to a histogram-dependent clipping or cutting.

berechnet, für das ein exemplarisches Beispiel in 8a gezeigt ist. Insbesondere gibt das Histogramm

in willkürlichen Einheiten an, in welcher Häufigkeit oder Frequenz ein vorbestimmter Höhenwert r in den Höhenbildern R -(0) ...., R -(3) vorkommt. In 8a ist das Histogramm

in einer Darstellung gezeigt, bei der die x-Achse den Höhenwerten r in mm und die y-Achse der Frequenz bzw. Häufigkeit des jeweiligen Höhenwerts r in willkürlichen Einheiten entspricht.In the sub-step 106b becomes an intermediate histogram

for which an exemplary example in 8a is shown. In particular, the histogram gives

in arbitrary units, in which frequency or frequency a predetermined height value r in the altitude images R - (0) ...., R - (3) occurs. In 8a is the histogram

in a representation in which the x-axis corresponds to the height values r in mm and the y-axis corresponds to the frequency of the respective height value r in arbitrary units.

merkliche Peaks für die interessierenden Oberflächen 28a-28c erkennen, wohingegen die durch Störungen verursachten Ausreißer lediglich mit niedrigerer Frequenz bzw. Häufigkeit vorhanden sind. Dementsprechend werden in dem Teilschritt 106b alle Höhenwerte R -(i)(x,y), die einen vorbestimmten Höhenschwellwert r_th überschreiten, unterdrückt bzw. auf Null gesetzt, was zu dem verbesserten Bild R ^ bzw. den verbesserten Höhenbildern R ^(0), ..., R ^(3) führt.

Like it out 8a can be seen leaves the intermediate histogram

significant peaks for the surfaces of interest 28a - 28c whereas the outliers caused by disturbances are only present at a lower frequency. Accordingly, in the substep 106b all altitude values R - (I) (X, y) which exceed a predetermined altitude threshold r _{th are} suppressed, resulting in the improved image R 1 and the enhanced altitude images, respectively R ^ (0) , ..., R ^ (3) leads.

überschreitet. Dabei kann beispielsweise der Schwellwert h_th derart gewählt werden, dass der Parameter r_th der Bodenberührungsfläche 28a entspricht. Bei geeigneter Wahl von h_th kann somit der höchste gültige Höhenwert automatisch aus dem Histogramm

berechnet werden, was eine effektive und insbesondere auch adaptive Ausreißerentfernung 106b liefert.The parameter r _th is related to the parameter h _th in the equation 4 such that it corresponds to the in 8a denotes the extreme right or the highest height value r, which has a predetermined frequency or frequency threshold h _th

exceeds. In this case, for example, the threshold value h _{th can be selected} such that the parameter r _{th of} the ground contact area 28a equivalent. With a suitable choice of h _th , the highest valid altitude value can thus be automatically obtained from the histogram

which is an effective and especially adaptive outlier distance 106b supplies.

For example, the choice of _hth may be automated by adjusting this threshold to a predetermined percentage of the absolute maximum of the intermediate histogram.

Returning to the 6 Be the ones by the noise removal 106 improved height images R ^ in a subsequent step 108 to superimpose the views juxtaposed in these images, thereby allowing recovery of lost data in one of the two views by the other view. This will be briefly in advance again 2 B With reference to which it has been explained that the pixel array 16 to half is used to the inspection volume 20 from a line of sight 24a to capture, and to another half from a different perspective 34b , The resulting brightness or gray level images taken from the pixel array 16 are included in the examples 3a shown. As can be seen, in the left half of the pixel array, the bottom of the bottle is picked up from a different view than the bottom of the pixel array right half. As can also be seen, the contour line of the left view is disturbed due to unwanted reflections. Of course, the disturbance is also reflected in the output signal of the extraction device 52 again or in the height image R. 9a shows, for example, the gray scale representation of the height image R, as in the construction of 2 B is obtained while 9b the corresponding noise-filtered height information R ^ represents. As it is by comparing the 9a and 9b can be seen leads the noise removal in step 106 Although to eliminate many disturbances in the height images R, but there are still areas where data has been lost, resulting in 9b on the black areas within the ellipse, which is the area 28a and 28c with gray scaled altitude information, can be seen. The step 108 now has the goal, from the two views left and right in 3a obtained height information of the 9b in the middle, and to manipulate the resulting height picture halves so that within them the positions (x, y) are aligned with each other, the x-axis of the pixel column number and thus the dimension transverse to the direction of movement 26 corresponds, and the y-axis of the dimension along the direction of movement 26 equivalent.

abgebildet, wobei M ein Medianfilter darstellt. Aus dem sich ergebenden geglätteten Bild B werden die Rand- bzw. Grenzpunkte beider Ansichten, d.h. in der linken und rechten Hälfte, in einem darauffolgenden Schritt 108d extrahiert. Dies wird durchgeführt, indem die Scanlinien aus B⁽⁰⁾, d.h. die Zeilen des für die 0-te Ebene erhaltenen grauskalierten Bildes B, in der linken und rechten Hälfte untersucht werden, um diejenigen äußersten linken und äu ßersten rechten Pixelpositionen B⁽⁰⁾(x,y) zu identifizieren, die eine Intensität gleich oder größer einer Benutzerdefinierbaren Schwelle B_th aufweisen.In order to restore or replenish the holey regions that result from data loss during profile acquisition, in step 108 both views aligned. For this purpose, in a sub-step 108a every noise-filtered height image R ^ ( 9b ) to a gray scaled image B

M is a median filter. The resulting smoothed image B becomes the boundary points of both views, that is, in the left and right halves, in a subsequent step 108d extracted. This is done by examining the scan lines from B ⁽⁰⁾ , that is, the lines of the gray scale image B obtained for the 0th plane, in the left and right halves, to those leftmost and outermost right pixel positions B ^(0). (x, y) having an intensity equal to or greater than a user definable threshold B _th .

The in step 108b detected or extracted edge pixels correspond to the boundary or the edge of the bottles 24 in the two views. Thus arise from the step 108d two sets of edge pixels, one for each view. The elements of a sentence function in a subsequent sub-step 108c as control points for, for example, a non-iterative elliptical fit process, as described, for example, in Fitzgibbon, A., Pilu, M., Fisher, R .: Direct Least Fitting of Ellipses, Intern. Conference on Pattern Recognition (1996) pp. 253-257. An implementation as described in Flusser, J. Halir, R .: Numerically stable direct least square fitting of ellipses. International Conference on Computer Graphics, Visulization and Interactive Digital Media (WSCG'98) 1 (1998), pp. 125-132 is also possible, the same being especially improved in terms of numerical stability and computational effort. Off step 108c result in two ellipses E _l and E _r , which approximate the edge or the boundary of the bottle in the left and right view, wherein the ellipses E _l and E _r in 9b With 109a and 109b you can see.

aufweist. Die Transformation A bildet den Inhalt der rechten Ellipse 109b auf die linke Ellipse 109a ab.On these ellipses is in a subsequent sub-step 108d an affine transformation to transfer the two ellipses 109a and 109b Into each other determines the shape

having. The transformation A forms the content of the right ellipse 109b on the left ellipse 109a from.

This transformation will now be in step 108e applied to the height images R ^ of the various channels, thereby obtaining aligned images R ~. This can be expressed as follows:

The application of the transformation A to the height images R ^ is in the sub-step 108e Performs and results in position-corrected height images R ~, which have only half the number of columns but are twice as many in number, since now under these location-corrected height images R ~ each two for a channel or a level are available.

Returning to 6 will after superimposing 108 the two views, a manipulation of the obtained position-corrected height images R ~ in step 110 performed such that the areas of interest 28a . 28c on the one hand and 28b on the other hand, the transparent object 24 extend substantially perpendicular to the height axis or height direction within these images. In other words, the step tries 110 compensate that due to positioning tolerances of the conveyor 22 in most cases the bottle bottoms 28 not exactly parallel to the xy plane or to the plane perpendicular to the illumination direction 32 which will prevent the presence of clear and clearly separated histogram peaks in a histogram of height images R ~.

For this reason, in step 110 the location of the surface of interest, and in particular the signal technically favorable outer surface 28a . 28c in particular, by fitting a plane into the elevation values representing the ground-contacting area 28a form. The necessary control points or interpolation points are obtained from the position corrected height images R ~ by applying the steps as shown in FIG 11 are extracted on each of the r rows (previously exemplified 50).

In a partial step 110a First, the information of the altitude images R ~ (0) and R ~ (4) merged. These elevation images, except for outliers, include, as in 9d are still recognizable, a height representation of the outer surface 28a . 28c Bottom of the bottle, with the level 0 of the left view and the level 4 of the right view in 9b corresponds, ie have been obtained from different perspectives.

wobei mit v_r = (v₀, ..., v_n) die rekonstruierten bzw. kombinierten Profile bezeichnet werden.The merge in the sub-step 110a serves to merge the information of both views. Since it is likely that data is missing in either of the two levels 0 and 4, in the substep 110a extract appropriate altitude values from the altitude images by selecting, for a given position (x, y), the valid, ie non-zero, altitude value from the one altitude image if the corresponding altitude value in the other plane is invalidated is, and in the case that for a given position (x, y) the height images of both planes have a valid height value, the corresponding altitude value is calculated by taking the average of these two altitude values. If there is no height information for a specific position x, y in both the one view and the other view, no reconstruction can be performed either. This is expressed by the following rule:
v _r = (v ₀ , ..., v _n ) with

where v _r = (v ₀ , ..., v _n ) the reconstructed or combined profiles are called.

wobei arg max V_l beispielsweise die Spaltennummer bzw. diejenige Komponente des Vektors V_l angibt, an der das Maximum von V_l liegt, und arg max V_r diejenige Komponente des Vektors V_r angibt, an der dort das Maximum vorliegt. In dem Satz C ~ sammeln sich somit Punkte im dreidimensionalen Raum, die in Koordinaten angegeben werden, von denen sich die erste auf die Richtung senkrecht zur Belichtungsrichtung 32 und senkrecht zur Bewegungsrichtung 26 in Einheiten von Spaltenummern bezieht, die zweite Komponente auf die Achse entlang der Bewegungsrichtung 26 in Einheiten der Aufnahmen bzw. der vorerwähnten Weginkremente und die dritte Komponente auf die Höhenrichtung r, wie beispielsweise in 2a angedeutet, in den vorerwähnten metrischen Einheiten.In a partial step 110b For each removed line v _r, a one-dimensional median filter is applied for the removal of isolated outliers. Thereupon will be in a sub-step 110c each Filtered line v _r subjected to a maximum search to find the maximum in the left half and the maximum in the right half of this line. Each maximum located on the left and arranged on the right is added as a three-dimensional vector to an initially empty set C ~, as it can be expressed as follows:

where arg max _l V indicates, for example, the column number or the component of the vector V _l, is at the maximum of V _l and V _r arg max that component of the vector V _r indicates there is present at the maximum. In the set C ~ thus accumulate points in three-dimensional space, which are given in coordinates, of which the first to the direction perpendicular to the exposure direction 32 and perpendicular to the direction of movement 26 in units of column numbers, the second component refers to the axis along the direction of movement 26 in units of the recordings or the aforementioned path increments and the third component on the height direction r, such as in 2a indicated in the aforementioned metric units.

To achieve a more robust levels of fit, is first in a sub-step 110d each element removed from the set C ~ that could potentially constitute an artifact, namely, each element p = (p _x, p _y, p _z) ^T ε C ~, whose height value p _z more than the standard deviation σ from the mean value μ over all height values p _{z of} the quantity C ~ deviates, which does not fulfill the following condition:
| p _z - μ | <σ,

Based on the remaining set of three-dimensional points in the set C ~ is in a sub-step 110e then the plane F: n x -d = 0 is computed, such as by a least squares fit or least squares method such as in Feddema, JT, Little, CQ: Rapid range data to geometric primitives. Intern. Conference on Robotics And Automation (1997), pp. 2807-2812, wherein in the Hessian normal form n shown, the normal vector is represented by the fit in the ground contact surface 28a received plane, d a distance of this plane from the origin, such as the point 50 in 2a , in direction n and x corresponds to a vector on a point of the plane.

wobei die dist(F_ref, F) den Normalenabstand zwischen F_ref und F an der Stelle (x,y) beschreiben soll. Nach der Ausrichtekorrektur in dem Teilschrift 110f sind die Höhenwerte ausgerichtet zu der xy-Ebene, was deutlich voneinander unterscheidbare Peaks für die Oberfläche in dem Histogramm der Höhenwerte über die so erhaltenen gekippten Höhenbilder R ‿(0), .., R ‿(7) ergibt.The roll and pitch angles of the ground touch surface plane thus determined encode the orientation of the bottle 24 in the room and are in a partial step 110f used to compute an image that aligns with the xy plane perpendicular to the height direction. For this reason, a new plane, namely the reference plane F _ref : n _ref x -d = 0, n _ref = (0,1,0) ^{T is} defined. Looking at the height picture R, one is only at its values R ~ (I) (X, y) interested and in the relationship between them. Since their positions (x, y) are not significant, it is in the section 110f sufficient to perform a shear transformation and not a more elaborate rotation transformation, wherein the shear transformation described in the part 110f can be used with

where the dist (F _ref , F) is to describe the normal distance between F _ref and F at the location (x, y). After the alignment correction in the part 110f are the elevation values aligned with the xy plane, which are clearly distinguishable peaks for the surface in the histogram of elevation values over the resulting tilted elevation images R ‿ (0) , .., R ‿ (7) results.

Returning to the 6 , will, after the height pictures in step 110 tilted or leveled, in one step 112 an assignment of the height values of the tilted height images to the areas of interest, namely the puncture area 28c on the one hand and the inner surface 28b the bottom of the bottle 28 on the other hand, based on the histogram over all images R ‿ performed.

erzeugt und dazu verwendet, um die Höhenwerte aus den Bildern R ‿ jeden für sich, d.h. individuell, der jeweiligen Fläche 28c oder 28b zuzuordnen, dem derselbe angehört bzw. entspricht. Ein Beispiel für das sich ergebene Histogramm ist in 8b gezeigt. Ein Vergleich mit dem Histogramm aus 8a ergibt ohne weiteres, dass vor der Orientierungskorrektur nach Schritt 110 in dem Histogramm der Peak für die Bodenberührungsfläche 28a einer geneigten Flasche nicht klar identifiziert werden kann bzw. mit dem Peak für den Einschnitt 28c verschmolzen ist, wohingegen dies in dem endgültigen Histogramm

nach der Ausrichtungskorrektur 110 sehr wohl möglich ist, da der Peak 113a, der an der Höhe der Bodenberührungsfläche 28a angeordnet ist, deutlich von dem Peak 113c getrennt ist, der der Einschnittfläche 28c zugeordnet ist.More specifically, based on the corrected height images R ‿ (0) , ..., R ‿ (7) a final histogram

generated and used to the height values from the images R ‿ each for themselves, ie individually, the respective area 28c or 28b to associate, to which it belongs or corresponds. An example of the resulting histogram is in 8b shown. A comparison with the histogram off 8a readily yields that before the orientation correction after step 110 in the histogram the peak for the ground contact area 28a a tilted bottle can not be clearly identified or with the peak for the incision 28c whereas this is in the final histogram

after the alignment correction 110 is very possible because the peak 113a at the height of the ground contact surface 28a is arranged, clearly from the peak 113c is separated, that of the incision surface 28c assigned.

extrahiert. Von diesem Punkt aus wird das Histogramm

nach links, d.h. zu kleineren r hin bzw. in eine Richtung der Höhenwerte r, die zu der Innenfläche 28b hin gerichtet ist, nach der Position r_ζ1 durchsucht, wo die Histogrammfrequenz bzw. -häufigkeit einen vorbestimmten Schwellwert

unterschreitet. Daraufhin wird das Gleiche in der anderen Suchrichtung, nämlich die Suchrichtung nach rechts in In the step 112 is therefore regarding the incision area 28c the position of the peak 113c , namely the position r _ζ from the final histogram

extracted. From this point, the histogram becomes

to the left, ie towards smaller r or in a direction of the height r, leading to the inner surface 28b is searched for the position r _ζ1 where the histogram frequency is a predetermined threshold

below. Thereupon, the same in the other search direction, namely the search direction to the right in

8b , from the peak position r _{ζ of} the peak 113c from which the position r _{ζr is} obtained, at which the frequency or the frequency in this search direction again differentiates the threshold h _ζ .

Based on the height _values r _ζl and r _ζr thus obtained, those height values from the height images R ‿ of the _puncturing surface then become 28c which are within the range between these values, whereby, taking into account those xy coordinates, a set of 3D points or a point cloud P _ζ for representing the puncture area 28c P _ζ = {(x, y, R ‿ ⁽ⁱ⁾ (x, y)) ^T : r _ζ1 <R ‿ ⁽ⁱ⁾ (x, y) _≦ r _ζr }

In the same way will step in 112 in terms of the peak 113b proceeded to the inside 28b concerns. Again, first the height position r _{ι of} the peak 113b determined, whereupon in the two search directions left and right thereof, the positions r _ιl and r _{ιl be} determined at which the histogram of r _ι from a certain threshold h _ι below, whereupon again those 3D points for the inside 28b obtained as a point cloud P _ι , by P ι = {(x, y, R ‿ (I) (x, y)) T : r ιl ≤ R ‿ (I) (x, y) ≤r ιr } can be represented.

The result of the step 112 is for example in 12 shown, but there in addition to the point clouds P _ζ and P _{ι have} also been applied to the points that the outer surface 28a or the ground contact surface 28a outside the puncture area 28c affect.

Returning to 6 gets in one step 114 an evaluation of the obtained point clouds is carried out to obtain the desired and necessary parameters for determining whether or not the measured bottle meets predetermined specifications. According to the present embodiment, the distances between the reference plane F _ref and the elements p from the point clouds P _ζ and P _ι calculated, which provides the basis for Paralallität- and Planaritätsmessparameter.

wobei die Argumente des Minimum und des Maximum-Operators in den soeben genannten Gleichungen den Normalenabstand des Punktes p zur Referenzebene F_ref : n_ref·p – d = 0 darstellen. Anders ausgedrückt, stellt ζ_min die minimale Höhe des Einstichs, bezogen auf die Referenzebene F_ref dar, während ζ_max die maximale Höhe des Einstichs bezogen auf die Referenzebene F_ref darstellt. ζ_obliquity ist definiert als die Differenz von ζ_max und ζ_min und entspricht somit der Schiefe des Einstichs 28c.It is important to distinguish between internal and external measurement parameters. The latter are not subject to refraction effects and are defined as follows:

where the arguments of the minimum and the maximum operator in the just mentioned equations represent the normal distance of the point p to the reference plane F _ref : n _ref * p -d = 0. In other words, ζ _{min represents} the minimum height of the groove relative to the reference plane F _ref , while ζ _{max represents} the maximum height of the groove relative to the reference plane F _ref . ζ _obliquity is defined as the difference between ζ _max and ζ _min and thus corresponds to the obliquity of the puncture 28c ,

wobei δ(p) = d_g(p)·(1 – ρ) gilt und ρ hier den Brechungseffekt modellieren soll. d_g(p) wird bestimmt bezogen auf eine Ebene, die in den Satz von Einschnittpunkten P_ζ gefittet worden ist. Tatsächlich aber hängt der Einfluss des Brechungseffekts nicht nur von dem Brechungsindex des Materials, sondern auch von der Flächenorientierung ab, so dass ein orientierungsabhängiger Korrekturfaktor ρ geeigneter wäre als ein lediglich konstanter Faktor.By contrast, for those on the inner surface 28d related measurement parameters, the refraction effects are taken into account. Since the refraction affects only that portion of the light passing through the glass, only the distance d _g (p), which is the perpendicular distance from p to the incision surface, has to be corrected accordingly 28c whereas the portion of the beam path through which the air passes remains unaffected. According to an exemplary embodiment of the present invention, the measurement parameters relating to the inner side thus result in:

where δ (p) = d _g (p) · (1 - ρ) and ρ here is to model the refraction effect. d _g (p) is determined relative to a plane which has been fitted into the set of entry points P _ζ . In fact, the influence of the refraction effect depends not only on the refractive index of the material, but also on the surface orientation, so that an orientation-dependent correction factor ρ would be more suitable than a merely constant factor.

The system described above 1 according to the variation 2 B was tested both under laboratory conditions and in a production line with a reference random sample of 66 bottles. Essentially, three types of tests were performed, which are briefly described below, before the results are presented.

The first experiment was designed to prove that the system provides nearly identical readings for identical bottles. For this reason, one and the same bottle was repeatedly supplied to the inspection system or the inspection device. The results are in 13a and show the performance of the delivered inspections. The repeated measurements for the minimum of the incision area averaged 1.25 mm with a standard deviation of 0.004 mm. The difference between the minimum and maximum measured value is 1.26 mm - 1.25 mm = 0.01 mm and represents the bandwidth of the measured values. If one compares the result with the theoretically achievable system resolution of 0.08 mm, this is correct The results for the maximum incision or the maximum incision depth show a comparable result with an average of 1.48 mm, a standard deviation of 0.01 mm and a width of 1.51 mm - 1.47 mm = 0, 04 mm.

Sensitivity as a function of the different heights of the bottles within the inspection volume has been the subject of the second test. Again, a bottle was repeatedly inspected, this time in a 6mm raised position, simulating the different heights of the bottles in which they can be presented to the inspection system within a production environment. The results are in 13b and slightly different from those from the first test or 13a , Here, the measurement for the minimum of the incision area averages 1.22 mm with a standard deviation of 0.004 mm and a bandwidth of 1.23 mm - 1.21 mm = 0.02 mm. Again, maximum cut depth readings show comparable results. The mean value is 1.43 mm, the standard deviation 0.011 mm and the bandwidth 1.47 mm - 1.44 mm = 0.03 mm.

In the third experiment, a sequence of known bottles was checked by the system and the errors in the measurements concerning the outer and inner surfaces are in 13c and 13d shown. The result shows that the outside measurement errors behave better than those related to the inside bottom surface; on average, the error of the incision area measurements is 0.03 mm for the minimum reading and 0.07 mm for the maximum reading, whereas the error for the inner tray readings is 0.16 mm for the minimum reading and 0.06 mm for the maximum reading. The variance represented by the standard deviation is 0.05 mm for the minimum reading and 0.04 mm for the maximum reading in the case for the external readings and 0.16 mm for the minimum readings and 0.15 mm for the maximum readings in the Trap of interior floor surface measurements. Due to the simplification of the refraction, the inner measurements generally behave slightly worse. As a lower limit for the accuracy of the inner measurement values, 0.2 mm could be extracted from the experiments, whereas the accuracy for the outer measurement values is 0.1 mm. Tests carried out on a conveyor belt in a production line give results comparable to those achieved by laboratory experiments.

The above embodiments present hence a technique that is capable of being made of glass Inspect bottles in a quick and contactless manner. In particular, a histogram-based algorithm has been described, the removal of unwanted Allows reflections, as well as an assignment of the height data to the inner and outer surface. Out the point clouds covering the surfaces metric measurements of planarity and parallelism were made to one Get fashion that is reproducible. The actually scored Accuracy was 0.20 mm for Outside measurements and 0.30 mm for those measurements that were related to the inner surface.

The Method is particularly suitable where the surfaces of the object are almost parallel aligned to the histogram-based approach works. Indeed be bottles, the bottoms show a strong skewness, correctly classified, but become for them no correct measured values were obtained. This is due to the design of the Algorithm based on the identification of concentrations of same or similar height values which give distinguishable peaks in the histogram. For bottles, the floors with strong skewness, these peaks disappear, causing an extraction of the surfaces becomes more difficult.

By the preliminary knowledge about the inspection volume obtained from the system design, The algorithm described is insensitive to noise caused by unwanted reflection on the object under inspection or on the components of the production line, such as. the conveyor belt, is effected.

The space available for integrating the system into an existing production environment affects the triangulation geometry, and in particular its baseline B (see 2a ). This in turn forms a major influence on the theoretically achievable resolution and thus also on the accuracy of the inspection system. A second limiting factor is provided by the optics relating to the focus area of the camera and the refractive projection lens which spans the light layer, since the physical thickness of the laser line and the sharpness of the image of the camera are an essential basis for the accuracy and resolution of the system.

Promising Improvements in accuracy are likely due to a more complicated modulation of the refraction effects and a pre-alignment the left and right view achievable, the latter the inaccuracies when extracting the control points for the elliptical fit would.

Once again, in other words, the above embodiments provide a way of quality checking glass bottles that allows to automatically evaluate 3D height data acquired by a light section sensor for the purpose of detecting discrepancies in the ground measurements of the transparent objects to set points , In particular, the embodiments provide a quick histogram-based approach to determining the point clouds that represent the areas of the ground that are the subject of the review. From these point clouds, parameters are derived automatically, which in turn are used for the classification of the inspected object. A major advantage of this approach is the high inspection speed, which meets the requirements of a 100% inspection in production lines running at a speed of 3-10 bottles per second. The above embodiments thus meet both the demand for a quick inspection, but also the demand for a high inspection accuracy, since for example the production of chemical-pharmaceutical glass containers require a high quality of production and thus a high percent reliability, or there a dimensionally accurate production according to customer-specific requirements is of central importance to the manufacturer. In other words, the Flascheninspizierer allows after 1 respectively. 1 . 2 B integrates 100% inspection into the production line with the aid of a 3D scanner designed to identify and document deviations from specified floor thickness or floor planarity setpoint values. By means of the 3D scanner, a complete scan of the surface to be tested is made possible. Digital measured values can be processed in real time and logged for quality management and provided in the form of digital output signals for connection to the production control. An integrated special camera realizes the fast data acquisition by means of light-slit methods. The procedure can be applied to both special glass and plastic bottle production.

• 1) Die Bodendicke darf nicht eine minimale Boden-Dicke unterschreiten, d. h. es soll gelten ζ_min – ι_max > d_min;
• 2) Die Bodendicke darf eine maximale Bodendicke nicht überschreiten, d.h. es soll gelten ζ_max – ι_min < d_max.
• 3) Die Schiefe des Innenbodens 28b darf eine maximale Schiefe nicht überschreiten, d. h. es soll gelten ι_obliquity < S_{max,Innenboden}.
• 4) Die Einstich-Tiefe darf eine minimale Einstich-Tiefe nicht unterschreiten, d. h. es soll gelten ζ_min > π_min;
• 5) Die Einstich-Tiefe darf eine maximale Einstich-Tiefe nicht überschreiten, d. h. es soll gelten ζ_max < π_max;
• 6) Der Einstich darf eine maximale Schiefe nicht überschreiten, d.h. es soll gelten ζ_obliquity < S_max,Einstich.

At this point should briefly again 6 To be received. The description of 6 essentially ends with the acquisition of the measurement parameters necessary for the classification of the bottles. These measured value parameters can, for example, be compared by a program with nominal values or with a nominal specification, which is also available on the computer 44 expires and, for example, is separate from the one that is also on the computer 44 runs and the surveying task after 6 takes over. For example, in the case of serum bottles, the specifications could provide that the measurements obtained by optical measurement of the bottoms of the bottles must meet the following criteria in order for the bottle to be considered good and not classified as waste:

• 1) The floor thickness must not fall below a minimum floor thickness, ie it should apply ζ _min - ι _max > d _min ;
• 2) The floor thickness must not exceed a maximum floor thickness, ie it should apply ζ _max - _min <d _max .
• 3) The obliquity of the inner bottom 28b may not exceed a maximum skewness, ie it should apply ι _obliquity <S _{max, inner bottom} .
• 4) The puncture depth must not be less than a minimum puncture depth, ie it should apply ζ _min > π _min ;
• 5) The puncture depth must not exceed a maximum puncture depth, ie it should apply ζ _max <π _max ;
• 6) The puncture must not exceed a maximum skewness, ie it should apply ζ _obliquity <S _{max, puncture} .

If one of these criteria is not met, the controller classifies 44 the corresponding bottle 24 As a committee, for example, after the surveying and classification, it discards it by disposing of it at a corresponding reject location following surveying and classification, whereas bottles that have been positively classified are passed on to another level of protection, such as through the sponsor 22 to a production stage for filling the bottles or the like.

In the following, let's look briefly at a possible realization of the system 1 respectively. 2 B or possible implementations of the individual components. With regard to the image recording components, for example, the image recording of the surface can be carried out with a high-resolution special camera for laser light-section method. As described above, the test object is conveyed by a conveyor 22 straight to a camera 46 passed, wherein the test object is illuminated by means of, for example, a high-power laser with line optics, and wherein for synchronizing the image recording with the object movement, a distance measuring system, such as a rotary encoder 40 , is used, which provides pulses at constant intervals, with each of which the recording of a line is triggered. In particular, the IVP Ranger M50 with 512 × 1.536 pixels, an HSSI high-speed camera connection and a lens mount CMount can now be used for the camera. The following types of light source are considered: StockerYale Magnum 2W 10 ° line optics.

The image evaluation electronics 44 consists, for example, of a computer on a PC platform. It handles, for example, the tasks of digital recording and evaluation of the camera images 6 , For example, an embodiment in the 19 "housing may be used. The computer can be equipped, for example, with plug-in cards for connecting the special camera mentioned above, such as the plug-in card HSSI adapter board for PCI connection. Furthermore, the computer can be equipped with a plug-in card for the connection of digital I / O signals, such as the following types: AddiData APCI-1500 (digital i / o) and APCI-1710 (function-programmable counter card). Evaluation electronics, operating elements and power supply for the camera, lighting can be accommodated in a control cabinet.

The software on the evaluation calculator 44 running, may include the following application-specific functions, namely a user interface for safe operation of the test station, program modules for the synchronization of the measurement process with the parts feeder, a database for storing the specific parameter sets, a test program synchronized in automatic measuring mode with the signals of the automatic parts feeder Checking all relevant surfaces of the test parts, generating the test result, logging the test result in a file, and outputting a corresponding signal to the digital I / O interface, a set-up mode in which alternatively the functions of a parameter ver management, image display and individual test, etc. The image representation can support the setting of the image acquisition unit. With the individual test, the station can be checked for its test performance. For example, the software could run on the Windows 2000 operating system.

The measurement of the surfaces described above is made possible by the projection of a laser line and detection at a triangulation angle, the detection in particular relates to the diffuse reflection on the outer and inner bottle bottom and not the total or specular reflection. One way to make this possible, for example, is to combine the laser beam source with egg ner expanding 10 ° degree optics, which then projects at a distance of 420 mm, an approximately 50 mm long laser line perpendicular to the glass bottom. In the 2d shown front surface mirrors, of which the mirrors 62a and 62c responsible for the line of sight, and the mirror 62b for splitting into two perspectives, can be of any kind. The camera, which views the laser line on the bottom of the bottle through these mirrors from two sides, can be arranged at an angle of approximately 55 ° C., as mentioned above. As mentioned above, the camera can be designed such that in particular it also performs the tasks of the extraction device 52 as is the case with the above-mentioned IVP Ranger M50. However, four lines or channels do not have to exist. It could also be only three lines, but at least two, which are extracted, for example, with an individual threshold from the gray value image of the camera.

The Calibrate the system from pixel coordinates to symmetric values can, as mentioned above, means Mechanical translation table with micrometer scale and a look-up or lookup table happen. In this case, for example, the output of the camera 8 bits or 16-bit pixel position information include, i. an 8/16 bit integer profile picture with height values on N channels of the Output image object. The output of the extraction device could be a 64-bit floating-profile image display with calibrated heights in, for example, mm.

Finally, reference is made to the following variation possibilities, which result in the above-described synchronous evaluation of the 2 or more laser lines, outside and inside, and the thus made possible measurement of the outer and inner ground with respect to flatness or thickness determination. With regard to image acquisition, the present invention is not limited to the configurations of 1 or 2 B limited. For example, it would also be possible to use a parallel recording from two viewing directions, in which the viewing directions are not mirror-symmetrical to the light plane. Furthermore, the present invention is not based on the illustrated evaluation algorithm 6 limited. In particular, the algorithm could be designed quite differently in that it is not first necessary to determine candidate pixel positions within each column as the possible contour line points. Rather, an algorithm could be used which determines the contour lines directly from the image or the two images of the two views or the like from the brightness image by means of another pattern recognition algorithm. The chosen algorithm, however, is advantageous in that it is simple and thus feasible online.

Furthermore, it could be provided that the steps in 6 . 7 . 10 and 11 partially omitted or varied in order. For example, the rescaling in step 104 also be done later. Furthermore, the illumination could also define a plane of light that is not necessarily perpendicular to the direction of movement.

In addition, the present invention is not limited to glass or plastic objects, as long as they are sufficiently transparent to the radiation used, the latter also does not have to be light, but could also be electromagnetic radiation of different wavelengths. In terms of the mirror 62a - 66c attention is also drawn to the possibility of alternative arrangements and orientations.

With regard to the implementation possibilities described above, it is also pointed out that the in 6 . 7 . 10 and 11 shown steps in the blocks of the flow charts individually or may be implemented in several subroutine routines. Alternatively, however, an implementation in the form of an integrated circuit is possible, in which these blocks are implemented, for example, as individual circuit parts of an ASIC.

Finally, the possibility of implementing the evaluation algorithm in software is mentioned again. The implementation may be on a digital storage medium, in particular a floppy disk or a CD with electronically readable control signals, which may interact with a programmable computer system such that the corresponding method is executed. In general, the invention thus also consists in a computer program product with a machine Baren carrier stored program code for performing the method according to the invention, when the computer program product runs on a computer. In other words, the invention can be realized as a computer program with a program code for carrying out the method when the computer program runs on a computer.

Claims (17)

Device for measuring a transparent object ( 24 ), with a beam plane generator ( 10 . 12 ) for generating a beam plane ( 14 ) containing the transparent object ( 24 ) cuts to a first profile line ( 54 ) on a front surface ( 28a . 28c ) of the transparent object ( 24 ) and a second profile line on a rear surface ( 28b ) of the transparent object, at the radiation of the beam plane ( 14 ) is scattered; a recording device ( 18 . 16 . 46 ) for picking up the at the first and second profile line ( 54 ) scattered radiation from a predetermined viewing direction ( 34 ; 34a . 34b ) obliquely to the beam plane ( 14 ) to obtain a profile line shot of the first and second profile lines; and an evaluation device ( 44 ) for evaluating the profile line recording, on the basis of which a survey result for the transparent object ( 24 ), which on the front surface ( 28a . 28c ) and the back surface ( 28b ) of the transparent object ( 24 ) is related. Device according to claim 1, further comprising: a movement device ( 22 ) for effecting a relative movement ( 26 ) between the beam plane ( 14 ) and the transparent object ( 24 ), so that the transparent object ( 24 ) the beam plane ( 14 ), the receiving device ( 14 . 16 . 46 ) is formed to produce a sequence of profile line recordings, while the transparent object ( 24 ) the beam plane ( 14 ) crosses. Device according to claim 2, wherein the beam plane generating device ( 10 . 12 ) is formed such that the radiation plane ( 14 ) perpendicular to a direction ( 26 ) of the relative movement between the beam plane ( 14 ) and the transparent object ( 24 ) stands. Device according to Claim 2 or 3, in which the receiving device ( 14 . 16 . 46 ) has the following feature: a pixel array ( 16 ) for generating a brightness image from on the pixel array ( 16 ) falling radiation; and an optic ( 18 ) for imaging the at the first and second profile line ( 54 ) scattered radiation on the pixel array. Device according to claim (4), further comprising: a determination device ( 52 ) for determining first pixel positions of the pixel array as candidates for locations ( 58 ), where the first profile line ( 54 ) on the pixel array ( 16 ) and second pixel positions of the pixel array ( 16 ) as candidates for locations, to the second profile line on the pixel array ( 16 ) is displayed. Apparatus according to claim 4, wherein the pixel array ( 16 ) is arranged such that a line perpendicular to a direction ( 26 ) of the relative movement and an irradiation direction ( 32 ), from the radiation for generating the beam plane ( 14 ) is substantially parallel to a first pixel line of the pixel array ( 16 ), which corresponds to either one pixel row or pixel column of the pixel array. Apparatus according to claim 6, further comprising: extracting means ( 52 ), for each second pixel line of the pixel array perpendicular to the first pixel line, extracting at least a first and a second pixel position within the respective second pixel line which serve as candidates for locations at which the first and second profile lines respectively on the pixel array ( 16 ) is displayed. Device according to one of claims 5 to 7, further comprising: a rescaling device ( 104 ) for rescaling each pixel position to a metric height value to obtain a first predetermined height image from the first pixel positions and a second predetermined height image from the second pixel positions forming a set of associated height images. Apparatus according to claim 8, further comprising: noise removal means (10); 106 ) to remove noise from the altitude images. Device according to claim 9, in which the optics ( 18 ) is formed to the radiation scattered by the first and second profile line next to the predetermined viewing direction ( 34a ) also from an alternative perspective ( 34b ) on the pixel array ( 16 ) or another pixel array such that a brightness image of an alternative view is obtained, wherein the extraction means and the rescaling means are adapted to obtain from the brightness image of the alternative view a first and second height image of the alternative view, the apparatus further comprising the following feature comprising: a superimposition device ( 108 ) for shifting the first and second predetermined height images and first and second height images of the alternative view relative to each other such that within the same the front and back surfaces represented by the first and second predetermined height images and the first and second height images of the alternative view, respectively the first and second predetermined height images and the first and second alternative view height images form a set of associated height images. Device according to one of claims 8 to 10, further comprising: a manipulation device ( 110 ) for manipulating the set of associated elevation images such that the front or rear surface represented by the associated elevation images extends substantially perpendicular to an elevation axis of the associated elevation images to obtain a set of location corrected elevation images. Device according to claim 11, further comprising: an allocation device for determining a height value histogram from the position-corrected height images and, dependent from the height value histogram, Assign altitude values the position corrected height images to the front surface or the back surface, around a first the front surface representing Point cloud and a second representing the back surface point cloud. Device according to claim 12, further comprising: An institution to evaluate the first and second point clouds to one or more the following values as at least part of the improvement result to calculate: a maximum distance of the front surface of a reference plane perpendicular to the elevation axis; a minimal Distance of the front surface from the reference plane; one of a crookedness of the front area dependent on the reference plane Value; a maximum distance of the rear surface of the reference plane; a minimum distance of the rear surface of the reference plane; one of a skewness of the rear surface relative dependent on the reference plane Value. Device according to one of claims 2 to 13, in which the movement device ( 22 ) is adapted to support a bottle of transparent material such that an underside of a bottom of the bottle forms the front surface and the inside of the bottom of the bottle forms the rear surface. Device according to claim 14, further comprising: a sorting facility to sort out the bottle to a reject or forward the bottle to a further processing location, depending on the survey result. Method for measuring a transparent object with the following steps: Generating a beam plane, the the transparent object intersects to form a profile line at one front surface of the transparent object and a second profile line at a rear area of the transparent object, at the radiation of the beam plane is scattered; Picking up the at the first and second profile line scattered radiation from a predetermined viewing direction to the beam plane, by a Profillinieaufnahme the first and the second profile line to obtain; and Evaluating profile taking to be based on the same a survey result for the transparent object get that on the front surface and the back surface the transparent object is related. Computer program with a program code for carrying out the The method of claim 16, wherein the computer program runs on a computer.