EP0453455A1

EP0453455A1 - Process and sewing machine for stitching two pieces of fabric together along the same length

Info

Publication number: EP0453455A1
Application number: EP19900901565
Authority: EP
Inventors: Reinhold Dobner; Robert Massen; Bernhard Mertel; Erich Willenbacher
Original assignee: GM Pfaff AG
Current assignee: GM Pfaff AG
Priority date: 1989-01-12
Filing date: 1990-01-11
Publication date: 1991-10-30
Also published as: JPH04506463A; WO1990008215A1

Abstract

Des images partiellement superposées des caractéristiques typiques de la structure de l'armure (structure de la contexture ou mailles) de chaque couche de tissu (7, 8) sont prises et enregistrées dans une mémoire-tampon sous forme de signaux numérisés d'image en valeurs de gris. On détermine la maximum de corrélation en calculant la fonction de corrélation mutuelle des signaux d'image superposés et on détermine à partir de cette valeur l'extension de l'avancement. Le degré d'allongement de chaque couche de tissu est déterminé en calculant la fonction d'auto-corrélation, l'orientation des fils étant détectée au préalable dans le cas où les fils de trame ou de chaîne d'un tissu ne seraient pas parallèles au sens d'avancement d'un tissu. Un signal de commande d'un agencement règle-point est dérivé de la comparaison entre l'extension d'avancement et le degré d'allongement des deux couches de tissu.Partially superimposed images of the typical characteristics of the structure of the weave (structure of the texture or meshes) of each layer of fabric (7, 8) are taken and recorded in a buffer memory in the form of digitized image signals in gray values. The maximum correlation is determined by calculating the mutual correlation function of the superimposed image signals and the extension of the advance is determined from this value. The degree of elongation of each layer of fabric is determined by calculating the autocorrelation function, the orientation of the threads being detected beforehand in the case where the weft or warp threads of a fabric are not parallel in the direction of advancement of a fabric. A control signal of a ruler-point arrangement is derived from the comparison between the advancement extension and the degree of elongation of the two layers of fabric.

Description

description

Process and sewing machine for sewing two layers of fabric together

The invention relates to a method and a sewing machine according to the preamble of the independent claims 1, 5 and 10 and claims 18 and 21.

DE-OS 35 25028, which is taken into account for the formation of the preamble, discloses a method for determining the feed size of two layers of fabric to be sewn together in the same length, in which, with the aid of two line cameras, which are arranged one behind the other, successively images of typical fabric from each fabric layer

Characteristics are created. By comparing the image data supplied by the first camera with that of the second camera, if the image data coincides, a pulse counting process initiated when the first image is created is terminated and the feed size of the corresponding material position is calculated from the sum of the pulses and a theoretical number of impulses resulting from slip-free feed. In the case of different feed sizes, an actuating signal for a sewing machine equipped with a differentiable upper and lower transport device can be formed from the difference, whereby the difference in the feed size can be corrected.

With the help of this method, good results can be achieved with layers of fabric of the same length. However, there are restrictions on the accuracy when the layers of fabric are stretched differently, which can be caused, for example, by improper handling during sewing or by different manufacturing and storage conditions, especially in the case of knitted and knitted fabrics. When sewing together fabric layers that are assigned to each other to the exact length, especially when Production of long seams, for example in the case of longitudinal seams on the sleeves and trouser legs of clothing, an accuracy of less than 0.1% is required, since e.g. B. with two equally long fabric cuts of 1 m length with 1% feed difference or 1% stretch difference results in a final error of 1 cm, which is not tolerable. The known method can only compensate for a mutual offset of the two layers of material caused by different feed rates. On the other hand, different strains of the two layers of fabric cannot be detected at all and therefore cannot be compensated for.

DE-OS 3416883 discloses a method and an arrangement for the continuous, contactless measurement of the two-dimensional shrinkage of textiles, in particular knitwear, during the manufacturing process, in which a section of the textile surface is recorded, digitized and digitized before and after the shrinkage with a television camera is saved. The one-dimensional auto-correlation function in the longitudinal and transverse directions is formed with the aid of a computing unit, after which the noise-free basic period in the longitudinal and transverse directions is determined from the position of the periodic maxima of this function. The longitudinal and transverse changes caused by shrinkage are determined from the change in this basic period in both directions.

With the help of this method, two-dimensional can be used

Measure shrinkage and thus size changes of textiles, but since with the help of two cameras arranged at a distance in the direction of travel of the textile web, the state before and after the shrinkage treatment is recorded, the method and in particular the device for carrying it out for the detection the stretching state of two layers of fabric in the area of the stitch formation point of a sewing machine is unsuitable. In addition, since the known method does not measure the feed size of the two Layers of fabric can be carried out, it is completely useless for sewing two layers of fabric together.

Finally, DE-PS 3346 163 discloses a method for pattern and structure-appropriate sewing together of two layers of fabric having the same pattern or the same pattern-like surface structure. With the help of a line-shaped sensor, images of each fabric layer are taken and digitized at the same time. By means of cross-correlation analysis of the image data, a data processing device calculates the relative position of the patterns of the two material layers and, depending on the result thereof, generates control commands for an actuating device in order to change the relative feed size of the feed means of the sewing machine such that the material layers overlap in a pattern. Since it must be accepted that the end edges of the layers of fabric do not coincide, this method is also unsuitable for sewing two layers of fabric together.

The invention has for its object to provide a method and a sewing machine for sewing any structured and possibly differently stretched layers of fabric. This object is achieved by the characterizing features of the independent claims 1, 5 and 10 and 18 and 21.

In the method according to the invention, the stitching of any structured layers of material takes place in such a way that the feed size of the layers of material is first determined from non-periodic signal portions of the images showing the binding structure of the respective fabric layer and a first signal is obtained from a comparison of the feed sizes and from the signal portions from along the

Direction of extension of periodically occurring structural elements (e.g. in the direction of the warp or the weft threads in the case of fabrics), the state of stretch of the fabric layers and taking into account the Angular length of the direction of extension of the periodically occurring structural elements, the expansion component running parallel to the feed direction is determined and a second signal is obtained from a comparison of the rectified expansion components, and a control signal for the actuating device assigned to the feed device is then formed from these signals. In this way, as a result of different stretching, layers of fabric of different lengths, which had previously been cut to the same length in the same stretching state, can now be sewn together again with the same length by sewing the fabric layers with the same number of stitches or threads per unit length. Since the binding structure of the fabric layers is recorded and evaluated in this process for determining the feed and the state of stretch, the universeTl process can also be used for single-color, unpatterned fabrics.

Claim 2 specifies a further embodiment of the method and claims 3 and 4 give alternative developments of the method according to claim 2.

In claim 5, a simplified method compared to the method according to claim 1 is specified, which is applicable provided that at least part of the image-specific, surface-specific features, for. B. in a fabric, the warp or weft threads extend substantially parallel to the feed direction. In this case, since the stretching state of the fabric layers is determined in the same direction as the feed size of the fabric layers and the angular determination of the extension position of surface-specific features is unnecessary, the evaluation of the image signal values after non-periodic

Signal portions for the feed rate determination and after periodically occurring signal portions for the elongation determination each take place from the same image signal values. Therefore needs for generation of the image signals to be evaluated according to two mutually different criteria, only one image recording system should be provided for each fabric layer.

For the determination of the feed size according to claim 6 as

Similarity function calculates the cross-correlation function, while the autocorrelation function is calculated for the determination of the elongation state according to claim 7.

One possibility for increasing the uniqueness when determining the correlation maximum is achieved according to claim 8 by appropriate preprocessing of the input data. If the approximate period of the material structure is known, this can be suppressed using an appropriately sized filter (band stop). The coefficients of this filter should preferably be dynamically adjustable in order to be adaptable to the prevailing material structure. Although this does not result in any gain in information or accuracy, it does facilitate the automatic determination of the correlation maximum since ambiguities are avoided.

While the irregularities and higher-frequency components in the signal of an image line are important for the feed measurement and the period of the fabric structure is filtered out, for the autocorrelation when determining the structure width, i.e. the elongation state of the fabric layers, it is precisely the periodic, regular signal component that contains the desired information. Therefore, in contrast to what has been described above, according to claim 9, the filtering operation for the input data is carried out by means of a band pass for the autocorrelation, all higher-frequency irregularities being suppressed. The coefficients of such bandpass filters should preferably also be adjustable in order to be able to be adapted to the structure width (mesh size or thread spacing) of the respective fabric layer. In the method according to claim 10, which, like the method according to claim 5, can be used provided that at least some of the surface-specific features that can be recorded in the image extend substantially parallel to the feed direction, the

Knowledge exploited that when calculating the similarity function carried out for determining the feed size, in addition to the main extreme value, several periodic secondary extreme values occur, the spacing of which corresponds to the spacing of the periodically occurring structural elements, namely the stitches or warp or weft threads. While the position of the main extreme value of the similarity function is still calculated for the determination of the feed size, the mutual distance of the secondary extreme values or the distance between the main extreme value and the two adjacent secondary extreme values is determined from the functional values of this similarity function. In contrast to the method according to claim 5, in this method the stretching state of the layers of material is not determined by calculating its own similarity function but by a corresponding evaluation of the calculation of a similarity function to be carried out anyway for determining the feed size. According to claim 11, this similarity function is the cross-correlation function.

If the measuring rate is sufficiently high, it can be assumed that the feed rate between two measurements can only change by a slight amount due to mechanical inertia. This can be used to operate the correlation computer in accordance with claim 12 in such a way that the new displacement maximum is only sought in the vicinity of the previous maximum. This noticeably reduces the computational effort since only a limited part of the overall cross-correlation function needs to be calculated. A further reduction in the computing effort is achieved in that the normalized cross-correlation function is calculated instead of the normalized one. Any fluctuations in the basic brightness of the image, such as those that can occur due to uneven lighting or uneven reflectivity of the layers of fabric, can have the effect that there is an overall decreasing or increasing course, the correlation maximum possibly being in the "valley floor". In order to be able to determine the maximum here nevertheless, a modified algorithm according to claim 13 is used for its calculation.

According to claim 14, the interpolation points of the calculated main maximum of the cross-correlation function and the two neighboring secondary maxima are smoothed by interpolation, whereby the accuracy of the position of the actual maximum is improved by a factor of about 5.

Since the direction of advance of the fabric layers is predetermined, i.e. H. is known, the brightness values are used for evaluation in only one direction. The size of the frame, i.e. H. of the measuring range is advantageously selected in such a way that, in the maximum advancing feed step, the image is at most 50% of the detail, ie. H. of the measuring range. With a corresponding change in the correlation algorithm, the shift could also be more than 50% of the image section and still be recognized, although with increasing shift a deterioration in the correlation values would have to be accepted.

To illuminate the fabric surfaces according to claim 16 Infrared! Not used. Infrared lighting offers several advantages, namely the elimination of color differences, since most of the common textile colors are infrared-permeable and therefore invisible in infrared light, the largely insensitivity to daylight and stray light effects, which makes shielding unnecessary and the avoidance of a disturbing light for the human eye. To avoid motion blur, it is advantageous if the infrared light source is operated stroboscopically in time with the image acquisition.

An advantageous embodiment of the sewing machine for carrying out the various methods according to claims 1, 5 or 10 results from claims 18 to 22.

The invention is explained with reference to four exemplary embodiments shown in the drawing. It shows:

1 is a side view of a sewing machine with two cameras and two lighting devices,

2 shows a block diagram of a signal processing device with which a cross-correlation analysis is carried out to determine the feed size and an auto-correlation analysis is carried out to determine the state of stretch of the fabric layers, at least some of the structural elements running essentially parallel to the feed direction,

Fig. 3 is a block diagram of a second embodiment of a signal processing device with which to determine the

Feed size a cross-correlation analysis and to determine the state of stretch of the fabric layers an evaluation of in the cross-correlation function contained periodic signal components is carried out, with at least some of the structural elements running essentially parallel to the feed direction

4 shows a diagram of a normalized cross-correlation function,

5-8 diagrams of the signal-related effects of various calculation steps in the calculation of an abnormal cross-correlation function,

9 is a block diagram of a third embodiment of a signal processing device in which a line sensor is used to determine the feed size and a surface sensor is used to determine the angular position of the structural elements of each fabric layer and the state of elongation.

Fig. 10 is a block diagram of a fourth embodiment of a signal processing device in which only one area sensor is provided for each fabric layer, which is for the

Determining the feed size can be read out line by line, and

11 shows a schematic representation of a fabric in which the warp and weft threads form first and second periodically occurring structural elements.

Exemplary embodiment 1 The sewing machine shown only partially in FIG. 1 has one

Base plate 1 and a head 2. In the head 2, the fabric presser rod 4 carrying the usual presser foot 3 and the needle bar 5 are received, the thread-guiding needle 6 of which is not shown Grapple works together. For advancing two layers of fabric 7, 8 to be connected to one another, the sewing machine has an upper fabric pusher 9 and a lower fabric pusher 10.

The lower fabric slide 10 is received by a carrier 11, the fork-shaped end of which engages around an eccentric 12, which is arranged on a shaft 13 mounted in the base plate 1 and which gives the fabric slide 10 a lifting movement per stitch-forming process. The other end of the carrier 11 is connected to a crank 14 which is fastened on a shaft 15 which is also mounted in the base plate 1.

The shaft 15 is driven by an adjustable drive mechanism, not shown, which, like the drive mechanism shown in DE-PS 3346 163 in FIG

15 designated shaft is constructed and works in the same way.

The presser bar 4 is at its lower end with a crossbar

16 provided which carries a pin 17. On the pin 17 a handlebar 18 is mounted, which is articulated by means of a hinge pin 19 with the upper slider 9. This is constantly pressed downwards by a spring-loaded ball 20 and receives its lifting movement from a lever 21 pivotally mounted on the crossbar 16, the free end of which engages under a roller 22 carried by two lateral bearing webs of the upper fabric slide 9. The other end of the lever 21 is connected to an angle lever 24 via an intermediate member 23.

The angle lever 24 is connected to an eccentric drive, not shown, which is shown in DE-PS 3346 163 in Fig. 3

Eccentric drive for driving the angle lever designated there corresponds to 48 and is used for lifting the upper fabric slider 9 in time with the stitch formation. An intermediate link 25, which is connected by a pivot pin 26 to a rocker arm 27, acts on the pin 19 to drive the upper fabric slide 9. The rocker arm 27 is connected to a drive mechanism, not shown, which, like the drive mechanism shown in DE-PS 3346 163 in FIG. 3, is constructed for the rocker arm designated 58 there and functions in the same way.

To the feed size of the upper fabric slide 9 relative to

To be able to change the feed size of the lower material slide 10, a schematically illustrated actuating device 28 is provided which, like the actuating device 80, is constructed from DE-PS 3346 163 and accordingly contains, among other things, a stepping motor, not shown here.

A line camera 30 and a lighting device 31 are arranged on a support 29 fastened to the front of the head 2.

Arranged below a glass plate 33 in front of the stitch formation point in the stitch plate 32 is a light guide bundle 34 which is surrounded by a light guide bundle 35 insulated from it. The inner light guide bundle 34 is connected to a line camera 36 and the outer light guide bundle 35 to an annular lighting device 37. Beneath the glass plate 33 there is an optical system, not shown, which enables targeted illumination of the measuring surface and which in turn images this on the end face of the inner light guide bundle 34. The light beams of the lighting devices 31, 37 are mutually shielded by an intermediate plate 38 separating the two layers of material 7, 8. A shaft 39 running synchronously with the shaft 15 FIG. 2 carries a pulse disk 40 which interacts with a pulse generator 41.

Each line camera 30, 36 has a line sensor 42 with rectangular diode elements 43. The diode elements 43 are aligned transversely to the direction of advance of the layers of material 7, 8.

The line camera 30 is assigned an A / D converter 44, which can be alternately connected to one of two image memories 46, 47 via an electronic switch 45. The image memory 46 is optionally connected via an electronic switch 48 to a digital filter 49 acting as a bandpass filter and a digital filter 50 acting as a bandstop filter. The image memory 47 is connected via an electronic switch 51 to a digital filter 52 which acts as a band stop. The filters 49, 50 and 52 are connected to a correlation computer 53.

The line camera 36 is connected in the same way as the line camera 30 to a circuit which consists of an A / D converter 44 ', a first electronic switch 45', two image memories 46 ', 47' and two further electronic switches 48 ', 51'. , acting as a band pass digital filter 49 ', two acting as a band-stop digital filters 50', 52 ^l and a correlation calculator 53 'is composed.

The two correlation computers 53, 53 'are connected to a comparator module 54, to which a feed control module 55 is connected. The feed control module 55 is connected to the stepping motor, not shown, of the actuating device 28.

The operation of the line scan cameras 30, 36, the lighting devices 31, 37, the switches 45, 45 ', 48, 48' and 51, 51 'and the comparator module 54 is controlled by a sequence controller 56, which is connected to the machine engine via the pulse generator 41 in FIG connection stands. The components or modules 44 - 53, 44 '- 53' and 54 - 56 together form a signal processing device 57.

Example 2

The sewing machine of this embodiment is designed in the same way as the sewing machine of the first embodiment. Furthermore, two line cameras are used, which are identical to the line cameras 30, 36 of the first exemplary embodiment and therefore have been given the corresponding reference numerals 30, 36.

In the block diagram of the second exemplary embodiment shown in FIG. 3, the line camera 30 is assigned an A / D converter 60 which can be alternately connected to one of two image memories 63, 64 via a digital filter 61 acting as a high-pass filter and an electronic switch 62. The image memories 63, 64 are connected to a correlation computer 65.

The line camera 36 is connected in the block diagram according to FIG. 3 in the same way as the line camera 30 with a circuit which consists of an A / D converter 60 ', a digital filter 61' acting as a high-pass filter, two image memories 63 ', 64' and a correlation calculator 65 '.

The two correlation computers 65, 65 'are connected to a comparator module 66 to which a feed control module 67 is connected. The feed control module 67 is connected to the stepping motor, not shown, of the actuating device 28.

The operation of the line scan cameras 30, 36, the lighting devices 31, 37, the switches 62, 62 'and the comparator module 66 is controlled by a sequence control 68 which is connected to the machine engine via the pulse generator 41. The components or Modules 60 - 65, 60 '- 65' and 66 - 68 together form a signal processing device 69.

Embodiment 3

The sewing machine of this exemplary embodiment is configured in the same way as the sewing machine of the first exemplary embodiment shown in FIG. 1 and therefore, like this, has the rocking lever 27 shown in FIG. 9, the actuating device 28 and the pulse disk 40 fastened on the shaft 39 and the pulse generator 41 on.

A schematically illustrated camera 70 and an illumination device 71 are assigned to the upper fabric layer 7. The camera 70 essentially consists of a housing 72, an optics 73, a partially transparent mirror 74 and a CCD line sensor 75 and a CCD area sensor 76.

A camera 77, also shown schematically, is assigned to the lower layer of fabric 8, which is constructed essentially like the camera 70 and therefore like this from a housing 78, an optical system 79, a partially transparent mirror 80 and a CCD line sensor 81 and a CCD area sensor 82 exists. A light guide bundle 83 is connected to the camera 77 and ends below the glass plate 33 shown in FIG. 1. The light guide bundle 83 is partially surrounded by a second light guide bundle 84, which is connected to an annular lighting device 85.

An A / D converter 86 is assigned to the line sensor 75 and can be connected alternately to one of two image memories 88, 89 via an electronic switch 87. The image memories 88, 89 are connected to a KKF computer 90 which calculates cross-correlation functions. The KKF computer 90 contains a digital filter, not shown, which serves as a bandstop filter. An A / D converter 91 is assigned to the area sensor 76 and is connected directly to an image memory 92. An angle calculation unit 93 and an AKF computer 94 calculating autocorrelation functions are connected to the image memory 92, the angle calculation unit 93 and the AKF computer 94 also being connected to one another. The AKF computer 94 contains a filter, not shown, which acts as a bandpass filter.

The line sensor 81 is connected in the same way as the line sensor 75 to a circuit which consists of an A / D converter 86 ', an electronic switch 87', two image memories 88 ', 89' and a KKF computer 90 '. Likewise, the area sensor 82, like the area sensor 76, is connected to a circuit which consists of an A / D converter 91 ', an image memory 92', an angle calculation unit 93 'and an AKF computer 94'.

The two KKF computers 90, 90 'and the two AKF computers 94, 94' are connected to a comparator module 95, to which a feed control module 96 is connected. The feed control module 96 is connected to the stepping motor, not shown, of the actuating device 28.

The operation of the cameras 70, 77, the lighting devices 71, 85, the switches 87, 87 'and the comparator module 95 is controlled by a sequence control 97 which is connected to the

Machine engine is connected. The components or modules 86-94, 86'-94 'and 95-97 together form a signal processing device 98.

Example 4

The sewing machine of this embodiment is configured in the same way as the sewing machine of the first embodiment shown in FIG. 1 and therefore, like this, has the one in FIG. 10 shown rocker arm 27, the actuating device 28 and the pulse disk 40 attached to the shaft 39 and the pulse generator 41.

A schematically represented camera 100 and an illumination device 101 are assigned to the upper fabric layer 7. The camera 100 essentially consists of a housing 102, an optical system 103 and a CCD area sensor 104.

The lower fabric layer 8 is assigned a camera 105, also shown schematically, which is constructed essentially like the camera 100 and therefore, like this, consists of a housing 106, an optics 107 and a CCD area sensor 108. A fiber optic bundle 109 is connected to the camera 105 and ends below the glass plate 33 shown in FIG. 1. The light guide bundle 109 is partially surrounded by a second light guide bundle 110, which is connected to an annular lighting device 111.

An A / D converter 112 is assigned to the camera 100 and can be connected alternately to one of two image memories 114, 115 via an electronic switch 113. The two image memories 114, 115 are connected to a KKF computer 116 which calculates cross-correlation functions. The KKF computer 116 contains a digital filter, not shown, which serves as a bandstop filter. An angle calculation unit 117 and an AKF calculator 118 calculating autocorrelation functions are also connected to the image memory 114, the angle calculation unit 117 and the AKF calculator 118 also being connected to one another. The AKF calculator 118 contains a filter, not shown, which acts as a bandpass filter.

In the same way as the camera 100, the camera 105 is connected to a circuit comprising an A / D converter 112 ', an electronic switch 113', two image memories 114 ', 115', one KKF calculator 116 ', an angle calculation unit 117' and an AKF calculator 118 '.

The two KKF computers 116, 116 'and the two AKF computers 118, 118' are connected to a comparator module 119, to which a feed control module 120 is connected. The feed control module 120 is connected to the stepping motor, not shown, of the actuating device 28.

The operation of the cameras 100, 105, the lighting devices 101, 111, the switches 113, 113 'and the comparator module 119 is controlled by a sequence controller 121, which is connected to the machine engine via the pulse generator 41. The components or modules 112-118, 112 '- 118' and 119-121 together form a signal processing device 122.

functionality

Example 1

In the first exemplary embodiment according to the block diagram according to FIG. 2, the area of the fabric layers 7, 8 to be imaged by the cameras 30, 36 is illuminated by the lighting devices 31, 37 at stroboscopic intervals with infrared light at intervals of the stitch period, which means that due to the meshes Knitted or knitted goods or the intersecting warp and weft threads of a fabric-induced rough surface structure of the fabric layers 7, 8, punctual differences in brightness occur on the upper side thereof. With the help of the cameras

30, 36 become simultaneous with the operation of the lighting devices

31, 37 cell-shaped images of the illuminated surface sections, each individual one of the rectangular diode elements 43 effecting an integral brightness measurement of the surface section section detected by it. The analog image data of the respective first image of the two cameras 30, 36 are after conversion in the A / D converter 44; 44 'via the switch 45; 45 'as a digital gray-scale image or as a signal profile in the image memory 46; 46 'cached. The second image taken by the cameras 30, 36 after a feed step of the fabric layers 7, 8, which is offset by the length of the feed step from the corresponding first image, is switched after the switch 45; 45 'also as a digital gray-scale image in the image memory 47; 47 'cached.

In the meantime until the next feed step is carried out and the next picture is taken, the image data are evaluated with regard to the feed size and the stretching state of the fabric layers 7, 8.

In order to determine the feed size, the cross-correlation function, abbreviated to KKF in the following, is used in the image memories 46, 47; 46 ', 47' performed cached image data. The maximum of the KKF is proportional to the size of the

Feed of the two layers of fabric 7, 8 between the first and second recording. This maximum of the KKF is due to the fact that although the layers of material 7, 8 consist of periodic structure elements, these structures do not occur at absolutely the same periodic intervals, but contain a relatively large superimposed noise component with broadband frequency components due to slight distortion and deformation. This is a non-periodic component that leads to a pronounced maximum of the KKF and allows a clear statement about the feed.

When calculating a standardized KKF, a function curve would result that is comparable to the diagram shown in FIG. 4. The standardized KKF results in an absolute maximum H der Correlation coefficient R _χy for the shift _index K _v , the distance to the zero point of which _corresponds to the size of the local offset of the two images taken with one and the same camera 30 or 36 in succession and thus the feed size of the layers of material 7, 8 between the exposures.

Since the calculation of the standardized KKF would take too long due to the large number of terms to be calculated, a simplified, non-standardized KKF is calculated. Before the image data for the calculation of the non-standardized KKF to the respective

Correlation computers are forwarded in the filters 50, 52; 50 ', 52', the periodic component resulting from the periodic arrangement of the structural elements is reduced, as a result of which the ordinate of the secondary maxima is reduced in the subsequent calculation of the KKF and thus the determination of the main maximum is made easier. The coefficients of the filters 50, 52; 50 ', 52' can be adjusted dynamically, which means that they can always be adapted to the prevailing fabric structure.

With the help of the two correlation computers 53, 53 ', the feed size of each fabric layer 7, 8 is calculated from the position of the main maximum H at the displacement index K _v and taking into account the image recording frequency of the cameras 30, 36. The instantaneous feed quantities of the two layers of material are then compared with one another in the comparator module 54 and a first signal is obtained from the comparison result, which is temporarily stored in a register of the feed control module 55.

For the subsequent calculation of the stretching state of the fabric layers 7, 8, the autocorrelation function, abbreviated as AKF, is calculated from the image data of an image by correlating the image data with itself in a known manner. Here, the fact is used that - as already mentioned - in the by the Cameras 30, 36 recorded images also contain periodically occurring image information, which result from the binding structure of the fabric layers _r, ie from the stitches in knitted or knitted goods or the intersecting warp and weft threads in fabrics. To calculate the AKF, switches 48, 51; 48 ', 51' switched, whereupon the image memories 46, 46 'are connected to the filters 49, 49' acting as a bandpass filter, while the image memories 47, 47 'are decoupled from the subsequent processing of the image data. The filters 49, 49 'reduce the higher-frequency irregularities, so that the periodic regular image information is more clearly visible. In the calculation of the AKF subsequently carried out by the correlation computer 53, 53 ', there is always a maximum of similarity if the image data of an image line are shifted from one another by exactly one or more periods, ie distances of the structural elements. From the position of these maxima, the mean period length can be calculated, which is a measure of the state of stretch of the fabric layers 7, 8. The instantaneous stretch dimensions of the two layers of material are then compared with one another in the comparator module 54 and a second signal is obtained from the comparison result.

The AKF can also be calculated simultaneously with the calculation of the KKF if the correlation computers are designed appropriately or if two correlation computers are used in parallel, so that the two signals are not generated in succession but at the same time.

In this case, however, for an accurate determination of the elongation state, at least some of the periodically occurring image information or structural elements should be aligned essentially parallel to the feed direction of the sewing machine and thus parallel to the alignment position or measurement direction of the line sensors 42, i. h that z. B. in a fabric either the warp threads or the weft threads run essentially parallel to the measuring direction. Otherwise, would two different frequencies fl and f2 occur in the AKF. These frequencies fl and f2 are the frequency of the warp and of the weft threads (or comparable structures) perpendicular thereto, which are a function of the angle a between the measuring direction (feed direction) and the direction of extension, e.g. B. the warp threads result.

Problems arise because the frequencies fl and f2 are not known separately, but only their product is contained in the AKF. The superimposition of the frequencies fl and f2 results in phase jumps in the measured resulting frequency, which can falsify the result when determining the state of stretch of the fabric layers.

The second signal and the temporarily stored first signal together form a measure of the effective feed size of the fabric layers 7, 8, d. H. for the number of structural elements per unit length.

The first and second signals are processed in the feed control module 55 to form a control signal for the actuating device 28, which adjusts the feed size of the upper fabric slide 9 in such a way that both layers of fabric 7, 8 are transported with the same effective feed rate, in which even when the stretching state was originally different always the same number of

Structural elements is moved under the needle 6 per feed step. If the layers of fabric to be sewn together had been cut to the same length with the same stretching state, they are now sewn together with the same length if their stretching state changes differently after cutting, for example during storage or during sewing.

Execution example 2 In the second exemplary embodiment according to the block diagram according to FIG. 3, the image data recorded by the two cameras 30, 36 and converted into digital gray-scale images in the A / D converters 60, 60 'is evaluated differently than in the first exemplary embodiment, by determining the The stretched state of the fabric layers 7, 8 does not carry out a calculation of the AKF, but the KKF is also evaluated with regard to the information relevant to the stretched state. It is assumed here that in the calculation of the KKF, in addition to the main maximum, there are several periodic secondary maxima, the spacing of which corresponds to the spacing of the periodically occurring structural elements.

When calculating the standardized KKF, which would result in a functional sequence corresponding to the diagram in FIG. 4, equally high secondary maxima N, the height of which would be lower than that of the main maximum H, would occur with an even spacing on both sides of the main maximum The distance between the secondary maxima N and the distance between the main maximum H and the two adjacent secondary maxima N is given by the letters s to S5.

Since in this case too, the calculation of the standardized KKF would take too long due to the large number of terms to be calculated, a simplified, non-standardized KKF is also calculated. Before the image data for the calculation of the non-standardized KKF are forwarded to the respective correlation computer, they are first pretreated in the filters 61, 61 ', which act as high-pass filters, in such a way that the periodic signal components are attenuated, but not completely suppressed. After passing through the filters 61 61 ', the image data of the respective first image are assigned to the associated image memory 63; 63 'and that of the next recording in each case the image memory 64; 64 ', whereupon they are stored in the corresponding correlation computer 65; 65 'are processed together. If the KKF is not normalized, fluctuations in the basic brightness of the image, e.g. B. may be caused by uneven lighting or reflectivity of the layers of material, to the effect that the KKF takes a decreasing or increasing course, the correlation maximum is in the "bottom". The solid line in FIG. 5 is an example of such an irregular course of an non-standardized KKF, the correlation coefficient R _{χy being} plotted on the ordinate and the displacement _index k on the abscissa. In order to determine the effective correlation maximum under these conditions, that is to say the main maximum proportional to the feed, an algorithm is used in which the quality of a maximum is related to its local environment.

The individual steps are as follows:

a First, all local maxima and minima of the function are determined, which results in a value curve shown by the dashed line in FIG. 5. Here, all maxima, including noise, etc. are also recorded.

b All maxima and minima are eliminated, the height or

Depth lies below a threshold, which is determined adaptively to the dynamics of the KKF, d. that is, the flanks of which fall below a predetermined minimum length. In this way, the curve shown in dashed lines in FIG. 6 results.

c Then a height is calculated for each valid maximum, which results from the sum of the amplitude difference to the neighboring minima.

d From these values, the maximum is determined which has the greatest height difference from the adjacent minima having. The main maximum H determined in this way and the associated minima are identified by arrows in FIG. 7.

e The exact location of the main maximum is then determined by

Interpolation further refined, using different interpolation methods depending on the shape of the maximum.

With the help of the two correlation computers 65, 65 ', the position of the main maximum H at the shift index k _v and below is also used

Taking into account the image recording frequency of the cameras 30, 36, the feed size of each fabric layer 7, 8 is calculated. The current feed sizes of the two layers of material are then compared in the comparator module 66 and a first signal is obtained from the comparison result, which is stored in a register of the

Feed control module 67 is temporarily stored.

For the calculation of the stretching state of the fabric layers 7, 8, the period length and thus the mutual distance between the periodically occurring structures causing the secondary maxima is determined by averaging the distances between adjacent maxima of the KKF, as a result of which the distances s- ^ to S7 indicated in FIG. 8 between two secondary maxima N or between the main maximum H and the neighboring secondary maxima N. "Outliers" can be eliminated by statistical methods.

When determining the period length or the period of the weave structure in accordance with the mesh or thread spacing, it must be taken into account that practically all occurring substances are an irregular and disturbed periodic structure, which makes displacement measurement possible. This means that the resulting KKF in addition to periodic maxima in the Rule also includes overlaid faults, which must not be included in the period measurement.

Other side effects occur when the tissue under consideration contains several frequencies, e.g. B. by ribbing or similar structures, which, however, in many cases have an integer frequency ratio to the thread period. In such cases, it must be ensured that the same period is always reproducibly measured.

Furthermore, as in embodiment 1, for an accurate

Determination of the elongation state at least some of the periodically occurring image information is essentially parallel to the measuring direction of the line sensors 42.

The instantaneous elongation dimensions of the two layers of material 7, 8 are then compared with one another in the comparator module 66 and a second signal is obtained from the comparison result.

The second signal and the temporarily stored first signal together form a measure of the effective feed size of the fabric layers 7, 8. The two signals are processed in the feed control module 67 to form a control signal for the actuating device 28, which sets the feed size of the upper fabric slide 9 so that both Layer of fabric 7, 8 can be transported with the same effective feed.

Example 3

With the aid of the line sensors 75, 81 of the cameras 70, 77, line-shaped images of the illuminated surface sections of the fabric layers 7, 8 are recorded simultaneously with the lighting devices 71, 85 which operate stroboscopically with the frequency of the stitch formation. After conversion in the A / D converter 86 or 86 ', the analog image data of the respective first image of the two line sensors 75, 81 are transmitted via the switch 87 or 87' as a digital gray value image or as a signal profile in the image memory 88 or 88 'cached. That after performing a

Feeding step of the layers of material 7, 8 by the line sensors 75, 81, the second image, which is offset by the length of the feed step from the corresponding first image, is also a digital gray value image in the image memory 89 and 89 after switching the switch 87 or 87 ' 'cached.

In the meantime until the next feed step is carried out and the next picture is taken, the temporarily stored image data are evaluated with regard to the feed size of the fabric layers 7 _r 8. This is done by calculating the cross-correlation function abbreviated with KKF in the KKF computers 90, 90 'with the image data of the assigned image memories 88, 89 and 88', 89 'for each fabric layer.

The maximum of the KKF is proportional to the size of the feed of the two layers of material 7, 8 between the first and second receptacles. This maximum of the KKF is due to the fact that although the layers of material 7, 8 consist of periodic structural elements S1, S2 (FIG. 11), these structures do not occur at absolutely the same periodic intervals, but rather a relatively large superimposed noise component due to slight warpage and deformation with broadband frequency components included. This is a non-periodic component that leads to a pronounced maximum of the KKF and allows a clear statement about the feed.

When calculating a standardized KKF, a function curve would result that is comparable to the diagram shown in FIG. 4. The standardized KKF results in an absolute maximum H der Correlation coefficients for the shift index K- V ,, whose distance from the zero point corresponds to the size of the local offset of the two images taken in succession with one and the same line sensor 75 or 81 and thus the feed size of the fabric layers 7, 8 between the images.

Since the calculation of the standardized KKF would take too long due to the large number of terms to be calculated, a simplified, non-standardized KKF is calculated. The KKF computer 90, 90 'is used to reduce the periodic component resulting from the periodic arrangement of the structural elements with the aid of the unspecified filter acting as a bandstop, whereby the ordinate of the secondary maxima is reduced in the subsequent calculation of the KKF and thus the determination of the Main maximum is made easy.

With the help of the two KKF computers 90, 90 ', the feed size of each fabric layer 7, 8 is calculated from the position of the main maximum H at the displacement index K _v and taking into account the image recording frequency of the cameras 70, 77. The instantaneous feed sizes of the two layers of material are then compared with one another in the comparator module 95 and a first signal is obtained from the comparison result, which is temporarily stored in a register of the feed control module 96.

Overlapping in time with the work of the line sensors 75, 81, with the help of the surface sensors 76, 82, flat images of the illuminated surface sections are recorded. After conversion in the A / D converter 91, 91 ', the analog image data of these images are temporarily stored as a matrix-like digital gray value image in the image memory 92 or 92', the gray value image consisting of one of the number of pixel rows and pixel columns of the CCD area sensor 76 or 82 dependent number of rows and columns distributed pixels. With the angle calculation unit 93 or 93 ', the angle a (FIG. 11) between the direction of extension R1 of first periodically occurring structural elements S1 (for example the warp threads of a fabric) and the direction of advance V of is now made from the image data of the image memory 92 or 92' Fabric layers 7, 8 determined. This is done in such a way that, starting from any fictitious point of contact within the data field contained in the image memory 92, 92 ', a multiplicity of angularly offset search beams, each passing through the defined point of contact, is placed over the data field and the signal values of all of each search beam pixels lying on it are added. Depending on the position of the point of contact and the structure of the layers of material, a maximum or minimum of the addition results now shows the direction of extension R1 of the first periodically occurring structural elements S1 of the respective material layer.

Then, using the AKF computers 94 and 94 'along the previously determined direction of extension R1 of the first structure elements S1, the mean distance or the period length P2 from the angle z. B. 90 ° to this second structural elements S2 (for example, the weft threads of a fabric) and thus the stretching state of the fabric layers 7, 8 parallel to the previously determined direction of extension Rl of the first structural elements S1 determined by the pixels running parallel to the direction of extension Rl of the pixels AKF abbreviated autocorrelation function is calculated. This calculation is carried out in a known manner in that the signal values or image data are correlated with themselves. The filters of the AKF computers 94 and 94 ', which are not specified in more detail and act as a bandpass filter, reduce the higher-frequency irregularities in the signal profile, so that the periodic regular image information is more clearly visible. When calculating the AKF, there is always a maximum of similarity if the image data is parallel to the previous one ascertained direction of extension Rl, selected for this calculation, are shifted against themselves by exactly one or more period lengths P2 of the second structure elements S2. From the position of these maxima, the mean period length P2 can be calculated, which is a measure of the state of elongation of the fabric layers 7, 8 in the direction of extension R1.

From the period length P2 of each fabric layer 7, 8, taking into account the determined angle a, the value P2 _{V of} the period length P2 running parallel to the feed direction _V and thus the state of stretching parallel to the feed direction V are determined. These elongation values of the two layers of material 7, 8 are compared with one another in the comparator module 95 and a second signal is obtained from the comparison result.

In the feed control module 96, the value of the controlled variable R is obtained from the first and second signals according to the formula R = Do / Du, by relating the throughput Do of the upper layer of material to the throughput Du of the lower layer of material. The throughput Do or Du is the measure of the effective feed of the respective fabric layer 7 or 8, d. H. the determined number of structural elements S2 per feed length, the throughput being calculated as follows:

Do or Du = determined feed length period length P2 / cos a

If the ratio Do / Du is not equal to 1, a control signal for the actuating device 28 is formed in the feed control module 96, which adjusts the feed size of the upper slider 9 in such a way that both layers of material 7, 8 are transported with the same throughput or the same effective feed in which the same number of structural elements S2 is always the same, even when the stretching state is originally different Feed step is moved under the needle 6. If the layers of fabric to be sewn together had been cut to the same length with the same stretching state, they are now sewn together with the same length if their stretching state changes after cutting, for example during storage or during sewing.

Since the computational effort for the determination of the state of elongation is much more time-consuming than the computational effort for the determination of the angle owing to the previously determined angle a

When the sewing machine is rotating at high speed and taking into account the computing speed that can currently be achieved using commercially available, industrially usable computers, the elongation state can only be calculated after about every tenth feeding step. Because of this U is during the

Stitch formation processes or feed steps during which no new stretch state was determined, the value of the controlled variable R is calculated from the current feed values and the last determined stretch value of each fabric layer. Since experience has shown that the stretching state of the fabric layers does not change abruptly over short distances, a sufficiently precise feed control is achieved despite the distance between the presence of new stretch values, for example ten steps.

Example 4

With the aid of the surface sensors 104, 108 of the cameras 100, 105, flat-shaped images of the illuminated surface sections of the fabric layers 7, 8 are recorded simultaneously with the lighting devices 101, 111 which operate stroboscopically with the frequency of the stitch formation.

The analog image data of the respective first image of the two area sensors 104, 108 are after conversion in the A / D converter 112 or 112 'via the switch 113 or 113' as a matrix-like digital gray-scale image in the image memory 114 or 114 ', each gray-scale image being distributed in rows and columns in a number dependent on the number of pixel rows and pixel columns of the area sensor 104 or 108 There are pixels.

The second image taken by the surface sensors 104, 108 after a feed step of the fabric layers 7, 8, which is offset by the length of the feed step relative to the corresponding first image, is also switched as a matrix-like digital gray value image after switching the switch 113 or 113 ' Image memory 115 or 115 'buffered. In the meantime until the execution of the next feed step and the taking of the next image, the image signals of a selected row of pixels 123 or 124 running parallel to the feed direction become the

Area sensors 104, 108 evaluated for determining the feed size. This is done in the same way as described in embodiment 3, in the KKF computers

116, 116 'with the image data of the pixel rows 123, 124 temporarily stored in the corresponding image memories 114, 115 and 114', 115 'for each material layer, the cross-correlation function abbreviated with KKF is calculated.

The feed variables of the two material layers determined by the cross-correlation are then compared with one another in the comparator module 119, and a first signal is obtained from the comparison result, as in exemplary embodiment 3.

The image data 114, 114 'are overlapped in time with the calculation of the KKF using the angle calculation units

117, 117 'in the same way as in embodiment 3, the angle a (FIG. 11) between the direction of extension R1 of first periodically occurring structural elements S1 and the direction of advance V of Fabric layers 7, 8 determined. Then, with the aid of the AKF computers 118, 118 ', the stretching state of the fabric layers 7, 8 is calculated in the same way as in the exemplary embodiment 3 along the previously determined direction of extension R1. The elongation values of the two layers of material are compared with one another in the comparator module 119 and a second signal is obtained from the comparison result.

In the feed control module 120, the value of the controlled variable R is determined from the first and second signals in the same way as in embodiment 3. If the ratio of the throughput of the two layers of material 7, 8 is not equal to 1, a control signal for the actuating device 28 is formed in the feed control module 120, which sets the ^' feed size of the upper slide valve 9 in such a way that both layers of material 7, 8 each have the same throughput or the same effective feed can be transported.

Since in this exemplary embodiment, similar to exemplary embodiment 3, the computational effort for determining the stretching state is very much greater than the computational effort for determining the feed size, and therefore the stretching state is only calculated after approximately every tenth • feed step, the stitch formation processes or feed steps are carried out during those , during which no new stretch state was determined, the value of the controlled variable R is calculated from the current feed values and the last determined stretch value of each fabric layer.

Claims

1. Method for sewing two layers of fabric together using one

Sewing machine, with a feed device with a lower and an upper feed means, the feed size of which can be changed relative to one another by at least one adjusting device, with at least one image sensor arranged in the feed direction in front of the stitch formation point for each fabric layer for detecting surface-specific features and a signal processing device which detects those of the image sensors compares the signals supplied with one another in order to control the actuating device in the sense of a relative change in the feed sizes of the feed means, as a function of the result of this comparison, characterized by the following steps:

a) images of the typical characteristics of the weave structure of each fabric layer are taken in rapid succession, partially overlapping images, digitized and temporarily stored as gray-scale image signals,

b) by calculating a first similarity function of the overlapping image signals, the extreme similarity value is determined and from this the distance between the two

Determines the feed size of each fabric layer in accordance with the position of the image, taking into account the image acquisition frequency,

c) the current feed quantities of the two layers of material are compared with one another or with a feed setpoint and a first signal is obtained from the comparison result, d) the angular position of periodically occurring first structural elements of each fabric layer with respect to the feed direction is determined from the image signals of an image of each fabric layer,

e) from the image signals of an image of each fabric layer in the direction of the previously determined angular position of the structural elements by determining a second similarity function, the elongation state of each fabric layer is determined by the average distance of im

Periodically occurring second structure elements running at an angle to the first structure elements is determined,

f) the elongation state of the two layers of material is compared with one another and / or with a predefinable elongation setpoint and a second signal is obtained from the comparison result and

g) the first and second signals are processed into a control signal for the actuating device.

2. The method according to claim 1, characterized in that for the calculation of the feed size of the fabric layers, image data of line-shaped images and for the determination of the angular length of the periodically occurring structural elements and for the calculation of the elongation state of the fabric layers, image data of sheet-like images are used.

3. The method according to claim 2, characterized in that the line-shaped and area-like distributed image data are generated separately. 2> 6

4. The method according to claim 2, characterized in that the line-shaped distributed image data are obtained from the surface-distributed image data.

5. The method according to the preamble of claim 1 for sewing two layers of fabric together, in which at least some of the surface-specific features detectable by the image sensors run essentially parallel to the feed direction, characterized by the following steps:

a) images of the typical characteristics of the weave structure of each fabric layer are taken in rapid succession, partially overlapping images, digitized and temporarily stored as gray value image signals,

b) by calculating a similarity function of the overlapping image signals, the extreme similarity value is determined and from its position corresponding to the distance between the two images, taking into account the

Image acquisition frequency determines the feed size of each fabric layer,

c) the current feed sizes of the two layers of material are compared with one another and a first signal is obtained from the comparison result,

d) by calculating a second similarity function from the image signals of one of the two images, the elongation state of each layer of material is determined by determining the average distance between periodically occurring structural elements, e) the elongation state of the two layers of material is compared with one another and / or with a predefinable elongation target value and a second signal is obtained from the comparison result and

f) the first and second signals are processed into a control signal for the actuating device.

6. The method according to one or more of claims 1 to 5, characterized in that the similarity function calculated for the determination of the feed size is the cross-correlation function.

7. The method according to one or more of claims 1 to 5, characterized in that the similarity function calculated for the determination of the elongation state is the autocorrelation function.

8. The method according to claim 6, characterized in that for

Determination of the cross correlation maximum from the image signals, a signal component containing the periodic component of the material structure is suppressed.

9. The method according to claim 7, characterized in that all higher-frequency irregularities are suppressed for the implementation of the autocorrelation.

10. The method according to the preamble of claim 1 for sewing two layers of fabric together, in which at least some of the surface-specific features detectable by the image sensors run essentially parallel to the feed direction, characterized by the following steps: a) images of the typical characteristics of the weave structure of each fabric layer are taken in rapid succession, partially overlapping images, digitized and temporarily stored as gray-scale image signals,

Image acquisition frequency determines the feed size of each fabric layer,

d) by evaluating the mutual distance of the secondary extreme values that occur periodically when structural elements are used to calculate the feed size and determine the extension function, the elongation state of each layer of material is determined,

e) the elongation state of the two layers of material is compared with one another and / or with a predefinable elongation setpoint and a second signal is obtained from the comparison result and

f) the first and second signals become one

Control signal for the actuator processed.

11. The method according to claim 10, characterized in that the similarity function is the cross-correlation function.

12. The method according to claim 6 or 11, characterized in that the new signal maximum is sought in the vicinity of the previous signal maximum when calculating the cross-correlation function.

13. The method according to claim 6, 11 or 12, characterized in that an abnormal cross-correlation function is calculated to shorten the computing method, the following steps being carried out in succession to determine the quality of a maximum based on the local environment:

a) all local maxima and minima are determined,

b) such maxima and minima are eliminated whose

Flanks fall below a predeterminable minimum length,

c) for each valid maximum a height is calculated, which results from the sum of the amplitude difference to the neighboring minima,

d) the maximum with the greatest height difference is used as the main maximum.

14. The method according to claim 13, characterized in that the support points of the calculated main maximum and the two adjacent secondary maxima of the cross-correlation function are smoothed by interpolation.

15. The method according to claim 6 and 11 to 14, characterized in that the images recorded successively for the cross-correlation function overlap substantially by not less than 50%.

16. The method according to one or more of the preceding claims, characterized in that the layers of fabric are not illuminated with infrared and only the infrared light component reflected by them is detected by the sensors.

17. The method according to one or more of the preceding claims, characterized in that the layers of material are illuminated stroboscopically.

18. Sewing machine for performing the method according to claims 1 to 3, characterized in that each layer of fabric (7; 8) has a camera (70; 77 or 100; 105) with an optical system (73; 79 or 103;

107) and an image sensor system (75, 76; 81, 82 and 104;

108) is assigned, which can take both line-shaped and area-shaped images, and that the cameras

(70; 77 and 100; 105) are connected to a signal processing device (98; 122) which has at least one A / D converter (86, 91; 86 ', 91' and 112, 112 ') for each camera at least two image memories (88, 89, 92; 88 ', 89', 92 'or 114, 115; 114', 115 ') and a first computer (90; 90' or 116; 116 ') for calculating the similarity function for the feed determination, as well as an angle calculation unit (93; 93 'or 117; 117') and a second computer (94; 94 'or 118; 118') connected to this for calculating the similarity function for the elongation determination , and also contains a comparator and a feed control module (95, 96 or 119, 120) which is connected to an actuating device (28) for at least one feed means (9) of the sewing machine.

19. Sewing machine according to claim 18, characterized in that each camera (70; 77) has a beam splitter (74; 80) and a line-shaped sensor (75; 81) and a flat sensor (76; 82).

20. Sewing machine according to claim 18, characterized in that each camera (100; 105) has only a single sensor (104; 108) which is flat and can be read out either in terms of area or line by the signal processing device (122).

21. Sewing machine for carrying out the method according to claims 5 and 10, characterized in that each fabric layer (7; 8) is assigned a single camera (30; 36) for obtaining the image data required for calculating the feed size and the stretching state, and each Camera (30; 36) contains only a substantially line-shaped sensor (42) which is aligned parallel to the feed direction of the sewing machine, and that the cameras (30; 36) with a

Signal processing device (57; 69) are connected, which for each camera have an A / D converter (44; 44 'or 60; 60'), each with at least two alternate image memories (46, 47; 46 ', 47') or 63, 64; 63 ', 64') and a computer (53; 53 'or 65; 65') for calculating the

Similarity functions for the feed and elongation determination, as well as a comparator and a feed control module (54, 55 or 66, 67), which is connected to an actuating device (28) for at least one feed means (9) of the sewing machine.

22. Sewing machine according to claims 18 to 21, characterized in that each camera (30; 36; 70; 77; 100; 105) has a lighting device consisting of stroboscopically controlled infrared diodes (31; 37; 71; 85; 101; 111 ) assigned.