Classifications

• • B—PERFORMING OPERATIONS; TRANSPORTING
  • B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
  • B65C—LABELLING OR TAGGING MACHINES, APPARATUS, OR PROCESSES
  • B65C9/00—Details of labelling machines or apparatus
  • B65C9/40—Controls; Safety devices
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/04—Architecture, e.g. interconnection topology
  • G06N3/045—Combinations of networks
• • G—PHYSICS
  • G06—COMPUTING; CALCULATING OR COUNTING
  • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
  • G06N3/00—Computing arrangements based on biological models
  • G06N3/02—Neural networks
  • G06N3/08—Learning methods

Definitions

• the invention relates to a method for operating a labeling system according to the preamble of claim 1, a labeling system with at least one labeling device according to the preamble of claim 15 and a data carrier with a training data set for use in such a method according to claim 16.
• the labeling systems in question here for labeling individual packs have at least one labeling device which is designed in particular as a price labeling device.
• the labeling device is equipped with at least one feed arrangement, one label dispensing arrangement and one label application arrangement as functional units, which are set up for labeling the individual packs in a labeling routine.
• the functional units are controlled in the labeling routine by means of a control arrangement.
• the feed arrangement is in particular a belt conveyor or a roller conveyor for moving the respective packs, it being possible for the moving packs to be labeled during operation.
• it is known to carry out the labeling routine with a chaotic pack feed, with different types of packs being fed to the feed arrangement in any desired sequence.
• the chaotic supply of packs usually requires at least partially automatic recognition of the respective pack with a classification in order to enable labeling related to a respective pack class.
• the packs can be automatically recognized, for example, based on weight values of the respective packs, with each pack class being assigned a weight range.
• Sensor arrangements such as cameras are also used for classification into the packaging classes, with the respective packaging class being inferred from the recorded images of the packaging, for example on the basis of the packaging geometry.
• the invention is based on the problem of specifying a method for operating a labeling system for labeling individual packs, with a particularly flexible pack supply being made possible.
• images of the respective packs are analyzed in an analysis routine by means of the control arrangement, with this analysis being used to derive a classification of the respective pack into a pack class, and the labeling device in the labeling routine being controlled as a function of the classification.
• the analysis routine be based on an application of a trained machine learning model to the images, which is carried out using the control arrangement.
• a classification of any image information based on a machine learning model can generally be computationally intensive. In the present case, however, it was recognized that with the labeling a recording of the images in a - 3 - is possible in a largely controlled environment, which significantly reduces the demands on the computing power in the analysis routine, even when using a machine learning model, and also allows targeted training of the machine learning model.
• the application of a machine learning model can even enable classification with high accuracy in real time, so that high process speeds can be achieved even with a chaotic product feed.
• the preferred refinements according to claims 2 and 3 relate to printing of the labels depending on the classification derived in the analysis routine. Weight-dependent pricing of the respective packs on the basis of a basic price assigned to the pack class is particularly preferred.
• the trained machine learning model is based on a trained neural network, such as a neural convolutional network.
• Neural convolution networks achieve particularly good results in image processing.
• a feature extractor can be used for the classification, with the classification being performed on the basis of the generated feature space. It is particularly advantageous here if at least one of these steps and preferably both steps are implemented by applying the trained machine learning model.
• a suggestion step is also provided in the analysis routine, with which suggested regions in the image are identified which - 4 - again be used in the classification step.
• This configuration can lead to a simplification of the analysis routine, particularly in the case of packs in which individual products are at least partially visible, for example packs with a transparent cover.
• Claims 10 and 11 relate to a learning routine based on a training data set.
• the training data record is at least partially derived from images of a preceding and/or ongoing labeling routine.
• a labeling routine with no chaotic pack feed can be used to build a large and relevant training data set. If respective packs of the same pack class are labeled here at least for a certain period of time, the annotation of the images for the training data record can be significantly simplified.
• a predefined distance can also be provided between the sensor arrangement and the respective packs, so that, for example, the effort associated with scaling the images in the analysis routine is reduced.
• control arrangement can implement central management of the packaging classes and/or central implementation of the analysis routine, for example in a cloud-based manner.
• a labeling system with at least one labeling device for labeling individual packs is claimed as such.
• the label - 5 - chaining system is set up in particular for carrying out the proposed method. Reference is made to all statements on the proposed method.
• 3 shows a schematic representation of the learning routine for the proposed method.
• the invention relates to a method for operating a labeling system 1 with at least one labeling device 2 for labeling individual packs 3.
• FIG. 1 shows a schematic representation of the labeling device 2 in a preferred embodiment as a price marking device.
• the labeling device 2 has at least one feed arrangement 4, a label dispensing arrangement 5, a label application arrangement 6 and a printer arrangement 7 as functional units, which are set up to carry out a labeling routine for the packs 3.
• further functional units of the labeling device 2 can also be provided.
• the functional units are controlled by a control arrangement 8 of the labeling system 1 in a labeling routine that includes the labeling of the individual packs 3 . 6
• the labeling routine provides that the respective packs 3 are transported by means of the feed arrangement 4, labels that can be removed from a material strip 9 are dispensed by means of the label dispensing arrangement 5, the dispensed label is applied to the respective pack 3 by means of the label application arrangement 6 and the from the strip of material 9 detachable or detached label is printed.
• the feed arrangement 4 is designed to transport respective packs.
• the feed arrangement 4 is preferably a belt conveyor or a roller conveyor, optionally also at least one robot arm, for moving the respective packs 3.
• the feed arrangement 4, here the belt conveyor has here and preferably at least one conveyor belt, via which the respective packs 3 are transported along a transport direction.
• the label dispensing arrangement 5 is also set up to dispense the label.
• the label is preferably detached from a strip of material 9 by means of the label dispensing arrangement 5 .
• a label that can be detached from a strip of material 9 means, in particular, a label whose adhesive surface is detachably attached to a carrier strip that forms the strip of material 9 and can be made of paper and/or plastic, for example. It is also possible for the label to be produced by separating a section from a printable or printed material strip 9, for example by cutting and/or tearing the material strip 9.
• labels designed as adhesive labels are used, which are already attached to the material strip 9 have an adhesive surface. The strip of material 9 is guided here over a dispensing edge 10, with which the labels are detached.
• the use of adhesive-free labels is also conceivable, which are only later provided with an adhesive surface or applied to an adhesive surface on the respective pack 3 .
• the labeling device 2 here in a common housing with the label dispensing arrangement 5, has the label applying arrangement 6 for applying the dispensed label to the respective pack 3.
• the label applicator assembly 6 for applying a label to the top of the package 3 here and preferably - 7 - have a stamp 11 on.
• the stamp 11 transfers the label to the surface of the pack 3 in one application movement.
• the stamp 11 is here and preferably designed as a pendulum stamp, which can be moved both linearly and pivoted.
• the stamp 11 has a suction foot as the stamp foot, preferably a suction and blowing foot, for sucking in and in particular also blowing off the label.
• the stamp 11 designed here as a pendulum stamp, carries out an application movement along the transport direction when the label is being transferred, in order to enable labeling of the pack 3 moved by means of the feed arrangement 4 . It is preferred here that the plunger 11 can also be adjusted in a direction orthogonal to the transport direction in order to enable the labels to be applied to different positions of the packs 3 orthogonal to the transport direction.
• the label can be applied in a touching manner, ie mechanically, by pressing the label onto the pack 3.
• the label can be applied without contact, for example by a suction and blowing foot of stamp 11 blowing the label off onto pack 3 by generating a compressed air blast directed towards pack 3, ie applying it pneumatically.
• the plunger 11 can also be a purely linear plunger, which can then only be moved linearly, possibly in a plurality of mutually orthogonal directions.
• a label suction arrangement 12 which transfers the detached labels to the stamp n.
• the label it is basically also conceivable for the label to be transferred directly onto the surface of the pack 3, in particular by means of a compressed air blast exerted on the label by the label suction arrangement 12, preferably by a blowing head. In this case, no stamp 11 is necessary to transfer the label.
• the printer assembly 7 is provided for printing the label, with a printing of the label in principle on the strip of material 9, after 8 detachment of the label from the strip of material 9, as well as before and/or after the label is applied to the respective pack 3.
• a printer arrangement 7 set up for thermal printing is provided.
• the printer arrangement 7 can also have a laser printer and/or an inkjet printer.
• the printer arrangement 7 is preferably integrated into the label dispensing arrangement 5, as shown, and prints the labels before, after and/or during dispensing.
• the control arrangement 8 takes care of the technical control tasks occurring in the labeling routine.
• the control arrangement 8 preferably has at least one computer device which is set up to control the functional units. 1 shows a local control unit 13 of the labeling device 2, which communicates with a cloud-based server 14 via a wired and/or wireless network, for example a local network, a mobile network and/or the Internet. A mobile device 15 is also provided, which also communicates with the other components of the control arrangement 8 via the network.
• Other variants of the control arrangement 8 are conceivable. For example, as an alternative to the illustrated control arrangement 8 with several components, only one local control arrangement 8 can be provided on the labeling device 2 .
• the labeling device 2 also has a sensor arrangement 16, which is preferably designed as an optical sensor arrangement and here and preferably as a camera. Images 17 of the respective packs 3 are recorded by means of the sensor arrangement 16 .
• the images 17 are therefore preferably camera images, in particular two-dimensional or three-dimensional image information of the respective pack 3.
• the camera can be designed as a color camera and in particular as a 3D camera.
• Further configurations of the sensor arrangement 16 are conceivable, for example with IR sensors or the like.
• the sensor arrangement 16 is arranged here and preferably on the feed arrangement 4, so that images 17 of the respective pack 3 on the feed arrangement 4, preferably when the pack 3 is moving, are recorded.
• Other configurations of the sensor arrangement 16 are also conceivable, which can record images 17 that are representative of the appearance of the packs 3, for example by means of laser scanning or the like.
• the proposed method focuses on a chaotic pack supply, packs 3 with different requirements being processed in the labeling routine.
• the images 17 of the respective packs 3 are analyzed in an analysis routine by means of the control arrangement 8 .
• a classification of the respective pack 3 into a pack class is derived by means of this analysis.
• a plurality of pack classes can be specified and stored in the control arrangement 8 .
• the respective package 3 is classified into at least one of these predefined package classes.
• the packaging classes can be assigned respective metadata, for example a product designation, an identification number and specifications relating to the labeling routine or the like.
• the labeling device 2 is controlled in the labeling routine depending on the classification. Accordingly, at least one aspect of the labeling routine can be carried out in different ways for packs 3 from different pack classes, and can be added and/or omitted. Preferably, the activation of the labeling device 2, which is dependent on the classification, is carried out by means of the control arrangement 8 without the intervention of an operator and is therefore carried out automatically.
• the analysis routine is based on an application of a trained machine learning model to the images, which is carried out by means of the control arrangement 8 . Accordingly, a model generated on the basis of a machine learning method is used, which is trained to classify the images 17 into one of the predefined pack classes.
• the label is particularly preferably printed by means of the printer arrangement 7 depending on the pack class of the respective packs.
• the printing is preferably carried out as a function of product information assigned to the packaging class.
• the product information can be general product-related information such as a product designation, a printed image specified for the packaging class or the like. 10 chen included. More preferably, in the design of the labeling system 1 for price marking of the packs 3, the printing is carried out using assigned price information, which is printed on the label in particular as a numerical value.
• a weighing arrangement 18 is also provided here as a functional unit, by means of which weight values for the individual packs 3 are determined.
• the printing of the label by means of the printer arrangement 7 can also be carried out here depending on the weight values of the respective packs 3 .
• the weight value or values that can be determined from this using the packaging class for example net weight, gross weight, tare and/or a weight range assigned to the weight, are printed on.
• the price information assigned to the packaging class preferably contains a basic price, which is used to calculate a weight-dependent packaging price, the respective label being printed with the packaging price determined from the weight value and basic price value by means of the printer arrangement 7 .
• the weighing arrangement 18 can be operated depending on the respective package class. For example, weighing parameters, such as weight ranges and/or division values, are assigned to the pack class, and the weighing arrangement 18 determines the weight values on the basis of the assigned weighing parameters.
• the label dispensing arrangement 5 is here and preferably equipped with a plurality of material strips 9 for dispensing different types of labels.
• a label type and/or one of the material strips 9 can be assigned to the packaging class.
• the label is dispensed by means of the label dispensing arrangement 5 in accordance with the label type associated with the pack class. Consequently, the label with the label type specific to the pack class is applied to the pack 3 via the label application arrangement 6 .
• the label application arrangement 6 can be used to apply the dispensed label to the respective pack 3 in accordance with an application specification assigned to the pack class.
• the application specification preferably indicates whether the label is applied without contact or pressed on, in particular pressed on with a specified pressure.
• the packing class can be a - 11
• the respective pack 3 is transported by means of the feed arrangement 4 according to a speed assigned to the pack class.
• the speed of the respective pack is adapted in particular to the transport path from the sensor arrangement 16 to the label application arrangement 6 and/or printer arrangement 7 .
• a sorting arrangement can also be provided as a functional unit, by means of which the individual packs 3 are sorted on the feed arrangement depending on the classification.
• the sorting can involve sorting out individual packs 3, for example removing them from the feed arrangement 4, which is effected, for example, by means of a blast of compressed air.
• a multi-path sorting arrangement can also be used, which distributes the packs 3 to different sorting paths, for example via one or more switches.
• the trained machine learning model used in the analysis routine is based on a trained neural network.
• the neural network can be a neural convolutional network.
• Neural convolution networks are known under the term "convolutional neural network" and in many cases allow a particularly effective image evaluation.
• the analysis routine is based on an image 17 recorded by means of the sensor arrangement 16, with which a pack 3 is detected. Images 17 of individual packs 3 on the feed arrangement 4 are preferably recorded by means of the sensor arrangement 16 so that overlapping of different packs 3 on the image 17 is avoided.
• a feature extractor 19 is applied directly or indirectly to the respective image to generate a feature space 20 .
• the feature space 20 known as "feature space” 12 shown with only one level, but preferably includes multiple levels.
• the respective pack 3 is classified into a pack class based on the feature space 20 .
• the pack 3 is preferably classified on the basis of an assignment based on the feature space 20 to a pack class belonging to a selection of predetermined pack classes (A, B . X). 2 shows, by way of example, that pack 3 is assigned a pack class B on the basis of image 17, which in turn is assigned the metadata (a, b, . . . x) already mentioned.
• the feature extractor 19 and/or the classification step 21 are based on the trained machine learning model, preferably on a trained neural network. Only a trained neural network 22 for classification is shown here. In particular, however, the feature extractor 19 and the classification step 21 can also be based jointly on the same trained neural network.
• suggestion regions 24 are identified in the image 17, which potentially contain sections of the pack 3.
• the suggested regions 24 can be used here in particular to identify regions in the image 17 which have individual products contained in the pack or also other subsections such as borders, existing labels or the like.
• the suggestion step 23 is preferably carried out using the trained machine learning model. Algorithms suitable for this are known under the term "region proposal”.
• the suggestion regions 24 are analyzed here for classification.
• the classification step 21 it is thus possible, among other things, to apply more complex calculations to targeted sections of the image.
• An example of a suitable algorithm for implementing the suggestion step 22 and the classification step 20 is R-CNN.
• the machine learning model is trained as part of the proposed method.
• the machine learning model is correspondingly trained in a learning routine on a training data set 25 by means of the control arrangement 8, which is shown in FIG.
• the machine learning model is trained here and preferably based on annotated images 26 of packs 3 .
• the trained machine learning model contains at least one weighting, preferably a set of parameters 28 that is representative of the weighting, as well as an application specification for how the weighting is to be implemented in the analysis routine.
• a parameter set 28 for the machine learning model is determined here and preferably via a neural network.
• the training data set 25 is at least partially derived from images 17 recorded by means of the sensor arrangement 16 in a preceding and/or ongoing labeling routine.
• this labeling routine a classification of the respective packs 3 can be specified, via which the images 17 are annotated.
• respective packs 3 of the same pack class are labeled at least for a certain period of time.
• the images 17 can also be annotated in the event of a chaotic supply of packs, for example if pack recognition, as mentioned in the introduction, is possible on the basis of the weight values or the like.
• the automatic annotation of the images 17 is preferably validated via a further classification, preferably via the weight values of the packs 3. In this case, only those images 17 are automatically annotated whose weight values fall into a weight class assigned to the packaging class. - 14 -
• an alignment arrangement 29 is also provided as a functional unit, by means of which the individual packs 3 are positioned on the feed arrangement 4 .
• the alignment arrangement 29 has at least one guide element 30, in this case guide elements 30 arranged on both sides of the feed arrangement 4, which at least partially adjoin and/or protrude into a conveying region of the feed arrangement 4.
• the guide elements 30 align the packs 3 by touching them.
• the guide elements 30 are preferably movable in order to allow adjustment of the alignment.
• the sensor arrangement 16 can be provided downstream of the alignment arrangement 29 and/or the alignment arrangement 29 on the feed arrangement 4 . Consequently, the images 17 can be determined on aligned packages, simplifying the analysis routine. However, it is also conceivable that the alignment by means of the alignment arrangement 29 is carried out depending on the classification of the packs 3 , with the alignment arrangement 29 being arranged downstream of the sensor arrangement 16 .
• the sensor arrangement 16 preferably has a predefined distance from the respective packs 3, which also leads to a simplification of the analysis routine.
• the sensor arrangement 16 is kept at a constant distance from the feed arrangement 4 and thus at a predefined distance from the underside of the packs 3 .
• the predefined distance preferably corresponds to the distance between the sensor arrangement 16 and the respective packs 3 in the labeling routine used for the training data set.
• an architecture of the machine learning model and/or the training step takes into account the predefined distance and/or the orientation.
• many scalings in the architecture and/or training of a machine learning model are taken into account, since it is not clear in which size and in which position the objects will ultimately appear in image 17.
• This problem has led to various solution approaches, which, however, are generally associated with increased computational complexity. In the present case, it is possible to save at least part of this complexity.
• control arrangement 8 can be designed as part of the labeling device 2 and/or cloud-based.
• the labeling system 1 preferably has a plurality of labeling devices 2 which are controlled by the control arrangement 8 .
• a cloud-based implementation enables the packaging classes and/or the trained machine learning model to be managed centrally.
• the training data set 25 can also be generated on the basis of the images 17 of one or more of the labeling devices 2 . It is also conceivable that at least part of the computationally intensive analysis routine is carried out in a cloud-based manner, here via the cloud server 14 .
• the labeling system 1 mentioned is claimed as such.
• the labeling system 1 is equipped with at least one labeling device 2 for labeling individual packs 3, in particular for pricing.
• the labeling device 2 has at least one feed arrangement 4, a label dispensing arrangement 5, a label application arrangement 6 and a printer arrangement 7 as functional units, with a control arrangement 8 of the labeling system 1 controlling the functional units in a labeling routine.
• the feed arrangement 4 transports the respective packs 3
• the label dispensing arrangement 5 donates labels that can be detached from a strip of material 9
• the label application arrangement 6 applies the dispensed label to the respective pack 3
• the printer arrangement 7 applies the label that can be detached from the strip of material 9 or detached label printed.
• the labeling device 2 has a sensor arrangement 16, preferably a camera, which records images 17 of the respective packs 3, with the control arrangement 8 analyzing the images 17 of the respective packs 3 in an analysis routine, using this analysis to classify the respective pack 3 into a pack class derives and carries out the activation of the labeling device 2 in the labeling routine depending on the classification. It is essential here that a trained machine learning model is stored in the control arrangement 8 and that the analysis routine is based on an application of the trained machine - 16 - learning model based on the pictures 17. Reference is made to the statements on the proposed method.
• a data carrier with a training data set 25 is claimed as such.
• the data carrier is intended for use in the proposed method and is generated by means of the learning routine mentioned.
• the training data record 25 is preferably stored on the data medium in a non-volatile manner. Reference is made to the statements on the proposed method.

Landscapes

• Engineering & Computer Science (AREA)
• Theoretical Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Health & Medical Sciences (AREA)
• Computing Systems (AREA)
• Biomedical Technology (AREA)
• Biophysics (AREA)
• Computational Linguistics (AREA)
• Data Mining & Analysis (AREA)
• Evolutionary Computation (AREA)
• Life Sciences & Earth Sciences (AREA)
• Molecular Biology (AREA)
• Artificial Intelligence (AREA)
• General Engineering & Computer Science (AREA)
• General Physics & Mathematics (AREA)
• Mathematical Physics (AREA)
• Software Systems (AREA)
• Health & Medical Sciences (AREA)
• Labeling Devices (AREA)
• Adhesives Or Adhesive Processes (AREA)
• Image Analysis (AREA)

Applications Claiming Priority (2)

Publications (1)

Family

ID=82019294

Family Applications (1)

Country Status (6)

Family Cites Families (10)

• 2021
  • 2021-05-12 DE DE102021112479.4A patent/DE102021112479A1/de active Pending
• 2022
  • 2022-05-11 AU AU2022271962A patent/AU2022271962A1/en active Pending
  • 2022-05-11 EP EP22729457.6A patent/EP4337546A1/de active Pending
  • 2022-05-11 CN CN202280034544.2A patent/CN117355465A/zh active Pending
  • 2022-05-11 WO PCT/EP2022/062810 patent/WO2022238494A1/de active Application Filing
  • 2022-05-11 CA CA3216918A patent/CA3216918A1/en active Pending

Also Published As

Similar Documents

Legal Events