Bale dewiring system, method for operating a bale dewiring system, computer program, and computer-readable storage medium

Abstract

The invention relates to a bale dewiring system (1) for dewiring bales (4; 5-7; 20-22) which have been wrapped with straps (2) in the manner of a cross strapping (3a) or an individual strapping (3b; 3c), comprising: a transport device (5), a supply shaft (9), a processing device (10), an image capturing device (11; 111), an artificial intelligence unit (13) and a control device (14), said image capturing device (11; 111) being designed to generate raw image data (RD) of a dynamic scenery (DS) of an observation portion (BA) of the transport device (5) on which the bales (4; 5-7; 20-22) are supplied.

Description

[0001] "Bale de-wiring system, method for operating a

[0002] Bale de-wiring system, computer program and computer-readable

[0003] Storage medium

[0004] The invention relates to a bale de-wiring system, a method for operating a bale de-wiring system, a computer program and a computer-readable storage medium.

[0005] From WO 2023 / 062054 Al, a device for cutting and removing a banding, in particular a banding wire, from a rectangular bale is known, which has a clamping and cutting device for clamping and cutting the banding, wherein the clamping and cutting device has an outer cylinder and an inner cylinder, wherein the outer cylinder and the inner cylinder each have at least one receiving groove for receiving the banding, wherein a rotary drive is provided for rotating the inner cylinder relative to the outer cylinder, wherein a clamping and cutting gap is formed between the outer cylinder and the inner cylinder for clamping and cutting the banding by rotating the inner cylinder relative to the outer cylinder via the rotary drive.

[0006] The object of the invention is to propose a bale de-wiring system, a method for operating a bale de-wiring system, a computer program and a computer-readable storage medium, by which an increasingly reliable, efficient and plant-friendly de-wiring process can be ensured.

[0007] This problem is solved by a bale de-wiring system, a method for operating a bale de-wiring system, a computer program, and a computer-readable storage medium according to claims 1, 5, 13, and 14. Advantageous embodiments, implementations, and further developments are the subject of the respective dependent claims. The bale de-wiring system according to the invention for de-wiring bales wrapped with bands in the manner of cross-wrapping or single-wrapping comprises: a transport device, a feed chute, a processing device, an image acquisition device, an artificial intelligence unit, and a control device, wherein the image acquisition device is configured to generate raw image data (RD) of a dynamic scene (DS) of an observation section of the transport device onto which the bales are fed.wherein the artificial intelligence unit comprises a classification model trained to process the raw image data (RD) such that, for example, it filters the data so that the path of the belt or belts is detected as parallel and / or transverse with respect to a transport direction, and wherein the control device is trained to activate a first cutting device and a second cutting device of the processing device, as well as a first removal device and a second removal device of the processing device, for processing the detected bale only if at least one belt with a parallel path, but no belt with a transverse path, is present in their respective working area. Such a bale de-wiring system makes it possible to operate the cutting devices and the removal devices according to the requirements, so that they are only activated when...if necessary. This protects the equipment. Furthermore, time can be saved when operating identical equipment sequentially, provided that only the equipment whose operation is required is running. By integrating a classification model, the operation of the bale de-wiring system becomes increasingly reliable with regard to the detection of the straps, thus also resulting in increased efficiency – downtime due to errors is avoided – and protection of the equipment – ​​it is increasingly impossible for bales that are still wrapped to leave the processing unit and cause malfunctions.

[0008] Furthermore, the bale de-wiring system can be equipped with a trigger device that activates the image acquisition device as soon as one of the bales being fed onto the conveyor in the direction of transport reaches a position suitable for detection by the image acquisition device. This protects the image acquisition alignment and the downstream electronic components, as continuous operation is avoided.

[0009] It may also be possible to design the classification model to extract a bale as a feature from the raw image data and / or to extract at least one area of ​​the bale as a feature from the raw image data and / or to extract the presence of at least one band on the extracted area or areas as a further feature from the raw image data.

[0010] This can increase the reliability in detecting the course of the existing belt(s), as it is also possible to check whether a bale is detected at all and / or whether at least an area of ​​a bale is detected and / or whether a belt or belts are detected at all.

[0011] Furthermore, the classification model can be implemented as a so-called machine learning model, specifically as an artificial neural network, and specifically as a deep learning model, and specifically as a convolutional neural network (CNN). A CNN is particularly suitable for 2D image data because it convolves learned features with input data and uses convolutional convolution (CD) layers. Furthermore, features can be directly extracted from images using a CNN. Thus, the relevant features can be learned when training with a collection of images.

[0012] The inventive method provides for a bale de-wiring system according to claims 1 and / or 2, in which raw image data (RD) of a dynamic scene (DS) of an observation section of the transport device on which the bales are fed is generated by means of the image acquisition device, wherein the raw image data (RD) are filtered by the classification model of the artificial intelligence unit, and the presence of at least one belt and its course with respect to the transport direction are extracted as features from the filtered raw image data (RD), wherein, depending on the result of the extraction, after the evaluated bale has been fed through the feed chute into the processing device, the control device activates a first cutting device and a second cutting device as well as a first removal device and a second removal device for processing the analyzed bale only ifif at least one belt is present in their respective work area.

[0013] This method protects the equipment. Furthermore, sequential operation of identical equipment can potentially save time, provided only the equipment required is in operation. Integrating a classification model makes the bale de-wiring system increasingly reliable in terms of strap detection, thus also increasing efficiency – downtime due to errors is avoided – and protecting the equipment – ​​it is increasingly unlikely that bales still wrapped will leave the processing unit and cause malfunctions.

[0014] It may also be possible to extract a frontal surface of the ball using the classification model as a surface.

[0015] This allows for a targeted limitation of the evaluation area of ​​a captured digital image.

[0016] Furthermore, it may be possible to extract a frontal surface and a side surface of the bale from the classification model. This allows for a more targeted narrowing of the evaluation area of ​​a captured digital image and, by focusing on two surfaces, increases the probability of the evaluation being correct.

[0017] It may also be possible to deactivate the processing unit and / or activate a reject device located upstream of the processing unit if a bale with an impermissible overwrap is detected. Naturally, such a procedure also includes issuing a warning message, after which the bale in question is manually removed from the transport unit.

[0018] Furthermore, it may be possible to classify a bale with a cross-hatched pattern formed by bands on one of its side surfaces and / or at least one band running transversely to the transport device on one of its side surfaces as an impermissible banding pattern for the control device. Such monitoring allows bales with impermissible banding to be reliably identified solely by evaluating one side surface.

[0019] Naturally, this requires that the front of the bale is aligned both transversely to the transport direction of the bale in the transport device and transversely to a conveying direction in the processing device.

[0020] Furthermore, it may be possible to carry out the procedure additionally with the following procedural steps:

[0021] Classification of the bale by the image evaluation unit into one of four classes; in the case of classification of the bale into class I, the processing unit is controlled after the bale has entered such that the strips are cut by a first cutting device of the processing unit and that the strips are removed by a first removal device of the processing unit; in the case of classification of the bale into class II, the processing unit is controlled after the bale has entered such that the strips are cut by a second cutting device of the processing unit and that the strips are removed by a second removal device of the processing unit;In the case of classification of the bale in class III, the processing device is controlled after the bale has entered the processing device in such a way that the strips are cut successively or simultaneously by the first cutting device and by the second cutting device of the processing device, and that the strips are removed successively or simultaneously by means of the first removal device of the processing device and by means of the second removal device of the processing device;If the bale is classified as Class IV, neither of the two cutting devices nor either of the two removal devices of the processing unit are activated. Such classification of the bales also allows for the easy derivation of subsequent statistics for the operating times of a first working unit (consisting of the first cutting device and the first removal device) and a second working unit (consisting of the second cutting device and the second removal device). This enables maintenance of the individual working units to be carried out as needed.

[0022] It may also be possible to extract a bale as a feature from the filtered raw image data and / or to extract at least one area of ​​the bale as a feature from the filtered raw image data. This can increase the reliability in recognizing the path of the existing band(s), as it allows for additional verification of whether a bale is detected at all and / or whether at least one area of ​​a bale is detected.

[0023] Furthermore, it may be possible to implement the classification model using a machine learning model, specifically by implementing the machine learning model as an artificial neural network, and in particular by implementing the artificial neural network as a so-called deep learning model, and in particular by implementing the deep learning model as a convolutional neural network (CNN). A CNN is particularly suitable for 2D image data because it convolves learned features with input data and uses 2D convolutional layers. Furthermore, features can be directly extracted from images using the CNN. Thus, the relevant features can be learned when training with a collection of images.

[0024] The computer program according to the invention comprises instructions which, when the computer program is executed by a computer, cause it to carry out the method according to one of claims 5 to 12.

[0025] The computer program according to claim 13 is stored on the computer-readable storage medium according to the invention.

[0026] For the purposes of the invention, a single strapping is understood to be a vertical or horizontal strapping of a bale, which is formed from one band as a single strapping or from several bands as a multiple strapping of a bale.

[0027] According to the invention, a bale comprises a multitude of recyclable waste materials compressed into a cuboid and additionally held together by at least one band, and in particular several bands, which encircle the cuboid. These bands can be made of metal or plastic. A bale de-wiring system is provided for cutting and removing such bands.

[0028] Further details of the invention are described in the drawing with reference to a schematically illustrated exemplary embodiment.

[0029] This shows:

[0030] Figure 1: a schematic view of a

[0031] Bale de-wiring system;

[0032] Figure 2: a view corresponding to Figure 1, showing three bales on the transport device, each with an impermissible strapping;

[0033] Figures 3a-3c: schematic views of the bales shown in Figure 1 on the transport device, as they are captured by the image capture device with their rear end face and their upper side face; Figures 4a-4c: schematic views of the bales shown in Figure 1 on the transport device, as they are captured by the optionally alternatively positioned image capture device with their rear end face.

[0034] Figure 1 shows a schematic view of a bale de-strawing machine 1 for de-strawing bales 4 strapped with bands 2 in the form of cross-strapping 3a, first single strapping 3b, or second single strapping 3c (vertical or horizontal strapping). The bands 2 are not all labeled, but only as examples. A first bale 6, a second bale 7, and a third bale 8 are shown on a transport device 5 of the bale de-strawing machine 1. The first bale 6 is shown in a first position 6-1 on the transport device 5 and additionally in a second position 6-2, which is its position after it has fallen through a feed chute 9 into a processing device 10.

[0035] The bales 4 are designed as cuboid blocks made of compressed recyclable waste, which are strapped by the bands 2.

[0036] The first bale 6 is strapped according to the cross-strapping method 3a by three straps 2, namely 2a, 2b, 2c of a first strapping group A and four further straps 2, namely 2d, 2e, 2f, 2g of a second strapping group B. With respect to a transport direction TR of the transport device 5, the three bales 6, 7 and 8 each have a front end face 6a, 7a, 8a oriented transversely to the transport direction TR and each have a rear end face 6b, 7b, 8b oriented transversely to the transport direction TR. Furthermore, the three bales 6, 7 and 8 each have four side faces 6c-6f, 7c-7f and 8c-8f oriented parallel to the transport direction TR.

[0037] During the continuous operation of the transport device 5, the first bale 6 tips over an edge K, so that it falls with its front end face 6a forward into the feed shaft 9 in a falling direction FR and enters the processing device 10 .

[0038] The second bale 7 is labelled according to the first single labelling 3b and the third bale 8 is labelled according to the second single labelling 3c.

[0039] The bale de-wiring system 1 further comprises an image acquisition device 11, which is designed as a digital camera 12, an artificial intelligence unit 13, and a control device 14. The image acquisition device 11 is configured to generate raw image data RD of a dynamic scenario DS of an observation section BA of the transport device 5, onto which the bales 4 are fed. The image acquisition device 11 can be activated by a trigger device TE whenever a bale 4 is again in the observation section BA.

[0040] The artificial intelligence unit 13 comprises a classification model 15 from which the raw image data RD are filtered to extract, as a first feature of the bale 8, and, as a second feature, at least one surface, and in particular the rear end face 8b and the upper side face 8c of the bale 8, and, as a further feature, the presence of at least one band 2 on the extracted surfaces 8b and 8c and the course of the band 2 or bands with respect to a transport direction TR. Alternatively, the classification model from which the raw image data RD are filtered can also be limited to extracting, as features, the presence of at least one band 2 and the course of the band 2 or bands with respect to a transport direction TR.

[0041] The control device 14 is designed to activate a first cutting device 16 and a second cutting device 17 of the processing device 10 as well as a first removal device 18 and a second removal device 19 of the processing device 10 for processing the detected bale 6 only if at least one belt 2 is present in their respective working area, depending on the extracted features.

[0042] Figure 1 shows the case in which both cutting devices 16, 17 and both removal devices 18, 19 are activated by the control device 14 for the bale 6, which is located in the processing device 10.

[0043] The first cutting device 16 cuts the strips 2a-2c on the first side surface 6c and, after cutting on the third side surface 6e, removes them from the first removal device 18, which can be done by pulling them off and / or winding them up.

[0044] The second cutting device 17 cuts the strips 2d-2g on the fourth side surface 6f and, after cutting on the second side surface 6d, removes them from the second removal device 19, which can be done by peeling and / or winding.

[0045] As an alternative to being mounted above the transport device 5, the image acquisition device 11 of the bale de-strawing system 1 can also be positioned above the feed chute 9. In this alternative position, the image acquisition device is designated with the reference symbol 111. With such a positioning, only the rear end face 6b of the bale 6 is detected. The control unit 14 controls the process according to the detected path of the belts 2.

[0046] - either the first cutting device 16 and the first removal device 18 on

[0047] - or the second cutting device 17 and the second removal device 19 on

[0048] - or both cutting devices 16, 17 and both removal devices 18, 19 on .

[0049] In the situation shown in Figure 1, the control unit 14 would control according to the third variant.

[0050] In the described alternative positioning of the image acquisition device 111, the trigger device, if present, is positioned in the area of ​​the feed chute 9.

[0051] The bales 6, 7, and 8 shown in Figure 1 on the transport device 5 are each encircled by bands 2 such that, after extracting their features, the control device 14 will permit their entry into the feed chute 9, since all of them have only bands 2 running in the transport direction TR on their side surfaces 6c-6f, 7c-7f, and 8c-8f. Based on the features extracted for the rear end faces 6b, 7b, and 8b, the control device 14 then controls the extent to which the processing device 10 is activated for each individual bale 6, 7, 8.

[0052] In contrast, Figure 2, which shows the bale de-strawing system 1 in a further simplified representation in the schematic view known from Figure 1, depicts a fourth bale 20, a fifth bale 21, and a sixth bale 22 moving in the transport direction TR on the transport device 5. The bales 20, 21, and 22 are encircled by belts 2 in such a way that the control device 14 (see Figure 1), after extracting their features, will not allow feeding into the feed chute 9, but will instead activate an optional rejection device 23, shown only in Figure 2. The rejection device 23 is designed as a roller conveyor, which can cover the feed chute 9 in such a way that, for example, the fourth bale 20 is guided onto a discharge device 24. The discharge device 23 and the removal device 24 can be components of the bale de-wiring system 1.

[0053] Bales 20, 21, and 22 all have bands 2 on all their side surfaces 20c-20f, 21c-21f, and 22c-22f, which run transversely to the transport direction TR. The cutting devices 16, 17 and the removal devices 18, 19 are not designed for cutting and removing such oriented bands 2. Accordingly, bales 20, 21, and 22 must be ejected, as already explained above.

[0054] In principle, it may also be intended that such an orientation of bales 20, 21 and 22 is avoided by a preceding process.

[0055] Instead of a discharge, which can also be done manually, the bale de-wiring system 1 can of course also include an automatic reorientation device 25 (see Figure 2), which is controlled by the control device 14 such that the bale in question is reoriented so that, after reorientation, it has bands 2 on all its side surfaces oriented only in the transport direction TR. The discharge device 23 and the discharge device 24 and / or the reorientation device 25 can also be included in the bale de-wiring system 1.

[0056] Naturally, a bale with intersecting bands on its end faces and all side faces must be manually or automatically rejected, as such a bale is unsuitable for the processing device even after reorientation. Figures 3a to 3c show schematic views of bales 6 (Figure 3a), 7 (Figure 3b), and 8 (Figure 3c) depicted on the transport device in Figure 1, showing how they are captured by the first image capture device 11 (see Figure 1) with their rear end faces 6b (Figure 3a), 7b (Figure 3b), and 8b (Figure 3c) and their upper side faces 6c (Figure 3a), 7c (Figure 3b), and 8c (Figure 3c).

[0057] Figures 4a to 4c show schematic views of the bales 6 (Figure 4a), 7 (Figure 4b) and 8 (Figure 4c) shown on the transport device in Figure 1, as they are captured by the alternatively positioned image capture device 111 (see Figure 1) with their rear end faces 6b, 7b, 8b.

[0058] Basically, the features extracted by the classification model 15 can be classified into one of four classes: I, II, III or IV.

[0059] For example, bale 8 is assigned to class I and, after bale 8 has entered, the processing device is controlled such that the bands 2 are cut by the first cutting device 16 of the processing device 10 and that the bands 2 are removed by the first removal device 18 of the processing device 10.

[0060] For example, bale 7 is assigned to class II and the processing device is controlled after bale 7 has entered such that the bands 2 are cut by the second cutting device 17 of the processing device 10 and that the bands 2 are removed by the second removal device 19 of the processing device 10.

[0061] For example, bale 6 is assigned to class III and the processing device 10 is controlled after bale 6 has entered such that the straps 2 are cut simultaneously by the first cutting device 16 and by the second cutting device 17 of the processing device 10, and that the straps 2 are first removed by the first removal device 18 of the processing device 10, and that the straps 2 are then removed by the second removal device 19 of the processing device 10.

[0062] For example, bale 20 is assigned to class IV. Accordingly, neither of the two cutting devices 16, 17 of the processing device 10 and neither of the two removal devices 18, 19 of the processing device 10 are activated, and bale 20 is ejected. Provided the

[0063] If bale 20 is turned over, it will also be assigned to a different class.

[0064] Reference character list

[0065] 1 bale de-wiring system

[0066] Volume 2

[0067] 2a-2c Band of A

[0068] 2d-2g band of B

[0069] 3a Cross-over

[0070] 3b first individual classification

[0071] 3c second single classification

[0072] 4 bales

[0073] 5 Transport equipment

[0074] 6 first bale

[0075] 6a front face of 6

[0076] 6b rear frontal surface of 6

[0077] 6c- 6f Side faces of 6

[0078] 6-1 first position of 6

[0079] 6-2 second position of 6

[0080] 7 second bales

[0081] 7a front face of 7

[0082] 7b rear frontal surface of 7

[0083] 7c-7 f Side faces of 7

[0084] 8 third bales

[0085] 8a front face of 8

[0086] 8b rear frontal surface of 8

[0087] 8c- 8f Side faces of 8

[0088] 9 Supply shaft

[0089] 10 Processing equipment

[0090] 11 Image capture device

[0091] 12 Digital camera

[0092] 13 artificial intelligence units

[0093] 14 Control device

[0094] 15 Classification model

[0095] 16 first cutting device

[0096] 17 second cutting device

[0097] 18 first remote device

[0098] 19 second removal device 20 fourth bale

[0099] 20c-20f Side surfaces of 20

[0100] 21 fifth bale

[0101] 21c-21f Side faces of 21

[0102] 22 sixth bale

[0103] 22c-22f Side faces of 22

[0104] 23 Discharge device

[0105] 24 Discharge device

[0106] 25 Reorientation facility

[0107] 111 alternative position of the image capture device 11

[0108] A first classification group from 2a-2c

[0109] B second classification group from 2d-2 f

[0110] BA Observation Section

[0111] DS dynamic scene

[0112] FR Fall direction from 4

[0113] K edge of 5

[0114] RD Raw Data

[0115] TE trigger device

[0116] TR Transport direction from 5

Claims

Claims 1. Bale de-wiring system (1) for de-wiring bales (4; 5-7; 20-22) strapped with bands (2) in the manner of cross-strapping (3a) or single-strapping (3b; 3c), comprising: a transport device (5), a feed chute (9), a processing device (10), an image acquisition device (11; 111), an artificial intelligence unit (13), and a control device (14), wherein the image acquisition device (11; 111) is configured to generate raw image data (RD) of a dynamic scene (DS) of an observation section (BA) of the transport device (5) on which the bales (4; 5-7; 20-22) are fed, wherein the artificial intelligence unit (13) comprises a classification model (15) configured to To process raw image data (RD) in such a way, for example, to filter the course of the belt (2) or belts (2) in relation to a transport direction (TR) is recorded as a parallel course and / or as a transverse course and - wherein the control device (14) is designed to activate a first cutting device (16) and a second cutting device (17) of the processing device (10) as well as a first removal device (18) and a second removal device (19) of the processing device (10) for processing the detected bale (4; 5-7; 20-22) only if at least one belt (2) with a parallel orientation but no belt with a transverse orientation is present in their respective working area.

2. Bale de-wiring system according to claim 1, characterized in that it comprises a trigger device (TE) by which the image acquisition device (11; 111) is activated as soon as one of the bales (4; 5-7; 20-22) fed on the transport device (5) in the transport direction (TR) has a has reached a suitable position for detection by the image acquisition device (11; 111).

3. Bale de-wiring system according to claim 1, characterized in that the classification model (15) is also configured to extract a bale (4; 5-7; 20-22) from the raw image data (RD) as a feature and / or to extract at least one area (6a-6f; 7a-7f; 8a-8f; 20c-20f; 21c-21f; 22c-22f) of the bale (4; 5-7; 20-22) from the raw image data (RD) as a feature and / or to extract from the raw image data (RD) as a further feature the presence of at least one band (2) on the extracted area (6a-6f; 7a-7f; 8a-8f; 20c-20f; 21c-21f; 22c-22f) or the extracted areas (6a-6f; 7a-7f;8a-8f; 20c-20f; 21c-21f; 22c-22f).

4. Bale de-wiring system according to at least one of the preceding claims, characterized in that the classification model is designed as a so-called machine learning model, wherein it is particularly provided to design the machine learning model as an artificial neural network, wherein it is particularly provided to design the artificial neural network as a so-called deep learning model, wherein it is particularly provided to design the deep learning model as a convolutional neural network (CNN).

5. Method for operating a bale de-wiring system according to one of claims 1 to 4, wherein raw image data (RD) of a dynamic scene (DS) of an observation section (BA) of the transport device (5) on which the bales (4; 5-7; 20-22) are fed are generated by the image acquisition device (11; 111), wherein the raw image data (RD) are filtered from the classification model (15) of the artificial intelligence unit (13), and the presence of at least one band (2) and its course in relation to the transport direction (TR) are extracted as features from the filtered raw image data (RD), and wherein, depending on the result of the extraction after the evaluated bale (4; 5-7; 20-22) has been fed through the feed chute (9) into the processing unit (10), the control device (14) will activate a first cutting device (16) and a second cutting device (17) as well as a first removal device (18) and a second removal device (19) for processing the analyzed bale (4; 5-7; 20-22) only if at least one band (2) is present in their respective working area.

6. Method according to claim 5, characterized in that an end face of the bale is extracted as a surface from the classification model (15).

7. Method according to claim 5, characterized in that the classification model (15) includes an end face (6b) as surfaces; 7b; 8b) and a side face (6c-6f; 7c-7f; 8c-8f; 10c-20f; 21c-21f; 22c-22f) of the ball (4; 5-7; 20-22) are extracted.

8. Method according to at least one of claims 5 to 7, characterized in that the processing device (10) is deactivated and / or that an ejection device (23) arranged in front of the processing device (10) is activated when a bale (4; 5-7; 20-22) is detected with an impermissible strapping.

9. Method according to claim 8, characterized in that a strapping in which the analyzed bale (4; 5-7; 20-22) on one of the side surfaces (6c-6f; 7c-7f; 8c-8f; 10c-20f; 21c-21f; 22c-22f) has a cross-hatching pattern formed by bands (2) and / or on one of the side surfaces (6c-6f; 7c-7f; 8c-8f; 10c-20f; 21c-21f; 22c-22f) has at least one band (2) running transversely to the transport device (TR), for which the control device (14) is assessed as an impermissible strapping.

10. A method according to at least one of claims 5 to 9 comprising the further steps: Classification of the bale by the image evaluation device into one of four classes (I; II, III; IV); in the case of classification of the bale (8) into class I. After the bale enters the processing unit (10), the processing unit is controlled such that the straps (2) are cut by a first cutting device (16) of the processing unit (10) and that the straps (2) are removed by a first removal device (18) of the processing unit (10); in the case of classification of the bale (7) into the class II. After the bale (7) has entered the processing unit, the processing unit (10) is controlled such that the straps (2) are cut by a second cutting unit (17) of the processing unit (10) and that the straps (2) are removed by a second removal unit (19) of the processing unit (10); in the case of classification of the bale (6) into the class III. After the bale (6) has entered the processing device, the processing device (10) is controlled such that the belts (2) are cut successively or simultaneously by the first cutting device (16) and by the second cutting device (17) of the processing device (10) and that the belts (2) are removed successively or simultaneously by means of the first removal device (18) of the processing device (10) and by means of the second removal device (19) of the processing device (10); in the case of a classification of the bale (20; 21; 22) into class IV, neither of the two cutting devices (16, 17) of the processing device (10) and neither of the two removal devices (18; 19) of the processing device (10) is activated.

11. Method according to claim 5, characterized in that a ball (4; 5-7; 20-22) is extracted as a feature from the filtered raw image data (RD) and / or that at least one area (6a-6f; 7a-7f; 8a-8f; 20c-20f; 21c-21f; 22c-22f) of the ball (4; 5-7; 20-22) is extracted as a feature from the filtered raw image data (RD).

12. Method according to claim 5, characterized in that the classification model is executed by a machine learning model, wherein it is particularly provided that the machine learning model is executed as an artificial neural network, wherein it is particularly provided that the artificial neural network is executed as a so-called deep learning model, wherein it is particularly provided that the deep learning model is executed as a convolutional neural network (CNN).

13. Computer program comprising instructions which, when executed by a computer, cause the computer to execute the method according to any one of claims 5 to 12.

14. Computer-readable storage medium on which the The computer program according to claim 13 is stored.

Images