Installation comprising a conveyor system having at least one conveyor device for determining and processing articles

Abstract

The invention relates to an installation comprising a conveyor system (1) for processing articles (2), comprising a supply conveying device (11) for supplying articles (2), an article determining device (13) for determining the supplied articles (2), a transferring device (15) for transferring the determined articles (2), and a processing device (17) for carrying out a respective operation associated with the determined article (2) on the basis of at least one piece of article information relating to the determined article (2). The article determining device (13) contains a computer system (20) comprising means for automatically determining articles (2).

Description

[0001] PLANT WITH A CONVEYOR SYSTEM WITH AT LEAST ONE CONVEYOR DEVICE FOR DETERMINING AND PROCESSING ARTICLES

[0002] The invention lies in the field of conveyor technology, in particular intralogistics, and relates to a system with a conveyor system with at least one conveyor device for determining and processing articles and an associated method.

[0003] In intralogistics, various items are processed in a conveyor system as part of warehouse management or order fulfillment. For this purpose, the items fed into the conveyor system are identified. This is done, for example, by reading information carriers such as labels, RFID tags, barcodes, QR codes, or chemical tags applied to the surface of the items. These information carriers are read, for example, at the beginning of the conveyor or along the conveyor path using appropriate readers.

[0004] However, such information carriers can be lost, damaged, or deteriorate and are therefore prone to errors. They must therefore be kept in good condition and regularly replaced.

[0005] Furthermore, the management of such information carriers, which includes, for example, the physical provision, programming or coding of the information carriers as well as the application of the information carriers to the articles, represents a considerable effort, which is also reflected in the operating costs.

[0006] Furthermore, for reliable use of such identification systems with information carriers, cross-industry standardization or harmonization of the information carriers and the associated hardware and software is generally required.

[0007] Furthermore, the use of labels or RFID tags generates waste, which is environmentally undesirable. Applying these information carriers to the items can also negatively impact their visual appearance. Additionally, the surface texture or geometry of the items can make application difficult or even impossible. Residue from information carriers or adhesive labels can also cause malfunctions in the conveyor system and lead to increased maintenance requirements.

[0008] The object of the invention is therefore to propose a solution that eliminates the aforementioned disadvantages associated with the use of readable information carriers.

[0009] The problem is solved by a system according to claim 1 and a method according to claim 16. The dependent claims, the description, and the figures include special embodiments and further developments as well as specifications of the invention.

[0010] The invention therefore relates to a system with a conveyor system for processing articles. The system comprises:

[0011] - a feeding device for conveying articles;

[0012] - an item identification device for identifying the conveyed items; - a conveying device for conveying the identified items; and - a processing device for executing each operation related to the identified item, depending on at least one piece of information about the identified item. The item identification device includes a computer system with means for the automated identification of items.

[0013] The invention is now characterized in that the article identification device comprises a camera system for creating images of articles, and a software application implemented in the computer system for identifying the articles, wherein the computer system comprises a database with images of already identified articles, and the software application for identifying the articles is designed to compare image information of the images taken by the camera system with image information of images stored in the database.

[0014] The images can be stored in the database in various ways. They can be stored as raw data, or they can be stored in a processed form, such as image information extracted from the images. This processing can be performed, for example, by a software application or an AI model.

[0015] If the database is, for example, a vector database, as described below, the image data can be stored in the vector database in the form of embeddings or vectors.

[0016] The images and their associated image information stored in the database for specific articles are each linked to at least one piece of article information. This article information could, for example, be the name of an article type. This article information is what makes it possible to identify the articles based on the stored images and their associated image information. These are also referred to as "labels" for the images or (feature) vectors. This at least one piece of article information can also be stored in the database.

[0017] Within the context of this invention, the term "article designation" means the assignment of the article to a specific article type. Such an article type can be, for example, a specific product, such as a specific garment of a particular size, a specific book, or a specific medication.

[0018] In the context of this invention, the term "specific article" means that the article in question has been determined using the article determination device according to the invention and by means of the method according to the invention.

[0019] Image information refers to a feature or aspect of the image. Image information therefore represents a visual characteristic of the filmed object, in this case, an article.

[0020] The inventive article identification is based on the understanding that articles fed into the conveyor system from the outside do not necessarily need to be uniquely identified in every case, as is possible, for example, using readable information carriers. In many applications, simply identifying the articles, i.e., assigning them to an article type, is sufficient.

[0021] The purpose of item identification is therefore primarily to assign an item to an item type, e.g., to a specific product. Since the system can also process multiple items of the same item type, i.e., several identical products, such as identical clothing items of the same size or identical books, item identification does not necessarily correspond to a one-to-one identification of the item.

[0022] The products mentioned can include, for example, consumer goods, pharmaceuticals, building materials or components, individual parts, or spare parts. Consumer goods can include, for example, clothing, food, electronic devices, or everyday necessities. The items or products can be packaged or unpackaged.

[0023] The items can also be packages containing several products, such as beverage containers. Furthermore, the items can also be palletized goods.

[0024] The software application is based in particular on a machine learning model, which is trained using machine learning software algorithms based on image recordings or image information about image recordings of specific articles stored in the database, in order to automatically determine the articles by comparing image information about image recordings of the camera system with image information about stored image recordings.

[0025] The term machine learning model refers to software that has been trained using machine learning software algorithms based on existing data, in this case image recordings, in order to, for example, extract image information from the image recordings using image analysis and to make certain decisions, such as article identification, automatically, i.e. without human intervention.

[0026] The image information can include, for example, image patterns that are recognized by the software algorithms when comparing the image recordings with image recordings stored in the database.

[0027] Image patterns can be specific image features or properties, such as a graphic, an outline, lettering, a color gradient, or an optically perceptible surface texture, etc.

[0028] The machine learning model is specifically a model based on artificial intelligence (AI model). The AI ​​model can be a generative AI model. For example, a generative AI model can evaluate (primary) image information from images stored in the database and generate new data or secondary image information for article identification. Image information therefore also includes image information derived from and further processed from images.

[0029] The AI ​​model is based in particular on a neural network.

[0030] Image matching can involve comparing or contrasting image information from the image recordings.

[0031] Image matching can involve comparing or juxtaposing complete image recordings using appropriate software algorithms.

[0032] During the matching process, the software application searches for matches between the images or image information. If, based on defined criteria, a sufficient match is found between the images or image information, the article can be identified. This is done by linking the article to be identified with at least one piece of article information related to the image(s) or image information stored in the database with which the match was achieved, creating a virtual representation.

[0033] Primary image information for the item in the image may include:

[0034] - Form of the article;

[0035] - Contours of the article;

[0036] - Footprint of the article;

[0037] - Dimensions of the article;

[0038] - Labels on the item;

[0039] - Edges of the item; - Surface color on the item;

[0040] - Surface structures or textures on the article;

[0041] - Surface markings on the item;

[0042] - Surface lettering on the article;

[0043] - Surface markings on the item;

[0044] - Image marks, word marks or combined marks on the article;

[0045] - Surface graphics on the article;

[0046] - Surface pattern on the item;

[0047] - Surface reflectivity;

[0048] - Number of products in one container.

[0049] Furthermore, the image information can also include attributes related to surface color, such as hue, saturation, brilliance, gloss, and color intensity.

[0050] The software application can use one or, in particular, several of the above-mentioned image information for article identification or comparison.

[0051] That is, the software application compares one and, in particular, several of the above-mentioned image information for at least one image of the article with image information for images that are stored in the database for specific articles in order to determine the article.

[0052] The software algorithms of the software application are specifically designed to derive image information, such as image patterns, from the image recordings of the articles, which are then used to search for images stored in the database.

[0053] The software application uses appropriate image analysis algorithms to search the database for identical or comparable image information or patterns. If the software finds a match between at least one image of the article and one or more images stored in the database, the article is assigned to the at least one article record linked to the relevant image(s) or their associated image information. Once assigned to this article record, the article is identified.

[0054] According to a further development of the invention, the database is designed as a vector database.

[0055] The image captures are stored in the vector database by the software application or by an AI model as so-called embeddings, i.e., vector representations of the image captures. Such vector representations are multidimensional vectors in which image information is represented in the form of numbers.

[0056] The image information is extracted from the image data using a so-called "feature extractor model" and converted into multidimensional vectors. For this purpose, the image information is extracted from the raw data of the image captures by the "feature extractor model" and translated into numerical values.

[0057] Such a "feature extractor model" is part of a AI model and is based in particular on a neural network, e.g. a Convolutional Neural Network, CNN.

[0058] Well-known vector databases include FAISS, Weaviate, and Pinecone.

[0059] A vector contains, in particular, all information, especially image information, about an article. A vector can, for example, represent a single image of an article. However, it is also possible for a vector to represent multiple images of the same article. In this way, the image information from several images can be converted into a single vector or embedded within the same vector. A vector then represents the image information about an article that was derived from one or more images.

[0060] It is also conceivable that in addition to image information extracted from image recordings, further information about the article in question, which is determined in other ways, such as the weight of the article, is converted and embedded together with the image information in the same vector.

[0061] For example, the well-known AI model CLIP (Contrastive Language-Image Pretraining) is able to link different types of information, such as text and image information.

[0062] A vector then represents not only image information about an article, which was obtained from one or more images, but also other information about the article.

[0063] The vector of a specific article is linked to at least one piece of article information about the article type, which allows the article to be determined.

[0064] The image information obtained or extracted from image recordings can, for example, be image information from the list revealed above.

[0065] The conversion of an image into a multidimensional vector occurs in several steps. Initially, the image is simply a matrix of pixel values, such as RGB color values ​​(raw data). Using the AI ​​model or neural network, image features or information are then extracted from the raw data.

[0066] The extracted image information is converted into (numerical) vectors, also known as embedding. Such vectors can have up to a thousand dimensions. The numerical values ​​of a vector thus represent the image information.

[0067] To identify an article, at least one image of it is taken. This image is then converted into a multidimensional vector, possibly incorporating other captured information about the article, as described above.

[0068] Using the vector generated from at least one image capture, appropriate algorithms, e.g. based on cosine similarity or Euclidean distance, are then used to search the database for similar vectors of already identified articles.

[0069] Determining similar vectors of already identified articles allows the determination of the article for a new vector.

[0070] According to a training course, the software application includes a "Support Vector Machine (SVM)". The SVM performs a classification on the captured vectors and serves to search for similar vectors for article identification.

[0071] According to a training course, the images undergo preprocessing before vector conversion. This preprocessing can include, for example, cropping the article from the image to eliminate irrelevant image elements, such as parts of the conveyor system, etc., which could interfere with or distort the image analysis. Preprocessing can also include normalization, such as resizing, orientation, or color correction. The goal of preprocessing is to ensure consistent image input for the model and to prevent distorting influences.

[0072] According to a training course, several images are taken of each item. Accordingly, multiple images are fed into the software application for item identification and comparison purposes.

[0073] For example, multiple images of each article can be taken from different perspectives or from different pages. Alternatively or additionally, multiple images of each article can be taken from different regions of the electromagnetic spectrum. These regions can lie within a segment of the electromagnetic spectrum ranging from ultraviolet to infrared. This can be achieved, for example, by using optical filters.

[0074] Depending on which part of the electromagnetic spectrum images are taken from, they can contain different image information. For example, an image taken in the UV spectrum may contain different information than an image taken in the visible spectrum. This is especially true if the items are marked with UV markings.

[0075] Taking multiple photos of the item results in more image information, which consequently increases the success rate in item identification.

[0076] The camera system contains at least one camera. The camera system can also contain multiple cameras, which can be used to take pictures of the item from different angles.

[0077] The at least one camera can be a conventional camera, a perspective camera, a depth camera (3D camera), or a video camera. Accordingly, the image capture can be a photograph or a video recording. Furthermore, the at least one camera can also be a line scan camera. These are particularly well-suited for this application because the items are moved past the camera by the conveyor motion.

[0078] The line scan camera is characterized by its ability to capture moving items line by line. A complete image is created by combining the individually captured image lines.

[0079] Unlike conventional cameras, the line scan camera has only a single line of pixels instead of a rectangular image sensor with many pixel rows.

[0080] The line scan camera has the advantage that it can create a complete image of the item, even if the image or field of view is small, i.e., even if the item is not completely within the camera's field of view.

[0081] For example, a line scan camera can film the item from below. Accordingly, the line scan camera is positioned below the conveyor track carrying the items. The line scan camera can film the items from below through the gap between two conveyors, such as belt conveyors. The line scan camera can thus create a complete image of the item from below, even if the field of view is relatively narrow and only captures a portion of the moving item at any given time.

[0082] It is also possible for a camera, such as a line scan camera, to film the items from above. The camera is positioned specifically above the conveyor belt carrying the items.

[0083] It is also possible for a camera, such as a line scan camera, to film the items from the side, viewed in the direction of transport. The camera is therefore positioned to the side of the conveyor path. Alternatively, a line scan camera can be positioned on each side, viewed in the direction of transport.

[0084] It is also possible for a camera, such as a line scan camera, to film the items from the front, viewed in the direction of transport. The camera can, for example, be positioned above the conveyor and its lens angled downwards with a directional component pointing against the direction of transport towards the moving item.

[0085] It is also possible for a camera, such as a line scan camera, to film the items from behind, viewed in the direction of travel. The camera can, for example, be positioned above the conveyor and its lens angled downwards with a directional component pointing towards the moving item in the direction of travel.

[0086] It is conceivable that the articles will be filmed by at least two of the following websites:

[0087] - viewed from a first side in the direction of delivery;

[0088] - viewed from one of the two sides opposite the first, in the direction of transport;

[0089] - from underneath;

[0090] - from above;

[0091] - viewed from the rear in the direction of conveyance;

[0092] - viewed from the front in the direction of conveyance.

[0093] Combinations of different camera types in one camera system are also possible.

[0094] An AI model is first trained using machine learning algorithms with images of specific items before it can be used for reliable item identification. This is done by creating images of each item type, storing them in the database, and linking them to at least one piece of information about the item type that allows for identification.

[0095] In a further development of the invention, the images captured by the camera system relating to specific articles are stored in the database. The software application, designed as an AI model, is then trained using machine learning algorithms with the newly stored images, e.g., (feature) vectors. The database can be continuously expanded with new images, e.g., (feature) vectors.

[0096] Accordingly, the computer system is designed to store the images captured by the camera system during operation in the database and to continuously train the AI ​​model with these images using machine learning algorithms. The images continuously stored in the database thus serve as training data for machine learning based on artificial intelligence.

[0097] This applies particularly to images of articles identified by the AI ​​model during operation. By continuously adding images to the database, the library of images used to train the AI ​​model becomes increasingly comprehensive. Consequently, the reliability of article identification by the AI ​​model also continuously improves.

[0098] It can be provided that the AI ​​model is trained only until a defined success rate is reached in identifying the articles, and that the AI ​​model is no longer trained once this success rate is reached. Such a success rate could, for example, be 99.9%. After reaching the defined success rate, the AI ​​model continues to work with the existing set of images stored in the database. According to a further development of the invention, each identified article is assigned an identification code after the system has identified it, allowing for the unique identification of the article.

[0099] Unique identification of items is necessary, for example, for tracking items in the conveyor system after item identification. Item tracking is required, for instance, to perform an operation related to the identified item following item identification.

[0100] Thus, a virtual representation of the conveyor system can be provided in the plant or its control unit to track the articles requested after the article determination device based on the assigned identification code.

[0101] The tracking of items takes place in the virtual representation of the conveyor system during their further processing within the system, based on the (virtually assigned) identification code. The identification code can be linked to at least one piece of item information, allowing the item type to be derived from the identification code.

[0102] Accordingly, the items are requested individually following the item identification system. For this purpose, conveyor systems with discrete conveying stations, such as tray conveyors or pocket conveyors, may also be used, which simplify tracking.

[0103] To track and locate the items, data such as control information, conveying speeds, and information from detection devices like light barriers are incorporated into the virtual model of the conveyor system along the item's path. The system therefore includes a control unit for controlling the conveyor system and its conveying components. The computer system is connected to the control unit and can also be integrated into it.

[0104] In particular, the article information assigned to the articles by the computer system during article determination is used by the control unit to control the conveying path and the delivery of the articles to the further processing facilities.

[0105] The system can be located, for example, in a logistics center or intralogistics operation and may be designed for sorting, picking or inserting items.

[0106] The items in question are primarily consumer goods. Consumer goods are defined as items intended for consumption or use. They can be finished products intended for an end consumer, or intermediate goods intended for further processing.

[0107] Consumer goods include, in particular: merchandise for sale, mail-order items, everyday items, and consumables. Consumer goods can also be semi-finished products, which are further processed, i.e., "consumed," in a subsequent production stage.

[0108] The articles are fed into the plant, in particular via a transfer device, and conveyed to the article determination device via the feed device.

[0109] Typically, various types of items are fed into the system and, consequently, into the item sorting device. The items are delivered, for example, by a transport vehicle such as a truck and fed into the system after unloading.

[0110] The items to be identified are conveyed by the item identification device, primarily in a horizontal position. During horizontal conveying, the items rest on a conveying surface or support surface. This conveying surface is formed, for example, by a driven conveying element such as the conveyor belt of a belt conveyor or the rollers of a roller conveyor.

[0111] Accordingly, the articles are conveyed by the infeed system, particularly in a horizontal position, to the article locating system. The infeed system specifically includes a horizontal conveyor for this purpose.

[0112] The conveying element for the horizontal conveying of the articles, in particular the conveying element of the feed device, can be a conveyor belt of a belt conveyor, a mat chain of a mat chain conveyor, a modular belt of a modular belt conveyor or the rollers of a roller conveyor.

[0113] The conveying element can also be formed by the conveying bowls of a bowl conveyor.

[0114] The infeed system can also include an overhead conveyor with conveyor pockets into which the items are fed to the item identification device. The conveying element is accordingly formed by the conveyor pockets of the overhead conveyor. According to this embodiment, the infeed system includes an unloading station where the items are unloaded from the conveyor pockets and transferred to a conveying device for horizontal conveying, so that the items can be conveyed horizontally through the item identification device.

[0115] The items can be dispensed at the unloading station, for example, downwards through the bottom of the bag, which can be opened. Such a suspended conveying device is described, for example, in WO 2018 / 202 512.

[0116] The system can generally comprise an overhead conveyor system in which the items are conveyed suspended in conveyor pockets and unloaded at an unloading station for identification purposes, particularly by gravity, and transferred to a horizontal conveyor. The items are conveyed and identified while lying down by the item identification device. Following the item identification device, the items are transferred from the horizontal conveyor back into conveyor pockets of an overhead conveyor system at a delivery station and transferred to a suspended conveyor system. This overhead conveyor system is, in particular, part of the downstream processing equipment.

[0117] In order for the camera system of the article identification device to be able to detect the articles individually and, if necessary, from different sides, it is advantageous if the articles are fed to the article identification device individually.

[0118] The conveying system can include a singulation device positioned upstream of the item determination unit when viewed in the conveying direction. This is particularly useful when items are fed in as a continuous stream, especially on an endless conveyor. The singulation device separates the items requested by the feeder before they reach the item determination unit. The singulation device can be part of the feeder.

[0119] The singulation device can include a discharge mechanism that releases the items individually, for example, by opening a barrier or a pivoting arm. Upstream of the discharge mechanism, the singulation device can form an accumulation section for piling up the accumulating items. According to one embodiment, the singulation device comprises a roller conveyor with pivoting roller modules, by means of which the items transported over the roller modules can be accelerated or decelerated in the conveying direction and, in particular, also deflected from the conveying direction. The roller modules each contain one or more rollers and are pivotable about a pivot axis that is arranged perpendicular to a support surface or conveying surface. The pivot axis is also arranged perpendicular to the axis of rotation of the rollers.

[0120] The roller feeder can thus have multiple roller modules embedded in a support table, the rollers of which project beyond the support surface of the table. The roller(s) of one roller module can be driven independently of the rollers of other roller modules. Likewise, the roller modules can be pivoted about the pivot axis independently of other roller modules.

[0121] Such a roller conveyor is described, for example, in EP 4299477 Al.

[0122] By accelerating or decelerating the items using a roller conveyor, individual items can be extracted from or retained within a stream of goods. Furthermore, by deflecting the items laterally from the conveyor direction, they can be positioned advantageously for optimal detection by the camera system, e.g., in a central position.

[0123] According to further training, the item identification system includes a positioning device that allows the orientation of the items within the system to be changed, enabling the camera system to detect the items from different sides. The positioning device is specifically designed to rotate the items around an axis perpendicular to the conveying or support surface. The positioning device can, for example, be a roller conveyor, as described above in connection with the singulation device. The roller modules can be aligned and the rollers driven in such a way that the item is rotated around an axis perpendicular to the conveying surface.

[0124] Such a positioning device is particularly useful when the camera system itself is not designed to capture the item with multiple cameras or with moving cameras from different sides. This is especially true when the camera is fixed in place.

[0125] The item identification device or its camera system can be designed to take at least one image during the movement or conveying of the item by the device. The item identification device can therefore be designed to identify the items during their conveying within the conveyor system, i.e., during conveyor operation.

[0126] However, it is also conceivable that the conveying movement of the articles is temporarily stopped when passing the article identification device in order to create at least one image recording by the camera system, or at least that the conveying speed is reduced so that at least one image recording of the article can be created, e.g. in a resting position or when moving slowly past.

[0127] The item identification device, or the space above the conveying surface or the item support surface in the area of ​​the item identification device, can be enclosed in a housing. The camera system is also arranged within the housing.

[0128] In a further training system, the infeed device or its (feeding) space above the conveying surface or the support surface of the items may also be enclosed by a housing. In a further training system, the onward conveying device or its (feeding) space above the conveying surface or the support surface of the items may also be enclosed by a tunnel-like housing.

[0129] The housing is designed in particular as a tunnel or conveyor tunnel through which the articles are conveyed.

[0130] The tunnel contains an entrance on the supply side and an exit on the supply side.

[0131] Shielding the items with a housing allows, for example, the creation of images under constant lighting conditions. This ensures that the items are uniformly exposed during the image capture. This leads to better results when extracting image information from the photographs.

[0132] The tunnel can be illuminated by means of a lighting device, e.g. to illuminate the items in the item identification device for the purpose of taking photographs.

[0133] The conveying device is located downstream of the item positioning device. The conveying device moves the items away from the item positioning device, particularly in a horizontal position. In this case, the items rest on the conveying surface of a conveyor. The conveying device specifically includes a horizontal conveyor.

[0134] The conveying element for horizontal conveying of the articles can be a conveyor belt of a belt conveyor, a mat chain of a mat chain conveyor, a modular belt of a modular belt conveyor, or the rollers of a roller conveyor. The conveying element can also be formed by the conveying trays of a tray conveyor.

[0135] Following identification and, where applicable, assignment of an identification code, the items are requested from a processing facility. The requesting facility is, in particular, part of the processing facility.

[0136] The processing facility can be an order processing facility through which orders, such as item orders, are processed.

[0137] In the further processing facility, an operation is performed on the article depending on at least one piece of article information for the specific article.

[0138] The operation can be an action, processing step, or measure related to the specific item. In particular, the operation is a physical action performed on the item.

[0139] The operation associated with the specific article can be:

[0140] - Insert;

[0141] - Packaging, palletizing;

[0142] - prepare for shipping;

[0143] - Send;

[0144] - Addressing;

[0145] - Check;

[0146] - Edit or

[0147] - a settlement

[0148] or a combination thereof.

[0149] According to a further development of the invention, the processing device comprises a plurality of dispensing points for the respective dispensing of the specific article at a dispensing point, depending on at least one piece of article information relating to the specific article. The operation related to the specific article here consists of conveying the specific articles, depending on the at least one piece of article information relating to the specific article, to a defined dispensing point within the conveying system or the article processing device and dispensing the articles at the dispensing point.

[0150] The delivery points can be assigned, for example, to sorting stations of a sorting facility, picking stations of a picking facility, shipping stations of a shipping facility, or storage stations of a storage facility.

[0151] According to a training course, the processing facility is a sorting facility for sorting the articles based on at least one piece of information about the specific article.

[0152] The specific items are delivered to the delivery points along a delivery route according to defined sorting criteria. These sorting criteria are based on at least one piece of information about the specific items.

[0153] For example, the system can be designed for sorting returns. The specific items are delivered to the respective drop-off points in the sorting facility based on their identified item type, and thus sorted according to that type.

[0154] According to a training course, the further processing facility is a picking facility for picking or assembling articles based on at least one article information about the articles.

[0155] The items are delivered and assembled along a delivery route, for example, based on an order received at the respective delivery points. Order picking is based on at least one piece of information about the specific items, which includes, for example, the item type.

[0156] The aforementioned processing equipment, such as sorting or picking equipment, can include a tray feeder, particularly a rotary feeder, such as a tilting tray feeder, for delivering the items to the delivery points. In this feeder, the specific items are conveyed into trays for further processing. The conveying elements are thus formed by trays or tilting trays arranged one behind the other along the conveying path. Tilting trays are characterized by the fact that they can be tilted to the side to deliver the items to the delivery points. Such a tilting tray feeder is described, for example, in EP 4286304 Al.

[0157] The aforementioned processing equipment, such as sorting or order picking equipment, can include a transverse belt conveyor, particularly a circular conveyor, for delivering the articles to the delivery points. In this conveyor, the conveying elements are formed by driven transverse belts. The articles are delivered via the transverse belts, which are driven perpendicular to the conveying direction. Such a transverse belt conveyor is described, for example, in DE 19845 527 A 1.

[0158] The aforementioned processing equipment, such as sorting or order picking equipment, may include a shoe conveyor, particularly a circular conveyor, with discharge elements that can be moved laterally across the conveyor surface, for delivering the items to the delivery points. A shoe conveyor is described, for example, in WO 2021 / 037609 Al.

[0159] The aforementioned processing equipment, such as sorting or picking equipment, can also include a suspended conveyor system with conveyor pockets, particularly one designed as a circular conveyor, in which the specific items are suspended in conveyor pockets for further processing, sorting, or picking. The items can be discharged at a discharge station, for example, downwards through the pocket bottom, which can be opened.

[0160] Such a suspended conveyor system is described, for example, in WO 2018 / 202 512.

[0161] The aforementioned processing equipment, such as sorting or picking equipment, can include a split-tray conveyor (also called a bombbay conveyor or flat-tray conveyor), particularly one configured as a circular conveyor, for delivering the items to the delivery points. With a split-tray conveyor, the items are discharged downwards via downward-swinging flaps.

[0162] According to further training, the processing facility is a storage facility for placing specific items in storage locations, the allocation of which depends on at least one piece of information about the specific item.

[0163] For example, the selection of a storage location can depend on at least one piece of information about the item. Frequently requested items are thus stored in locations that are easily accessible, while less frequently requested items are stored in locations with longer access times.

[0164] The storage locations can also be of different sizes, so that the allocation of items to storage locations depends on the item size, which in turn can be derived from at least one piece of item information.

[0165] In both cases, identifying the item is necessary for optimal allocation of storage spaces. According to further training, a processing unit is a processing device for executing a processing step on the item, depending on at least one piece of information about the item.

[0166] For example, different processing steps can be performed depending on the type of item. These processing steps could include, for instance, packaging the item. Accordingly, depending on the item information, the items can undergo processing steps at different processing stations.

[0167] Viewed in the direction of conveyance, the system can include a reject device for removing unidentifiable items after the item identification unit. The reject device can, for example, be integrated into the downstream processing unit.

[0168] According to another embodiment, the processing unit is a billing unit for billing the items determined by the item identification unit, based on at least one piece of information about the specific item. The billing unit may include a payment unit for cashless payment, particularly electronic or digital payment, for an item order or purchase. The billing unit may also include a printer unit for printing documents, especially billing documents.

[0169] The system can be designed to process orders or purchases of articles, in which the articles are identified according to the invention, and subsequently, price information is assigned to each article based on at least one piece of article information. The processing of orders or purchases is carried out by the processing system based on this price information. The price information can be contained within the at least one piece of article information. Alternatively, the price information can be stored in a database, such as that of the processing system, and linked to the at least one piece of article information for article identification in order to determine the price.

[0170] According to a training course, the system includes a weighing device for automatically weighing the items requested by the feed system.

[0171] The weighing device can be positioned upstream of the processing equipment in the conveying direction. Alternatively, the weighing device can be part of the processing equipment.

[0172] The weight determined by the weighing device for each item can be incorporated into the software application as additional information for item identification. In this case, the weighing device can be part of the item identification system.

[0173] The weight determined by the weighing device for the items can be used as additional information by a billing system to determine the price of an item with a weight-dependent price.

[0174] The inventive system with conveyor system is not only used in intralogistics, but can also be designed, for example, as a cash register system with a billing device for settling item combinations or purchases.

[0175] Instead of reading a barcode or QR code, the items of a purchase are determined and billed in the cash register system using the inventive system and the inventive method.

[0176] Such a point-of-sale (POS) system can be used, for example, in retail stores, particularly supermarkets. Thanks to this type of system, information carriers such as barcodes or price tags can be dispensed with, making the POS system more flexible. For instance, if there are price changes to products, such as during promotions, only the corresponding price information for each item needs to be updated in the computer system's database. There's no need to cover existing information carriers, such as barcodes, with updated information.

[0177] The invention also relates to a method for processing articles in a aforementioned plant according to the invention.

[0178] The process is characterized by the fact that at least one image of an item is captured using the camera system. Item identification is based on images of previously identified items stored in the database. The software application implemented in the computer system compares the image information of the at least one image captured by the camera system with image information from the images stored in the database. The item identification is then performed based on this image comparison.

[0179] Following the article identification, the further processing unit performs one operation related to the identified article, depending on at least one piece of information about the identified article.

[0180] The identification of articles and the comparison of image recordings are particularly associated with computational operations, which are used especially in valuations and which are performed by software algorithms within the software application. The present invention has the potential to significantly simplify processes in logistics centers or in retail by eliminating the need to attach information carriers to the articles and to manage these information carriers for article identification. Article identification is carried out exclusively by the external appearance of the article, which is captured by means of image recordings. This external appearance generally contains sufficient (image) information for article identification.

[0181] Another aspect of the present invention relates to a system for detecting defective articles in a conveyor system of the plant.

[0182] According to this aspect, the plant also includes a conveyor system for processing articles. The conveyor system contains:

[0183] - a feeding device for the feeding of articles;

[0184] - an article inspection device for identifying defective articles, and - a rejection device for rejecting defective articles.

[0185] During the handling of items in logistics centers, damage to the goods occasionally occurs. This can range from damaged packaging to damage to the contents themselves. In many cases, the damage is visible from the outside. For example, the packaging may be torn, or the surface of the damaged packaging or the damaged item may show scratches or be deformed, e.g., dented.

[0186] According to this further aspect of the invention, such defective articles can now be identified using the same inventive method and the same technical means as for article identification. Accordingly, the above disclosure relating to article identification, particularly with regard to the technical equipment, also applies to this aspect of the invention, provided that the features do not relate to article identification itself.According to the invention, the article inspection device comprises a computer system and a camera system for creating images of articles and a software application implemented in the computer system for automatically identifying defective articles, wherein the computer system includes a database with images of already identified articles, and the software application for identifying defective articles is designed to compare image information of the images taken by the camera system with image information of images stored in the database.

[0187] The images stored in the database are primarily of (pre-defined), undamaged items. The software application compares the at least one image taken by the camera with the stored images of the same item type.

[0188] During the comparison process, the software application searches for discrepancies between the images or image information. If a discrepancy is detected, it allows for the identification of defective items. The criteria for the type and extent of the discrepancies that lead to the identification of a defective item can be predefined.

[0189] It can also be provided that the database contains, as an alternative or supplement, images of (pre-defined) items that have been assessed as defective. By comparing the images or their associated image information, the software application identifies a defective item if there is a match between the images or image information. The image information for the images of defective items stored in the database could, for example, include typical damage patterns.

[0190] To identify defective items, the items to be inspected must first be defined. This is the only way to ensure that only images of the same item type are compared. According to a further aspect of the invention, the items to be inspected are accordingly predefined. The computer system already has at least one piece of information relevant to the item being inspected, enabling item identification. Specifically, the item to be inspected is linked to this at least one piece of information in the virtual representation of the plant or conveyor system.

[0191] Since the article must be determined in advance, the method or the technical means for identifying defective articles, according to the further aspect of the invention, can also be implemented in the article identification device described above.

[0192] This means that the item identification device is designed not only to identify the item but also to detect defective items following item identification. The item identification device then corresponds to an item identification and testing device.

[0193] The software application can also be based on a machine learning model to identify defective items. This model is trained using machine learning algorithms based on images stored in the database, in order to identify defective items by comparing image information from the camera system with image information from stored images.

[0194] In principle, the characteristics of the machine learning model as revealed above in connection with the article determination device also apply here.

[0195] According to further training, the computer system can also be designed here to store image recordings captured by the camera system in the database and to continuously train the machine learning model with the image recordings captured by the camera system using machine learning algorithms.

[0196] These can be photographs of specific, undamaged items. Alternatively or additionally, these can be photographs of specific, defective items.

[0197] The database can be a vector database, as described in detail above in connection with article determination. The images are stored in the vector database as embeds or vectors. For the structure, functionality, and operation of the vector database, please refer to the disclosure above.

[0198] According to a training course, the computer system is designed to assign an identification code to each defective item and to track the defective items using the identification code in a virtual representation of the conveyor system until they are removed.

[0199] According to a further training, a singulation device is arranged in the conveying direction upstream of the article inspection device for singulating the articles conveyed by the feeder device as an article stream to the article inspection device.

[0200] The feeding device is designed for the horizontal feeding of articles to the article testing device and includes in particular a horizontal conveyor, such as a belt conveyor.

[0201] The articles to be determined are conveyed by the article testing device, particularly in a lying position.

[0202] The (defective) items are conveyed from the item inspection unit to a rejection unit, primarily by means of a conveying device. This conveying is typically done horizontally, for example, using a horizontal conveyor such as a belt conveyor. The associated process for handling items in the system described above is characterized by the fact that at least one image of each item is captured by the camera system. Using the software application implemented in the computer system, defective items are then identified and rejected at the rejection unit. This is done by comparing the image information of the at least one image captured by the camera system with the image information of the images stored in the database.

[0203] According to a further development of the invention, the images captured by the camera system of defective and / or undamaged articles are stored in the database. A machine learning model is then trained using machine learning algorithms with the stored images.

[0204] The machine learning model is, in this context, primarily a model based on artificial intelligence (AI model). The AI ​​model can be a generative AI model. For example, a generative AI model can analyze (primary) image information from images stored in the database and generate new data or secondary image information to identify defective items. Image information therefore also includes image information derived from and further processed from image recordings.

[0205] The images stored in the database are each linked to at least one piece of article information stored in the database for the purpose of identifying the articles.

[0206] According to a training course, each item requiring inspection is assigned an identification code. These items, or rather their identification codes, are linked to at least one piece of information stored in the database for item identification purposes. Defective items are tracked, using their identification codes, in the virtual model of the plant or conveyor system until they are removed.

[0207] The software application compares one or more of the following image information from at least one image related to the article with corresponding image information from images in the database:

[0208] - Form;

[0209] - Contours;

[0210] - Dimensions;

[0211] - Labels;

[0212] - edges;

[0213] - Surface color;

[0214] - Surface structure or texture;

[0215] - Surface markings;

[0216] - Surface lettering;

[0217] - Surface markings;

[0218] - Figurative marks, word marks, combined marks;

[0219] - Surface graphics;

[0220] - Surface patterns.

[0221] According to a training course, several images of the articles are taken from different perspectives, especially from different article pages, and the images are fed into the software application to identify defective articles.

[0222] For this purpose, a camera system or cameras can be used, as described above in connection with article identification.

[0223] The identification of defective items or the comparison of image recordings is particularly associated with calculation operations, which are used especially in valuations, and which are carried out using software algorithms of the software application.

[0224] The inventive method of identifying and removing defective items from a system or conveyor system is intended to prevent defective items from being delivered from a logistics center to clients or customers. This should reduce the number of returns.

[0225] The invention will now be explained in more detail with reference to preferred embodiments, which are illustrated in the accompanying figures. These schematically depict:

[0226] Figure 1: a top view of a system according to the invention in a first embodiment;

[0227] Figure 2: a perspective view of an article identification device;

[0228] Figure 3: a perspective view of a roller conveyor that can be used as a singulation device, positioning device or conveyor switch;

[0229] Figure 4: a top view of a dispensing device with multiple dispensing points; Figure 5: a top view of a system designed as a cash register system according to the invention;

[0230] Figure 6: a top view of a further embodiment of a system designed as a cash register system according to the invention;

[0231] Figure 7: a side view of a system according to the invention in a further embodiment;

[0232] Figure 8: a flowchart of the inventive process;

[0233] Figure 9: a side view of a camera system for the inventive device.

[0234] In principle, identical parts in the figures are designated with the same reference numerals. For the sake of clarity, certain features are not shown in the figures. The described embodiments are exemplary of the subject matter of the invention and do not have a limiting effect. The conveying system 1 of the inventive system according to Figure 1 includes a feed device 11 in the form of a belt conveyor, on which the articles 2 are conveyed in a stream of articles in conveying direction F to an article determining device 13. The feed device 11 includes a singulation device 12.1 arranged in conveying direction F upstream of the article determining device 13, at which the articles 2 to be determined are singulated from the article stream and conveyed individually to the article determining device 13.

[0235] The singulation device 12.1 comprises a roller conveyor with a plurality of roller modules 50 embedded in a conveyor table 53, each module having several driveable rollers 51 that project beyond the support surface of the conveyor table 53 (see also Figure 3). The rollers are rotatable about a geometric axis of rotation D located in the conveying plane. By rotating the driven rollers 51, the articles 2 are moved or conveyed in the conveying plane via the roller conveyor.

[0236] The roller modules 50 can be pivoted about a geometric pivot axis S oriented perpendicular to the conveyor plane. By pivoting the roller modules 50, the direction of movement of the articles 2 in the conveyor plane can be changed.

[0237] To separate the articles 2 from the article stream, individual articles 2 are accelerated in the conveying direction F via the rollers 51 of the roller modules 50, so that they move faster in the conveying direction compared to the articles 2 of the article stream.

[0238] At the same time, the individual articles 2 are deflected towards the center of the conveyor by selectively pivoting the roller modules 50 around the pivot axis S, so that the individual articles 2 are conveyed centrally into the article determination device 13.

[0239] The article determination device 13 includes a positioning device 12.2 with a roller conveyor, which is designed identically to the roller conveyor of the singulation device 12.1. For article determination purposes, the singulated articles 2 are directed onto the support surface of the positioning device 12.2.

[0240] The article identification device 13 comprises a camera system 30 with at least one camera 31 (see also Figure 2). Figure 2 shows, by way of example, a conventional still camera 31, a video camera 32, and a perspective camera 33.

[0241] The at least one camera 31 creates at least one image recording of the article 2 to be determined. The article determination device 13 is equipped with a lighting device 34 for illuminating the articles 2.

[0242] To create multiple images of different sides of the article 2, the article 2 to be determined is rotated on the roller conveyor, in particular around an axis perpendicular to the conveyor plane, by means of appropriately aligned roller modules 50.

[0243] The article identification system 13 further comprises a computer system 20 with a database 21 and a processing unit 22, on which a software application in the form of an AI model 23 is implemented. The database 21 contains images of previously identified articles 2, which the AI ​​model was trained on. This means that the AI ​​model knows the article type for each image in the database. For this purpose, the images are linked to at least one piece of article information, which is also stored in the database. This at least one piece of article information allows the article type to be determined. The article type defines what kind of product the article is, e.g., a specific garment of a certain size, a specific medication of a certain package size, or a specific book.

[0244] The at least one image of the article 2 in the article identification device 13, created by the at least one camera 31, or image information from the at least one image, are compared with image recordings stored in the database or with image information about image recordings stored in the database by the KL model.

[0245] If the image comparison or the comparison of the associated image information results in a match based on the criteria of the AI ​​model, then the associated at least one piece of article information is assigned to Article 2 and Article 2 is thereby determined.

[0246] The at least one image of the now identified article 2 can be stored in database 21 and used for further training of the AI ​​model.

[0247] The system also includes a control unit 60 for controlling the conveying system and its associated conveying equipment. The control unit 60 or the computer system 20 assigns a (virtual) identification code to the specific article 2, by means of which the specific article can be uniquely identified and tracked on its way through the conveying system 1.

[0248] In the control unit 60, a virtual representation of the plant or conveyor system 1 is implemented for tracking specific articles 2. Tracking the articles 2 within conveyor system 1 is necessary to move them through conveyor system 1 in a targeted manner, based on at least one piece of information assigned to each article, and, for example, to deliver them to the correct dispensing point.

[0249] After the article 2 has been identified by the article determination device 13 and assigned an identification code, it is conveyed out of the article determination device 13 by the conveyor rollers 51 of the positioning device 12.2 and transferred to a conveying device 14. The identified articles 2 are then conveyed via the conveying device 14 to a processing device 17. Figure 1 shows three exemplary embodiments of processing devices 17.

[0250] A first and second embodiment relate to a sorting 17.1 and picking device 17.2, which are shown together in Figure 1. These will be explained in more detail in connection with Figure 4.

[0251] A third embodiment relates to a storage facility 17.3 and a fourth embodiment relates to a shipping facility 17.4.

[0252] The four embodiments have in common that these dispensing points are equipped with collection points or collection areas for collecting or receiving the dispensed articles 2. The specific articles 2 are dispensed at the corresponding dispensing points by the control device 60, depending on the associated at least one piece of article information.

[0253] The conveyor system 1 contains a conveyor diverter 15 downstream of the article determination station 13, by means of which the determined articles 2 are assigned to one of the three further processing facilities 17.1-17.3 mentioned.

[0254] The conveyor diverter 15 is designed as a vertical conveyor, as already described above in connection with the singulation device. The diverter function is performed by the swiveling roller modules 50, which, depending on their swivel position around the pivot axis S, deflect the articles 2 to the left or to the right, or convey them straight ahead.

[0255] Figure 4 shows an embodiment of a processing device, such as can be used in a sorting device 17.1 or a picking device 17.2. This device includes a discharge conveyor 55, which is designed as a roller conveyor, as described above in connection with the singulation device 11. Discharge points 56 are provided on both sides of the discharge conveyor 55, at which articles 2 can be discharged laterally. From the discharge points 56, conveyor chutes 57 extend laterally at an angle, over which the articles 2 slide down to collection points 58. The sorted or picked articles 2 are collected at the collection points 58.

[0256] The dispensing function is performed by the swiveling roller modules 50, which, depending on their swivel position around the swivel axis S, deflect the articles 2 to the left or right at the dispensing points 56 and thus dispense them laterally or - without dispensing - convey them straight ahead.

[0257] At the end of the discharge conveyor 55, a discharge station 59 is arranged for the discharge of the unspecified articles 2.

[0258] The storage facility 17.3, as indicated in Figure 1, has a large number of storage locations to which the articles 2 are delivered for storage.

[0259] The shipping facility 17.4, as indicated in Figure 1, has a plurality of shipping stations where the delivered items 2 of a shipping order are collected, possibly packed or palletized and made ready for shipment.

[0260] Figure 3 shows a versatile roller conveyor, here in its function as a singulation device 12.1. The roller conveyor contains a plurality of roller modules 50, each with four driveable rollers 51, embedded in a conveyor table 53. The conveyor table 53 forms a support surface 52. The arrows P shown indicate the direction of movement of the articles 2 moved over the corresponding roller modules 50 and serve to position the singulated articles 2 centrally in the conveying plane. Figures 5 and 6 show a system according to the invention implemented as a checkout system 70. The checkout system 70 can be used, for example, in a supermarket. The purchases, corresponding to the articles 2, are placed from a shopping cart or shopping basket onto a belt conveyor 11 as is known. The belt conveyor 11 corresponds to the infeed device, which requests the articles 2 in the conveying direction F of the article determination device 13.

[0261] In the conveying direction F upstream of the article determination device 13, a singulation device 12.1 in the form of a roller conveyor is arranged, as already described in detail in connection with Figure 1 and Figure 3. The articles 2 are singulated from the article stream in the singulation device 12.1.

[0262] The individual articles 2 are then conveyed to the article determination device 13 and determined as already described above in connection with Figure 1 and Figure 2.

[0263] The article identification device 13 replaces the conventional scanning system for reading barcodes and QR codes.

[0264] The article identification device 13 includes a positioning device 12.2 for rotating the articles in order to create images of different article sides. The structure and function of the article identification device 13 have already been described in detail above in connection with Figures 1 and 2.

[0265] The cash register system 70 also includes a weighing device 16 for the automated weighing of the items 2. The weighing device 16 is arranged along the conveyor line for the items 2, following the item identification device 13. Weight information is particularly important for unpackaged items 2 with weight-dependent prices, such as vegetables or fruit, in order to determine the item price. The cash register system 70 further includes a settlement device 17.6 for settling the invoice for the purchased items 2. For this purpose, the settlement device 17.6 links at least one item information piece for the specific item 2 with a price information stored with the item information to determine the item price. Based on this price information and, if applicable, the item weights, the settlement device 17.6 calculates the purchase amount due.

[0266] The billing unit 17.6 also includes a payment terminal for cashless, i.e. electronic, payment of purchases.

[0267] The items 2 of a purchase are requested after item determination and invoicing and collected in a collection bay 71 in the usual manner.

[0268] In the collection bay, the purchased items are received from the buyer and packed. The cash register system 70 can include several such collection bays 71. Accordingly, the cash register system 70 can contain a conveyor switch 15 upstream of the collection bays 71, by means of which the items 2 of a purchase are directed into an empty collection bay 71.

[0269] The conveyor switch 15 is designed as a roller conveyor, as described above. The switch function is performed by the pivoting roller modules 50, which, depending on their pivot position around the pivot axis S, deflect the articles 2 to the left or convey them straight ahead.

[0270] According to the embodiment shown in Figure 6, the space above the conveying surface, at least in the area of ​​the item identification device 13, is shielded from the outside by a conveying tunnel 65. The items 2 are guided through this conveying tunnel 65 as they pass the item identification device 13. The conveying tunnel 65 has the advantage that the camera system 30 and, in particular, the lighting device 34 are shielded from the outside, thus preventing, for example, disruptive glare from the lighting. Furthermore, the conveying tunnel 65 enables better light management, which in turn leads to better image quality. In addition, the conveying tunnel 65 prevents disruptive influences or even unauthorized external interference that could impair item identification.

[0271] Figure 7 shows another embodiment of a system according to the invention. The system includes a feeder 11 for feeding the articles 2 to an article identification device 13. The feeder 11 includes a suspended feeder 64 with conveying pockets 66, into which the articles 2 to be identified are conveyed to an unloading station 91. At the unloading station 91, the articles 2 to be identified are individually discharged from the conveying pockets 66 and transferred to a belt conveyor 10. The articles 2 are thus transferred to a horizontal conveying system. The unloading station 91 includes an opening device 81 for opening the conveying pockets 66 in order to discharge articles 2 from the conveying pockets 66.

[0272] In the present example, the pocket bottom 68 can be moved from a closed position to an open position by means of the opening device 81, in particular pivoting, so that the articles 2 can fall downwards onto the belt conveyor 10 through a bottom opening released by the pocket bottom 68.

[0273] The dispensing of the items 2 from the conveyor bags 66 can be done individually. In this case, no additional singulation device is necessary for singulating the items 2. Instead, the items 2 can be directly requested to the item determination device 13.

[0274] However, several items 2 can be dispensed from a conveyor bag 66 at once. Accordingly, a singulation device, e.g. of the type described above, is provided downstream of the unloading station 91, in particular for singulating the items 2.

[0275] The items fall downwards from the conveyor pockets 66 at the unloading station 91, primarily due to gravity.

[0276] Furthermore, the system includes the aforementioned article identification device 13 with a camera system 30 and a computer system 20 for identifying the articles 2.

[0277] Furthermore, the system in the area of ​​the article identification device 13 includes a positioning device 12.2 for positioning, i.e., rotating the articles 2, in order to create images of different sides of the articles. The positioning device 12.2 is a roller conveyor as already described above.

[0278] Following the article determination, the articles 2 are conveyed by a conveyor system 14 to a processing facility 17.

[0279] The processing device 17, in the present embodiment, is an overhead conveyor 17.5 with conveyor pockets 66 suspended along a conveyor path. The articles 2 are transferred from the belt conveyor of the processing device 14 to the conveyor pockets 66 of the overhead conveyor 17.5 at a loading station 90 for further processing. For this purpose, the loading station 90 includes a pocket opening device 81 for opening the previously folded conveyor pockets 66. The pocket opening device 81 has an opening element that is movable, in particular pivotable, from below against the pocket bottom 68 and pushes it upwards. This causes the pocket bottom 68 to fold from an approximately vertical orientation to an approximately horizontal orientation. The transport pocket 66 is now open, and the articles 2 can fall from above through the pocket opening into the open conveyor pocket 66.When the conveyor bag 66 is opened, it is also pressed against a bag stop 82, so that the conveyor bag 66 cannot move during the opening process and the dispensing of the item.

[0280] The loading station 90 also includes a holding and release device 83, by means of which the conveyor pocket 66 is held in place during the transfer of the articles 2. As soon as the article 2 has been dispensed into the conveyor pocket 66, the holding and release device 83 releases the conveyor pocket 66 for further conveying. The holding and release device 83 may include switchable mechanical elements that hold the conveyor pocket 66 in place. The holding and release device 83 is connected to the control unit 60.

[0281] The conveyor pockets 66 are equipped with readable information carriers 67 for the purpose of tracking the articles 2 in the processing unit 17 or overhead conveyor 17.5. Thus, when an article 2 is transferred to a conveyor pocket 66, the information carrier 67 is read by a reading device 85 for the purpose of identifying the conveyor pocket 66. For this purpose, the reading device 85 reads, in particular, an identification code stored in the information carrier 67. The reading device 85 is connected to the control unit 60.

[0282] For the purpose of tracking an article 2 dispensed into the conveyor bag 66, the control device 60 links the identification code of the conveyor bag 66 with the identification code of the article 2 transferred into the respective conveyor bag 66. For this purpose, it is specifically provided that only a single article 2 is dispensed into a conveyor bag 66 at a time.

[0283] The overhead conveyor 17.5 now conveys the articles 2 in the conveyor pockets 66, depending on associated article information, in particular to a specific delivery point, which is arranged along a delivery line with several delivery points. Tracking of the articles 2 in the downstream processing unit 17 or overhead conveyor 17.5 is carried out via a virtual representation of the system in the control unit 60, as described above.

[0284] Figure 8 shows a flowchart for the inventive method for article determination.

[0285] The articles 2 to be identified are fed in step S1 by means of the feeder 11 and separated in the subsequent step S2 by means of the singulation device 12.1 and fed individually to the article identification device 13. In order to create images of different sides of the articles 2, the articles 2 are positioned accordingly in the article identification device 13 (step S4) by means of the positioning device 12.2 by rotating the articles 2 on the conveyor surface (step S3).

[0286] In step S5, the created images and their associated image information are compared with images and their associated image information stored in database 21. In step S6, the articles 2 are determined from this comparison according to the inventive method.

[0287] For tracking purposes, each of the identified articles 2 is assigned a (virtual) identification code in step S7. Following article identification, in step S9, the articles 2 are transferred to a processing unit 17 for the execution of an operation on the article 2. The articles 2 are tracked in the processing unit 17 using a virtual representation of the system (step S8).

[0288] The camera system shown in Figure 9 comprises a first line-scan camera 35a, which is arranged below the conveyor track and whose optics are directed upwards through the conveyor gap 38 between two belt conveyors 36, 37. The first line-scan camera 35a accordingly films the underside of the article 2 conveyed above it from below.

[0289] The camera system in question further comprises a second line camera 35b, which is arranged above the conveyor belt and whose optics are directed downwards. The second line camera 35b accordingly films the top side of the article 2 conveyed below it from above.

[0290] Furthermore, the camera system in question comprises a third and fourth line camera 35c, which, viewed in the conveying direction F, are arranged laterally to the conveyor and, in particular, opposite each other. The optics of the laterally arranged line cameras 35c are each directed from the side towards the moving article 2. The third and fourth line cameras 35c each film one side of the article 2 as it is conveyed between them.

[0291] The line scan cameras 35a-35c film the article 2 moving past them line by line, whereby an image of the article 2 is created from the image lines. This allows the creation of images of the underside of the article 2, even if it can only be filmed through a conveyor gap 38, which provides a narrow field of view.

Claims

PATENT CLAIMS 1. Plant with a conveying system (1) for processing articles (2), in particular consumer articles (2), comprising: - a feeding device (11) for feeding articles (2); - an article determination device (13) for determining the requested articles (2); - a forwarding device (15) for forwarding the specific articles (2), and - a further processing device (17) for the respective execution of an operation related to the specific article (2) depending on at least one article information relating to the specific article (2), wherein the article identification device (13) includes a computer system (20) with means for automatically identifying articles (2), characterized in that The article identification device (13) comprises a camera system (30) for creating images of articles (2) and a software application (23) implemented in the computer system (20) for identifying the articles (2), wherein the computer system (20) comprises a database (21) with images of previously identified articles (2), and the software application (23) is designed for identifying the articles (2) by comparing image information of the images taken by the camera system (30) with image information of images stored in the database (21).

2. System according to claim 1, characterized in that the software application (23) is based on a machine learning model which is generated using machine learning algorithms based on data stored in the database The image recording system is trained to determine the articles (2) by comparing image information from the camera system's image recordings (30) with image information from stored image recordings.

3. System according to claim 2, characterized in that the computer system (20) is designed to store the image recordings captured by the camera system (30) in the database (21) and to continuously train the machine learning model with the image recordings captured by the camera system (30) using machine learning algorithms.

4. System according to one of claims 1 to 3, characterized in that the database is a vector database, and the computer system, in particular an AI model implemented in the computer system (20), is designed to convert the image recordings, or image information from the image recordings, into multidimensional vectors and to store the image recordings, or the image information extracted from the image recordings, in the form of vectors in the vector database.

5. System according to one of claims 1 to 4, characterized in that the computer system (20) is designed to assign an identification code to each of the specific articles (2) and to track the specific articles (2) by means of the identification code in a virtual image of the conveyor system (1).

6. Plant according to one of claims 1 to 5, characterized in that a singulation device (12.1) for singulating the articles (2) conveyed by the feeder device (11) as an article stream to the article determination device (13) is arranged in the conveying direction (F) upstream of the article determination device (13).

7. Plant according to one of claims 1 to 6, characterized in that the further processing device (17) forms a plurality of dispensing points (56) for the respective dispensing of a specific article (2) at a dispensing point (56) depending on the at least one article information for the specific article (2).

8. Plant according to one of claims 1 to 7, characterized in that the further processing device (17) is a sorting device (17.1) for sorting the specific articles (2) based on the at least one article information for the specific article (2).

9. Plant according to one of claims 1 to 7, characterized in that the further processing device (17) is a picking device (17.2) for picking the specific articles (2) based on the at least one article information about the specific article (2).

10. Plant according to one of claims 1 to 7, characterized in that the further processing device (17) includes a storage device (17.3) for storing the specific articles (2) in storage locations, the allocation of which is based on at least one article information for the specific article (2).

11. Plant according to one of claims 1 to 7, characterized in that the further processing device (17) is a processing device for carrying out a processing step on the specific article (2) depending on the at least one article information about the specific article (2).

12. Plant according to one of claims 1 to 7, characterized in that the further processing device (17) includes a settlement device (17.6) for settling the amounts determined by the article determination device (13). Article (2) depending on the fact that at least one article information is available for the specific article (2).

13. System according to one of claims 1 to 12, characterized by a weighing device (16) for weighing the articles (2) requested by the requesting device (11).

14. System according to one of claims 1 to 13, characterized in that the system is a cash register system (70) with a billing device (17.6) for billing certain articles (2).

15. System according to one of claims 1 to 14, characterized in that the software application (23) is designed to identify specific defective articles (2) by comparing image information of the images taken by the camera system (30) with image information of images stored in the database (21), and the system in particular includes an ejection device (59) for ejecting defective articles (2).

16. Method for processing articles (2), in particular consumer articles (2), in a system according to any one of claims 1 to 15, characterized in that at least one image of an article (2) is created by means of the camera system (30), and by means of the software application implemented in the computer system (20), based on image data stored in the database (21) relating to specific articles (2), an article identification is carried out by comparing image information of the at least one image taken by the camera system (30) with image information of image data stored in the database (21), and by means of the further processing device (17) an operation related to the specific article (2) is carried out in each case dependency on which at least one article information for the specific article (2) is executed.

17. Method according to claim 16, characterized in that the image recordings of certain articles (2) captured by the camera system (30) are stored in the database (21), and a machine learning model (23) is trained with the stored image recordings using machine learning algorithms.

18. Method according to one of claims 16 to 17, characterized in that the database is a vector database, and image information is extracted from the image recordings by means of an AI model and converted into multidimensional vectors and stored in the vector database.

19. Method according to one of claims 16 to 18, characterized in that the image recordings stored in the database (21) are each linked with at least one article information stored in the database for determining the articles (2).

20. Method according to one of claims 16 to 18, characterized in that each of the specific articles (2) is assigned an identification code and the specific articles (2) are tracked in the virtual image of the conveyor system (1) by means of the identification code.

21. Method according to one of claims 16 to 20, characterized in that by means of the software application (23) one or more of the following image information of the at least one image recording of the article (2) is compared with corresponding image information of image recordings in the database (21): - Form; - Contours; - Dimensions; - Labels; - edges; - Surface color; - Surface structure or texture; - Surface markings; - Surface lettering; - Surface markings; - Figurative marks, word marks, combined marks; - Surface graphics; - Surface patterns; - Number of products in one container.

22. Method according to one of claims 16 to 21, characterized in that the operation relating to the specific article (2) consists of conveying the specific articles (2) to a specific delivery point (56) within the conveying system (1) depending on the at least one article information relating to the specific article and delivering the articles (2) at the delivery point (56).

23. Method according to any one of claims 16 to 22, characterized in that the operation relating to the specific article (2) is a: - Order picking; - Sorting; - Deposit; - Packaging, such as palletizing; - Prepare for shipping; - Send; - Addressing; - Edit or - a settlement of the definite article (2) is.

24. Method according to one of claims 16 to 23, characterized in that the further processing device is a billing device (17.6), and the articles (2) are determined via the article determination device (13) and price information is assigned to the determined articles (2) depending on the at least one article information for the determined articles (2), and the articles (2) are billed electronically individually or as an article compilation by means of the billing device (17.6).

25. Method according to one of claims 16 to 24, characterized in that the articles (2) are automatically weighed by means of the weighing device (16) and the weight of the articles (2) is: - is fed into the software application (23) for article determination and / or - is fed into the billing unit (17.6) for the calculation of a weight-based price for a specific item (2).

26. Method according to one of claims 16 to 25, characterized in that several image recordings of the articles (2) are created from different perspectives, in particular from different sides of the articles, and the image recordings are fed into the software application for article identification.

27. Method according to one of claims 16 to 26, characterized in that, by means of the software application implemented in the computer system (20), based on image recordings of certain articles (2) stored in the database (21), at least one image recording taken by the camera system (30) is compared with image information of image recordings stored in the database (21). Defective articles (2) are identified and removed at an ejection device (59).

28. Plant with a conveying system (1) for processing articles (2), in particular consumer articles (2), comprising: - a feeding device (11) for feeding articles (2); - an article testing device (13) for identifying defective articles (2) and - a diverting device (59) for diverting defective articles (2), characterized by the fact that The article inspection device (13) comprises a computer system (20) and a camera system (30) for creating images of articles (2) and a software application (23) implemented in the computer system (20) for automatically identifying defective articles (2), wherein the computer system (20) comprises a database (21) with images of previously identified articles (2), and the software application (23) is designed to compare image information of the images taken by the camera system (30) with image information of images stored in the database (21) for the purpose of identifying defective articles (2).

29. System according to claim 28, characterized in that the software application (23) is based on a machine learning model which is trained using machine learning algorithms based on image recordings stored in the database in order to identify defective articles (2) by comparing image information of image recordings of the camera system (30) with image information of image recordings stored in the database (21).

30. System according to claim 29, characterized in that the computer system (20) is designed to store image recordings captured by the camera system (30) in the database (21) and to continuously train the machine learning model with the image recordings captured by the camera system (30) using machine learning algorithms.

31. System according to one of claims 28 to 30, characterized in that the database is a vector database, and the computer system, in particular an AI model implemented in the computer system (20), is designed to convert the image recordings into multidimensional vectors and to store the image recordings in the form of vectors in the vector database.

32. System according to one of claims 28 to 31, characterized in that the computer system (20) is designed to assign an identification code to each of the specific articles (2) and to track the defective articles (2) by means of the identification code in a virtual image of the conveyor system (1) until they are removed.

33. Plant according to one of claims 28 to 32, characterized in that a singulation device (12.1) for singulating the articles (2) conveyed by the feed device (11) as an article stream to the article testing device (13) is arranged in the conveying direction (F) upstream of the article testing device (13).

34. Method for processing articles (2), in particular consumer articles (2), in a system according to one of claims 28 to 33, characterized in that at least one image of an article (2) is created by means of the camera system (30), and by means of the software application implemented in the computer system (20), based on image recordings of certain articles (2) stored in the database (21), by comparing image information with that of the camera system (30) Defective articles (2) are identified and removed at the ejection device (59) from the recorded image, at least one image with image information relating to images stored in the database (21).

35. Method according to claim 34, characterized in that the image recordings of defective and / or undamaged articles (2) captured by the camera system (30) are stored in the database (21), and a machine learning model (23) is trained with the stored image recordings using machine learning algorithms.

36. Method according to one of claims 34 to 35, characterized in that the database is a vector database, and image information is extracted from the image recordings by means of an AI model and converted into multidimensional vectors and stored in the vector database.

37. Method according to one of claims 34 to 36, characterized in that the image recordings stored in the database (21) are each linked with at least one article information stored in the database for determining the articles (2).

38. Method according to one of claims 34 to 37, characterized in that each defective article (2) is assigned an identification code and the defective article (2) is tracked in the virtual image of the conveyor system (1) until it is removed by means of the identification code.

39. Method according to one of claims 34 to 38, characterized in that by means of the software application (23) one or more of the following image information of the at least one image recording of the article (2) is compared with corresponding image information of image recordings in the database (21): - Form; - Contours; - Dimensions; - Labels; - edges; - Surface color; - Surface structure; - Surface markings; - Surface lettering; - Surface markings; - Figurative marks, word marks, combined marks; - Surface graphics; - Surface patterns; - Number of products in one container.

40. Method according to one of claims 34 to 39, characterized in that several image recordings of the articles (2) are created from different perspectives, in particular from different sides of the articles, and the image recordings are fed into the software application to identify defective articles (2).

Images