Improved sorting method and system

Abstract

Managing an input item flow (10) comprising a flow of a plurality of items (11, 12, 13) comprises recording a sequence of two-dimensional images of the input item flow (10). Individual items (11, 12, 13) in the images are identified and a respective type of item is determined for each identified item. A sequence of molecular absorption spectra of the input item flow (10) is recorded and a sequence of material indicators is determined, using the spectra. A sequence of information markers of the input item flow (10) is recorded. An association between each identified item (11, 12, 13) and at least one material indicator and at least one information marker is determined. Using the type of item for each item (11, 12, 13) and the association of a material indicator and an information marker with each item, a grading is made of each item into a class. Each item (11, 12, 13) may then be sorted into a respective output item flow (41, 42, 43) based on the grading class.

Description

[0001] IMPROVED SORTING METHOD AND SYSTEM

[0002] Technical field

[0003] The present disclosure relates to a method of managing an input item flow comprising a flow of a plurality of items and a corresponding system The managing of the input item flow may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow.

[0004] Background

[0005] Throughout a wide range of industries identification, detection, classification and sorting of various objects are frequently required and desired.

[0006] In its simplest form, manual identification of objects by a person may be employed to advantage when a limited number of objects are to be identified, sorted and classified. The person in question may then, based on his / her knowledge identify and classify the objects concerned. This type of manual identification is however monotonous and prone to errors. Also, the experience level of the operator will significantly influence the results of the operation performed by the operator. Moreover, manual identification of the above kind suffers from low identification speeds.

[0007] In industry, identification, sorting and classification of bulk objects is therefore often performed by machines where the bulk objects are supplied in form of a continuous object stream. Such machines are generally faster than an operator and can operate for longer periods of time, hence offering an enhanced overall throughput. Machines of this kind are for instance used in agriculture for fruits and vegetables, and in recycling for identifying and sorting objects and materials that are to be recycled.

[0008] Machines of the above kind generally has some form of sensor that is used for identifying the objects of interest. For instance, an optical sensor in form of a spectral sensor may be employed to determine the quality of harvested fruits and vegetables. Similarly, a spectral sensor may be employed to determine the material of objects that are to be recycled.

[0009] In some contexts it is of great importance to separate hazardous items from non-hazardous items, for example separation of potentially poisonous items in a context where food-grade products are to be sorted from items that are not of food-grade. An example of such a context is the sorting of household waste that comprises food containers that are to be recycled. Needless to say, sorting of household waste entails sorting of enormous amounts of items and an associated requirement with such sorting is thus sorting speed. It is therefore desirable to enable sorting of such items in a single sorting procedure, avoiding a need for subjecting items of two or more sequential sorting procedures. Prior art sorting systems and methods do not provide an optimal solution to such a combined need of sorting accuracy in terms of hazardous and non-hazardous items with sorting speed.

[0010] Summary

[0011] An object is to overcome the drawbacks as summarized above and according to a first aspect there is provided a method of managing an input item flow, said input item flow comprising a flow of a plurality of items. The managing of the input item flow may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow. The method comprises recording a sequence of two-dimensional images of the input item flow. Individual items in the sequence of two-dimensional images are identified and, using the sequence of two-dimensional images, a respective type of item is determined for each identified item. A sequence of molecular absorption spectra of the input item flow is recorded and a sequence of material indicators is determined, using the sequence of molecular absorption spectra. A sequence of information markers of the input item flow is recorded.

[0012] An association between each identified item and at least one material indicator and at least one information marker is determined. Using the respective determined type of item for each identified item and the association of at least one material indicator and at least one information marker with each identified item, a grading is made of each identified item into one class among a plurality of classes.

[0013] Each identified item in the input item flow may then, in some embodiments, be sorted into a respective output item flow based on the grading class determined for each respective identified item. Such a method thus performs a single-stage procedure of determining what kind of items are received in the input item flow and thereby enabling accurate and fast sorting of the items. By utilizing three parameters, i.e. types of items determined from two-dimensional images, the material of the items from molecular spectra and a third parameter obtained by recording information markers on the items, accuracy in the grading into grading classes is obtained without requiring one or more further sorting procedures. The method thus provides an efficient way of sorting items such as food grade containers in a flow of items comprising nonfood grade items.

[0014] The determination of a grading of each identified item into one class among a plurality of classes may comprise determining that each identified item is any of a food item, a non-food item, a hazardous material item, a non-hazardous material item.

[0015] That is, in a context where it is necessary to sort items into more or less healthy and non-healthy grading classes it is typically vital that the accuracy of the sorting procedure is high. Such a context may be one wherein food grade items are to be separated from non-food grade items. Due to strict regulatory requirements on recycling of food-grade items, it remains a most complex recycling task, in particular, to obtain from a high throughput sorting and recycling process such food-grade items because of the risks associated with recycled items potentially containing toxic chemicals that could be dangerous to human health.

[0016] The recording of a sequence of information markers may comprise recording information encoded by a spatial pattern on at least some items in the plurality of items. For example, the recording of a sequence of information markers may comprise recording any one or more of infrared radiation, ultra violet (UV) light, fluorescence light, a barcode, a quick-response (QR) code, serialization codes, a radio-frequency identification (RFID) signal, a Near-field communication (NFC) signal, chemical encoders or codes, electromagnetic encoders or codes, and density-based encoders or codes.

[0017] In other words, in addition to performing imaging and spectroscopy of the items in the input item flow, the third parameter may be in a form that provides detailed information relating to the items, not only on a “collective” level relating to many similar or identical items, but also on an “individual” level where each item may be uniquely identified. It is known to those skilled in the art that electromagnetic radiation visible to the human eye constitutes a narrow bandwidth within the electromagnetic spectrum, specifically spanning wavelengths between 380 and 780 nanometres. However, sensor technology advanced to the acquisition of images across a much wider range of wavelengths, including both multi-spectral and hyperspectral images and commonly referred to as multi- spectral imaging (MSI) and hyperspectral imaging (HIS) or collectively as imaging spectroscopy. In imaging spectroscopy, e.g., in HSI, the spectrum for each pixel is measured at different wavelengths and to provide more information on observed items, each pixel is broken down into many different spectral bands. Typically, imaging spectrometers operate in the 0.4 to 5 pm wavelength range, capturing the visible and solar-reflected infrared spectrum (i.e., near-infrared or NIR, and shortwavelength infrared or SWIR) from the observed items.

[0018] In some embodiments, the recording of a sequence of information markers may comprise obtaining the spatial pattern from the recorded two-dimensional images. For example, the recording of the sequence of information markers may comprise recording any one or more of a barcode, a QR code, serialization codes, and color codes.

[0019] That is, in such embodiments the third parameter is in a form that is possible to identify in two-dimensional image, typically in a visual wavelength range. The images of the items in the flow of items are thus used, in addition to the determination of object type, also for identifying and providing the third parameter. It will thus be appreciated by those skilled in the art that imaging spectroscopy often results in a multi-dimensional and large data set, which requires image processing techniques for extracting and analysing useful information. Image processing techniques have evolved to a level of computational efficiency that, if used in the present context, will optimize the accuracy and speed of the grading into grading classes and thereby further enable an accurate and fast sorting of the items. In some embodiments, the recording of a sequence of information markers may comprise recording temperature and in some embodiments, the recording of a sequence of information markers may comprise recording water content.

[0020] By obtaining information about temperature and / or water content it is possible to detect Refuse derived fuel (RDF) in a waste stream.

[0021] It is known to those skilled in the art that RDF is an alternative solid fuel which could be derived from domestic or industrial solid waste. The demand for RDF is growing worldwide to reduce dependency on fossil fuels and landfills. The production requires a high recovery rate for specific materials to ensure a consistent calorific value and high-quality end-product. Detecting and analysing waste for overall combustibility is thus a means of generating a source of alternative fuels from various waste streams. E.g., a high throughput screening unit capable of continuously monitoring critical values such as calorific, chlorine and water content of the waste stream enables alternative fuel preparation.

[0022] Embodiments of the method may be implemented in various ways, using processing and control devices, for example in a way where specific algorithms are performed in sequence and / or in parallel. Thus, in various embodiments, the identification of individual items may be performed using an item identification algorithm on the sequence of two-dimensional images, outputting an item identification data structure. The determination of a respective type of item for each identified item may be performed using a type determination algorithm on the item identification data structure, outputting an item type data structure. The determination of a sequence of material indicators may be performed using a material determination algorithm on the sequence of molecular absorption spectra, outputting an item material data structure. The determination of an association between each identified item and at least one material indicator and at least one information marker may be performed using an item material association algorithm, outputting an item material association data structure, and the determination of a grading of each identified item into one class among a plurality of classes may be performed using a grading algorithm, inputting the item type data structure, the item material data structure and the item material association data structure and outputting a grading data structure configured for use by the sorting of each identified item in the input item flow into a respective output item flow.

[0023] In other embodiments, the method may be implemented in various ways, using processing and control devices, for example in a way where artificial intelligence (Al) and machine learning (ML) tools are employed. Thus, in various embodiments, the identification of individual items, the determination of a respective type of item for each identified item, the determination of a sequence of material indicators, the determination of an association between each identified item and at least one material indicator and at least one information marker, and the determination of a grading of each identified item into one class among a plurality of classes are performed by a neural network trained with training data comprising recorded sequences of two-dimensional images of input item flows similar to the item flow, recorded sequences of molecular absorption spectra of input item flows similar to the item flow and recorded sequences of information markers of input item flows similar to the item flow.

[0024] Implementations where specific algorithms are performed in sequence and / or in parallel may advantageously be utilized when a limited amount of training data is available.

[0025] However, to unlock the full potential of Al and ML in sorting and recycling processes and systems, implementations where Al and ML tools are employed may advantageously be utilized where in sorting tasks that cannot be solved with conventional sorting methods. Typically, deep learning technology extracts numerous material characteristics of individual objects to classify them for sorting. For example, deep learning-based applications for waste wood, PE-silicone cartridges and other materials set new performance standards for objects once considered difficult to sort. Furthermore, artificial neural networks are trained with thousands of images to create a pool of information that improves performance. Delivering high-purity and user-defined fractions, this innovation can upgrade material output and create new revenue streams. In fact, design of a system may be done automatically through using artificial neural network technology. It will be appreciated by those skilled in the art that although the initial training effort is higher, such a system design with automated training through the use of artificial neural networks is capable of extending the scope of what can be detected. Such a system has the further advantage of unifying to one neural network and achieve an end-to-end network solution. For example, in the event that a sorting task dictates the identifying of objects with a high visual contamination level into different grades such as food- and non-food grades, the method and system according to the invention will achieve this in a single sorting procedure, avoiding a need for subjecting items of two or more sequential sorting procedures in different sorting systems.

[0026] In a second aspect there is provided a system for managing an input item flow, said input item flow comprising a flow of a plurality of items. The managing of the input item flow may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow. The system of the second aspect comprises an imaging system configured to record a sequence of two- dimensional images of the input item flow, a spectrometer system configured to record a sequence of molecular absorption spectra of the input item flow, a recording system configured to record a sequence of information markers of the input item flow, and a controller. The controller is configured to identify individual items in the sequence of two-dimensional images, determine a respective type of item for each identified item, using the sequence of two-dimensional images. The controller is further configured to determine a sequence of material indicators, using the sequence of molecular absorption spectra, determine an association between each identified item and at least one material indicator and at least one information marker, and determine, using the respective determined type of item for each identified item and the association of at least one material indicator and at least one information marker with each identified item, a grading of each identified item into one class among a plurality of classes.

[0027] The system of the second aspect may, in some embodiments, further comprise a sorting arrangement configured to be controlled by the controller and sort each identified item in the input item flow into a respective output item flow based on the grading class determined by the controller for each respective identified item. Such a system is thus configured to perform a single-stage procedure of determining what kind of items are received in the input item flow and thereby enabling accurate and fast sorting of the items. By utilizing three parameters, i.e. types of items determined from two-dimensional images, the material of the items from molecular spectra and a third parameter obtained by recording information markers on the items, accuracy in the grading into grading classes is obtained without requiring one or more further sorting procedures.

[0028] In various embodiments of the system according to the second aspect, the recording system configured to record a sequence of information markers may comprise any one or more of an infrared thermometer, a micro-bolometer, a thermal camera, a color camera, a barcode reader, a QR code reader, an RFID reader, an NFC reader, an UV light detector, and / or fluorescence detector, a spectroscopy detector, an electromagnetic detector and / or and electromagnetic inductor, and a dual energy X-ray detector.

[0029] In a third aspect there is provided a system for managing an input item flow, said input item flow comprising a flow of a plurality of items. The managing of the input item flow may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow. The system of the third aspect comprises an imaging system configured to record a sequence of two-dimensional images of the input item flow, a spectrometer system configured to record a sequence of molecular absorption spectra of the input item flow, and a controller. The controller is configured to identify individual items in the sequence of two- dimensional images, determine a respective type of item for each identified item, using the sequence of two-dimensional images and determine a sequence of information markers by obtaining a spatial pattern from the recorded two- dimensional images, and determine an association between each identified item and at least one material indicator and at least one information marker.

[0030] The controller is further configured to determine a sequence of material indicators, using the sequence of molecular absorption spectra, determine, using the respective determined type of item for each identified item and the association of at least one material indicator and at least one information marker with each identified item, a grading of each identified item into one class among a plurality of classes.

[0031] The system of the third aspect may, in some embodiments, further comprise a sorting arrangement configured to be controlled by the controller and sort each identified item in the input item flow into a respective output item flow based on the grading class determined by the controller for each respective identified item.

[0032] Such a system is thus configured to perform a single-stage procedure of determining what kind of items are received in the input item flow and thereby enabling accurate and fast sorting of the items. By utilizing three parameters, i.e. types of items determined from two-dimensional images, the material of the items from molecular spectra and a third parameter obtained by recording information markers on the items, accuracy in the grading into grading classes is obtained without requiring one or more further sorting procedures. Moreover, such a system has an advantage in that it is capable of obtaining third parameter information via the images obtained by means of the imaging system. The system according to the third aspect thus overcomes the drawbacks of the prior art without the need for hardware devices for obtaining the third parameter information and thereby reducing the complexity and cost of the system.

[0033] In a fourth aspect there is provided a non-transitory computer-readable storage medium having stored thereon instructions for implementing the method as summarized above, when executed on a device having processing capabilities.

[0034] Such a non-transitory computer-readable storage medium has corresponding effects and advantages as the method of the first aspect and the systems of the second and third aspects.

[0035] Brief description of the drawings

[0036] The above and other aspects of the present invention will now be described in more detail, with reference to appended figures. The figures should not be considered limiting; instead, they are used for explaining and understanding. Like reference numerals refer to like elements throughout.

[0037] Figure 1a schematically illustrates a system for sorting a plurality of items, figure 1 b schematically illustrates a system for sorting a plurality of items, and figure 2 is a flow chart of a method, figure 3 schematically illustrates a method in a neural network, figure 4a schematically illustrates a sequence of two-dimensional images, figure 4b schematically illustrates a sequence of molecular absorption spectra, and figure 4c schematically illustrates items with information markers.

[0038] Detailed description

[0039] The present invention will now be described more fully hereinafter with reference to the accompanying drawings, in which currently preferred embodiments of the invention are shown. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided for thoroughness and completeness, and to fully convey the scope of the invention to the skilled person.

[0040] Reference will now be made to figure 1a and figures 4a-c, which illustrates an embodiment of a system 1 for managing an input item flow 10, said input item flow 10 comprising a flow of a plurality items 11 , 12, 13. As will be described below, the managing of the input item flow 10 may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow 10. The system 1 comprises an imaging system 41 configured to record a sequence of two-dimensional images 120 of the input item flow 10, a spectrometer system 42 configured to record a sequence of molecular absorption spectra 130 of the input item flow 10, a recording system 43 configured to record a sequence of information markers 140 of the input item flow 10, and a controller 40.

[0041] The controller 40 typically comprises processing circuitry, memory circuitry and signal input / output circuitry connected to the imaging system 41 , the spectrometer system 42 and the recording system 43 as well as a sorting arrangement 22.

[0042] The controller 40 is configured to identify, in the sequence of two- dimensional images 120, individual items 11 , 12, 13, determine, using the sequence of two-dimensional images 120, a respective type of item for each identified item 11 , 12, 13, determine, using the sequence of molecular absorption spectra 130, a sequence of material indicators, determine an association between each identified item 11 , 12, 13 and at least one material indicator and at least one information marker 140, and determine, using the respective determined type of item for each identified item 11 , 12, 13 and the association of at least one material indicator and at least one information marker 140 with each identified item 11 , 12, 13, a grading of each identified item 11 , 12, 13 into one class among a plurality of classes.

[0043] An optional sorting arrangement 22 is configured to be controlled by the controller 40 and sort each identified item 11 , 12, 13 in the input item flow 10 into a respective output item flow 41 , 42, 43 based on the grading class determined by the controller 40 for each respective identified item 11 , 12, 13. The sorting arrangement 22 may thus be configured to eject jets 23 of, e.g., a gas such as air or any other suitable fluid, the jets 23 being controlled in terms of, e.g., spatial focusing and flow in order to push the items 11 , 12, 13 along respective paths 41 , 42, 43 into sorting containers 31 , 32, 33 where each sorting container 31 , 32, 33 is located and configured such that they receive a respective class of items graded as described above and pushed by a respective jet 23.

[0044] The imaging system 41 may be in the form of a high resolution camera configured to operate in the visual wavelength range. For example, the imaging system 41 may thus be configured to record images having a resolution of at least 12 megapixels and being sensitive to light in a wavelength range from about 380 to about 750 nanometers.

[0045] The spectrometer system 42 may comprise a light source and be configured to receive and analyse light which is reflected and / or scattered by the items 11 , 12, 13. The spectrometer system 42 may comprise a spectrometer manufactured by Tomra which is able to cope with the required rate of the item flow 10 and the characteristics of the items 11 , 12, 13 to be analysed.. The spectrometer may thus be configured to analyse light in the wavelength interval 400 - 1000 nm. The spectrometer may be configured to analyse light in the wavelength interval 500 - 1000 nm. The spectrometer may be configured to analyse light in the wavelength interval 1000 - 1900 nm. The spectrometer may be configured to analyse light having a wavelength above 900 nm. The spectrometer may be configured to analyse light in the wavelength interval 1900 -2500 nm. The spectrometer may be configured to analyse light in the wavelength interval 2700 - 5300 nm. The spectrometer may be configured to analyse light in the wavelength interval 900 1700 nm. The spectrometer may be configured to analyse light in the wavelength interval 700-1400 nm. The spectrometer may analyse visible light. The spectrometer may analyse near-infrared (NIR) light. The spectrometer may analyse infrared (IR) light.

[0046] The recording system 43 configured to record a sequence of information markers 140 may comprise any one or more of an infrared thermometer, a microbolometer, a thermal camera, a color camera, a barcode reader, a QR code reader, an radio RFID reader, an NFC reader an UV light detector, and / or fluorescence detector, a spectroscopy detector, an electromagnetic detector and / or and electromagnetic inductor, and a dual energy X-ray detector.

[0047] An infrared thermometer and a micro-bolometer may be configured to record thermal radiation emitted from the items 11 , 12, 13 and thereby provide information regarding, e.g., temperature and water content of the items 11 , 12, 13.

[0048] A barcode reader and a QR code reader are capable of recording information printed on the items 11 , 12, 13 that is of textual and numerical character.

[0049] Similarly, an RFID reader and an NFC reader are capable of recording information encoded in RFID and NFC circuitry embedded in the items 11 , 12, 13 that is of textual and numerical character.

[0050] An UV light detector may be configured with detecting circuitry that is capable of recording information printed on the items 11 , 12, 13, such printed information being invisible to the naked eye and therefore less distracting in comparison with, e.g. barcodes and QR codes.

[0051] Turning now to figure 1 b and with continued reference to figures 4a-c, an embodiment will be described of a system 2 for managing an input item flow 10, said input item flow 10 comprising a flow of a plurality items 11 , 12, 13. The managing of the input item flow 10 may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow 10. The system 2 comprises an imaging system 41 configured to record a sequence of two- dimensional images 120 of the input item flow 10, a spectrometer system 42 configured to record a sequence of molecular absorption spectra 130 of the input item flow 10, and a controller 40.

[0052] The controller 40 typically comprises processing circuitry, memory circuitry and signal input / output circuitry connected to the imaging system 41 , the spectrometer system 42 and a sorting arrangement 22.

[0053] The controller 40 is configured to identify, in the sequence of two- dimensional images 120, individual items 11 , 12, 13, and determine, using the sequence of two-dimensional images 120, a respective type of item for each identified item 11 , 12, 13. The controller 40 is further configured to determine a sequence of information markers 140 by obtaining a spatial pattern from the recorded two-dimensional images 120, determine, using the sequence of molecular absorption spectra 130, a sequence of material indicators, and determine an association between each identified item 11 , 12, 13 and at least one material indicator and at least one information marker 140.

[0054] The controller is further configured to determine, using the respective determined type of item for each identified item 11 , 12, 13 and the association of at least one material indicator and at least one information marker 140 with each identified item 11 , 12, 13, a grading of each identified item 11 , 12, 13 into one class among a plurality of classes.

[0055] An optional sorting arrangement 22 is configured to be controlled by the controller 40 and sort each identified item 11 , 12, 13 in the input item flow 10 into a respective output item flow 41 , 42, 43 based on the grading class determined by the controller 40 for each respective identified item 11 , 12, 13.

[0056] The imaging system 41 and the spectrometer system 42 may be configured in more detail as described above in connection with figure 1a. With regard to the controller 40 in the system 2 of figure 1 b, in addition to the configuration described above, it may be configured to analyse spatial pattern from the recorded two- dimensional images 120 such as barcodes, QR codes and other spatial information printed on the items 11 , 12, 13. Turning now to figure 2, and with continued reference to figures 1a-b and figures 4a-c, a method of managing an input item flow 10, said input item flow 10 comprising a flow of a plurality items 11 , 12, 13, will be described in some detail. The managing of the input item flow 10 may involve management in terms of detecting, analysing, grading and / or sorting items in the input item flow 10. The method, performed by a system 1 , 2 as exemplified in figures 1a and 1 b, comprises a recording step S205 wherein a sequence of two-dimensional images 120 of the input item flow 10 is recorded.

[0057] An identifying step S210 comprises identification of individual items 11 , 12, 13 in the sequence of two-dimensional images 120.

[0058] A determining step S215 comprises determining a respective type of item for each identified item 11 , 12, 13, using the sequence of two-dimensional images 120.

[0059] A recording step S220 comprises recording a sequence of molecular absorption spectra 130 of the input item flow 10.

[0060] A determining step S225 comprises determining a sequence of material indicators, using the sequence of molecular absorption spectra 130.

[0061] A recording step S230 comprises recording a sequence of information markers 140 of the input item flow 10.

[0062] A determining step S235 comprises determining an association between each identified item 11 , 12, 13 and at least one material indicator and at least one information marker 140.

[0063] A determining step S240 comprises determining a grading of each identified item 11 , 12, 13 into one class among a plurality of classes, using the respective determined type of item for each identified item 11 , 12, 13 and the association of at least one material indicator and at least one information marker 140 with each identified item 11 , 12, 13.

[0064] An optional sorting step S245 comprises sorting each identified item 11 , 12, 13 in the input item flow 10 into a respective output item flow 41 , 42, 43 based on the grading class determined for each respective identified item 11 , 12, 13.

[0065] The determination, in the determining step S235, of a grading of each identified item 11 , 12, 13 into one class among a plurality of classes may comprise determining that each identified item 11 , 12, 13 is any of a food item, a non-food item, a hazardous material item and a non-hazardous material item.

[0066] The recording, in the recording step S215, of a sequence of information markers 140 may comprise recording information encoded by a spatial pattern on at least some items in the plurality of items 11 , 12, 13. For example, the recording, in recording step S215, of a sequence of information markers 140 may comprise recording any one or more of:

[0067] - infrared radiation,

[0068] - UV light and / or fluorescence light,

[0069] - a barcode,

[0070] - a QR code,

[0071] - serialization codes,

[0072] - an RFID signal,

[0073] - an NFC signal, color codes,

[0074] - chemical encoders or codes,

[0075] - electromagnetic encoders or codes, and

[0076] - density-based encoders or codes.

[0077] In some embodiments, the recording, in recording step S215, of a sequence of information markers 140 may comprise obtaining the spatial pattern from the recorded two-dimensional images 120. In addition to, e.g., barcodes and QR codes, information markers that may be obtained from two-dimensional images 120 include colors and color codes, serialization codes, textures, pictures and symbols such as logos etc.

[0078] In some embodiments, the recording, in recording step S215, of a sequence of information markers 140 may comprise recording temperature and / or recording water content.

[0079] As exemplified in figure 2, the method may be performed under the control of the controller 40 by way of instructions for implementing the method stored on a non-transitory computer-readable storage medium 41. Such implementing instructions may be in the form of dedicated algorithms where the identification in identifying step S210 of individual items 11 , 12, 13 is performed using an item identification algorithm on the sequence of two-dimensional images 120, outputting an item identification data structure, the determination in determining step S215 of a respective type of item for each identified item 11 , 12, 13 is performed using a type determination algorithm on the item identification data structure, outputting an item type data structure, the determination in determining step S225 of a sequence of material indicators is performed using a material determination algorithm on the sequence of molecular absorption spectra 130, outputting an item material data structure, the determination in determining step S235 of an association between each identified item 11 , 12, 13 and at least one material indicator and at least one information marker 140 is performed using an item material association algorithm, outputting an item material association data structure, and the determination in determining step S240 of a grading of each identified item 11 , 12, 13 into one class among a plurality of classes is performed using a grading algorithm, inputting the item type data structure, the item material data structure and the item material association data structure and outputting a grading data structure configured for use by the sorting S245 of each identified item 11 , 12, 13 in the input item flow 10 into a respective output item flow 41 , 42, 43.

[0080] Alternatively, referring to figure 3, and with continued reference to figures 1a-b and figures 4a-c, the method may be performed under the control of the controller 40 by way of implementing instructions that result from a deep learning / machine learning (ML) artificial intelligence (Al) process wherein the identification S210 of individual items 11 , 12, 13, the determination S215 of a respective type of item for each identified item 11 , 12, 13, the determination S225 of a sequence of material indicators, the determination S235 of an association between each identified item 11 , 12, 13 and at least one material indicator and at least one information marker 140, and the determination S240 of a grading of each identified item 11 , 12, 13 into one class among a plurality of classes are performed by a neural network 300 trained with training data comprising recorded sequences of two-dimensional images of input item flows similar to the item flow 10, recorded sequences of molecular absorption spectra of input item flows 10 similar to the item flow 10 and recorded sequences of information markers of input item flows similar to the item flow 10. The person skilled in the art realizes that the present invention by no means is limited to the preferred embodiments described above. On the contrary, many modifications and variations are possible within the scope of the appended claims.

[0081] Additionally, variations to the disclosed embodiments can be understood and effected by the skilled person in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.

Claims

CLAIMS1. A method of managing an input item flow (10), said input item flow (10) comprising a flow of a plurality items (11 , 12, 13), the method comprising:- recording (S205) a sequence of two-dimensional images (120) of the input item flow (10), identifying (S210), in the sequence of two-dimensional images (120), individual items (11 , 12, 13), determining (S215), using the sequence of two-dimensional images (120), a respective type of item for each identified item (11 , 12, 13), recording (S220) a sequence of molecular absorption spectra (130) of the input item flow (10), determining (S225), using the sequence of molecular absorption spectra (130), a sequence of material indicators,- recording (S230) a sequence of information markers (140) of the input item flow (10), determining (S235) an association between each identified item (11 , 12, 13) and at least one material indicator and at least one information marker (140), and determining (S240), using the respective determined type of item for each identified item (11 , 12, 13) and the association of at least one material indicator and at least one information marker (140) with each identified item (11 , 12, 13), a grading of each identified item (11 , 12, 13) into one class among a plurality of classes.

2. The method according to claim 1 , comprising:- sorting (S245) each identified item (11 , 12, 13) in the input item flow (10) into a respective output item flow (41 , 42, 43) based on the grading class determined for each respective identified item (11 , 12, 13).

3. The method according to claim 1 or claim 2, wherein the determination (S235) of a grading of each identified item (11 , 12, 13) into one class among a plurality of classes comprises determining that each identified item (11 , 12, 13) is any of:- a food item,- a non-food item,- a hazardous material item,- a non-hazardous material item.

4. The method according to any one of claims 1 to 3, wherein the recording (S215) of a sequence of information markers (140) comprises recording information encoded by a spatial pattern on at least some items in the plurality of items (11 , 12, 13).

5. The method according to claim 4, wherein the recording (S215) of a sequence of information markers (140) comprises recording any one or more of:- infrared radiation,- ultra violet, UV, light and / or fluorescence light,- a barcode,- a quick-response QR code,- serialization codes,- a radio-frequency identification RFID signal,- a Near-field communication NFC signal,- color codes,- chemical encoders or codes,- electromagnetic encoders or codes, and- density-based encoders or codes.

6. The method according to claim 4, wherein the recording (S215) of a sequence of information markers (140) comprises obtaining the spatial pattern from the recorded two-dimensional images (120).

7. The method according to claim 6, wherein the recording (S215) of a sequence of information markers (140) comprises recording any one or more of:- a barcode,- a quick-response, QR, code- serialization codes, and- color codes.

8. The method according to any one of claims 1 to 3, wherein the recording (S215) of a sequence of information markers (140) comprises recording temperature.

9. The method according to any one of claims 1 to 3, wherein the recording (S215) of a sequence of information markers (140) comprises recording water content.

10. The method according to any one of claims 1 to 9, wherein:- the identification (S210) of individual items (11 , 12, 13) is performed using an item identification algorithm on the sequence of two-dimensional images (120), outputting an item identification data structure,- the determination (S215) of a respective type of item for each identified item (11 , 12, 13) is performed using a type determination algorithm on the item identification data structure, outputting an item type data structure,- the determination (S225) of a sequence of material indicators is performed using a material determination algorithm on the sequence of molecular absorption spectra (130), outputting an item material data structure,- the determination (S235) of an association between each identified item (11 , 12, 13) and at least one material indicator and at least one information marker (140) is performed using an item material association algorithm, outputting an item material association data structure, and- the determination (S240) of a grading of each identified item (11 , 12, 13) into one class among a plurality of classes is performed using a grading algorithm, inputting the item type data structure, the item material data structure and the item material association data structure and outputting a grading data structure configured for use by the sorting (S245) of each identified item (11 , 12, 13) in the input item flow (10) into a respective output item flow (41 , 42, 43).11 . The method according to any one of claims 1 to 7, wherein:- the identification (S210) of individual items (11 , 12, 13),- the determination (S215) of a respective type of item for each identified item (11 , 12, 13),- the determination (S225) of a sequence of material indicators,- the determination (S235) of an association between each identified item (11 ,12. 13) and at least one material indicator and at least one information marker(140), and- the determination (S240) of a grading of each identified item (11 , 12, 13) into one class among a plurality of classes are performed by a neural network (300) trained with training data comprising recorded sequences of two-dimensional images of input item flows similar to the item flow (10), recorded sequences of molecular absorption spectra of input item flows (10) similar to the item flow (10) and recorded sequences of information markers of input item flows similar to the item flow (10).

12. A system (1) for managing an input item flow (10), said input item flow (10) comprising a flow of a plurality items (11 , 12, 13), the system (1 ) comprising:- an imaging system (41) configured to record a sequence of two-dimensional images (120) of the input item flow (10),- a spectrometer system (42) configured to record a sequence of molecular absorption spectra (130) of the input item flow (10),- a recording system (43) configured to record a sequence of information markers (140) of the input item flow (10), and- a controller (40) configured to: identify, in the sequence of two-dimensional images (120), individual items (11 , 12, 13), determine, using the sequence of two-dimensional images (120), a respective type of item for each identified item (11 , 12, 13), determine, using the sequence of molecular absorption spectra (130), a sequence of material indicators, determine an association between each identified item (11 , 12, 13) and at least one material indicator and at least one information marker (140), determine, using the respective determined type of item for each identified item (11 , 12, 13) and the association of at least one material indicator and at least one information marker (140) with each identified item (11 , 12, 13), a grading of each identified item (11 , 12, 13) into one class among a plurality of classes.

13. The system (1) according to claim 12, wherein the recording system (43) configured to record a sequence of information markers (140) comprises any oneor more of:- an infrared thermometer,- a micro-bolometer,- a thermal camera,- a color camera,- a barcode reader,- a quick-response, QR, code reader,- a radio-frequency identification RFID reader,- a Near-field communication NFC reader,- an ultra violet, UV, light detector and / or fluorescence detector,- a spectroscopy detector,- an electromagnetic detector and / or and electromagnetic inductor, and- a dual energy X-ray detector.

14. A system (2) for managing an input item flow (10), said input item flow (10) comprising a flow of a plurality items (11 , 12, 13), the system (1 ) comprising:- an imaging system (41 ) configured to record a sequence of two-dimensional images (120) of the input item flow (10),- a spectrometer system (42) configured to record a sequence of molecular absorption spectra (130) of the input item flow (10),- a controller (40) configured to: identify, in the sequence of two-dimensional images (120), individual items (11 , 12, 13), determine, using the sequence of two-dimensional images (120), a respective type of item for each identified item (11 , 12, 13), determine a sequence of information markers (140) by obtaining a spatial pattern from the recorded two-dimensional images (120), determine, using the sequence of molecular absorption spectra (130), a sequence of material indicators, determine an association between each identified item (11 , 12, 13) and at least one material indicator and at least one information marker (140), determine, using the respective determined type of item for each identified item (11 , 12, 13) and the association of at least one material indicator and at leastone information marker (140) with each identified item (11 , 12, 13), a grading of each identified item (11 , 12, 13) into one class among a plurality of classes.

15. The system (1 , 2) according to any one of claims 12 to 14, comprising:- a sorting arrangement (22) configured to be controlled by the controller (40) and sort each identified item (11 , 12, 13) in the input item flow (10) into a respective output item flow (41 , 42, 43) based on the grading class determined by the controller (40) for each respective identified item (11 , 12, 13).

16. A non-transitory computer-readable storage medium (41 ) having stored thereon instructions for implementing the method according to any one of claims 1 to 11 , when executed on a device having processing capabilities.

Images