Method for sorting waste combining analysis of chemical composition and use of markers

Abstract

The invention relates to a method for sorting objects, comprising the steps of: depositing objects on a conveyor belt (BCV); acquiring images of a first zone (Z1) of the belt; acquiring detection signals indicative of the detection of a material and / or colour for adjacent cells (CL) of a second zone (Z2) of the belt; processing the detection signals to determine detected attributes; detecting and decoding markers in the images to determine specified attributes and identifiers; determining a first setpoint for each cell associated with one object identifier; determining a second setpoint for each cell associated with material and / or colour attributes; and determining a combined sorting decision depending on the first and second sorting setpoints of the cell and on a set of combination rules.

Description

[0001] DESCRIPTION

[0002] TITLE: Waste sorting process combining chemical composition analysis and the use of markers

[0003] Technical field

[0004] The present invention relates to the field of sorting waste, in particular of the plastic or fibrous packaging type, on the basis of different criteria such as material, color, shape, use, viscosity, and the presence of carbon of biological origin and recycled material.

[0005] State of the art

[0006] Generally, waste sorting relies on different technologies for direct detection of the chemical composition of the objects to be sorted. These technologies use different detection tools that are often combined together for a more detailed diagnosis. These tools may, for example, include spectrometers in the near or mid-infrared range for the molecular analysis of organic products (plastics, cellulosic or biological products), color cameras or spectrometers in the visible range for analyzing the color of objects; inductive detectors characterizing the presence of metals, X-ray detectors or laser-induced breakdown spectroscopy (LIBS) for analyzing atomic composition.

[0007] There are also image processing systems that include shape recognition algorithms, particularly those using artificial intelligence. However, such systems are not well-suited to processing waste that may be highly degraded, for example, deformed and soiled.

[0008] Despite the existence of a wide variety of detection tools, it is not always possible to determine other properties important for recycling, such as the food or non-food use of the object, the presence of hazardous content, the presence of a layer partially or completely masking the object (case of multi-layer packaging), the type and shape of the object, the presence of additives in the material constituting the object, the viscosity index, particularly in the case where the object is made of HDPE (high-density polyethylene), the rate of recycled material and the proportion of biogenic carbon contained in the object.

[0009] It has already been proposed to add chemical tracers to the objects to be sorted to characterize a desired property. This therefore allows indirect detection of the property based on a declarative principle. However, the use of such tracers has drawbacks. Indeed, the meaning of such tracers requires the approval of an agreement between the actors concerned. The association between tracers and objects raises risks of error or fraud. The coding capacity is limited, typically 1 bit of information for each tracer. Tracers must ultimately be removed from the objects during recycling to avoid compromising the next sorting / recycling cycle.

[0010] We also know the international barcode system, which assigns a unique identifier to each commercial packaging reference. This identifier provides access to all the product's technical data, and therefore potentially to all the attributes relevant to recycling. However, reading a barcode requires orienting the barcode in front of a reading device. This marking technique is therefore not suitable for the automatic sorting of waste presented randomly on a conveyor. A new version of barcodes currently being standardized, based on QR codes, has the same disadvantage of being present in only one location on the packaging.

[0011] RFID chip marking systems are also known, which can assign a unique identifier to each product. A global reading is possible, without orienting the product. However, RFID chips are significantly expensive compared to the cost of packaging, and do not allow the object to be precisely located on the sorting belt.

[0012] In the 2010s, a new type of marker was proposed for packaging. These are digital watermarks, referred to as "markers" in the following. They consist of adding a printed or embossed (molded or engraved) pattern to a label or the packaging of the object. Whether printed or engraved, these patterns have a very high resolution, in the order of 75 to 150 dpi, or pixels measuring 0.1 to 0.2 mm on each side. The variation in the grayscale of a print or the depth of the relief is calculated so that the marker is imperceptible to the human eye, but is systematically seen by a fast camera with appropriate lighting (pulsed or not, monochrome or white).

[0013] The European consortium Holy Grail 2.0 launched a program in 2020 to industrially validate the use of such markers for packaging in Europe. This program is based on technologies developed by industrial players such as Digimarc® and Filigrade®.

[0014] Compared to chemical tracers, markers do not require any external material input, and do not leave any trace during the recycling phase. In addition, their information richness is incomparable: a pattern measuring 1 to 2 cm on each side can contain 10 to 50 bits of information depending on the technological solutions chosen. In comparison, systems with three different chemical tracers, equivalent to 3 bits, can only code seven different situations of objects thus marked (eight combinations are possible, but the total absence, coded 0,0,0, does not allow to distinguish marked and unmarked objects), which is largely insufficient given the number of types of objects to be marked. The information richness of markers is therefore comparable to that of a bar code.Unlike barcodes, these markers are formed in many places on the surface of the object, or even on its entire surface, so that they can be read automatically without having to present a specific face of the object to the reading device.

[0015] They therefore remain visible regardless of the orientation and condition of the object, even when crushed and partially masked, or even soiled. These properties make it possible to consider the use of such markers for optical sorting before recycling.

[0016] Summary

[0017] The present invention proposes to produce a detection tunnel based on a row of cameras, placed above a sorting belt, and which is combined with a conventional optical sorting system, using physical sensors such as spectrometers in the near infrared (N IR) and / or visible ranges, to determine the composition of the material.

[0018] Thus, the present invention provides a method for sorting marked and unmarked objects in the same operation, combining direct detection and indirect detection by the detection of markers providing information not detectable by direct detection. Indeed, it is desirable to be able to sort objects such as packaging to be recycled on the basis of complex combinations of criteria such as material, color, shape, use, viscosity, and the presence of carbon of biological origin and recycled material, knowing that the markers are not necessarily present on the objects to be sorted, or if they are present, they can be detected only on a part of the surface of the objects.

[0019] Embodiments relate to a method for sorting objects comprising steps of: depositing objects to be sorted on a belt of a conveyor belt; acquiring an image stream of objects present on a first area of ​​the belt extending over a width of the belt; acquiring material and / or color detection signals for each cell of a set of adjacent cells located in a second area extending over the width of the belt; for each of the cells, processing the detection signals to determine detected material and / or color attributes; detecting markers in the image stream of the first area; for each detected marker, decoding the marker to determine an object identifier; for each decoded marker,assigning the object identifier determined by decoding the marker to the cells covered by the decoded marker; determining specified attributes for each object identifier determined from a decoded marker; for each cell associated with an object identifier, determining a first sorting instruction based on sorting criteria and the specified attributes; for each cell, determining a second sorting instruction based on the sorting criteria and the detected attributes; for each cell, determining a combined sorting decision based on the first and second sorting instructions of the cell and a set of combination rules. In this way, it is possible to detect in a stream of objects to be sorted marked objects mixed with unmarked objects, by applying sorting criteria that can combine several values ​​of specified object attributes,which can be obtained from object identifiers resulting from the decoding of markers present in the acquired image stream and from material and / or color detection signals. The combination of marker decoding and material detection of the objects to be sorted offers many advantages, particularly in terms of detection accuracy. Indeed, in the majority of cases, the detected markers do not entirely cover an object. This is particularly the case when the objects are not marked over their entire surface, when the label is partly torn off, or when the algorithm is unable to detect all the markers (calculation times too long, marked surface too damaged or too dirty, tilted object, etc.).

[0020] Furthermore, the detected material may not correspond to that declared by the markers, for example when the same mold is used for several materials, or in the event of a marker printing error. In the event of a contradiction, priority is given to the material and / or color detection signals, the measurement of which is real, and not declarative unlike the data resulting from the decoding of a marker.

[0021] Even if marker technology is widely adopted, there will always be unmarked products present in a stream of marked objects to be sorted. By using a material detection device, it is possible to make combined sorting decisions on these objects. For example, sorting criteria can be configured to purify a stream of marked objects of contaminants consisting of unmarked objects.

[0022] This combination also reduces redundancy and the required computing power. Indeed, the algorithms used to detect and decode markers use a high level of redundancy between successive images, which results in significant computing power. By reducing the image capture frequency while ensuring the decoding of at least one marker per object, and by using material and / or color detection signals to complete the object mapping, it is possible to have equally high detection efficiency, despite a reduced image capture frequency.

[0023] According to one embodiment, the method comprises steps of extending the combined sorting decision of cells associated with an object identifier to neighboring cells not associated with an object identifier, by applying a set of extension rules. and extending the combined sorting decision to neighboring cells not associated with the detection of a marker,

[0024] According to one embodiment, the method comprises steps of controlling a sorting member to direct each object present on the strip towards an output selected according to the combined sorting decisions of the cells opposite the sorting member. The precision obtained in the control of the sorting member applies regardless of the output of the sorting system selected for each marker, which makes it possible to ensure that the sorting command applied to all the cells corresponding to an object in its entirety is consistent and therefore to improve the purity of the sorting.

[0025] According to one embodiment, according to the set of combination rules, the combined sorting decision for a cell is: in accordance with one of the first and second sorting instructions for the cell if these are identical, assigned to one of the outputs of the sorting system in the event of a contradiction between the first and second sorting instructions for the cell, the output of the sorting system being determined according to the sorting criteria, in accordance with the first sorting instruction of the cell if the second sorting instruction is indeterminate due to an absence of significant detection of material and / or color in the cell, in accordance with the second sorting instruction of the cell, if the second sorting instruction of the cell is determined and if the sorting criteria, taking into account the detected attributes, specify that the second sorting instruction has priority, and determined according to the sorting criteria and the combined sorting decision of an adjacent cell,if the first sorting instruction is undetermined due to the absence of detection of a marker in the cell.,

[0026] Thus, all cases of values ​​of the first and second sorting instructions are processed taking into account the specificities of the detection modes by image capture and processing and by material and / or color detection.

[0027] According to one embodiment, the method comprises a first extension operation comprising steps consisting of: selecting at least one cell of a first extension rank, not associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and which is adjacent to a core of at least one cell associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and for which the material specified by the marker corresponds to the material detected by the sensor assembly for the adjacent cell and the selected cell; and assigning to the combined sorting decision of each selected cell a value equal to that of a cell of the core.

[0028] Such an extension operation makes it possible to increase the number of cells affected by the sorting commands to be activated for an object to be sorted, and therefore to improve the precision of the sorting commands, and thus to reduce sorting command errors.

[0029] According to one embodiment, the method comprises a subsequent extension operation comprising steps consisting of: selecting at least one cell belonging to a higher extension rank, not associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and which is adjacent to at least one cell belonging to a lower extension rank, and for which the material detected by the sensor assembly corresponds to that of the selected cell; and assigning to the combined sorting decision of each cell selected for the higher extension rank a value equal to that of the cell belonging to the lower extension rank.

[0030] This arrangement makes it possible to further increase the number of cells affected by the sorting commands to be activated for an object to be sorted, for better precision of the sorting command and therefore better efficiency of the sorting of the object to be sorted.

[0031] According to one embodiment, several extension operations are performed up to a maximum number of extension operations defined for each output of the sorting system.

[0032] This arrangement makes it possible to further increase the number of cells affected by the sorting commands to be activated for an object to be sorted, for better sorting efficiency of the object to be sorted, while limiting the risk of overflowing onto adjacent objects to be sorted which could lead to a sorting error.

[0033] According to one embodiment, the maximum number of extension operations is defined for each identifier or each value of a specified attribute.

[0034] According to one embodiment, the method comprises the implementation of a detection table comprising for each cell:

[0035] - the first sorting instruction,

[0036] - the second sorting instruction,

[0037] - the combined sorting decision,

[0038] - the extension rank, and

[0039] - the object identifier from a marker decoding.

[0040] According to one embodiment, determining the specified attributes from each object identifier determined using a decoded marker comprises a step of searching for the object identifier in a local database storing object identifiers capable of being extracted from the specified markers and attributes associated with each object identifier.

[0041] According to one embodiment, the method comprises steps of: receiving sorting criteria; and updating a database storing object identifiers capable of being extracted from the specified markers and attributes associated with each object identifier, the updating of the database comprising determining a sorting instruction for each object identifier based on the sorting criteria and the specified attributes associated with the object identifier, the first sorting instruction for each cell associated with an object identifier being the sorting instruction associated with the object identifier in the database.

[0042] By implementing a database of marked objects, it is possible to determine the first sorting instruction and the specified attributes determined using an object identifier by simply reading the database. According to one embodiment, the method comprises steps consisting of: deleting from the database the object identifiers which, during a certain period, have not been read to determine the first sorting instruction.

[0043] This arrangement makes it possible to simplify the operations of searching for objects in the database, using their identifiers to retrieve the associated sorting instructions and updating the sorting instructions stored in the database according to the sorting criteria.

[0044] According to one embodiment, the sorting criteria relate to one or more of the following specified or detected attributes: the nature of the material, the color, the shape, the use, the viscosity of the material, a percentage of biogenic material in the object, a percentage of recycled material in the object.

[0045] Thus, many sorting criteria can be taken into account, allowing complex sorting.

[0046] According to one embodiment, the image stream is acquired by illuminating the first area of ​​the strip alternately under two lighting colors, such that each object point moving through the first area is illuminated at least once by each of the lighting colors, the two lighting colors being located respectively at the two ends of the visible spectrum.

[0047] In this way, most markers present on product packaging can be detected regardless of the packaging color with lighting that is simpler to construct and more powerful for the same total number of sources (LEDs) than if three lighting colors had to be provided.

[0048] According to one embodiment, the images of the image stream are acquired alternately under two different lighting colors, so that each cell is illuminated successively under the two lighting colors.

[0049] According to one embodiment, one of the two lighting colors has a wavelength less than 460 nm and the other of the two lighting colors has a wavelength greater than 630 nm.

[0050] Embodiments may also relate to a sorting system, comprising: a belt conveyor comprising a belt on which objects to be sorted are placed, a set of cameras for acquiring a stream of images of objects present on a first area of ​​the belt extending over a width of the belt, a set of sensors for acquiring material and / or color detection signals for each cell of a set of adjacent cells located in a second area extending over the width of the belt, an image processing unit for detecting markers in the stream of images, a signal processing unit for determining detected attributes determined from the material and / or color detection signals, a sorting member for directing each object present on the belt towards an outlet of the sorting system, and a sorting control unit configured to implement the method defined above.

[0051] According to one embodiment, the system comprises a lighting system comprising two sets of monochrome sources alternately illuminating the first zone.

[0052] Brief description of the figures

[0053] The present invention will be better understood with the aid of the following description of exemplary embodiments with reference to the appended figures, in which identical reference signs correspond to structurally and / or functionally identical or similar elements.

[0054] Figure 1 is a schematic side view of an object sorting system, according to one embodiment,

[0055] Figure 2 is a schematic view of a detection and analysis device of the sorting system, according to one embodiment,

[0056] Figure 3 is a schematic view of an image of objects detected on a belt of a conveyor of the sorting system, according to one embodiment,

[0057] Figure 4 is a schematic view of an image of objects detected on the conveyor belt of the sorting system, according to one embodiment,

[0058] Figure 5 represents the contents of a sorting table at a stage of a sorting method, according to one embodiment,

[0059] Figure 6 represents an example of the contents of the sorting table at another stage of the sorting method, according to one embodiment,

[0060] Figure 7 represents the contents of the sorting table at a next step of the sorting method, according to one embodiment,

[0061] Figure 8 represents the contents of the sorting table at a next step of the sorting method, according to one embodiment,

[0062] Figure 9 represents the contents of the sorting table at a next step of the sorting method, according to one embodiment,

[0063] Figure 10 represents another example of a control table of a sorting member, according to one embodiment,

[0064] Figure 11 is a schematic side view of an object sorting system, according to another embodiment.

[0065] Detailed description

[0066] Figure 1 represents an object sorting system, according to one embodiment. The sorting system comprises a belt conveyor BCV comprising a belt B1 on which objects OB to be sorted are deposited. The belt B1 is driven in translation in a direction D1 to drive the objects into a tunnel formed by a roof T1 arranged above the belt B1. The sorting system also comprises a detection device DTD comprising a set of cameras IS1-IS6 and a set of sensors IVS for detecting material and / or color. The detection device DTD is arranged so as to observe an area of ​​the belt B1 downstream of an area Z0 for depositing the objects to be sorted OB. The set of cameras IS1-6 can be associated with lighting devices VL1, VL2 which can be focused, to illuminate an area Z1 of the belt B1 observed by the cameras under different wavelengths. The IVS sensor assembly observes a Z2 area of ​​the B1 band downstream of the Z1 area.The sorting system also comprises a sorting member BLS arranged at an outlet end of the conveyor BCV. The sorting member BLS comprises, for example, a plurality of compressed air ejection nozzles, aligned in a direction L1 perpendicular to the direction D1 and parallel to the plane of the belt B1, each nozzle being controlled by a respective solenoid valve. When one of these valves is open, the air ejected by the nozzle thus opened pushes the object located above the nozzle towards the outlet S1, otherwise the objects arriving at the outlet end of the conveyor BCV fall (under the effect of gravity) towards the outlet S0.

[0067] The IVS sensor assembly may comprise at least one of the following devices: a spectrometer operating in the near or mid-infrared wavelength range and / or in the visible range, a LIBS spectrometer, one or more color cameras for detecting the colors of objects on the B1 band, or more generally for recognizing the appearance of objects (shape, texture, color, etc.), inductive detectors for characterizing the presence of metals, an X-ray emitter / detector assembly, and a laser profilometry or triangulation system for detecting black objects or objects insensitive to the detection technologies mentioned above.

[0068] The sorting system may also comprise a turbine BL for driving air above the belt B1 at a speed corresponding substantially to the speed of movement of the belt B1 in the direction D1, so that the lightest objects OB deposited on the belt are also driven in the direction D1 at the same speed. The air supplied by the turbine BL is guided into the tunnel formed by the roof T1.

[0069] Figure 2 represents the DTD detection device associated with an analysis device of the sorting system, according to one embodiment. Figure 2 shows the IVS sensor assembly, the IS1-IS6 camera assembly, the associated VL1, VL2 lighting devices, as well as the exit zone of the strip B1 with the BLS sorting member comprising the NZ nozzles, and the respective EV solenoid valves. Figure 2 shows in particular the field observed by each of the IS1-IS6 cameras along the width of the B1 strip (L1 direction). As illustrated, the areas of the B1 strip observed by the IS1-IS6 cameras overlap slightly. Each IS1-IS6 camera is connected to an image processing unit PC1-PC6.

[0070] The IVS sensor assembly is connected to a PCS processing unit for signals from the sensor assembly. The PC1-PC6 and PCS processing units communicate via an NS network with a sorting control unit MPC. A TL terminal can be used to control the MPC control unit. The VL1, VL2 lighting devices include, for example, light-emitting diode (LED) bars, placed on either side of the detection axes of the IS1-IS6 cameras.

[0071] According to a known technique, the IS1-IS6 cameras are monochrome and the LED bars include two or three sets of monochrome LEDs, each covering a specific color in the visible spectrum. The lights are switched on sequentially, following a cycle controlled by a specially designed LM electronic box. Image acquisition is synchronous between all the IS1-IS6 cameras, and also synchronous with one of the lighting colors, the whole being controlled by the LM box. The image capture frequency is adjustable between 30 and 300 fps ("frames per second"). This is the total number of images, all colors combined. By successively activating three colors at the same frequency, 100 images per second are obtained in each color at 300 fps. The use of different colors maximizes contrasts for the detection of markers on objects detected on the B1 band.For example, if a marker is printed in yellow on a white background, it is barely visible to the naked eye, but with blue lighting, it appears black on a white background with a monochrome camera. It is not necessary to know the color that illuminated the detected object for decoding markers.

[0072] According to another known technique, the illumination of the Z1 zone observed by the IS1-IS6 cameras can be achieved by white LEDs, pulsed or not, and the IS1-IS6 cameras used are RGB color cameras, i.e. with three channels, red, green and blue. Each image capture therefore provides three monochrome sub-images, i.e. one per channel. This solution allows a higher acquisition frequency per color channel, but at the cost of lower spatial resolution, because the pixels of each channel are juxtaposed on the reading circuit, and therefore shared between the three channels.

[0073] It has been observed by the inventors that two different lighting colors are sufficient to detect all markers appearing on product packaging. Furthermore, measurements show that the best contrasts are obtained by using lighting colors alternately at both ends of the visible spectrum, namely blue lighting for the lower limit (at a wavelength below 460 nm), and red or near-infrared lighting for the upper limit (at a wavelength above 630 nm). The use of another lighting color does not provide any significant benefit. According to one embodiment, the detection of the markers is carried out by illuminating the zone Z1 alternately using blue lighting and red or near-infrared lighting. This solution is very advantageous, because the lighting power of an LED bar can be distributed between two colors instead of three.This ensures 50% more power per color for the same number of light sources.

[0074] In the example of Figure 2, the camera array comprises six IS1-IS6 cameras for a strip width B1 of 1.2 m. Each camera has a field at the strip of 23 cm along the L1 axis by 14 cm along the D1 axis. Each camera is connected to a respective image processing unit PC1-PC6. Of course, other configurations are possible, in particular depending on the computing power of the image processing units. Thus, for example, several processing units can be connected to one camera, or conversely, several cameras can be connected to one processing unit.

[0075] Each PC1-PC6 image processing unit is configured to analyze the images provided by the IS1-IS6 camera to which it is connected, in order to detect object markers present in each image. The PC1-PC6 image processing units do not communicate with each other, but transmit messages to the MPC sorting control unit at the end of each image processing.

[0076] According to one embodiment, each of the messages thus transmitted to the MPC control unit comprises:

[0077] - a moment of image capture,

[0078] - the coordinates (X, Y) in the image of each marker detected in the image, and

[0079] - an identifier corresponding to each of the detected markers.

[0080] The IVS sensor assembly is for example equipped with a scanner which periodically scans the entire width of the strip B1 (along the direction L1) in the zone Z2. The PCS processing unit connected to the IVS sensor assembly periodically provides data in the form of an acquisition zone divided into one or more lines (along the direction L1) of a few hundred adjacent rectangular cells. Thus an acquisition line comprises for each cell, a number indicating the position (along the width L1 of the strip B1) of the cell (and possibly an acquisition line number), and a code indicating a classification of the detected attribute, for example a material, a color, .... The resolution of the IVS sensor assembly can be coarser than that of the IS1-IS6 cameras, since the cells of the IVS sensor assembly can have several millimeters on each side.However, this resolution is sufficient to create a "chemical image" where the color of each point can represent a given family of objects. Furthermore, this resolution is not necessarily linked to the number of EV solenoid valves in the BLS sorting device. The scanning period of the IVS sensor assembly is independent of that of the IS1-IS6 cameras. Alternatively, the IVS sensor assembly may include a hyperspectral camera viewing the entire width of the B1 strip, without physical movement to scan a detection line extending across the width of the strip. Nevertheless, each line is acquired according to a constant acquisition period and providing data for each cell.

[0081] The PC1-PC6 and PCS processing units and the MPC control unit can be synchronized by known protocols such as, for example, TCP / IP or UDP protocols, with an accuracy of less than a millisecond. Furthermore, since the workload of the MPC control unit is relatively low, it is conceivable that the processing performed by the PCS processing unit and the MPC control unit is handled by a single processing unit.

[0082] The image processing and marker detection algorithms implemented in the PC1-PC6 processing units require relatively high computing power due to the high number of images to be processed per second and the high resolution of each of these images. Even if these processes can be optimized by parallelizing the calculations using specialized processors (GPU type), the marker decoding time can remain long depending on the images. Consequently, it can be expected that the detection algorithm does not necessarily decode all the markers of an image, but stops at the first marker found and decoded in the image, or at the first marker found in a predefined sub-part of the image.

[0083] Thus, regardless of the detail of the operation of the marker decoding software, the markers are not always all decoded. Figure 3 represents an image of the band B1, the objects C1, C2 detected by the processing units PC1-PC6, the markers M1 represented in the form of crosses associated with a sorting command to the output S1, the markers M2 represented in the form of circles associated with a sorting command to the output S0, and the markers M3 represented in the form of diamonds which have not been decoded. Thanks to the redundancy between successive images, the undecoded markers M3 can be decoded during the processing of the following images.

[0084] It turns out that the execution time of an image processing algorithm can be variable. The total time between acquisition and delivery of a result by each processing unit PC1 PC6 can fluctuate for example between 20 and 300 ms. If the distance between the cameras IS1-IS6 and the nozzle bar NZ with the solenoid valves EV is set to a value between 1 and 2 m, and the moving speed of the belt B1 can be between 2 and 5 m / s, it is possible that an image processing result is delivered too late to the MPC control unit to be evaluated. If this situation occurs, a non-blocking error message can be issued by the PCS control unit.

[0085] The sorting control unit MPC combines the data received from the image processing units PC1-PC6 and the processing unit PCS, and controls the sorting unit BLS. To do this, the control unit MPC stores the received data in a detection table which contains all the useful information for combining the data, and controlling the sorting unit BLS. The control unit MPC performs image processing to improve sorting and deduces sorting orders which it sends to the solenoid valves EV. The detection table can be saved in a file, for later analysis.

[0086] The MPC sorting control unit also includes a user interface accessible from the TL terminal, and presenting a set of menus for configuring sorting criteria, including for example:

[0087] - a menu for selecting sorting attributes,

[0088] - a menu for adjusting the sorting of marked objects,

[0089] - a menu for setting object families, and

[0090] - optionally, a graphic display of a sorting test, in the form of a map representing the detection table.

[0091] The information encoded in the markers can be of two types, depending on the coding capacity of the marker. If the coding capacities of the markers are reduced (10 to 15 bits per marker), it can be considered to directly encode a list of sorting and recycling attributes, i.e. directly the few properties to be used to determine sorting instructions. For example, the property "food use" or not can be encoded on 1 bit, the property "viscosity level" (MFI) can be encoded on 2 bits for "very low", "low", "medium" and "high", a material code (plastic resins) can be encoded from 1 to 7, on 3 bits.

[0092] If the coding capacity of the markers is sufficient, for example 42 bits as for barcodes, or more, it is more advantageous to code a unique object identifier, also called SKU ("Store Keeping Unit"). The object identifiers used can be those standardized by the GTIN ("Global Trade Identification Number"), and used by the international barcode system. Of course, other choices are possible, provided that they uniquely identify the objects. This object identifier serves as an entry key to access a complete technical sheet of the object. This sheet can be accessed via a remote information system, for example via the Internet, and provides in particular a list of sorting / recycling attributes.

[0093] When an OB object passes over the B1 band, only a few milliseconds are available between its detection and the sorting decision being made. The time available may therefore be insufficient to access remote information and make the sorting decision.

[0094] According to one embodiment, the sorting system comprises a local database SD connected to the sorting control unit MPC and providing a sorting instruction for each marked object identifier to be sorted, for example output S0 / output S1. The local database SD can also store a list of values ​​of specified object attributes for each object identifier, the sorting instruction being modifiable and determined according to sorting criteria stored by the sorting control unit MPC and selected by the user.

[0095] Each time a marker is decoded, the object identifier extracted from the marker is used to access the local database SD. If this identifier is not already present in the local database, a remote database is queried to obtain in a few seconds or minutes the list of specified object attributes corresponding to the extracted identifier. The local database SD is then updated with the identifier, the specified object attributes obtained and possibly the sorting instruction determined on the basis of the sorting criteria stored by the sort control unit MPC. Thus, the first object(s) associated with an identifier not appearing in the local database SD are not sorted (sent to the output SO), but the following objects associated with this identifier can be sorted correctly.

[0096] If the object identifier extracted from the marker is present in the local SD database, a corresponding sorting instruction is read and compared with the sorting instruction resulting from the analysis of the signals from the IVS sensor assembly. A combined sorting decision is established on the basis of these two sorting instructions by the MPC processing unit which controls the EV solenoid valves. Thus, access to the SD database can be performed only once for a given sorting recipe, and only to read the ejection instructions. This arrangement avoids unnecessary calculations or disk accesses.

[0097] According to one embodiment, the object identifiers that have not been searched for during a certain period to determine the first sorting instruction are deleted from the SD database. This arrangement makes it possible to simplify the operations of searching for objects in the database, from their identifiers to retrieve the associated sorting instruction and updating the sorting instructions stored in the database according to the sorting criteria. For this purpose, a read occurrence counter can be associated with each object identifier stored in the database. This counter can be incremented by one for an object each time the object identifier is searched for in the database to determine the first sorting instruction. This counter can be reset to zero from time to time, for example at regular intervals (for example every week).Objects associated with a zero occurrence counter before resetting are deleted from the SD database.

[0098] According to one embodiment, the object attribute selection menu is presented on the screen of the TL terminal. An example of such a menu is presented in the following table 1:

[0099] Table 1

[0100] The possible values ​​of each object attribute are predefined. An example of these possible values ​​is shown in the following Table 2:

[0101] Table 2

[0102] Following validation of the selections made in Table 1, the "marked objects" menu for selecting material determined by a marker is displayed. The columns of the "marked objects" menu are selected in the sorting attributes selection menu (Table 1) and include an additional "Output" column which allows you to specify a sorting instruction based on the attributes specified in the other columns. The "marked objects" menu is for example as presented in the following Table 3: Table 3

[0103] (1), (2), (3), (4), (5): the boxes in table 3 thus referenced contain all the possible choices for the value of the corresponding attribute appearing in the header of the corresponding column, these choices being listed in table 2. Table 3 is displayed when the attributes "Color", "Shape", "Use", "Viscosity", "% biogen." and "% recycled" have been selected in the attribute selection table (table 1).

[0104] It is possible to duplicate a line of the "marked products" menu following the selection in a box belonging to the line and one of the columns "Color", "Shape", "Use", "Viscosity", "% biogenic" and "% recycled" of a value different from "All", to allow to specify another sorting instruction for the same material but having other values ​​for the attributes thus selected.

[0105] According to one embodiment, upon validation of the "marked products" menu, each of the first sorting instructions specified in the "Output" column is entered into the local SD database for each marked object identifier corresponding to the attribute values ​​indicated in the row of the table corresponding to the sorting instruction. It is therefore not necessary to reread the choices made in the menus during sorting, only the local SD database being consulted according to the object identifier resulting from the decoding of a marker in an image.

[0106] The "object families" menu for selecting materials and colors, detected by the IVS sensor assembly, includes for each material indicated in the "Material" line of table 2, and possibly each color indicated in the "Color" line of table 2, a set of sorting criteria allowing a sorting instruction to be determined.

[0107] An example of the "object families" menu is shown in the following table 4:

[0108] Table 4

[0109] * if the marker is non-priority or if no marker is detected

[0110] If the "color" attribute is not selected in the object attribute selection menu (Table 1), the "Detected color" column of the "object families" menu is not displayed. The user is prompted to select one of the alternatives indicated in each box of the "Detected color", "Priority marker" and "Output" columns of the "object family" menu. Following the selection of a color as "clear" or "colored" in a row, this row is duplicated to allow specifying another sorting instruction for the same material but having a different value of the "color" attribute. Thus, it is possible to specify a different sorting instruction for two object families made of the same material, for example two PET families such as "Light PET" and "Colored PET". If the "Priority marker" box is selected (set to "yes"), the combined sorting decision is determined by the first instruction specified in the "Output" column of the "marked objects" menu.The sorting instruction specified in the "Output" box of the "Object Family" menu determines the combined sorting decision in cases where no marker has been detected, and the "Priority Marker" box contains the value "no".

[0111] According to one embodiment, the detection table is configured to map the entire space located between the area observed by the cameras IS1-IS6 and the ejection line at the downstream end of the strip B1. The detection table has the form of a two-dimensional array, consisting of records each representing a rectangular cell whose height (along the direction D1) is linked to a time increment dependent on the acquisition time of a line of cells of the IVS sensor assembly. For example, if a line of cells is acquired every 4 ms, and the strip moves at a speed of 3 m / s, the cells of the array have a height of 12 mm. However, the choice of the time increment can also be conditioned by the ejection technology. The width of each cell represented by the detection table is chosen so that the cells have a shape close to a square.This width can also be linked to the spacing between the NZ ejection nozzles of the BLS sorting device. The position of the objects to be sorted present in the detection table is updated at each time increment, according to a circular buffer scheme. The principle of a circular buffer is known to those skilled in the art and is not re-explained here.

[0112] The detection table accumulates and combines the results from all IS1-IS6 cameras and all marker images, applying a logical OR to all the results provided by the PC1-PC6 and PCS processing units, except in the special case where contradictory results appear in the same recording. With significant redundancy between images depending on the frame rate, the same cell can be designated several times. Each recording of the detection table can account for these redundancies. In case of contradictory results, one can for example decide to invalidate the cell for any future action, and note a negative redundancy value to trace the event.

[0113] Given the width in the D1 direction of the images taken by the IS1-IS6 cameras, the image capture rate and the translation speed of the strip B1, a point of the strip can be seen several times. Such redundancy creates a significant computational load. It is therefore desirable to limit this redundancy without significantly reducing the detected areas. Furthermore, if the cells corresponding to the records of the detection table are more spaced than the markers, each cell can contain several markers which can therefore lead to different sorting commands. According to one embodiment, one of the outputs SO, S1, for example output S1, is provided to receive selected objects, while the other output S0 receives all the unselected objects. In the event of a contradiction, output S0 is favored so as not to pollute the objects sent to output S1.

[0114] According to one embodiment, each record of the detection table stores the following parameters:

[0115] - a first sorting instruction (towards output SO / S1) determined on the basis of the detected and decoded markers,

[0116] - a second sorting instruction (towards output SO / S1) determined on the basis of detected attributes (material, color, etc.), obtained by processing the signals from the IVS sensor assembly,

[0117] - a combined sorting decision (towards output SO / S1),

[0118] - an extension row,

[0119] - an object identifier resulting from a marker decoding,

[0120] - values ​​of detected attributes (material, color), and

[0121] - a redundancy count value of the detected markers.

[0122] The values ​​of the detected attributes, provided by the PCS processing unit, and the corresponding second sorting instruction, are stored in the records of the detection table taking into account the number of rows of the table corresponding to the distance between the observation zone of the IS1-IS6 cameras and the observation zone of the IVS sensor assembly.

[0123] The combination of data from the PC1-PC6 image processing units and data from the PCS signal processing unit from the IVS sensor assembly allows the detection of marking errors. Indeed, it is possible that the material detected by the IVS sensor assembly for a cell in the detection table does not correspond to that deduced from the decoded markers. This situation can occur, for example, if the same mold is used for several materials, or if there have been errors in printing the markers. In the event of a contradiction, priority can be given to the IVS sensor assembly, whose measurement is real and not declarative.

[0124] The combination of data from the PC1-PC6 image processing units and data from the PCS processing unit improves the determination of the surfaces to be ejected and therefore the EV solenoid valves to be controlled for sorting each object. Indeed, in many cases, the detected markers do not entirely cover the object. This is the case when the objects are not marked over their entire surface, or when the label is partially torn off. This is also the case when the image processing fails to detect all the markers (calculation times too long, marked surface too damaged or too dirty, tilted object, etc.). In this case, the surface of the object taken into account for ejection can be extended to all the cells covered by the object detected by the IVS sensor assembly, from cell nuclei centered on the decoded markers.Positive extensions are thus defined around cores containing markers specifying the ejection of the object. Similarly, negative extensions are defined around cores containing markers specifying not to eject the object. The EV solenoid valves will be activated on all cells covered by an object if the extension is positive, and will not be activated on all cells covered by the object if the extension is negative.

[0125] Combining data from the PC1-PC6 image processing units with data from the PCS processing unit also allows for sorting of unmarked objects. Even if marker technology is widely adopted, there will always be unmarked products mixed in with marked objects. By using the IVS sensor array, combined sorting decisions can be made for these objects. In addition, a stream of marked objects can be purified of contaminants consisting of unmarked objects.

[0126] Combining data from the PC1-PC6 image processing units with data from the PCS processing unit also reduces redundancy and the required computing power. Working at a frame rate of 300 images per second requires significant downstream computing power. By decoding at least one marker per object and having the PCS processing unit complete the object mapping, it is possible to achieve similar sorting efficiency, despite the reduced frame rate.

[0127] According to one embodiment, the MPC control unit is configured to determine the first and second sorting instructions for each cell from the data provided in the messages transmitted by the processing units PC1-PC6 and PCS. For example, the first and second sorting instructions may take the values ​​1 (sorting to output S1), -1 (sorting to output S0) or 0 in the case where no data has been provided for the cell by the processing units PC1-PC6 and PCS. Then, the MPC control unit is configured to determine the combined sorting decision of each cell on the basis of the combination of the first and second sorting instructions determined for this cell. For this purpose, the MPC control unit may use the following instruction fusion table: Table 5

[0128] Table 5 provides possible values ​​for the combined sorting decision for a cell in the form of codes, depending on the first sorting instruction (instruction 1) and the second sorting instruction (instruction 2), as well as depending on the criteria selected in the "object families" menu (table 4). Table 5 also distinguishes between cases where the material and / or color detected for a cell by the IVS sensor assembly is consistent or not ("Compliant product" / "Non-compliant product") with the material and / or color specified by the marker detected for the cell. Output S0 is selected (no opening of the corresponding EV solenoid valve(s)) in the absence of data that could lead to selecting output S1. Note that the different codes indicated in table 5 are intended in particular to provide additional information on the value of the combined sorting decision that can be used for statistical purposes.

[0129] In the representation of Table 5: the code "X" indicates that the corresponding cell is to be sorted to output S1 because the material detected by the IVS sensor assembly is a material to be sorted with "Priority Marker" set to "yes" and the first sorting instruction indicates S1; the code "Y" indicates that the corresponding cell is to be sorted to output S0 because there is an inconsistency between the marking associated with the corresponding cell and the material or color detected by the IVS sensor assembly, while priority was given to the detection of markers, this inconsistency possibly resulting from a marking error; the code "O" indicates that the combined sorting decision of the corresponding cell is determined solely on the basis of the first sorting instruction of the cell, which indicates output S0, because the material detected by the IVS sensor assembly is a material to be sorted with "Priority Marker" set to "yes";

[0130] The code "A" indicates that the corresponding cell is to be sorted to output S1 due to the priority given to detection by the IVS sensor assembly; the code "B" indicates that no marker was detected in the corresponding cell, but that output S1 was selected by the second sorting instruction; The code "C" indicates that the corresponding cell is to be sorted to output S0 due to the priority given to detection by the IVS sensor assembly; the code "D" indicates that no marker was detected in the corresponding cell, but that output S0 was selected by the second sorting instruction; the code "E" indicates that the corresponding cell is to be sorted to output S1 due to the detected marker, the IVS sensor assembly having detected nothing, this situation can occur in the presence of a black object or a material or color detection hole;

[0131] The code "F" indicates that the corresponding cell is to be sorted towards the S0 output due to the detected marker, the IVS sensor assembly having detected nothing, this situation can occur in the presence of a black object or a material or color detection hole;

[0132] The codes "M" and "N" are to be used when priority has been given to the first sorting instruction and no marker has been detected in the corresponding cell. Thus, the code "N" is to be used if the second instruction for the cell selects output S1, and the code "M" is to be used if the second instruction for the cell selects output S0; and the code indicates that nothing was detected in the corresponding cell, probably due to the absence of an object in the cell, with output S0 then being selected by default.

[0133] The first step performed determines whether the material detected by the IVS sensor assembly is a material to be sorted according to the marker ("Priority marker") or not, as specified for each detectable material by the "object family" menu. If the material is not to be sorted according to the marker, then the first instruction has no impact and the combined decision is then equal to the decision of the second instruction. Thus, if the second instruction is at S1 and at least one marker has been detected (regardless of the first instruction), the combined sorting decision receives the code "A". If the second instruction is at S1 and no marker has been detected, the combined sorting decision receives the code "B".

[0134] If the second instruction is at S0 and at least one marker has been detected (regardless of the first instruction), the combined sorting decision receives the code "C". If the second instruction is at S0 and no marker has been detected, the combined sorting decision receives the code "D".

[0135] If the material is to be sorted according to the marker, then a conformity test is performed. This test consists of determining whether the material detected by the IVS sensor assembly for the cell corresponds to the material indicated by the marker detected for the cell. If the detected material and the material specified by the marker are not consistent, there is a material non-conformity problem and the combined sorting decision receives the code "Y". On the contrary, if the material detected by the IVS sensor assembly is consistent with that specified by the marker, then the cell is considered compliant. In this case, the combined decision applied to the cell corresponds to the first instruction. Thus, if the first instruction is at S1, then the combined sorting decision receives the code "X", otherwise (first instruction at SO) the combined sorting decision receives the code "O".If no marker has been detected, we then apply the second instruction, because it is the only one at our disposal: the combined sorting decision receives the code "N" if the second instruction is at S1, and the code "M" if the second instruction is at SO.

[0136] If no material has been detected by the IVS sensor set, then the combined sorting decision corresponds to the first instruction: the combined sorting decision receives the code "E" if the first instruction is at S1, and the code "F" if the first instruction is at SO. If no material or marker has been detected for a cell, then no sorting information is available for the cell, so the combined sorting decision receives the code

[0137] In an example sorting recipe, the sorting criteria selected in the "object families" menu are shown in the following table 6:

[0138] Table 6

[0139] The first row of Table 6 (after the header row) indicates that light-colored PETs as detected by the IVS sensor set are to be sorted according to the marker. The goal is to separate clear food-grade PET from clear non-food-grade PET, which can only be achieved using the attributes provided by the markers. Thus, the combined sorting decision will be determined by the marker for the "Light PET" cells. The first row allows clear PETs to be sorted according to the markers and directed to the S1 output when they are not marked, as they are assumed to be food-grade. The second row of Table 6 allows colored PETs, with or without a marker, to be directed to the SO output, as in this case they are contaminants. The third row of Table 6 allows sleeved PETs to be sorted according to the markers, as the color under the sleeve is not detectable by the IVS sensor set.The attributes provided by the marker are used to direct the object to exit S1 if it is a clear food-grade PET, and otherwise to exit SO. If no marker is detected, exit SO is specified in order to consider the object as a contaminant. The fourth line of Table 6 is used to direct other plastics to exit SO regardless of their color, with or without a marker, because they are contaminants. The fifth line of Table 6 is used to direct fibrous materials to exit SO regardless of their color, with or without a marker, because they are contaminants.

[0140] In the example of the sorting recipe partly specified in Table 6, the sorting criteria selected in the "marked objects" menu are shown in the following Table 7: Table 7

[0141] The sorting criteria thus selected allow the selection of output S1 for light-colored PET objects in the shape of a bottle and for food use, and light-colored sleeved PET objects for food use. All other combinations have their output specified at SO because they are considered contaminants.

[0142] Figure 5 illustrates an example of a detection table and shows in particular an example of the distribution of the combined sorting decisions in the CL cells of the detection table. Figure 5 shows three objects more or less separated by "empty" codes, namely an object OB1 to be sorted towards the output S1 (code "X") at the top left, an object OB2 to be sorted towards the output SO (code "O") at the bottom right, and which touches the first object, and an unmarked object OB3 (code "N") and not contiguous to the objects OB1, OB2, at the top right.

[0143] In object OB1, some cells received the code "X" due to the detection of a marker consistent with the material or color detected for these cells by the IVS sensor set. The other cells of the object received the code "M" due to the absence of a marker in the corresponding cells, the material and / or color detected by the IVS sensor set in these cells being assigned to the SO output. In object OB3, the cells received the code "N" due to the absence of a marker in the corresponding cells, the material and / or color detected by the IVS sensor set in the cell being assigned to the S1 output. In object OB2, some cells received the code "O" for example because it is a non-food product packaging and the sorting criteria specified to sort only food packaging to the SO output.The other cells of the object were assigned the code "M" due to the absence of marker in the corresponding cells, the material and / or color detected by the IVS sensor assembly in these cells being assigned to the SO output.

[0144] According to one embodiment, the MPC control unit executes an extension algorithm to extend the selection of cells for each of the outputs SO, S1, to the maximum number of cells covered by the objects, in order to improve the control accuracy of the sorting member BLS. This algorithm processes only the cells having received the code "M" or "N", i.e. the cells whose detected material is to be sorted according to the marker ("Priority marker"), but for which no marker has been detected. Furthermore, these extensions are carried out only on the cells which are neighbors of a cell having received the code "X" or "O" and if the material detected by the IVS sensor assembly for the cell at "N" or "M" is identical to that of the neighboring cell at "X" or "O".

[0145] Two types of extensions are applied. In so-called "positive" extensions, the areas of cells that have received the "X" code for selecting output S1 are enlarged by assigning a non-zero extension rank to cells that have received the "N" or "M" code and that are adjacent to a cell that has received the "X" code. In so-called "negative" extensions, the areas of cells that have received the "O" code for selecting output S0 are enlarged by assigning a non-zero extension rank to cells that have received the "N" or "M" code and that are adjacent to a cell that has received the "O" code. Thus, the extension rank assigned to a cell that has received the "X" or "O" code is always 0. The extension rank assigned to a cell adjacent to a cell that has received the "X" or "O" code is 1. The extension rank assigned to a cell adjacent to a cell that has received the rank n is n+1.The adjacency relationship applied here can be, for example, a 4-connectivity (we take into account the neighboring cells located at the top, bottom, left, right of the cell considered) or 8 (we also take into account all the neighboring cells located diagonally from the cell considered), depending on the number of cells considered adjacent to a cell located outside the edges of the area considered.

[0146] According to one embodiment, several expansion iterations are executed by traversing the detection table at each iteration. Each iteration adds a new cell rank with an expansion rank incremented by one around the already expanded areas. Each time an expansion rank is assigned to a cell, the detection table record corresponding to the cell may also receive the marker identifier of the lower expansion rank cell to which it is neighbored.

[0147] If two objects of the same material are contiguous as illustrated in Figure 4, there is a risk that the extension ranks assigned by successive iterations will overflow from a marked object into a neighboring unmarked object. Indeed, Figure 4 represents an image taken by cameras IS1-IS6 comprising a marked object C3 and an unmarked object C4. The assignment of extension ranks to the marked object risks extending the area occupied by the marked object into that of the marked object, for example up to the broken line extending into the object C4. As a result, cells covered by the unmarked object of the object C4 will be ordered for sorting like those covered by the marked object C3. The unmarked object (C4) therefore risks being sent by mistake to the same output as the marked object (C3).

[0148] The concept of numbered extension ranks allows this risk to be controlled, and a maximum extension rank can be defined, for example, in one of the menus displayed by the TL terminal. This maximum extension rank can be different for positive and negative extensions. It can also be specific to each material detected or specified by the markers or to each marked product identifier. In this case, a column can be added to the "object family" menu and / or the "marked objects" menu to allow the specification of the maximum extension rank.

[0149] Each maximum extension rank thus defines the number of iterations to be executed for the corresponding extension type or for the corresponding product type. One of the menus displayed by the TL terminal can specify that one or other of the outputs SO, S1 is preferred. Thus, if output S0 is preferred, negative extensions are performed before positive extensions at each iteration. It can be planned not to assign a non-zero extension rank to a cell when the cell identifier field is already filled. This ensures that the cell is assigned to the object of the same family, whose marker is closest in the images.

[0150] Figure 6 illustrates the contents of the detection table after the execution of a first iteration of the extension algorithm on the table illustrated by Figure 5. In the representation of Figure 6, the cells having received the code "N" or "M" and the extension rank 1 are noted by the code "#" if they are adjacent to a cell having received the code "X" (positive extension) and by the code "e" if they are adjacent to a cell having received the code "O" (negative extension).

[0151] Figure 7 illustrates the contents of the detection table after running a second iteration of the extension algorithm on the table illustrated in Figure 6. In the representation of Figure 7, cells that have received the code "N" or "M" and the extension rank 2 are noted by the code "#2" if they are adjacent to a cell that has received the code "#" and by the code "e2" if they are adjacent to a cell that has received the code "e".

[0152] Figures 8 and 9 represent the contents of the detection table after the execution of a third and a fourth iteration of the extension algorithm on the table illustrated by Figure 7, in the case where the maximum rank is set to 2 for positive extensions, and at least 4 for negative extensions. In the representation of Figure 8, the cells having received the code "N" or "M" and the extension rank 3 are noted by the code "e3" if they are adjacent to a cell having received the code "e2". In the representation of Figure 9, the cells having received the code "N" or "M" and the extension rank 4 are noted by the code "e4" if they are adjacent to a cell having received the code "e3".

[0153] At the end of the execution of all the required iterations of the extension algorithm, the remaining cells assigned the code "N" or "M" belong to objects not containing markers, such as the object OB3 located at the top right in Figures 5 to 9, and to areas of a marked object, which were not reached by the extension process. Indeed, the cells assigned the code "N" or "M" located beyond the maximum extension rank with respect to the "X" or "O" kernels do not receive a non-zero extension rank. Such an area appears at the bottom left of the object OB1. However, a large majority of the surface of the object OB1 is already selected for the output S1. The risk of sorting errors is therefore low. The choice of maximum extension ranks has several objectives. First of all, it is to ensure good sorting efficiency of the objects to be directed to the output S1 thanks to the positive extensions.This is also to avoid overflowing onto unmarked neighboring objects, and therefore to avoid mistakenly directing them to the wrong output. In the case of negative extensions, this is to avoid an untimely orientation towards output S1 of an object that should be sent to output S0, but which would not be completely covered by the extensions.

[0154] According to one embodiment, the MPC control unit is configured to successively read each line (extending along the direction L1) of the detection table and to open the EV solenoid valves corresponding to the cells of the read line, associated with the combined sorting decisions set to "X", "N", "A", "B", "E", and to the extension ranks "# <n>" (n="", 2, 3, ..., maximum positive extension rank), in order to send the corresponding object to the output S1, the EV solenoid valves corresponding to the cells of the read line, associated with the combined sorting decisions at "Y", "O", and to the extension ranks "e <n>" (n="", 2, 3, ..., maximum negative extension rank) remaining closed. The cells whose EV solenoid valves will be successively opened line by line (extending in the L1 direction) are shown in gray in Figure 9. Figure 10 shows a sequence of EV solenoid valve command lines obtained from the extension table of Figure 9 by replacing the combined sorting decision codes with the corresponding output S1 or S0, with output S0 being selected by default. The command table of Figure 10 indicates for each cell whether output S1 is selected, with empty cells in the table indicating that output S0 is selected. If the number of cells in the width of the table of Figure 10 is equal to the number of EV solenoid valves in the BLS sorting unit, each cell in this table defines the command of a solenoid valve.

[0155] According to one embodiment, the MPC control unit is configured to archive the lines (extending along the L1 direction) of the detection table after reading and controlling the EV solenoid valves. Each line thus archived can be indexed by the time elapsed since the start of a sorting sequence following the definition of the sorting criteria. To limit the volume of data thus archived, it may be decided not to save the lines of empty cells.

[0156] This archiving allows for a posteriori analysis of the performance of the sorting system. The archiving of the rows of the detection table can be used to determine the areas occupied by the sorted objects on belt B1, by calculating the number of cells assigned to each marker ID. This processing can make it possible to estimate the number of different objects that have been processed. This estimation can be based on an average detected area recorded per object ID (or SKU). This estimation can also be carried out by a connectivity analysis between cells corresponding to the same marker ID. This estimation is all the more reliable when the marker IDs are detailed. However, errors can occur in the presence of a cluster of objects in contact with the same marker ID, which risks being interpreted as a single object.To limit the impact of these errors, a maximum area per object identifier can be defined, which makes it possible to deduce whether it is a cluster or a single object.

[0157] Furthermore, discontinuous marking of objects, which would not be compensated by the extensions carried out (discontinuity of detection of markers and detection data from the IVS sensor set) may appear. The same object then appears as several islands of cells. Classic image processing algorithms, such as "closing", can be implemented to group the islands if they are the same object.

[0158] It will be clear to those skilled in the art that the present invention is susceptible to various variant embodiments and various applications. In particular, the invention is not limited to a sorting system with two outlets, but can also be applied to a system comprising three or more outlets. Thus, Figure 11 shows a sorting system which differs from the sorting system of Figure 1 in that it comprises three outlets S0, S1, S2 and a second sorting member BL1 arranged at an outlet end of the conveyor BCV which can be activated to direct an object to be sorted towards outlet S2. When the two sorting members BLS, BL1 are inactive, the objects to be sorted are directed towards outlet S0. In this case, the menus for specifying the sorting criteria (tables 3 and 4) and the merging table (table 5) are adapted by introducing a third alternative value (S2) in the "Output" column.

[0159] It should also be noted that the control of the sorting device is optional insofar as the process can be limited to taking inventory of the objects placed on the conveyor. Indeed, the combined sorting decision determines a set of parameters relating to each detected object, namely parameters concerning in particular the material, color, shape and use of the object.

[0160] Furthermore, other rules for combining the first and second sorting instructions to determine the combined sorting decision may be applied as needed. In particular, different priorities applied to the sorting instructions may be implemented without departing from the scope of the present invention.

[0161] The execution of the extension algorithm is also optional, knowing that this algorithm simply allows to enlarge the areas subject to the sorting commands and therefore to reduce sorting errors. In addition, the contours of the objects present on the B1 band can be detected by the implementation of known image processing, for example based on one or more artificial intelligence or neural network models. The extension algorithm can also be carried out further upstream on the images from the IS1-IS6 cameras. It is also possible to use artificial intelligence software to recognize other aspects of the objects such as shape, texture, color, etc.). The order of the two sets of sensors described can be reversed. Indeed, the system can operate equally well with the IVS sensor set placed upstream in the direction of movement of the B1 band, before the IS1-IS6 cameras placed downstream.

[0162] The implementation of the SD database is also optional, knowing that the determination of the first sorting instruction can be carried out without this local database, by accessing a remote database to determine the specified attributes (linked to the detected marking identifier) ​​and by applying the sorting criteria to these attributes.

[0163] It should also be noted that detecting and decoding markers using alternating illumination with only two colors located respectively at the ends of the visible spectrum, constitutes a separate invention from the method of sorting objects. Thus, the present description also covers a method of reading markers on an object, comprising steps of: illuminating the object alternately with two colors located respectively at the ends of the visible spectrum, acquiring images of the object when it is illuminated under each of the two illumination colors, detecting one or more markers on the object in the images where the object is illuminated under one of the two illumination colors, and decoding each detected marker.

[0164] Image acquisition can be performed using one image sensor sensitive to both lighting colors, or using two sensors sensitive to both lighting colors respectively.< / n> < / n>

Claims

CLAIMS 1. A method for sorting objects comprising steps consisting of: placing objects to be sorted on a belt (B1) of a belt conveyor (BCV); acquiring a stream of images of objects present on a first zone (Z1) of the belt extending over a width of the belt; acquiring material and / or color detection signals for each cell of a set of adjacent cells (CL) located in a second zone (Z2) extending over the width of the belt; for each of the cells, processing the detection signals to determine detected material and / or color attributes; detecting markers in the stream of images of the first zone; for each detected marker, decoding the marker to determine an object identifier; for each decoded marker, assigning the object identifier determined by the decoding of the marker to the cells covered by the decoded marker;determining specified attributes for each object identifier determined from a decoded marker; for each cell associated with an object identifier, determining a first sorting instruction based on sorting criteria and the specified attributes; for each cell, determining a second sorting instruction based on the sorting criteria and the detected attributes; for each cell, determining a combined sorting decision based on the first and second sorting instructions of the cell and a set of combination rules.; 2. The method of claim 1, comprising steps of extending the combined decision of sorting cells associated with an object identifier to neighboring cells not associated with an object identifier, by applying a set of extension rules.

3. Method according to claim 1 or 2, comprising steps consisting of controlling a sorting member (BLS, BL1) to orient each object present on the strip towards an output (S0, S1, S2) selected as a function of the combined sorting decisions of the cells opposite the sorting member.

4. Method according to one of claims 1 to 3, in which, according to the set of combination rules, the combined sorting decision for a cell (CL) is: in accordance with one of the first and second sorting instructions for the cell if these are identical, assigned to one of the outputs (SO, S1, S2) of the sorting system in the event of a contradiction between the first and second sorting instructions for the cell, the output of the sorting system being determined according to the sorting criteria, in accordance with the first sorting instruction of the cell if the second sorting instruction is undetermined due to an absence of significant detection of material and / or color in the cell, in accordance with the second sorting instruction of the cell, if the second sorting instruction of the cell is determined and if the sorting criteria taking into account the detected attributes specify that the second sorting instruction has priority, and determined according to the sorting criteria and the combined sorting decision of an adjacent cell, if the first sorting instruction is undetermined due to an absence of detection of a marker in the cell.

5. Method according to one of claims 1 to 4, comprising a first extension operation comprising steps consisting of: selecting at least one cell (CL) of a first extension rank, not associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and which is adjacent to a core of at least one cell associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and for which the material specified by the marker corresponds to the material detected by the sensor assembly (IVS) for the adjacent cell and the selected cell; and assigning to the combined sorting decision of each selected cell a value equal to that of a cell of the core.

6. Method according to claim 5, comprising a subsequent extension operation comprising steps consisting in: selecting at least one cell (CL) belonging to a higher extension rank, not associated with an object identifier, for which the sorting criteria assign a priority to the first instruction, and which is adjacent to at least one cell belonging to a lower extension rank, and for which the material detected by the sensor assembly (IVS) corresponds to that of the selected cell; and assigning to the combined sorting decision of each cell selected for the higher extension rank a value equal to that of the cell belonging to the lower extension rank.

7. Method according to claim 6, in which several extension operations are carried out up to a maximum number of extension operations defined for each output (SO, S1, S2) of the sorting system.

8. The method of claim 7, wherein the maximum number of extension operations is defined for each identifier or each value of a specified attribute.

9. Method according to one of claims 5 to 8, comprising the implementation of a detection table comprising for each cell: - the first sorting instruction, - the second sorting instruction, - the combined sorting decision, - the extension rank, and - the object identifier from a marker decoding.

10. Method according to one of claims 1 to 9, in which the determination of the specified attributes from each object identifier determined using a decoded marker, comprises a step of searching for the object identifier in a local database (SD) storing object identifiers capable of being extracted from the markers and the specified attributes associated with each object identifier.

11. Method according to one of claims 1 to 10, comprising steps consisting of: receiving sorting criteria; and updating a database (SD) storing object identifiers capable of being extracted from the specified markers and attributes associated with each object identifier, the updating of the database comprising the determination of a sorting instruction for each object identifier according to the sorting criteria and the specified attributes associated with the object identifier, the first sorting instruction for each cell (CL) associated with an object identifier being the sorting instruction associated with the object identifier in the database.

12. Method according to one of claims 1 to 11, in which the sorting criteria relate to one or more of the following specified or detected attributes: the nature of the material, the color, the shape, the use, the viscosity of the material, a percentage of biogenic material in the object, a percentage of recycled material in the object.

13. Method according to one of claims 1 to 12, in which the image stream is acquired by illuminating the first zone of the strip (B1) alternately under two lighting colors, so that each point of the object moving through the first zone is illuminated at least once by each of the lighting colors, the two lighting colors being located respectively at the two ends of the visible spectrum.

14. Method according to one of claims 1 to 13, in which the images of the image stream are acquired alternately under two different lighting colors, so that each cell (CL) is illuminated successively under the two lighting colors, one of the two lighting colors being able to have a wavelength less than 460 nm and the other of the two lighting colors being able to have a wavelength greater than 630 nm.

15. Sorting system, comprising: a belt conveyor (BCV) comprising a belt on which objects to be sorted (OB1, OB2, OB3) are placed, a set of cameras (IS1-IS6) for acquiring a stream of images of objects present on a first zone (Z1) of the belt extending over a width of the belt, a set of sensors (IVS) for acquiring material and / or color detection signals for each cell of a set of adjacent cells (CL) located in a second zone (Z2) extending over the width of the belt, an image processing unit (PC1-PC6) for detecting markers in the stream of images, a signal processing unit (PCS) for determining detected attributes determined from the material and / or color detection signals, a sorting member (BLS, BL1) for directing each object present on the belt towards an output (SO, S1, S2) of the sorting system sorting,and a sorting control unit (MPC) configured to implement the method according to one of claims 1 to 14., 16. Sorting system according to claim 15, comprising a lighting system comprising two sets of monochrome sources (VL1, VL2) alternately illuminating the first zone (Z1).

Images