Method for evaluating quality of sorting line item stream

Abstract

The invention relates to a method (1) for evaluating at least one quantitative parameter of a stream of items in an item sorting line (2) comprising a conveyor belt of the stream of items, each item being identifiable by at least one intrinsic characteristic, the sorting line (2) comprising sorting means adapted to sort items according to at least one target intrinsic characteristic; the method (1) comprises: a step (10) of receiving at least one item stream image; a step (11) of detecting an item in said at least one image; a step (12) of classifying each detected item by a preset list of item types; a step (13) of determining, for each detected item, the inherent properties of the classified item, according to its classification in said list of item types; and-a step (14) of calculating at least one value of a quantitative parameter of the stream of items as a function of a proportion of items in the stream of items having an intrinsic characteristic corresponding to the at least one target intrinsic characteristic.

Description

[0001] The invention relates to a method for evaluating the quality of a flow of items in a sorting line.

[0002] The present invention relates to an item sorting line, such as a plastic packaging, carton packaging sorting line or any other type of sorting line. However, it is more conventional to sort these items according to the materials they are made of. Therefore, although the present invention requires that the items can be detected on the sorting line as described below, in the following description, material sorting or item sorting can be equally mentioned.

[0003] In the field of material sorting, a well-known problem is the assessment of the purity of the material flow coming out of a sorting line.

[0004] Since the sorted material stream is subsequently recycled, high mass purity is important.

[0005] The normal practice is to conduct random quality checks on material samples. Operators take samples or unpack packages (because some materials are packed in packages for shipment), manually sort the materials, and calculate the overall purity.

[0006] However, this approach is not only cumbersome but also not particularly reliable because it is sensitive to possible operator error and is based on a small number of samples.

[0007] To achieve an automated characterization of item flows, especially waste flows, we consulted the paper by Gilles Pagès, Victor Reutenauer: Precision of caracterization of paper for recycling.2019.hal-02388073. However, the solution proposed in the paper for paper does not achieve optimal accuracy, especially for other materials. In fact, the paper describes the characterization of the types of paper in a paper flow, taking into account the total area covered by the different types of paper on the conveyor belt. This is feasible in this case because the density values ​​of different types of paper are relatively uniform. However, for other materials, the density variations can be much greater, which makes this model unreliable. For example, a crushed plastic bottle has a smaller volume than an intact one and occupies less space on the conveyor belt, even though both have the same mass.

[0008] Therefore, there is a need for a method for evaluating the purity of an item flow that can be applied to any type of items in a sorting line.

[0009] To this end, we propose a method for evaluating at least one quantitative parameter of a flow of objects in an object sorting line, the sorting line comprising a conveyor belt of a flow of objects, each object being identifiable by at least one intrinsic characteristic, the sorting line comprising sorting means suitable for sorting objects according to at least one target intrinsic characteristic; the method comprising the following steps:

[0010] - A step of receiving at least one image of a flow of objects;

[0011] - a step of detecting an object in said at least one image;

[0012] - a step of classifying each detected object by a preset list of object types;

[0013] - for each detected object, a step of determining, according to its classification in said list of object types, the intrinsic characteristics of the classified object; and

[0014] - A step of calculating at least one value of a quantitative parameter of said flow of objects as a function of the total count of detected objects and the count of detected objects having an intrinsic characteristic corresponding to said at least one target intrinsic characteristic.

[0015] Therefore, we adopt a method that can reliably calculate quantitative parameters of item flows and is applicable to any type of item flow.

[0016] Advantageously and without limitation, the inherent property comprises the constituent material of the article. This inherent property is most suitably exploited when sorting articles in a sorting line for a recyclable material stream.

[0017] Advantageously and without limitation, the quantitative parameter is selected from the group consisting of calorific value, mass purity or recyclable utilization. Thus, a quantitative parameter with a strong correlation can be determined. Mass purity is a preferred quantitative parameter in the implementation of the present invention, because it is the main parameter related to the quality of the material flow. Calorific value or recyclable utilization are two other quantitative parameters, which, although usually more complicated to analyze, can be obtained relatively reliably by the present invention.

[0018] Advantageously and without limitation, the above-mentioned determination step comprises executing an expert system adapted to determine the intrinsic property of each object based on the object type and the target intrinsic property. Thus, the intrinsic property of each object can be determined reliably and timely.

[0019] Advantageously and non-restrictively, the method performs the step of collecting at least one value of another quantitative parameter for each item type in a preset list of item types; and then performs the step of calculating at least one value of the quantitative parameter of the item flow based on the proportion of items in the item flow having inherent characteristics corresponding to the at least one target inherent characteristic and the value of the another quantitative parameter for each item type in the list.

[0020] Thus, when calculating a quantitative parameter of the flow of objects, another quantitative parameter of each object detected is taken into account, thereby increasing the accuracy of the assessment.

[0021] Advantageously and without limitation, said further quantitative parameter of each item type comprises its mass. In fact, the weight of each item is particularly relevant, especially, but not exclusively, when mass purity is used as the quantitative parameter.

[0022] Advantageously but not limitingly, the step of acquiring at least one value of said another quantitative parameter comprises, for each item type in said list of item types, calculating an average of a plurality of measured values ​​of said another quantitative parameter of a plurality of items belonging to said item type.

[0023] Therefore, we associate to each item type, also referred to in the specification as a class of items, an average value of a quantitative parameter, for example, the average weight of the class of items. This enables an improved estimation of the quantitative parameters of the item flow.

[0024] Advantageously and non-limitingly, each item type in said list of preset item types comprises a plurality of values ​​of said further quantitative parameter corresponding to a plurality of measured values ​​of said further quantitative parameter for a plurality of different items belonging to the same item type; and

[0025] The step of calculating at least one value of a quantitative parameter of the article flow executes a statistical module so as to obtain a simulated quantitative parameter, wherein the simulated quantitative parameter depends on the proportion of articles in the article flow having an inherent characteristic corresponding to the at least one target inherent characteristic, and a plurality of values ​​of the other quantitative parameter for each article type in the preset article type list.

[0026] In this way, the method adopts a statistical approach capable of improving the quantitative parameters of the article flow, taking into account previously acquired measurements of multiple quantitative parameters for each article type.

[0027] Advantageously and without limitation, the step of calculating at least one value of the quantitative parameter of the flow of articles comprises executing the statistical module a plurality of times so as to obtain a confidence interval for the value of the quantitative parameter of the flow of articles from the distribution of the results.

[0028] Thus, by repeating the statistical module multiple times, a distribution of results can be derived, making it possible to determine a confidence factor for the obtained evaluation value, thereby improving the quality of the estimation of the quantitative parameters of the article flow.

[0029] Advantageously, the statistical module comprises a sampling statistical module with reset, such as a bootstrap sampling method.

[0030] Therefore, when we have a large number of quantitative parameter measurements of items in each item type, we can use statistical methods that are particularly relevant.

[0031] Advantageously and without limitation, the receiving step may include receiving at least one item flow image from a remote server (eg, a cloud server), a server installed at the sorting site, a portable data storage device, or any other known data providing means.

[0032] Advantageously and without limitation, the sorting line comprises means for acquiring images of the flow of objects conveyed.

[0033] Advantageously and without limitation, the image acquisition device comprises an optical camera, or an infrared or near-infrared acquisition device, or an X-ray acquisition device.

[0034] Therefore, the appropriate type of image acquisition device can be selected according to the type of sorting line and target item to provide the most relevant detection information for the image.

[0035] The invention also relates to a device for evaluating at least one quantitative parameter of a flow of objects in an object sorting line, the object sorting line comprising a conveyor belt for conveying a flow of objects, each of which can be identified by at least one intrinsic characteristic, the sorting line comprising sorting means suitable for sorting the objects according to at least one target intrinsic characteristic; the device comprising:

[0036] - means for receiving at least one image of a flow of objects;

[0037] - means for detecting an object in said at least one image;

[0038] - means for classifying each detected object according to a preset list of object types;

[0039] - means for determining, for each detected object, the intrinsic characteristics of the classified object according to its classification in said list of object types; and

[0040] - means for calculating at least one value of a quantitative parameter of the flow of objects as a function of the proportion of objects in the flow of objects having an intrinsic characteristic corresponding to said at least one target intrinsic characteristic.

[0041] The invention also relates to an article sorting line comprising a conveyor belt for conveying a flow of articles, wherein each article is identifiable by at least one intrinsic characteristic, the sorting line comprising a sorting member suitable for sorting the articles according to at least one target intrinsic characteristic, characterised in that the sorting line comprises an evaluation device as described above.

[0042] Advantageously and without limitation, the sorting line comprises a device for collecting images of the conveyed stream of articles. Specifically, the device for collecting images of the conveyed stream of articles comprises an optical camera, or an infrared or near-infrared collecting device, or an X-ray collecting device.

[0043] Other features and advantages of the present invention can be found by reading the following description of three specific embodiments of the present invention in conjunction with the accompanying drawings, which are only for illustrative purposes and are not intended to limit the present invention, wherein:

[0044] [ Figure 1 ] is a flow chart of the method described in the first embodiment of the present invention;

[0045] [ Figure 2 ] is a view of an item sorting line as seen from a camera used in the method of the present invention;

[0046] [ Figure 3 ] is a view of an item sorting line seen from a camera used in the method of the present invention, in which items are detected and the detected items are marked with a bounding box; and

[0047] [ Figure 4 ] is a flow chart of the method described in the first and second embodiments of the present invention.

[0048] The material sorting line 2 comprises a conveyor belt 20 suitable for conveying a flow 21 of objects to be sorted.

[0049] The sorting line 2 generally comprises a device for sorting materials (not shown) conveyed on a first conveyor belt, and then conveying the sorted materials to a material packaging device.

[0050] These bales are usually cubical bales weighing several hundred kilograms and are ready for transport to recycling centers. As is known to all, the requirements for the purity of the material flow are particularly important. For example, in the sorting of plastic packaging, people try to obtain transparent PET bales with a mass purity of more than 98%.

[0051] In order to automatically assess the purity of the sorted material flow, a camera is installed after the sorting device and suspended above the conveying material flow before being stored in bulk or packed into material packs for shipment.

[0052] The purpose of this specification is not to describe the different ways of installing and organizing a sorting line on a single conveyor belt or on multiple conveyor belts. The invention can be applied to any type of sorting line, provided that an image of the flow of objects can be acquired after the sorting device and before they are packaged into material packages.

[0053] In order to evaluate at least one quantitative parameter of the object flow of the sorting line, a camera is mounted above the conveyor belt of the object flow, the purity of which is to be evaluated.

[0054] like Figure 2 and Figure 3 As shown, the camera captures continuous images of the sorted object flow and transmits the images to a device for evaluating the quantitative parameters.

[0055] The evaluation device is a computer, which is defined as an electronic component including a processor, a memory, a storage device and an input / output device, which is particularly capable of receiving images captured by the camera and displaying, printing, transmitting or even recording the obtained evaluation purity.

[0056] In order to evaluate at least one quantitative parameter of the article flow, the evaluation device uses method 1 described below.

[0057] In the main embodiment of the invention, the quantitative parameter evaluated is the mass purity of the product stream 21 .

[0058] However, the invention is not limited to this single quantitative parameter. In particular, the quality of a material flow can be assessed by other types of quantitative parameters, such as its recyclability or calorific value.

[0059] The method 1 employed enables the assessment of at least one quantitative parameter, here the mass purity of the material flow.

[0060] However, alternative embodiments of the invention may be provided, in particular wherein the quantitative parameter may be the calorific value, the renewable utilization or other parameters available to the person skilled in the art.

[0061] Each object may be identified by at least one intrinsic property, here its material, and the sorting line comprises sorting means adapted to sort the objects according to at least one target intrinsic property, such as an expected material.

[0062] Also, for example, in this embodiment, a sorting line suitable for sorting waste is used, where the waste refers to plastic packaging, so as to separate various different plastic materials such as transparent PET packaging, dark PET packaging, opaque PET packaging and PE / PP / PS mixtures.

[0063] As we all know, polyethylene terephthalate is abbreviated as PET.

[0064] Generally speaking, the main sorting line leads to a plurality of secondary conveying lines 2, which convey the sorted different article flows to the packaging components.

[0065] In this case, the method can be implemented on each secondary line so as to determine the purity of each product stream produced according to the target material corresponding to the material to be transported by the secondary line, thereby producing a product stream that meets the specifications of the recycling device.

[0066] To this end, method 1 first comprises step 10 of receiving an image stream from the above-mentioned camera.

[0067] For each analyzed image, the method performs an object detection step 11 .

[0068] To do this, we use an object detection module, here referred to as You Only Look Once (commonly abbreviated as YOLO).

[0069] This real-time object detection module is known for its original research paper: You Only Look Once: Unified, Real-Time Object Detection, Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp.779-788.

[0070] This type of object detection is particularly suitable for real-time detection, especially for detecting objects in video streams.

[0071] Indeed, if the method is described herein as continuous image acquisition, the invention may be applied to a video stream viewed as a rapid succession of images.

[0072] In a video stream, it may be beneficial that the same item detected between two images is not counted twice.

[0073] Advantageously, object detection methods are suitable not only for identifying an object in an image, but also for associating an object with the same object identified in a previous image, thereby linking the same object in two images and assigning it a unique identifier.

[0074] This tracking method (tracking in English) can track the flow of items in real time.

[0075] However, if all items are counted multiple times in a similar manner, then multiple counts of the same item may not be a problem and, therefore, there is no need to implement a real-time tracking step. Thus, tracking of items in the stream of items is not essential for the proper functioning of the method.

[0076] Furthermore, the present invention is not limited to such real-time tracking. Intermittent capture at regular time intervals can also be implemented so that each image includes a new set of objects conveyed by the object flow and all objects in the object flow appear in the captured image. However, unlike real-time analysis, this sequential method may result in detection errors.

[0077] The present invention is not limited to the YOLO target detection module, and any other detection method may also be used.

[0078] However, the YOLO module uses a convolutional neural network, which has high performance that enables particularly powerful real-time detection.

[0079] In this embodiment, the YOLO module can obtain a bounding box 30 including the detected object for each analysis image, such as Figure 3 shown.

[0080] After obtaining the image bounding box 30, a classification step 12 is performed for each detected object.

[0081] The captured image is then transmitted, with the coordinates of the detected object indicated in the image in the form of the coordinates of the bounding box 30 where the detected object is located.

[0082] In an alternative implementation of the invention, an image segmentation step is performed in order to separate each bounding box 30 from the image.

[0083] Each detected object can then be sent to the classification module by simply inputting the portion of the image contained in the associated bounding box 30.

[0084] In the main embodiment of the present invention, the classification step 12 uses another neural network module, which is the ResNetm module, which is well known from the following scientific paper: He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778. 2016.

[0085] Specifically, a 50-layer deployment called ResNet-50 can be adopted.

[0086] However, the present invention is not limited to this single classification module, and other neural network-based deployments may be used, and other methods such as linear regression, decision tree, or any other method known to those skilled in the art may be used. Thus, each object detected in the image is assigned an object category.

[0087] For example, a category list of items in a plastic packaging item stream, also called an item type list, might include the following:

[0088]

[0089]

[0090] Then, a step 13 of determining the intrinsic property of the object according to its category is performed. In this embodiment, the intrinsic property of the object is its material.

[0091] This determination step 13 is implemented by executing an expert system adapted to determine the material of an object based on its category.

[0092] However, it is not always possible to accurately determine the material based on the category of the item.

[0093] Therefore, the expert system also takes the target material into consideration.

[0094] For example, when an opaque bottle is detected, it is not always possible to determine exactly whether it is a PE / PP / PS blend or a dark PET. To carry out this detection, the expert system can determine, based on the target material, whether the material is located on a sorting line or on a secondary line as described above. In fact, if the item is located on a dark PET target material line, it is most likely this material. On the contrary, if it is a so-called mixed line, in particular receiving a PE / PP / PS blend, then the material that makes up the plastic bottle is most likely one of these materials.

[0095] PE-PP-PS mixture refers to a product line of bottles and cans made of high-density polyethylene PE, polypropylene PP and polystyrene PS.

[0096] For example, based on the previously obtained categories, the expert system is able to determine the materials in the material list, including:

[0097] 1 EMR (Equipped Household Recyclable Packaging) 2 Metal 3 Paper 4 Polyethylene / polypropylene / polystyrene blend 5 Light color PET 6 Dark PET 7 Residue 8 Sterile Brick Materials for Food 9 unknown

[0098] These materials on the list can be combined within the expert system. In fact, the main purpose is to determine whether the object is an object made of the target material. Therefore, the material list can be reduced and simplified according to the production line where the method is implemented.

[0099] As we all know, an expert system consists of an inference engine, a set of rules, and a knowledge base.

[0100] The rationale for using an expert system here is that associating a class of items with their materials solves a logical problem, and an expert system can solve this problem relatively effectively if the knowledge base is sufficient and the rules are well formulated.

[0101] Other methods may also be used, for example, machine learning methods, such as linear regression method, decision tree method or other suitable machine learning methods.

[0102] Therefore, the category and material of each item detected in the image can be derived.

[0103] Then a calculation step 14 is performed, which is to calculate at least one quantitative parameter value of the item flow (here its mass purity) based on the proportion of items in the item flow having an intrinsic property (its material) corresponding to the at least one target intrinsic property (here the target material of the sorting line).

[0104] To this end, according to a main embodiment of the present invention, this step includes calculating the mass purity ratio of the article flow by calculating the proportion of articles made of the target material in the article flow.

[0105] Therefore, we assess the quality purity of the item stream as the proportion of items made of the target material in the total number of items detected.

[0106] This primary approach works well for sorting lines that sort plastic or paper materials, such as plastic packaging, because the weight of each item is roughly the same, especially when considering the total mass of the packaged material.

[0107] According to a second embodiment of the invention, the method 1 ′ differs from the first embodiment in the calculation step 14 .

[0108] Indeed, in the second embodiment, before the calculation step 14, a mass assignment step 15 of each detected object is performed.

[0109] To this end, in method 1 ', a collection step 40 is initially performed, in which a plurality of objects of the same category are weighed. In a second embodiment, the collection step also calculates the average value of the masses of the objects of the same category. In other words, we calculate the average weight of each category of objects. The more objects are weighed, the more reliable the average mass value is.

[0110] In the calculation step 14, the ratio of the mass of the object made of the target material to the total mass of the detected object is calculated.

[0111] In other words, mass purity is calculated by dividing the sum of the weights of items made from the target material by the sum of the masses of all items in the stream of inspected items.

[0112] Thus, particularly reliable mass purities can be achieved, but relatively long mass formation steps are required beforehand and a large amount of analytical data needs to be acquired.

[0113] According to a third embodiment of the invention, the method 1 ′ differs from the first and second embodiments in the calculation step 14 .

[0114] In a third embodiment, the purpose of the calculation step 14 is no longer to determine a single mass purity value, or any other quantitative parameter value of the sorted article stream, but to associate this value with a confidence interval.

[0115] To this end, a collection step 40 is previously performed, by which the weights of a plurality of objects associated with a category in the previous category list are collected. Unlike the second embodiment of the invention, this collection step does not calculate an average value, but a set of weights of different weights.

[0116] To this end, multiple objects belonging to the same category are weighed and each weight value is recorded and associated with the object category. The greater the number of objects weighed, the greater the mass associated with a certain object category.

[0117] Therefore, in each mass distribution step 15 of the inspected object executed before the calculation step, the mass of the material package is simulated on the one hand, and the mass of the object made of the target material is simulated on the other hand, so as to calculate the mass purity of the material package by resetting the sampling statistical inference method.

[0118] Here we use the bootstrap sampling method which is well known in the statistical community.

[0119] The method is based on randomly sampling a mass value for each inspected item from all masses associated with each item class.

[0120] The more mass data associated with each item category, the better the quality of this statistical method. On the other hand, if fewer weight values ​​are stored, for example, the number of mass values ​​per item category is less than 100, the advantage of this method decreases.

[0121] Mass purity is calculated as the ratio of the total simulated mass of items made from the target material to the simulated mass of the material pack (that is, the mass of all detected items).

[0122] This simulation is repeated a number of times, for example 100-2000 iterations, so that the simulation is multiplied.

[0123] The number of iterations will also be calibrated based on the number of weights associated with each item category. The greater the number of weights, the greater the number of iterations.

[0124] Therefore, the confidence interval of the purity value of the detected product flow mass can be calculated by the obtained distribution of simulated purity.

Claims

1. A method (1, 1') for evaluating at least one quantitative parameter of a flow of objects (21) in an object sorting line (2), the sorting line comprising a conveyor belt (20) for a flow of objects, each object being identifiable by at least one intrinsic characteristic, the sorting line (2) comprising sorting means adapted to sort the objects according to at least one target intrinsic characteristic; the method (1, 1') comprising the following steps: - a step (10) of receiving at least one image of the flow of objects; - a step (11) of detecting an object in said at least one image; - a step of classifying each detected object according to a preset list of object types (12); - for each detected object, a step (13) of determining the intrinsic characteristics of the classified object according to its classification in said list of object types; and - A step (14) of calculating at least one value of a quantitative parameter of said flow of objects as a function of the total count of detected objects and the count of detected objects having an intrinsic characteristic corresponding to said at least one target intrinsic characteristic.

2. The method (1, 1') according to claim 1, characterized in that The inherent properties include the materials of which the item is composed.

3. The method (1, 1') according to any one of claims 1 or 2, characterized in that The quantitative parameter is selected from calorific value, mass purity or recyclability.

4. The method according to any one of claims 1 to 3, characterized in that The determining step (13) comprises executing an expert system adapted to determine the intrinsic properties of each object based on the object type and the target intrinsic properties.

5. The method (1') according to any one of claims 1 to 4, characterized in that The method performs a step (40) of acquiring at least one value of another quantitative parameter for each item type in a preset list of item types; Then, a step (14) is performed to calculate at least one value of the quantitative parameter of the object flow based on the proportion of objects in the object flow having an intrinsic characteristic corresponding to the at least one target intrinsic characteristic and the value of the other quantitative parameter for each object type in the list.

6. The method (1') according to claim 5, characterized in that The further quantitative parameter of each item type comprises its mass.

7. The method according to claim 5 or 6, characterized in that: The step (40) of acquiring at least one value of the further quantitative parameter comprises, for each item type in the list of item types, calculating an average of a plurality of measured values ​​of the further quantitative parameter of a plurality of items belonging to the item type.

8. The method (1') according to any one of claims 5 or 6, wherein each item type in the preset list of item types comprises a plurality of values ​​of the further quantitative parameter corresponding to a plurality of measured values ​​of the further quantitative parameter for a plurality of different items belonging to the same item type; and The step (14) of calculating at least one value of a quantitative parameter of the article flow executes a statistical module so as to obtain a simulated quantitative parameter, the simulated quantitative parameter depending on the proportion of articles in the article flow having an intrinsic characteristic corresponding to the at least one target intrinsic characteristic, and a plurality of values ​​of the other quantitative parameter for each article type in the preset article type list.

9. The method according to claim 8, characterized in that The step (14) of calculating at least one value of the quantitative parameter of the flow of objects comprises executing the statistical module a plurality of times, thereby obtaining a confidence interval for the value of the quantitative parameter of the flow of objects based on the distribution of the results.

10. The method according to claim 9, characterized in that The statistics include a sampling statistics module with reset, such as a self-service sampling method.

11. The method according to any one of claims 1 to 10, characterized in that: The receiving step (10) may include receiving at least one item flow image from a remote server or a server installed at the sorting site, a portable data storage device 12. The method according to any one of claims 1 to 10, characterized in that: The sorting line (2) comprises means for capturing images of the conveyed flow of objects.

13. The method according to claim 12, characterized in that The device for collecting images of the conveyed article flow includes an optical camera, or an infrared or near-infrared collecting device, or an X-ray collecting device.

14. A device for evaluating at least one quantitative parameter of a flow of objects (21) in an object sorting line (2), the object sorting line comprising a conveyor belt (20) for conveying the flow of objects, each object being identifiable by at least one intrinsic characteristic, the sorting line (2) comprising sorting means adapted to sort the objects according to at least one target intrinsic characteristic; the device comprising: - means for receiving at least one image of a flow of objects; - means for detecting an object in said at least one image; - means for classifying each detected object according to a preset list of object types; - means for determining, for each detected object, the intrinsic characteristics of the classified object according to its classification in said list of object types; as well as - means for calculating at least one value of a quantitative parameter of the flow of objects as a function of the proportion of objects in the flow of objects having an intrinsic characteristic corresponding to said at least one target intrinsic characteristic.

15. An article sorting line comprising a conveyor belt (20) for conveying a stream of articles, each article being identified by at least one intrinsic characteristic, the sorting line (2) comprising sorting means adapted to sort the articles according to at least one target intrinsic characteristic, characterized in that The sorting line comprises an evaluation device as claimed in claim 14 .

Images