Automatic device for collecting and separating garbage

Abstract

Automatic device (D) for collecting and separating garbage (R), made with only one material or more materials, comprising a hollow-inside support frame (1), for letting said device (D) rest on a surface, in which at least one opening (11) is obtained for the insertion of said one or more garbage (R) and at least one first (51) and one second container (52), which are able to house the separated garbage (R), comprising an acquisition and processing unit (2), arranged in said support frame (1) which is able to: acquire the images of said one or more garbage (R), detect in said acquired images predetermined visual features of said one or more garbage (R), associate with said visual features a category which said materials belong to, classify said garbage (R) based on said association, send signals corresponding to said classification, and a bearing plate (3,3'), equipped with at least one movable portion (31, 31', 32, 32', 33, 33', 34, 34'), arranged elevated above said at least one first (51) and one second (52) container, which is able to receive said one or more garbage (R) coming from said opening (11), which is able to receive said corresponding signals from said acquisition and processing unit (2) and which is able to move said at least one portion (31, 31', 32, 32', 33, 33', 34, 34'), for the passage of said one or more garbage (R) towards the respective at least one first (51) and second container (52), based on said corresponding signals, at least one actuator member (4), arranged in said support frame (1), elevated above said bearing plate (3, 3'), which is able to receive said corresponding signals from said acquisition and processing unit (2) and to cause the passage of said one or more garbage (R) from said at least one portion (31, 31', 32, 32', 33, 33', 34, 34') towards the respective at least one first (51) and second container (52), based on said corresponding signals.

Description

Field of application

[0001] The present invention relates to a device for collecting and separating garbage, in particular a device which is able to collect the garbage and automatically separate it based on the material which is present in the garbage.

[0002] These devices are present in everyday life, but they do not allow the automatic garbage separation, rather they are solely used for the collection thereof.

[0003] More particularly, the invention relates to a device, which is able to recognize the object inserted for the collection, recognize the material which forms that object so as to categorize and finally separate it in an appropriate container based on the obtained and processed data.

[0004] The device can be mainly used in public places even outdoors, where there is a little attention to recycling and garbage management.Prior art

[0005] At present, the increase in garbage production represents one of the main environmental challenges, making the management thereof a priority for environmental sustainability. Therefore, devices used for the garbage selective collection are already known.

[0006] As a consequence of a raising awareness among the population on the issue of garbage selective collection, the use of a multiplicity of devices for collecting garbage, one for each type of material that has to be separated, has become widespread.

[0007] Although substantially responding to the field needs, the devices according to the prior art still have some drawbacks which are unsolved to date, related in particular to human error during the garbage separation based on the regulations in force and on the need for a plurality of containers, one for each material that has to be separated.

[0008] An aim of the present invention is to provide a device for collecting garbage which is able to receive more garbage at the same time and collect it separately based on the composition material, so as to make the selective collection more efficient and sustainable, improving people's quality of life.

[0009] A further aim of the present invention is to comply with the regulations which provide penalties in case of wrong separation and billing based on the amount of non-recyclable garbage being produced.

[0010] Yet, an aim of the present invention is to provide a device for collecting garbage, in which there is no need to frequently empty the garbage contained therein.

[0011] Finally, an aim of the present invention is to provide a device for collecting garbage which is reliable, robust and safe.

[0012] The technical problem underlying the present invention is therefore to solve the above-mentioned drawbacks of the prior art, providing a device which allows the collected garbage to be separated and deposited in the suitable container.Summary of the invention

[0013] The solution idea underlying the present invention is to provide an automatic device for collecting and separating garbage, based on the recognition of the category which the garbage item belongs to.

[0014] Based on the above solution idea, the technical problem is solved by an automatic device for collecting and separating garbage, made of only one material or more materials, comprising a hollow-inside support frame for letting said device rest on a surface, in which at least one opening is obtained for the insertion of said one or more garbage and a collection section comprising at least one first and one second container, which are able to house the separated garbage, comprising an acquisition and processing unit, arranged in said support frame which is able to: acquire the images of said one or more garbage, detect in said acquired images predetermined visual features of said one or more garbage, associate with said visual features a category which said materials belong to, classify said garbage based on said association, send signals corresponding to said classification and a bearing plate, equipped with at least one movable portion, arranged elevated above said at least one first and one second container, which is able to receive said one or more garbage coming from said opening, which is able to receive said corresponding signals from said acquisition and processing unit and which is able to move said at least one portion, for the passage of said one or more garbage towards the respective at least one first and second container, based on said corresponding signals, and at least one actuator member, arranged in said support frame, elevated above said bearing plate, which is able to receive said corresponding signals from said acquisition and processing unit and to cause the passage of said one or more garbage from said at least one portion towards the respective at least one first and second container, based on said corresponding signals.

[0015] Furthermore, according to the invention, said at least one movable portion of said bearing plate is a first, second, third and fourth side panel, being arranged along the four sides of the perimeter of said bearing plate and able to slide along a vertical axis, to allow said garbage to fall in said collection section.

[0016] Still according to the invention, said at least one movable portion of said bearing plate is geometrically one quarter of said bearing plate, which is able to slide along a vertical axis, and corresponds to the size of the relevant collection section, in particular of the corresponding container.

[0017] Always according to the invention, said actuator member comprises a telescopic arm, which is able to move along said vertical axis away from and towards said bearing plate and / or in a horizontal plane which is perpendicular to said vertical axis, and a push element associated with said telescopic arm, which is able to contact the garbage which is present on the bearing plate and push it towards the collection section.

[0018] Preferably according to the invention, said push element of said actuator member is a blade.

[0019] Furthermore, according to the invention, said actuator member is a delta arm or a Cartesian arm.

[0020] Still according to the invention, said push element of said actuator member is an air element, such as an aspirator, which is able to suck the garbage, or a blower, which is able to push the garbage.

[0021] Always according to the invention, a compression device, which is able to compress the garbage which is present in said collection section, is arranged in said collection section.

[0022] Preferably according to the invention, said compression device comprises a piston and a pair of rods, anchored to each other at one end, wherein said piston is able to move along said vertical axis, causing the movement of said pair of rods in a first vertical plane or in a second vertical plane.

[0023] Furthermore, according to the invention, said device comprises at least one sensor for detecting the loading of said collection section.

[0024] A further object of the present invention is a method for operating an automatic device for collecting and separating garbage, made of only one material or more materials, comprising a hollow-inside support frame, for letting said device rest on a surface, in which at least one opening is obtained for the insertion of said one or more garbage and a collection section comprising at least one first and one second container, which are able to house the separated garbage, comprising the following steps: (110). locating and recognizing said garbage inserted into said opening and sending corresponding signals, by means of an acquisition and processing unit; (120). sorting said garbage in said collection section, by means of an actuator member based on said corresponding signals.

[0025] Furthermore, according to the invention, said locating and recognizing step 110. comprises the following sub-steps: (111). locating said garbage item on a bearing plate arranged inside said device by means of a bounding box; (112). recognizing said garbage item from the bounding box and associating percentages of the material which it is made of by means of a convolutional neural network. Still according to the invention, said sorting step 120. comprises the following sub-steps: (121). training a convolutional neural network to associate an image of a garbage item with at least one percentage of the material contained therein; (122). separating the garbage in said collection section by means of said actuator member, based on said association. Always according to the invention, said method comprises an initialization step 101. in which, by means of the acquisition and processing unit, the image of said bearing plate in pixels is detected and the real size of said bearing plate in millimetres is mapped, according to a preset conversion factor, so that a portion in millimetres of said bearing plate corresponds to each pixel, and said image is stored as a background image.

[0026] Preferably according to the invention, the method comprises the step of acquiring successive images of said garbage comparing them with said background image to determine the presence of a garbage item on said bearing plate.Brief description of the drawings

[0027] Figure 1 shows a side sectional view of the automatic device D according to the present invention; Figure 2 shows a front sectional view of the automatic device D according to the present invention; Figure 3 shows a front partial sectional view of the automatic device D according to the present invention; Figure 4 shows a perspective view of the device of figure 1; Figure 5 shows a top schematic view of a first embodiment of a bearing plate of the device of figure 1; Figure 6 shows a perspective view of the first embodiment shown in figure 5; Figure 7 shows a top schematic view of a second embodiment of a bearing plate of the device of figure 1; Figure 8 shows a block diagram of the operation of the device of figure 1. Detailed description

[0028] With reference to figures 1 and 2 of the attached drawings, an automatic device for collecting and separating garbage R which can be made of only one material or more materials is indicated with D, mainly comprising a hollow-inside support structure 1, in which at least one opening 13 is obtained for the insertion of the garbage R, an image acquisition and processing unit 2, arranged in said support structure 1, for the recognition of the garbage R, a bearing plate 3, which is able to receive the garbage R and route it, also by means of at least one actuator member 4, in appropriate containers arranged in a collection section 5.

[0029] Said support structure 1 has a substantially parallelepiped elongated shape, whose prevailing dimension extends along a vertical axis Z, with respect to the ground, of a Cartesian system formed by a first horizontal axis X, which is perpendicular with respect to said vertical axis Z, and by a second horizontal axis Y, which is perpendicular with respect both to said first horizontal axis X and to said vertical axis Z, from an upper portion 11, towards a base portion 12.

[0030] The dimensions of the support structure 1 can be approximately of 600 mm in width, i.e. along said first horizontal axis X, 700 mm in depth, i.e. along said second axis Y and 1,600 mm in height, i.e. along said vertical axis Z.

[0031] With reference to figures 2-4, an opening 13 is obtained on said support structure 1, to let the garbage R in.

[0032] A chute 131, which is able to convey the garbage R towards the inside of the support structure 1, is associated with said opening 13.

[0033] A rear opening, not shown in the figures, is also obtained on said support structure 1 to access inside the device D in case of maintenance.

[0034] Said acquisition and processing unit 2, which is able to acquire the images and process them, is arranged in said upper portion 11, as will be described in detail below.

[0035] In particular, said acquisition and processing unit 2 comprises a microcontroller associated with a processing unit 21 equipped with an operating program or code of the device D and at least one camera 22, preferably two.

[0036] Without departing from the scope of protection of the present invention, it is additionally possible to insert a spectroscopic sensor of the NIR - near infrared type, or the like, whose output is processed through a known technique such as sensor fusion, as a supplement to the image acquisition and processing unit 2.

[0037] Said acquisition and processing unit 2 can also comprise one or more sensors, such as presence sensors for example, for detecting the garbage R inside the support structure 1.

[0038] The presence sensors can be for example infrared or ultrasonic sensors.

[0039] Referring in particular to figure 5, said bearing plate 3 is substantially a platform on which the garbage R fall and rest, after being inserted into said opening 13, and it is arranged elevated above said collection section 5.

[0040] Said bearing plate 3 is constantly framed by said camera 22.

[0041] Said bearing plate 3 comprises at least one movable portion for the passage of the garbage.

[0042] In particular, in a first embodiment represented in figures 5 and 6, said bearing plate 3 is equipped with a first movable portion or side panel 31, a second side panel 32, a third side panel 33 and a fourth side panel 34, arranged along the four sides of the perimeter of the bearing plate 3.

[0043] Said first 31, second 32, third 33 and fourth 34 side panels are able to slide along said vertical axis Z, to allow the garbage to fall in the collection section 5.

[0044] Said actuator member 4, schematically represented in figures 1-3 and 6, is internally fixed in said upper portion 11 of said support structure 1, and it is able to push the garbage R lying on the bearing plate 3 and let it fall towards the collection section 5, based on the category thereof.

[0045] Preferably, said actuator member 4 is a robotic arm of the delta type having four degrees of freedom on said first horizontal axis X, second horizontal axis Y and vertical axis Z, and on itself, and it is moved by four electric motors, in particular known motors of the "stepper motor nema 17" type characterized by a good speed and equally good movement precision.

[0046] Referring in particular to figure 6 said actuator member 4 comprises a telescopic arm 41 and a push element 42. Said telescopic arm 41 is able to move along said vertical axis Z away from and towards said bearing plate 3. Said telescopic arm 41 is also able to move in the horizontal plane XY, formed by said first horizontal axis X and said second horizontal axis Y. Said push element 42 is associated with said telescopic arm 41, so as to contact the garbage R which is present on the bearing plate 3 and push it towards the collection section 5.

[0047] Said push element 42 is preferably a blade or a paddle. However, said push element 42 can also be replaced by an air element, such as an aspirator, which is able to suck the garbage R, or a blower, which is able to push the garbage R.

[0048] Without departing from the scope of protection of the present invention, said actuator member 4 can be a Cartesian arm or another device which is suited to move objects.

[0049] Said collection section 5, shown in figures 1-6, is arranged below said bearing plate 3, along said vertical axis Z and is able to collect the garbage R which is made to fall from said bearing plate 3.

[0050] Said collection section 5 comprises at least one first container 51 and one second container 52. In this described embodiment, said collection section 5 comprises, in addition to said first 51 and second 52 container, a third container 53 and a fourth container 54.

[0051] The volumetric collection capacity of the containers is modulable, based on the materials to be collected. Preferably the available total volume is of 300 litres divided into said first 51, second 52, third 53 and fourth 54 containers; each one with a volume of 75 litres each.

[0052] For example, it is possible to provide two containers for paper collection and two containers for plastic collection. Without departing from the scope of protection of the present invention, it is possible to combine all the typologies of containers, so as to collect the garbage R made of all the existing materials, also based on the different local regulations in force for the garbage selective collection.

[0053] In an alternative embodiment, it is possible to insert a member for the initial collection, not shown in the figures, for temporarily collecting the garbage R in a part of the support structure 1, so as not to fill the bearing plate 3 and to allow time for the sorting device D to sort all the garbage R.

[0054] As can be seen in figure 4, said first 51, second 52, third 53 and fourth 54 containers are so shaped as to be tilting in pairs in the first vertical plane XZ, to allow a flap opening, the extraction of the suitably collected garbage R and the emptying of the containers themselves.

[0055] Without departing from the scope of protection of the present invention said first 51, second 52, third 53 and fourth 54 containers can be also tilting in pairs in the second vertical plane YZ.

[0056] A respective side panel 31, 32, 33, 34 overlooks each container 51, 52, 53, 54, so that the garbage falls from the bearing plate 3 just in the corresponding container. By way of example, the first side panel 31 is arranged at the first container 51, the second side panel 32 is arranged at the second container 52, the third side panel 33 is arranged at the third container 53 and the fourth side panel 34 is arranged at the fourth container 54.

[0057] Furthermore, said collection section 5 comprises at least one sensor 55 arranged in each of said first 51, second 52, third 53 and fourth 54 containers, for detecting the loading thereof. Said sensor 55 is a sensor of the ultrasonic type. Said sensors 55 are able to monitor the loading of said first 51, second 52, third 53 and fourth 54 containers and to send an alert signal to the acquisition and processing unit 2, so as to advise the manager of the automatic device D that a predetermined loading threshold of each container has been reached. The sensors 55 are ultrasonic sensors.

[0058] In addition, it is possible to insert weight sensors, or load cells, not shown in the figure, which are able to measure the weight of the garbage R which is present inside the containers, calibrated so as to advise the manager that a predetermined loading threshold of each container has been reached.

[0059] The set loading threshold both for said sensors 55 and for said weight sensors is modifiable at will by the manager.

[0060] For the container of wet or organic garbage R, it is also possible to perform a time control so as to avoid the emission of unpleasant odours.

[0061] Said collection section 5 further comprises a compression device 6 which is able to compress the garbage R in order to reduce the total volume thereof. Said compression device 6 comprises in turn a piston 61 and a pair of rods 62, anchored to each other at one end. Said piston 61 is able to move along said vertical axis Z, thus causing the movement of the pair of rods 62 in a first vertical plane XZ or in a second vertical plane YZ, and the subsequent compaction of the garbage R. Said compression device 6 thus allows the continuous operation of the automatic device D during the garbage compaction, since the shape of the pair of rods 62 allows the continuous fall of the garbage R without accumulation in the collection section 5 which would prevent the collection of the next garbage R.

[0062] A second embodiment of said bearing plate 3' is shown in figure 7, comprising a first portion 31', a second portion 32', a third portion 33' and a fourth portion 34', which are all movable.

[0063] Each portion 31', 32', 33' and 34' is geometrically one quarter of the bearing plate 3' and it corresponds to the size of the relevant collection section 5, in particular of the corresponding container 51, 52, 53 54, so that the garbage falls just in the target space. The garbage R is thereby constrained to fall just in the collection container and it does not fall in empty spaces inside the support structure 1.

[0064] The operation of the automatic device D, which is the object of the present invention, is the following.

[0065] When a user has to dispose a garbage item R, he inserts it through the opening 13. The garbage item R runs through the chute 131, until it rests on the bearing plate 3.

[0066] The garbage R resting on the bearing plate 3, is thus processed according to a method 100. At first, it is located and recognized, in a locating and recognizing step 110 and then separated, in a separating step 120. These two steps continuously alternate, firstly the position and the typology of the garbage R which is present on the bearing plate 3 are detected and then the actuator member 4 is operated, to separate a certain garbage item R, automatically selected and classified. The process is repeated until all the garbage R is correctly disposed in the collection section 5.

[0067] In the locating and recognizing step 110 the garbage R is located and individually classified, finding the position and the occupied area on said bearing plate 3, while in the separating step 120, the garbage R is sorted.

[0068] The locating and recognizing step 110 is performed by using convolutional neural networks (CNN), in particular by means of computer vision algorithms which are able to perform an object detection, which is able to locate a garbage item R, return the corresponding bounding box and associate a label with each bounding box, so that the garbage R can be classified with the corresponding label.

[0069] The algorithms used in the locating and recognizing step 110 are of the Multi-Stage detectors type which provide the bounding boxes.

[0070] The algorithms used in the classification and separating step 120 are of the Single-Stage detectors type, so as to locate and classify the garbage item R in parallel, at the same time.

[0071] The parameters entered in the algorithms installed in the devices are in a limited number, so as to occupy little space in the memory of the devices on which they are installed.

[0072] Without departing from the scope of protection of the present invention, the artificial intelligence models might be replaced with equivalent models, that is models with a different structure but which can similarly carry out the same task, allowing the continuous update to the current state of the art in the above scope, a possible replacement of the CNN is the Vision Transformer model.

[0073] The training data are both open-source datasets which are available on the web and proprietary datasets, internally collected to reinforce the less performing cases in terms of recognition precision.

[0074] The proprietary dataset comprises known images, coming from known sources of the open-source type and new images. The known images have been accurately selected from different datasets, not only limited to the garbage but also to objects of daily use. This choice is driven by the need to develop a machine learning algorithm which is able to recognize the objects, regardless of the original classification thereof. In order to increase the model ability to generalize in different scenarios, the backgrounds have been diversified in the images outside the device D, while for the images inside the device D, a uniform background reproducing the internal environment of the device D has been set, for ease of setup and control.

[0075] The objects with extremely different sizes in a single frame, the separation of adjacent or overlapping objects, and the recognition of transparent materials or of colours which blend with the background are difficult to detect. Therefore, the dataset has been enriched with direct shots on a recognition plate, using a smooth and isolated lighting to minimize environmental interferences.

[0076] The classification dataset is divided into two main subsets. The first subset S1 comprises a wide range of existing datasets, which can be accessed for commercial use, from which the relevant images have been extracted and cleaned. The cleaning process has involved the removal of non-relevant categories and of low-quality images, which are unsuitable for the training of effective algorithms. For some categories, a more specific remapping was required, for example, the conversion of images generically labelled as "plastic" in precise sub-categories such as "PET plastic bottle". Moreover, images containing multiple garbage have been cut out to create new images focused on single objects, thus improving the precision of the classification model.

[0077] The second subset S2 consists of proprietary images, accurately taken to cover gaps detected in the model testing step. Each object has been selected to represent specific recognition difficulties, such as the change in shape, size and material. For example, reflective metal objects and articles composed of translucent plastic have been photographed under controlled conditions to ensure that the model can effectively distinguish them under real operating conditions. The images have been taken with different angles and lighting configuration to test the model robustness under different lighting conditions and visual perspectives.

[0078] Preliminarily, the method 100 provides an initialization step 101, in which the automatic device D is preparing to operate. In particular, the automatic device D is calibrated and all the working parameters are set and all the components are arranged in the required position to perform the next operations, and then the automatic device D enters standby mode.

[0079] The acquisition and processing unit 2 detects the image of the bearing plate 3 in pixels and maps the real size of the bearing plate 3 in millimetres, according to a preset conversion factor, so that a portion in millimetres of the bearing plate 3 corresponds to each pixel. Moreover, the image of the empty bearing plate 3 is acquired and stored as a background, for further processing operations, said image will be defined hereafter as background image Is.

[0080] This step is not required when different methods, based on artificial intelligence, are used.

[0081] Afterwards, the locating and recognizing step 110 begins, during which the garbage which is present on the bearing plate 3 is identified and recognized.

[0082] In particular, in a first subset 111, the detection and the localization of the presence of the garbage item R on the bearing plate 3 occur. For this purpose, a time pulse or trigger enables at preset time intervals the acquisition of the images by means of said camera 22. Starting from the background image Is, the next images I F which are compared with the background image Is by means of a background subtraction algorithm are acquired.

[0083] As soon as the camera 22 acquires a next image I F in which there is a garbage item R, a bounding box B which defines the outlines of the garbage item R is created on the image.

[0084] An alternative to the above-described method can provide the use of a presence sensor, for example an ultrasonic or passive infrared - PIR - one, arranged inside the support structure 1, by which the presence of the garbage item R is directly detected.

[0085] When the number of bounding boxes B is higher than 0, it means that a garbage item R has been found and it is possible to proceed with the recognition thereof through the following second sub-step 112.

[0086] Still another alternative to the above-described method is represented by a method based on the artificial intelligence instructed to recognize the garbage R starting from the image detected by the camera 22 and to output the bounding boxes of the garbage R or, as an alternative, to segment the image.

[0087] Afterwards, in the second sub-step 112 the garbage R located in the first sub-step 111 is recognized or classified, i.e. it is determined of which material the garbage item R is composed.

[0088] For this second sub-step 112 a convolutional neural network is used, in which for each cut-out image of the single garbage item R at the input, the value of the percentage of probability that the garbage item R of the image is made of a certain material is provided at the output, for example: metal = 0,7; glass = 0,1; paper = 0,2.

[0089] Alternatively, it is possible to associate a label with a product brand, or with a specific object. The allocation of the labels is dynamic and variable.

[0090] The use of the percentages associated with each category of garbage R allows to classify the garbage R in a higher way, since, for example, if a garbage item R is classified with 90% of metal and 10% of dangerous garbage, it is possible to separate the garbage item R as dangerous and not as metal. Or, if a garbage item R is classified with 60% of plastic and 40% of paper, it is possible to separate the garbage item R as unselective.

[0091] These settings can be modified in a flexible manner.

[0092] Finally, if all the categories have a low recognition percentage, it is possible to allocate the category "other" and arrange the collection container "other".

[0093] After identifying and classifying the garbage item R in step 110, the separating step 120 is performed, in which the actuator member 4 receives the instructions from said acquisition and processing unit 2 for the correct separation of the garbage R. The instructions are in the form of commands and coordinates, which are similar to the known GCODE algorithm, used by 3D printers.

[0094] The separating step 120 preliminarily provides a first training sub-step 121, using simulation libraries in which the environment, i.e. the internal part of the support structure 1, the actuator member 4 and the garbage R are modelled, in a finite number, trying to represent as much as possible all the existing garbage R.

[0095] The first training sub-step 121 provides the determination of a reward function, i.e. of the value function. The two important aspects of the training step are the correctness of the separation of the garbage item R and the routing time of the garbage item R in the collection section 5.

[0096] The separation correctness takes priority with respect to the routing time, therefore in the reward function, the separation correctness will have a higher score with respect to the routing time.

[0097] For the reward function a negative score is also determined if the routing time exceeds a certain threshold value. The greater the routing time will be with respect to the threshold value, the greater the negative score which is subtracted from the final score will be.

[0098] If the garbage is correctly separated, the penalty will have the value 0. If the garbage is wrongly separated, in addition to the negative score of the routing time, a further negative value is also added, for example a value which is equal to - 1000.

[0099] Afterwards, in a second separating sub-step 122, the acquisition and processing unit 2 performs an algorithm which calculates the garbage item R arranged closer to the respective movable side panel 31, 32, 33, 34, based on the percentages of material received from said second recognizing sub-step 112, on the settings of the collection section 5, and determines the straight line from the centre of the garbage item R to the closest side panel, turning this information into commands for the actuator member 4.

[0100] In particular, having the bounding box available, that is the object position, it is possible to calculate the centre thereof, thus to calculate the shortest path to reach the selected container 5.

[0101] The container 5 is selected based on the probability matrix and on the database of the separation rules activated on that container in the initial configuration in which the specific mapping of the containers 5 is set. Since the single container 5 has a given position which is fixed in space it is possible to calculate the distance thereof between each garbage item R and the correct disposal container 5; the garbage item R which will have a shorter path as the distance, that is, the closest one will be the candidate to be moved in the related container 5.

[0102] Due to the bounding box the arm 41 will be able to position itself on the opposite side of the garbage item R with respect to the container 5 and to move the garbage item R until it reaches the specific container 5.

[0103] The main algorithms are the calculation of the distance between two points in the plane, the calculation of the centre of a bounding box, the calculation of the opposite side of the bounding box with respect to the container 5.

[0104] By means of these calculations, it is possible to transfer through a GCODE protocol the instructions to the arm 41.

[0105] It is also possible to transfer the series of points through which the arm 41 has to move, by means of a programming interface.

[0106] The arm 41 will position itself parallel to the opposite side on the bounding box of the garbage item R with respect to the container 5, and positioning the arm 41 from said point on said side and the rotation of the blade 42 parallel to the side of the garbage item R, the arm 41 will move until the specific container 5, through the delta interface.

[0107] Moreover, the movement in case of coincident paths is also managed.

[0108] In fact, a deviation on a third point next to the bearing plate 3, 3' is provided, calculated as the closest free angle, in case there is another garbage item R on the path to reach the container 5 of each garbage item R which is present on the bearing plate 3, 3', otherwise the free garbage item R with the shorter distance will be discarded, since a path passing through a garbage item R is considered as invalid, since colliding with another garbage item R on the bearing plate 3, 3', not considering as correct paths which are incident with those of other garbage R, the shorter path between the garbage item R and the container 5 is another non-incident path.

[0109] An alternative mode for implementing the separating step 120 is to use an artificial intelligence algorithm which provides the points to be communicated to the actuator member 4, in order to obtain a more efficient sorting.

[0110] The automatic device D comprises a plurality of sensors for collecting and transmitting information about the amount, quality and typology of the collected garbage R.

[0111] Said data can be used and then processed with different aims such as: optimizing the collection, monitoring the consumptions and the production levels, monitoring the actual recycling activities estimating the real recycling impact. Moreover, the data analysis can assist in identifying the areas with greater pollution problems and in developing public policies to face these problems. The selective collection can thereby become more efficient and sustainable, improving people's quality of life.

[0112] The directly detectable or derived and analysable data are for example: number of separated garbage R; number of garbage R for each category; number of compressions performed by the compression device 6; number of openings of the containers of the collection section 5; number of average emptying operations of the collection section 5; number of produced per capita garbage R; kgs of collected garbage R; total weight of garbage R per category; average weight of garbage R per category and the like.

[0113] Although only selected embodiments have been chosen to illustrate the present invention, it will be apparent to the persons skilled in the art from this disclosure that several changes and modifications can be made herein without departing from the scope of the invention as defined in the attached claims.

Examples

first embodiment

[0042]In particular, in a first embodiment represented in figures 5 and 6, said bearing plate 3 is equipped with a first movable portion or side panel 31, a second side panel 32, a third side panel 33 and a fourth side panel 34, arranged along the four sides of the perimeter of the bearing plate 3.

[0043]Said first 31, second 32, third 33 and fourth 34 side panels are able to slide along said vertical axis Z, to allow the garbage to fall in the collection section 5.

[0044]Said actuator member 4, schematically represented in figures 1-3 and 6, is internally fixed in said upper portion 11 of said support structure 1, and it is able to push the garbage R lying on the bearing plate 3 and let it fall towards the collection section 5, based on the category thereof.

[0045]Preferably, said actuator member 4 is a robotic arm of the delta type having four degrees of freedom on said first horizontal axis X, second horizontal axis Y and vertical axis Z, and on itself, and it is moved by four elect...

second embodiment

[0062]said bearing plate 3' is shown in figure 7, comprising a first portion 31', a second portion 32', a third portion 33' and a fourth portion 34', which are all movable.

[0063]Each portion 31', 32', 33' and 34' is geometrically one quarter of the bearing plate 3' and it corresponds to the size of the relevant collection section 5, in particular of the corresponding container 51, 52, 53 54, so that the garbage falls just in the target space. The garbage R is thereby constrained to fall just in the collection container and it does not fall in empty spaces inside the support structure 1.

[0064]The operation of the automatic device D, which is the object of the present invention, is the following.

[0065]When a user has to dispose a garbage item R, he inserts it through the opening 13. The garbage item R runs through the chute 131, until it rests on the bearing plate 3.

[0066]The garbage R resting on the bearing plate 3, is thus processed according to a method 100. At first, it is located...

Field of application

[0001] The present invention relates to a device for collecting and separating garbage, in particular a device which is able to collect the garbage and automatically separate it based on the material which is present in the garbage.

[0002] These devices are present in everyday life, but they do not allow the automatic garbage separation, rather they are solely used for the collection thereof.

[0003] More particularly, the invention relates to a device, which is able to recognize the object inserted for the collection, recognize the material which forms that object so as to categorize and finally separate it in an appropriate container based on the obtained and processed data.

[0004] The device can be mainly used in public places even outdoors, where there is a little attention to recycling and garbage management.Prior art

[0005] At present, the increase in garbage production represents one of the main environmental challenges, making the management thereof a priority for environmental sustainability. Therefore, devices used for the garbage selective collection are already known.

[0006] As a consequence of a raising awareness among the population on the issue of garbage selective collection, the use of a multiplicity of devices for collecting garbage, one for each type of material that has to be separated, has become widespread.

[0007] Although substantially responding to the field needs, the devices according to the prior art still have some drawbacks which are unsolved to date, related in particular to human error during the garbage separation based on the regulations in force and on the need for a plurality of containers, one for each material that has to be separated.

[0008] An aim of the present invention is to provide a device for collecting garbage which is able to receive more garbage at the same time and collect it separately based on the composition material, so as to make the selective collection more efficient and sustainable, improving people's quality of life.

[0009] A further aim of the present invention is to comply with the regulations which provide penalties in case of wrong separation and billing based on the amount of non-recyclable garbage being produced.

[0010] Yet, an aim of the present invention is to provide a device for collecting garbage, in which there is no need to frequently empty the garbage contained therein.

[0011] Finally, an aim of the present invention is to provide a device for collecting garbage which is reliable, robust and safe.

[0012] The technical problem underlying the present invention is therefore to solve the above-mentioned drawbacks of the prior art, providing a device which allows the collected garbage to be separated and deposited in the suitable container.Summary of the invention

[0013] The solution idea underlying the present invention is to provide an automatic device for collecting and separating garbage, based on the recognition of the category which the garbage item belongs to.

[0014] Based on the above solution idea, the technical problem is solved by an automatic device for collecting and separating garbage, made of only one material or more materials, comprising a hollow-inside support frame for letting said device rest on a surface, in which at least one opening is obtained for the insertion of said one or more garbage and a collection section comprising at least one first and one second container, which are able to house the separated garbage, comprising an acquisition and processing unit, arranged in said support frame which is able to: acquire the images of said one or more garbage, detect in said acquired images predetermined visual features of said one or more garbage, associate with said visual features a category which said materials belong to, classify said garbage based on said association, send signals corresponding to said classification and a bearing plate, equipped with at least one movable portion, arranged elevated above said at least one first and one second container, which is able to receive said one or more garbage coming from said opening, which is able to receive said corresponding signals from said acquisition and processing unit and which is able to move said at least one portion, for the passage of said one or more garbage towards the respective at least one first and second container, based on said corresponding signals, and at least one actuator member, arranged in said support frame, elevated above said bearing plate, which is able to receive said corresponding signals from said acquisition and processing unit and to cause the passage of said one or more garbage from said at least one portion towards the respective at least one first and second container, based on said corresponding signals.

[0015] Furthermore, according to the invention, said at least one movable portion of said bearing plate is a first, second, third and fourth side panel, being arranged along the four sides of the perimeter of said bearing plate and able to slide along a vertical axis, to allow said garbage to fall in said collection section.

[0016] Still according to the invention, said at least one movable portion of said bearing plate is geometrically one quarter of said bearing plate, which is able to slide along a vertical axis, and corresponds to the size of the relevant collection section, in particular of the corresponding container.

[0017] Always according to the invention, said actuator member comprises a telescopic arm, which is able to move along said vertical axis away from and towards said bearing plate and / or in a horizontal plane which is perpendicular to said vertical axis, and a push element associated with said telescopic arm, which is able to contact the garbage which is present on the bearing plate and push it towards the collection section.

[0018] Preferably according to the invention, said push element of said actuator member is a blade.

[0019] Furthermore, according to the invention, said actuator member is a delta arm or a Cartesian arm.

[0020] Still according to the invention, said push element of said actuator member is an air element, such as an aspirator, which is able to suck the garbage, or a blower, which is able to push the garbage.

[0021] Always according to the invention, a compression device, which is able to compress the garbage which is present in said collection section, is arranged in said collection section.

[0022] Preferably according to the invention, said compression device comprises a piston and a pair of rods, anchored to each other at one end, wherein said piston is able to move along said vertical axis, causing the movement of said pair of rods in a first vertical plane or in a second vertical plane.

[0023] Furthermore, according to the invention, said device comprises at least one sensor for detecting the loading of said collection section.

[0024] A further object of the present invention is a method for operating an automatic device for collecting and separating garbage, made of only one material or more materials, comprising a hollow-inside support frame, for letting said device rest on a surface, in which at least one opening is obtained for the insertion of said one or more garbage and a collection section comprising at least one first and one second container, which are able to house the separated garbage, comprising the following steps: (110). locating and recognizing said garbage inserted into said opening and sending corresponding signals, by means of an acquisition and processing unit; (120). sorting said garbage in said collection section, by means of an actuator member based on said corresponding signals.

[0025] Furthermore, according to the invention, said locating and recognizing step 110. comprises the following sub-steps: (111). locating said garbage item on a bearing plate arranged inside said device by means of a bounding box; (112). recognizing said garbage item from the bounding box and associating percentages of the material which it is made of by means of a convolutional neural network. Still according to the invention, said sorting step 120. comprises the following sub-steps: (121). training a convolutional neural network to associate an image of a garbage item with at least one percentage of the material contained therein; (122). separating the garbage in said collection section by means of said actuator member, based on said association. Always according to the invention, said method comprises an initialization step 101. in which, by means of the acquisition and processing unit, the image of said bearing plate in pixels is detected and the real size of said bearing plate in millimetres is mapped, according to a preset conversion factor, so that a portion in millimetres of said bearing plate corresponds to each pixel, and said image is stored as a background image.

[0026] Preferably according to the invention, the method comprises the step of acquiring successive images of said garbage comparing them with said background image to determine the presence of a garbage item on said bearing plate.Brief description of the drawings

[0027] Figure 1 shows a side sectional view of the automatic device D according to the present invention; Figure 2 shows a front sectional view of the automatic device D according to the present invention; Figure 3 shows a front partial sectional view of the automatic device D according to the present invention; Figure 4 shows a perspective view of the device of figure 1; Figure 5 shows a top schematic view of a first embodiment of a bearing plate of the device of figure 1; Figure 6 shows a perspective view of the first embodiment shown in figure 5; Figure 7 shows a top schematic view of a second embodiment of a bearing plate of the device of figure 1; Figure 8 shows a block diagram of the operation of the device of figure 1. Detailed description

[0028] With reference to figures 1 and 2 of the attached drawings, an automatic device for collecting and separating garbage R which can be made of only one material or more materials is indicated with D, mainly comprising a hollow-inside support structure 1, in which at least one opening 13 is obtained for the insertion of the garbage R, an image acquisition and processing unit 2, arranged in said support structure 1, for the recognition of the garbage R, a bearing plate 3, which is able to receive the garbage R and route it, also by means of at least one actuator member 4, in appropriate containers arranged in a collection section 5.

[0029] Said support structure 1 has a substantially parallelepiped elongated shape, whose prevailing dimension extends along a vertical axis Z, with respect to the ground, of a Cartesian system formed by a first horizontal axis X, which is perpendicular with respect to said vertical axis Z, and by a second horizontal axis Y, which is perpendicular with respect both to said first horizontal axis X and to said vertical axis Z, from an upper portion 11, towards a base portion 12.

[0030] The dimensions of the support structure 1 can be approximately of 600 mm in width, i.e. along said first horizontal axis X, 700 mm in depth, i.e. along said second axis Y and 1,600 mm in height, i.e. along said vertical axis Z.

[0031] With reference to figures 2-4, an opening 13 is obtained on said support structure 1, to let the garbage R in.

[0032] A chute 131, which is able to convey the garbage R towards the inside of the support structure 1, is associated with said opening 13.

[0033] A rear opening, not shown in the figures, is also obtained on said support structure 1 to access inside the device D in case of maintenance.

[0034] Said acquisition and processing unit 2, which is able to acquire the images and process them, is arranged in said upper portion 11, as will be described in detail below.

[0035] In particular, said acquisition and processing unit 2 comprises a microcontroller associated with a processing unit 21 equipped with an operating program or code of the device D and at least one camera 22, preferably two.

[0036] Without departing from the scope of protection of the present invention, it is additionally possible to insert a spectroscopic sensor of the NIR - near infrared type, or the like, whose output is processed through a known technique such as sensor fusion, as a supplement to the image acquisition and processing unit 2.

[0037] Said acquisition and processing unit 2 can also comprise one or more sensors, such as presence sensors for example, for detecting the garbage R inside the support structure 1.

[0038] The presence sensors can be for example infrared or ultrasonic sensors.

[0039] Referring in particular to figure 5, said bearing plate 3 is substantially a platform on which the garbage R fall and rest, after being inserted into said opening 13, and it is arranged elevated above said collection section 5.

[0040] Said bearing plate 3 is constantly framed by said camera 22.

[0041] Said bearing plate 3 comprises at least one movable portion for the passage of the garbage.

[0042] In particular, in a first embodiment represented in figures 5 and 6, said bearing plate 3 is equipped with a first movable portion or side panel 31, a second side panel 32, a third side panel 33 and a fourth side panel 34, arranged along the four sides of the perimeter of the bearing plate 3.

[0043] Said first 31, second 32, third 33 and fourth 34 side panels are able to slide along said vertical axis Z, to allow the garbage to fall in the collection section 5.

[0044] Said actuator member 4, schematically represented in figures 1-3 and 6, is internally fixed in said upper portion 11 of said support structure 1, and it is able to push the garbage R lying on the bearing plate 3 and let it fall towards the collection section 5, based on the category thereof.

[0045] Preferably, said actuator member 4 is a robotic arm of the delta type having four degrees of freedom on said first horizontal axis X, second horizontal axis Y and vertical axis Z, and on itself, and it is moved by four electric motors, in particular known motors of the "stepper motor nema 17" type characterized by a good speed and equally good movement precision.

[0046] Referring in particular to figure 6 said actuator member 4 comprises a telescopic arm 41 and a push element 42. Said telescopic arm 41 is able to move along said vertical axis Z away from and towards said bearing plate 3. Said telescopic arm 41 is also able to move in the horizontal plane XY, formed by said first horizontal axis X and said second horizontal axis Y. Said push element 42 is associated with said telescopic arm 41, so as to contact the garbage R which is present on the bearing plate 3 and push it towards the collection section 5.

[0047] Said push element 42 is preferably a blade or a paddle. However, said push element 42 can also be replaced by an air element, such as an aspirator, which is able to suck the garbage R, or a blower, which is able to push the garbage R.

[0048] Without departing from the scope of protection of the present invention, said actuator member 4 can be a Cartesian arm or another device which is suited to move objects.

[0049] Said collection section 5, shown in figures 1-6, is arranged below said bearing plate 3, along said vertical axis Z and is able to collect the garbage R which is made to fall from said bearing plate 3.

[0050] Said collection section 5 comprises at least one first container 51 and one second container 52. In this described embodiment, said collection section 5 comprises, in addition to said first 51 and second 52 container, a third container 53 and a fourth container 54.

[0051] The volumetric collection capacity of the containers is modulable, based on the materials to be collected. Preferably the available total volume is of 300 litres divided into said first 51, second 52, third 53 and fourth 54 containers; each one with a volume of 75 litres each.

[0052] For example, it is possible to provide two containers for paper collection and two containers for plastic collection. Without departing from the scope of protection of the present invention, it is possible to combine all the typologies of containers, so as to collect the garbage R made of all the existing materials, also based on the different local regulations in force for the garbage selective collection.

[0053] In an alternative embodiment, it is possible to insert a member for the initial collection, not shown in the figures, for temporarily collecting the garbage R in a part of the support structure 1, so as not to fill the bearing plate 3 and to allow time for the sorting device D to sort all the garbage R.

[0054] As can be seen in figure 4, said first 51, second 52, third 53 and fourth 54 containers are so shaped as to be tilting in pairs in the first vertical plane XZ, to allow a flap opening, the extraction of the suitably collected garbage R and the emptying of the containers themselves.

[0055] Without departing from the scope of protection of the present invention said first 51, second 52, third 53 and fourth 54 containers can be also tilting in pairs in the second vertical plane YZ.

[0056] A respective side panel 31, 32, 33, 34 overlooks each container 51, 52, 53, 54, so that the garbage falls from the bearing plate 3 just in the corresponding container. By way of example, the first side panel 31 is arranged at the first container 51, the second side panel 32 is arranged at the second container 52, the third side panel 33 is arranged at the third container 53 and the fourth side panel 34 is arranged at the fourth container 54.

[0057] Furthermore, said collection section 5 comprises at least one sensor 55 arranged in each of said first 51, second 52, third 53 and fourth 54 containers, for detecting the loading thereof. Said sensor 55 is a sensor of the ultrasonic type. Said sensors 55 are able to monitor the loading of said first 51, second 52, third 53 and fourth 54 containers and to send an alert signal to the acquisition and processing unit 2, so as to advise the manager of the automatic device D that a predetermined loading threshold of each container has been reached. The sensors 55 are ultrasonic sensors.

[0058] In addition, it is possible to insert weight sensors, or load cells, not shown in the figure, which are able to measure the weight of the garbage R which is present inside the containers, calibrated so as to advise the manager that a predetermined loading threshold of each container has been reached.

[0059] The set loading threshold both for said sensors 55 and for said weight sensors is modifiable at will by the manager.

[0060] For the container of wet or organic garbage R, it is also possible to perform a time control so as to avoid the emission of unpleasant odours.

[0061] Said collection section 5 further comprises a compression device 6 which is able to compress the garbage R in order to reduce the total volume thereof. Said compression device 6 comprises in turn a piston 61 and a pair of rods 62, anchored to each other at one end. Said piston 61 is able to move along said vertical axis Z, thus causing the movement of the pair of rods 62 in a first vertical plane XZ or in a second vertical plane YZ, and the subsequent compaction of the garbage R. Said compression device 6 thus allows the continuous operation of the automatic device D during the garbage compaction, since the shape of the pair of rods 62 allows the continuous fall of the garbage R without accumulation in the collection section 5 which would prevent the collection of the next garbage R.

[0062] A second embodiment of said bearing plate 3' is shown in figure 7, comprising a first portion 31', a second portion 32', a third portion 33' and a fourth portion 34', which are all movable.

[0063] Each portion 31', 32', 33' and 34' is geometrically one quarter of the bearing plate 3' and it corresponds to the size of the relevant collection section 5, in particular of the corresponding container 51, 52, 53 54, so that the garbage falls just in the target space. The garbage R is thereby constrained to fall just in the collection container and it does not fall in empty spaces inside the support structure 1.

[0064] The operation of the automatic device D, which is the object of the present invention, is the following.

[0065] When a user has to dispose a garbage item R, he inserts it through the opening 13. The garbage item R runs through the chute 131, until it rests on the bearing plate 3.

[0066] The garbage R resting on the bearing plate 3, is thus processed according to a method 100. At first, it is located and recognized, in a locating and recognizing step 110 and then separated, in a separating step 120. These two steps continuously alternate, firstly the position and the typology of the garbage R which is present on the bearing plate 3 are detected and then the actuator member 4 is operated, to separate a certain garbage item R, automatically selected and classified. The process is repeated until all the garbage R is correctly disposed in the collection section 5.

[0067] In the locating and recognizing step 110 the garbage R is located and individually classified, finding the position and the occupied area on said bearing plate 3, while in the separating step 120, the garbage R is sorted.

[0068] The locating and recognizing step 110 is performed by using convolutional neural networks (CNN), in particular by means of computer vision algorithms which are able to perform an object detection, which is able to locate a garbage item R, return the corresponding bounding box and associate a label with each bounding box, so that the garbage R can be classified with the corresponding label.

[0069] The algorithms used in the locating and recognizing step 110 are of the Multi-Stage detectors type which provide the bounding boxes.

[0070] The algorithms used in the classification and separating step 120 are of the Single-Stage detectors type, so as to locate and classify the garbage item R in parallel, at the same time.

[0071] The parameters entered in the algorithms installed in the devices are in a limited number, so as to occupy little space in the memory of the devices on which they are installed.

[0072] Without departing from the scope of protection of the present invention, the artificial intelligence models might be replaced with equivalent models, that is models with a different structure but which can similarly carry out the same task, allowing the continuous update to the current state of the art in the above scope, a possible replacement of the CNN is the Vision Transformer model.

[0073] The training data are both open-source datasets which are available on the web and proprietary datasets, internally collected to reinforce the less performing cases in terms of recognition precision.

[0074] The proprietary dataset comprises known images, coming from known sources of the open-source type and new images. The known images have been accurately selected from different datasets, not only limited to the garbage but also to objects of daily use. This choice is driven by the need to develop a machine learning algorithm which is able to recognize the objects, regardless of the original classification thereof. In order to increase the model ability to generalize in different scenarios, the backgrounds have been diversified in the images outside the device D, while for the images inside the device D, a uniform background reproducing the internal environment of the device D has been set, for ease of setup and control.

[0075] The objects with extremely different sizes in a single frame, the separation of adjacent or overlapping objects, and the recognition of transparent materials or of colours which blend with the background are difficult to detect. Therefore, the dataset has been enriched with direct shots on a recognition plate, using a smooth and isolated lighting to minimize environmental interferences.

[0076] The classification dataset is divided into two main subsets. The first subset S1 comprises a wide range of existing datasets, which can be accessed for commercial use, from which the relevant images have been extracted and cleaned. The cleaning process has involved the removal of non-relevant categories and of low-quality images, which are unsuitable for the training of effective algorithms. For some categories, a more specific remapping was required, for example, the conversion of images generically labelled as "plastic" in precise sub-categories such as "PET plastic bottle". Moreover, images containing multiple garbage have been cut out to create new images focused on single objects, thus improving the precision of the classification model.

[0077] The second subset S2 consists of proprietary images, accurately taken to cover gaps detected in the model testing step. Each object has been selected to represent specific recognition difficulties, such as the change in shape, size and material. For example, reflective metal objects and articles composed of translucent plastic have been photographed under controlled conditions to ensure that the model can effectively distinguish them under real operating conditions. The images have been taken with different angles and lighting configuration to test the model robustness under different lighting conditions and visual perspectives.

[0078] Preliminarily, the method 100 provides an initialization step 101, in which the automatic device D is preparing to operate. In particular, the automatic device D is calibrated and all the working parameters are set and all the components are arranged in the required position to perform the next operations, and then the automatic device D enters standby mode.

[0079] The acquisition and processing unit 2 detects the image of the bearing plate 3 in pixels and maps the real size of the bearing plate 3 in millimetres, according to a preset conversion factor, so that a portion in millimetres of the bearing plate 3 corresponds to each pixel. Moreover, the image of the empty bearing plate 3 is acquired and stored as a background, for further processing operations, said image will be defined hereafter as background image Is.

[0080] This step is not required when different methods, based on artificial intelligence, are used.

[0081] Afterwards, the locating and recognizing step 110 begins, during which the garbage which is present on the bearing plate 3 is identified and recognized.

[0082] In particular, in a first subset 111, the detection and the localization of the presence of the garbage item R on the bearing plate 3 occur. For this purpose, a time pulse or trigger enables at preset time intervals the acquisition of the images by means of said camera 22. Starting from the background image Is, the next images I F which are compared with the background image Is by means of a background subtraction algorithm are acquired.

[0083] As soon as the camera 22 acquires a next image I F in which there is a garbage item R, a bounding box B which defines the outlines of the garbage item R is created on the image.

[0084] An alternative to the above-described method can provide the use of a presence sensor, for example an ultrasonic or passive infrared - PIR - one, arranged inside the support structure 1, by which the presence of the garbage item R is directly detected.

[0085] When the number of bounding boxes B is higher than 0, it means that a garbage item R has been found and it is possible to proceed with the recognition thereof through the following second sub-step 112.

[0086] Still another alternative to the above-described method is represented by a method based on the artificial intelligence instructed to recognize the garbage R starting from the image detected by the camera 22 and to output the bounding boxes of the garbage R or, as an alternative, to segment the image.

[0087] Afterwards, in the second sub-step 112 the garbage R located in the first sub-step 111 is recognized or classified, i.e. it is determined of which material the garbage item R is composed.

[0088] For this second sub-step 112 a convolutional neural network is used, in which for each cut-out image of the single garbage item R at the input, the value of the percentage of probability that the garbage item R of the image is made of a certain material is provided at the output, for example: metal = 0,7; glass = 0,1; paper = 0,2.

[0089] Alternatively, it is possible to associate a label with a product brand, or with a specific object. The allocation of the labels is dynamic and variable.

[0090] The use of the percentages associated with each category of garbage R allows to classify the garbage R in a higher way, since, for example, if a garbage item R is classified with 90% of metal and 10% of dangerous garbage, it is possible to separate the garbage item R as dangerous and not as metal. Or, if a garbage item R is classified with 60% of plastic and 40% of paper, it is possible to separate the garbage item R as unselective.

[0091] These settings can be modified in a flexible manner.

[0092] Finally, if all the categories have a low recognition percentage, it is possible to allocate the category "other" and arrange the collection container "other".

[0093] After identifying and classifying the garbage item R in step 110, the separating step 120 is performed, in which the actuator member 4 receives the instructions from said acquisition and processing unit 2 for the correct separation of the garbage R. The instructions are in the form of commands and coordinates, which are similar to the known GCODE algorithm, used by 3D printers.

[0094] The separating step 120 preliminarily provides a first training sub-step 121, using simulation libraries in which the environment, i.e. the internal part of the support structure 1, the actuator member 4 and the garbage R are modelled, in a finite number, trying to represent as much as possible all the existing garbage R.

[0095] The first training sub-step 121 provides the determination of a reward function, i.e. of the value function. The two important aspects of the training step are the correctness of the separation of the garbage item R and the routing time of the garbage item R in the collection section 5.

[0096] The separation correctness takes priority with respect to the routing time, therefore in the reward function, the separation correctness will have a higher score with respect to the routing time.

[0097] For the reward function a negative score is also determined if the routing time exceeds a certain threshold value. The greater the routing time will be with respect to the threshold value, the greater the negative score which is subtracted from the final score will be.

[0098] If the garbage is correctly separated, the penalty will have the value 0. If the garbage is wrongly separated, in addition to the negative score of the routing time, a further negative value is also added, for example a value which is equal to - 1000.

[0099] Afterwards, in a second separating sub-step 122, the acquisition and processing unit 2 performs an algorithm which calculates the garbage item R arranged closer to the respective movable side panel 31, 32, 33, 34, based on the percentages of material received from said second recognizing sub-step 112, on the settings of the collection section 5, and determines the straight line from the centre of the garbage item R to the closest side panel, turning this information into commands for the actuator member 4.

[0100] In particular, having the bounding box available, that is the object position, it is possible to calculate the centre thereof, thus to calculate the shortest path to reach the selected container 5.

[0101] The container 5 is selected based on the probability matrix and on the database of the separation rules activated on that container in the initial configuration in which the specific mapping of the containers 5 is set. Since the single container 5 has a given position which is fixed in space it is possible to calculate the distance thereof between each garbage item R and the correct disposal container 5; the garbage item R which will have a shorter path as the distance, that is, the closest one will be the candidate to be moved in the related container 5.

[0102] Due to the bounding box the arm 41 will be able to position itself on the opposite side of the garbage item R with respect to the container 5 and to move the garbage item R until it reaches the specific container 5.

[0103] The main algorithms are the calculation of the distance between two points in the plane, the calculation of the centre of a bounding box, the calculation of the opposite side of the bounding box with respect to the container 5.

[0104] By means of these calculations, it is possible to transfer through a GCODE protocol the instructions to the arm 41.

[0105] It is also possible to transfer the series of points through which the arm 41 has to move, by means of a programming interface.

[0106] The arm 41 will position itself parallel to the opposite side on the bounding box of the garbage item R with respect to the container 5, and positioning the arm 41 from said point on said side and the rotation of the blade 42 parallel to the side of the garbage item R, the arm 41 will move until the specific container 5, through the delta interface.

[0107] Moreover, the movement in case of coincident paths is also managed.

[0108] In fact, a deviation on a third point next to the bearing plate 3, 3' is provided, calculated as the closest free angle, in case there is another garbage item R on the path to reach the container 5 of each garbage item R which is present on the bearing plate 3, 3', otherwise the free garbage item R with the shorter distance will be discarded, since a path passing through a garbage item R is considered as invalid, since colliding with another garbage item R on the bearing plate 3, 3', not considering as correct paths which are incident with those of other garbage R, the shorter path between the garbage item R and the container 5 is another non-incident path.

[0109] An alternative mode for implementing the separating step 120 is to use an artificial intelligence algorithm which provides the points to be communicated to the actuator member 4, in order to obtain a more efficient sorting.

[0110] The automatic device D comprises a plurality of sensors for collecting and transmitting information about the amount, quality and typology of the collected garbage R.

[0111] Said data can be used and then processed with different aims such as: optimizing the collection, monitoring the consumptions and the production levels, monitoring the actual recycling activities estimating the real recycling impact. Moreover, the data analysis can assist in identifying the areas with greater pollution problems and in developing public policies to face these problems. The selective collection can thereby become more efficient and sustainable, improving people's quality of life.

[0112] The directly detectable or derived and analysable data are for example: number of separated garbage R; number of garbage R for each category; number of compressions performed by the compression device 6; number of openings of the containers of the collection section 5; number of average emptying operations of the collection section 5; number of produced per capita garbage R; kgs of collected garbage R; total weight of garbage R per category; average weight of garbage R per category and the like.

[0113] Although only selected embodiments have been chosen to illustrate the present invention, it will be apparent to the persons skilled in the art from this disclosure that several changes and modifications can be made herein without departing from the scope of the invention as defined in the attached claims.

Examples

first embodiment

[0042]In particular, in a first embodiment represented in figures 5 and 6, said bearing plate 3 is equipped with a first movable portion or side panel 31, a second side panel 32, a third side panel 33 and a fourth side panel 34, arranged along the four sides of the perimeter of the bearing plate 3.

[0043]Said first 31, second 32, third 33 and fourth 34 side panels are able to slide along said vertical axis Z, to allow the garbage to fall in the collection section 5.

[0044]Said actuator member 4, schematically represented in figures 1-3 and 6, is internally fixed in said upper portion 11 of said support structure 1, and it is able to push the garbage R lying on the bearing plate 3 and let it fall towards the collection section 5, based on the category thereof.

[0045]Preferably, said actuator member 4 is a robotic arm of the delta type having four degrees of freedom on said first horizontal axis X, second horizontal axis Y and vertical axis Z, and on itself, and it is moved by four elect...

second embodiment

[0062]said bearing plate 3' is shown in figure 7, comprising a first portion 31', a second portion 32', a third portion 33' and a fourth portion 34', which are all movable.

[0063]Each portion 31', 32', 33' and 34' is geometrically one quarter of the bearing plate 3' and it corresponds to the size of the relevant collection section 5, in particular of the corresponding container 51, 52, 53 54, so that the garbage falls just in the target space. The garbage R is thereby constrained to fall just in the collection container and it does not fall in empty spaces inside the support structure 1.

[0064]The operation of the automatic device D, which is the object of the present invention, is the following.

[0065]When a user has to dispose a garbage item R, he inserts it through the opening 13. The garbage item R runs through the chute 131, until it rests on the bearing plate 3.

[0066]The garbage R resting on the bearing plate 3, is thus processed according to a method 100. At first, it is located...

Claims

1. An automatic device (D) for collecting and separating garbage (R), made of only one material or more materials, comprising a hollow-inside supporting frame (1), for letting said device (D) rest on a surface, in which at least one opening (11) is obtained for the insertion of said one or more garbage (R) and a collection section (5) comprising at least one first (51) and one second container (52), which are able to house the separated garbage (R), characterized in that it comprises: - an acquisition and processing unit (2), arranged in said supporting frame (1) which is able to: acquire the images of said one or more garbage (R), detect in said acquired images predetermined visual features of said one or more garbage (R), associate with said visual features a category which said materials belong to, classify said garbage (R) based on said association, send signals corresponding to said classification; and - a bearing plate (3,3'), equipped with at least one movable portion (31, 31', 32, 32', 33, 33', 34, 34'), arranged elevated above said at least one first (51) and one second (52) container, which is able to receive said one or more garbage (R) coming from said opening (11), which is able to receive said corresponding signals from said acquisition and processing unit (2) and which is able to move said at least one portion (31, 31', 32, 32', 33, 33', 34, 34'), for the passage of said one or more garbage (R) towards the respective at least one first (51) and one second container (52), based on said corresponding signals; - at least one actuator member (4), arranged in said supporting frame (1), elevated above said bearing plate (3, 3'), which is able to receive said corresponding signals from said acquisition and processing unit (2) and to cause the passage of said one or more garbage (R) from said at least one portion (31, 31', 32, 32', 33, 33', 34, 34') towards the respective at least one first (51) and one second container (52), based on said corresponding signals.

2. The automatic device (D) according to claim 1, characterized in that said at least one movable portion (31, 32, 33, 34) of said bearing plate (3) is a first (31), second (32), third (33) and fourth (34) side panel, arranged along the four sides of the perimeter of said bearing plate (3) and able to slide along a vertical axis (Z), to allow said garbage (R) to fall in said collection section (5).

3. The automatic device (D) according to claim 1, characterized in that said at least one movable portion (31', 32', 33', 34') of said bearing plate (3') is geometrically one quarter of said bearing plate (3'), which is able to slide along a vertical axis (Z), and corresponds to the size of the relevant collection section (5), in particular of the corresponding container (51, 52, 53, 54).

4. The automatic device (D) according to any one of the previous claims, characterized in that said actuator member (4) comprises a telescopic arm (41), which is able to move along said vertical axis (Z) away from and towards said bearing plate (3, 3') and / or in a horizontal plane (XY) which is perpendicular to said vertical axis (Z), and a push element (42) associated with said telescopic arm (41), which is able to contact the garbage (R) which is present on the bearing plate (3, 3') and push it towards the collection section (5).

5. The automatic device (D) according to the previous claim, characterized in that said pushing element (42) of said actuator member (4) is a blade.

6. The automatic device (D) according to any one of the previous claims, characterized in that said actuator member (4) is a delta arm or a Cartesian arm.

7. The automatic device (D) according to any one of the previous claims, characterized in that said pushing element (42) of said actuator member (4) is an air element, such as an aspirator, which is able to suck the garbage (R), or a blower, which is able to push the garbage (R).

8. The automatic device (D) according to any one of the previous claims, characterized in that a compression device (6), which is able to compress the garbage (R) which is present in said collection section (5), is arranged in said collection section (5).

9. The automatic device (D) according to the previous claim, characterized in that said compression device (6) comprises a piston (61) and a pair of rods (62), anchored to each other at one end, wherein said piston (61) is able to move along said vertical axis (Z), causing the movement of said pair of rods (62) in a first vertical plane (XZ) or in a second vertical plane (YZ).

10. The automatic device (D) according to any one of the previous claims, characterized in that it comprises at least one sensor (55) for detecting the loading of said collection section (5).

11. A method (100) for operating an automatic device (D) for collecting and separating garbage (R), made of only one material or more materials, comprising a hollow-inside support frame (1), for letting said device (D) rest on a surface, in which at least one opening (11) is obtained for the insertion of said one or more garbage (R) and a collection section (5) comprising at least one first (51) and one second container (52), which are able to house the separated garbage (R), according to any one of the previous claims 1-10, comprising the following steps: (110). locating and recognizing said garbage (R) inserted into said opening (11) and sending corresponding signals, by means of an acquisition and processing unit (2); (120). sorting said garbage (R) in said collection section (5), by means of an actuator member (4) based on said corresponding signals.

12. The method (100) according to the previous claim, characterized in that said locating and recognizing step 110. comprises the following sub-steps: (111). locating said garbage item (R) on a bearing plate (3, 3') arranged inside said device (D) by means of a bounding box; (112). recognizing said garbage item (R) from the bounding box and associating percentages of the material which it is made of by means of a convolutional neural network.

13. The method (100) according to any one of claims 11 or 12, characterized in that said sorting step (120) comprises the following sub-steps: (121). training a convolutional neural network to associate an image of a garbage item (R) with at least one percentage of the material contained therein; (122). separating the garbage in said collection section (5) by means of said actuator member (4), based on said association.

14. The method (100) according to any one of claims 11-13, characterized in that it comprises an initialization step 101. in which, by means of the acquisition and processing unit (2), the image of said bearing plate (3) in pixels is detected and the real size of said bearing plate (3) in millimetres is mapped, according to a preset conversion factor, so that a portion in millimetres of said bearing plate (3), corresponds to each pixel, and said image is stored as a background image (Is).

15. The method (100) according to the previous claim, characterized in that it comprises a step of acquiring further images (IF) of said garbage (R) and comparing them with said background image (Is) to determine the presence of a garbage item (R) on said bearing plate (3, 3').

Images