Method and system for waste tracking and handling in a waste processing facility

Abstract

The present disclosure relates to a computer-implemented method and an electronic system for locating waste objects (O) in a waste processing facility (5), the method comprising the steps of: receiving an image of the waste storage area (51); processing (S102), using a waste object recognition module, the received image of the waste storage area (51) to identify one or more waste objects (O) in the image, and generate a representation of each identified waste object (O) and a location, in the image, of each identified waste object (O); determining a three-dimensional position of each identified waste object (O) in the waste storage area (51); and storing the three-dimensional position the identified waste objects (O), the three-dimensional position of a particular identified waste object (O) linked to the representation of the particular identified waste object (O).

Description

[0001] P29129PC00 12.12.2025

[0002] 1

[0003] METHOD AND SYSTEM FOR WASTE TRACKING AND HANDLING IN A WASTE PROCESSING FACILITY

[0004] FIELD OF THE DISCLOSURE

[0005] The present disclosure relates to a computer-implemented method and a system for locating waste objects in a waste processing facility.

[0006] BACKGROUND OF THE DISCLOSURE

[0007] Waste processing facilities receive waste from a large number of waste sources, the waste typically being deposited or dumped into a waste storage area, such as a waste pit, for intermediary storage prior to further processing. The further processing may include, for example, incineration of the waste.

[0008] The waste in the waste storage area is typically carried to the incinerator, or a hopper which feeds the incinerator, by means of a bucket or grab crane mounted above the waste storage area. In existing waste processing facilities, the crane operator constantly surveils the waste storage area, ensuring that waste items unsuitable for incineration are identified and separated. Such waste items include items too bulky to be incinerated, or hazardous items such as gas tanks.

[0009] Recently, systems have been developed in which cameras monitor the waste storage area to identify waste which may be hazardous. The images from the cameras are processed using vision systems designed to detect hazardous waste.

[0010] A drawback of such known systems is that they are only able to identify waste which is located on the surface of the waste storage area.

[0011] SUMMARY OF THE DISCLOSURE

[0012] It is an object of this invention to provide methods and systems for locating waste objects in a waste storage area of a waste processing facility. In particular, the invention relates P29129PC00 12.12.2025

[0013] 2 to methods and systems for locating waste objects in a waste storage area of a waste processing facility which overcome one or more drawbacks of the prior art, in particular in that they enable the locating of a waste object which may be covered and obscured by other waste objects.

[0014] According to the present invention, these objects are achieved through the features of the independent claims. In addition, further advantageous embodiments follow from the dependent claims and the description.

[0015] The invention relates to a computer-implemented method for locating waste objects in a waste processing facility. The method comprises a plurality of steps, some of which may be option. The steps include receiving, for example, from a waste storage area sensor system configured to monitor waste in the waste storage area, an image of the waste storage area. The steps include processing, using a waste object recognition module, the received image of the waste storage area. The processing using the waste object recognition module comprises identifying one or more waste objects in the image. The processing comprises generating, as an output, a representation of at least one identified waste object and a location, in the image, of the at least one identified waste object. The steps include determining, using the location of the at least one identified waste object in the image, a position of the at least one identified waste object in the waste storage area, for example a three-dimensional position. The steps include storing the position of the at least one identified waste object, the position linked to the representation of the at least one identified waste object.

[0016] The invention also relates to a computer-implemented method for locating waste objects in a waste processing facility. The method comprises a plurality of steps. The steps include initializing a three-dimensional model of a waste storage area in a waste processing facility, the three-dimensional model defining physical boundaries of the waste storage area. The steps include receiving, from a waste storage area sensor P29129PC00 12.12.2025

[0017] 3 system configured to monitor waste in the waste storage area, an image of the waste storage area. The steps include processing, using a waste object recognition module, the received image of the waste storage are. The processing comprises identifying one or more waste objects in the image. The processing comprises generating, as an output, a representation of each identified waste object and a location, in the image, of each identified waste object. The steps of the method include determining, using the location of each identified waste object in the image and the three-dimensional model of the waste storage area, a three-dimensional position of each identified waste object in the waste storage area. The steps include storing the three-dimensional position of the identified waste objects, the three-dimensional position of a particular identified waste object linked to the representation of the particular identified waste object.

[0018] By storing the three-dimensional position, the invention thereby retains the position of the waste object even if the waste object is covered by new waste. This increases the reliability and quality of locating and tracking the waste objects over time.

[0019] In an embodiment, the method further comprises generating a control signal for controlling an actuator of the waste processing facility, using the stored three- dimensional position of the identified waste objects. The actuator of the waste processing facility may be, for example, a crane of the waste processing facility. The control signal may be configured to control the movement or action of the crane of the waste processing facility, for example to grab and / or avoid grabbing the one or more identified waste objects.

[0020] In an embodiment, the method comprises receiving, from the waste storage area sensor system, a sequence of images. The method comprises processing, using the waste object recognition module, the sequence of images. The method comprises tracking the three-dimensional position of each identified waste object over time, by matching the representation of the identified waste object in a particular image in the sequence of P29129PC00 12.12.2025

[0021] 4 images of the waste storage area to a representation of a previously identified waste object in a previous image in the sequence of images. Thereby, the embodiment allows for tracking a particular identified waste object across a sequence of images, to enable more robust tracing and logging of particular identified waste objects.

[0022] In an embodiment, the method comprises changing, in particular updating, the generated three-dimensional position of a particular identified waste object if the particular identified waste object is found at a different position in a current image than in a previous image. The method comprises not updating, but rather maintaining, the generated three- dimensional position of the particular identified waste object if the waste object is found at the same position in the current image as in the previous image. The method may further comprise not updating, but rather maintaining, the generated three-dimensional position if the waste object is not identified in the current image (for example, due to being covered or otherwise obscured).

[0023] In other words, the method provides for temporal waste object persistence or permanence, tracking a particular waste object across a sequence of images and reidentifying the waste object in each image in which it appears. One of the advantages of this method is that it allows tracking of a particular identified waste object across time, for example to be able to determine how long a particular identified waste object remains in the waste storage area, how often it is moved, etc. Further, it allows for re-identification of a particular waste object which may be temporarily obscured or covered by other waste objects, which avoids inadvertent double-counting.

[0024] In an embodiment, the method further comprises processing, using the waste object recognition module, the received image of incoming waste, wherein the waste object recognition module is further configured to classify each waste object into one or more of the following waste classes: oversized, hazardous, or combustible. The method comprises filtering the identified waste objects by one or more of the waste classes. The P29129PC00 12.12.2025

[0025] 5 method comprises generating a three-dimensional map of the filtered waste objects, using the three-dimensional position of the filtered waste objects. This allows a crane operator to avoid or target particular areas of the waste storage area, allows the crane operator to easily see and trace particular waste objects, for example for easier removal from the waste storage area.

[0026] In an example, the three-dimensional map may be used to generate or determine “grab” or “no grab” areas, to guide the crane operator or an automated crane control system.

[0027] The three-dimensional map also be used to generate a list of oversized and / or hazardous waste objects present in the waste storage area.

[0028] The waste objects may be further classified according to material type (e.g., wood, metal, plastic), size (e.g., absolute size or volume, and / or relative size, for example as compared to a crane bucket), hardness / softness (e.g. substantially incompressible or compressible), or shape (e.g., a spherical or rounded object, a square or rectangular object, or an elongated object).

[0029] In an embodiment, the method comprises processing, using the waste object recognition module, the received images of the waste storage area. The processing uses the received images of the storage area as input. The processing, using the waste object recognition module, comprises identifying one or more waste objects in the one or more images, wherein the waste objects include one or more aggregate waste materials comprised of a plurality of particles of a similar material. The processing comprises generating, as an output, a representation of each identified aggregate waste material and generating a location, in the one or more images, of each identified aggregate waste material. The processing comprises determining, using the location of each identified aggregate waste material in the image and the three-dimensional model of the waste storage area, a three-dimensional space occupied by each identified aggregate waste P29129PC00 12.12.2025

[0030] 6 material in the waste storage area. The processing comprises storing the three- dimensional space of the identified aggregate waste materials, the three-dimensional space of a particular identified aggregate waste material linked to the representation of the particular identified aggregate waste material.

[0031] The three-dimensional space occupied by each identified aggregate waste material may be updated over time using physically based modeling, in particular using particle modelling.

[0032] In an embodiment, the method further comprises receiving a three-dimensional position of a crane bucket of a crane of the waste processing facility. The method comprises receiving a grabbing waste signal, the grabbing waste signal indicative of the crane bucket grabbing waste, in particular grabbing waste objects. The method comprises determining one or more grabbed waste objects grabbed by the crane bucket, by matching the three-dimensional position of the crane bucket with the three-dimensional position of waste objects. The method comprises receiving a dropping waste signal, the dropping waste signal indicative of the crane bucket dropping waste. The method comprises updating the three-dimensional position of the grabbed waste objects using the three-dimensional position of the crane bucket associated with the dropping waste signal.

[0033] The updated three-dimensional position of the grabbed waste may be in the waste storage area or outside it, for example in a shredder, hopper, or incinerator.

[0034] In addition to the grabbing waste signal and the dropping waste signal, the crane signal may include: a shredding waste signal, a mixing waste signal, a dropping waste into an incinerator signal, or a dropping waste into an auxiliary container signal. P29129PC00 12.12.2025

[0035] 7

[0036] In an embodiment, the method further comprises receiving one or more of the following crane action states: a drop for incineration action state or a waste rejection action state, the drop for incineration action state indicative of the crane bucket dropping waste for incineration and the waste rejection action state indicative of the crane bucket dropping waste into a reject area. The method comprises determining one or more dropped waste objects dropped by the crane bucket by matching the three-dimensional position of the crane bucket with the three-dimensional position of waste objects at a time-point indicated by the drop for incineration signal. The method comprises updating a status associated with dropped waste objects according to the crane action state. The list of oversized and / or hazardous objects may be updated based on the crane signal.

[0037] In an embodiment, the method comprises receiving, from a crane control system, the three-dimensional position of a crane bucket of the crane. The method comprises receiving, from the crane control system, at least one of the following crane action states: the grabbing waste action state or the dropping waste action state.

[0038] In an embodiment, the method further comprises processing, using a crane tracking module, the received image of the waste storage area. The processing comprises identifying, using the crane tracking module, the crane bucket in the image. The processing comprises generating as an output a location of the identified crane bucket in the image and generating at least one of the following crane action states using the identified crane bucket: the grabbing waste action state or the dropping waste action state. The processing comprises determining, using the location of the crane bucket in the image and the three-dimensional model of the waste storage area, a three- dimensional position of the crane bucket in the waste storage area.

[0039] The advantages of considering the crane bucket include that it allows tracking of waste objects even when the waste object may be at least temporarily not visible, for example due to being carried by the bucket. It also allows to track an “in-flight” movement of the P29129PC00 12.12.2025

[0040] 8 waste object by associating the crane bucket’s movement with the movement of the waste object.

[0041] In an embodiment, the method further comprises receiving, from an exhaust gas analyzer configured to analyze exhaust gas of the waste processing facility, a gas analyzer signal at a reading time-point. The method comprises determining, using the gas analyzer signal, the presence of one or more pollutants in the exhaust gas at the reading time-point. The method comprises determining incinerated waste objects dropped for incineration within a defined time period prior to the reading time-point. The method comprises associating the one or more pollutants with one or more incinerated waste objects.

[0042] In an embodiment, the method further comprises receiving, from an auxiliary sensor system configured to monitor incoming waste in the waste processing facility, images of incoming waste. The method comprises processing, using the waste object recognition module, the received images of incoming waste to identify one or more waste objects in the images of the incoming waste. The method comprises assigning the one or more identified waste objects to a three-dimensional position in the waste storage area.

[0043] The identified waste objects may be assigned to the three-dimensional position using a defined position of the entry port of the incoming waste in the three-dimensional model of the waste storage area. In case of a plurality of defined entry ports (e.g., multiple unloading bays), the defined entry port may be selected / identified from amongst the plurality of defined entry ports using an identifier of the auxiliary sensor system, or by using the waste storage area sensor system which may be configured to detect the incoming waste stream.

[0044] The incoming waste stream may flow on a conveyer belt or an inclined surface, for example. P29129PC00 12.12.2025

[0045] 9

[0046] In case of a plurality of defined entry ports (e.g., multiple unloading bays), the defined entry port may be selected / identified, by the processor, from amongst the plurality of defined entry ports using an identifier of the auxiliary sensor system, or by using the waste storage area sensor system, which may be configured to detect the incoming waste stream.

[0047] The images from the auxiliary sensor system may be used to classify the waste objects.

[0048] In an embodiment, the method further comprises receiving a waste provider identifier of a waste provider associated with a particular delivery time-point, the waste provider providing the waste objects in the incoming waste stream to the waste processing facility. The method comprises associating the one or more identified waste objects with the waste provider identifier, using a time-point associated with the images of incoming waste and the delivery time-point. Thereby, any identified non-compliant, hazardous or oversized waste objects may be associated with the waste provider.

[0049] In an embodiment, the method further comprises associating one or more pollutants with the waste provider identifier, using the association between the one or more identified waste objects and the waste provider identifier and, upon the identified waste objects having been dropped for incineration, the association between the one or more pollutants and one or more incinerated waste objects. This allows the tracing of pollutants to the waste provider, using the waste provider identifier, enabling the identification of waste providers who may inadvertently be providing hazardous or polluting waste for incineration.

[0050] In an embodiment, the method further comprises training the waste object recognition module using a training dataset and a machine learning algorithm, wherein the training dataset comprises a plurality of images of waste storage areas with classified waste objects. P29129PC00 12.12.2025

[0051] 10

[0052] In addition to the method described herein, the present invention also relates to an electronic system comprising a processor configured to perform one of the methods as described herein.

[0053] In an embodiment, the electronic system further comprises a waste storage area sensor system configured to monitor waste in the waste storage area, the waste storage area sensor system connected to the processor and configured to transmit, to the processor, the sequence of images of the waste storage area.

[0054] In an embodiment, the electronic system further comprises an auxiliary sensor system configured to monitor incoming waste in the waste processing facility, the auxiliary sensor system connected to the processor and configured to transmit, to the processor, the images.

[0055] The present invention also relates to a computer program and a computer program product comprising computer-readable instructions configured to control a processor of an electronic system as described herein such that the processor performs the method as described herein.

[0056] The present invention also relates to a non-transitory memory storing computer readable instructions configured to control a processor of an electronic system as described herein such that the processor performs the method as described herein.

[0057] BRIEF DESCRIPTION OF THE DRAWINGS

[0058] The herein described disclosure will be more fully understood from the detailed description given herein below and the accompanying drawings, which should not be considered limiting to the invention described in the appended claims. The drawings in which: P29129PC00 12.12.2025

[0059] 11

[0060] Fig. 1 shows a schematic illustration of a waste processing facility, including a waste storage area implemented as a waste pit, an unloading area an incinerator, and a computer connected to a waste storage area sensor system, an auxiliary sensor system, an exhaust gas analyzer, and crane;

[0061] Fig. 2 shows a block diagram of an electronic system according to the invention comprising a computer, waste storage area sensor system and an optional auxiliary sensor system;

[0062] Fig. 3 shows a block diagram illustrating a waste object recognition module, in particular the inputs and outputs;

[0063] Fig. 4 shows a block diagram illustrating a crane tracking module, in particular the inputs and outputs;

[0064] Fig. 5 shows a schematic illustration of a three-dimensional model of a waste storage area, including waste objects, a waste surface, and an entry point of an unloading bay;

[0065] Fig. 6 shows a flow diagram illustrating a method comprising a series of steps for locating waste objects in a waste storage area; and

[0066] Fig. 7 shows a flow diagram illustrating a method comprising a series of steps for tracking waste objects in a waste storage area which have been moved by a crane bucket.

[0067] DESCRIPTION OF THE EMBODIMENTS

[0068] Reference will now be made in detail to certain embodiments, examples of which are illustrated in the accompanying drawings, in which some, but not all features are shown. Indeed, embodiments disclosed herein may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will satisfy applicable legal P29129PC00 12.12.2025

[0069] 12 requirements. Whenever possible, like reference numbers will be used to refer to like components or parts.

[0070] Figure 1 shows a schematic diagram of a waste processing facility 5. The waste processing facility 1 includes a waste storage area 51 in which waste W is deposited and temporarily stored prior to further processing. The waste storage area 51 may be realized as a waste pit or any other defined area, zone or space in which the waste W is collected.

[0071] The waste W may be deposited into the waste storage area 51 via one or more unloading bays 52, which are inclined slopes onto which a dump truck can deposit the incoming waste IW such that it slides into the waste storage area 51. The incoming waste IW can alternatively or additionally be provided into the waste storage area 51 by cranes, conveyer belts, or other mechanisms.

[0072] The waste W is comprised of waste objects O which are physical objects that reside in the waste storage area 51. The waste objects O may comprise solid waste, including items discarded by the public, such as household trash, food scraps, packaging materials, or garden waste. The waste objects O may comprise industrial waste, for example including scrap metals, concrete, wood, metals and insulation. The waste objects O may comprise electronic waste, including electronic devices such as computers, laptops, televisions, and smartphones. The waste objects O may include hazardous waste such as batteries, paints, chemicals, gas canisters, and medical waste.

[0073] The waste objects O may be in the form of items which are individually identifiable, such as pieces of metal, wood, gas canisters and so on. The waste objects O may further comprise aggregate waste W in which the individual pieces of waste W are no longer identifiable owing to their small size. Aggregate waste comprises materials such as broken concrete, brick and masonry, sand, ceramics, crushed stone, wood chips or P29129PC00 12.12.2025

[0074] 13 sawdust, or other waste W which has been ground, blended, chopped up or otherwise turned into small pieces.

[0075] The waste objects O may be classified into one or more of the classes or categories defined above. In particular, the waste classes may define waste which is unsuitable for incineration, such as hazardous waste, and oversized waste too large for the incinerator.

[0076] The waste objects O may be classified into labeled and pre-defined categories. The waste objects O, however, may additionally or alternatively be classified into categories established or defined by the waste object detection module during training (in particular using machine learning) of the waste object detection module in a label free or unsupervised manner, in which the waste object detection module may be configured to establish a classification system including a set of classes of waste objects O. The set of classes may be established to allow for a defined distinction between the waste objects O based on visually apparent features of the waste objects O and / or based on an inferred type and / or an inferred properties of the waste objects O.

[0077] The waste objects O may additionally or alternatively be classified using a separate pretrained image classification model which is pre-trained to classify images or objects featured in the images into one or more pre-defined categories or classes. For example, a two-stage object detection pipeline may be used including the trained waste object detection module and in addition including a separate pre-trained image classification model.

[0078] In an example, if the waste object detection module detects a waste object O however fails to classify the waste object O with a sufficiently high likelihood, the image or part thereof including the waste object O may be provided to a separate model, in particular a second model trained on a larger training dataset of images and labels, for example a zero shot model, a Visual Language Model (VLM), a YOLO- World, CLIP, or SAM3 which P29129PC00 12.12.2025

[0079] 14 may be computationally more expensive to run during inference, however can provide a more robust or accurate classification of a larger number of waste objects O.

[0080] The waste objects O in the waste storage area 51 may be forwarded for further processing depending on the type or class of the waste objects O. For example, a particular waste object O may be forwarded for incineration in the incinerator 54, or may be separated and moved to a separate area due to it being unsuitable for incineration.

[0081] The waste processing facility 51 may comprise an incinerator 54 for incinerating waste objects O. The waste objects O may be fed or provided to the incinerator 54 by a feeder 53, which may be implemented as an inclined slope, a hopper, or other conveyance means, which convey waste objects O from the feeder 53 to the incinerator 54. The incinerator 54 further comprises a chimney 55 for removing exhaust gasses.

[0082] The waste processing facility 5 includes a crane with a crane bucket 56. The crane is designed to move the waste W, in particular the waste objects O, within the waste storage area 51 using the crane bucket 56. The crane is further designed to move the waste W from the waste storage area 51 to the feeder 53. The crane may be implemented as an overhead crane with a gantry system for moving the crane bucket 56 across the waste storage area 56.

[0083] The crane may be operated by a human operator. The crane may, additionally or alternatively, be operated automatically by a crane control system operatively connected to the crane, which can move the crane bucket, cause the crane bucket to open and close, or otherwise perform defined routines, methods or processes as described herein. The crane control system may be implemented by a computer 11. In particular, the crane bucket 56 may comprise one or more crane bucket actuators 56 for controlling the operation of the crane bucket 56, operatively connected to the computer 11. P29129PC00 12.12.2025

[0084] 15

[0085] The waste processing facility 5 further includes an exhaust gas analyzer 14 designed to analyze the exhaust gases from the incinerator. The exhaust gas analyzer 14 is designed to detect chemicals in the exhaust gas, for example CXHX, H2S, SO2, CO, CO2, NOX, NO, and NO2. The exhaust gas analyzer 14 is further configured to detect heavy metals, organic compounds, and particulates.

[0086] For example, the exhaust gas analyzer 14 comprises an infrared and / or ultraviolet sensor which may detect gases such as CO and SO2. The exhaust gas analyzer 14 may comprise an electrochemical sensor to detect NOX. The exhaust gas analyzer 14 may comprise a flame ionization detector to detect volatile organic compounds (VOCs).

[0087] The exhaust gas analyzer 14 is operatively connected to the computer 11 and is configured to transmit a gas analyzer signal to the computer 11 indicative of the analyzed exhaust gas, in particular indicating any gasses, heavy metals, organic compounds, and / or particulates present in the exhaust gas.

[0088] The waste processing facility 5 comprises a waste storage area sensor system 12 (also referred to herein as a waste sensor system (WSS) 12). The WSS 12 is operatively connected to the computer 11.

[0089] The WSS 12 is fixedly installed in the waste processing facility 5 at a defined physical location. The physical location may be stored during configuration or commissioning of the WSS 12. The physical location in particular defines the relative location of the WSS 12 with respect to the waste storage area 51.

[0090] The WSS 12 may comprise a number of sensors and is designed to monitor the waste storage area 51. The WSS 12 may, for example, be mounted or fixedly attached above or beside the waste storage area 12 and preferably can survey the entire waste storage area 51 at once. In an embodiment, the WSS 12 is implemented as a single integrated P29129PC00 12.12.2025

[0091] 16 device. In another embodiment, the WSS 12 comprises a plurality of distributed devices, wherein each device may be arranged to survey parts of the waste storage area 51 not visible to other devices.

[0092] The sensors include optical systems designed to image the waste storage area 51 in the visible, infra-red and / or ultraviolet optical spectra. The optical systems are designed to record and transmit a sequence of images. In an embodiment, the optical systems are designed for stereographic imaging, thereby allowing for more accurate determinations of distance to the waste objects O. The determination of the distance may be performed in the optical system itself, or in an external device using output from the optical system.

[0093] The WSS 12 may further comprise a ranging system such as radar and / or lidar for ranging measurements, in particular for determining the distance, position, size and / or shape of waste objects O. The WSS 12 may further be configured to determine the distance to the crane bucket 56 and / or the position of the crane bucket 56.

[0094] The waste processing facility 5 may further comprise one or more auxiliary sensor systems 13. The auxiliary sensor systems 13 may comprise similar sensors to the WSS 12. The auxiliary sensor systems 13 are operatively connected to the computer 11. The auxiliary sensor systems 13 are arranged in other areas of the waste processing facility 5. The auxiliary sensor systems 13 may transmit an auxiliary sensor system identifier. The auxiliary sensor system identifier may be associated or linked with a particular unloading bay 52 in which the auxiliary sensor systems 13 is arranged.

[0095] In an embodiment, auxiliary sensor systems 13 are arranged in the unloading bays 52. For example, a single auxiliary sensor system 13 is arranged in each unloading bay 52. The auxiliary sensor systems 13 are arranged to image incoming waste IW before it is deposited into the waste storage area 51. Thereby, individual waste objects O may be recognized more easily, as the waste objects O may be scattered, thinly layered or P29129PC00 12.12.2025

[0096] 17 individualized on the slope, chute, or other conveyance system which transports the incoming waste IW into the waste storage area W.

[0097] The auxiliary sensor systems 13 may further be configured to image a garbage truck 6 which provides the incoming waste IW. In particular the auxiliary sensor systems 13 may be configured to image an identifier of the garbage truck 6, for example a license plate, a logo or a name. Thereby, the original source of the incoming waste IW may be identified.

[0098] Figure 2 shows a block diagram of an electronic system 1 according to an embodiment. The electronic system 1 is preferably implemented entirely at the waste processing facility 5. In some embodiments, however, the electronic system 1 may be distributed, in particular in that one or more components are implemented remotely. For example, some processors may be implemented on a remote server, for example a cloud-based computing platform.

[0099] The electronic system 1 comprises a processing unit 11. The processing unit 11 includes a processor 111 , a memory 112, and a communication interface. The processing unit 11 is operatively connected to the other electronic components of the electronic system 1 using the communication interface.

[0100] Depending on the embodiment, the processing unit 11 , respectively, comprises a system on a chip (SoC), a central processing unit (CPU), and / or other more specific processing units such as a graphical processing unit (GPU), application specific integrated circuits (ASICs), reprogrammable processing units such as field programmable gate arrays (FPGAs), as well as processing units specifically configured to accelerate certain applications, such as artificial intelligence (Al) accelerators for accelerating neural network and / or machine learning processes. P29129PC00 12.12.2025

[0101] 18

[0102] The memory 112 comprises one or more volatile (transitory) and or non-volatile (non- transitory) storage components. The storage components may be removable and / or nonremovable, and can also be integrated, in whole or in part with the processor 111. Examples of storage components include RAM (Random Access Memory), flash memory, hard disks, data memory, and / or other data stores. The memory 112 has stored thereon computer program code configured to control the processing unit 11 of the electronic system 1 , such that the electronic system 1 performs one or more steps and / or functions as described herein. The memory 112 may additionally be configured to store data received from the sensor systems 12, 13. The memory 112 is further configured to store the waste object recognition module 2 and the crane tracking module 3.

[0103] Depending on the embodiment, the computer program code is compiled or non-compiled program logic and / or machine code. As such, the processing unit 11 is configured to perform one or more steps and / or functions. The computer program code defines and / or is part of a discrete software application. One skilled in the art will understand, that the computer program code can also be distributed across a plurality of software applications. The software application is installed in the processing unit 11. Alternatively, the computer program code can also be retrieved and executed by the processing unit 11 on demand. In an embodiment, the computer program code further provides interfaces, such as APIs (Application Programming Interfaces), such that functionality and / or data of the processing unit 11 can be accessed remotely, such as via a client application or via a web browser. In an embodiment, the computer program code is configured such that one or more steps and / or functions are not performed in processing unit 11 but in a remote server at a different location to the processing unit 11 , e.g. in a cloud-based computer system.

[0104] In particular, the processing unit 11 is configured to perform the steps and / or functions as defined by the waste object recognition module 2 and the crane tracking module 3 as described herein. P29129PC00 12.12.2025

[0105] 19

[0106] The operative connection between the electronic components as described herein comprises the use of a data connection mechanism, relates to a mechanism that facilitates data communication between two modules, devices, systems, or other entities. The data connection mechanism is a wired connection across a cable or system bus, or wireless connection using direct or indirect wireless transmissions. The data connection mechanism may be implemented, for example using a CAN bus or using Ethernet / IP, and may comprise the use of networks such as a local area network (LAN) or public networks such as the Internet.

[0107] In particular, the communication interface of the processing unit 11 is connected at least with the WSS 12 using the data connection mechanism.

[0108] In an embodiment, the processing unit 11 is connected directly to the WSS 12 using the data connection mechanism. In this embodiment, the processing unit 11 is located at or near the location of the WSS 12, such as in the waste processing facility 5. Alternatively, the processing unit 11 is located remotely from the waste processing facility 5 and is connected to electronic components of the waste processing facility 5, for example the WSS 12, over the Internet using a communication interface, such as a local gateway, connected to the WSS 12.

[0109] In an embodiment, in addition to being connected to the WSS 12, the processing unit 11 is also connected to a remote server via the Internet, using the communication interface. The connection to the remote server enables the processing unit 11 to exchange data with the remote server.

[0110] As described above with reference to Figure 1 , the WSS 12 is designed to image the waste W in the waste storage area 51. The electronic system 1 may further comprise one or more auxiliary sensor systems 12, an exhaust gas analyzer 14, or a crane control system 15. The exhaust gas analyzer 14 transmits a gas analyzer signal and the crane P29129PC00 12.12.2025

[0111] 20 control system 15 transmits one or more crane messages indicative of a three- dimensional position of the crane and / or a crane action state.

[0112] Figure 3 shows a block diagram illustrating a waste object recognition module 2, in particular the inputs 21 and outputs 22 of the waste object recognition module 2. The waste object recognition module 2 is, in an embodiment, implemented in the processing unit 11 of the electronic system in software. In other words, the waste object recognition module 2 is embodied as one or more software applications, libraries, functions and / or routines and may include application data or other data files in the processing unit 11. The waste object recognition module 2 may comprise compiled and / or non-compiled software code. The person skilled in the art will understand that the waste object recognition module 2 may, additionally or alternatively, also be implemented as a distributed software module in which at least some parts are remotely, for example in a remote server.

[0113] The input 21 to the waste object recognition module 2 includes one or more images of the waste W as recorded by the WSS 12 and / or the auxiliary sensor systems 13. In particular, the images 211 may comprise a sequence of colour images of the waste W. The inputs 21 may include, beyond the images 211 , metadata of the images, in particular a time-stamp of each image. The images 211 may be provided as two dimensional vectors, for example.

[0114] The waste object recognition module 2 is configured to process the input 21 .

[0115] The processing of the input 21 may comprise using a pre-processing sub-module of the waste object recognition module 2, configured to, for example, resize the images to a defined size, performing automated contrast and colour adjustment, or otherwise making the input images suitable for further analysis. P29129PC00 12.12.2025

[0116] 21

[0117] For substantial processing of the input 21 , the waste object recognition module 2 may comprise known image processing libraries, also referred to as machine vision libraries, such as OpenCV, or Scikit-lmage. These known general-purpose image processing libraries are pre-configured to perform basic image recognition tasks, such as detecting, recognizing, or tracking objects and are therefore also suitable for use in identifying waste objects in the images. In other words, the waste object recognition module 2 may include high level or generalist image processing libraries to perform the common and well understood task of identifying the presence of one or more objects in an image 211 of the input 21 and / or providing an indication of the locations of the objects in the image 211 , and / or defining a size or a bounding box of the object in the image 211.

[0118] The waste object recognition module 2, in particular the image processing libraries, may be augmented, configured, trained and / or fine-tuned to improve the detection of waste objects O in the images 211 of the waste storage area 51 and / or the unloading bay 52. For example, OpenCV supports model execution of modules trained using machine learning.

[0119] Additionally, or alternatively, further specialist sub-modules or models of the waste object recognition module 2 may be arranged down-stream of the one or more general-purpose image processing libraries described above. Such specialist sub-modules may also be configured, trained and / or fine-tuned to improve the detection of waste objects O in the images 211 and / or to classify waste objects or determine further properties of the waste objects, such as their calorific value, their size, or their suitability for incineration.

[0120] In a variant, the waste object recognition module 2 includes a neural network trained to detect and / or classify waste objects 2 in the images 211.

[0121] For example, the neural network may in particular comprise a convolutional neural network comprising a plurality of layers. For example, the neural network comprise one P29129PC00 12.12.2025

[0122] 22 or more of the following models or algorithms: the YOLO Algorithm for Object Detection; the Faster R-CNN object detection model which uses a region proposal network (RPN) with a convolutional neural network (CNN) model; a ll-Net; and / or a DETR (DEtection TRansformer), which is a deep learning model for object detection.

[0123] The configuration, training and / or fine-tuning of the image processing library(s), submodules and / or neural network may include the use of a training dataset. The training dataset includes a large number of classified (labelled) images of waste, in which individual waste objects are classified. In other words, each image in the training dataset includes identified waste objects, in particular their location, size, bounding box, and / or class. Preferably, the training dataset includes at least 1000 such classified images of a waste storage area and / or an unloading bay. In particular, the training dataset may include at least one thousand images of a waste storage area similar to the waste storage area of the waste processing facility.

[0124] The training dataset may further comprise a generic dataset of images and labels of objects in the images. Thereby, the training dataset may include a wide variety of objects which may not typically be present in waste, allowing the trained waste object recognition module 2 to detect a wider variety of objects. The generic dataset of images and labels may be generated by larger general model, for example a so-called zero shot model.

[0125] The complete training dataset is split into three sets, in particular for training, validation and testing. The training part of the training dataset is used to update tunable parameters of the waste object recognition module 2, in particular the neural network, during a training phase, while the validation and testing part of the training dataset is used to test the performance of the waste object recognition module 2 on previously unseen labelled images. This ensures that the waste object recognition module 2 does not become overfitted and may be applied or generalized to new, unseen images in operation. P29129PC00 12.12.2025

[0126] 23

[0127] Throughout the training of the waste object recognition module 2, in particular the neural network, extensive hyperparameter searches may be used, comprising adjusting learning rates, the number of network layers, and even selecting and testing different architectures or types of neural networks from amongst those mentioned herein. Thereby, robust and accurate predictions of the waste objects handled in the bunker may be generated.

[0128] The configuration, training and / or fine-tuning of the waste object recognition module 2 includes the use of machine-learning algorithms, in particular the use of supervised learning algorithms and the training dataset.

[0129] For example, neural network based deep learning models are trained for object detection and segmentation using the training dataset and classified (labelled) data.

[0130] The waste object recognition module 2 is configured to process the one or more images 211 in the input 21 and to generate an output 22. The output 22 may comprise one or more of the following: an identifier 221 of waste object(s) in the input 21 , a representation 222 of waste object(s) in the input 21 , a waste class 223 or classifier of waste object(s) in the input 21 , and location(s) 224 of the waste object(s) in the input 21.

[0131] The identifier 221 is designed to uniquely identify a particular waste object and may be an alpha-numeric string. The identifier 221 is linked or associated to the other properties or outputs of the waste object recognition module 2 related to the particular waste object.

[0132] The waste object recognition module 2 is designed to match waste objects across different images, giving each waste object a single unique identifier 221. Thereby, the tracking and tracing of an individual waste object 221 is simplified. The identifier 221 may further be associated with a time-point or time-stamp of each image 211 in which a particular waste object is identified. The identifier 221 may further be associated with a P29129PC00 12.12.2025

[0133] 24 waste source, i.e. , the originator of the waste object and / or the deliverer of the waste object to the waste processing facility.

[0134] The identifier 221 is therefore a unique and persistent identifier linked to a particular waste object over time.

[0135] The representation 222 of the waste object is a digital representation of the physical appearance of the waste object, allowing for the waste object to be matched and tracked across multiple images. The representation 222 may be implemented as an embedding vector, generated for example as part of the image processing. For example, the embedding vector may be generated as part of processing, of the image 211 , using a neural network. The embedding vector is in particular an n-dimensional vector in an n- dimensional feature space.

[0136] The embedding vector, may for example, be generated by a convolutional neural network (CNN), in particular by a penultimate layer of a trained CNN. The embedding vector may alternatively be generated by a transformer encoder.

[0137] The representation 222 is designed to be at least partially invariant to spatial transformations of the waste object and / or partial obscurations of the waste object, allowing for the waste object to be matched across images even if the waste object has turned, rotated, is lying in another position or is partially obscured by another object.

[0138] More generally, the representation 222 is or may be used as a unique identifier of a waste object or a unique digital fingerprint of the waste object. The representation 222 is designed to allow for re-identification and tracking of a particular waste object and is also unique for each waste object. P29129PC00 12.12.2025

[0139] 25

[0140] Thereby, the representation 222, in particular in combination with the identifier 221 , allows to distinguish between two objects with a substantially identical optical appearance at different locations.

[0141] In an embodiment, the representation 222 may be used to generate the identifier 221 , for example as a digest or hash of the representation 222.

[0142] In an embodiment, the representation 222 may function as the identifier 221. In other words, a separate identifier 221 may not be necessarily required, as the representation 222 is itself unique for every waste object.

[0143] The representations 222 may be matched using a similarity function. In particular, a similarity function may take as an input two representations 222 and compute a value indicative of the level of similarity between the two representations 222. If the level of similarity is above a defined threshold, then the two representations 222 may be considered to relate to the same waste object. For example, the similarity function may be implemented by computing an inner product (e.g., a cosine similarity) between the two representations 222 implemented as embedding vectors.

[0144] In another example, the similarity function may compute a distance or degree of spatial separation between the two representations 222 using a distance metric. In particular, a norm such as an L2 norm may be used to compute a distance between the two embedding vectors 222. If the computed distance is below a defined threshold, then the two representations 222 may be considered to relate to the same waste object.

[0145] In another example, a trained model may be used which takes as an input the two representations 222 and provides as an output a likelihood of whether the two representations 222 are related to the same waste object. The trained model in particular may perform a weighting of features or parts of the representations 222. The weighting P29129PC00 12.12.2025

[0146] 26 provides the trained model with the capability to match two representations 222 of the same waste object even when the orientation of the waste object has changed.

[0147] The waste class 223 of the waste object comprises at least one or more of the following classes: oversized, hazardous, or combustible. Oversized in particular relates to the size of the hopper or incinerator, indicating that the waste object is too large to fit and / or be incinerated. Further classes are described herein. The waste class 223, in addition to discrete classes or categories, may further include continuous outputs, for example an estimated calorific content of the waste object, a size of the waste object, a volume of the waste object, or a weight of the waste object.

[0148] The waste object recognition module 2 may further be configured to identify aggregate waste. In particular, the waste object recognition module 2 may be configured to identify aggregate waste and define, for particular aggregate waste, a singular waste object having a defined volume, shape, and / or location 224. Thereby, a singular aggregate waste object, may be defined.

[0149] The location 224 of the waste object describes the two-dimensional coordinates of the waste object, for example its center or the edges of its bounding box, in the input image 211. The location 224 is used as described herein to determine the three-dimensional position of the waste object in the waste storage area.

[0150] Figure 4 shows a block diagram illustrating a crane tracking module 3, in particular the inputs 31 and outputs 32. The crane tracking module 3 is, in an embodiment, implemented in the processing unit of the electronic system in software. In other words, the crane tracking module 3 is embodied as one or more software applications, libraries, functions and / or routines and may include application data or other data files in the processing unit 11. The crane tracking module 3 may comprise compiled and / or noncompiled software code. The person skilled in the art will understand that the crane P29129PC00 12.12.2025

[0151] 27 tracking module 3 may, additionally or alternatively, also be implemented as a distributed software module in which at least some parts are remotely, for example in a remote server.

[0152] The input 21 to the crane tracking module 3 comprises one or more images 31 of the waste W as recorded by the WSS 12 and / or the auxiliary sensor systems 13, as described above with reference to the waste object recognition module 2 of Figure 3.

[0153] The crane tracking module 3 is designed to detect the crane bucket in the images, and to provide as an output 3 the location of the crane bucket in the image and further to determine an action state of the crane bucket.

[0154] The crane tracking module 3 may be trained using machine learning, in particular using a training dataset of images including a crane bucket in a variety of action states, with the images including labels related to the location of the crane bucket in the image and its action state.

[0155] The action state of the crane bucket includes at least whether it is open or closed, and / or the degree of its opening, whether it is opening or closing, etc. The action state may further comprise whether the crane bucket is holding a waste object, whether the crane bucket is mixing the waste, or whether the crane bucket is transporting the waste. The action state may further define where the waste is being transported to, for example whether the waste is being transported for incineration, for example of depositing in the incinerator feed. The action state may further define whether waste was dropped. The action state may further be indicative of a waste sprinkling.

[0156] The location of the crane bucket in the image is defined as the two-dimensional location, in the image, of the crane bucket, in particular its center and / or a bounding box of the crane bucket. P29129PC00 12.12.2025

[0157] 28

[0158] The crane tracking module 3 may be implemented as part of the waste object recognition module 2, or share at least some processing steps with the waste object recognition module 2, for example the pre-processing steps and / or general-purpose object recognition steps.

[0159] The crane tracking module 3 includes one or more image processing libraries as described above, configured, trained and / or fine-tuned to detect the crane bucket and its action state.

[0160] Figure 5 shows a schematic illustration of a three-dimensional model M of a waste storage area, including waste objects O, a waste surface S, and an entry point E of an unloading bay.

[0161] The three-dimensional model M is a data structure, data file, database and / or data table stored in the processing unit, in particular in the memory. The three-dimensional model M corresponds to the waste storage area of the waste processing facility and may include one or more adjacent areas or spaces, for example a feeder (such as a hopper), an auxiliary or reject area which may store separated waste, an unloading bay, etc. The three-dimensional model M may be considered to form a digital twin to the waste storage area. In other words, the physical design and characteristics of the waste storage area are digitally reproduced in the model. The three-dimensional model M may be generated or configured based on pre-defined dimensions of the physical waste storage area.

[0162] In an embodiment, three-dimensional model M of the waste storage area is implemented as a list, set, table, database or collection of three-dimensional positions of waste objects. In particular, the list of waste objects identified to be in the waste storage area, along with their three-dimensional positions, defines the three-dimensional model of the waste storage area. P29129PC00 12.12.2025

[0163] 29

[0164] In other words, the three-dimensional model M may be expressed as a table of three- dimensional coordinates, each of which correspond to three-dimensional positions within the waste storage area and optionally adjacent areas. Table entries in the table of three- dimensional coordinates are linked to specific waste objects, for example by including a link, field, entry or other association between the three-dimensional coordinate and an identifier and / or a representation of the specific waste object.

[0165] The three-dimensional coordinates may have a defined spatial resolution.

[0166] The three-dimensional model M defines a three-dimensional volume corresponding to the waste storage area. The three-dimensional model M may be defined by a number of vertices, edges, polygons, voxels, etc.

[0167] Specifically, the three-dimensional model M may be a raster model in which the three- dimensional volume is divided or separated into a number of cells arranged in a grid structure. Was objects may occupy one or more cells in the model.

[0168] The three-dimensional model M may alternatively be a vector-based model in which the waste objects are assigned or associated with a particular three-dimensional coordinate within the model M.

[0169] The three-dimensional model M may further define one or more entry points E, in particular the location of the one or more entry points E.

[0170] Each waste object in the three-dimensional model has a defined position. The waste objects further include identifiers, in particular identifiers as provided or generated by the waste object recognition module.

[0171] The three-dimensional model M may be continually updated according to data provided by the waste object recognition module, in particular in that new waste objects may be P29129PC00 12.12.2025

[0172] 30 added to the model, incinerated or separated waste objects removed from the model, and the positions of waste objects which have moved or shifted being updated in the model. The updated model M may be stored in the memory.

[0173] Depending on the embodiment, the three-dimensional model may include physics models, in particular regarding collision and / or penetration detection between waste objects, and kinematic modelling of movement.

[0174] Thereby, the location of waste objects may additionally be predicted based on movements of other objects. In particular, the trajectory of waste objects falling into the waste storage area through the entry point E may be modelled such as to predict their resting place.

[0175] The three-dimensional model M may further separately define, or define depending on the present waste objects, the surface S of the waste.

[0176] The three-dimensional model M may further define or store the position of the WASS within the waste storage area.

[0177] Figure 6 shows a flow diagram of a method 100 according to an embodiment of the invention. The method 100 includes a number of steps S100 - S109, at least some of which are optional. In particular, steps S108 and S09 are optional. The method may be performed by the processing unit 11. The method may be performed, at least in part, by an external or remote computing device, such as a remote server, connected to the WASS.

[0178] The method, or particular steps, may be performed on demand, for example step S101 may be triggered by the reception of a new image. The method may alternatively be performed continually as indicated by the arrows in the Figure linking the individual steps. P29129PC00 12.12.2025

[0179] 31

[0180] In step S100, the three-dimensional model M of the waste storage area is initialized. Initializing the model M may comprise retrieving it from non-volatile memory and reading it into volatile memory. The initialization of the model M may comprise configuring the model M based on pre-defined parameters of the waste storage area. In an embodiment, this step is optional and may be omitted.

[0181] In step S101 , one or more images are received from the WASS.

[0182] In step S102, the images are processed by the waste object recognition module as described above. In particular the images are processed such that the waste objects in the images are identified in step S103, the representation of the waste objects is generated in step S104, and the location of the waste objects in the image is generated in step S105. The order or sequence of these steps may be interchanged.

[0183] While the order between steps S104 and S105 may be interchanged, the image processing of step S102 and the identification of the waste objects of step S103 take place prior to the representation being generated and the location determined.

[0184] In step S106, the three-dimensional position of the waste objects identified in step S104 is determined. The three-dimensional position is determined using the two-dimensional location of the waste objects in the image, along with a pre-defined three-dimensional position of the WASS.

[0185] The mentioned image processing libraries or computer vision libraries discussed herein, and / or photogrammetry, may be used to determine the three-dimensional position from the two-dimensional location.

[0186] Further, the three-dimensional orientation of the WASS, in other words, the direction in which the sensors are pointed, may be taken into account to determine the three- dimensional position of the waste objects. Further, the three-dimensional position of the P29129PC00 12.12.2025

[0187] 32 waste objects may be determined using a known field of view (FoV) of the WASS and monocular depth estimation techniques. Depending on the embodiment, further information provided by the WASS may be used to determine the three-dimensional position of waste objects, including using depth sensors of the WASS (such as radar or lidar), or using stereo vision.

[0188] In an embodiment, the waste storage area sensor system does not comprise LIDAR or another ranging sensor to directly measure spatial information. Rather, the waste storage area sensor system uses only one or more cameras to provide one or more images of the waste storage area.

[0189] The spatial information, in particular the three-dimensional location of the waste object, may be generated using a stereo vision model, in particular if two or more cameras are included, and / or a monocular vision model, in particular if only a single camera is included. The monocular vision model may be a trained model.

[0190] In particular, the three-dimensional position of the waste objects in the three-dimensional model M is determined.

[0191] In step S107, the three-dimensional position of the waste objects is stored. In particular, the three-dimensional position of the waste objects in the three-dimensional model M is stored in the memory and linked to the identifier of the waste object and / or the representation of the waste object.

[0192] Steps S101 to S107 may be performed repeatedly, in particular for each image or batch of images received from the WASS. Thereby, the three-dimensional position of waste objects may be kept current and up-to-date.

[0193] The three-dimensional model M or parts thereof, in particular including the position, classification and / or identification of waste objects may be transmitted in a message. For P29129PC00 12.12.2025

[0194] 33 example, the message may be transmitted and then displayed to an operator of the waste processing facility, providing the operator with a current map, model, view, rendering, list or other representation of the waste objects currently in the waste storage area.

[0195] The operator may filter, select, or otherwise manipulate the received three-dimensional model M or part thereof such as to identify particular waste objects, or filter waste objects according to their class, for example filtering out waste objects not suitable for incineration due to their class indicating hazardous, incombustible, oversize or otherwise unsuitability for incineration.

[0196] In step S108, the three-dimensional position of a particular waste object is tracked, using a plurality of instances of the three-dimensional model M over time, or using a plurality of three-dimensional positions of waste objects over time.

[0197] The plurality of three-dimensional positions of a particular waste object may be determined using the WSS 12 and / or the auxiliary sensor system 13. For example, a particular waste object in the incoming waste identified by the auxiliary sensor system 13 arranged in the unloading bay may be matched to a particular waste object in the waste in the waste storage area identified by the WSS 12. A physics based model may be used to calculate or predict a trajectory or landing position of the particular waste object entering the waste storage area from the unloading bay.

[0198] The waste object identified in a particular image is tracked by matching the representation of the waste object to a representation of a waste object in an earlier image. In particular, the representations are compared, and representations which match each other well, in other words, are close enough to within a defined threshold, are considered to relate to the same waste object. In particular, for representations implemented as embedding vectors, a distance (for example a Euclidean distance, P29129PC00 12.12.2025

[0199] 34 though other geometric metrics are also possible) may be calculated between the vectors, and a distance below a defined threshold may correspond to a positive match.

[0200] T racking the position over time of waste objects is useful for many reasons. For example, it allows for accurate statistics on the number of individual waste objects present in the waste storage area at any given time point. This is possible only with tracking as any current image will only reveal waste objects on the surface and not those which were once on the surface but are now covered. Beyond the number of individual waste objects, statistics on the class and properties of waste objects can be accurately determined. Further, the time that an individual waste object is present in the waste storage area prior to being incinerated may be determined, for example.

[0201] Further, tracking the position over time allows for more effective automated control of the waste storage area, as it can be checked and confirmed that tracked waste objects moved by the crane were moved correctly by comparing the position of the same waste object (identified by a unique identifier) before and after moving by the crane.

[0202] Additionally, a time-point of incineration may be determined or at least estimated, based on a time-point at which the waste objects are provided to the feeder or otherwise moved for incineration. By associating this time-point with a gas analyzer signal from the gas analyzer, the waste object may be associated with any particular pollution determined by the gas analyzer.

[0203] Further, by associating the waste object with a particular source of the waste object, for example by identifying the garbage truck 6, hazardous waste deposited may be traced back to its original source. Further, by associating the waste object with the gas analyzer signal from the gas analyzer, the original source of any pollutants emitted during incineration may be traced back to its original source. P29129PC00 12.12.2025

[0204] 35

[0205] Optionally, the gas analyzer signal may be used to identify a waste object which gave rise to the gas analyzer signal. For example, if the gas analyzer signal is indicative of mercury being present, this may be indicative of old paint containing mercury having been incinerated. The suspected waste object(s) considered to have likely been the cause of the gas analyzer signal may further be linked to a burn-time of the suspected waste object(s), thereby providing a more reliable prediction of which waste objects provided to the incinerator were related to the gas analyzer signal.

[0206] In an example, the three-dimensional model may be used to generate or determine “grab” or “no grab” areas of the waste storage area, to guide the crane operator or an automated crane control system.

[0207] The three-dimensional model may also be used to generate a list of oversized and / or hazardous waste objects present in the waste storage area.

[0208] In step S109, the message or a control signal may be generated based on, using or depending on the three-dimensional position of at least one of the waste objects.

[0209] The control signal is defined to control an actuator of the waste processing facility, such as the crane. For example, the control signal may be transmitted to a crane control system, such that the crane may be controlled depending on the three-dimensional position of one or more waste objects. For example, the crane may be controlled to selectively grab or avoid particular waste objects, in particular according to the waste class of the particular waste object. For example, the crane may be controlled to grab and remove from the waste storage area a waste object that is hazardous, oversized and / or otherwise unsuitable for incineration.

[0210] In a variant, the control signal is evaluated to determine whether the control signal would cause the crane or the crane bucket to grab waste in an area designated “no grab” or an P29129PC00 12.12.2025

[0211] 36 area which may for other reasons be designated out of bounds or to be avoided. In particular, one or more virtual fences may be defined, preferably in the three-dimensional model, and the control signal is evaluated to determine whether the crane bucket would enter a fenced off area. If the evaluation establishes that the control signal would lead the crane bucket to a fenced off or “no grab” area, the control signal is then either rejected or modified such that the crane bucket does not enter the fenced off or “no grab” area.

[0212] Figure 7 shows a flow diagram of a method 200 according to an embodiment of the invention. The method 200 includes a number of steps S200 - S209, at least some of which are optional. The method 200 may be performed by the processing unit 11. The method may be performed, at least in part, by an external or remote computing device, such as a remote server, connected to the WASS.

[0213] The steps do not necessarily need to be performed in the sequence described, and some sets or subsequences of steps may be performed continually or upon receipt of image(s), in particular steps S201 to S205, while subsequent steps may be performed only if particular conditions are met, in particular depending on the action state of the crane bucket.

[0214] In an embodiment, steps S201 to S205 may be omitted at least in part, and the position and / or action state of the crane bucket may be received from a crane control system.

[0215] In step S200, one or more images of the waste storage area are received, similar to step S109 described above with reference to Figure 6.

[0216] In step S201 , the one or more received images are processed using the crane tracking module as described herein. In particular, the one more images are processed to identify, in step S202, the crane bucket in the image, in particular to identify whether a crane bucket is visible in the one or more images. In step S203, the two-dimensional location P29129PC00 12.12.2025

[0217] 37 of the crane bucket in the image is determined and in step S204, the action state of the crane bucket is generated and provided, by the crane tracking module, as an output.

[0218] In step S205, the three-dimensional position of the crane bucket in the waste storage area is determined, for example as a three-dimensional position in the three-dimensional model M. The three-dimensional position of the crane bucket is determined in a similar manner to how the three-dimensional position of waste objects is determined in step S106 described above. In particular, the position, orientation and / or field of view of the WASS which imaged the crane bucket is used to determine the three-dimensional position of the crane bucket. Additionally or alternatively, ranging sensors in the WASS, such as lidar or radar, stereo vision, and / or machine learning algorithm based approaches may be used.

[0219] In step S206, a grabbing waste action state is identified. The grabbing waste action state is identified depending on the crane bucket grabbing one or more waste objects. The grabbing waste action state may be determined by the crane bucket closing at or near the surface of the waste or at a particular position coincident, corresponding, or otherwise being within range of one or more waste objects. The particular position may be determined based on the image and / or by using the position of the waste objects as included in the three-dimensional model.

[0220] In step S207, the grabbed waste objects are determined, for example by identifying waste objects in an image recorded of the crane bucket closing, the waste objects identified using the waste object recognition module. Additionally or alternatively, the grabbed waste objects may be determined by identifying, using the three-dimensional model, waste objects which have a three-dimensional position which have a position near (more specifically, are in a defined three-dimensional volume related to the volume of waste the crane bucket can grab) the crane bucket. P29129PC00 12.12.2025

[0221] 38

[0222] In step S208, a dropped waste action state is determined as a generated output of the crane tracking module. The three-dimensional position of the crane bucket is further determined.

[0223] In step S209, the three-dimensional position of grabbed waste objects is updated, using the three-dimensional position of the crane bucket at a time-point corresponding to the crane bucket opening and dropping the waste objects. A physics based model may be used to calculate the trajectories of or kinematics of the dropped waste objects.

[0224] In an embodiment, in which it is determined that the crane bucket slowly opens while moving, thereby spreading or sprinkling waste objects over a surface of the waste storage area, the updated three-dimensional position of the grabbed waste objects may be determined by assuming that the waste objects are spread evenly onto an area underneath the crane bucket as it slowly opens.

[0225] In particular, if the grabbed waste objects comprise an aggregate waste object relating to aggregate waste, the aggregate waste object may have, after being grabbed, dropped, spread and / or mixed an updated shape. Additionally or alternatively, the singular aggregate waste object may be split into two or more separate waste objects, for example if only part of the aggregate waste object is grabbed and moved, or if the aggregate waste object is grabbed and dropped partly at different positions. The two or more separate waste objects may be determined by considering conservation of volume and / or mass and by taking into account the positions at which the crane bucket opens and drops, or otherwise mixes, at least part of the aggregate waste object.

[0226] In particular, if the waste objects comprise aggregate waste, then conservation of mass or volume may be used to determine the depth or thickness of the waste spread onto the surface of the waste storage area. Specifically, the thickness or depth may be calculated P29129PC00 12.12.2025

[0227] 39 by dividing the known or assumed volume of the crane bucket by the total area swept or covered by the crane bucket during the spreading or sprinkling action state.

[0228] The updated three-dimensional position of the grabbed waste may be in the waste storage area or outside it, for example in a feeder, shredder, hopper, or incinerator.

[0229] In addition to the grabbing waste state and the dropping waste signal, the crane action state may be: a shredding waste state, a mixing waste state, a sprinkling waste state a dropping waste into an incinerator state, or a dropping waste into a reject area (for example, an auxiliary container) state.

[0230] The above-described embodiments of the disclosure are exemplary and the person skilled in the art knows that at least some of the components and / or steps described in the embodiments above may be rearranged, omitted, or introduced into other embodiments without deviating from the scope of the present disclosure.

Claims

P29129PC00 12.12.202540CLAIMS1. A computer-implemented method for locating waste objects (O) in a waste processing facility (5), the method comprising the steps of: initializing (S100) a three-dimensional model (M) of a waste storage area (51) in a waste processing facility (5), the three-dimensional model (M) defining physical boundaries of the waste storage area (51); receiving (S101), from a waste storage area sensor system (12) configured to monitor waste (W) in the waste storage area (51), an image (311) of the waste storage area (51); processing (S102), using a waste object recognition module (2), the received image (311) of the waste storage area (51), the processing comprising: identifying (S103) one or more waste objects (O) in the image (311), and generating (S104, S105), as an output, a representation (222) of each identified waste object (O) and a location (224), in the image (311), of each identified waste object (O); determining (S106), using the location of each identified waste object (O) in the image (311) and the three-dimensional model (M) of the waste storage area (51), a three-dimensional position of each identified waste object (O) in the waste storage area (51); and storing (S107) the three-dimensional position the identified waste objects (O), the three-dimensional position of a particular identified waste objectP29129PC00 12.12.202541(O) linked to the representation (222) of the particular identified waste object (O).

2. The method according to claim 1 , wherein the method further comprises: generating (S109) a control signal for controlling an actuator of the waste processing facility (5), using the stored three-dimensional position of the identified waste objects (O).

3. The method according to one of claims 1 or 2, wherein the method comprises: receiving (S101), from the waste storage area sensor system (12), a sequence of images (311); processing (S102), using the waste object recognition module (2), the sequence of images (311); and tracking (S108) the three-dimensional position of each identified waste object (O) overtime, by matching the representation (222) of the identified waste object (O) in a particular image (311) in the sequence of images (311) of the waste storage area (51) to a representation (222) of a previously identified waste object (O) in a previous image (311) in the sequence of images (311).

4. The method according to one of claims 1 to 3, wherein the method further comprises: processing (S102), using the waste object recognition module (2), the received image (311) of incoming waste (W), wherein the waste object recognition module (2) is further configured to classify each waste objectP29129PC00 12.12.202542(O) into one or more of the following waste classes (223): oversized, hazardous, or combustible; filtering the identified waste objects (O) by one or more of the waste classes (223); and generating a three-dimensional map of the filtered waste objects (O), using the three-dimensional position of the filtered waste objects (O).

5. The method according to one of claims 1 to 4, wherein the method further comprises: processing (S102), using the waste object recognition module (2), the received images (311) of the waste storage area (51), the processing comprising: identifying (S104) one or more waste objects (O) in the one or more images (311), wherein the waste objects (O) include one or more aggregate waste materials comprised of a plurality of particles of a similar material; and generating (S105), as an output, a representation (222) of each identified aggregate waste material and generating (S105) a location (224), in the one or more images (311), of each identified aggregate waste material; and determining (S106), using the location of each identified aggregate waste material in the image (311) and the three-dimensional model (M) of the waste storage area (51), a three-dimensional space occupied by each identified aggregate waste material in the waste storage area (51); andP29129PC00 12.12.202543 storing (S107) the three-dimensional space of the identified aggregate waste materials, the three-dimensional space of a particular identified aggregate waste material linked to the representation of the particular identified aggregate waste material.

6. The method according to one of claims 1 to 5, wherein the method further comprises: receiving a three-dimensional position of a crane bucket (56) of a crane; receiving a grabbing waste signal, the grabbing waste signal indicative of the crane bucket (56) grabbing waste; determining (S207) one or more grabbed waste objects (O) grabbed by the crane bucket (56), by matching the three-dimensional position of the crane bucket (56) with the three-dimensional position of waste objects (O); receiving a dropping waste signal, the dropping waste signal indicative of the crane bucket (56) dropping waste; and updating (S209) the three-dimensional position of the grabbed waste objects (O) using the three-dimensional position of the crane bucket (56) associated with the dropping waste signal.

7. The method according to claim 6, the method further comprising: receiving one or more of the following crane action states: a drop for incineration action state or a waste rejection action state, the drop for incineration action state indicative of the crane bucket (56) droppingP29129PC00 12.12.202544 waste for incineration and the waste rejection action state indicative of the crane bucket (56) dropping waste into a reject area; determining one or more dropped waste objects (O) dropped by the crane bucket (56) by matching the three-dimensional position of the crane bucket (56) with the three-dimensional position of waste objects (O) at a time-point indicated by the drop for incineration action state; and updating a status associated with dropped waste objects (O) according to the crane action state.

8. The method according to claim 7, wherein the method comprises: receiving, from a crane control system, the three-dimensional position of a crane bucket (56) of the crane; and receiving, from the crane control system, a crane message indicative of at least one of the following crane action states: the grabbing waste action state or the dropping waste action state.

9. The method according to one of claims 6 to 8, wherein the method further comprises: processing (S201), using a crane tracking module (3), the received image (311) of the waste storage area (51), the processing comprising: identifying (S202) the crane bucket (56) in the image (311), and generating (S203) as an output a location of the identified crane bucket (56) in the image (311) and generating (S204) at least oneP29129PC00 12.12.202545 of the following crane action states using the identified crane bucket (56): the grabbing waste action state or the dropping waste action state; and determining (S205), using the location of the crane bucket (56) in the image (311) and the three-dimensional model (M) of the waste storage area (51), a three-dimensional position of the crane bucket (56) in the waste storage area (51).

10. The method according to one of claims 6 to 9, wherein the method further comprises: receiving, from an exhaust gas analyzer (14) configured to analyze exhaust gas of the waste processing facility (5), a gas analyzer signal at a reading time-point; determining, using the gas analyzer signal, the presence of one or more pollutants in the exhaust gas at the reading time-point; determining incinerated waste objects (O) dropped for incineration within a defined time period prior to the reading time-point; and associating the one or more pollutants with one or more incinerated waste objects (O).

11. The method according to one of claims 1 to 10, wherein the method further comprises:P29129PC00 12.12.202546 receiving, from an auxiliary sensor system configured to monitor incoming waste (IW) in the waste processing facility (5), images of incoming waste (IW); processing, using the waste object recognition module (2), the received images of incoming waste to identify one or more waste objects (O) in the images of the incoming waste (IW); and assigning the one or more identified waste objects (O) to a three- dimensional position in the waste storage area (51).

12. The method according to claim 11 , wherein the method further comprises: receiving a waste provider identifier of a waste provider associated with a particular delivery time-point, the waste provider providing the waste objects (O) in the incoming waste stream to the waste processing facility (5); and associating the one or more identified waste objects (O) with the waste provider identifier, using a time-point associated with the images of incoming waste (IW) and the delivery time-point.

13. The method according to claim 12, wherein the method further comprises: associating one or more pollutants with the waste provider identifier, using the association between the one or more identified waste objects (O) and the waste provider identifier and, upon the identified waste objects (O) having been dropped for incineration, the association between the one or more pollutants and one or more incinerated waste objects (O).P29129PC00 12.12.20254714. The method according to one of claims 1 to 13, wherein the method further comprises: training the waste object recognition module (2) using a training dataset and a machine learning algorithm, wherein the training dataset comprises a plurality of images of waste storage areas with classified waste objects.

15. An electronic system (1) comprising a processor (111) configured to perform the method according to one of claims 1 to 14.

16. The electronic system (1) according to claim 15, further comprising a waste storage area sensor system (12) configured to monitor waste (W) in the waste storage area (51), the waste storage area sensor system (12) connected to the processor (111) and configured to transmit, to the processor (111), the sequence of images (311) of the waste storage area (51).

17. The electronic system (1) according to one of claims 15 or 16, further comprising an auxiliary sensor system (13) configured to monitor incoming waste (IW) in the waste processing facility (5), the auxiliary sensor system (13) connected to the processor (111) and configured to transmit, to the processor (111), the images (311).

18. A computer program product comprising computer-readable instructions configured to control a processor (111) of an electronic system (1) according to one of claims 9 to 11 such that the processor (111) performs the method according to any one of claims 1 to 14.

Images