Material sorting system and process

Abstract

A system and method allow for automated, and highly targeted detection of specific pieces of material such as woodchip within a mass flow. The system comprises a conveying means 101; a sensor arrangem

Description

FIELD This invention relates generally to the field of material sorting systems, and more specifically, devices and methods for detecting and removing chosen materials, particularly for waste wood. BACKGROUND In many industries, there exists a need to detect, and sort between materials of an incoming feed of materials. For example, in various recycling processes, there exists a need to sort incoming materials to ensure that only material that is suitable for a given recycling process is admitted to the process. If some unsuitable material is allowed to pass through the process, then it might cause damage to systems in the process, or simply result in a lower quality recycled material. Some example processes are wood recycling processes. In Europe alone, approximately 67-million-ton waste wood is generated every year, and 90% of it is incinerated, thereby releasing the carbon that is stored within. As opposed to incinerating such waste wood, recycling processes can transform such wood into new raw materials, enabling a sustainable use of wood products. The input mass flow into such recycling process can include waste wood that has been shredded to woodchips or wood pieces (including solid wood, MDF, chipboards, OBF, etc.), mixed with metal pieces, insulation materials, rocks, glass, etc. from the demolition process. For effective recycling, this flow should be separated into one or more streams of materials suitable for recycling, and other materials. There do exist devices and systems such as those described in EP3624958B1, which relates to the separation of wood-based materials, such as pieces of wood, shavings, or woodchips, from other non-wood materials, such as plastic materials, rubber, metal materials, or inert materials, such as glass, stones, rocks, or pieces of brick, as a preliminary operation prior to making wood-based panels. However, such systems rely at least partially on a difference between specific weights of materials to be sorted, and therefore are unsuitable for sorting materials which are very similar by weight. In addition, materials are detected on a conveyor belt. As such, it is not possible to detect any flaws or impurities that exist on the underside of the material, lying against the conveyor belt. It would therefore be beneficial to provide a flexible system for detecting and sorting between materials, including those which have very similar physical characteristics. SUMMARY According to a first aspect, there is provided method for detecting at least one characteristic of an object, such as a woodchip. The method comprises imparting a velocity on the object by a conveying means such that, when the object leaves the conveying means, it falls in a trajectory through a detection area to a first collection zone; and detecting at least one characteristic of the object as it falls through the detection area by a sensor arrangement located between the conveying means and the first collection zone. As would be appreciated, if the conveying means imparts a horizontal velocity to the object, the trajectory may be a ballistic trajectory due to gravity. In other examples, the trajectory maybe a vertical trajectory due to gravity. Advantageously, such a method allows for characteristics of the objects to be detected when the object is in free fall. As such, it is possible to detect all sides of the object. This allows for more robust detection of, for example, impurities in the input material, or constituent parts of the material. For example, waste woodchips might have been produced, in part, from a material that had been painted on one side. For a recycling process, it might be desirable to sort the chips that have paint on them from the clean woodchips. Typical systems would be unable to detect such impurities when the woodchips are lying face down on a surface. However, by detecting the woodchip in freefall, then all sides of the woodchip can be simultaneously detected. The sensor arrangement may comprise a plurality of sensors, for example optical sensors. Each optical sensor may use visible or short-wave infrared (SWIR) wavelengths or hyperspectral imaging. The sensors may be arranged such that, collectively, they have line of sight of all external surfaces of the object as it falls through the detection area. For example, one sensor may be arranged to have a different line of sight at a different angle or from a different direction to the line of sight of another sensor. The angle may be measured relative to a horizontal or vertical plane. The sensor arrangement may comprise at least one upper sensor and at least one lower sensor, which may be optical sensors. Such sensors may have line of sight of all external surfaces of the object, as it falls through the detection area. The at least one characteristic of the object may be a surface characteristic. At least one pair of sensors comprising an upper sensor and a lower sensor may be directed toward each other along the same line of sight. The line of sight may be angled relative to vertical and horizontal planes, which provides optimal sensing of all external surfaces of the object when it is falling in a ballistic trajectory. Accordingly the sensors may be able to detect any impurities that exist anywhere on the surface of the object. Each sensor may use visible or SWIR wavelengths or hyperspectral imaging. The sensor arrangement may comprise a combination of these different types of sensors, which may help to increase the precision of sorting models used to process the sensor data. The optical sensors may comprise line-scan cameras. The method may comprise using a further detection means to aid detection of the at least one characteristic of the object. The detection means may be arranged so as to detect at least one characteristic of an object as the object falls through the detection area, between the conveying means and the first collection zone. Like the sensor arrangement, the further detection means may comprise at least one upper sensor and at least one lower sensor to view all sides of the object. The further detection means may comprise different sensors to the sensor arrangement. For example, if the sensor arrangement comprises one or more visible light / line-scan cameras, the further detection means may comprise hyperspectral / SWIR cameras (or vice versa). Whilst, in visible wavelengths, such impurities may be very similar to the underlying material, in hyperspectral / SWIR cameras, the impurity may turn luminous compared to the background material, and thus be easier to detect, which increases the correct detection rate of the system. The method may comprise providing EM radiation to the detection area using an EM source, such as illuminating the detection area, or the object in the detection area, with visible light from one or more light sources. The EM source may be configured to provide EM radiation at wavelengths according to the spectral range(s) of the sensor arrangement. Illuminating the object with the appropriate wavelengths enables more accurate detection of surface characteristics of the object by the sensors. This may help to increase the precision of sorting models used to process the sensor data. The object may be part of an input mass flow of a plurality objects and the method may comprise spreading the input mass flow across the conveying means. The conveying means may comprise a conveyor belt. The method may comprise arranging a plurality of the objects on the conveyor belt in a single layer, for example using a vibration table to spread the objects out across the width of the conveyor belt. The single layer ensure that objects are not on top of each other, which could result in impurities being hidden from the sensor arrangement in the detection area if the sensor line of sight is blocked by another object. For example, a layer of paint on a woodchip may be hidden from the sensor arrangement by another woodchip that was covering the painted woodchip on the conveyor belt. The method may comprise pre-sorting the input mass flow so as to ensure relatively homogenous geometry of each of the plurality of objects. Such a homogenous flow may improve mass flow, detection, classification, and removal of objects through the separation process. The pre-sorting may comprise using magnets and / or a metal detector to identify metallic objects in the mass flow for removal. The method may further comprise using dust management means, such as an air suction means (e.g., a pump) to remove air and dust from the detection area and / or pressurised air nozzles to blow dust off surfaces of sensors and / or EM sources. This may be performed periodically at set intervals, or automatically in response to a detected parameter that indicates an accumulation of dust, such as a predetermined reduction in light received by the at least one sensor. The dust management means helps to maintain effective, stable detection of impurities by the sensor arrangement by reducing the impact of dust on the sensors and any lights. Dust management can be a particular challenge when handling materials such as wood, and particularly wood that has been shredded into woodchips. Dust can be stuck to the woodchips or originate from the woodchip material itself and loosen during transportation or other handling. Without good dust management it can be difficult to maintain a good sorting quality over time due to the impact of dust on the sensors and lights. Free-floating dust in the detection area can settle over the sensor apertures, blocking or blurring the sensors’ sight lines and reducing accurate and reliable detection. Dust can also settle over the light sources and reduce the illumination provided, which can also reduce the accuracy and reliability of the sensors. It is therefore important to reduce the presence of dust in the air in the detection area. In addition, the mass flow of objects by the conveying means can drag air into the detection area which may create an overpressure and produce turbulence that throws dust around. As well as removing dust, air suction can help to reduce air turbulence in the detection area and so further reduce dust swirling in the air. Furthermore, the fine dust can be volatile and present a fire hazard, particularly in the presence of heat from the light sources. Dust can also be detrimental to the health of personnel who could breathe the dust in if it is not removed and contained. When the method is used for waste wood, the dust may contain substances such as paint and plastic particles that could be particularly harmful when breathed in. It is therefore important to reduce the amount of dust in the system from a health and safety perspective. The method may comprise identifying the object for removal based upon a detection of at least one characteristic of the object; and selectively instructing an ejection means to impart an ejection force on the object for removal within a removal area, the removal area being between the detection area and the first collection zone, thereby deviating the object for removal from its (e.g., ballistic) trajectory to an altered trajectory responsive to the detection of at least one characteristic of the object, the altered trajectory ending at a second collection zone. As such, not only does the method allow for robust and effective detection of a characteristic (such as a surface characteristic) of an object, but it may also allow for automated, and highly targeted sorting of specific pieces of material based on the detected characteristics. The ejection means may comprise a plurality of ejection means, each of the plurality of ejection means being operable to selectively impart an ejection force on a respective ejection zone within the removal area. The step of selectively instructing the ejection means to impart an ejection force on the object may further comprise: determining, in use, the (e.g., ballistic) trajectory of the object, and which ejection zone the trajectory passes through, and responsive to a detection of the at least one characteristic of the object; and selectively instructing at least one of ejection means to operate to impart an ejection force on the object as the object falls through its respective ejection zone, thereby deviating the object from its trajectory to an altered trajectory, the altered trajectory ending at the second collection zone. The use of a plurality of ejection means allows for the method to be applied to a mass flow comprising many pieces of closely packed material. The method allows for a characteristic (such as a surface characteristic) of each piece of material of the mass flow to be identified. If it is identified that a given piece of material should be removed from the mass flow, then the path along which that piece of material will fall may be calculated, and a respective ejection means activated at the moment where the identified piece of material is passing through its ejection zone. This allows for targeted removal of only the identified pieces of material from the wider mass flow. When the objects / material are arranged in a single layer on a conveyor belt, this also prevents inadvertent ejection of more objects than just the object identified for removal. For example, if an object having impurities (e.g., a woodchip with a paint layer) is on top of a clean object (e.g., a clean woodchip), this may result in both objects being ejected, and so may undesirably increase the number of clean objects being discarded. In contrast, ensuring the objects are in a single layer on the conveyor belt helps to ensure that no objects are hidden from the sensor arrangement or inadvertently discarded when passing through the detection and ejection zones. The ejection means may comprise a plurality of pneumatic nozzles, and / or a plurality of flaps. Using air suction as part of the dust management means provides further advantages when the ejection means is air-based, as this may be an additional source of turbulence. According to another aspect, there is provided a method of sorting waste wood, comprising the above method, wherein the object is a woodchip. According to another aspect, there is provided a system for implementing the method described above. In this respect, there is provided a system for detecting at least one characteristic of an object, such as a woodchip, comprising, a conveying means; a sensor arrangement for detecting at least one characteristic of an object; and a first collection zone. The conveying means is, in use, configured to impart a velocity on an object such that, when the object leaves the conveying means, it falls due to gravity through a detection area to the first collection zone. The sensor arrangement is arranged so as to detect at least one characteristic of an object as the object falls through the detection area, between the conveying means and the first collection zone. Such a sensor arrangement allows for characteristics of the objects to be detected when the object is in free fall. As such, it is possible to detect all sides of the object. This allows for more robust detection of, for example, impurities in the input material, or constituent parts of the material. The sensor arrangement may comprise a plurality of sensors, for example optical sensors. Each optical sensor may use visible or short-wave infrared (SWIR) wavelengths or hyperspectral imaging. The sensors may be arranged such that, collectively, they have line of sight of all external surfaces of the object, as it falls through the detection area. For example, one sensor may be arranged to have a different line of sight at a different angle or from a different direction to the line of sight of another sensor. The angle may be measured relative to a horizontal or vertical plane. The sensor arrangement may comprise at least one upper sensor and at least one lower sensor, which may be optical sensors. Such sensors may have line of sight of all external surfaces of the object, as it falls through the detection area. The at least one characteristic of the object may be a surface characteristic. For example, at least one pair of sensors comprising an upper sensor and a lower sensor may be directed toward each other along the same line of sight. The line of sight may be angled relative to vertical and horizontal planes, which provides optimal sensing of all external surfaces of the object when it is falling in a ballistic trajectory. Accordingly the sensors may be able to detect any impurities that exist anywhere on the surface of the object. Each optical sensor may use visible or SWIR wavelengths or hyperspectral imaging. The sensor arrangement may comprise a combination of these different types of sensors, which may help to increase the precision of sorting models used to process the sensor data. The optical sensors may comprise line-scan cameras. The system may comprise a further detection means configured to aid detection of the at least one characteristic of the object. The detection means may be arranged so as to detect at least one characteristic of an object as the object falls through the detection area, between the conveying means and the first collection zone. Like the sensor arrangement, the further detection means may comprise at least one upper sensor and at least one lower sensor to view all sides of the object. The further detection means may comprise different sensors to the sensor arrangement. For example, if the sensor arrangement comprises one or more visible light / line-scan cameras, the further detection means may comprise hyperspectral / SWIR cameras (or vice versa). Whilst, in visible wavelengths, such impurities may be very similar to the underlying material, in hyperspectral / SWIR cameras, the impurity may turn luminous compared to the background material, and thus be easier to detect, which increases the correct detection rate of the system. The system may comprise an EM source configured to provide EM radiation to the object as it falls through detection area. The EM source may be configured to provide EM radiation at wavelengths according to the spectral range(s) of the sensor arrangement, for example a light source to provide visible light. Illuminating the object with the appropriate wavelengths enables more accurate detection of surface characteristics of the object by the sensors. This may help to increase the precision of sorting models used to process the sensor data. The object may be part of an input mass flow of a plurality objects and the system may comprise a spreading means configured to spread the input mass flow across the conveying means. The conveying means may comprise a conveyor belt. The method may comprise arranging a plurality of the objects on the conveyor belt in a single layer, for example using a vibration table to spread the objects out across the width of the conveyor belt. The single layer ensure that objects are not on top of each other, which could result in impurities being hidden from the sensor arrangement in the detection area if the sensor line of sight is blocked by another object. For example, a layer of paint on a woodchip may be hidden from the sensor arrangement by another woodchip that was covering the painted woodchip on the conveyor belt. The system may comprise magnets and / or a metal detector arranged to identify metallic objects in the mass flow for removal. The system may further comprise dust management means, such as air suction means (e.g., a pump), to remove air and dust from the detection area and / or pressurised air nozzles to blow dust off surfaces of sensors and / or EM sources. This may be performed periodically at set intervals, or automatically in response to a detected parameter that indicates an accumulation of dust, such as a predetermined reduction in light received by the at least one sensor. The dust management means helps to maintain effective, stable detection of impurities by the sensor arrangement by reducing the impact of dust on the sensors and any lights. The system may further comprise an ejection means being operable to impart an ejection force within a removal area, the removal area being between the detection area and the first collection zone; a second collection zone; and a processing means configured to, in use, identify an object for removal based upon the detection of at least one characteristic of the object, and selectively instruct the ejection means to operate to impart an ejection force on the object for removal as the object falls through the removal area, thereby forcing the object to the second collection zone. As such, not only does the system allow for robust and effective detection of a characteristic (such as a surface characteristic) of an object, but it may also allow for automated, and highly targeted sorting of specific pieces of material, based on the detected characteristics. The ejection means may comprise a plurality of ejection means, each of the plurality of ejection means being operable to selectively impart an ejection force on a respective ejection zone within the removal area, and the processing means may be further configured to determine, in use, the (e.g., ballistic) path of the object, and which ejection zone the path passes through, and responsive to a detection of the at least one characteristic of the object by the sensor arrangement, selectively instruct at least one of ejection means to operate to impart an ejection force on the object as the object falls through its respective ejection zone. The use of a plurality of ejection means allows for the system to be utilised with an incoming mass flow comprising many pieces of material. The system allows for a characteristic (such as a surface characteristic) of each piece of material of the mass flow to be identified. If it is identified that a given piece of material should be removed from the mass flow, then the path along which that piece of material will fall may be calculated, and a respective ejection means activated at the moment where the identified piece of material is passing through its respective ejection zone. This allows for targeted removal of only the identified pieces of material from the wider mass flow. The ejection means may comprise a plurality of pneumatic nozzles, and / or the ejection means may comprise a plurality of flaps. BRIEF DESCRIPTION OF THE DRAWINGS Certain examples of the disclosure will now be described, by way of example only, with reference to the accompanying drawings, in which: Figure 1A shows an example system for detecting objects within, and sorting an input mass flow; Figure 1B shows in more detail subsection A illustrated in Figure 1A; Figure 2 shows another example system for detecting objects within, and sorting an input mass flow; and Figure 3 shows a flow chart of an example sorting process. DETAILED DESCRIPTION Whilst the description herein mainly refers to the detection of and sorting of woodchips, it would be appreciated that the systems and methods discussed herein may be used in detecting and sorting other materials. As would be understood, the term “mass flow” refers to a “flow” of distinct pieces of material to be sorted. Taking the example of sorting woodchips, the flow may comprise a collection of individual woodchips, where each woodchip might comprise some impurities and it is desirable to detect the woodchips with impurities, and separate them from the flow. It is also foreseen that the flow could comprise, for example, distinct pieces of different materials, where it is desirable to sort the pieces by what material they comprise. It would be also appreciated herein that references to a ballistic or vertical trajectory, and free fall does not exclude presence of other forces, for example normal drag / air resistance. Such terms are instead used to indicate a stable, predictable path by which an object would fall given the conditions imparted on it by a system (such as the object’s initial velocity), and the further effects of gravity and / or other fields as it falls. Deviation from these trajectories indicates that an extra force is imparted on the object at a point on its fall path, thereby deviating it from that path. A system 100 for detecting objects within, and sorting an input mass flow 200 may be seen in Figures 1A and 1B, where Figure 1B shows in more detail the detection and removal process seen in detail A of Figure 1A. As can be seen, the system 100 comprises a conveying means 101, such as a conveyor belt, that is configured to, in use, receive an input mass flow 200 and move the input mass flow through the sorting system 100. The conveying means 101 is arranged such that, at an end 101a of the conveying means, the material thereon falls off under the effect of gravity. The velocity imparted on the material by the conveying means 101 results in the material falling off the end 101a following a ballistic trajectory, such as trajectory B or C. This trajectory may be changed by changing the speed of the conveying means. Unless acted upon by an external force during its fall, the system is configured such that, in normal operation, the ballistic trajectory leads material of the input mass flow to a first collection zone 112. In other examples, instead of a conveyor belt which may impart a horizontal velocity to the material and thus a ballistic trajectory under freefall, the conveying means may provide a vertical trajectory. For example, the conveying means may comprise a container with an opening in the base, and the material may be allowed to fall vertically through the opening. The system 100 is configured such that the ballistic trajectory B,C of material falling from the conveying means 101 passes through a detection area D and removal area E, before reaching a collection zone 110, 112. As the material is free falling (i.e. falling under the influence of gravity, with no physical contact to other driving means), all external surfaces of the material are visible, as it passes through the detection area. The system 100 comprises at least one sensor arrangement 103 configured to detect characteristics of objects falling though the detection area, for example, whether the objects comprise a coating, the material of the object, and / or whether there exist other impurities on the surface of the object. The sensor arrangement 103 may comprise a plurality of sensors 103a, 103b, such as at least one upper sensor 103a and at least one lower sensor 103b directed towards one another (e.g. along line of sight 103c) such that they can detect several sides of falling objects simultaneously. This allows the sensor arrangement 103 to have line of sight of all external surfaces of the object as it falls through the detection area D. Due to the ballistic trajectory B, C of the objects, it is advantageous to arrange the upper and lower sensors 103a, 103b to have a line of sight 103c that is angled relative to vertical and horizontal planes for optimal detection of all external surfaces of the objects. One or more of the sensors 103a and 103b may be optical sensors, such as linescan cameras, using visible or short-wave infrared (SWIR) wavelengths. One or more of the sensors 103a and 103b may use hyperspectral imaging. Having a combination of different types of sensors, which may help to increase the precision of sorting models used to process the sensor data. The signals captured by the sensors can be utilised by an algorithm to identify and characterise the material that passes through the detection area D, and also to remove material within the removal area E. The at least one sensor arrangement may further comprise one or more lights 103d configured to illuminate the mass flow in the line of sight of the at least one sensor 103a, 103b, in order to improve detection of the material. The lights 103d are configured to provide EM radiation at wavelengths according to the spectral range(s) of the sensor arrangement. When sorting materials such as wood, a lot of dust can be generated which can obscure the sensors and lights and thus reduce the sensors’ detecting ability. The dust can also be volatile and present a fire hazard, particularly in the presence of heat from the light sources. Therefore, the system 100 may comprise a dust management system configured to reduce the amount of dust in the detection area D. In order to prevent dust accumulation on surfaces of the at least one sensor 103a, 103b and / or the one or more lights 103d, the dust management system may comprise an air source, such as one or more high-pressure air nozzles arranged to blow pressurised air on the at least one sensor 103a, 103b and / or the one or more lights 103d to remove dust that may have accumulated on the surfaces. The blowing of air may be performed periodically at set intervals, or automatically in response to a detected parameter that indicates an accumulation of dust. For example, in response to a predetermined reduction in light received by the at least one sensor 103a, 103b. The dust management system may additionally or alternatively comprise an air suction system 116 for extracting air and dust from the detection area D. The air suction system 116 may comprise one or more air channels extending through a housing of the sensor arrangement 103 to remove dust and air from the detection area D at the end 101a of the conveying means 101 where the objects enter freefall. A relatively high air suction rate of more than 3000 m3 per hour may be applied to remove dust from the system 100 and collect or vent it elsewhere. This air suction also helps to reduce air turbulence and overpressure created by the mass flow of objects along the conveying means 101, which may be particularly pronounced in the detection area D. In other examples, the air suction system 116 may remove air and dust at other locations along the conveying means 101, for example at the other end of the conveying means 101 adjacent the input of the mass flow. The conveying means 101 is configured to ensure a stable speed of all objects leaving the conveyor belt and out into free fall, thereby ensuring a predictable, stable condition for subsequent sorting. The speed of the conveying means may be adjusted in order to optimize how long the objects are stable in the air, and thereby optimise the ballistic trajectory of the objects as they leave the conveying means. Higher speeds help to increase the rate at which the material is processed, but too high a speed may cause air resistance to affect the trajectory of the objects, and may reduce the accuracy of the sensor detection. The speed also needs to not be too fast for the identification / detection, decision and ejection processes (discussed further below) to be carried out. Speeds in the range of 1-2.5 m / s for the conveying means have been found to be suitable for a good balance between processing a lot of material quickly, and sorting the material with good accuracy. Owing to the detection during free fall, and contrary to typical systems which detect the input material solely whilst it lies on a conveyor belt and therefore are only able to detect a single side of the material, the present system 100 is capable of detecting all sides of an object. This allows for more robust detection of, for example, impurities in the input material, or constituent parts of the material. For example, waste woodchips might have been produced, in part, from a material that had been painted on one side. For a recycling process, it might be desirable to sort the chips that have paint on them from the clean woodchips. Typical systems would be unable to detect such impurities, when they are lying face down on a surface. However, by detecting the woodchip in freefall, then all sides of the woodchip can be simultaneously detected. The output from the sensor arrangement(s) 103a, 103b are analysed by a processing means 108 to recognize and categorize the material, and thereby identify material from the mass flow that should be removed. For example, images from the sensors (e.g. from both sides of the material) may be transferred to a processing means 108, and image detection may be performed to identify a piece of material, and also at least one characteristic of that piece of material (e.g. the presence of impurities on the surface of the identified piece of material), and therefore whether that the identified material piece should be removed from the mass flow. For example, the processing means 108 may identify that one piece of material has an impurity (e.g. a woodchip with paint on at least one of its surfaces). The identification of surface characteristics may be based on reflection characteristics of the object surface. For example, a painted surface of a woodchip will reflect EM radiation differently to an unpainted surface of a woodchip. Examples of impurities that may be identified on woodchips include paint, discolourations due to age / use, and embedded or adhered non-wood materials such as plastic and rubber. However, the identified surface characteristics of the objects are not limited to impurities. For example, the output from the sensor arrangement may be used to distinguish between different types or forms of wood, such as chipboard, OSB, veneer, plywood, MDF, HDF and wood fibre insulation. The identification / detection of a characteristic of a given object within a mass flow may be performed by an algorithm using Al based methods that are trained to recognize the materials contained in the input mass stream to be sorted. The detection, categorisation and decision processes are conducted at high speed while the material is in a ballistic trajectory. In order to separate, and therefore sort, objects within the mass flow following detection of the characteristics of each object, the system 100 may further comprise ejection means 105. The ejection means 105 is configured to remove one or multiple objects out of the mass flow by any suitable method, such as air pressure (e.g. where the ejection means 105 comprises ejection air nozzles), or by mechanical methods (e.g. where the ejection means 105 comprises mechanical flaps). The ejection means 105 operates within the removal area E, which is an area downstream of the detection area D, where objects may be removed from the mass flow based on the detection of their characteristics by the sensor arrangement 103. In other words, the ejection means 105 allows for removal of an object from its ballistic trajectory by the ejection means 105 selectively imparting an external force on a given piece of material (e.g. a piece of material that comprises an impurity). The external force changes the trajectory of a given piece of material from the free fall trajectory such as B or C, to an altered trajectory such as F (as seen in Figure 1B). The altered trajectory leads the separated material to fall into a second collection zone 110, separated from the first collection zone 112. In this way, the present invention may separate desired materials from the mass flow, and therefore result in the collection of only desired material. In the example of sorting woodchips, then it might be that only “clean” woodchips are allowed to pass to one collection zone, and that woodchips with impurities are collected in another collection zone. As would be appreciated, the ejection means 105 may cover a number of distinct ejection zones, which together might make up a total removal area E, and therefore be configured to remove only selected objects from a given part of the mass flow. For example, the ejection means 105 may comprise a plurality of air nozzles covering a plurality of distinct ejection zones. The ejection zone(s) together may span across the same width as the conveying means (or more), and therefore forming a removal area covering the entire width of the mass flow, such that any pieces of the material of the mass flow can be selectively removed by the ejection means 105, by activating an ejection means 105 at the moment that the detected object is passing through the ejection zone associated with the ejection means. Based on the output of the sensor(s), the processing means 108 identifies pieces of material that should be removed from the mass flow. For example, the processing means may recognise a number of pieces of material within the mass flow, calculate material quality for each piece of material, and if the material quality is below a threshold (e.g. if paint is identified on one side of the material), then the processing means may identify that that piece of material should be removed from the flow. Alternatively, it might be determined that it is more efficient to “remove” only the desired material, for example, if the desired material makes up only a minority of the incoming mass flow. In addition, various thresholds may be set within the processing means 108, according to the desired sorting results. As would be appreciated, an incoming mass flow might contain many different materials and / or types of impurities, and the processing means may be configured to only remove certain materials from the mass flow. In addition, tolerance levels may be set for each material / type of impurity - i.e. the most amount of that material / impurity would be acceptable, and sort the mass flow accordingly. The decision to eject an object from the mass flow may be performed by an algorithm using Al-based methods that are trained to recognize acceptable and unacceptable materials, i.e., whether or not materials have impurities and need to be removed. The detection, categorisation and decision processes are conducted at high speed while the material is in a ballistic trajectory. The processing means 108 may use computer vision and Al to detect and eject unwanted materials. An Al model can be trained by capturing and identifying images of various object types and surfaces that are expected. Based on a large number of images, a decision model can be created that enables classification of imaged objects into acceptable and unacceptable categories, which enables a decision to accept or reject the object to be made. The more images of different objects and impurities used to train the Al model, the higher the decision-making accuracy. As the material follows a generally ballistic trajectory B, C (which may be calculated based upon the speed of the conveyor belt, and therefore the speed of the individual pieces of the mass flow, and gravity), it is possible to calculate where the material will fall, and therefore, which ejection zone it will pass through and when. Therefore, based on the time at which the piece material passed through the sensor line of sight 103c, the location of the piece material in the mass flow, the horizontal speed with which it is travelling (e.g., as defined by the conveyor belt), and gravity, the processing means 108 calculates the ballistic trajectory of the given piece of material and instructs the ejection means 105 to activate a given ejection zone, when the identified object is passing that ejection zone. This allows for automated, and highly targeted removal of specific pieces of material from the mass flow. As would be appreciated, whilst only two collection zones are illustrated, there could be many separate collection zones, and the ejection means configured to direct different material to different collection zones, depending on the material detected. For example, the ejection means may comprise several ejection means pointing in different directions in order to impart different forces on selected falling material, thereby pushing them towards other collection areas. Additionally or alternatively, the ejection means may be configured such that it can impart different amounts of force on falling material, thereby affecting its trajectory in different amounts, urging the material towards different collection zones. Again taking the example of woodchips, then the input mass flow might contain “clean” woodchips, woodchips with paint on them, as well as other materials. These could be sorted into three or more separate collection zones. The first and second collection zones 112, 110 may be separated by an adjustable separator 106. The separator may be moved (for example up and down, and / or left and right) so as to adjust the collection zones. For example, if the separator is moved up (and therefore closer to the free fall ballistic path), the amount of deviation from the free fall ballistic path that the materials must experience so as to fall to the second collection zone 110 is less. This may be adjusted, for example, based on the material of the mass flow. The system 100 may comprise a spreading means, configured to spread the input mass flow 200 across the conveying means 101, such as a vibrator table. As such, the input mass flow 200 may be spread out from a relatively narrow flow to more even distribution across the conveying means 101. The means for spreading the materials may ensure that none, or only a few constituent parts of the mass flow material are on top of each other, thereby aiding later detection, and sorting of the material. The system 100 may comprise a further detection means 102 configured to aid detection of the materials and / or impurities on the material in the mass flow. As seen in Figure 1, the detection means 102 may be directed at the conveying means 101, having a line of sight 102c such that the detection means 102 can detect materials that lie on the conveying means 101. However, the further detection means 102 may be used within the at least one sensor arrangement 103 as a supplement to the other sensors 103a, 103b, such as line-scan cameras. The detection means 102 may comprise an optical sensor configured to observe a spectrum of wavelengths over one or more channels for the purpose of object detection and material identification, for example a hyperspectral camera and / or a short-wave infrared (SWIR) camera. This information may be used to supplement the information gathered by the sensors 103a, 103b, thereby aiding in the detection of desired materials by the processing means 108. In some cases an impurity may be subtler than others. For example, only a small part of woodchips may have paint on them, and / or the colour of the paint may be very similar to that of the wood (e.g. worn yellow paint similar to the colour of the underlying wood, worn grey paint on worn wood that has turned grey from water and sun (wear), and thin layers of white where the colour of the underlying wood shines through). Whilst, in visible wavelengths, such impurities may be very similar to the underlying material, in hyperspectral / SWIR cameras, the impurity may turn luminous compared to the background material, and thus be easier to detect, which increases the correct detection rate of the system 100. The detection means 102 may further comprise one or more lights 102d, configured to illuminate the line of sight of the detection means, thereby allowing for more accurate detection. The lights 102d are configured to provide EM radiation at wavelengths according to the spectral range(s) of the sensor arrangement. As for the sensor arrangement 103, the dust management system 114 may include additional high-pressure air nozzles arranged to blow pressurised air on the sensors of the detection means 102 to reduce the accumulation of dust. An alternative system 300 for sorting an input mass flow 200 may be seen in Figure 2. The system 300 is similar to the system 100 of Figures 1A and 1B, and like reference numerals indicate like features. However, and as may be seen, rather than the detection means 102 being directed at the conveying means 101, in system 300, the detection means 302 is arranged in a similar general location to the sensor arrangement 303 (which may correspond to sensor arrangement 103 in Fig. 1A), and are similarly configured to detect the mass flow when it is in freefall. In this way, the detection means may comprise an upper detection means 302a, and a lower detection means 302b opposing one another along a line of sight 302c. This enables the detection means 302 to detect multiple sides of the material in the mass flow simultaneously, similarly to sensors 303a and 303b. Again, the detection means 302 may comprise an optical sensor configured to observe a spectrum of wavelengths over one or more channels for the purpose of object detection and material identification, for example a hyperspectral camera and / or a short-wave infrared (SWIR) camera, and the processing means 108 may identify material based on the output of both the sensors 303a, 303b and detection means 302a and 302b. It has been found that, a combination of a sensor arrangement 303 comprising a camera configured to generally detect wavelengths in the visual spectrum, and a further detection means 302 configured to detect hyperspectral wavelengths advantageously allows for rapid and highly accurate detection of desired objects in freefall. Having a combination of different types of sensors has been found to help to increase the precision and accuracy of sorting models used to process the sensor data. As some types of impurities may be more easily detected using certain wavelengths, using different types of sensors for imaging all surfaces of an object increases the detection accuracy. As shown in Figure 2, the lines of sight 302c, 303c may be configured to cross the ballistic path at the same point, allowing for complete (and early) detection of pieces of material. The lines of sight 302c, 303c are thus at different angles, both angled with respect to horizontal and vertical planes. The sensors 103a, 103b, 303a, 303b, and / or detection means 102, 302a, 302b may comprise a variable set of sensors, which may be selected based upon the materials that are to be sorted. In addition, further sensors may be added. The sensors and / or detection means may comprise a plurality of sensors, each having a group of materials that it is configured to detect and validate the content of. As would be appreciated, the sensors may be turned on and off based on which combination of sensors create the right output quality for a given mass flow, and materials therein. For example, certain paints / coatings may be best recognised by optical sensors configured to sense only certain wavelengths. When it is known that a given mass flow contains material having such coatings, then it may be desirable to use such sensors. If the mass flow contains a lot of metal, then it might be desirable to add / utilise metal detectors for removing metal, and / or documenting the presence of metal in the mass flow for control purposes. Such detectors may be added anywhere on the system - for example, configured to detect materials when the material is on the conveying means, and / or configured to detect materials in free fall. As described above, the systems discussed herein to effectively detect and separate materials in a mass flow. In one example, the system and process has been proven effective when sorting waste wood that has been shredded to woodchips or wood pieces, where the chips / pieces contain different types of wood (solid wood, MDF, chipboards, OBF, etc.) mixed with metal pieces, insulation materials, rocks, glass, etc. from the demolition process. However, as would be appreciated, the logic applies to any sorting process. In order to ensure optimal operation of such a sorting system, it may be desirable to ensure that the geometry of the pieces of material of the incoming mass flow are relatively consistent. For example, this may result from specific considerations in preparation for shredding, and / or in the shredding process itself. For example, in the example for sorting woodchips, a process may be applied so as to produce a relatively homogenous geometry of woodchips, such as woodchips of <100mm in length, such as 20-80mm or 20-60mm in length, 10-40mm in width, and / or 3-20mm in thickness. Keeping the woodchip length below 100mm helps to avoid blockages in the system. Such a homogenous flow may improve mass flow, detection, classification, and removal of objects through the separation process. One such way to achieve such a homogenous flow is with a combination of shredding and sifting techniques. Advantageously, these shredding and sifting processes may also ensure removal of at least some unwanted materials such as metal, plastic, rocks, insulation, and unwanted sizes of wood (over and under) before the sorting process. As such, effective shredding and sifting may reduce volume of unwanted objects and decreases the number of steps necessary later in the process - or simply decrease the amount of material that needs to be removed by the system. The geometry of the objects may impact the precision of the sorting model. If the objects are too small, this may increase the amount of acceptable material ejected along with impure material. The geometry of the objects also impacts the sorting volume. For a given system setup, the size of detection area is fixed - for example by the width of the conveyor belt, and the field of view of the sensors. Therefore, to vary the volume of material that can be sorted in a given period of time, the speed of the conveying means and / or the thickness of the objects needs to be varied. For example, increasing the average thickness of woodchips from 3mm to 9mm would give a 3x increase the average volume of material sorted per unit time. The preprocessing stage can be adapted to give the required thickness of the objects, for example shredding the woodchips to give a certain thickness. Optimising the geometry of the objects also helps to improve sensor visibility of the objects and the ejection precision by considering the amount of space between the objects provided by the conveying means. The smaller the objects, the more empty space needed to separate the objects, which lowers the average volume of material sorted per unit time. For example, increasing the average dimensions of the woodchips from 10mm to 50mm by length and from 3mm to 10mm by width through the pre-processing will increase the average volume of material sorted per unit time. An example preparation process is illustrated in Figure 3. In order to even further enhance the effectiveness of the sorting system, such as the system 100 or system 300, there may be performed a pre-sorting process 400 of an incoming mass flow 200. Such a process may be automated, or performed manually before shredding. The process may remove any unwanted objects from the mass flow as soon as possible. For example, the process might be performed at a demolition / building site, or other collection point, and material can be sorted at that point based on its level of contaminations. Solid wood materials without any treatment (for example, without paint or pressure treatments) can be separated out and later shredded separately. Major pieces of loose waste materials like; plastic, metal, insulation, rocks that can be taken out manually with little effort may be removed. Smaller pieces stuck to the materials is left on. It is advantageous to remove heavier objects, for example over 75g, such as large metal objects and rocks, from the mass flow because these objects may be to large / heavy for the ejection means to reliably handle them. Similarly, solid wood materials with external treatment (paint) may be separated out and later shredded separately. Again, major pieces of loose waste materials like; plastic, metal, insulation, rocks that can be taken out manually with little effort may be removed. Additionally or alternatively, wood panel boards can be separated out and later shredded separately. Again, all major pieces of waste materials like; plastic, metal, insulation, rocks that can be taken out manually with little effort may be removed. Pressure treated wood materials may also be separated out. Again, all major pieces of waste materials like; plastic, metal, insulation, rocks that can be taken out manually with little effort may be removed. The pre-sorted mass flow may then be shredded, and sifted as described above, and then, the pre-sorted and shredded mass flow may proceed to a feeding step 500, for feeding the material to a sorting system (such as system 100 or 300) for sorting 600. The feeding system may be configured to remove loose pieces of metal and / or non-metal objects having metal embedded therein from the pre-sorted mass flow. One or more magnets may be used to remove ferrous metal objects or non-ferrous metal objects having ferrous metal embedded within them. For example, the feeding system may comprise a hopper that feeds a conveying means (such as a conveyor belt) with an over band magnet, such that the magnet may remove loose metal pieces lying on top of the mass flow. One or more metal detectors may be used to identify metallic objects in the mass flow. The metal detectors use the objects’ influence on a high frequency EM field as the objects pass the metal detector to identify the presence of metal and can be very sensitive to register even very small variations caused by small amounts of metal. The conveying means may feed into a system of at least one metal detectors / removers, configured to detect and / or remove remaining metal pieces, ferrous and non-ferrous. Additionally / alternatively a metal detector / remover (such as a tunnel detector) may be added downstream of the sorter, thereby removing metal after sorting. The signals from the one or more metal detectors may be input to a system trained to differentiate between signals caused by metal and background “noise” in the system to ensure accurate and precise detection of metallic objects. A mechanical valve may be used to divert metallic objects out of the mass flow. The feeding system may be configured to measure the amount of mass in the mass flow. The conveying means may feed over to another conveying means (such as a conveyor belt) having a built in scale to measure weight and mass of the incoming mass flow, and a drum magnet at the end to remove metal that are at the bottom of the mass flow. The resulting material may then be fed to a sorting system as an input mass flow, such as mass flow 200 into system 100, 300. It would also be appreciated that the system 100, 300 may operate with an unsorted input mass flow. In short, the systems discussed herein provide can provide characterisation and / or sorting of an input mass flow based on: • a set of mechanical devices (e.g. conveyors, spreading means) that ensure a stable and predictable condition of the mass flow for characterisation / sorting, • a set of sensors that captures signals that can be used by an algorithm (e.g. implemented in a processing means). • recognizing and categorizing the material based on the signals captured by the sensors, • and an algorithm (e.g. in the processing means) that identifies individual pieces of the mass flow, characterises them and can initiate removal of certain categories of material using a set of ejector mechanisms, if sorting is desired. The algorithms are software based and can utilize artificial intelligence (Al) methods that are trained to recognize the materials contained in the product stream to be sorted. The detection, categorisation and decision processes are conducted at high speed while the material is in a ballistic trajectory, thereby allowing for complete detection of all surfaces of the material. The optical sensors can use a plurality of light wavelengths to generate images to create a set of renderings representing the surface of a falling object. The individual images can be combined to create a model of the object’s surface, which may represent all sides of the object across all detected wavelengths combined. An Al model may be used to analyse the images to determine properties of the object, and can calculate the best decision to take regarding activating the ejection means based on the image analysis. The combinations of described processes and technology discussed above have been discussed in the specific example when sorting woodchips created of waste wood, where there are remnants of unwanted substances from the demolition process, it would be appreciated that the systems and methods discussed herein equally may be applied to any sorting process. It will be appreciated by those skilled in the art that the disclosure has been illustrated by describing one or more specific examples, but is not limited to these examples; many variations and modifications are possible within the scope of the accompanying claims.

Claims

1. A method for detecting at least one characteristic of an object, comprising: imparting a velocity on the object by a conveying means (101) such that, when the object leaves the conveying means (101), it falls in a trajectory (B, C) through a detection area (D) to a first collection zone (112); and detecting at least one characteristic of the object as it falls through the detection area (D) by a sensor arrangement (103; 303) located between the conveying means (101) and the first collection zone (112).

2. The method of claim 1, wherein the sensor arrangement (103; 303) comprises at least one upper sensor (103a; 303a) and at least one lower sensor (103b; 303b) and, optionally, the at least one upper sensor (103a; 303a) and the at least one lower sensor (103b; 303b) are optical sensors.

3. The method of claim 1 or 2, further comprising using a detection means (102; 302) to aid detection of the at least one characteristic of the object.

4. The method of claim 3, wherein the detection means (302) is arranged so as to detect at least one characteristic of an object as the object falls through the detection area (D), between the conveying means (101) and the first collection zone (112).

5. The method of claim 4, wherein the detection means (302) comprises at least one upper sensor (302a) and at least one lower sensor (302b).

6. The method of claim 4, wherein the sensor arrangement (103; 303) comprises at least one upper sensor (103a; 303a) and at least one lower sensor (103b; 303a) that are one of visible light sensors, SWIR sensors, and hyperspectral sensors, and the detection means (302) comprises at least one upper sensor (302a) and at least one lower sensor (302b) that are another of visible light sensors, SWIR sensors, and hyperspectral sensors.

7. The method of any preceding claim, further comprising providing EM radiation to the object as it falls through the detection area (D) using an EM source (102d, 103d).

8. The method of any preceding claim, further comprising blowing pressurised air onto a sensing surface of the sensor arrangement (103; 303) to remove dust.

9. The method of any preceding claim, further comprising removing dust from the detection area (D) by air suction means (116).

10. The method of any preceding claim, wherein the object is part of an input mass flow (200) of a plurality of objects, and optionally the method further comprises spreading the input mass flow (200) across the conveying means (101).

11. The method of claim 10, further comprising pre-sorting the input mass flow (200) so as to ensure relatively homogenous geometry of each of the plurality of objects.

12. The method of any preceding claim, the method further comprisingidentifying the object for removal based upon a detection of at least one characteristic of the object; andselectively instructing an ejection means (105) to impart an ejection force on the object for removal within a removal area (E), the removal area (E) being between the detection area (D) and the first collection zone (112), thereby deviating the object for removal from its trajectory (B, C) to an altered trajectory (F) responsive to the detection of at least one characteristic of the object, the altered trajectory (F) ending at a second collection zone (110).

13. The method of claim 12, wherein the ejection means (105) comprises a plurality of ejection means (105), each of the plurality of ejection means (105) being operable to selectively impart an ejection force on a respective ejection zone within the removal area (E), and wherein the step of selectivelyinstructing the ejection means (105) to impart an ejection force on the object further comprises:determining the trajectory (B, C) of the object and which ejection zone the trajectory (B, C) passes through, and responsive to a detection of the at least one characteristic of the object; andselectively instructing at least one of the plurality of ejection means (105) to operate to impart an ejection force on the object as the object falls through its respective ejection zone, thereby deviating the object from its trajectory (B, C) to an altered trajectory (F), the altered trajectory (F) ending at the second collection zone (110).

14. A method of sorting waste wood, comprising the method of claim 12 or 13, wherein the object is a woodchip.

15. A system (100) for detecting at least one characteristic of an object, comprising:a conveying means (101);a sensor arrangement (103) for detecting at least one characteristic of an object; anda first collection zone (112);wherein the conveying means (10) is, in use, configured to impart a velocity on an object such that, when the object leaves the conveying means (101), it falls due to gravity through a detection area (D) to the first collection zone (112);wherein the sensor arrangement (103) is arranged so as to, in use, detect at least one characteristic of an object as the object falls through the detection area (D), between the conveying means (10) and the first collection zone (112).

16. The system of claim 15, wherein the sensor arrangement (103) comprises at least one upper sensor (103a) and at least one lower sensor (103b) and, optionally, the at least one upper sensor (103a) and the at least one lower sensor (103b) are optical sensors.

17. The system of claim 15 or 16, further comprising a detection means (102; 302) configured to aid detection of the at least one characteristic of the object.

18. The system of claim 17, wherein the detection means (102) is arranged so as to detect at least one characteristic of an object as the object falls through the detection area (D), between the conveying means (101) and the first collection zone (112).

19. The system of claim 18, wherein the detection means (302) comprises at least one upper sensor (303a) and at least one lower sensor (303b).

20. The system of claim 18, wherein the sensor arrangement (103) comprises at least one upper sensor (103a) and at least one lower sensor (103b) that are one of visible light sensors, SWIR sensors, and hyperspectral sensors, and the detection means (302) comprises at least one upper sensor (303a) and at least one lower sensor (303b) that are another of visible light sensors, SWIR sensors, and hyperspectral sensors.

21. The system of any of claims 15 to 20, further comprising an EM source (102d, 103d) configured to provide EM radiation to the object as it falls through the detection area (D).

22. The system of any of claims 15 to 21, further comprising at least one pressurized air nozzle configured to blow pressurised air onto a sensing surface of the sensor arrangement (103) to remove dust, and / or further comprising air suction means configured to remove dust from the detection area.

23. The system of any of claims 15 to 22, further comprising:an ejection means (105) being operable to impart an ejection force within a removal area (E), the removal area (E) being between the detection area (D) and the first collection zone (112);a second collection zone (110); anda processing means (108) configured to, in use, identify an object for removal based upon the detection of at least one characteristic of the object,and selectively instruct the ejection means (105) to operate to impart an ejection force on the object for removal as the object falls through the removal area (E), thereby forcing the object to the second collection zone (110).

24. The system of claim 23, wherein the ejection means (105) comprises a plurality of ejection means (105), each of the plurality of ejection means (105) being operable to selectively impart an ejection force on a respective ejection zone within the removal area (E), andwherein the processing means (108) is further configured to determine, in use, a trajectory (B, C) of a detected object, and which ejection zone the trajectory (B, C) passes through, and responsive to a detection of the at least one characteristic of the object by the sensor arrangement (103), selectively instruct at least one of the ejection means (105) to operate to impart an ejection force on the object as the object falls through its respective ejection zone.

25. The method or system of any preceding claim, wherein the ejection means (105) comprises a plurality of pneumatic nozzles, and / or wherein the ejection means (105) comprises a plurality of flaps.

Images