阅读说明：本技术 具有胶囊识别的饮料制备机器 (Beverage preparation machine with capsule recognition ) 是由 M·鲁杰罗 S·巴拉西 于 2018-06-15 设计创作，主要内容包括：本发明涉及用于制备和分配饮料的机器,该饮料是诸如茶、咖啡、巧克力热饮、巧克力冷饮、奶、汤或婴儿食品,该机器包括：胶囊识别模块,其用于识别在胶囊识别位置处被插入所述机器中的胶囊,该胶囊识别模块包括用于捕获处于所述胶囊识别位置的所述胶囊的至少一部分的图像的相机；其中所述胶囊识别模块包括神经网络计算装置,所述神经网络计算装置被配置为基于由所述相机捕获的所述胶囊的至少一部分的图像来确定多个预定义胶囊类型中的所述胶囊的类型。(The present invention relates to a machine for preparing and dispensing beverages such as tea, coffee, hot chocolate, cold chocolate, milk, soup or baby food, comprising: a capsule identification module for identifying a capsule inserted into the machine at a capsule identification location, the capsule identification module comprising a camera for capturing an image of at least a portion of the capsule at the capsule identification location; wherein the capsule identification module comprises a neural network computing device configured to determine a type of the capsule of a plurality of predefined capsule types based on an image of at least a portion of the capsule captured by the camera.)

1. Machine for preparing and dispensing beverages such as tea, coffee, hot chocolate, cold chocolate, milk, soup or baby food, such machine comprising:

-an extraction unit for extracting a beverage ingredient capsule to form the beverage, for example a unit having a first part and a second part relatively movable between a distant position for inserting and/or removing a capsule and a close position for securing and extracting such capsule, such as a close position in which the first part and the second part delimit an extraction chamber, optionally at least one of the parts having a capsule opener, for example one or more capsule piercers, and/or at least one of the parts having an opening for inflow of a liquid to be mixed with ingredients contained in such capsule;

-a control unit for controlling the extraction unit to extract such capsules, such as a control unit powered by a power source, e.g. via an electric wire;

an outlet for dispensing the beverage formed by extraction of such capsules to a user receptacle, such as a cup or mug, located in a receptacle placement area, such as on a receptacle support, such as an external placement support on which such a machine is located, or a machine support, such as a movable or removable machine support, for collecting the beverage,

-a capsule identification module for identifying a capsule inserted into the machine at a capsule identification location, the capsule identification module comprising a camera for capturing an image of at least a portion of the capsule at the capsule identification location;

characterized in that the capsule identification module comprises a neural network computing device configured to determine the type of the capsule of a plurality of predefined capsule types based on an image of at least a portion of the capsule captured by the camera.

2. The machine of claim 1, wherein the plurality of predefined capsule types includes a capsule type corresponding to a capsule unknown to the capsule identification module.

3. The machine of one of the preceding claims, wherein the neural network computing device comprises a computing device, such as a microcontroller, microprocessor or any other suitable computing device, and a neural network computer program that implements a neural network when run on the computing device.

4. The machine of the preceding claim, wherein the neural network is a convolutional neural network.

5. The machine of one of the preceding claims, further comprising a network interface, such as a wired or wireless network interface.

6. The machine of the preceding claim, wherein the neural network computing device is configurable and/or can be updated remotely through the network interface.

7. The machine of any preceding claim, wherein the capsule identification module comprises at least one light source, such as at least one LED, for illuminating a capsule at the capsule identification location, and a diffuser for diffusing light of the at least one light source towards the capsule identification location.

8. Machine according to the preceding claim, the diffuser forming a conical cavity extending from the camera to the capsule recognition position and opening towards the capsule recognition position.

9. Method of configuring a neural network computing device of a machine according to one of the preceding claims, the method comprising:

-training a neural network computer program external to the neural network computing device by inputting a number of images of the capsule of the predetermined type until the neural network correctly determines the type of each next capsule image with a probability above a predetermined threshold;

-copying the trained neural network computer program into the neural network device of the beverage preparation machine.

10. Method of updating a neural network computing device of a machine according to one of the preceding claims, the method comprising:

-training a neural network computer program outside the neural network computing device by inputting a number of images of the capsule of the predetermined type, including a new previously unknown type, until the neural network correctly determines the type of each next capsule image with a probability above a predetermined threshold;

-copying the trained neural network computer program into the neural network device of the beverage preparation machine.

11. The method according to the preceding claim, the copying step being performed over a data network, for example over the internet, for example from a remote server to the neural network computing device.

12. A neural network computer program for determining a type of a capsule of a plurality of predefined capsule types based on an image of at least a portion of the capsule captured by a camera of the beverage preparation machine.

13. The neural network computer program of the preceding claim, which implements a convolutional neural network when the neural network computer program is run on a computing device.

14. The neural network computer program of the preceding claim, wherein the convolutional neural network comprises at least one convolutional layer and at least one fully-connected classification layer.

15. The neural network computer program of claim 13, wherein the convolutional neural network comprises three convolutional layers followed by a pooling layer and two layers of fully-connected classifiers.

Technical Field

The field of the invention relates to beverage preparation machines using capsules with ingredients for the beverage to be prepared. The field of the invention relates in particular to beverage preparation machines using capsules and configured to automatically identify the type of capsule inserted into the machine, for example to adapt beverage preparation parameters to the identified type of capsule.

For the purposes of this specification, "beverage" is intended to include any liquid substance that is edible to humans, such as tea, coffee, hot or cold chocolate drinks, milk, soup, baby food, and the like. "capsule" is intended to include any pre-portioned beverage ingredient, such as a flavoring ingredient, placed in a closed package, particularly an airtight package, made of any material (e.g., plastic wrap, aluminum wrap, recyclable wrap, and/or biodegradable wrap) and can be of any shape and configuration, including a soft pod or rigid cartridge containing the ingredient. The capsule may contain a quantity of ingredients for preparing a single-serving beverage or multiple-serving beverage.

Background

Some beverage preparation machines use capsules containing the ingredients to be extracted or to be dissolved and/or the ingredients stored and dosed automatically in the beverage preparation machine or added when preparing the beverage. Some beverage machines have a filling device comprising a pump for a liquid, usually water, which is used to pump the liquid from a water source, which is cold or actually heated by a heating device, such as a heating block or the like.

In the field of coffee preparation in particular, machines have been widely developed in which a capsule containing beverage ingredients is inserted in a brewing device.

Brewing devices have been developed that facilitate the insertion of "fresh" capsules and the removal of used capsules. Generally, the brewing device comprises two parts that are relatively movable from a configuration for inserting/removing the capsule to a configuration for brewing the ingredients in the capsule.

The movable part of the brewing device may be actuated manually, as disclosed in WO 2009/043630, WO 01/15581, WO 02/43541, WO 2010/015427, WO 2010/128109, WO 2011/144719 and WO 2012/032019. Various handle configurations are disclosed in EP 1867260, WO 2005/004683, WO 2007/135136, WO 2008/138710, WO2009/074550, WO 2009/074553, WO 2009/074555, WO 2009/074557, WO 2009/074559, WO2010/037806, WO 2011/042400, WO 2011/042401 and WO 2011/144720. Integration of such arrangements into a beverage machine is disclosed in WO2009/074550, WO 2011/144719, EP2014195046, EP2014195048 and EP 2014195067.

The movable part of the brewing device can be actuated by a motor. Such a system is disclosed, for example, in EP 1767129. In this case, the user does not have to manually turn the brewing device on or off. The brewing device has a capsule insertion channel provided with a safety door assembled to a movable part of the brewing device via a switch for detecting an undesired presence of a finger in the channel during closure thereof and preventing the finger from being injured by squeezing. Alternative lids for the capsule insertion channel are disclosed in WO 2012/093107 and WO 2013/127906. Different motorized systems are disclosed in WO 2012/025258, WO 2012/025259 and WO 2013/127476.

To allow a user to interact with such machines to provide operating instructions to the machines or to obtain feedback therefrom, various systems have been disclosed in the art, such as those mentioned in the following references: AT 410377, CH 682798, DE 4429353, DE 20200419, DE 202006019039, DE 2007008590, EP 1448084, EP 1676509, EP 08155851.2, FR 2624844, GB 2397510, US 4,377,049, US 4,458,735, US 4,554,419, US 4,767,632, US 4,954,697, US5,312,020, US5,335,705, US5,372,061, US5,375,508, US5,645,230, US5,685,435, US5,731,981, US5,836,236, US5,959,869, US6,182,555, US6,354,341, US6,759,072, US 2007/0157820, WO 97/25634, WO 99/50172, WO 2004/030435, WO 2004/030438, WO 2006/063645, WO 2006/090183, WO 2007/003062, WO2007/003990, WO 2008/104751, WO 2008/138710, WO 2008/138820, WO 2010/003932, WO2011/144720 and WO 2012/032019.

To facilitate the operation of the machine, the capsules supplied to the machine may be automatically identified and then automatically handled and extracted, as disclosed for example in WO 2012/123440.

There is still a need to improve beverage dispensing by reliably automatically identifying the machine of the capsule.

Disclosure of Invention

The present invention relates to a machine for preparing beverages. The beverage preparation machine may be a domestic or non-domestic machine.

The machine can be used for preparing coffee, tea, chocolate, cocoa, milk, soup, baby food, etc.

Beverage preparation typically involves mixing a variety of beverage ingredients, such as water and powdered milk; and/or brewing beverage ingredients, such as brewing ground coffee or tea with water. One or more of such ingredients may be provided in loose and/or agglomerated powder form and/or in liquid form, in particular in concentrate form. A carrier or diluent liquid (e.g., water) may be mixed with such ingredients to form a beverage. Typically, a predetermined quantity of beverage corresponding to a portion (e.g., a serving) is formed and dispensed upon request by a user. The volume of this portion may be in the range of 25mL to 200mL, even up to 300mL or 400mL, for example the volume of a filled cup, depending on the type of beverage. The beverage formed and dispensed may be selected from ristrettos, espresso, lungo, cappuccino, lattemacchiato, latte, american coffee, tea, etc. In particular, the coffee machine may be configured to dispense espresso coffee, for example with an adjustable volume of 20ml to 60ml per serving; and/or for dispensing lungo coffee, for example in a volume in the range of 70ml to 150ml per serving.

The machine of the invention has an extraction unit for extracting the beverage ingredient capsule in the extraction chamber to form the beverage. The unit has a first part and a second part relatively movable between a distant position for inserting and/or removing capsules into and/or from the extraction chamber and a close position for fixing and extracting such capsules in the extraction chamber. In the closed position, the first and second parts generally define an extraction chamber.

The capsule may comprise a capsule body, for example a substantially straight, conical or disc-shaped body. The capsule may have a circular peripheral annular flange, e.g. a flexible or rigid flange, extending from a peripheral portion, e.g. an edge or surface of the capsule body. The capsule may contain flavour ingredients for preparing tea, coffee, hot chocolate, cold chocolate, milk, soup or baby food.

At least one of the first and second parts may define a cavity for receiving, for example, the ingredient within the capsule, such as a tapered cavity (e.g., a conical or pyramidal cavity) or a straight cavity (e.g., a cylindrical or trapezoidal cavity) or a disc-shaped cavity. Such cavities may extend along an axis that is substantially collinear with the direction of relative movement of the first and second members. An extraction chamber is then delimited on one side by such a cavity.

The other of these first and second parts may be delimited by another cavity and/or comprise an extraction plate, such as a plate provided with piercing elements for opening the flow-through face of the capsule or a non-invasive plate for cooperating with a pre-opened or self-opening flow-through face of the capsule.

Examples of extraction units are disclosed in WO 2008/037642 and WO 2013/026843.

At least one of these components may have a capsule opener, such as one or more capsule piercers.

The capsule may also include a self-opening mechanism. Self-opening capsules are disclosed, for example, in CH 605293 and WO 03/059778.

When using a closed capsule, the first and second parts may comprise a capsule opener, such as a blade and/or a tearing tool, e.g. a plate with a tearing profile, e.g. from Nespresso^TMPanels with tear profiles known to the machines or disclosed in EP 0512470, EP 2068684 and WO 2014/076041 and the references cited therein.

At least one of the parts may have an opening for the inflow of a liquid to be mixed with the ingredients contained in such a capsule.

The machine comprises a control unit for controlling the extraction unit to extract such capsules. The control unit may be powered by a power source, for example via an electrical cord.

The machine has an outlet for dispensing a beverage formed by extraction of such capsules to a user container, such as a cup or mug, located in a container placement area.

The preparation method of the flavoring beverage comprises the following steps: a carrier liquid, such as water, is circulated (using a liquid driver, e.g. a pump) into the capsule to flavor the liquid by exposure to the flavor ingredient held in the capsule, e.g. along an extraction direction, which may be substantially parallel to the direction of relative movement of the first and second parts or substantially parallel to the longitudinal or central direction of extraction.

For example, a user receptacle may be placed on the receptacle support to collect a beverage.

The container support may be formed by placing a support on the outside of such a machine.

The container support may be formed by a machine-included support, such as a movable or removable machine support.

The container placement area may be associated with a machine container support for supporting such user containers located below the outlet. The support may be associated with a drip tray, such as a drip tray supporting the support; and/or is movable vertically relative to the housing below the outlet and/or is movable away from the housing below the outlet, thereby enabling placement of user receptacles of different heights below the outlet. Examples of suitable container supports are disclosed in EP 0549887, EP 1440639, EP 1731065, EP 1867260, US5,161,455, US5,353,692, WO 2009/074557, WO 2009/074559, WO2009/135869, WO 2011/154492, WO 2012/007313, WO 2013/186339, WO 2016/096705, WO2016/096706 and WO 2016/096707.

In embodiments, the outlet may be fixed to or formed from or mounted on or in:

-a machine head having a deployed position in which the outlet is located above the container placement area and a collapsed position in which the outlet is retracted into the outer machine main housing, such as a machine head driven by at least one of the first and second components or by an actuator controlled by the control unit to move inwardly into and outwardly out of the main housing; and/or

A movable beverage guide having a beverage dispensing configuration to dispense beverage to the container placement area and a beverage stop configuration to prevent dispensing of beverage to the container placement area, e.g. by draining residual beverage from the guide to the waste container via a guide edge, such as a beverage guide driven between the dispensing configuration and the stop configuration by at least one of the first and second parts or by one (or the above) machine head or by an actuator controlled by the control unit.

For example, the machine is provided with a machine head, as disclosed in WO 2017/037212 and WO 2017/037215.

Examples of suitable waste containers for carrying out the invention are disclosed in EP 1867260, WO 2009/074559, WO2009/135869, WO 2010/128109, WO 2011/086087, WO 2011/086088, PCT/EP2017/050237 and WO 2017/037212.

The directional liquid guide may be defined entirely in the main body and/or the machine head.

Details of directional liquid guides suitable or adapted for carrying out the present invention are disclosed in WO 2006/050769, WO 2012/072758, WO 2013/127907, WO 2016/083488 and WO 2017/037212.

The extraction unit may comprise a capsule feeder for feeding capsules to the extraction chamber, the feeder having a capsule dispenser with a release configuration releasing such capsules from the feeder towards the extraction chamber and a holding configuration holding such capsules away from the extraction chamber.

The capsule dispenser may be formed by a mechanical and/or magnetic capsule door, such as for example a capsule holder having a shape complementary and matching at least a part of the outer shape of such a capsule.

The capsule holder may have a capsule door, such as a pivotable and/or translatable capsule door, movable between a position obstructing transfer towards the extraction chamber and a position not obstructing transfer towards the extraction chamber.

The capsule holder may have an actuator, such as an actuator controlled by a control unit, for changing from the holding configuration to the release configuration or vice versa.

The capsule holder may be formed by at least a portion of the first part and/or the second part of the extraction unit, for example by a surface and/or an edge of the first part and/or the second part hindering the transfer of the capsule towards the extraction chamber when the first part and the second part are in the close position and/or when they are in an intermediate position between the far position and the close position. The capsule holder is then actuated simultaneously with the first and second parts of the extraction unit.

The capsule dispenser may be changed from the release configuration to the holding configuration immediately after releasing the capsule towards the extraction chamber, such that a passage towards the extraction chamber is provided only when the capsule needs to be released.

The capsule feeder may comprise a channel for guiding such capsules to the extraction chamber in a predetermined capsule orientation for entry into the extraction chamber, such as a channel associated with a capsule holder for holding such capsules between the first and second parts in the remote position before relatively moving the first and second parts to their close position.

The interaction between the first and second parts (and optionally the capsule guiding channel) and the ingredient capsule may be of the type disclosed in WO 2005/004683, WO 2007/135135, WO 2007/135136, WO 2008/037642 and WO 2013/026856.

When the first and second parts are in or move towards the distant position, the control unit may control the capsule dispenser to release such capsules from the feeder so as to enter such capsules into the extraction chamber when the first and second parts return to their close position.

The control unit may control the capsule dispenser to hold such capsules in the feeder and away from the extraction chamber when the first and second parts are:

-in or relatively moved to an approach position; or

-in a distant position and about to move relatively to a close position, so that the time left for such a capsule if it is released from the dispenser is not sufficient to receive the capsule into the extraction chamber before the first and second parts reach the close position.

The capsule feeder may comprise or be associated with a capsule sensor connected to a control unit configured to bring the capsule dispenser into or maintain its holding configuration when the capsule sensor senses that there is no such capsule on or at the capsule dispenser. Examples of capsule sensors are disclosed, for example, in WO 2012/123440, WO 2014/147128, WO 2015/173285, WO 2015/173289, WO 2015/173292, WO 2016/005352 and WO 2016/005417.

The control unit may be configured to control the actuator such that after a predetermined period of time the first part and the second part are moved by the actuator: moving from the close position to the distant position and from the distant position to the close position, the predetermined time period starting for example from a beverage preparation triggering event such as for example capsule detection, capsule recognition, user actuation of a user interface of the machine, etc. or a combination thereof, for example the predetermined time period is in the range of 3 to 15 seconds, such as 5 to 12 seconds, for example 7 to 10 seconds.

Examples of such components that are relatively moved by an actuator (e.g. a motor) are disclosed in EP 1767129, WO 2012/025258, WO 2012/025259, WO2013/127476 and WO 2014/056641.

For example, the first and second components can be relatively movable by an actuator from a proximate position to a distal position and/or vice versa substantially along a vertical axis.

The machine may comprise a liquid supply for supplying liquid, such as water, into the extraction chamber, the liquid supply being connected to and controlled by the control unit to supply such liquid into the extraction chamber and to interrupt such supply, the operation of interrupting such supply being performed automatically and/or manually via a user interface connected to the control unit, and/or upon detection of removal of such container by the detection arrangement. For example, the liquid supply comprises one or more of: a source of the liquid, such as a liquid tank or a liquid connector for connecting to an external liquid provider; one or more liquid tubes for directing such liquid to the extraction chamber; a liquid driver for driving such liquid into the extraction chamber, such as a pump, for example an electromagnetic pump (reciprocating piston pump) or a peristaltic pump or a diaphragm pump; and thermal regulators, e.g., heaters and/or coolers, such as in-line thermal regulators, e.g., in-line flow regulators, for thermally regulating such liquids.

Examples of suitable liquid sources, such as tanks or connectors, are disclosed in WO 2016/005349, EP2015194020.2, PCT/EP2017/050237 and references cited therein.

The thermal regulator may be a water heater, a heating block or an On Demand Heater (ODH), for example of the ODH type disclosed in EP 1253844, EP 1380243 and EP 1809151.

Examples of pumps and their incorporation in beverage machines are disclosed in WO 2009/150030, WO 2010/108700, WO 2011/107574 and WO 2013/098173.

The control unit may be configured to control the liquid supply so as to:

-automatically supplying said liquid into the capsule extraction chamber when the first and second parts have reached their close position and the capsule is contained in the capsule extraction chamber, as the first and second parts move from the distant position to the close position, so as to mix said liquid with the ingredients contained in the capsule and form a beverage dispensed via the outlet, optionally after sensing the supply of such capsule to the unit by a capsule sensor (or sensors described above); and/or

-automatically supplying the liquid into the extraction chamber when the first and second parts have reached their close position and no capsule is contained in said chamber, in order to rinse or clean at least a part of the unit and optionally the outlet, the liquid supplier being for example configured to supply the liquid at a rinsing or cleaning temperature which is different from the temperature at which such liquid forms a beverage, for example by brewing.

In a particular embodiment, it is also contemplated to deliver a frozen or chilled beverage.

The control unit may be configured to control the liquid supply to not automatically supply liquid into the extraction chamber when the first and second parts have reached their close position and no capsule (e.g. detected or identified) is contained in the extraction chamber. For example, the control unit is configured to control the liquid supply to supply liquid into the extraction chamber upon sensing a corresponding manual user input on a user interface connected to the control unit.

The control unit may have an extraction end management program that automatically runs when the liquid supply is interrupted (for example, when a predetermined extraction process is ended or detected as faulty) to perform the following operations:

-moving the first and second parts relatively to each other immediately to their remote positions in order to remove any capsule from between the first and second parts; or

-keeping the first and second parts in the close position for a predetermined period of time (e.g. in the range of 1 to 5 seconds, such as 2 to 3 seconds) to allow a manual request (e.g. via a user interface connected to the control unit) to supply an additional amount of liquid into the extraction chamber via the liquid supply, and, in the absence of such a manual request within the predetermined period of time, relatively moving the first and second parts to their distant position in order to remove any capsules from between the first and second parts, e.g. moving such capsules into an old capsule collector formed by one (or the above) waste container.

For example, the first and second parts may be held into their remote positions for a predetermined period of time, such as a period in the range of 1 to 6 seconds, for example 2 to 4 seconds, before moving them into their close positions, thereby allowing a new capsule to be inserted between the first and second parts before relatively moving them into their close positions and be received in the extraction chamber for extraction of the new capsule.

Thus, a user may request that two (or more) beverages (e.g., double espresso) be dispensed into the same user receptacle.

According to the invention, a machine for preparing and dispensing beverages such as tea, coffee, hot chocolate, cold chocolate, milk, soup or baby food comprises:

-an extraction unit for extracting a beverage ingredient capsule to form the beverage, for example a unit having a first part and a second part relatively movable between a distant position for inserting and/or removing a capsule and a close position for securing and extracting such capsule, such as a close position in which the first part and the second part delimit an extraction chamber, optionally at least one of the parts having a capsule opener, for example one or more capsule piercers, and/or at least one of the parts having an opening for inflow of a liquid to be mixed with ingredients contained in such capsule;

a control unit for controlling the extraction unit to extract such capsules, such as a control unit powered by a power source, for example via an electric wire;

an outlet for dispensing a beverage formed by extraction of such a capsule to a user receptacle, such as a cup or mug, located in a receptacle placement area, such as on a receptacle support, such as an external placement support on which such a machine is located, or a machine support, such as a movable or removable machine support, for collecting said beverage,

-a capsule identification module for identifying a capsule inserted into the machine at a capsule identification position, the capsule identification module comprising a camera for capturing an image of at least a portion of this capsule at the capsule identification position;

wherein the capsule identification module comprises a neural network computing device configured to determine a type of the capsule of a plurality of predefined capsule types based on an image of at least a portion of the capsule captured by a camera.

The plurality of predefined capsule types includes, for example, a capsule type corresponding to a capsule unknown to the capsule identification module.

The computing means comprise, for example, a computing means such as a microcontroller, microprocessor or any other suitable computing means, and a neural network computer program which implements a neural network when run on the computing means. The neural network is, for example, a convolutional neural network.

In an embodiment, the machine further comprises a network interface, such as a wired or wireless network interface. The neural network computing device is then, for example, configurable and/or can be updated remotely via a network interface.

In an embodiment, the capsule identification module comprises at least one light source, e.g. at least one LED, for illuminating the capsule at the capsule identification position and a diffuser for diffusing the light of the at least one light source towards the capsule identification position. The diffuser preferably forms a conical cavity extending from the camera to the capsule recognition position and opening towards the capsule recognition position.

According to the invention, the method of configuring the neural network computing device of such a machine comprises:

-training a neural network computer program external to the neural network computing device by inputting a number of images of a capsule of a predetermined type until the neural network correctly determines the type of each next capsule image with a probability above a predetermined threshold;

-copying the trained neural network computer program into a neural network device of the beverage preparation machine.

According to the invention, the method of updating the neural network computing means of such a machine comprises:

-training a neural network computer program outside the neural network computing device by inputting a number of images of a capsule of a predetermined type, including a new previously unknown type, until the neural network correctly determines the type of each next capsule image with a probability above a predetermined threshold;

-copying the trained neural network computer program into a neural network device of the beverage preparation machine.

The step of copying the trained neural network computer program into the neural network computing device of the machine may be performed over a data network, such as over the internet, such as from a remote server to the neural network computing device.

The invention also relates to a neural network computer program for determining a type of a capsule of a plurality of predefined capsule types based on an image of at least a portion of the capsule captured by a camera of a beverage preparation machine.

The neural network computer program preferably implements a convolutional neural network when run on a computing device. In an embodiment, the convolutional neural network comprises at least one convolutional layer and at least one fully-connected classification layer. In an embodiment, the convolutional neural network includes, for example, three convolutional layers, each followed by a pooling layer, and two layers of fully-connected classifiers.

The use of a well-trained neural network to identify the type of capsule allows for a rapid determination of the capsule type.

Furthermore, an image comparison process in the beverage preparation machine is avoided, which requires high computational resources and large storage space for storing sample images per capsule type.

Drawings

The invention will now be described with reference to the schematic drawings, in which:

figure 1 is a perspective view of a machine according to the invention;

figure 2 is a cross-sectional view of a capsule identification module according to an embodiment of the invention;

figure 3 is an exploded view of the capsule identification module of figure 2;

figure 4 is a cross-sectional view of a capsule identification module according to another embodiment of the invention;

figure 5 is an exploded view of the capsule identification module of figure 4.

Detailed Description

Fig. 1 shows an exemplary embodiment of a beverage maker 1 according to the present invention for preparing and dispensing a beverage, such as tea, coffee, a hot chocolate drink, a cold chocolate drink, milk, soup or baby food. The ingredients may be supplied in the form of ingredient capsules, for example of the type described in the above section entitled "technical field".

The machine 1 comprises an extraction unit for extracting a beverage ingredient capsule to form a beverage. The extraction unit has, for example, a first part and a second part, which are preferably located inside the machine housing, and therefore not visible in fig. 1, and are relatively movable between a distant position for inserting and/or removing the capsule and a close position for fixing and extracting the capsule (such as a close position in which the first part and the second part delimit an extraction chamber). For example, at least one of the parts has a capsule opener, such as one or more capsule piercers, and/or at least one of the parts has an opening for the inflow of a liquid to be mixed with the ingredients contained in the capsule.

The machine 1 comprises a control unit, preferably located inside the machine housing, for controlling the extraction unit to extract the capsule. The control unit may be powered by a power source (e.g. in a known manner via an electrical cord) or by a direct current power source (e.g. a battery such as a car battery or a portable battery or a machine battery).

The machine 1 has an outlet 16 for dispensing the beverage formed by extraction of such capsules to a user receptacle (not shown), such as a cup or mug, located in a receptacle placement area 17 for collection of the beverage.

In an embodiment, the outlet 16 is e.g. fixed to or formed by or mounted on or in the machine head 10, wherein the outlet 16 is located above the container placement area 17.

The extraction unit includes an actuator configured to relatively move the first member and the second member between relatively distant positions and relatively close positions thereof. The actuator is connected to and controlled by a control unit of the machine to move the first and second parts relative to each other.

The control unit is connected to input means for activating and/or controlling the extraction unit. According to the invention, the input means comprise, for example, a user interface 3, such as a touch screen.

Referring to fig. 2, the extraction unit may comprise a capsule feeder 2 for feeding capsules to the extraction chamber, and a capsule identification module 4 for identifying the type of capsule fed to the extraction chamber. The capsule recognition module 4 recognizes the type of such capsule, for example, when the capsule is in a capsule recognition position 20 within the capsule feeder 2 and before it is fed to an extraction chamber, which is not shown in the figures and is located, for example, below the capsule recognition position 20.

The machine preferably further comprises a capsule sensor, such as a mechanical and/or optical sensor (not shown), for sensing the presence of a capsule proximate and/or located at the capsule identification location 20.

The capsule feeder 2 may have holding means (not shown) to hold the capsule at the capsule identification position 20. The holding means typically has a release configuration for releasing the capsule from the capsule recognition position 20 towards the extraction chamber, e.g. under the influence of gravity, and a holding configuration for holding the capsule away from the extraction chamber, e.g. above the extraction chamber, at the capsule recognition position 20.

The capsule feeder 2 may have a channel for guiding the capsule to the extraction chamber into a predetermined capsule orientation for entering the extraction chamber through the capsule identification position. The passage for example comprises a guide groove 21 into which the flange of the capsule is inserted while the capsule travels through the passage, thereby keeping the capsule in a predetermined orientation, preferably adapted for the capsule identification module 4 to identify the capsule and/or to insert the capsule into the extraction chamber.

When the first and second parts of the extraction unit are in the distant position or when they are moved towards the distant position, the control unit may control the holding means to release the capsule from the capsule recognition position 20 so as to bring the capsule into the extraction chamber when the first and second parts are returned to their close position. The holding means comprise, for example, the surfaces and/or edges of the first part and/or the second part of the extraction unit.

The capsule identification module 4 is preferably connected to the control unit and is configured to identify a type among the predetermined capsule types of the capsule located at the capsule identification position 20.

As explained in more detail further below, capsule identification module 4 identifies the type of capsule by capturing an image of the capsule and feeding it as input to a trained neural network computing device.

The control unit is for example configured to control the liquid supply of the machine according to a liquid supply program associated with the capsule type, such as a liquid supply program having one or more adjusted supply liquid parameters selected from liquid temperature, flow rate, pressure and volume, which are constant or variable during extraction of the identified capsule. For example, the capsule type may be selected from a plurality of known predetermined capsule types that can be extracted in an extraction chamber of the machine.

The capsule recognition module 4 is preferably positioned along the passage of the feeder 2 at the level of the capsule recognition position 20.

During use of the machine of the invention, the following steps may be performed:

-placing a container in a container placement area;

-inserting the capsules into the capsule feeder 2;

-identifying the type of capsule by the capsule identification module 4;

-automatically, semi-automatically or manually relatively moving the first and second parts of the extraction unit to their remote positions;

-supplying the capsules to an extraction chamber;

-relatively moving the first and second parts to their close position to position the capsule in the extraction chamber;

-applying extraction parameters determined on the basis of the identified type of capsule to extract the capsule located in the extraction chamber to prepare the beverage; and

dispensing the prepared beverage to the container via the outlet 16.

According to the invention, the capsule identification module 4 is configured to determine the type of capsule inserted into the beverage preparation machine at the capsule identification position 20 by capturing an image of said capsule and processing said image by the neural network computing means.

Beverage preparation machines generally allow different types of capsules to be extracted in order to prepare different beverages and/or different beverage styles. The different types of capsules extractable in the extraction chamber correspond, for example, to different ingredients contained therein and/or different ingredient conditioning.

In an embodiment, each type of capsule corresponds to a specific type of coffee, which is different from the coffee contained in the other types of capsules, such as, but not limited to, one or more of the origin, the degree of roasting, the level of grinding, the quantity of coffee contained in the capsule and/or its caffeine content. Alternatively or in combination thereof, the different types of capsules extractable in the beverage preparation machine 1 correspond to ingredients for preparing different beverages, such as, for example, coffee, milk, soup, baby milk powder, tea, cold drinks, etc.

Preferably, each type of capsule is associated with a particular aspect of the capsule, such as a color or combination of colors, particular characters and/or drawings, etc. formed on the surface of the capsule, so that, for example, a user may visually distinguish between different types of capsules. The neural network means of the capsule recognition module 4 comprises a neural network program, preferably previously trained, to recognize the type of capsule based on a digital image of at least a portion of said capsule. The neural network device typically includes a data processor for running a neural network program.

The machine 1 may be configured to extract each capsule using preparation parameters specific to a specific type of capsule. For example, the preparation parameters include one or more of: carrier liquid temperature, carrier liquid volume, extraction time, carrier liquid pressure, carrier liquid type, multiple successive preparation stages, etc. The preparation parameters used with each type of capsule extractable in the machine 1 are preferably stored in an internal or external data storage means connected or connectable to the control unit and/or to the capsule identification module 4. The appropriate preparation parameters for a specific capsule are selected based on the capsule type determined by the capsule identification module 4 and the control unit uses these preparation parameters to control the extraction of the identified capsule.

The machine 1 may also be configured to store and/or transmit information about each capsule type extracted in the machine to an external terminal and/or server, for example to monitor capsule consumption on the machine 1.

Referring to fig. 2, the capsule identification module 4 comprises a light source (e.g., one or more white LEDs 40 or any other suitable light source, preferably having a known and well-defined spectrum), a camera 41 (e.g., a CCD camera), and a neural network computing device (not shown in the figure, e.g., in the form of a microcontroller, microprocessor, or another suitable computing device configured to run a neural network computer program). The capsule identification module 4 preferably also comprises a controller (not shown in the figure, for example but not exclusively an ASIC or a programmable microcontroller) for controlling the light source, the camera 41 and the neural network computing device. Alternatively, the neural network computing device and the controller of the capsule identification module are the same component or integrated into a single component. The controller of the capsule identification module 4 is for example configured to switch the light source on and off and/or for receiving and processing signals from the camera 41. The light source, camera 41, neural network computing device and module controller are preferably attached, e.g. soldered, to an electronic board 42, typically a PCB, and provide them with the necessary power and data connections and/or interconnections in a known manner. The controller is preferably connected to and controlled by a control unit of the machine 1.

Preferably, the capsule identification module 4 further comprises a light guide 43 and a diffuser 44 for guiding light emitted by the light source towards the capsule identification position 20 and/or for preferably defining the light received by the camera 41 as light reflected by a capsule located at the capsule identification position 20 in order to avoid sensing stray light, e.g. ambient light.

Preferably, the light source is an extended light source. In an embodiment, the light source comprises, for example, a plurality of LEDs 40, preferably equally distributed around the camera 41. In the illustrated embodiment, the light source comprises, for example, four LEDs 40, only two of which are schematically shown in fig. 2, distributed around a camera 41. However, other types and/or distributions of extended light sources are also possible within the framework of the invention, such as for example a ring light source surrounding the camera 41.

The light guide 43 is, for example, in the form of a substantially transparent cover, which is associated with, for example at least partially covers, the light source and/or the camera 41. The cover comprises e.g. openings or other guiding means to guide light through the diffuser 44 to and from the capsule identification location 20. In the embodiment schematically illustrated in fig. 2, the light guide 43 for example comprises light guiding protrusions 430, each protrusion 430 extending from an LED 40 of the light source towards the diffuser 44 for guiding light emitted by the LED 40 into the diffuser 44.

The light guide 43 and/or the diffuser 44 are preferably configured to avoid that light emitted by the light source directly reaches the capsule identification position 20. In the embodiment shown in fig. 2, the light guiding protrusions 430 are for example configured such that a major part of the guided light is guided into the diffuser 44 against a portion of the inner wall of the diffuser where the light will be reflected, thus being reflected in a direction transverse to the longitudinal axis of the diffuser 44 extending from the camera 41 to the capsule position identification 20. The ends of the light-guiding protrusions 430 on the sides of the diffuser 44 are inclined, for example, at 45 °, so that most of the guided light is reflected on the inclined surface and directed towards the diffuser 44 at an angle substantially perpendicular to the longitudinal axis of the diffuser 44.

In the illustrated example, the diffuser 44 comprises a preferably light-colored, for example white, opaque element forming a truncated cone shaped cavity opening towards the capsule identification position 20. However, other shapes are also possible for the diffuser within the framework of the invention. However, the diffuser preferably forms a conical cavity extending from the camera 41 to the capsule recognition position, preferably opening towards the capsule recognition position.

The diffuser 44 is preferably configured to avoid reflections on the capsule located at the capsule identification position 20 and/or in the capsule identification module, which may lead to illumination failure of the capsule and thus to capsule identification failure. The inner surface of the diffuser 44 is, for example, textured in order to improve the diffusion of light travelling from the light source to the capsule identification position 20.

The diffuser 44 optionally comprises openings and/or protruding elements that interrupt and/or deviate the light rays emitted from the light sources in the direction of the capsule identification position 20. Alternatively or in combination thereof, the diffuser is made of a transparent or translucent material and the light from the light source is guided through the wall of the diffuser, e.g. by a light guide, so that the light is refracted out of the wall of the diffuser in a distributed manner and into the cavity in front of the capsule identification position.

Fig. 4 schematically illustrates an exemplary embodiment of a diffuser 44 comprising protruding elements 440 formed on its inner surface in order to prevent light (e.g. light guided by the light guide 43 and exiting the light guide 43 through the light guiding protrusions 430) from directly reaching the capsule identification position 20.

In the illustrated embodiment, the light guide 44 forms a semi-ellipsoidal cavity that opens toward the capsule recognition location 20, the camera 41 preferably being located at an end opposite the capsule recognition location 41.

In the illustrated embodiment, the protruding elements 440 at least partially cover the ends of the light guiding protrusions 430 of the light guide 43. The purpose of the protruding elements 440 is to block light from refracting out of the beveled ends of the light guiding protrusions 430 in a direction close to the longitudinal axis of the diffuser 44 to avoid such light from reaching the capsule identification location 20 directly, i.e. without first being reflected at least once by the inner walls of the diffuser 44. However, the protruding elements 440 of the diffuser 44 preferably comprise openings configured to allow light refracted towards the inner wall of the diffuser 44 to enter the diffuser cavity.

Optionally, the machine 1 comprises a capsule detector for detecting capsules located at or near the capsule identification location 20. The capsule detector is for example comprised in the capsule identification module 4, preferably attached to, for example welded on, the electronic board 42. However, other deployments of the capsule detector are possible within the framework of the invention. The capsule detector may be of any suitable type, for example, a presence detector and/or a motion detector, such as an Infrared (IR) detector, an inductive and/or resistive detector, a mechanical switching element, or the like. The capsule detector is controlled, for example, by the controller of the capsule identification module 4 or directly by the control unit of the machine.

In an embodiment, when a capsule is approached and/or placed on the capsule feeder 2 at the capsule recognition position 20, the capsule detector detects the presence of the capsule and sends a corresponding signal to the controller and/or the control unit of the machine, which signal activates the light source 40 to illuminate at least a portion of the surface of the capsule located at the capsule recognition position 20. The camera 41 is then activated to capture an image of at least a portion of the illuminated capsule. The captured digital image is then fed to and processed by a neural network computing device in order to determine the type of capsule. In other embodiments, for example, if the machine does not comprise any capsule detector, the capsule identification module 4 (in particular the light source 40 and the camera 41) is activated, for example, by user actuation of a user interface of the machine, for example by actuation of a beverage selector.

Preferably, the camera 41 is a CCD sensor or any suitable camera providing a digital image (e.g. a color or grayscale image) of at least a portion of the capsule. The images are fed as input data to a neural network computing device, which will run a neural network computer program using the input data in order to determine the type of capsule, wherein the parameters of the neural network, in particular its synaptic weights and deviations, have been previously set during training of the neural network implemented by the neural network computer program.

The neural network for determining capsule type according to the invention is trained, for example manually and/or automatically, on a training station comprising a capsule feeder for feeding capsules of various types to one or more capsule recognition positions, each capsule recognition position of the training station having a camera for taking a picture of at least a portion of each capsule fed at the corresponding capsule recognition position. Preferably, the illumination and image capturing conditions in the capsule recognition position of the training station are similar to the illumination and image capturing conditions at the capsule recognition position in the beverage preparation machine. The captured image is then fed as input to a computing device running the neural network computer program of the present invention. The output of the program (i.e. the neural network) for each image is compared with the actual type of the corresponding sample capsule, and the result of the comparison is fed back to the program to cause the program to adjust the neural network parameters, in particular the weights and/or deviations of one or more synaptic connections.

Once the desired rate of successful recognition is achieved, the neural network is preferably considered to be fully trained, and the neural network program with the adjusted parameters may be, for example, copied and loaded into a neural network computing device integrated or to be integrated into the beverage preparation machine of the present invention.

The neural network computer program of the present invention may be updated with a new training session to identify new types of capsules according to the above-described scheme, wherein the capsules comprise new types of capsules. Once the neural network learns a new type of capsule, the computer program with updated parameters may be copied and loaded into the neural network computing device of the functional machine to update its capsule identification module to a newer version. Optionally, one of the various identified capsule types corresponds to a capsule that is not known by the machine.

The training of the neural network computer program of the present invention is preferably performed on a training station external to the beverage preparation machine. However, the conditions under which the image of the sample capsule is captured are preferably similar or even identical to the conditions in the machine.

Examples of suitable architectures for neural networks implemented by the neural network computer program of the present invention are summarized in the following table:

table 1: examples of suitable neural network architectures

According to this example, the neural network is a convolutional neural network comprising twelve layers, where the first convolutional layer has an input size of 128 × 128 × 1. The neural network thus typically expects a 128 pixel by 128 pixel gray scale image as its input and performs 11-way classification. The neural network comprises, for example, five basic layers: three convolutional layers followed by two fully connected layers. Each convolutional layer uses a kernel of size 3 x 1 and step 1 with no padding to generate 32 output profiles. Each convolution is followed by a ReLU non-linear and maximally pooled sub-sampling operation. The output of the feature extraction section is then fed to a two-tier fully-connected classifier, such as a LogSoftMax classifier. The hidden layer of the classifier is constructed, for example, from sixty-four neurons and also uses the ReLU activation function.

Preferably, three-step pre-processing is performed on the image data inputs before they are forwarded through the neural network of the present invention. In a first pre-processing step, the data is preferably centered around zero by subtracting the average activation calculated over the entire training set of each pixel. In a second pre-processing step, the data is divided by the global standard deviation of the data set in order to normalize the range of input values. In the third and final preprocessing step, local contrast normalization is performed with a thirteen by thirteen gaussian weighted window in order to enlarge the edges in the picture.

However, other neural network architectures and/or data pre-processing are also possible within the framework of the present invention in order to achieve a reliable capsule type identification based on a digital image of the capsule or at least a part thereof.