TECHNICAL FIELD

The present invention relates to a classification system configured to identify an article, and a checkout counter comprising said classification system.

BACKGROUND

In today’s stores many different articles, such as food products, hygiene articles, etc. may be purchased which all have different sizes and shapes. Normally, a checkout operator handles each article manually and makes sure that the article is associated with the correct pricing at checkout for payment by the customer. This is traditionally done either by scanning a barcode attached to the article, manually inputting the PLU-code, manually inputting the price or a combination of those.

However, the environment for a checkout operator is often very stressful due to the constant need of identifying all articles as fast as possible in order to reduce the long queues. This is especially true for articles that need to be manually inputted either by price or PLU-code, since the checkout operator does not always know the codes and prices by heart and needs to check up the information in for example a book.

Fully automatic checkout counters are becoming an alternative for retail stores and supermarkets, thus completely removing the need of a checkout operator. However, automatic checkout counters are not always suitable, such as in the case of smaller convenience stores.

There is thus a need for improvement in the field of manned checkout counters. More specifically, there is a need for improved checkout counters allowing for better working conditions for the checkout operators.

SUMMARY

Accordingly, the present invention preferably seeks to mitigate, alleviate or eliminate one or more of the above-identified deficiencies in the art and disadvantages singly or in any combination and solves at least the above-mentioned problems by providing a classification system for identification of articles that generates a list of possible article identities which helps the checkout operator to efficiently identify unidentified articles.

In a first aspect, a classification system for identification of articles in a checkout counter is provided. The classification system comprises a barcode reading system being arranged to identify an article by scanning a readable barcode of an article, and a controller being in operative communication with a display. If the barcode reading system is unable to correctly identify the article, the controller is configured to generate at least one list of possible article identities for said article. The controller is configured to communicate the at least one list of possible article identities to the display which is configured to display said at least one list. The display is arranged to be used by a checkout operator for manual identification of the article based on the list of possible article identities.

The list of possible article identities may comprise article identities being determined, by the controller, to have the highest probability of corresponding to the actual article.

The classification system may further comprise a memory unit configured to store article identities.

The classification system may further comprise a weight sensor configured to weigh the articles.

In one embodiment, the classification system further comprises at least one sensor arranged to at least in part identify the articles that the barcode reading system is unable to correctly identify.

The controller may be configured to generate a list of possible article identities based on the output from the at least one sensor. The at least one sensor may be a weight sensor.

In one embodiment, the controller is configured to generate a list of possible article identities based on statistical analysis.

The classification system may further comprise at least one sensor arranged to at least in part identify the article which the barcode reading system is unable to identify, wherein the controller is configured to generate a first list of possible article identities based on the output from the at least one sensor and to generate a second list of possible article identities based on statistical analysis.

In one embodiment the display is arranged in the classification system. In an alternative embodiment the display is arranged in a POS-system. In yet one embodiment the display is arranged in a handheld device.

In a second aspect a checkout counter is provided. The checkout counter comprises a classification device according to the first aspect.

In a third aspect, a method for classifying an article in a checkout counter is provided. The method comprises the steps of scanning the article using a barcode reading system and identifying articles being provided with a readable barcode, and if the barcode reading system is unable to correctly identify an article: generating at least one list of possible article identities, displaying the at least one list of possible article identities; and selecting, by manual input from a checkout operator, the correct article among a plurality of displayed article identities.

The method may further comprise adding the selected article to a registration account.

If the article is correctly identified by the barcode reading system, the article may be added to a registration account.

If the barcode reading system is unable to correctly identify the article, the method may further comprise stopping at least one conveyor belt before displaying the at least one list of possible article identities.

The weight of the article may be used to calculate the price before adding the article to the registration account. This is true for articles that have a price which are dependent on its weight.

In one embodiment, the method further comprises the step of weighing the article, and if the weight is incorrect in relation to the article identity determined by the barcode reading system. The method further comprises generating at least one list of possible article identities, displaying the at least one list of possible article identities; and selecting, by manual input from a checkout operator, the correct article among a plurality of displayed article identities. In a fourth aspect a checkout counter system configured to be operated by a single checkout operator is provided. The checkout counter comprises at least two checkout counters and at least one classification system according to the first aspect.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the invention will be described in the following; reference being made to the appended drawings which illustrate non-limiting examples of how the inventive concept can be reduced into practice.

Fig. la schematically shows a top view of a checkout counter according to an embodiment;

Fig. lb schematically shows a top view of a checkout counter according to an embodiment;

Fig. lc schematically shows a top view of a checkout counter according to an embodiment;

Fig. 2a is a schematic workflow of a method of a classification system according to an embodiment;

Fig. 2b is a schematic workflow of a method of a classification system according to an embodiment;

Fig. 2c is a schematic workflow of a method of a classification system according to an embodiment;

Fig. 3a schematically shows a display of a classification system according to an embodiment;

Fig. 3b schematically shows a display of a classification system according to an embodiment;

Fig. 3c schematically shows a display of a classification system according to an embodiment;

Fig. 4 schematically shows a classification system according to an

embodiment;

Fig. 5a schematically shows a top view of a checkout counter system according to an embodiment; Fig. 5b schematically shows a top view of a checkout counter system according to an embodiment;

Fig. 5c schematically shows a top view of a checkout counter system according to an embodiment;

Fig. 5d schematically shows a top view of a checkout counter system according to an embodiment;

Fig. 5e schematically shows a top view of a checkout counter system according to an embodiment; and

Fig. 5f schematically shows a top view of a checkout counter system according to an embodiment.

DETAILED DESCRIPTION OF EMBODIMENTS

Figs la-c shows a checkout counter 100 comprising a conveyor system 10 for transporting articles 3 from a loading area to a packing area, and a classification system 30 through which articles 3 pass for automatic or semi-automatic identification.

Preferably, a checkout operator 5, or store attendant, is positioned somewhere in conjunction to the checkout counter 100 for providing manual input when needed, as will be described more herein.

The conveyor system 10 extends from the loading area, and ensures article transport through the classification system 30. The direction of motion of the article 3 on the conveyor belts is shown in Figs la-c by arrow X. The conveyor system 10 comprises a loading conveyor 12 on which articles 3 may be placed by a customer (not shown), and a second conveyor 14 which is arranged in series with the loading conveyor 12. Preferably, a third conveyor 16 is provided in series after the second conveyor 14. Customers generally tend to place articles 3 very close to each other on the loading conveyor 12, or even in a stacked manner. The second conveyor 14 may be arranged so as to ensure that the articles 3 are separated from each other longitudinally, i.e. in the direction of movement. For this, the second conveyor 14 may be operated at a higher speed than the loading conveyor 12. In order to make sure that the articles 3 being loaded are aligned laterally, the loading conveyor 12, the second conveyor 14 and/or third conveyor 16 are tilted in the lateral direction with respect to a horizontal plane.

The checkout counter 100 may further make use of a customer divider bar 7 for separating articles 3 belonging to one customer from the next one in line. As will be described more in detail later on, the classification system 30 may be able to identify the customer divider bar 7 and thus realize that all articles 3 being associated with a specific customer have been scanned.

As seen in Figs la-c the classification system 30 comprises a barcode reading system 32 for scanning articles 3 and for identifying articles 3 being provided with a readable barcode and a controller 20. The classification system 30 may further comprise a weight sensor 31 for weighing the articles 3. The controller 20 is in operative communication with a display 40, wherein the display either is arranged as part of the classification system (as is seen in Fig. la) or as a part of a separate POS-system (as in Figs. 5d-e). The different parts will now be described more in detail.

Preferably, the weight sensor 31 comprises one conveyor scale comprising one conveyor part and one weight unit connected thereto which automatically conveys the article 3, weighs it and transmit the information of the weight to a database in the controller 20. In this way the need of manual transport of the article 3 over the weight unit is removed. In some embodiments, one or several sensors may be connected to the checkout counter 100 for controlling the conveyor scale. The weight sensor 31 may be arranged either in the loading conveyor 12, the second conveyor 14 or the third conveyor 16.

In one embodiment, the weight of the article 3 is subsequently used by the controller 20 together with the article identity, which is determined either by the barcode reading system 32 or manually by the checkout operator 5, for verifying that the article on the weight sensor 31 corresponds with the weight information stored in the system. In this embodiment, the weight sensor 31 is used as control measurement.

In one embodiment, as shown in Fig. lb, the weight sensor 31 is connected both to the controller 20 as described with reference to Fig. la and a POS-system 50.

The POS-system 50 performs a certified weighing. Certified weighing might be required in some environments. The certified weighing may have a greater accuracy than the weight information gathered by the controller 20. For articles not having a barcode and where the price of the article is dependent on its weight, the POS-system will use the certified weighing to determine the price of the article. Hence, in this embodiment and for articles whose price is dependent on its weight, the information transmitted by the controller 20 to the POS-system 50 does not comprise information of the weight and/or price of the article.

In the situation where the controller 20 finds a barcode on an article where its price is dependent on its weight, the weight will be transmitted as a part of the barcode.

As seen in the system shown in Fig. lb, the POS-system 50 may comprise a display 55. This display 55 may be in communication with the controller 20, wherein the controller 20 may be configured to communicate article identities on the display 55 of the POS-system. In such an arrangement the display 40 in the classification system is not needed (as illustrated by the dotted lines).

In yet one embodiment, as shown in Fig. lc, the weight sensor 31 is arranged to perform the certified weighing without the need of a POS-system. The information transmitted to the POS-system 50 from the controller 20 will, for articles whose price is dependent on its weight, comprise information relating to both the price and/or the weight of the article.

The barcode reading system 32 is arranged in conjunction to the loading conveyor 12, the second conveyor 14 and/or the third conveyor 16. The barcode reading system 32 scans all articles 3 and identifies those articles 3 being provided with a readable barcode. As will be described later, articles 3 having no barcode or a defect barcode will be identified by other means. The barcode reading system 32 may comprise at least one camera for providing still or moving images. The barcode reading system 32 may be connected to an image processing unit, possibly realized by means of the controller 20, which image processing unit allows the read barcode to be checked against pre-stored article identities.

The barcode reading system 32 may further comprise a second camera and possibly several cameras to be able to see the article 3 from different angles for achieving the highest possible reliability when detecting the barcode. The other camera, and if applicable a further cam era/cam eras, is/are arranged to record an image or images which will be used by the image processing unit for analysis of a barcode reading. It is preferred if the barcode can be read regardless of the position of the barcode on the article 3, or the position of the article 3 on the conveyor system 10. Preferably, the barcode reading system 32 comprises at least four cameras, arranged above, below and on both sides of the conveyor system 10.

In the case where the barcode reader system 32 is unable to correctly identify the article 3, for example due to a defect barcode or due to a missing barcode, it is beneficial if the checkout operator 5 receives information of possible article identities in order to speed up his/her manual identification of the article 3. This information is presented as at least one list in the display 40;55, where the checkout operator 5 selects the correct article identity among a plurality of article identities. The at least one list may be generated by the controller 20, which then communicates the list to the display arranged in conjunction to the checkout operator 5. This speeds up the manual identification process significantly, since the checkout operator in prior art systems would have to manually input the PLU-code or, if the article is arranged with a barcode, scan the barcode.

In the case where the barcode reader system 32 succeeds in identifying the article 3, no article list is shown and no manual input is needed. The display 40;55 is thus arranged to show information of article identities in order to assist the checkout operator to manually identify the article 3 only when needed. This provides a more efficient checkout counter, where the time required to identify articles 3 is reduced since the staff i) does not have to manually scan the articles 3 having a barcode, and ii) does not have to manually identify and write down the PLU-code or price for articles 3 not having a barcode.

The display 40;55 may be in the form of a touch sensitive display having a user-friendly interface or a non-touch display being controlled by a keyboard, numpad, buttons or similar device allowing for manual input. The display 40; 55 could also be a touch sensitive display which additionally could be controlled by for example a keyboard and/or buttons arranged on the display 40;55 or as an external device.

In one embodiment the display 40;55 is a touch sensitive display configured to display and operate one or more virtual keys on the touch display. The display 40;55 is arranged to display graphical objects such as menu items, article identities etc., as will be described more in detail with reference to Figs. 3a-c.

In one embodiment the display 40;55 is arranged in the checkout counter 100 as a stationary part. The display 40;55 may for example be arranged in conjunction to the conveyor system 30 and/or in conjunction to the barcode reader system 32. In an alternative embodiment the display 40;55 is an external device which could be handheld. The handheld external device may for example be a mobile communication terminal such as a mobile phone, a tablet computer, a personal digital assistant, or generally any hand-held, user-carried or user-worn device capable of displaying information and communicating with other devices. The communication between such an external device and the controller 20 may be performed by a communication interface, which will soon be described further.

As previously stated, the classification system 30 further comprises a controller 20. The controller 20 may be implemented as one or more processors (CPU) or programmable logic circuits (PLC), which is connected to or comprises a memory 22. The memory 22 may be implemented using any commonly known technology for computer-readable memories such as ROM, RAM, SRAM, DRAM, FLASH, DDR, SDRAM or some other memory technology. The memory 22 is configured to store article identities. The controller 20 is at least in communication with the weight sensor 31, the barcode reading system 32, and the display 40;55 in order to at least in part identify the article 3.

Methods of classifying articles

Before turning into a detailed description of other possible parts of the classification system 30 a general method of classifying articles will be described in conjunction to Fig. 2a.

A new customer starts arranging its articles onto the loading conveyor 12 of the checkout counter. The checkout counter 100 may recognize 210 a new customer manually by the checkout operator 5 pressing a designated button, preferably arranged in conjunction with the display 40;55, or alternatively the checkout counter 100 may automatically recognize a new customer. The automatic recognition may for example be done using a customer divider bar 7 being arranged with a machine-readable medium. The machine-readable medium may for example be a barcode, a RFID tag/label, magnetic disks, digital watermarks, cards, tapes, and drums, punched cards and paper tapes, optical discs, barcodes and/or magnetic ink characters. The machine-readable medium is preferably arranged on the customer divider bar 7. Additionally, or alternatively, the customer divider bar 7 is recognized by its weight. In a preferred embodiment, the customer divider bar 7 is arranged with a barcode.

The automatic recognition of the customer divider bar 7 may additionally or alternatively be done using pattern recognition. Hence, the customer divider bar 7 may be arranged with a specific pattern that allows the system to identify it.

Once a customer has placed all his/her articles on the loading conveyor 12, the customer or the checkout operator 5 places a customer divider bar 7 on the loading conveyor 12, after his/her articles. The classification system 30 identifies the customer divider bar 7 and recognizes that all articles have been scanned for that particular customer.

Once a new customer is recognized, the checkout counter 100 initiates movement of the conveyor belts, if these were not already moving.

The classification system 30 then tries to identify 220 the articles using the barcode reading system 32. If the article is provided with a readable barcode, the article will be identified and its article identity will be added 270 to a registration account of the customer.

If the classification system 30 is unable to correctly identify the article by reading the barcode, the controller 20 is configured to generate at least one article list 240 showing article identities having a high probability of corresponding to the actual article. The at least one article list may be generated in several ways. The at least one article list may for example be generated using other sensors present in the classification system 30, as will be described further with reference to Fig. 4. The one or additional sensor(s) may for example be one or several cameras, used possibly together with the weight sensor 31 of the classification system 30, in order to identify possible article identities that then are presented on the display 40;55. Additionally, or alternatively, the list may be generated using statistical analysis.

The at least one article list is shown 250 on the display 40; 55. The checkout operator 5 views the alternative article identities and selects, from the plurality of article identities, the correct article identity from the list so that the article becomes identified. This speeds up the manual identification process significantly, since the checkout operator 5 in prior art systems would have to manually write in the PLU-code in the system or, if the article is arranged with a barcode, scan the barcode.

Once the checkout operator 5 has identified the article having a defect or no barcode, the article is added 270 to the registration account of the customer. The classification system 30 will thus continue to try to automatically identify the rest of the articles belonging to the customer.

In one embodiment, some articles may be associated with a flag in the classification system 30 which transmits a stop signal to the controller 20 which is configured to stop 230 the movement of the conveyor system 10 if the flag is detected. The controller 20 would then notify, by showing information on the display 40; 55, the checkout operator 5 of the flag and after the operator 5 has approved the flag, the controller 20 is configured to transmit a start signal to the conveyor system 10 in order to start 260 the movement of the conveyor belts if the belts where previously stopped. This is preferably used to prevent articles with age restrictions, such as alcoholic beverages, to be purchased by minors, but can also be used for manual verification of e.g. products sold by weight or piece and/or products sold in multipacks.

When all articles of the customer has been identified, the system recognizes that the customer is finished, and recognizes 210 a new customer. In one embodiment, the checkout counter 100 may further be equipped with a buffering mode, which enables the recognition 210 of a new customer after all articles of the previous customer have been identified, but before the previous customer has finished the payment operation, in order to speed up the checkout process. In this embodiment, it is preferable for the controller 20 to allow articles to be added 270 to the registration account of the new customer, but prevent these registered articles to be transmitted to the point of sale system until after the previous customer has finished the payment operation, to avoid accidental payment of another customer’s articles.

In a situation where the classification system 30 fails to automatically identify the article, the system may facilitate for the checkout operator 5 in more ways than displaying at least one list of possible article identities. In addition to displaying at least one list, the movement of the belts of the conveyor system 100 may be altered.

In one embodiment, a stop signal is transmitted to the controller 20 which is configured to stop 230 the movement of the conveyor system 10 if the classification system 30 fails in automatically identifying the article. Once the checkout operator 5 has identified the article having a defect or no barcode, the controller 20 is configured to transmit a start signal to the conveyor system 10 in order to start 260 the movement of the conveyor belts if the belts where previously stopped.

Additionally or alternatively, the speed of the conveyor system 10 is reduced if the classification system 30 fails in automatically identifying the article. Once the checkout operator 5 has identified the article having a defect or no barcode, the controller 20 is configured to transmit a signal to the conveyor system 10 in order to increase 206 the speed of the conveyor belts if the speed of the belts where previously reduced.

In one embodiment, the speed of the conveyor system 10 is always such that the staff is capable of manually identifying an article by selecting from the list of article identities shown on the display 40;55 while the conveyor belts 10 are moving.

Fig. 2b illustrates a further method of identifying articles which are weight- priced articles. If an article was automatically identified in step 220, weight information is received 280 and the price for the article is calculated. If the article instead was identified by a checkout operator 5 identifying the article from an article list, the weight information is received and the price is calculated as soon as the attendant has identified the article. It should be noted that the weighing of the article could be done before or after the step of trying to identify the article using the barcode reading system 32. The scanning of the barcode and the weighing of the article could also be performed simultaneously. Fig. 2c illustrates an embodiment where the classification system 30 takes into account the fact that articles may be bought in multipack. The customer may have chosen a multipack of six water bottles, where he/she, or someone else, has removed two bottles so that the multipack only consists of four bottles. The barcode may be set for the whole multipack of six bottles, and in prior art systems the customer would then have to pay for all six bottles although he/she only buys four bottles. This issue is prevented in the method shown in Fig. 2c.

The checkout counter 100 recognize 210 a new customer, as has previously been described. Once a new customer is recognized, the checkout counter 100 initiates movement of the conveyor belts, if these were not already moving. The articles are then being weighed 215 by the weight sensor 31. In a next step, 220, the classification system 30 tries to identify 220 the articles using the barcode reading system 32. If the article is identified by the barcode reading system 32, the controller 20 is configured to check 224 if the weight is correct in relation to the article identity. If the article is identified by the barcode reading system 32 and the weight is correct, its article identity will be added 270 to registration account of the customer.

If the article is identified by the barcode reading system 32 but the weight is incorrect in relation to the article identity, the classification system 30 will generate a list of possible articles so that the checkout operator 5 may select the correct article identity and price manually. In the case of the water bottles, the list may comprise a single water bottle, a multipack of water bottles comprising two bottles, a multipack of water bottles comprising three bottles, a multipack of water bottles comprising four bottles and multipacks comprising other numbers of bottles - thus allowing the checkout operator 5 to choose the correct amount of bottles.

If the classification system 30 fails in reading the barcode, the controller 20 is configured to generate 240 at least one article list showing article identities having a high probability of being the article which the barcode reading system 32 failed to identify. The checkout operator 5 views the alternative article identities and chooses the correct article from the list, so that it becomes identified. Once the checkout operator 5 has identified the article having a defect or no barcode, the article is added 270 to the registration account of the customer. The classification system 30 will then continue to try to automatically identify the rest of the articles belonging to the customer. When all articles of the customer have been identified, the system recognizes that the customer is finished, and recognizes 210 a new customer.

The display

As has previously been described, the display 40;55 displays information of possible article identities when the barcode reader system 32 has failed to identify an article. The information is preferably presented in at least one article list. The information in the at least one article list is then used by the checkout operator 5, whom selects the correct article identity among a plurality of article identities, preferably being at least two article identities, shown on the display 40;55. This speeds up the manual identification process significantly, since the checkout operator 5 in prior art systems would have to manually write in the PLU-code in the system, or if the article is arranged with a barcode, scan the barcode.

Fig. 3a-b illustrates a display 40;55. In Fig. 3a, a display is shown in a situation where the articles are identified by the barcode reader system 32 and no manual input is needed and Fig. 3b-c illustrates a situation where manual input is needed.

In one embodiment the display 40;55 is configured to be arranged in an idle mode and an active mode. The display is preferably less bright when it is arranged in the idle mode compared to the active mode. The display 40;55 is driven in its idle mode when no input from the checkout operator 5 is needed, and is driven in its active mode once the system requires input from the checkout operator 5. In one embodiment the display 40;55 is configured to be set from the idle mode to the active mode when the barcode reader system 32 has failed to identify an article. Hence, the display 40;55 will stay in its idle mode when articles are identified using the barcode reader system 32.

This is both energy efficient and works as an alert for the checkout operator 5 that manual input is needed once the display lights up.

In Fig. 3a, the display 40;55 is shown having two menu items 42a-b. The menu items may for example be a pause button in order to pause the movement of the conveyor system 10 and/or a button for finishing the process of the customer (thus all articles of the customer have been identified, and the customer shall receive payment information). Some menu items may only be visible in a situation where the articles are not identified by the barcode reader system 32 and manual input is needed. In some embodiments the display 40;55 may then display a button for manually inserting a PLU code. This would be used in the situation where the barcode reader system 32 fails to identify the article, and where the controller 20 does not provide a list of article identities showing the actual article.

Fig. 3b-c illustrates different information presented on the display 40;55 in a situation where the barcode reader system 32 has failed to identify an article. In Fig. 3b the display shows two article lists 44, 46, LIST 1 and LIST 2, each having a number of article identities. In this example each list contains five article identities, however it should be understood that any number of article identities could be shown. Preferably, each article list contains between 2 and 10 article identities. If the number of article identities is too high, the longer time it takes for the checkout operator 5 to choose the correct identity. On the other hand, if too few article identities are shown the risk that the display 40;55 does not show the correct identity increases.

In this embodiment the first article list 44, LIST 1, is generated based on a statistical analysis. The statistical analysis may be based on different kind of information, for example based on the article identity which most often fails to be identified by the barcode reader system 32 and/or based on the most bought articles not being arranged with barcodes. In some embodiments the first article list 44, LIST 1, is constructed using both information about the article identity that most often fails in combination with the information of the most popular articles. The statistical information is gathered by the controller 20 which presents said information to the checkout operator 5 in an article list. The list may contain a fixed number of article identities shown, or be dynamically arranged.

The article identities may be sorted in different order depending on the user settings of the classification device 30. For example, the article identity that most often fails to be identified by the barcode reader system 32 may be presented at the top, having the second most often failed article identity below, and so on. Another example is to display the top article identities in alphabetical order, or to sort the article identities based on its category (i.e. different fruits are arranged together, separated from candy).

The second list 46, LIST 2, is generated based on information gathered by the one or plurality of additional sensors, as will be described with reference to Fig. 4, in the classification system 30. Based on the output from the one or more sensors, which at least in part identifies the article, the controller 20 generates a list of possible article identities. The list may contain a fixed number of article identities, as in Fig. 3b, or be of dynamical range as in Fig. 3c. In the case of a dynamical list, the number of article identities shown thus vary depending on the probability that the article identity is in fact the article. The article identities are preferably shown in an order relating to the probability that said identity represents the article in question, hence the article identity having the highest identification probability may be presented at the top, the article identity having the second highest identification probability is presented below, and so on. However, the article identities could be arranged in any other order, as previously described.

In one embodiment, the second list 46, LIST 2, is generated based on a combination of information gathered by the one or plurality of additional sensors and statistical analysis.

In Fig. 3c, the identification probability is shown for each article identity, showing the article identity having the highest probability at the top.

Although Fig. 3b-c illustrates a display 40;55 showing two article lists, it should be understood by a person skilled in the art that the display could show a dynamic number of lists. For example, the display could dynamically show a single list, three lists, four lists etc. depending on the configuration of the classification system 30. The display 40;55 could either display a list generated by statistical analysis and/or a list generated by at least in part identifying the article using the sensor arrangement in the classification system 30.

In an embodiment where the classification system 30 does not comprise any more identification sensors than the barcode reading system 32 and the weight sensor 31, the display 40;55 will be arranged to show at least one list generated by statistical analysis. It is beneficial if the classification system 30 alerts the checkout operator 5 that manual input is needed. As will be described further below, this notification can be done by the display and/or by communicating to an external device such as mobile communication terminal.

In one embodiment, the classification system 30 comprises an alarm system for providing feedback to the checkout operator 5. When the system 30 fails to identify an article using its barcode reader, a signal is transmitted to the alarm system, which in turn triggers an alarm to the checkout operator 5 signaling that manual input is needed. The alarm may for example be a sound, a blinking lamp, a bright color, a vibration, vibration pattern or the like. Additionally, or alternatively, the display 40;55 may display an visible information text that manual input is needed. The display 40;55 may comprise a loudspeaker, a lamp or similar device for gaining the attention of the checkout operator 5.

In one embodiment the classification system 30 further comprises a

communication interface (not shown). The communication interface may be configured to send information regarding the article identification to a mobile communications terminal (not shown) of a checkout operator 5 or other store personnel. The mobile communications terminal may be a mobile phone, a tablet computer, a personal digital assistant, or generally any hand-held, user-carried or user-worn device capable of communicating with other devices. This may be beneficial if the display 40;55 of the checkout counter 100 is not monitored by a checkout operator 5 at all times. This may for example be the case during times when there are few customers in the store and the checkout operator 5 can spend his/her time with other duties or in a situation where the checkout operator 5 is helping another customer. When the checkout counter 100 needs manual input from a checkout operator 5, the system 30 alerts the attendant by use of the communication interface.

The communication interface may be a wireless radio frequency interface such as a Bluetooth™ or a WiFi (IEEE802.1 lb standard) link. The communication interface may also be a wired interface. The communication interface may be configured to send an alarm as a text-message, Bluetooth signal, email or the like to a communication terminal of the checkout operator 5, informing him/her of the detected condition. The communication interface is preferably connected to or in communication with the controller 20.

The controller 20 is in operative communication with the display 40; 55. The display 40; 55 could thus either be part of the classification system 30 or being a part of a POS-system 50. The communication between the display 40; 55 and the controller 20 could be wired or wireless. Hence, a communication interface as has been described above could be implemented both in the display 40; 55 and in the controller 20.

In one embodiment, the classification system 30 is controllable by a mobile communications terminal of a checkout operator 5 or other store personnel. The checkout operator 5 may use his/her communication terminal to choose the correct article from the article list being displayed on the mobile terminal. Hence, in this embodiment the display of the communication system might be seen as the display 40; 55 of the classification system.

Additional sensor/ s) for generating an article list

Fig. 4 illustrates further components that might be present in the classification system 30 in order to generate an article list showing article identities having a high probability of being the unidentified article. In order to generate a list for the checkout operator 5 to choose from, the classification system 30 may comprise one or a plurality of sensor(s). The system 30 may comprise one or several of: a contour sensor 35 and /or a symbol reading sensor 36 which uses optical character recognition and (machine) text interpretation and/or a color texture sensor 38 and/or a color histogram sensor 37 and/or a first spectroscopy sensor 33, and/or a second spectroscopy sensor 34 and/or an object sensor 39. The symbol reading sensor 36 is from hereon called OCR which is a generally known abbreviation of the English expression“Optical Character

Recognition”.

The first spectroscopy sensor 33 may be an infrared spectroscopy sensor, from hereon denoted as a NIR sensor which is detecting wavelengths from approximately 780 nm to 2500 nm. The memory unit 22 comprises one or several first signatures created by the first NIR sensor 33 or another NIR sensor 34, each of which first signatures is connected to a corresponding article identity. The first signatures may be created directly at the checkout counter by using the first NIR sensor 33, a second NIR sensor 34, or by storing signatures created by a NIR sensor not connected to the checkout counter in said memory.

When a NIR sensor 33, 34 is used on a certain kind of article, e.g. a specific type of apple, a first signature will be received which may be coupled to the article and which may be denoted as a specific article identity in the memory unit 22, like e.g. the name of the article. Each type of article creates a unique first signature which may be coupled to the identity of the article. The first NIR sensor 33 is arranged to create a second signature connected to the article when an article is placed before, on or after the weight sensor 31. The controller 20 is subsequently arranged to compare the second signature to the first signature in order to identify the article as an existing article identity in the memory unit 22.

The second spectroscopy sensor 34 may be a VIS sensor 33, 34. The VIS sensor 33, 34 is a spectrometer comprising a light source and a VIS camera, from hereon called a VIS sensor 33, 34, the VIS sensor 33, 34 is detecting wavelengths from approximately 200 nm to 1100 nm. The spectrum thus overlaps the wavelengths of visual light which extends from 400 nm to 660 nm. Experiments have shown that, at the device according to the invention, the classification device comprising a color texture sensor 38 and/or a color histogram sensor 37 and/or a VIS sensor 33, 34 does not operate satisfactory when the VIS sensor 33, 34 is operating in the complete frequency interval 200 nm - 1100 nm since there is a conflict between the color sensors 37, 38 and the VIS sensor 33, 34 in the interval of visual light, i.e. between 400 nm and 660 nm.

The VIS sensor 33, 34 is therefore active in the intervals between 200 nm and 400 nm and between 660 nm and 1100 nm when it is combined with the color texture sensor 38 and/or the color histogram sensor 37. If the color texture sensor 38 and the color histogram sensor 37 are disconnected the VIS sensor 33, 34 may however operate in the complete frequency interval between 200 nm and 1100 nm since there is no conflict. The controller 20 is programmed to control the sensors to achieve optimal efficiency of the classification

The symbol reading sensor 36 is connected to a computer/image processing unit which uses an algorithm using information from images from the existing camera or cameras of the device. For articles, which substantially can be unambiguously identified by means of symbol reading, it will be sufficient if the symbol reading sensor 36, OCR, identifies a symbol and/or a text which then unambiguously identifies the article. Examples of articles which may be identified by only using a symbol reading sensor 36, OCR, are pre-packaged packages where the customer is not required to perform any procedure, such as refilling or any other procedure. Example of articles where it is not enough with the symbol reading sensor 36, are some bulk articles where the quantity of the article, i.e. weight, is not known. Further properties of the article may be necessary and may require symbol reading and/or weight and/or color histogram and/or color texture and/or contour. It shall be mentioned that“contour” is defined as a two dimensional projection of a three dimensional object.

Certain articles are thus more difficult to identify than others and depending on the article one or several of the included sensors of the classification device are required.

The contour sensor 35 comprises a camera for providing still or moving images and may preferably be a linear camera which reads a horizontally projected surface or a linear camera in combination with an object sensor 39 which consists of a vertical light curtain for reading the vertical projection. The contour sensor 35 is connected to an image processing unit where the contour, i.e. a two dimensional projection of a three dimensional object, is checked against the properties in the database.

The symbol reading sensor 36 comprises a camera for providing still or moving images. The symbol reading sensor 36 is connected to an image processing unit where the symbol is checked against the properties in the database.

The color texture sensor 38 comprises a camera for providing still and moving image. The color texture sensor 38 is connected to an image processing unit where the color texture is checked against the properties in the database. The image processing unit comprises an algorithm which calculates where a certain color is present in the image. One common algorithm is“Weibull color texture algorithm”, but other algorithms may also be considered.

The color histogram sensor 37 comprises a camera for providing still and moving pictures. The color ratio in the image is usually illustrated by means of a representation, a so-called histogram. A histogram is generated by examination of all pixels of the image, and the number of pixels having a specific color value are summarized.

The above mentioned image processing units may consist of one or several units and may comprise one or several computers with software capable of performing the above mentioned analyses. The classification device may comprise one or several cameras that are included in the above mentioned sensors.

One example of an embodiment is that the contour sensor 35 comprises a first camera positioned in a way that the contour is read when the article passes the camera.

A linear camera is suitable since the reading then occurs during the conveying of the article between two conveyor belts or over a translucent surface. It is also suitable that the classification device comprises a second camera and possibly several cameras to be able to see the article from different angles for achieving the highest possible reliability when detecting text and images. The other camera, and if applicable a further camera/cameras, is arranged to record an image or images which will be used by the image processing unit for analyses of color histogram, color texture and OCR. One further alternative is that the classification device comprises only the first camera and the second camera where the second camera is optically connected to one or several lenses that observe the article from different angles and where the image processing unit analyses the images from corresponding angles.

The NIR sensor 33, 34 operates in such way that infrared light illuminates the article and the reflecting infrared light from the article is being analyzed with reference to phase displacement caused by surface ratio/surface properties and chemical bonds at the article which creates a reflection spectrum. NIR sensors 33, 34 are known per se by prior art.

As mentioned above NIR is a shortening of the English term“Near InfraRed Spectroscopy” and comprises a light source for near infrared light and a NIR camera that may register near infrared light. Near infrared light typically has a wavelength of 580-2500 nm, or preferably 780-1750 nm. The wavelength has shown to be suitable for analyzing bulk material, fruit and vegetables. In this context“NIR” may include the light source and the NIR camera, i.e. the complete NIR arrangement for analyzes. However,“NIR sensor” may only include the sensing equipment, e.g. the light guiding probe and the spectrometer.

By analyzing a known article with a NIR sensor 33, 34 a unique reflection spectrum is received which may be connected to the article. The reflection spectrum may either be used directly as a signature connected to the article or the reflection spectrum is processed to create the signature. An article in a store may look different at different occasions, e.g. an article may grow old (eventually fruit will rotten) and the article may be packed in one or several plastic bags or the article may be solitary or in a group, or be arranged in different orientations; natural variations of the article occurs also, etc. The environment for a checkout counter 100 may also be different in different stores, e.g. different amount of light, color, etc. All these parameters provide that a NIR spectrum of a certain article in a certain environment at a certain occasion does not necessarily match with another NIR spectrum of the article in another environment at another occasion. To be able to use a NIR sensor 33, 34, the first signature has to match the second signature at a certain degree of accuracy such that the controller 20 is able to identify the article by a comparison. It is thus an advantage if the first signature is created in the same environment as the second signature. Since the second signature is created at the checkout counter 100 during use, it is an advantage if the first signature is created during the same conditions.

At the creation of the first and second signature the surroundings will be considered by means of a background spectrum, i.e. an empty checkout counter 100, or an empty conveyor belt. When analyzing an article the background spectrum is known and the controller 20 may consider it in different ways.

The VIS sensor 33, 34 is a spectrometer comprising a light device suitable for the mentioned wavelengths and a VIS camera capable of registering light of the wavelengths between 200 nm and 1100 nm. Similar to the NIR sensor 33, 34 the VIS sensor 33, 34 uses the change in wavelength when light is partly absorbed by or reflected by an article. The VIS sensor 33, 34 is particularly suitable for analyzing different shades of brown, which makes it suitable for analyzing bread which is normally hard to classify by means of any of the other sensors. The different shades of brown are detectable by the VIS sensor 33, 34. In this context“VIS sensor” may include the light source as well as the VIS camera, i.e. the complete VIS device for analyzing. However, the VIS sensor 33, 34 may also be a separate device, not connected to the light source, but including a light guiding probe and a spectrometer.

By analyzing a known article by means of a VIS sensor 33, 34 a unique reflection spectrum, VIS spectrum, is received, which may be coupled to the article. The reflection spectrum may either be used directly as a signature for the article, or the reflection spectrum may be processed for creating the signature.

The embodiment with several sensors such as described above, arranged to generate possible article identities to be shown in a list or a plurality of lists are thus designed on a number of combinations comprising a partial set of sensors, where it will be sufficient that one of the combinations provides a possible article identity. The combinations may be predetermined or arbitrarily selected. Sensors may be switch on, i.e. be activated, in sequences in order to find beneficial combinations or a partial set of sensors or all sensors may be active until one of the combinations provide a possible article identity.

The article identity may be determined by means of checking a database comprising properties of a number of articles. Example of properties may include weight, size, color, shape, contour and/or marking.

The sensors may preferably be placed completely or partly in a tunnel shaped construction which shields a part of the conveyor belt and therefore improves the security by preventing unauthorized people from the possibility to affect the

classification process.

System learning

The classification system for identification of articles may be subject to training, or learning, in order to improve the accuracy of the generation of possible article identities.

For example, the at least one sensor arranged to generate possible article identities, may be activated during predetermined training sessions, in which a store attendant picks articles in a consecutive order. For each picked article type, the attendant manually scans a barcode on the article or inputs the PLU-code for at least one article for a secure identification.

Additionally, or alternatively, the classification system 30 may be trained, or learned, during normal use of the system. The classification system 30 will always receive the correct article identity, due to the manual input from the checkout operator, and the system can thus use this information to better identify the article the next time by creating at least one list of possible article identities that have a higher probability of being the unidentified article. The accuracy and/or number of articles shown in the list(s) that is shown to the checkout operator will thus be improved over time.

In yet one embodiment the algorithm generating the list of article identities in the classification system 30 is continuously learned by using the raw data from the at least one sensor. In this situation, the algorithm preferably uses deep learning methods, such as Convolutional neural network or Fully Connected neural network.

In the example where the at least one sensor is a NIR-sensor, a self-learning functionality could be seen as follows. The classification system has thus a self-learning functionality in which the first signature is created by programming the memory unit 22 with an article identity whereafter the article is transported through the checkout counter 100 during circumstances similar to use, i.e. circumstances for the checkout counter 100 which refers to customer use. To consider the mentioned variations the article is transported several times through the checkout counter 100 and in different variations, e.g. with one or several bags and/or solitary or in a group, etc. Each time the article is transported through the checkout counter 100 and a NIR sensor 33, 34 is analyzing the article a first signature is created, which means that each article identity may be connected to a large amount of first signatures such that the controller 20 will be able to identify the article when comparing it to the second signature and one or more of the first signatures. During learning the first NIR sensor 33 may be arranged to perform the analyses, or a second NIR sensor 34 will be connected. The learning does not need to be performed at the exact location where the checkout counter 100 will be used but may rather be performed at another location. In addition to or in alternative to deep learning, the algorithm may be trained using classifications using k-nearest neighbor’s algorithm (K-NN) and/or statistical networks such as Bayesian Belief Network (BBN).

Payment

In one embodiment, once all the articles of the customer are correctly identified, the controller 20 transmits a transaction signal to a POS-system 50. The POS-system 50 can retrieve the registration account to allow the customer to finish the payment transaction. The payment transaction may be initiated manually by the checkout operator 5, for example by pressing a dedicated button on the display 40; 55. Additionally, or alternatively, the payment may be initiated automatically by the controller 40 of the classification system 30, for example by the classification system 30 recognizing a customer divider bar 7 signaling that all articles for that customer has been scanned.

As previously mentioned, in some embodiments the POS-system 50 could be a part of the classification system 30. In other embodiments the classification system 30 may be part of the POS-system 50. The POS-system 50 and the classification system 30 could also be two separate systems being in operative communication with each other.

Checkout system

One of the several benefits of having a classification system for identification of articles that is configured to display at least one list of possible article identities for the article that the barcode reading system 32 failed to identify, will now be further described with reference to Figs. 5a-c.

Figs. 5a-c illustrate embodiments where two checkout counters 100, 200 are arranged in conjunction with each other so as to form a checkout system 300. Since the identification process is very easy and fast for the checkout operator 5, thanks to the novel classification system 30, the checkout operator 5 will have time to handle a further checkout counter simultaneously. This reduces the number of checkout operators 5 needed (thus reducing the costs), while still keeping the personal touch and service to the customer. The checkout systems 300 comprises at least one classification system 30 for identification of articles arranged on the two checkout counters 100, 200. In the embodiment in Fig. 5a, a single classification system is used to identify the articles arranged on the two checkout counters 100, 200. Hence, only one controller 20 and one display 40 is used to generate and display the at least one list of possible article identities. The single display 40 may have a split screen showing two separate interfaces, one for each checkout counter 100, 200.

In Fig. 5b, the checkout system 300 comprises two separate displays 40’, 40”, one for each checkout counter 100, 200. The displays 40’, 40” could be arranged adjacent to each other to facilitate for the checkout operator 5. This embodiment could be seen as having one classification system 30 and an additional display 40” or as two classifications systems having a shared controller 20.

In the embodiment illustrated by Fig. 5c, the checkout system 300 comprises two classification systems. The first classification system comprises a first controller 20’ and a first display 40’ configured to classify articles being arranged on the first checkout counter 100. The second classification system comprises a second controller 20” and a second display 40” configured to classify articles being arranged on the second checkout counter 200. The displays 40’, 40” could be arranged adjacent to each other to facilitate for the checkout operator 5. The controllers 20’, 20” could be arranged in operatively communication with each other.

In the embodiments shown in Figs. 5d-e, the relationship between the checkout system 300 and a POS-system 50 is shown. The controller is in operative

communication with at least one display 40; 55, either by having a display 40 being part of the classification system or by having a display 55 being part of a POS-system 50.

In Fig. 5d, the POS-system 50 is in communication with the controller 20 and the weight sensors 31. The POS-system performs the certified weighing of the articles that have a price being dependent on weight, and calculates the price of the identified article based on the article identity (received from the controller 20) and the weight information. In this embodiment the POS-system 50 displays the results on one or several displays 55’, 55” being a part of the POS-system. Hence, in this embodiment the classification system does not itself comprise display(s) but uses the display(s) of the POS-system.

It should be understood that this configuration with the POS-system could be implemented in the situations as shown in Figs. 5a and 5b as well.

In Fig. 5e, the POS-system 50 is only in communication with the controller 20. Hence, the weight sensors 31 will in communication with the controller 20 perform the certified weighing when needed. For articles that have a price being dependent on its weight, the information transmitted to the POS-system thus comprises information relating to both weight, price and the article identity. In this embodiment, the classification system does not itself comprise a display(s) but uses the display(s) of the POS-system.

It should be understood that this configuration with the POS-system could be implemented in the situations as shown in Figs. 5a and 5b as well.

In Fig. 5f, the POS-system 50 is in communication with the controller 20 and either direct or indirect in communication with the one or more displays 40’, 40”. The weight sensors 31 will in communication with the controller 20 perform the certified weighing when needed. For articles that have a price being dependent on its weight, the information transmitted to the POS-system thus comprises information relating to both weight, price and the article identity. Moreover, the classification system comprises one or several displays 40’, 40” that are used to shown the article identity and the information retrieved from the POS-system. Hence, in this situation the displays are a part of the classification system and not the POS-system. The POS-system may transmit the information either directly to the displays or by transmitting the information to the controller 20.

It should be understood that this configuration with the POS-system could be implemented in the situations as shown in Figs. 5a and 5b as well.

In the embodiments of Figs. 5a-f, the conveyor systems 100, 200 are designed so that their respective packing area are in parallel to each other. However, it should be understood that their respective conveyor systems 100, 200 could be completely separated.