JP2011118653A - Injustice degree recording device, method for controlling injustice degree recording device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an injustice degree recording device capable of easily detecting an injustice concerning a means accompanied by an accomplice, a method of controlling the injustice degree recording device and a program. <P>SOLUTION: This injustice degree recording device is configured of: an injustice occurrence pattern storage part 71 for storing an injustice occurrence pattern in which basic scores as the weighting of each of one or more injustice occurrence elements assumed in a store are described; a customer detection part 77 for detecting a customer in case of accounting processing by an accounting processing part 51; an injustice occurrence detection part 72 for detecting the occurrence/non-occurrence of one or more injustice occurrence elements in case of accounting processing; an injustice degree calculation part 73 for calculating the degree of injustice of a customer by totaling the basic scores corresponding to the injustice occurrence elements whose occurrence has been detected by the injustice occurrence detection part 72; and an injustice degree recording part 74 for recording the calculation result of the injustice degree calculation part 73 in a database 23 for recording as well as the identification information of the customer detected by the customer detection part 77. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a fraud recording device, a fraud recording device control method, and a program for detecting fraudulent acts by store clerk and customers in restaurants such as supermarkets and family restaurants.

  Conventionally, Patent Document 1 has been proposed as a system for detecting fraud at the time of accounting processing by a store clerk in a retail store or a restaurant. The system described in Patent Document 1 includes a POS (Point Of Sales) terminal (POS terminal), a camera that captures an image around the POS terminal, a photo journal recorder, and a search terminal. The photo journal recorder records journal data (transaction history data), image data around the POS terminal captured by the camera, and operation history (operation history data) of the POS terminal by the store clerk in association with each other. . Then, the administrator designates predetermined search conditions such as time and POS terminal operation information using the search terminal, thereby displaying journal data and image data that match the search conditions on the display.

JP 2005-115504 A

  However, the system described in Patent Document 1 only displays a list of recorded data (accounting processing contents and captured images at that time) that match the search conditions specified by the administrator. In order to detect whether or not it has been received, the administrator must check the recorded data step by step. In addition, since it is necessary to check a large amount of recorded data, there is a problem in that an illegal act is not detected. As described above, the conventionally proposed system has a problem in that it cannot reliably detect an illegal act. Further, in recent years, the methods of fraudulent acts have become diversified and complicated, and it is impossible to accurately determine whether or not the fraudulent acts have been performed only with the captured images obtained by photographing the periphery of the POS terminal. As an example, the accomplice may pretend to be a good customer, and another person's receipt may be handed over to the accomplice clerk (operator) to perform a cancellation process (registration minus process). In such a case, it is difficult to detect fraud from a temporary captured image. In addition, since accomplices who perform such techniques are likely to repeat the crime, it is possible to identify the accomplices by checking the recorded data over a long period of time. Takes time and effort. As described above, it has been very difficult to detect fraudulent behaviors with accomplices.

  In view of the above problems, an object of the present invention is to provide a fraud recording apparatus, a fraud recording control method, and a program capable of easily detecting fraudulent actions involving accomplices.

  The fraud level recording apparatus of the present invention includes a customer detection unit that detects a customer at the time of accounting processing by the accounting processing unit, a status data detection unit that detects status data in a store at the time of accounting processing, and detection of the status data detection unit Based on the result, a fraud level calculation unit that calculates the fraud level of the customer, and a fraud level recording unit that records the calculation result of the fraud level calculation unit in the database together with the customer identification information detected by the customer detection unit, It is characterized by having.

  The fraud level recording apparatus control method of the present invention is a fraud level recording apparatus control method provided with an accounting processing unit and a database inside or outside, and detects a customer during accounting processing by the accounting processing unit. Steps, a status data detection step for detecting status data in the store at the time of accounting processing, a fraud level calculation step for calculating customer fraud based on the detected status data, and a calculation result by the fraud level calculation step And a fraud level recording step of recording in the database together with the customer identification information detected by the customer detection step.

  According to these configurations, based on the in-store situation data at the time of accounting processing, the customer's fraud level (high possibility of fraudulent conduct as an accomplice) is calculated, and the calculated result is Since the information is recorded together with the identification information, the user (such as an administrator) can grasp the possibility that the customer is an accomplice from the record. As a result, it is possible to easily detect fraudulent actions even for complicated techniques involving accomplices.

  The fraud level recording device described above further includes a fraud occurrence pattern storage unit that stores fraud occurrence patterns describing a basic score as a weighting for each of one or more fraud occurrence elements assumed in the store, The data detection unit detects occurrence / non-occurrence of one or more fraud occurrence elements, and the fraud level calculation part sums the basic points corresponding to the fraud occurrence elements detected by the situation data detection part. It is preferable to calculate the degree of fraud.

According to this configuration, the degree of fraud can be easily calculated by summing the basic points corresponding to one or more fraud elements in which occurrence has been detected.
Note that fraudulent elements include predetermined key operations by the accounting processing unit, inconsistencies in receipts during exception processing and checkout processing, inconsistencies in target customers during exception processing and checkout processing, and predetermined voice keywords during exception processing. Non-detection, change in cash drawer open / close state, intrusion of the store clerk into the designated area, execution of accounting processing in the situation where there is no person other than the store clerk operating the accounting processing section around the checkout counter, suspicious behavior of the store clerk, It includes at least one element among changes in the biological signal.

  In the fraud level recording apparatus described above, the accounting processing unit further includes an exception processing detection unit that detects that an exception process corresponding to a part of the accounting process has occurred, and the fraud level calculation unit includes the exception processing unit. It is preferable to function on the condition that exception processing is detected by the detection unit.

  In general, fraudulent acts are often executed along with exception processing. According to this configuration, since the fraud level is calculated only when exception processing occurs, fraud can be efficiently detected.

  In the fraud recording apparatus described above, exception processing includes “cancellation processing” for canceling input data during accounting processing of the accounting processing unit, “cancellation processing” for canceling the accounting processing after accounting processing, "Return processing" to return the product, "Discount processing" to discount the product, "Discount processing" to discount the product, and "Exchange processing" to exchange money, At least of “inspection process” for reconciling cash in cash drawer and sales, “withdrawal / refund process” for withdrawal or refund, and “reissue process” for reissuing receipt It preferably includes one treatment.

  According to this configuration, “cancellation process”, “cancellation process”, “return process”, “discount process”, “discount process”, “exchange process”, “inspection process”, “withdraw / refund process”, It is possible to detect fraudulent acts associated with “reissue processing”.

  In the fraud level recording apparatus described above, the exception processing detection unit detects the type of exception processing, and the fraud occurrence pattern storage unit stores fraud occurrence patterns for each type of exception processing, and calculates the fraud level. The unit preferably extracts the fraud occurrence pattern corresponding to the type of exception processing detected by the exception processing detection unit from the fraud occurrence pattern storage unit, and calculates the fraud level.

  According to this configuration, since the fraud occurrence pattern is stored for each type of exception processing, a more accurate fraud can be calculated using the fraud occurrence pattern suitable for the type of exception processing.

  In the fraud recording device described above, if it is determined from the number of records recorded in the database that the number of exception processing executions for the same customer exceeds a predetermined number, it is described in the fraud occurrence pattern. A fraud occurrence pattern rewriting unit that increases the basic score of at least a part of one or more fraud occurrence elements that have been set, and the fraud degree calculation unit calculates the fraud degree based on the fraud occurrence pattern after rewriting. It is preferable to do.

  In general, customers who repeat exception handling are likely to be accomplices. According to this configuration, for a customer who repeats exception processing, the basic score of fraud elements is increased, resulting in an increase in the degree of fraud calculated as a result. Therefore, it is possible to more reliably detect a malicious fraud that repeats exception processing.

  In the fraud level recording apparatus described above, it is preferable that the fraud level calculation unit further includes a warning unit that issues a warning when a fraud level greater than a predetermined value is calculated.

  According to this configuration, since a warning is issued at the time of accounting processing in which the target customer is likely to be an accomplice, it is possible to notify the store clerk and manager (manager, store manager, etc.) to that effect.

  In the fraud level recording apparatus described above, it is preferable that the fraud level recording apparatus further includes a warning unit that issues a warning when the cumulative total of the fraud levels of the same customer recorded in the database exceeds a predetermined value.

According to this configuration, even if a customer has not been warned in the past, if the cumulative fraud rate exceeds a predetermined value, it may be an accomplice, so a clerk or administrator (manager or manager) Such as the store manager). Thereby, the vicious accomplice who has escaped fraud detection with the advanced fraud technique can be detected more reliably.
The warning unit preferably issues a warning even when the past fraud level is not recorded in the database and the predetermined fraud level calculation result exceeds a predetermined value.

  In the fraud recording apparatus described above, the warning unit preliminarily outputs a warning document including the detection result of the customer detection unit and the status data detected by the status data detection unit using an email or a Web application. It is preferable to notify a designated administrator.

  According to this configuration, the administrator who has received the notification is based on the detection result (imaging data, etc.) of the customer detection unit and the situation data (occurrence / non-occurrence of each fraud element) detected by the situation data detection unit. , Can understand the situation with customers and accounting. In addition, since an e-mail or Web application is used, a warning document can be reliably notified regardless of the location of the administrator.

  In the fraud recording apparatus described above, the customer identification information is preferably customer biometric information.

  According to this configuration, the customer can be reliably identified by using the biological information. As biometric information, it is preferable to use facial features, voice (voice print), fingerprint, palm shape, retina, iris, blood vessel (vein pattern), ear shape, and the like. As described above, by using the physical feature (static information), it is not necessary to monitor the behavior, and an individual can be easily specified.

  In the fraud recording apparatus described above, the database is preferably shared by a plurality of fraud recording apparatuses.

  According to this configuration, since the database can be shared among a plurality of fraud recording devices and a plurality of stores, even the accomplices who repeat crimes at a plurality of stores can detect the fraud. Can do.

  A program according to the present invention causes a computer to execute each step in the control method for the fraud recording apparatus described above.

  By executing this program, it is possible to realize a control method for a fraud recording apparatus that can easily detect fraudulent behaviors associated with accomplices.

1 is a system configuration diagram of a store monitoring system according to a first embodiment. It is a functional block diagram of the store monitoring system concerning a 1st embodiment. It is a figure which shows the list of fraud patterns. It is a figure which shows the recording content of the database for recording. It is a figure which shows the structure of a fraud element. It is a figure which shows the specific example of a fraud element. It is a figure which shows the example of generation | occurrence | production order definition of fraud pattern A1. It is a figure which shows an example of the detection result with respect to fraud pattern A1. It is a flowchart which shows the injustice degree recording process of a shop monitoring system. It is a functional block diagram of the store monitoring system concerning a 2nd embodiment. It is a figure which shows an example of the log | history of fraud. It is a figure which shows the example of generation | occurrence | production order definition of fraud pattern A2. It is a figure which shows an example of the detection result with respect to fraud occurrence pattern A2. It is a flowchart which shows a part of fraud recording process of the shop monitoring system which concerns on 2nd Embodiment.

  Hereinafter, a fraud level recording apparatus, a fraud level recording apparatus control method, and a program according to the present invention will be described with reference to the accompanying drawings. In each embodiment shown below, the case where the fraud recording device of the present invention is applied to a store monitoring system is illustrated. The store monitoring system is constructed to detect fraudulent acts by store clerk and customers in a retail store such as a supermarket.

[First Embodiment]
FIG. 1 is a system configuration diagram of a store monitoring system SY according to the first embodiment. As shown in the figure, the store monitoring system SY includes a POS terminal 1 that performs an accounting process, a receipt printer 2 that issues a receipt (not shown), a handy scanner 3 that reads a barcode attached to or printed on a product, A card reader 4 that performs non-contact reading of the clerk card C, a clerk wristband 5 that is worn on the wrist of the clerk (operator), a customer display 6 that displays product information, and a head of the clerk. A headset 7, a receipt scanner 8 that reads receipts, a cash register monitoring camera 9 that monitors the vicinity of a cash register counter, a shop monitoring camera 11 that monitors the inside of the shop, and a shop management server 12 that supervises the shop monitoring system SY. . The POS terminal 1, the receipt printer 2 and the handy scanner 3, the receipt printer 2 and the card reader 4, the clerk wristband 5, the customer display 6, the headset 7 and the receipt scanner 8 are each connected to a serial interface (for example, USB (Universal Serial Bus) ) Interface) or wireless LAN. The POS terminal 1, the receipt printer 2, the cash register monitoring camera 9, the in-store monitoring camera 11, and the store management server 12 are connected by an in-store LAN 13 (wired LAN or wireless LAN).

  On the other hand, the store management server 12 is connected to the mobile phone company 31 via a network NT such as the Internet. The illustrated manager mobile phone 33 is possessed by a manager (manager, store manager, etc.) having a predetermined authority in the store, and the store clerk wearing the headset 7 via the telephone line 32 and the mobile phone company 31. It is possible to call. The manager's mobile phone 33 has an electronic mail function, and can receive an electronic mail (warning document) transmitted from the store management server 12.

  Although FIG. 1 shows a case where there is one POS terminal 1, when a plurality of cashier counters 10 are installed in the store, the POS terminal 1 and the like are provided for each cashier counter 10. The “degree of fraud recording device” in the claims includes the receipt printer 2 and the store management server 12 among the above-described components.

  The POS terminal 1 is installed in the cashier counter 10 and performs accounting processing, a data hook driver 42 that outputs an operation result of the POS terminal 1 to the receipt printer 2, a cash drawer 43 that stores cash, have. In addition, although not specifically shown, it has a general register configuration. The handy scanner 3 reads a product barcode on which a product code or the like is recorded, and outputs the read result to the POS terminal 1.

  The card reader 4 reads a clerk ID from a clerk card C in which a wireless IC (RFID: Radio Frequency Identification or the like) is incorporated. The store clerk in the store is required to possess the store clerk card C. The store clerk wristband 5 incorporates a sweat sensor 45 that measures the amount of sweat of the wearer and a pulse sensor 46 that measures the pulse rate of the wearer. A store clerk operating the POS terminal 1 in the cashier counter 10 is required to wear the store clerk wristband 5.

  The customer display 6 has a function of acquiring a customer's voice by a customer microphone 6a in addition to a general function as a customer display for displaying a product name and a product price to a customer. The customer microphone 6 a is a small sound collecting microphone, and is provided on the peripheral portion of the customer display 6 toward the customer side. The headset 7 is also called an income, and includes a clerk microphone 7a and headphones (not shown). A store clerk operating the POS terminal 1 is required to wear the headset 7. The receipt scanner 8 reads a receipt for the purpose of determining receipt authenticity.

  The receipt printer 2 has a function of measuring the temperature in the register counter 10 by a thermometer 44 in addition to a function of issuing a receipt based on the accounting information output from the POS terminal 1. Further, the receipt printer 2 inputs various information from serially connected peripheral devices and outputs them to the store management server 12. In this way, by inputting / outputting various information with the receipt printer 2 as the main (the receipt printer 2 filters the acquired various information and outputs necessary information), the POS backbone network (POS terminal 1) Network) is not affected. Further, when applying the present invention to an existing POS system, it is not necessary to change the POS backbone network itself.

  The cashier monitoring camera 9 is provided on the ceiling of the cashier counter 10 and the like, inside the cashier counter 10 (such as around the store clerk and the POS terminal 1), on the cashier counter 10 (products subject to accounting processing, cash exchange, etc.) An image is taken in front of the cashier counter 10 (such as a customer). The in-store monitoring cameras 11 are provided at a plurality of locations in the store and monitor the entrance / exit of the store clerk to / from a predetermined area or a predetermined room. As the cash register monitoring camera 9 and the store monitoring camera 11, a CCD camera or a PTZ (Pan Tilt Zoom) camera can be used.

  The store management server 12 includes a facial feature database 21 that stores the facial features of the store clerk, a keyword database 22 that stores the keywords of the store clerk and customers spoken during the accounting process, And a recording database 23 for recording the degree of injustice. The store management server 12 calculates the degree of fraud of the store clerk and the customer based on the situation data in the store detected by each of the above devices and the reference result of the face feature amount database 21 and the keyword database 22. The degree of injustice is recorded in the recording database 23. Further, when the calculated degree of fraud exceeds a predetermined value, a warning is given to an operator or an administrator.

  Next, the functional configuration of the store monitoring system SY will be described with reference to FIG. The POS terminal 1 has an accounting processing unit 51 as a main functional configuration. The accounting processing unit 51 has the POS application 41 as a main part, and executes accounting processing including settlement processing and exception processing. The processing results are output to the receipt printer 2 as transaction information.

  In the present embodiment, exception processing refers to predetermined processing that is likely to be related to fraud. Specifically, “return processing” for returning a product, “misregistration processing (cancellation processing)” for canceling the accounting processing after accounting processing, “discount processing” for discounting the product, “Discount process” for discounting the product, “Cancel process” for canceling the input data during the accounting process, “Exchange process” for exchanging money, and cash and sales in the cash drawer 43 It refers to “inspection processing” for collation, “withdrawal / refund processing” for withdrawal or refund, and “reissue processing” for reissuing a receipt. Each of these processes is distinguished as an exception process type.

  The card reader 4 has a clerk ID reading unit 52 as a main functional configuration. The store clerk ID reading unit 52 performs short-range wireless communication with the store clerk card C, and reads the store clerk ID from the store clerk card C existing around the cashier counter 10. The store clerk wristband 5 includes a sweating amount measurement unit 53 and a pulse rate measurement unit 54 as main functional configurations. The sweating amount measuring unit 53 is realized by the sweating sensor 45, and the pulse rate measuring unit 54 is realized by the pulse sensor 46.

  The customer microphone 6a provided on the customer display 6 has a customer voice acquisition unit 55 as a main functional configuration. The customer voice acquisition unit 55 acquires a customer voice signal and outputs it to the receipt printer 2. The clerk microphone 7a provided in the headset 7 has a clerk voice acquisition unit 56 as a main functional configuration. The clerk voice acquisition unit 56 acquires the clerk's voice signal and outputs it to the receipt printer 2.

  The receipt scanner 8 has a receipt reading unit 57 as a main functional configuration. The receipt reading unit 57 reads the receipt and obtains product information (transaction number, receipt issue date, product name, quantity, product unit price, total price, etc.) printed on the receipt. The cashier monitoring camera 9 includes a cashier imaging unit 58 as a main functional configuration. The cash register imaging unit 58 captures an image around the cash register counter 10 and outputs the imaging result to the store management server 12. The in-store monitoring camera 11 has an in-store imaging unit 59 as a main functional configuration. The in-store imaging unit 59 captures an in-store image including the behavior (movement) of the clerk or customer in the store, and outputs the captured result to the store management server 12.

  The receipt printer 2 includes a transaction information acquisition unit 61, an exception processing detection unit 62, a clerk ID detection unit 63, a biological signal detection unit 64, a voice recognition unit 65, a reading result acquisition unit 66, a printing unit 67, and a main functional configuration. An operator warning unit 68 is provided. The transaction information acquisition unit 61 acquires the transaction information output from the POS terminal 1 (accounting processing unit 51).

  Based on the transaction information acquired by the transaction information acquisition unit 61 or the detection result of the data hook driver 42, the exception processing detection unit 62 detects that the exception processing has occurred in the transaction processing unit 51 and its type. The fraud level calculation unit 73, the fraud level recording unit 74, the operator warning unit 68, and the administrator warning unit 75, which will be described later, function on the condition that exception processing is detected by the exception processing detection unit 62.

  The clerk ID detection unit 63 detects the clerk ID from the reading result of the clerk ID reading unit 52, and outputs the detection result to the fraud occurrence detection unit 72 (described later). The number of salesclerks present around the cashier counter 10 is specified by the detected number of salesclerk IDs, and the salesclerk operating the POS terminal 1 can be specified by the detected salesclerk ID. Note that the clerk ID detection unit 63 may be realized by a plurality of card readers 4 having directivity. According to this configuration, it is possible to clearly distinguish and detect the clerk performing the accounting operation and the other clerk.

  The biological signal detection unit 64 detects a biological signal that is a measurement result of the sweating amount measurement unit 53 and the pulse rate measurement unit 54, and outputs the detection result to the fraud occurrence detection unit 72. The voice recognition unit 65 acquires voice data from the customer voice acquisition unit 55 and the store clerk voice acquisition unit 56 and performs voice recognition. The voice recognition unit 65 has two dedicated recognition units for independently recognizing customer voices and store clerk voices. Each dedicated recognition unit includes an acoustic analysis unit, an acoustic model, and a language. A model, a word dictionary, and a text conversion unit are included (all not shown). Each dedicated recognition unit uses these to convert the acquired speech signal into text information. The text information is output to the fraud occurrence detection unit 72.

  The reading result acquisition unit 66 detects product information that is the reading result of the receipt reading unit 57 and outputs the detection result to the fraud occurrence detection unit 72. The printing unit 67 prints the transaction information acquired by the transaction information acquisition unit 61 on receipt paper and issues a receipt. The operator warning unit 68 notifies the operator that there is a high possibility that an illegal act has been performed when a fraud level exceeding a predetermined value is calculated by the fraud level calculation unit 73 (described later). As a notification method, display on an operator display (not shown), sound generation, LED lighting / flashing, vibration by a vibrator, and the like are conceivable.

  Store management server 12 has fraud occurrence pattern storage part 71, fraud occurrence detection part 72, fraud degree calculation part 73, fraud degree recording part 74, and manager warning part 75 as main functional composition. The fraud occurrence pattern storage unit 71 stores fraud occurrence patterns in which one or more fraud occurrence elements assumed in the store are defined. Here, the “fraud element” refers to a predetermined key operation by the accounting processing unit 51, a mismatch of receipt reading results (product information) at the time of exception processing and payment processing, a target customer at the time of exception processing and payment processing Discrepancy, non-detection of predetermined voice keyword during exception processing, change in cash drawer 43 open / close state, intrusion of store clerk into a predetermined area (changing room, etc.), no other store clerk or customer around register counter 10 Refers to the execution of accounting processing in the situation, suspicious behavior of store clerk (looking around), change of store clerk's biological signal, etc.

  As shown in FIG. 3, fraud patterns are classified and stored for each exception processing type. In the list of fraud occurrence patterns shown in FIG. 3, “fraud technique” indicates a technique of fraudulent behavior assumed when the corresponding exception processing type is executed. In the figure, an example in which “exception processing type” and “illegal trick” have a 1: 1 relationship is shown, but an N: 1 or 1: N relationship may be used. In addition, as shown in “illegal utterance pattern”, a plurality of fraud patterns are associated with each exception processing type.

  The fraud occurrence detection unit 72 detects occurrence / non-occurrence of fraud occurrence elements defined in the fraud occurrence pattern. In addition, for the fraud element that has detected the occurrence, the precision is detected as necessary. The matching rate refers to an intermediate value (a value greater than 0% and less than 100%) where “occurrence” is 100% and non-occurrence is 0%. The fraud level calculation unit 73 extracts a fraud occurrence pattern corresponding to the type of exception processing detected by the exception processing detection unit 62 from the fraud occurrence pattern storage unit 71, refers to the fraud occurrence pattern, and determines the fraud level. calculate. Specifically, out of the fraud occurrence elements described in the fraud occurrence pattern, the basic score (= abnormal value) corresponding to the fraud occurrence elements detected by the fraud occurrence detection unit 72 is summed up to calculate the fraud level. Is calculated. Note that for fraud occurrence elements whose relevance rate is detected by the fraud occurrence detection unit 72, the corresponding base score is multiplied by the relevance rate, and the multiplication result is used as an abnormal value. Details will be described later. The fraud level recording unit 74 records the calculation result of the fraud level calculation unit 73 in the recording database 23.

  As shown in FIG. 4, various information is recorded in the recording database 23 for each accounting process (accounting process number). Specifically, fraud occurrence date and time, transaction data, target cash register ID (POS terminal 1 ID), cashier person ID (clerk ID), exception processing type, detection result of each fraud element (situation data), customer Information and fraud levels are recorded. The customer information includes a customer ID specified by using a membership card, a credit card, electronic money, and the like, a facial feature amount obtained from an imaging result of the cashier monitoring camera 9, and a voice acquisition result of the customer microphone 6a. Refers to biological information such as voiceprints obtained from Further, the type of exception processing indicates the detection result of the exception processing detection unit 62, and the degree of fraud indicates the calculation result of the fraud level calculation unit 73.

  The administrator warning unit 75 issues a warning to the administrator when the fraud level calculation unit 73 calculates a fraud level exceeding a predetermined value. In this embodiment, a warning document is transmitted to the manager's mobile phone 33 by e-mail. The e-mail includes items recorded in the recording database 23 described above. Note that it is preferable that the manager's mobile phone 33 (manager) can talk to the headset 7 (clerk) after receiving the e-mail. According to this configuration, the administrator can directly check the status of the accounting process with the operator, and can take appropriate measures against fraudulent acts. Further, when a PTZ camera is employed as the cashier monitoring camera 9, it is preferable that the shooting direction and zoom of the cashier monitoring camera 9 can be controlled based on the operation of the manager's mobile phone 33. According to this configuration, the situation around the cashier counter 10 can be confirmed without bothering the operator. Further, as a warning method for the manager, a mobile terminal connected to a Web application or a store LAN may be used instead of an e-mail. As described above, by using an electronic mail or a Web application, it is possible to reliably notify the warning content regardless of the location of the administrator. Moreover, although there is a limit of whereabouts by using a portable terminal, a communication charge can be reduced.

  Next, the structure of the fraud element will be described with reference to FIG. As described above, the fraud occurrence element is defined in the fraud occurrence pattern, and the user uses the operation unit (not shown) provided in the receipt printer 2 or the store management server 12 to describe the details. Can be defined. As shown in the figure, the fraud elements include information on “classification number”, “element number”, “detection device”, “determination content”, and “weighting”. The “classification number” is group identification information when the fraud elements are classified into a plurality of groups, and specific examples include a to e in the figure. The user searches the fraud occurrence element to be incorporated into the fraud occurrence pattern using the classification number as an index. The “element number” is identification information uniquely assigned to each fraud element. The “detection device” is a device for detecting the fraud element, and specifically includes m1 to m8 shown in the figure. As will be described later with reference to FIG. 6, “determination content” refers to “determination item” and “determination condition”, and “weighting” refers to “basic score” assigned to the fraud element and conformity “Coefficient” for calculating the rate.

  A plurality of “detection devices” can be associated with fraud elements that can be detected by a plurality of detection devices. For example, if the fraudulent element is a content that limits the number of salesclerks around the cashier counter 10, the number of salesclerks can be detected by the card reader 4 (m1) and the cashier monitoring camera 9 (m6). Two detection devices may be associated with each other. Also, the “determination content” and “weighting” can be arbitrarily set by the user.

  Here, a specific example of the fraud element will be described with reference to FIG. For example, the fraudulent element of “element number: E1” is detected by the data hook driver 42 (see FIG. 1) (m8) incorporated in the POS terminal 1, and the POS is detected by the operation information output from the data hook driver 42. If the pressing of the cancel key in the terminal 1 can be detected, “occurrence” is detected by the fraud occurrence detection unit 72 as “POS exception key pressed”. In this case, the store clerk ID of the operator is detected using the data hook driver 42 or the card reader 4, and the store clerk ID is recorded in the recording database 23 as a detection result. Note that “the POS exception key has been pressed” does not use the data hook driver 42, but the accounting information output from the POS terminal 1, the print data for receipt printing, and the display for display on the customer display 6. You may detect based on data.

  The fraudulent occurrence element of “element number: E2” is detected by the cashier surveillance camera 9 (m6), and when the similarity is determined to be 60% or less by the same person determination, “buyer and exception applicant” “Occurrence” is detected by the fraud occurrence detection unit 72. Further, in the example of the figure, since the basic score is defined as “20” and the coefficient is defined as “1-image similarity”, when the image similarity is 20%, the relevance rate (the purchaser and the exception applicant) Is different = probability of becoming True) is 80%, and the abnormal value is the number of basic points × the matching rate = 16. Whether or not they are the same person is determined based on the facial feature amount of the customer recorded in the recording database 23 for each accounting process (see FIG. 4). That is, the facial feature amount of the target customer at the time of exception application (exception processing) (the facial feature amount calculated by analyzing the imaging result of the cashier monitoring camera 9 and normalizing the image of the facial part) And the facial feature quantity of the target customer at the time of product purchase (during the scanning process) specified from the receipt or the like presented at the time of exception application, and the degree of similarity is determined.

  Other generation elements are as shown in the figure. As a simple supplement, for “element number: E3”, “element number: E5”, and “element number: E6”, a moving area is detected from the imaging result of the cashier monitoring camera 9 (m6), and the moving area is detected. Therefore, the determination is made by repeating the steps of detecting the movement of the bill, the opening / closing of the human face, the hand and the drawer. For “element number: E5”, it is determined from the monitoring results of the card reader 4 and the cashier monitoring camera 9 that the number of salesclerks around the cashier counter 10 is determined to be 1 and the number of customers is 0. It becomes.

  For “element number: E4”, the in-store monitoring camera 11 arranged in the vicinity of the entrance of the changing room uses the facial feature amount of the operator (stored in the facial feature amount database 21 in association with the clerk ID). When a face feature amount similar to is detected, “occurrence” is detected. For “element number: E4”, the card reader 4 may be installed in the vicinity of the entrance of the changing room, and the clerk card C may be read by the card reader 4 to detect the clerk ID.

  In FIG. 6, only six fraud elements E1 to E6 are illustrated, but the following elements can be added as fraud elements. For example, “the exception applicant is silent and in front of the cash register” may be detected from the voice acquired by the customer microphone 6a (m3). This is because a customer usually utters a phrase such as “The receipt is incorrect” at the time of an exception application, and it is considered that the possibility of fraud is high if the voice is silent. In this case, it is determined that “the exception applicant is silent and in front of the cash register” on the condition that a predetermined voice keyword (for example, “mistake”) cannot be recognized. The voice keyword can be defined by selecting one or more keywords from the keyword database 22 (see FIG. 1). Similarly, it may be detected from the sound acquired by the store clerk microphone 7a (m3) that the store clerk is performing a cash register operation silently.

  Note that the customer microphone 6a for collecting the customer's voice and the clerk microphone 7a for collecting the clerk's voice may not be provided independently, but may be collected by one microphone. In this case, a store clerk voice print storage unit (not shown) for storing the store clerk's voice print is provided in the receipt printer 2 or the store management server 12. Then, the voice recognition unit 65 compares the acquired voice voice print with the voice print stored in the store clerk voice print storage unit, and discriminates between the store clerk's voice and the customer's voice according to whether or not they match. . According to this configuration, it is not necessary to provide the clerk and customer microphones independently. Thereby, the cost reduction of the store monitoring system SY can be achieved.

  Further, “a large difference between the original data of the occurrence of the exception and the processing data” may be detected from the reading result of the receipt scanner 8 (m5) as a fraud occurrence factor. This is to cope with a technique for performing a return processing using a forged receipt. In this case, “occurrence / non-occurrence” and “matching rate” are detected from the comparison result between the product information (receipt description item) read at the time of product purchase and the product information read at the time of exception application.

  Further, it may be detected from the reading result of the card reader 4 (m1) that “there is only one store clerk around the cashier counter” as a fraud occurrence factor. This is because fraudulent acts are usually performed when there are no other clerk around. In this case, “occurrence / non-occurrence” is detected according to whether or not the number of salesclerk IDs acquired by the card reader 4 is 1 or less. Further, as a fraud occurrence factor, “there is no customer in front of the checkout counter” may be detected from the imaging result of the checkout monitoring camera 9 directed at the checkout. In this case, “occurrence / non-occurrence” is detected according to whether or not face detection has been performed from the imaging result of the cashier monitoring camera 9.

  Further, as a fraud occurrence factor, “the store clerk is sweating” may be detected from the measurement result of the sweat sensor 45 (m2). This is because when a cheating is performed, a change occurs in the clerk's biological signal. In this case, “occurrence / non-occurrence” is detected based on the increase in the amount of sweating before and after the exception processing. However, when the thermometer 44 detects that the temperature difference before and after the exception process is large (when it suddenly becomes hot), the precision is calculated in consideration of the temperature difference (coefficient based on the temperature difference). It is preferable. Moreover, you may detect from a measurement result of the pulse sensor 46 (m2) that a salesclerk's palpitation is intense as a fraud occurrence factor. In this case, “occurrence / non-occurrence” is detected based on the amount of increase in the pulse rate before and after the exception processing. However, when the movement amount of the operator before the exception process is large from the monitoring result of the cashier monitoring camera 9, it is preferable to obtain the matching rate in consideration of the movement amount (coefficient based on the movement amount).

  Next, the generation order of fraud patterns will be described with reference to FIG. This figure shows a definition example of a fraud occurrence pattern (described as “pattern A1”). Pattern A1 corresponds to a technique in which the store clerk performs a cancellation process using a receipt that the customer did not receive, dresses cash equivalent to the sales amount of the receipt, and hides the dressed cash in the locker in the changing room It is. As shown in the figure, the fraud occurrence pattern can define “occurrence order” of each fraud occurrence element and “interval between each element”. Here, the “occurrence order” is defined by the described numbers, but it is not always necessary to define all fraud occurrence elements (the fraud occurrence elements may be included regardless of the occurrence order, FIG. 12). See black circle). In addition, the fraudulent occurrence element whose “occurrence order” is blank (hyphen display) means that it is not a detection target. Further, “interval with each element” means a time limit. For example, when “element number: E1” occurs (the POS exception key is pressed) within 10 seconds from the occurrence of “element number: E5” (the store clerk crawls) Assume that fraudulent elements occur in the specified order of occurrence.

  On the other hand, as shown in the lower part of the figure, each fraud occurrence pattern can be defined with a “matched addition value”. This “matched addition value” is a value added to the total value of abnormal values when the “occurrence order” defined in the fraud occurrence pattern matches the occurrence order detected by the fraud occurrence detection unit 72. It is. That is, the total value of the abnormal value total value and the coincidence addition value is calculated as the degree of fraud.

  Next, the detection result of the fraud occurrence pattern (pattern A1) and the calculation result of the fraud level based on the detection result will be described with reference to FIG. As described above, the fraud occurrence detection unit 72 detects occurrence / non-occurrence for each fraud occurrence element. In addition, for the fraud elements in which “coefficient” is defined as “weighting” (see FIG. 6), the precision is detected. For fraud elements in which “occurrence” is detected and the relevance rate is not detected, the number of basic points is calculated as an abnormal value. Further, the fraud occurrence element in which “occurrence” is detected and the relevance rate is detected (in the example of the figure, “element number: E6”) is calculated as the number of basic points × the relevance rate = abnormal value. Further, although not present in the example of the figure, the abnormal value is set to “0” for the fraud element in which “non-occurrence” is detected. In addition, the example in the figure shows a case where the “occurrence order” defined in the fraud occurrence pattern matches the occurrence order detected by the fraud occurrence detection unit 72. Therefore, the degree of fraud is calculated by adding the coincidence addition value to the sum of the abnormal values of all fraud elements defined in the pattern A1. In the present embodiment, when the degree of fraud exceeds a predetermined value “60”, a warning is given by the operator warning unit 68 and the administrator warning unit 75. Therefore, in the example shown in the figure, since the degree of fraud exceeds the predetermined value “60”, a warning is issued.

  Next, the flow of the fraud level recording process of the store monitoring system SY will be described with reference to the flowchart of FIG. When the accounting process is started (S10), the store monitoring system SY determines whether or not exception processing has occurred (S11, exception processing detection unit 62), and determines that exception processing has occurred. (S11: Yes), a fraud occurrence pattern corresponding to the type of exception processing is extracted from a number of fraud occurrence patterns stored in the fraud occurrence pattern storage unit 71 (S12, see FIG. 3). Further, the detection result of the fraud occurrence detection unit 72 is acquired (S13), the total value of the abnormal values is reset to “0” (S14), and the basic score × the precision = The abnormal value is added to the total value of the abnormal values (S15). Thereafter, it is determined whether or not calculation has been performed for all fraud elements in the same fraud pattern (S16). If there is an uncalculated fraud element (S16: No), the process returns to S15.

  On the other hand, when calculation is performed for all fraud elements (S16: Yes), the occurrence order is checked (S17). Here, when the occurrence order defined in the fraud occurrence pattern matches the occurrence order detected by the fraud occurrence detection unit 72 (S18: Yes), the coincidence addition value is added to the total value of the abnormal values. The fraud level is calculated (S19). If the occurrence order does not match (S18: No), the sum of abnormal values is calculated as the degree of fraud (S20). As described above, S12 to S20 are steps executed by the fraud degree calculation unit 73.

  Thereafter, the fraud level is recorded in the recording database 23 for the fraud occurrence pattern that is the target of fraud level calculation (S21, fraud level recording unit 74). Also, it is determined whether the patterns are matched (whether the occurrence order is matched) or whether the calculation has been completed for all fraud patterns that are associated with the same exception processing type (S22). If not (S22: No), the process returns to S14 to repeat the fraud degree calculation process. Further, in the case of S22: Yes, it is determined whether or not there is a fraud pattern that has a fraud degree exceeding a predetermined value (S23). If there is a fraud pattern that exceeds a predetermined value (S23: Yes), A warning is issued by the operator warning unit 68 and the administrator warning unit 75 (S24). Further, when there is no fraud pattern exceeding the predetermined value (S23: No), and when exception processing has not occurred (S11, for example, during the settlement process), the fraud level recording process is performed without warning. Exit.

  In S24, when the administrator warning unit 75 transmits an e-mail, the detection result of each fraud occurrence element and the calculated fraud level are notified for one or more fraud occurrence patterns whose fraud levels exceed a predetermined value. . Further, when there are a plurality of fraud patterns that have a fraud level exceeding a predetermined value, notification may be made only for the fraud pattern having the highest fraud level.

  As described above, according to the store monitoring system SY according to the first embodiment, the degree of fraud is calculated by summing the abnormal values corresponding to one or more fraud elements that have been detected by the fraud occurrence detection unit 72. Since the calculation result is recorded in the recording database 23, it is possible to grasp the possibility of fraud from the record. As a result, it is possible to accurately determine the presence or absence of a fraudulent act even for a complex act of fraud. In addition, since the degree of fraud is calculated by taking into account the accuracy rate of fraud elements and the order of occurrence of fraud elements, more accurate fraud levels can be calculated. Further, since a warning is issued when the degree of fraud exceeding a predetermined value is calculated, it is possible to notify the store clerk or manager.

  Further, since the fraud occurrence pattern is stored in the fraud occurrence pattern storage unit 71 for each exception processing type, a more realistic fraud level is calculated using the fraud occurrence pattern suitable for the exception processing type. be able to. In addition, since a plurality of fraud occurrence patterns can be stored for each exception processing type, it is possible to deal with a number of illegal acts. In general, fraudulent acts are often executed along with exception processing. Therefore, as in this embodiment, fraudulent activity is calculated and recorded only when exception processing occurs, so that fraud can be efficiently performed. Actions can be detected.

  In the above-described embodiment, the operator warning unit 68 and the administrator warning unit 75 indicate warning modes such as a warning destination and a warning method according to the calculated degree of fraud and the type of exception processing detected at that time. May be variable. “Variable warning destination” means increasing or decreasing the number of e-mail transmission destinations, switching whether or not there is a warning to the operator, and the like. “Variation of warning method” means that in the case of voice warning, the voice pattern, volume, number of warnings, warning interval, etc. are changed. In the case of a display warning, it means that the display content, display color, display frequency, display interval, and the like are variable. In the case of a notification warning (e-mail notification, etc.), it means that the notification content, notification method, number of notifications, etc. are changed. Furthermore, changing warning means such as voice, display, and notification is also included in changing the warning method. Note that the correspondence between the degree of fraud and the warning destination or warning method may be changed by the user using an operation unit provided in the receipt printer 2 or the store management server 12. According to this configuration, it is possible to issue a warning in a warning mode (a warning destination and a warning method) suitable for the type of exception processing. In addition, by changing the warning mode according to the type of exception processing, the operator or the administrator can easily guess the method of fraud that has occurred from the warning mode.

  In the above-described embodiment, the fraud level is calculated and recorded only when exception processing occurs. However, the fraud level may be calculated and recorded during all accounting processes. According to this configuration, the recorded data can be used for store clerk operation education and customer trend analysis.

  In the above embodiment, the detection result and the reading result by the card reader 4, the clerk wristband 5, the customer microphone 6 a, the clerk microphone 7 a, and the receipt scanner 8 are sent to the store management server 12 via the receipt printer 2. Although it is configured to transmit, it may be configured to transmit directly to the store management server 12 via the in-store LAN 13. That is, each unit other than the printing unit 67 and the operator warning unit 68 of the receipt printer 2 may be provided in the store management server 12. Conversely, each part of the store management server 12 may be provided in the receipt printer 2.

  In the above-described embodiment, the administrator warning unit 75 uses an e-mail or a Web application to issue a warning. However, as with the operator warning unit 68, display on the display, sound generation, and LED lighting. / Alternative warning may be performed using blinking, vibration by a vibrator, or the like.

[Second Embodiment]
Next, a second embodiment of the present invention will be described with reference to FIGS. This embodiment makes it a subject to detect the trick of the clever technique which makes a customer an accomplice among the fraud which occurs in a store. Therefore, only differences from the first embodiment will be described. In the present embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted. Moreover, the modification applied about the component similar to 1st Embodiment is applied similarly about this embodiment.

  FIG. 10 is a functional block diagram of the store monitoring system SY according to the second embodiment. Compared with the first embodiment, the store management server 12 of the present embodiment includes an fraud occurrence pattern rewriting unit 76 and a customer detection unit 77. First, the customer detection unit 77 detects a customer at the time of transaction processing based on the imaging result of the cash register imaging unit 58. Specifically, an imaging result is acquired, a difference between frames and a background difference are extracted, and a moving region is detected from the extraction result. Then, face detection is performed from the detected moving area, and when the face is detected, it is determined that the customer exists there. Then, the fraud level recording unit 74 detects the customer identification information (customer information such as customer ID, face feature amount and voiceprint) detected by the customer detection unit 77 and the fraud occurrence detection unit 72 (situation data detection unit). The result (situation data in the store) is associated with the fraud degree calculated by the fraud degree calculation unit 73 and recorded in the recording database 23 (see FIG. 4).

  On the other hand, the fraud occurrence pattern rewriting unit 76 refers to the recording database 23, and executes the exception processing execution count for the same customer (in this embodiment, recording is performed only during exception processing, so the exception processing execution count = recording). When it is determined that the number exceeds a predetermined number, the basic score is increased for at least a part of one or more fraud occurrence elements described in the fraud occurrence pattern. In this way, when the basic score is rewritten by the fraud occurrence pattern rewriting unit 76, the fraud degree calculation unit 73 calculates the fraud degree based on the fraud occurrence pattern after rewriting. Details will be described later.

  Further, the operator warning unit 68 and the administrator warning unit 75 of the present embodiment issue a warning when the cumulative total of the fraud degrees of the same customer recorded in the recording database 23 exceeds a predetermined value. In the first embodiment described above, a warning is given when the fraud degree calculated by the fraud degree calculation unit 73 exceeds a predetermined value. However, in this embodiment, the fraud calculated for one accounting process is performed. The difference is that a warning is issued not only when the degree exceeds a predetermined value but also when the cumulative total of past fraud degrees exceeds a predetermined value.

  Here, with reference to FIG. 11, the necessity determination of the warning based on the cumulative fraud level will be described. The figure shows a table recording fraud history for three customers. The table corresponds to a part of information recorded in the recording database 23 (see FIG. 4). In the example shown in the figure, for example, for the customer of “customer ID: 0123”, the fraud degree calculated during the first exception processing (the highest value when fraud degrees are calculated for a plurality of fraud occurrence patterns) is 25. The fraud level calculated during the second exception processing is 42, and the cumulative value of the fraud level has exceeded a predetermined value (here, “60”) during the second exception processing. Indicates what has been done. Similarly, for the customer of “customer ID: 0321”, since the cumulative value of the inaccuracy exceeds a predetermined value during the third exception process, a warning is given during the third exception process. Furthermore, for the customer of “customer ID: 0225”, the fraud degree calculated at the first exception processing is 77, so the cumulative value is 77, and the predetermined value is exceeded, so a warning is issued at the first exception processing. Indicates what has been done. For customers who have been warned based on the cumulative value of fraud, a warning is always issued when exception processing occurs. However, the accumulated value may be reset when a warning is issued according to user settings.

  Also, the predetermined value X for determining the fraud level calculated for one exception process may be different from the predetermined value Y for determining the cumulative value of fraud levels. However, the relationship between the predetermined value X and the predetermined value Y is that X ≦ Y. In the example shown in the figure, the customer ID is used as the customer identification information. However, other customer information (biological information such as a facial feature amount and a voiceprint) may be used.

  Next, rewriting of the fraud occurrence pattern will be described with reference to FIGS. FIG. 12 is a diagram illustrating an example of defining the order of occurrence of fraud patterns (pattern A2). The pattern A2 corresponds to a technique in which the store clerk uses a receipt that the customer has not received to wear cash corresponding to the sales amount of the receipt, and passes the dressed cash to the customer who is the accomplice. It is assumed that “matched addition value” of pattern 2 is defined as “15”. Note that “black circle” described in “occurrence order” means that the fraudulent occurrence element is a detection target, but the occurrence order is not defined. In addition, “element number: E2” is defined as “interval with each element”, and the imaging result within a predetermined time before and after the time when the receipt is recognized from the imaging result of the registration monitoring camera 9 Based on this, the “occurrence” is detected.

  FIG. 13 is a diagram illustrating a detection result for the fraud occurrence pattern (pattern A2) and a fraud level calculation result. When the detection result shown in the column “occurrence / non-occurrence, relevance ratio” in the figure is obtained, the coincidence addition value “15” is added to the total value “17” of the abnormal values to obtain the fraud degree “32”. Calculated. Therefore, in the first embodiment, warning based on the fraud occurrence pattern is not performed. However, in the present embodiment, when the number of records recorded in the recording database 23 for the target customer exceeds a predetermined number, the basic score is rewritten for a specific fraud element. In this embodiment, “element number: E2” is defined as a rewrite target, and the basic score is rewritten from “20” to “100”. There may be a plurality of fraud elements to be rewritten, and it may be defined for each fraud pattern. The basic score after rewriting of each fraud occurrence element may be set in common for all fraud occurrence patterns, or may be set for each fraud occurrence pattern. In the example shown in the figure, the total value of abnormal values is changed to “73” by rewriting the basic score. Therefore, by adding the coincidence addition value “15”, the fraud degree is calculated as “88”, and a warning is issued because it exceeds the predetermined value. In addition, the pattern A1 shown in FIG. 8 and the pattern A2 shown in FIG. 13 have the same basic score assigned to each fraud occurrence element, but it is possible to define different basic scores for each fraud occurrence pattern. May be.

  Next, the flow of the fraud level recording process of the store monitoring system SY according to the second embodiment will be described with reference to the flowchart of FIG. This flowchart is executed following the flowchart shown in FIG. If it is determined in S23 of the flowchart shown in FIG. 9 that there is a fraud occurrence pattern in which the calculated degree of fraud exceeds a predetermined value (S23: Yes), a warning is issued by the operator warning unit 68 and the administrator warning unit 75. Perform (S24). On the other hand, when it is determined that there is no fraud pattern in which the calculated fraud degree exceeds the predetermined value (S23: No), it is determined whether or not the cumulative value of past fraud degrees exceeds the predetermined value (S32). If it exceeds (S32: Yes), a warning is given (S24).

  If the cumulative value of past fraud levels does not exceed the predetermined value (S32: No), then it is determined whether or not the number of executions of the exception process of the target customer exceeds the predetermined number (S33). If the number exceeds (S33: Yes), the basic score of the specific fraud element is rewritten (S34). Note that S33 and S34 are steps executed by the fraud occurrence pattern rewriting unit 76. Thereafter, the fraud level is recalculated and recorded for the fraud pattern including the specific fraud element (S35, fraud level calculation unit 73, fraud level recording unit 74). S35 corresponds to S14 to S22 of the flowchart shown in FIG. As a result, if there is a fraud pattern in which the calculated degree of fraud exceeds a predetermined value (S36: Yes), a warning is given (S24). Further, when there is no fraud occurrence pattern in which the calculated fraud degree exceeds the predetermined value (S36: No), and when the number of executions of the exception processing of the target customer does not exceed the predetermined number (S33: No), a warning is issued. The fraud level recording process is terminated without performing it.

  As described above, according to the store monitoring system SY according to the second embodiment, the calculated degree of fraud is recorded together with the customer identification information, so that the possibility that the customer is an accomplice can be estimated from the record. . As a result, it is possible to easily detect fraudulent actions even for complicated techniques involving accomplices. In addition, when it is determined from the number of records recorded in the recording database 23 that the number of times that the exception processing has been executed for the same customer exceeds a predetermined number, one or more described in the fraud occurrence pattern Since the basic score is increased for at least some of the fraud elements, it is possible to reliably detect malicious customers who repeat fraudulent acts as accomplices. Furthermore, even if the customer has not been warned in the past, if the cumulative fraud level exceeds a predetermined value, a warning is given because there is a possibility that the customer is an accomplice. In this way, by performing fraudulent recording, it is possible to detect a technique (a technique involving an accomplice) that is difficult to determine by one accounting process.

  The recording database 23 described in the above embodiment may be shared among a plurality of stores. According to this structure, the fraudulent act can be detected also about the accomplice who repeats a crime in a some store.

  Further, in the above embodiment, when the number of executions of exception processing exceeds a predetermined number, the basic score of a specific fraud element is rewritten, but the coefficient may be rewritten instead of the basic score. Further, instead of rewriting the weights of specific fraud elements such as the basic score and coefficients, a method of rewriting the coincidence addition value or further adding the specific addition value may be employed. In addition, a plurality of thresholds may be provided, and the number of basic points, coefficients, coincidence addition value, and specific addition value may be increased in stages, or based on a predetermined calculation formula (algorithm) using the number of executions of exception processing as a parameter The degree of fraud may be calculated.

  In the above embodiment, a warning is issued when the cumulative value of the degree of fraud exceeds a predetermined value. However, the number of exception processing executions exceeds the predetermined number, and the cumulative value of the degree of fraud exceeds a predetermined value. A warning may be given in case. In addition, the user may be able to set conditions for performing warnings, conditions for rewriting fraud patterns, and items to be rewritten in that case (the above-mentioned basic points, coefficients, coincidence addition values, specific addition values, etc.).

  Further, in the above embodiment, when exception processing is performed, fraud pattern rewriting and fraudulent cumulative value calculation are performed, but only when a predetermined exception processing type occurs. Also good. In other words, only when the exception processing detection unit 62 determines that a predetermined exception processing type has occurred, the fraud occurrence pattern may be rewritten and the total fraud degree may be calculated. According to this configuration, fraud detection can be performed efficiently and effectively by targeting exception processing types that are likely to be used as fraudulent techniques involving accomplices.

  In the above embodiment, it is described that the facial feature amount or the voiceprint is used as the customer information. However, a fingerprint, a palm shape, a retina, an iris, a blood vessel (vein pattern), an ear shape, or the like may be used. As described above, by using the physical feature (static information), it is not necessary to monitor the behavior, and an individual can be easily specified.

  In the above embodiment, the customer detection unit 77 detects the presence / absence of a customer based on the imaging result of the cashier monitoring camera 9, but may detect the presence / absence of the customer using an infrared sensor. Alternatively, the customer may have a membership card in which a wireless IC is incorporated, and the presence or absence of the customer may be detected depending on whether the membership card is detected by a card reader. In this case, the card reader is connected to the receipt printer 2 or the in-store LAN 13.

  As mentioned above, although 1st Embodiment and 2nd Embodiment were shown, it is possible to provide each step of the flowchart (Frequency recording process of store monitoring system SY) shown in these each embodiment as a program. Further, the program can be provided by being stored in various recording media (CD-ROM, flash memory, etc.). That is, a program for causing a computer to function as each component of the store monitoring system SY and a recording medium on which the program is recorded are also included in the scope of rights of the present invention.

  In each of the above embodiments, the fraud level recording apparatus of the present invention is used for preventing fraud in a store. However, the fraud level recording device of the present invention may be applied to a system for other purposes (nurturing and managing store clerks). good. Moreover, you may implement | achieve each part of the shop management server 12 with a POS system or a WWW server. In addition, the system configuration and processing steps of the store monitoring system SY can be appropriately changed without departing from the gist of the present invention, regardless of the above-described embodiment.

  DESCRIPTION OF SYMBOLS 1 ... POS terminal 2 ... Receipt printer 3 ... Handy scanner 4 ... Card reader 5 ... Clerk wristband 6 ... Customer display 6a ... Customer microphone 7 ... Headset 7a ... Clerk microphone 8 ... Receipt scanner 9 ... Registration monitor camera 10 ... cashier counter 11 ... in-store monitoring camera 12 ... store management server 13 ... in-store LAN 23 ... database for recording 33 ... manager's mobile phone 45 ... sweat sensor 46 ... pulse sensor 62 ... exception processing detector 68 ... operator warning section 71 ... Fraud occurrence pattern storage unit 72 ... Fraud occurrence detection unit 73 ... Fraud degree calculation unit 74 ... Fraud degree recording unit 75 ... Manager warning unit 76 ... Fraud occurrence pattern rewriting unit 77 ... Customer detection unit SY ... Store monitoring system

Claims (13)

A customer detection unit for detecting a customer during accounting processing by the accounting processing unit;
A situation data detection unit for detecting situation data in the store at the time of the accounting process;
Based on the detection result of the situation data detection unit, a fraud level calculation unit that calculates the fraud level of the customer;
A fraud level recording apparatus comprising: a fraud level recording unit that records a calculation result of the fraud level calculation unit in a database together with customer identification information detected by the customer detection unit. A fraud occurrence pattern storage unit that stores fraud occurrence patterns describing a basic score that is a weight for each of one or more fraud occurrence elements assumed in the store,
The situation data detection unit detects occurrence / non-occurrence of each of the one or more fraud elements,
The fraud degree according to claim 1, wherein the fraud degree calculation unit calculates the fraud degree by summing up the basic points corresponding to the fraud occurrence elements detected by the situation data detection unit. Recording device. The accounting processing unit further includes an exception processing detection unit that detects that an exception processing corresponding to a part of the accounting processing has occurred,
3. The fraud level recording apparatus according to claim 2, wherein the fraud level calculation unit functions on condition that an exception process is detected by the exception process detection unit.   The exception processing includes “cancellation processing” for canceling input data during accounting processing of the accounting processing unit, “cancellation processing” for canceling the accounting processing after the accounting processing, and for returning the product. "Return processing", "discount processing" for discounting the product, "discount processing" for discounting the product, and "exchange processing" for money exchange, in the cash drawer At least one of “inspection processing” for matching money with sales, “withdrawal / refund processing” for withdrawing or refunding, and “reissue processing” for reissuing receipts The fraud level recording apparatus according to claim 3, further comprising: The exception processing detection unit detects a type of exception processing,
The fraud occurrence pattern storage unit stores the fraud occurrence pattern for each exception processing type,
The fraud level calculation unit extracts the fraud occurrence pattern corresponding to the type of exception processing detected by the exception processing detection unit from the fraud occurrence pattern storage unit, and calculates the fraud level. The injustice degree recording apparatus according to claim 3 or 4. When it is determined from the number of records recorded in the database that the number of times that the exception processing has been executed for the same customer exceeds a predetermined number, one or more frauds described in the fraud occurrence pattern A fraud occurrence pattern rewriting unit that increases the basic score of at least some of the occurrence elements is further provided,
The fraud level recording apparatus according to claim 3, wherein the fraud level calculation unit calculates the fraud level based on the fraud occurrence pattern after rewriting.   The fraud level recording apparatus according to claim 6, further comprising a warning unit that issues a warning when the fraud level is calculated by the fraud level calculation unit to be equal to or greater than a predetermined value.   6. The warning unit according to claim 1, further comprising a warning unit that issues a warning when the cumulative degree of fraud of the same customer recorded in the database exceeds a predetermined value. Fraud recording device.   The warning unit notifies a predetermined administrator of a warning document including the detection result of the customer detection unit and the status data detected by the status data detection unit using an e-mail or a Web application. The fraud recording apparatus according to claim 7 or 8, characterized in that:   The fraud recording apparatus according to claim 1, wherein the customer identification information is biometric information of the customer.   11. The fraud level recording apparatus according to claim 1, wherein the database is shared by a plurality of fraud level recording apparatuses. A method for controlling a fraud recording apparatus having an accounting processing unit and a database inside or outside,
A customer detection step of detecting a customer during accounting processing by the accounting processing unit;
A status data detection step of detecting status data in the store during the accounting process;
A fraud degree calculating step of calculating the fraud degree of the customer based on the detected situation data;
A fraud level recording step of recording a fraud level recording step in which the calculation result of the fraud level calculation step is recorded in the database together with customer identification information detected by the customer detection step.   A program for causing a computer to execute each step in the control method for a fraud recording apparatus according to claim 12.

Images