METHOD AND APPARATUS FOR DETECTING FRAUD ATTEMPTS IN REVERSE VENDING MACHINES

• US 20140147005A1
• Filed: 06/22/2012
• Published: 05/29/2014
• Est. Priority Date: 06/24/2011
• Status: Active Grant

• Current Assignee: Tomra Systems ASA
• Original Assignee: Tomra Systems ASA
• Inventors: Kavli, Tom ystein, Njstad, Johnny, Sther, Geir
• Granted Patent: US 9,189,911 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06Q 20/4016 involving fraud or risk level assessment in transaction processing
• CPC Class Codes: G06T 7/0002 Inspection of images, e.g. flaw detection
• CPC Class Codes: G06T 7/0004 Industrial image inspection
• CPC Class Codes: G07F 7/00 Mechanisms actuated by objects other than coins to free or to actuate vending, hiring, coin or paper currency dispensing or refunding apparatus
• CPC Class Codes: G07F 7/06 by returnable containers , i.e. reverse vending systems in which a user is rewarded for returning a container that serves as a token of value, e.g. bottles
• CPC Class Codes: G07F 7/0609 by fluid containers, e.g. bottles, cups, gas containers
• CPC Class Codes: G07G 1/0054 with control of supplementary check-parameters, e.g. weight or number of articles
• CPC Class Codes: Y02W 30/60 Glass recycling

Abstract

A reverse vending machine, including: a chamber adapted to receive an object returned to the reverse vending machine; a plurality of cameras arranged around the perimeter of the chamber for viewing said object; a transparent or translucent plate arranged such that the cameras in use view the object obliquely through the transparent or translucent plate; and means adapted to couple light into the plate such that the light undergoes total internal reflection in the plate. Also, a method of detecting dirt in a reverse vending machine.

Claims

1. A method for detecting fraud attempts in a reverse vending machine, the reverse vending machine comprising at least one image sensor and at least one data processing unit, the method comprising:

using said at least one image sensor to obtain a set of one or more images of a returned item inserted into the reverse vending machine;

using at least one data processing unit to apply a first feature extraction algorithm to extract a first feature from at least a first subset of said set of one or more images;

using at least one data processing unit to apply a second feature extraction algorithm to extract a second feature from i) at least a second subset of said set of one or more images or ii) data received from an additional sensor; and

using at least one data processing unit to determine whether a position or a motion of said first feature is inconsistent with a position or motion of said second feature.

2. A method according to claim 1, wherein said second feature extraction algorithm is applied to data received from an ultrasonic sensing device included in said reverse vending machine.

3. A method according to claim 1, wherein said second feature extraction algorithm is applied to data received from a laser ranging device included in said reverse vending machine.

4. The method of claim 1, wherein:

obtaining said set of one or more images includes obtaining at least two images obtained at least two different points in time.

5. The method of claim 4, wherein:

said first feature extraction algorithm is used to extract a feature from an image obtained at a first point in time, said feature being indicative of a relative position of an object at said first point in time;

said second feature extraction algorithm is used to extract a feature from an image obtained at a second point in time, said feature being indicative of a relative position of said object at said second point in time; and

said determining includes estimating a velocity of said object based on said relative positions and the time interval between said first and second point in time, and comparing said estimated velocity with an expected velocity.

6. The method of claim 5, wherein said expected velocity is any velocity which is positive relative to the direction a returnable item normally travels through the reverse vending machine, and said determining indicates inconsistent motion if said estimated velocity is a velocity which is negative relative to the direction a returnable item normally travels through the reverse vending machine.

7. The method of claim 5, wherein said expected velocity is based on the velocity at which a conveyor arranged to transport returned items through said reverse vending machine is operating, and said determining indicates inconsistent motion if said estimated velocity is substantially different from said expected velocity.

8. The method of claim 5, wherein said expected velocity is based on a determined estimated velocity resulting from the application of said first feature extraction, said second feature extraction and said determining on a previous set of images, and the method further comprises using the result of said determining to update the expected velocity prior to the application of said first feature extraction, said second feature extraction and said determining on a subsequent set of images.

9. The method of claim 5, wherein:

said first feature extraction algorithm includes applying edge detection along two axes and, based on said edge detection, identifying a region of interest in said first image; and

said second feature extraction algorithm includes searching for a region in said second image which is highly correlated with said region of interest in said first image.

10. The method of claim 9, wherein said first image is obtained at a current point in time and said second image is obtained at a previous point in time.

11. The method of claim 1, wherein:

said first feature extraction algorithm is used to extract a first feature indicative of a relative position of a first object associated with said first feature from at least two images obtained at least two different points in time;

said second feature extraction algorithm is used to extract a second feature indicative of a relative position of a second object associated with said second feature from at least two images obtained at least two different points in time; and

said determining includes estimating relative velocities for said first and said second object based on said relative positions and time intervals between the different points in time, and indicating an inconsistent motion of said first object relative to said second object if there is a substantial difference between said first and said second estimated velocity.

12. The method of claim 11, wherein said first feature extraction algorithm is applied to images obtained from a first image sensor and said second feature extraction algorithm is applied to images obtained from a second image sensor, and where said first feature is representative of a barcode and said second feature is representative of a shape of a returned item.

13. The method of claim 1, wherein:

said first feature extraction algorithm includes detecting the presence and position of a thin object in at least one image from said first subset of images;

said second feature extraction algorithm includes detecting a position of a returned item from at least one image from said second subset of images or from said sensor data; and

said determining includes comparing the position of the detected thin object relative to the position of the returned item.

14. The method of claim 13, wherein detecting the presence and position of a thin object includes using a foreground-background segmentation method to generate a processed image, classifying each pixel in said processed image as either an object pixel or a background pixel, and classifying each object pixel as either “

normal”

or “

thin”

.

15. The method of claim 14, wherein said foreground-background segmentation method includes dividing a current image by a reference image.

16. The method of claim 1, wherein:

said first feature extraction algorithm includes detecting the presence and position of at least one characteristic feature of the shape of a returned item from at least two images in said first set of images, said at least two images being obtained at different points in time;

said second feature extraction algorithm includes detecting a position of a returned item from at least two images from said second subset of images or from said sensor data; and

said determining indicates an inconsistency if said characteristic feature has a relative position with respect to the overall shape of the returned item which differs substantially in said at least two images from said first set of images.

17. The method of claim 16, wherein detecting the presence and position of at least one characteristic feature of the shape of a returned item includes using a foreground-background segmentation method to generate a processed image, classifying each pixel in said processed image as either an object pixel or a background pixel, and detecting a feature as a peak in the 1st derivative of a curve tracing the edge between object pixels and background pixels.

18. The method of claim 17, wherein said foreground-background segmentation method includes dividing a current image by a reference image.

19. The method of claim 13, wherein said first subset of images are obtained using an image sensor which is configured to view a returned item silhouetted against a background including a retroreflector, and where a light source and a beam splitter are configured such that said light source and said image sensor are optically co-located as seen from the retroreflector.

20. A reverse vending machine with an opening through which items can be inserted into the interior of said machine, at least one image sensor configured to obtain images of at least a part of an inserted item, and at least one data processing unit, the reverse vending machine comprising:

at least one data processing unit configured to receive a set of one or more images of a returned item inserted into the reverse vending machine from said at least one image sensor;

at least one feature extraction unit configured to apply a first feature extraction algorithm to extract a first feature from at least a first subset of said set of one or more images and apply a second feature extraction algorithm to extract a second feature from i) at least a second subset of said set of one or more images, or ii) data received by said at least one data processing unit from an additional sensor; and

at least one data processing unit configured to determine whether a position or a motion of said first feature is inconsistent with a position or a motion of said second feature.

21. The reverse vending machine of claim 20, further comprising an ultrasonic sensor configured to obtain sensor data from inside said interior and provide said data to said at least one data processing unit;

and wherein said feature extraction unit is configured to apply said second feature extraction algorithm to data received from said ultrasonic sensor.

22. The reverse vending machine of claim 20, further comprising a laser ranging device configured to obtain sensor data from inside said interior and provide said data to said at least one data processing unit;

and wherein said feature extraction unit is configured to apply said second feature extraction algorithm to data received from said laser ranging device.

23. The reverse vending machine of claim 20, wherein said at least one data processing unit is configured to control said at least one image sensor to obtain images at least two different points in time.

24. The reverse vending machine of claim 23, wherein said at least one feature extraction unit is configured to apply said first feature extraction algorithm to an image obtained at a first point in time and to apply said second feature extraction algorithm to an image obtained at a second point in time;

and

said at least one data processing unit is configured to estimate a velocity of an object present in said interior based on the relative positions of said first and said second features, and to compare said estimated velocity with an expected velocity.

25. The reverse vending machine of claim 23, wherein said at least one feature extraction unit is configured to apply said first feature extraction algorithm and said second feature extraction algorithm to images obtained at least two different points in time;

and

said at least one data processing unit is configured to estimate relative velocities for a first and a second object associated with said first and said second feature, respectively, and to indicate an inconsistent motion of said first object relative to said second object if there is a substantial difference between said first and said second estimated velocity.

26. The reverse vending machine of claim 25, wherein said at least one image sensor includes at least one first image sensor with a relatively narrow field of view in order to capture images of barcodes attached to returned items at least partly present in said interior, and at least one second image sensor with a relatively wide field of view in order to capture images of the shape of returned items at least partly present in said interior.

27. The reverse vending machine of claim 20, further comprising:

a retroreflector, a beam splitter and a light source; and

wherein at least one image sensor is configured to view a returned object silhouetted against said retroreflector, and said light source and said beam splitter are configured such that said light source and said image sensor are optically co-located as seen from the retroreflector.

28. The reverse vending machine of claim 20, wherein the feature extraction unit comprises one or more components chosen from the group consisting of:

a set of computer program instructions stored on a computer readable media and configured to enable the at least one data processing unit to perform feature extraction functions, a field programmable gate array (FPGA), and an application specific integrated circuit (ASIC).