SYSTEM AND METHOD FOR SUSPECT SEARCH

• US 20140328512A1
• Filed: 12/18/2013
• Published: 11/06/2014
• Est. Priority Date: 05/05/2013
• Status: Active Grant

• Current Assignee: Qognify Ltd. (PSBU Midco SARL)
• Original Assignee: Nice Systems Limited (Nice Ltd)
• Inventors: YEHEZKEL, Raanan Yonatan, GOLDNER, Vladimir, BOUDOUKH, Guy, BLUMSTEIN-KOREN, Guy, GURWICZ, Yaniv, GIRMONSKY, Doron
• Granted Patent: US 9,471,849 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06F 18/2178 based on feedback of a supervisor
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06V 10/7784 based on feedback from supervisors
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

A system and method for detecting an object of interest. A system and method may generate a first signature for an object of interest based on an image of the object of interest. A system and method may generate a second signature for a candidate object based on an image of the candidate object. A system and method may calculate a similarity score by relating the first signature to the second signature and may determine the image of the candidate object is an image of the object of interest based on the similarity score.

Claims

1. A method of detecting an object of interest, the method comprising:

generating a first signature for the object of interest based on an image of the object of interest;

generating a second signature for a candidate object based on an image of a candidate object;

calculating a similarity score by relating the first signature to the second signature; and

determining if the image of the candidate object is an image of the object of interest based on the similarity score.

2. The method of claim 1, wherein the first signature is generated based on an image acquired by a first camera and wherein the second signature is generated based on an image acquired by a second camera.

3. The method of claim 1, wherein the first and second signatures include a respective first and second sets of characteristics and wherein calculating the similarity score includes relating the first set to the second set.

4. The method of claim 1, wherein generating a signature includes:

defining a curve on an object in an image such that the curve traverses regions associated with an object in the foreground of the image;

placing a plurality of patches along the curve with some overlap between the patches;

determining attributes of all of the patches;

identifying sets of patches having common attributes; and

generating the signature based on at least one set of patches.

5. The method of claim 1, comprising identifying an object of interest in an image by identifying background and foreground pixels in an image wherein identifying background and foreground pixels comprises:

defining a first region in the image, the first region including an image of the object of interest and background pixels;

defining a second region in the image, the second region including the first region;

designating pixels included in the second region and excluded by the first region as background pixels; and

based on attributes of the background pixels, identifying additional background pixels in the first region.

6. The method of claim 1, wherein the image of the object of interest is acquired at a first location and the image of the candidate object is acquired at a second location.

7. The method of claim 1, wherein the image of the object of interest is acquired using a first field of view (FOV) and the image of the candidate object is acquired using a second FOV, wherein the second FOV does not overlap the first FOV.

8. The method of claim 1, comprising detecting motion in a set of images and searching for an object of interest in an image only in areas where motion is detected.

9. The method of claim 1, comprising determining, for a first and second horizontal portions of an image a respective first and second object sizes and searching for an object of interest in the first and second horizontal portions based on the respective first and second object sizes.

10. The method of claim 1, wherein the signature of the object of interest is generated based on a characteristic of a device operated by the object of interest.

16. The method of claim 1, wherein the signatures of the object of interest and candidate object are generated such that they are invariant to at least one of:

a scaling of an image, an orientation of the object in an image, a light intensity in an image, an angle and a deformation of the object in an image and a color temperature of a light source.

19. The method of claim 2, comprising searching an image from the second camera based on determining that the RO traveled from a location of the first camera to a location of the second camera using license plate recognition (LPR).

21. The method of claim 1, comprising generating the first and second signatures based on the user ID attached to messages sent from devices operated by the object of interest and the candidate object, wherein the messages are related to a social network;

defining a score based on the similarity of the first and second signatures; and

associating the candidate object with the object of interest based on the similarity score.

22. The method of claim 21, comprising:

capturing a message sent by the object of interest or by the candidate object that was associated to the object of interest, the message related to a social network;

using data in the message to access information in a social network, the information related to the object of interest or to the candidate object that was associated to the object of interest; and

generating a profile of the object of interest based on the information.

11. A method of detecting an object of interest, the method comprising:

generating signatures for the object of interest and for a plurality of candidate objects based on images of the object and of the plurality of candidate objects;

selecting from the plurality of candidate objects, a set of candidate objects based on a set of similarity scores respectively associated with the set of candidate objects, the similarity scores calculated based on the signatures;

presenting images of the set of candidate objects to a user;

receiving input from the user, the input indicating candidate objects in the set that match the object of interest; and

based on the input;

modifying the signature of the object of interest;

modifying at least one parameter used for generating the signature, and

modifying at least one parameter used for calculating a similarity score.

12. The method of claim 11, comprising using a signature of a candidate object indicated as matching the object of interest to search for additional candidate objects.

13. The method of claim 11, comprising extending the set of candidate objects that match the object of interest by:

selecting, from a set of candidate objects that are already indicated by the user as matching the object of interest, a candidate object and using a signature of that selected candidate object to identify additional candidate objects, wherein that selected candidate object is one of;

a candidate object that is least similar to the object of interest;

a candidate object that was acquired at a time that is the furthest from the time the image of the object of interest was acquired;

a candidate object that was acquired at a location that is the farthest from the location of the object of interest; and

a candidate object selected by the user.

14. The method of claim 11, comprising presenting the set of candidate objects based on at least one of:

a time the images were acquired, a location where the images were acquired and a similarity of the candidate objects to the object of interest.

15. The method of claim 11, comprising sorting the images of the candidate objects according to a similarity of the candidate objects to the object of interest.

17. The method of claim 11, comprising:

obtaining signals from a plurality of devices in a plurality of regions;

determining, based on the signals, a set of devices that were present in at least two locations where the object of interest was present and at the same times the object of interest was present at those locations;

associating each of the devices in the set with a score based on the number of presences of the device at a locations and at the times where and when the object of interest was present;

associating a device with the object of interest based on the score; and

generating the signature for the object of interest based on information related to the device.

18. The method of claim 11, comprising:

using a set of signatures of objects of interest and candidate objects which were determined by the user as matching and a set of objects of interest and candidate objects which were determined by the user as not matching to train a classifier and using the classifier to identify matching candidate objects for a given objects of interest.

20. The method of claim 11, comprising:

determining a time period during which to start searching for the object of interest;

determining an initial location of the object of interest from where to begin the search;

determining a plausible area based on the initial location and the time period; and

searching for a set of candidate objects from a plurality of candidate objects at the plausible area.