Automated searching for probable matches in a video surveillance system

• US 20070025593A1
• Filed: 04/04/2006
• Published: 02/01/2007
• Est. Priority Date: 04/05/2005
• Status: Active Grant

• Current Assignee: 3VR Security Incorporated (Identiv, Inc.)
• Original Assignee: 3VR Security Incorporated (Identiv, Inc.)
• Inventors: Vallone, Robert, Russell, Stephen, Frederick, Timothy, Haupt, Gordon, Fleischer, Stephen
• Granted Patent: US 8,130,285 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

A technique for searching for probable matches in a video surveillance system is disclosed. A new event, such as a face captured in an image set, is matched against other events in a database of events. A similarity score is generated based on the difference between the new event and other events in the database. The similarity score may be weighted by information external to the image sets. Because of limited system resources, an association between a new event and every other event in the system may not be kept. Thus, when searching for probable matches of a particular event, some events that are related to the particular event may not be initially selected. Such events may be associated with an event in a first set of events that are associated with the particular event. Therefore, a second set of events is selected that are associated with the first set of events.

Claims

1. A method of determining probable matches between a target event and a set of events, wherein an event is an image set associated with an object, comprising performing a machine-executed operation involving instructions, wherein the machine-executed operation is at least one of:

A) sending said instructions over transmission media;

B) receiving said instructions over transmission media;

C) storing said instructions onto a machine-readable storage medium; and

D) executing the instructions;

wherein said instructions are instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of;

generating a plurality of similarity scores based on comparisons between a feature set of the target event and feature sets of the set of events, wherein a feature set is the set of identifying information extracted from an image set;

generating a plurality of weighted similarity scores by weighting the similarity scores based on information external to the feature sets; and

based on the plurality of weighted similarity scores, determining a set of two or more events from the set of stored events that best match the target event.

2. The method of claim 1, wherein the information includes at least one of the following:

how long an event has been stored, date and time of an event, one or more cameras that captured an event, and physical location of the one or more cameras.

3. A method comprising performing a machine-executed operation involving instructions, wherein the machine-executed operation is at least one of:

A) sending said instructions over transmission media;

B) receiving said instructions over transmission media;

C) storing said instructions onto a machine-readable storage medium; and

D) executing the instructions;

wherein said instructions are instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of;

identifying a set of events captured in images from one or more cameras;

for each event in the set of events, generating a set of one or more similarity scores;

wherein each similarity score in the set of similarity scores that is associated with an event (a) is relative to another event, and (b) reflects how similar the event is to the other event;

determining probable matches to a target event by

determining that a first subset of events are probable matches to the target event based, at least in part, on a similarity between a feature set of target event and feature sets of events in said first subset; and

determining that a second subset of events are probable matches to the target event based on similarity scores between events in the first subset and events in the second subset.

4. The method of claim 3, wherein:

the instructions include instructions for storing features sets for only a subset of the events; and

when the second subset is determined, at least one of the events in the second subset does not have a stored feature set.

5. The method of claim 4, wherein the step of determining that a second subset of events are probable matches includes determining that a particular event is a probable match to the target event without comparing the feature set of the target event to any feature set of the particular event.

6. The method of claim 5, wherein, for each event in the set of events, the instructions include instructions for:

determining whether said each event is associated with a feature set; and

if said each event is not associated with a feature set, generating a new feature set for said each event and associating the new feature set with said each event.

7. The method of claim 3, wherein:

the instructions include instructions for associating a scored-event list with each event in the set of events; and

the scored-event list associated with each event indicates one or more other events and a similarity score between said each event and each event in said one or more other events.

8. The method of claim 7, wherein associating a scored-event list with each event in the set of events is performed by associating no more than a specified maximum number of events with each event.

9. The method of claim 7, wherein associating a scored-event list with each event in the set of events comprises including, in the scored-event list for a particular event, only other events that have similarity scores above a particular threshold relative to said particular event.

10. The method of claim 7, wherein, for each event in the set of events, the instructions include instructions for:

determining whether a preceding event in the set of events that precedes said each event is indicated by the scored-event list of said each event; and

if the preceding event is not indicated by the scored-event list of said each event;

generating a similarity score between the preceding event and said each event; and

updating the scored-event list of said each event to include an entry to the preceding event.

11. The method of claim 3, wherein the similarity score between each event in the first subset and the target event are above a first threshold and each similarity score between an event in the second subset and an event in the first subset are above a second threshold.

12. The method of claim 3, wherein:

a first threshold represents a first maximum number of events in the first subset of events that best match the target event according to similarity score; and

a second threshold represents a second maximum number of events in the second subset of events that best match an event in the first subset of events according to similarity score.

13. The method of claim 3, wherein each event in the set of events is associated with at least one of the following criteria:

the date and time of the event, information that indicates one or more cameras that captured the event, and physical location of the one or more cameras.

14. The method of claim 13, wherein each similarity score in the set of similarity scores is weighted based on the criteria.

15. The method of claim 3, wherein the step of determining that a first subset of events are probable matches to the target event includes:

inspecting a scored-event list associated with the target event to identify any matching events that were created before the target event; and

inspecting scored-event lists associated with events created after the target event to identify any matching events that were created after the target event.