Unsupervised learning of events in a video sequence
First Claim
1. A method of learning events contained within a video sequence, the method comprising the steps of:
- providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence; and
analyzing the feature vector to determine a number of possible event candidates.
1 Assignment
0 Petitions
Accused Products
Abstract
Methods and systems for the unsupervised learning of events contained within a video sequence, including apparatus and interfaces for implementing such systems and methods, are disclosed. An illustrative method in accordance with an exemplary embodiment of the present invention may include the steps of providing a behavioral analysis engine, initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object located within an image sequence, and then analyzing the feature vector to determine a number of possible event candidates. The behavioral analysis engine can be configured to prompt the user to confirm whether an event candidate is a new event, an existing event, or an outlier. Once trained, a testing/operational phase mode of the behavioral analysis engine can be further implemented to detect the occurrence of one or more learned events, if desired.
-
Citations
53 Claims
-
1. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence; and
analyzing the feature vector to determine a number of possible event candidates. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19)
-
-
20. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence, the behavioral analysis engine including an actor feature database, a frame feature database, and a programmable event library;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence;
analyzing the feature vector to determine a number of possible event candidates;
prompting a user to confirm whether a detected event candidate is a new event; and
storing the new event within the event library if the detected event candidate is confirmed by the user. - View Dependent Claims (21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36)
-
-
37. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence; and
employing a clustering routine to analyze the feature vector to determine a number of possible event candidates.
-
-
38. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence;
applying a time-consistency filtering routine to the image sequence; and
analyzing the feature vector to determine a number of possible event candidates.
-
-
39. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a testing phase mode within the behavioral analysis engine; and
detecting the occurrence of one or more events learned by the behavioral analysis engine. - View Dependent Claims (40, 41)
-
-
42. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a testing phase mode within the behavioral analysis engine; and
detecting the occurrence of one or more learned events selected by a user.
-
-
43. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a testing phase mode within the behavioral analysis engine;
detecting the occurrence of one or more events learned by the behavioral analysis engine;
determining the probability that an event has occurred; and
outputting a response to a user if the probability that the event has occurred is greater than a confidence threshold value.
-
-
44. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence;
analyzing the feature vector to determine a number of possible event candidates;
prompting a user to confirm whether a detected event candidate is a new event;
storing the new event within an event library if the detected event candidate is confirmed by the user;
initiating a testing phase mode within the behavioral analysis engine;
comparing features contained within the feature vector against one or more learned events contained within the event library;
determining the probability that an event has occurred; and
outputting a response to the user if the probability that the event has occurred is greater than a confidence threshold value.
-
-
45. A method of learning events contained within a video sequence, the method comprising the steps of:
-
providing a behavioral analysis engine adapted to learn new events contained within an image sequence;
initiating a training phase mode within the behavioral analysis engine and obtaining a feature vector including one or more parameters relating to an object disposed within the image sequence;
analyzing the feature vector to determine a number of possible event candidates; and
wherein said step of analyzing the feature vector to determine a number of possible event candidates is performed automatically using a clustering routine employed by the behavioral analysis engine.
-
-
46. A video surveillance system, comprising:
-
at least one image sensor adapted to produce an image sequence within a field of view; and
a behavioral analysis engine adapted to learn new events contained within the image sequence. - View Dependent Claims (47, 48, 49, 50, 51, 52)
-
-
53. A video surveillance system, including:
-
at least one image sensor adapted to produce an image sequence within a field of view;
a computer configured to run a behavioral analysis engine for automatically learning new events contained within the image sequence; and
a graphical user interface.
-
Specification