Behavioral recognition system
View Patent ↗Embodiments of the present invention provide a method and a system for analyzing and learning behavior based on an acquired stream of video frames. Objects depicted in the stream are determined based on an analysis of the video frames. Each object may have a corresponding search model used to track an object's motion frame-to-frame. Classes of the objects are determined and semantic representations of the objects are generated. The semantic representations are used to determine objects' behaviors and to learn about behaviors occurring in an environment depicted by the acquired video streams. This way, the system learns rapidly and in real-time normal and abnormal behaviors for any environment by analyzing movements or activities or absence of such in the environment and identifies and predicts abnormal and suspicious behavior based on what has been learned.
1. A method for processing video image data, the method comprising:
identifying, by operation of one or more processors, one or more objects depicted in a sequence of video frames capturing a scene;
generating, from the sequence of video frames, a plurality of information streams characterizing one or more of the identified objects; and
generating, from the plurality of information streams, by a machine learning engine, one or more object classifications to assign to objects depicted in the sequence of video frames, wherein the machine learning engine is configured to derive patterns of behavior engaged in by objects having a common object classification and to identify instances of the patterns of behavior engaged in by objects depicted in the sequence of video frames as said instances occur.
2. The method of claim 1 , wherein the plurality of information streams includes a stream characterizing kinematics of the at least one object, as depicted in the sequence of video frames.
3. The method of claim 1 , wherein the plurality of information streams includes a stream characterizing an appearance of the at least one object, as depicted in the sequence of video frames.
4. The method of claim 1 , wherein the plurality of information streams includes a stream specifying a feature based classification of the at least one object based on kinematics data and appearance data derived for the at least one object derived from the sequence of video frames.
5. The method of claim 1 , wherein identifying one or more objects depicted in a sequence of video frames comprises:
generating a background model of the scene; and
segmenting one or more foreground objects depicted in each of the sequence of video frames based on the generated background model.
6. The method of claim 5 , further comprising, tracking one or more of the segmented foreground object across a plurality of the sequence of video frames, and wherein one of the plurality of information streams includes a tracked trajectory of one or more of the segmented objects.
7. The method of claim 1 , further comprising issuing at least one alert indicating an occurrence of one of the identified patterns of behavior engaged in by one of the identified objects.
8. The method of claim 1 , wherein the machine learning engine is further configured to determine, from the plurality of information streams, whether an occurrence of one of the identified patterns of behavior is normal or anomalous event, relative to prior analysis of the indentified objects depicted in the sequence of video frames, as represented by of the plurality of information streams generated from the sequence of video frames.
9. A non-transitory computer-readable storage medium containing a program, which, when executed on a processor is configured to perform an operation, comprising:
identifying, by operation the processor, one or more objects depicted in a sequence of video frames capturing a scene;
generating, from the sequence of video frames, a plurality of information streams characterizing one or more of the identified objects;
generating, from the plurality of information streams, by a machine learning engine, one or more object classifications to assign to objects depicted in the sequence of video frames, wherein the machine learning engine is configured to derive patterns of behavior engaged in by objects having a common object classification and to identify instances of the patterns of behavior engaged in by objects depicted in the sequence of video frames as said instances occur.
10. The computer-readable storage medium of claim 9 , wherein the plurality of information streams includes a stream characterizing kinematics of the at least one object, as depicted in the sequence of video frames.
11. The computer-readable storage medium of claim 9 , wherein the plurality of information streams includes a stream characterizing an appearance of the at least one object, as depicted in the sequence of video frames.
12. The computer-readable storage medium of claim 9 , wherein the plurality of information streams includes a stream specifying a feature based classification of the at least one object based on kinematics data and appearance data derived for the at least one object derived from the sequence of video frames.
13. The computer-readable storage medium of claim 9 , wherein identifying one or more objects depicted in a sequence of video frames comprises:
generating a background model of the scene; and
segmenting one or more foreground objects depicted in each of the sequence of video frames based on the generated background model.
14. The computer-readable storage medium of claim 13 , wherein the operation further comprises, tracking one or more of the segmented foreground object across a plurality of the sequence of video frames, and wherein one of the plurality of information streams includes a tracked trajectory of one or more of the segmented objects.
15. The computer-readable storage medium of claim 9 , wherein the operation further comprises issuing at least one alert indicating an occurrence of one of the identified patterns of behavior engaged in by one of the identified objects.
16. The computer-readable storage medium of claim 9 , wherein the machine learning engine is further configured to determine, from the plurality of information streams, whether an occurrence of one of the identified patterns of behavior is normal or anomalous event, relative to prior analysis of the indentified objects depicted in the sequence of video frames, as represented by of the plurality of information streams generated from the sequence of video frames.
17. A system, comprising:
a video input source;
a processor; and
a memory storing:
a computer vision engine, wherein the computer vision engine is configured to:
identify, by operation of one or more processors, one or more objects depicted in a sequence of video frames capturing a scene, and
generate, from the sequence of video frames, a plurality of information streams characterizing one or more of the identified objects; and
a machine learning engine, wherein the machine learning engine is configured to:
generate, from the plurality of information streams, one or more object classifications to assign to objects depicted in the sequence of video frames, wherein the machine learning engine is configured to derive patterns of behavior engaged in by objects having a common object classification and to identify instances of the patterns of behavior engaged in by objects depicted in the sequence of video frames as said instances occur.
18. The system of claim 17 , wherein the plurality of information streams includes a stream characterizing kinematics of the at least one object, as depicted in the sequence of video frames.
19. The system of claim 17 , wherein the plurality of information streams includes a stream characterizing an appearance of the at least one object, as depicted in the sequence of video frames.
20. The system of claim 17 , wherein the plurality of information streams includes a stream specifying a feature based classification of the at least one object based on kinematics data and appearance data derived for the at least one object derived from the sequence of video frames.
21. The system of claim 17 , wherein identifying one or more objects depicted in a sequence of video frames comprises:
generating a background model of the scene; and
segmenting one or more foreground objects depicted in each of the sequence of video frames based on the generated background model.
22. The system of claim 21 , wherein the computer vision engine is further configured to track one or more of the segmented foreground object across a plurality of the sequence of video frames, and wherein one of the plurality of information streams includes a tracked trajectory of one or more of the segmented objects.
23. The system of claim 17 , wherein the machine learning engine is further configured to issue at least one alert indicating an occurrence of one of the identified patterns of behavior engaged in by one of the identified objects.
24. The system of claim 17 , wherein the machine learning engine is further configured to determine, from the plurality of information streams, whether an occurrence of one of the identified patterns of behavior is normal or anomalous event, relative to prior analysis of the indentified objects depicted in the sequence of video frames, as represented by of the plurality of information streams generated from the sequence of video frames.