Inter-trajectory anomaly detection using adaptive voting experts in a video surveillance system
View Patent ↗A sequence layer in a machine-learning engine configured to learn from the observations of a computer vision engine. In one embodiment, the machine-learning engine uses the voting experts to segment adaptive resonance theory (ART) network label sequences for different objects observed in a scene. The sequence layer may be configured to observe the ART label sequences and incrementally build, update, and trim, and reorganize an ngram trie for those label sequences. The sequence layer computes the entropies for the nodes in the ngram trie and determines a sliding window length and vote count parameters. Once determined, the sequence layer may segment newly observed sequences to estimate the primitive events observed in the scene as well as issue alerts for inter-sequence and intra-sequence anomalies.
1. A computer-implemented method for evaluating objects detected in a video stream, the method comprising:
detecting a first foreground object and a second foreground object in a single video stream, the first foreground object being different from the second foreground object;
while the first foreground object and the second foreground object are present in the single video stream:
building a first trajectory characterizing the first foreground object in a series of successive frames of the single video stream, and
building a second trajectory characterizing the second foreground object in the series of successive frames of the single video stream; and
determining an anomalous interaction between the first foreground object and the second foreground object based on observing the first trajectory relative to the second trajectory.
2. The method of claim 1 , further comprising generating an alert message indicating an occurrence of the anomalous interaction between the first foreground object and the second foreground object.
3. The method of claim 1 , wherein the first trajectory comprises a sequence of vectors storing kinematic data derived from observations of the first foreground object in the single video stream.
4. The method of claim 3 , wherein the sequence of vectors are mapped to nodes of a self organizing map (SOM) and wherein nodes of the SOM are clustered using an adaptive resonance theory (ART) network to generate a sequence of SOM nodes.
5. The method of claim 4 , wherein determining the anomalous interaction comprises determining a probability of observing the sequence of SOM nodes of the first foreground object in conjunction with a probability of observing a sequence of SOM nodes associated with the second trajectory.
6. The method of claim 1 , wherein the first foreground object and the second foreground object are classified as being instances of an object type sharing similar microfeature vectors characterizing the first foreground object and the second foreground object in the series of successive frames of the single video stream.
7. The method of claim 1 , wherein the determining of anomalous interaction is also based on prior observations of foreground objects in the single video stream.
8. A non-transitory computer storage medium, which, when executed on a processor, performs an operation for evaluating objects detected in a video stream, comprising:
detecting a first foreground object and a second foreground object in a single video stream, the first foreground object being different from the second foreground object;
while the first foreground object and the second foreground object are present in the single video stream:
building a first trajectory characterizing the first foreground object in a series of successive frames of the single video stream, and
building a second trajectory characterizing the second foreground object in the series of successive frames of the single video stream;
determining an anomalous interaction between the first foreground object and the second foreground object based on observing the first trajectory relative to the second trajectory.
9. The non-transitory computer storage medium of claim 8 , further comprising generating an alert message indicating an occurrence of the anomalous interaction between the first foreground object and the second foreground object.
10. The non-transitory computer storage medium of claim 8 , wherein the first trajectory comprises a sequence of vectors storing kinematic data derived from observations of the first foreground object in the single video stream.
11. The non-transitory computer storage medium of claim 10 , wherein the sequence of vectors are mapped to nodes of a self organizing map (SOM) and wherein nodes of the SOM are clustered using an adaptive resonance theory (ART) network to generate a sequence of SOM nodes.
12. The non-transitory computer storage medium of claim 11 , wherein determining the anomalous interaction comprises determining a probability of observing the sequence of SOM nodes of the first foreground object in conjunction with a probability of observing a sequence of SOM nodes associated with the second foreground object.
13. The non-transitory computer storage medium of claim 8 , wherein the second trajectory comprises a sequence of vectors storing kinematic data derived from observations of the second foreground object in the single video stream.
14. The non-transitory computer storage medium of claim 8 , wherein the determining of anomalous interaction is also based on prior observations of foreground objects in the single video stream.
15. A video surveillance system, comprising:
a video source configured to provide a single input video stream captured by a video camera;
a processor; and
a memory containing a program, which, when executed on the processor is configured to perform an operation for evaluating objects detected in the single input video stream, the operation comprising:
detecting a first foreground object and a second foreground object in the single input video stream, the first foreground object being different from the second foreground object,
while the first foreground object and the second foreground object are present in the single video stream:
building a first trajectory characterizing the first foreground object in a series of successive frames of the single video stream, and
building a second trajectory characterizing the second foreground object in the series of successive frames of the single video stream;
determining an anomalous interaction between the first foreground object and the second foreground object based on observing the first trajectory relative to the second trajectory.
16. The system of claim 15 , further comprising generating an alert message indicating an occurrence of the anomalous interaction between the first foreground object and the second foreground object.
17. The system of claim 15 , wherein the first trajectory comprises a sequence of vectors storing kinematic data derived from observations of the first foreground object in the single input video stream.
18. The system of claim 17 , wherein the sequence of vectors are mapped to nodes of a self organizing map (SOM) and wherein nodes of the SOM are clustered using an adaptive resonance theory (ART) network to generate a sequence of SOM nodes.
19. The system of claim 18 , wherein determining the anomalous interaction comprises determining a probability of observing the sequence of SOM nodes of the first foreground object in conjunction with a probability of observing a sequence of SOM nodes associated with the second foreground object.
20. The system of claim 15 , wherein the second trajectory comprises a sequence of vectors storing kinematic data derived from observations of the second foreground object in the single input video stream.
21. The system of claim 15 , wherein the determining of anomalous interaction is also based on prior observations of foreground objects in the single video stream.