Detecting anomalous trajectories in a video surveillance system
View Patent ↗Techniques are disclosed for determining anomalous trajectories of objects tracked over a sequence of video frames. In one embodiment, a symbol trajectory may be derived from observing an object moving through a scene. The symbol trajectory represents semantic concepts extracted from the trajectory of the object. Whether the symbol trajectory is anomalous may be determined, based on previously observed symbol trajectories. A user may be alerted upon determining that the symbol trajectory is anomalous.
1. A computer-implemented method for analyzing a sequence of video frames depicting a scene captured by a video camera, the method comprising:
passing each of a plurality of kinematic data vectors derived from analyzing a foreground object detected in the sequence of video frames to a plurality of adaptive resonance theory (ART) networks, each ART network modeling a subset of kinematic data parsed from the kinematic data vectors;
identifying, in each ART network, a cluster to which each respective subset of kinematic data parsed from the plurality of kinematic data vectors is mapped;
generating a primitive trajectory based on the clusters identified in each of the ART networks; and
generating a symbol trajectory from the primitive trajectory.
2. The computer-implemented method of claim 1 , wherein the primitive trajectory includes a sequence of labels assigned to the clusters in a respective ART network, wherein the sequence corresponds to an order in which the plurality of kinematic data vectors are passed to the respective ART network.
3. The computer-implemented method of claim 1 , wherein the symbol trajectory is generated using latent semantic analysis (LSA) and singular value decomposition (SVD) applied to the primitive trajectory.
4. The computer-implemented method of claim 1 , further comprising:
computing a probability of observing the symbol trajectory in the scene.
5. The computer-implemented method of claim 4 , wherein the probability of observing the symbol trajectory in the scene is computed based on a plurality of symbol trajectories previously observed in the scene.
6. The computer-implemented method of claim 4 , wherein the probability of observing the symbol trajectory in the scene is computed using at least one of a Markov model, a hidden Markov model, and probabilistic latent semantic analysis.
7. The computer-implemented method of claim 4 , further comprising:
upon determining that a probability of observing the symbol trajectory falls below a threshold value, alerting a user upon of the video surveillance system of an occurrence of an anomalous trajectory.
8. A computer-readable storage medium containing a program which, when executed by a processor, performs an operation for analyzing a sequence of video frames depicting a scene captured by a video camera, the operation comprising:
passing each of a plurality of kinematic data vectors derived from analyzing a foreground object detected in the sequence of video frames to a plurality of adaptive resonance theory (ART) networks, each ART network modeling a subset of kinematic data parsed from the kinematic data vectors;
identifying, in each ART network, a cluster to which each respective subset of kinematic data parsed from the plurality of kinematic data vectors is mapped;
generating a primitive trajectory based on the clusters identified in each of the ART networks; and
generating a symbol trajectory from the primitive trajectory.
9. The computer-readable storage medium of claim 8 , wherein the primitive trajectory includes a sequence of labels assigned to the clusters in a respective ART network, wherein the sequence corresponds to an order in which the plurality of kinematic data vectors are passed to the respective ART network.
10. The computer-readable storage medium of claim 8 , wherein the symbol trajectory is generated using latent semantic analysis (LSA) and singular value decomposition (SVD) applied to the primitive trajectory.
11. The computer-readable storage medium of claim 8 , wherein the operation further comprises:
computing a probability of observing the symbol trajectory in the scene.
12. The computer-readable storage medium of claim 11 , wherein the probability of observing the symbol trajectory in the scene is computed based on a plurality of symbol trajectories previously observed in the scene.
13. The computer-readable storage medium of claim 11 , wherein the probability of observing the symbol trajectory in the scene is computed using at least one of a Markov model, a hidden Markov model, and probabilistic latent semantic analysis.
14. The computer-readable storage medium of claim 11 , wherein the operation further comprises:
upon determining that a probability of observing the symbol trajectory falls below a threshold value, alerting a user upon of the video surveillance system of an occurrence of an anomalous trajectory.
15. A system, comprising:
a video input source configured to provide a sequence of video frames, each depicting a scene;
a processor; and
a memory containing a program, which, when executed on the processor is configured to perform an operation for analyzing the scene, as depicted by the sequence of video frames captured by the video input source, the operation comprising:
passing each of a plurality of kinematic data vectors derived from analyzing a foreground object detected in a sequence of video frames to a plurality of adaptive resonance theory (ART) networks, each ART network modeling a subset of kinematic data parsed from the kinematic data vectors,
identifying, in each ART network, a cluster to which each respective subset of kinematic data parsed from the plurality of kinematic data vectors is mapped,
generating a primitive trajectory based on the clusters identified in each of the ART networks, and
generating a symbol trajectory from the primitive trajectory.
16. The system of claim 15 , wherein the primitive trajectory includes a sequence of labels assigned to the clusters in a respective ART network, wherein the sequence corresponds to an order in which the plurality of kinematic data vectors are passed to the respective ART network.
17. The system of claim 15 , wherein the symbol trajectory is generated using latent semantic analysis (LSA) and singular value decomposition (SVD) applied to the primitive trajectory.
18. The system of claim 15 , further comprising:
computing a probability of observing the symbol trajectory in the scene.
19. The system of claim 18 , wherein the probability of observing the symbol trajectory in the scene is computed based on a plurality of symbol trajectories previously observed in the scene.
20. The system of claim 18 , wherein the probability of observing the symbol trajectory in the scene is computed using at least one of a Markov model, a hidden Markov model, and probabilistic latent semantic analysis.
21. The system of claim 18 , further comprising:
upon determining that a probability of observing the symbol trajectory falls below a threshold value, alerting a user upon of the video surveillance system of an occurrence of an anomalous trajectory.