Object behavior anomaly detection using neural networks
AssigneeNVIDIA Corporation
InventorsMilind Naphade, Shuo Wang
Life of this patent
Estimating the term…
Abstract
In various examples, a set of object trajectories may be determined based at least in part on sensor data representative of a field of view of a sensor. The set of object trajectories may be applied to a long short-term memory (LSTM) network to train the LSTM network. An expected object trajectory for an object in the field of view of the sensor may be computed by the LSTM network based at least in part an observed object trajectory. By comparing the observed object trajectory to the expected object trajectory, a determination may be made that the observed object trajectory is indicative of an anomaly.
Claims20
A method comprising:
- determining, based at least in part on sensor data representative of a field of view of a sensor, a set of object trajectories;
- applying the set of object trajectories to a long short-term memory (LSTM) network to train the LSTM network;
- computing, by the LSTM network and based at least in part on one or more observed object trajectories, an expected object trajectory for an object in the field of view of the sensor;
- and
- determining that an object trajectory is indicative of an anomaly based at least in part on a comparison of the object trajectory to the expected object trajectory.
A method comprising:
- determining an object trajectory representative of a first series of positions of an object in an environment;
- applying the object trajectory into an inference component that uses a machine learning model to identify a deviation of the object trajectory from an expected object trajectory, the deviation identified based at least in part on differences between the first series of positions and a second series of positions representing the expected object trajectory;
- determining that the object trajectory is indicative of an anomaly based at least in part on the deviation;
- and
- generating a message indicative of the anomaly.
A method comprising:
- obtaining an object trajectory, including a series of positions of an object in an environment and corresponding timestamps;
- inputting the object trajectory into an inference module that uses a normal object behavior model to identify an extent of deviation of the object trajectory from normal object behavior, the normal object behavior model trained over a time period to adapt to object behavior observed in the environment over the time period;
- and
- based on the extent of deviation of the object trajectory from normal object behavior, determine the object trajectory is an anomaly compared to the normal object behavior.
Dependencies are inferred from the claim text.