Feature-based tracking of moving objects

Patent metadata

Field	Value
Publication number	US-9823344-B2
Application number	US-201414563215-A
Country	US
Kind code	B2
Filing date	Dec 8, 2014
Priority date	Dec 8, 2014
Publication date	Nov 21, 2017
Grant date	Nov 21, 2017

Abstract

Systems and methods are provided for tracking moving objects. A sensor system is configured to measure a position of each of a plurality of moving objects as a set of measurements. A track assignment component is configured to accept the set of measurements from the sensor system and assign a subset of the measurements to respective tracks. The tracks represents the motion of respective objects. The track assignment component represents a cost calculation component configured to provide a cost matrix from at least a change point detection analysis of a feature associated with the set of measurements and an object assignment component configured to assign the subset of the measurements to respective tracks. A filtering component is configured to update respective associated states of tracks representing the plurality of moving objects according to the set of measurements and the assignment of the subset of the measurements to respective tracks.

First claim

Having described the invention, we claim:

1. A system for tracking moving objects comprising:
a sensor system configured to measure a position of each of a plurality of moving objects as a set of measurements;
a track assignment component configured to accept the set of measurements from the sensor system and assign a subset of the measurements to respective tracks, a given track representing the motion of an object of the plurality of moving objects, the track assignment component comprising:
a cost calculation component configured to provide a cost matrix from at least a change point detection analysis of a feature associated with the set of measurements, the feature associated with the set of measurements being modelled as a non-exponential distribution; and
an object assignment component configured to assign the subset of the measurements to respective tracks; and

a filtering component configured to update respective associated states of tracks representing the plurality of moving objects according to the set of measurements and the assignment of the subset of the measurements to respective tracks.

2. The system of claim 1 , wherein the change point detection analysis utilizes Bayesian online change point detection.

3. The system of claim 1 , wherein the filtering component comprises a Kalman filter.

4. The system of claim 1 , wherein the filtering component comprises a non-linear filter.

5. The system of claim 1 , wherein the feature associated with the set of measurements is a radar cross section.

6. The system of claim 1 , wherein the feature associated with the set of measurements is modelled as a Rice distribution.

7. The system of claim 1 , wherein an underlying process model of the change point detection algorithm is evaluated using variational Bayesian inference.

8. The system of claim 1 , wherein the track assignment component assigns the subset of the measurements to respective tracks according to an N-dimensional data assignment process such that the track assignment jointly finds a maximum a posteriori probability estimate of the association vectors over a sliding window of the last N−1 time steps, where N is a positive integer greater than one.

9. The system of claim 8 , wherein the track assignment component assigns the subset of the measurements to respective tracks according to a 3-dimensional data assignment process such that the track assignment jointly finds a maximum a posteriori probability estimate of the association vectors over a sliding window of the last two time steps.

10. A method for tracking moving objects comprising:
receiving, at a radar tracking processor, a first set of measurements from a radar system representing a plurality of moving objects, each measurement having an associated location and a value for at least one feature;
determining a set of likelihood values, representing expected values for a second set of measurements, from the first set of measurements via a change point detection algorithm;
receiving a second set of measurements representing the plurality of moving objects;
determining a lowest cost assignment of the second set of measurements to respective tracks representing the plurality of moving objects from the set of likelihood values according to an N-dimensional data assignment process such that the lowest cost assignment jointly finds a maximum a posteriori probability estimate of association vectors over a sliding window of the last N−1 time steps, where N is a positive integer greater than one; and
updating a state associated with the tracks according to assigned measurement values.

11. The method of claim 10 , wherein determining the set of likelihood values from the first set of measurements via the change point detection algorithm comprises determining the set of likelihood values from the first set of measurements via a Bayesian online change point detection algorithm.

12. The method of claim 11 , wherein an underlying process model of the Bayesian online change point detection algorithm is evaluated using variational Bayesian inference.

13. The method of claim 10 , wherein an underlying process model of the at least one feature associated with the first set of measurements is not part of an exponential family of models with a conjugate prior.

14. The method of claim 10 , where the at least one feature includes a radar cross-section.

15. The method of claim 10 , wherein determining the lowest cost assignment of the second set of measurements to respective tracks representing the plurality of moving objects from the set of likelihood values comprises determining the lowest cost assignment according to an N-dimensional data assignment process such that the track assignment jointly finds a maximum a posteriori probability estimate of the association vectors over a sliding window of the last N−1 time steps, where N is a positive integer greater than one.

16. A non-transitory computer readable medium, storing executable instructions readable by an associated processor to track at least respective positions of a plurality of moving objects, the executable instructions comprising:
a track assignment component configured to receive a set of measurements representing at least a position and a radar cross-section of each of the plurality of moving objects and assign a subset of the measurements to respective tracks, a given track representing the motion of an object of the plurality of moving objects, the track assignment component comprising:
a cost calculation component configured to provide a cost matrix from at least a Bayesian online change point detection analysis of the radar cross-section associated with the set of measurements, an underlying process model of the change point detection algorithm being evaluated using variational Bayesian inference; and
an object assignment component configured to assign the subset of the measurements to respective tracks; and

a filtering component configured to update respective associated states of tracks representing the plurality of moving objects according to the set of measurements and the assignment of the subset of the measurements to respective tracks.

17. The non-transitory computer readable medium of claim 16 , wherein the track assignment component assigns the subset of the measurements to respective tracks according to an N-dimensional data assignment process such that the track assignment jointly finds a maximum a posteriori probability estimate of the association vectors over a sliding window of the last N−1 time steps, where N is a positive integer greater than one.

18. The non-transitory computer readable medium of claim 16 , wherein the filtering component comprises a Kalman filter.

Assignees

• Turner Ryan D
• Bottone Steven
• Stanek Clay J
• Northrop Grumman Systems Corp

Inventors

• Turner Ryan D
• Bottone Steven
• Stanek Clay J

Classifications

• G01S13/726 Primary

  Multiple target tracking · CPC title