Data-driven delta-generalized labeled multi-bernoulli tracker

1. A method of tracking a plurality of objects, comprising:

receiving unlabeled measurement data identifying a plurality of targets corresponding to the plurality of objects;

generating a multi-target likelihood function using a persistent target density, a birth target density, and a clutter density;

using the multi-target likelihood function to identify persistent targets in the plurality of targets in the measurement data to associate the persistent targets with persistent tracks for the plurality of objects to update the persistent tracks and to identify birth targets in the plurality of targets in the measurement data that are not associated with the persistent tracks to initiate a new track for each birth target.

2. The method of claim 1 , wherein each measurement in the unlabeled measurement data that is not associated with a persistent target or clutter generates a birth target track in a first time step and further comprising making each birth target in the first time step into a persistent target for a second time step.

3. The method of claim 1 , wherein the persistent target density and the multi-target likelihood function are for a first time step, and further comprising:

generating a joint posterior density from the persistent target density, the birth target density, and the multi-target likelihood function;

performing a time-update on the joint posterior density to generate a predicted density for a second time step; and

changing variables to make the predicted density for the second time step into a persistent target density for the second time step.

4. The method of claim 3 , wherein generating the joint posterior density comprises generating the joint posterior density through the Bayes update.

5. The method of claim 3 , wherein performing the time-update on the joint posterior density comprises propagating the joint posterior density forward in time according to a multi-target Chapman-Kolmogorov equation.

6. The method of claim 1 further comprising displaying the persistent tracks and the new tracks.

7. The method of claim 1 , wherein the plurality of objects are selected from a plurality of objects in space, a plurality of objects on the ground, a plurality of objects in the air, or a plurality of objects under water.

8. The method of claim 1 , further comprising:

receiving sensor data, wherein the sensor data comprises video images of the plurality of objects in a scene; and

identifying the images of the plurality of objects in the sensor data as the targets to generate the measurement data.

9. An apparatus, comprising:

a data processing system configured to receive unlabeled measurement data identifying a plurality of targets corresponding to a plurality of objects; and

a data-driven delta-GLMB tracker implemented in the data processing system and configured to:

generate a multi-target likelihood function using a persistent target density, a birth target density, and a clutter density;

use the multi-target likelihood function to identify persistent targets in the plurality of targets in the measurement data to associate the persistent targets with persistent tracks for the plurality of objects to update the persistent tracks and to identify birth targets in the plurality of targets in the measurement data that are not associated with the persistent tracks to initiate a new track for each birth target.

10. The apparatus of claim 9 , wherein each measurement in the unlabeled measurement data that is not associated with a persistent target or clutter generates a birth target track in a first time step and wherein the data-driven delta-GLMB tracker is further configured to make each birth target in the first time step into a persistent target for a second time step.

11. The apparatus of claim 9 , wherein the persistent target density and the multi-target likelihood function are for a first time step, and wherein the data-driven delta-GLMB tracker is further configured to:

generate a joint posterior density from the persistent target density, the birth target density, and the multi-target likelihood function;

perform a time-update on the joint posterior density to generate a predicted density for a second time step; and

change variables to make the predicted density for the second time step into a persistent target density for the second time step.

12. The apparatus of claim 11 , wherein the data-driven delta-GLMB tracker is configured to generate the joint posterior density through the Bayes update.

13. The apparatus of claim 11 , wherein the data-driven delta-GLMB tracker is configured to perform the time-update on the joint posterior density by propagating the joint posterior density forward in time according to a multi-target Chapman-Kolmogorov equation.

14. The apparatus of claim 9 further comprising a display generator configured to generate a track display for displaying the persistent tracks and the new tracks.

15. The apparatus of claim 9 , wherein the plurality of objects are selected from a plurality of objects in space, a plurality of objects on the ground, a plurality of objects in the air, or a plurality of objects under water.

16. The apparatus of claim 9 , wherein:

the receiver is configured to receiving sensor data, wherein the sensor data comprises video images of the plurality of objects in a scene; and further comprising

a target identifier configured to identify the images of the plurality of objects in the sensor data as the targets to generate the measurement data.

17. A method of identifying birth targets for tracking a plurality of objects, comprising:

receiving measurement data identifying a plurality of targets corresponding to the plurality of objects in a first time step;

generating a multi-target likelihood function using a persistent target density for the first time step, a birth target density, and a clutter density;

using the multi-target likelihood function to identify the birth targets for the first time step in the plurality of targets in the received measurement data, wherein the birth targets are targets that are not associated with already identified persistent tracks;

generating a joint posterior density from the persistent target density for the first time step, the birth target density, and the multi-target likelihood function;

performing a time-update on the joint posterior density to generate a predicted density for a second time step; and

changing variables to make the predicted density for the second time step into a persistent target density for the second time step.

18. The method of claim 17 , further comprising:

using the multi-target likelihood function to identify persistent targets in the plurality of targets in the measurement data to associate the persistent targets with persistent tracks for the plurality of objects to update the persistent tracks and to initiate a new track for each birth target in the plurality of targets in the measurement data that is not associated with a persistent track.

19. The method of claim 17 , wherein generating the joint posterior density comprises generating the joint posterior density through the Bayes update.

20. The method of claim 17 , wherein performing the time-update on the joint posterior density comprises propagating the joint posterior density forward in time according to a multi-target Chapman-Kolmogorov equation.