METHOD AND APPARATUS FOR THE TRACKING OF MULTIPLE OBJECTS

• US 20140324339A1
• Filed: 04/30/2014
• Published: 10/30/2014
• Est. Priority Date: 04/30/2013
• Status: Active Grant

• Current Assignee: Baselabs GmbH
• Original Assignee: Baselabs GmbH
• Inventors: Adam, Christian, Mattern, Norman, Richter, Eric, Schubert, Robin
• Granted Patent: US 8,954,272 B2
• Time in Patent Office: Days
• US Class Current: 701/519
• CPC Class Codes: G01S 13/726 Multiple target tracking

Abstract

Tracking systems using multi-object-tracking (MOT) combine data from a plurality of sensors (101) to derive a state estimate of one or more objects (103) are described. The interfaces to the sensors (101) are such that different ones of the sensors (101) can be used. The tracking (102) in the system consists of different blocks for state estimation (104), object'"'"'s existence estimation (106), and data association (105). To utilize the data of a sensor, the sensor data and the system which is observed is modeled using the Bayesian filtering framework. All interfaces of the system, i.e. the interfaces between the sensors (101) and the MOT system as well as of the blocks in the MOT are of a characteristic that ensures that only compatible models are used with each other already when designing the system

Claims

1. A system for tracking one or more movable objects comprising:

one or more sensors recording object measurement data and passing the recorded object measurement data to a memory;

a measurement space located in the memory for storing a plurality of measurement vectors representative of the recorded object measurement data, wherein the measurement space has a plurality of dimensions, the dimensions having a dimension identifier;

a state space located in the memory for storing a plurality of state vectors representative of the state of the object, wherein the state space has a plurality of dimensions, the dimensions having a dimension identifier;

a measurement noise space located in the memory for storing a plurality of measurement noise vectors representative of uncertainties in the object measurement data;

a system model having a system deterministic part and a system probabilistic part and being stored in the memory, the system model being representative of the one or more movable objects;

a measurement model having a measurement deterministic part and a measurement probabilistic part and being stored in memory; and

a processor configured to update the system model based on a previous system model and using at least an updated one of the measurement vectors, the measurement noise vectors or the state vectors.

2. The system of claim 1, wherein the processor comprises a plurality of Bayes filter implementations to update the system model.

3. The system of claim 1, wherein the system model comprises a plurality of tracks and the processor is configured to associate one or more of the object measurement data with one or more of the plurality of tracks and to assign probability that the assigned one of the measurement data is correctly assigned to a correct one of the plurality of tracks.

4. The system of claim 1, further comprising a gating configured to review the recorded object measurement data and pass to the measurement space only those object measurement data falling within a threshold level.

5. A method for tracking one or more movable objects comprising:

recording object measurement data of the one or more movable objects;

creating in memory measurement vectors representative of the recorded object measurement data;

creating in memory state vectors representative of the state of the object;

creating in memory measurement noise vectors representative of uncertainties in the object measurement data;

accessing a system model having a system deterministic part and a system probabilistic part and a measurement model having a measurement deterministic part and a measurement probabilistic part; and

updating the system model using at least an updated one of the measurement vectors, the measurement noise vectors or the state vectors.

6. The method of claim 5, further comprising eliminating all of the object measurement data falling outside of a threshold value.

7. The method of claim 5, further comprising associated at least one or more of the object measurement data to one or more tracks in the system model and calculating the probability that the assigned one of the object measurement data is correctly assigned to the correct one of the tracks.

8. A vehicle assistance system comprising system for tracking one or more movable objects in the neighbourhood of a vehicle comprising:

one or more sensors recording object measurement data and passing the recorded object measurement data to a memory;

a measurement space located in the memory for storing a plurality of measurement vectors representative of the recorded object measurement data, wherein the measurement space has a plurality of dimensions, the dimensions having a dimension identifier;

a state space located in the memory for storing a plurality of state vectors representative of the state of the object, wherein the state space has a plurality of dimensions, the dimensions having a dimension identifier;

a measurement noise space located in the memory for storing a plurality of measurement noise vectors representative of uncertainties in the object measurement data;

a system model having a system deterministic part and a system probabilistic part and being stored in the memory, the system model being representative of the one or more movable objects;

a measurement model having a measurement deterministic part and a measurement probabilistic part and being stored in memory; and

a processor configured to update the system model based on a previous system model and using at least an updated one of the measurement vectors, the measurement noise vectors or the state vectors.