Updating intensities in a PHD filter based on a sensor track ID

Patent metadata

Field	Value
Publication number	US-11175142-B2
Application number	US-201414448803-A
Country	US
Kind code	B2
Filing date	Jul 31, 2014
Priority date	Jul 31, 2014
Publication date	Nov 16, 2021
Grant date	Nov 16, 2021

Abstract

In one embodiment, a method of tracking multiple objects with a probabilistic hypothesis density filter is provided. The method includes obtaining measurements corresponding to a first object with at least one sensor, the at least one sensor providing one or more first track IDs for the measurements. A Tk+1 first predicted intensity is generated for the first object based on a Tk first track intensity. A Tk+1 measurement from a first sensor of the at least one sensors is obtained, the first sensor providing a second track ID for the Tk+1 measurement. The second track ID is compared to the one or more first track IDs, and the Tk+1 first predicted intensity is selectively updated with the Tk+1 measurement based on whether the second track ID matches any of the one or more first track IDs to generate a Tk+1 first measurement-to-track intensity for the first object.

First claim

What is claimed is:

1. A tracking method comprising:
tracking multiple objects in an environment around a vehicle with a probabilistic hypothesis density filter using one or more processing devices that are configured to perform a method comprising:
obtaining statistics of a measurement vector, based on one or more sensor measurements corresponding to a first object with at least one sensor onboard the vehicle;
obtaining, by the at least one sensor, a first track identifier (ID) for the one or more sensor measurements, wherein the first track ID is correlated across time to the one or more sensor measurements;
generating a T k first track intensity for the first object based on the statistics of the measurement vector, the T k first track intensity including a weight, a state mean vector, and a state covariance matrix of statistics of a track of the first object at time T k ;
generating a T k+1 first predicted intensity for the first object based on the T k first track intensity, the T k+1 first predicted intensity corresponding to time T k+1 ;
obtaining statistics of a T k+1 measurement vector, based on a T k+1 sensor measurement from a first sensor of the at least one sensor, the first sensor providing a second track ID for the T k+1 sensor measurement, wherein the T k+1 sensor measurement corresponds to time T k+1 ;
comparing the second track ID to the first track ID;
updating the T k+1 first predicted intensity with the statistics of the T k+1 measurement vector if the second track ID matches the first track ID to generate a T k+1 first measurement-to-track intensity for the first object at time T k+1 ;
generating a new intensity having values equal to the statistics of the T k+1 measurement vector if the second track ID does not match the first track ID;
generating a T k+1 second track intensity for a second object based on the new intensity, the T k+1 second track intensity corresponding to time T k+1 ;
generating a T k+2 second predicted intensity based on the T k+1 second track intensity, the T k+2 second predicted intensity corresponding to time T k+2 ;
associating the second track ID with the T k+2 second predicted intensity;
obtaining statistics of a T k+2 measurement vector, based on a T k+2 sensor measurement from the first sensor, the first sensor providing a third track ID for the T k+2 sensor measurement, wherein the T k+2 sensor measurement corresponds to time T k+2 ;
comparing the third track ID to the second track ID;
computing a statistical distance between the T k+2 second predicted intensity and the statistics of the T k+2 measurement vector; and
selectively updating the T k+2 second predicted intensity with the statistics of the T k+2 measurement vector based on whether the third track ID matches the second track ID and based on whether the statistical distance is less than a threshold to generate a T k+2 second measurement-to-track intensity for the second object at time T k+2 ;

wherein the probabilistic hypothesis density filter is operative to maintain a single intensity for each object being tracked.

2. The method of claim 1 , further comprising:
computing a statistical distance between the T k+1 first predicted intensity and the statistics of the T k+1 measurement vector; and
wherein updating includes updating the T k+1 first predicted intensity with the statistics of the T k+1 measurement vector if the statistical distance is less than a threshold.

3. The method of claim 2 , wherein:
if either the second track ID does not match the first track ID or the statistical distance is greater than the threshold, not updating the T k+1 first predicted intensity with the statistics of the T k+1 measurement vector.

4. The method of claim 1 , further comprising:
generating a T k+1 first track intensity based on the T k+1 first measurement-to-track intensity;
associating the first track ID with the T k+1 first track intensity;
generating a T k+2 first predicted intensity for the first object based on the T k+1 first track intensity, the T k+2 first predicted intensity corresponding to time T k+2 ;
obtaining statistics of a T k+2 measurement vector, based on a T k+2 sensor measurement from the first sensor, the first sensor providing a fourth track ID for the T k+2 sensor measurement, wherein the T k+2 sensor measurement corresponds to time T k+2 ;
comparing the fourth track ID to the first track ID; and
selectively updating the T k+2 first predicted intensity with the statistics of the T k+2 measurement vector based on whether the fourth track ID matches the first track ID to generate a T k+2 first measurement-to-track intensity for the first object at time T k+2 .

5. The method of claim 1 , wherein updating includes:
computing a statistical distance between the T k+1 first predicted intensity and the statistics of the T k+1 measurement vector;
if the second track ID matches the first track ID and the statistical distance is less than a threshold:
updating the T k+1 first predicted intensity with the statistics of the T k+1 measurement vector to generate a T k+1 first measurement-to-track intensity; and
associating only the second track ID with the T k+1 first measurement-to-track intensity;

pruning the T k+1 first predicted intensity such that the T k+1 first predicted intensity is no longer used for tracking the first object;
generating a T k+1 track intensity from the T k+1 first measurement-to-track intensity; and
associating only the second track ID with the T k+1 track intensity.

6. The method of claim 1 , wherein each track ID of the first track ID and the second track ID is one of an identifier assigned by a sensor of the at least one sensor to a set of measurements obtained by that sensor that are correlated across time, or an international civil aviation organization (ICAO) aircraft address received by a sensor of the at least one sensor and corresponding to a respective measurement for that sensor.

7. A tracking system comprising:
one or more processing devices onboard a vehicle;
a plurality of sensors onboard the vehicle and operatively coupled to the one or more processing devices; and
one or more data storage devices onboard the vehicle and including instructions which, when executed by the one or more processing devices, cause the one or more processing devices to track multiple objects in an environment around the vehicle with a probabilistic hypothesis density filter, wherein the instructions cause the one or more processing devices to:
obtain a plurality of sensor measurements corresponding to a first object with at least one sensor onboard the vehicle;
obtain, by the at least one sensor, a first track identifier (ID) for the plurality of sensor measurements, wherein the first track ID is correlated across time to the plurality of sensor measurements;
generate a T k first track intensity for the first object based on measurement statistics from the plurality of sensor measurements, the T k first track intensity including a weight, a state mean vector, and a state covariance matrix of statistics of a track of the first object at time T k ;
generate a T k+1 first predicted intensity for the first object based on the T k first track intensity, the T k+1 first predicted intensity corresponding to time T k+1 ;
obtain a T k+1 sensor measurement from a first sensor of the plurality of sensors, the first sensor providing a second track ID for the T k+1 sensor measurement, wherein the T k+1 sensor measurement corresponds to time T k+1 ;
compare the second track ID to the first track ID;
selectively update the T k+1 first predicted intensity with T k+1 measurement statistics from the T k+1 sensor measurement, based on whether the s

Assignees

• Honeywell Int Inc

Inventors

• Bageshwar Vibhor L
• Elgersma Michael Ray
• Euteneuer Eric A

Classifications

• G16C10/00 Primary

  Computational theoretical chemistry, i.e. ICT specially adapted for theoretical aspects of quantum chemistry, molecular mechanics, molecular dynamics or the like · CPC title

• G01D9/32

  there being a common recording element for two or more variables · CPC title

• G01D9/30

  there being a separate recording element for each variable, e.g. multiple-pen recorder · CPC title

• G01S13/933

  of aircraft or spacecraft · CPC title

• G01S13/867

  Combination of radar systems with cameras · CPC title