Weapon fire detection and localization algorithm for electro-optical sensors

Patent metadata

Field	Value
Publication number	US-10389928-B2
Application number	US-201615234184-A
Country	US
Kind code	B2
Filing date	Aug 11, 2016
Priority date	Aug 11, 2016
Publication date	Aug 20, 2019
Grant date	Aug 20, 2019

Abstract

A method is disclosed for detecting and locating a blast, including muzzle flash, created by the launch of a projectile from a gun barrel, rocket tube or similar device, generally associated with weapons fire. The method is used in conjunction with electro-optical imaging sensors and provides the azimuth and elevation from the detecting sensor to the launch location of the blast and also provides the weapon classification.

First claim

What is claimed is:

1. A method of hostile fire detection and localization based on a weapons fire detection imaging sensor system having detection sensors to provide a respective video output of the detection sensor as a sensor output, and a processor to process said sensor outputs and compute hostile fire detection, the method comprising the steps of:
processing by the processor said provided respective video output for each detection sensor to process a method to determine detection in the respective detection sensor, said method to determine detection in the respective detection sensor comprising the steps of:
reading an output video frame of the provided video output and preprocessing said output video frame to differentiate signal from background,
tracking the signal as differentiated over time,
determining the location of peak signal intensity by analyzing one or more signal consistently differentiated,
determining the time at which the signal starts and the time at which it stops by analyzing the location of peak signal per differentiated signal,
extracting temporal profile of signal at peak spatial location and temporal profile of signal in neighborhood of peak spatial location per differentiated signal,
measuring duration and shape features of said temporal profiles,
comparing said measurements against known target measurements, and
associating multiple signals and declaring detection in a selected sensor output;

comparing detections from multiple sensor outputs;
classifying signal against known threat signatures, wherein said classifying signal against known signatures comprises the steps of:
for each detection, classify the detection as either weapons fire or false alarm based on extracted signature features;
if the detection is classified as false alarm, then updating a false alarm database to reflect the recently determined false alarm; and
if the detection is classified as weapon fire, then reclassifying the detection to determine the class of weapon fire, said reclassifying the detection to determine the class of weapon comprising the steps of:
analyzing a class containing known target signatures of a weapons fire class and a false alarm class derived from a database of false alarms to determine which features provide the greatest distinction between two classes,
calculate the distance between the detection and the weapons fire class based on the features which provide the greatest distinction between the two classes,
comparing the calculated distance to a threshold,
if the distance threshold is exceeded, then detection is declared a false alarm, and the false alarm class is updated with the new false alarm, and
if the calculated distance is within the distance threshold, the detection is declared as weapon fire;

declaring as a weapon fire event the associated image sensor detections which have been classified as a weapon fire; and
outputting azimuth location, elevation location, weapon classification, and time of firing for each declared weapon fire event, wherein, the distinction, or distance, between the two classes, is called a distance metric, wherein a feature distance is calculated using the following formula, where μ i and σ i are the mean and standard deviation value of a given feature for the i th class and abs is the absolute value:

Feature
⁢

⁢
Distance

=

abs
⁡

(

μ
1

-

μ
2

)

1
/
2

⁢

(

σ
1
2

+

σ
2
2

)

.

2. The method of hostile fire detection and localization according to claim 1 , wherein determining detection in the respective detection sensor includes independently determining a weapons fire detection per video output of the respective detection sensor as an imaging sensor detection, wherein a single weapons fire event may generate imaging sensor detections on multiple sensors.

3. The method of hostile fire detection and localization according to claim 1 , wherein comparing detections from multiple sensor outputs includes the steps of:
analyzing imaging sensor detections across multiple sensors to determine if they were generated by a single hostile fire event, wherein detection locations, detection event time, and characteristics of detected signatures are used to determine whether multiple detections were generated by the single hostile fire event; and
if the multiple imaging sensor detections are found to have been generated by a single hostile fire event, a single hostile fire detection is created.

4. The method of hostile fire detection and localization according to claim 1 wherein classifying signal against known threat signatures includes the steps of:
calculating features from temporal and intensity profiles of a hostile fire detection to classify the hostile fire detection as either of a weapon fire event or a false alarm; and
subdividing a weapon fire event into at least anti-tank guided missile, recoilless rifle, and rocket propelled grenade weapon fire events.

5. The method of classifying a hostile fire detection and localization according to claim 1 , wherein said extracting temporal profile of signal at peak spatial location and temporal profile of signal in neighborhood of peak spatial location per differentiated signal includes the steps of:
extracting the signal between the measured start and stop times; and
extracting a pixel-based signal from pixels surrounding the pixel in which the peak signal was found, using the measured start and stop times.

6. The method of classifying a hostile fire detection and localization according to claim 1 , wherein said measuring duration and shape features of said temporal profile includes analyzing the extracted signals for duration and shape characteristics.

7. The method of classifying a hostile fire detection and localization according to claim 1 , wherein said comparing said measurements against known target measurements includes comparing the duration and shape characteristics of the extracted signals against known target characteristics.

8. The method of classifying a hostile fire detection and localization to claim 1 , wherein said associating multiple signals and declaring detection in a selected sensor output includes the steps of:
if multiple signals are found to have characteristics corresponding to known targets, analyzing those signals to

Assignees

• Us Army

Inventors

• Brown Jeremy B
• Hutchison Iii John E
• Davis Jami H
• Gabonia Joshua K

Classifications

• G01S3/78 Primary

  using electromagnetic waves other than radio waves · CPC title

• H04N23/45

  for generating image signals from two or more image sensors being of different type or operating in different modes, e.g. with a CMOS sensor for moving images in combination with a charge-coupled device [CCD] for still images · CPC title

• G06V20/00

  Scenes; Scene-specific elements (control of digital cameras H04N23/60) · CPC title

• G06T7/73

  using feature-based methods · CPC title

• G06T2207/10048

  Infrared image · CPC title