Fused EO/IR detection & tracking of surface targets: Flight demonstrations

At Fusion 2007 we presented real-time methods for fused video imaging, detection and tracking of multiple ground targets, and showed results of tower-based ground tests. This paper summarizes the use of these methods in proof of concept flight tests, undertaken as a collaborative effort among severa...

Full description

Saved in:
Bibliographic Details
Main Authors: Waxman, A., Fay, D., Ilardi, P., Arambel, P., Xinzhuo Shen, Krant, J., Moore, T., Gorin, B., Tilden, S., Baron, B., Lobowiecki, J., Jelavic, R., Verbiar, C., Antoniades, J., Baumback, M., Hass, D., Edwards, J., Henderson, S., Chester, D.
Format: Conference Proceeding
Language:English
Subjects:
Collaboration
Gas detectors
Hyperspectral imaging
Hyperspectral sensors
image fusion
Image sensors
Infrared detectors
multi-target tracking
neural networks
Optical imaging
Sensor fusion
sensor resource management
System testing
target detection
Target tracking
Video tracking
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:At Fusion 2007 we presented real-time methods for fused video imaging, detection and tracking of multiple ground targets, and showed results of tower-based ground tests. This paper summarizes the use of these methods in proof of concept flight tests, undertaken as a collaborative effort among several business units of BAE Systems North America. Utilizing a Sonoma commercial RGB/MWIR 4-band turret mounted on a King Air 90, we integrated our existing image co/geo-registration methods with the image fusion, target learning & detection, multi-target tracking, and sensor/turret resource management methods reported previously, implemented them on real-time commercial processors, and conducted two test flights over water and land. We illustrate results of these flight tests that demonstrate proof of concept capabilities, showing examples of air-to-surface detection and tracking of multiple vessels over an extended field of regard, ground vehicles (both stationary and moving), and multiple dismounts walking on a beach. The approach does not require stabilization of the background in order to detect moving targets because detection is all spectral-based. Targets can be detected as spectral anomalies with respect to an adaptive local background, and then learned on-the-fly as a target class. Alternatively, targets can be designated by an operator and learned in real-time. This approach applies to both day & night fused video imaging over land & water, and can be extended to include other imaging sensors such as SWIR & LWIR, and coupled with hyperspectral fingerprinting for feature-aided tracking.