@CONFERENCE\{IMM2005-03665,
    author       = "A. Morgenstjerne and B. Karlsen and J. Larsen and H. B. S{\o}rensen and K. B. Jakobsen",
    title        = "A comparative and combined study of {EMIS} and {GPR} detectors by the use of Independent Component Analysis",
    year         = "2005",
    keywords     = "Independent component analysis. Unsupervised detection, discrimination, and classification. {GPR} and {EMIS} detection of UXOs and landmines.",
    booktitle    = "Proceedings of the 2005 Detection and Remediation Technologies for Mines and Mine-Like Targets, AeroSense 2005",
    volume       = "",
    series       = "",
    editor       = "",
    publisher    = "SPIE",
    organization = "",
    address      = "Orlando, {FL}",
    url          = "http://www2.compute.dtu.dk/pubdb/pubs/3665-full.html",
    abstract     = "Independent Component Analysis (ICA) is applied to classify unexploded ordnance (UXO) on laboratory {UXO} 
test-field data, acquired by stand-off detection. The data are acquired by an Electromagnetic Induction Spectroscopy
(EMIS) metal detector and a ground penetrating radar (GPR) detector. The metal detector is a {GEM}-{3,} 
which is a monostatic sensor measuring the response of the environment on a multi-frequency constant wave
excitation field (300 Hz to 25 kHz), and the {GPR} detector is a stepped-frequency {GPR} with a monostatic bow-tie
antenna (500MHz to 2.5GHz). For both sensors the in-phase and the quadrature responses are measured at
each frequency. The test field is a box of soil where a wide range of UXOs are placed at selected positions. The
position and movement of both of the detectors are controlled by a {2D-}scanner. Thus the data are acquired
at well-defined measurement points. The data are processed by the use of statistical signal processing based
on {ICA}. An unsupervised method based on {ICA} to detect, discriminate, and classify the UXOs from clutter is
suggested. The approach is studied on {GPR} and {EMIS} data, separately and compared. The potential is an
improved ability: to detect the UXOs, to evaluate the related characteristics, and to reduce the number of false
alarms from harmless objects and clutter."
}