Camp Beale Live-Site UXO Data Inversion and Classification Using Advanced EMI Models

• Main Authors: Shubitidze, Fridon, Shamatava, Irma, Miller, Jon, Keranen, Joe, Bijamov, Alex, Barrowes, Benjamin
• Other Authors: DARTMOUGH COLLEGE HANOVER NH THAYER SCHOOL OF ENGINEERING
• Format: Text
• Language: English
• Published: 2011
• Subjects: Magnetic & Electric Fld Detection & Detectors; Ammunition and Explosives; *EXPLOSIVES DETECTION; *UNEXPLODED AMMUNITION; EIGENVALUES; ELECTROMAGNETIC INTERFERENCE; MAGNETIC ANOMALY DETECTION; SYMPOSIA; BRIEFING CHARTS; ONVMS(ORTHONORMALIZED VOLUME MAGNETIC SOURCE); GAUSSIAN MIXTURE ALGORITHM; JD(JOINT DIAGONALIZATION) ALGORITHM; UXO(UNEXPLODED ORDNANCE); CAMP BEALE(CALIFORNIA); UXO CLASSIFICATION; Beale; Ida; Cold Regions Research and Engineering Laboratory
• Online Access: http://www.dtic.mil/docs/citations/ADA554384; http://oai.dtic.mil/oai/oai?&verb=getRecord&metadataPrefix=html&identifier=ADA554384

The advanced EMI and statistical classification models are applied to the cued data sets of the Metal Mapper and two next-generation portable sensors: MPV and 2x2 3D TEMTADS. The advanced models combine: (1) the joint diagonalization (JD) algorithm for estimating the number of potential anomalies from the measured data without inversion, (2) the orthonormalized volume magnetic source (ONVMS) model for representing the EMI responses and extracting the intrinsic parameters (feature vector) of the targets, and (3) the Gaussian Mixture algorithm that utilizes the extracted features to classify buried objects as targets of interest (TOI) or not. The inversion and classification schemes of these advanced models consist of the following steps: (i) build the multi-static-response (MSR) data matrix by combining the Tx and Rx data points of the advanced sensors; (ii) apply the JD to the MSR data matrix to determine its eigenvalues; (iii) estimate the data quality and the number of potential targets, based on the eigenvalues; (iv) study the temporal decay of the eigenvalues to identify the signal to noise ratio (SNR); (v) invert all data sets using the ONVMS-Differential Evolution algorithm; (vi) apply the semi-supervised GM clustering algorithm to the inverted total ONVMS to determine the clusters of anomalies; (vii) select anomalies from each cluster to build a custom training list (viii) request the ground truth for the selected targets; (ix) use the obtained ground truth to score the unknown targets using the GM weights for the ONVMS clusters; and (x) submit the final dig-list to the ESTCP office for independent scoring. In this presentation the data inversion processing and discrimination schemes of the advanced EMI models will be reviewed, and the classification results scored by the Institute for Defense Analyses (IDA) will be presented for Camp Beale, CA cued data sets of both MM and portable EMI sensors. Prepared in collaboration with the US Army Cold Regions Research and Engineering Laboratory (CRREL) and Sky Research. Presented at the Partners in Environmental Technology Technical Symposium & Workshop, Washington, DC, 29 Nov-1 Dec 2011. Sponsored by SERDP and ESTCP. U.S. Government or Federal Rights License