Data Modeling, Feature Extraction, and Classification of Magnetic and EMI Data, ESTCP Discrimination Study, Camp Sibert, AL. Demonstration Report

摘要

An ESTCP-sponsored demonstration was conducted at the Former Camp Sibert, Alabama, where the objective was to discriminate potentially hazardous 4.2' mortars from non-hazardous shrapnel, range and cultural debris. Nine different discrimination techniques were tested and used data from the MTADS magnetometer array, a Geonics EM61 cart, the MTADS EM61 array and the Geonics EM63. Discrimination was achieved using either rule-based or statistical classification of feature vectors extracted from dipole or polarization tensor model fits to detected anomalies. Each of the EM methods utilized a size and time-decay based feature vector. All nine different methods were successful at discriminating the 4.2' mortars from non-hazardous times. The EM based methods were more effective than the magnetometer. The higher SNR, denser coverage and more precise positioning of the MTADS EM61 array data resulted in fewer false- positives than the EM61 cart. When depth constraints from the magnetometer data were used to constrain the EM fits through cooperative inversion, discrimination performance was improved. The most effective discrimination technique used the cooperatively inverted fits to the EM63 cued interrogation data. All UXO were recovered before a single false-positive was excavated.