Improved Analysis Algorithms for UXO Filler Identification
Abstract
This project addressed the Statement of Need Number UXSON-04-02 for Innovative Technology for Identification of Filler Material in Recovered UXO. SAIC and Duke University investigated several advanced data analysis algorithms and techniques to apply to Pulsed Elemental Analysis with Neutrons (PELAN) data. The data was collected on shells filled with inert and explosive materials, and chemical simulants. The goal of this investigation was to improve the performance, reliability, and robustness of the PELAN decision-making process and to make it easier to train the PELAN system using target libraries. These studies have provided considerable improvements in performance over the previous methods used to analyze the PELAN spectra and the decision-making process. The focus of this effort concentrated on unexploded ordnance (UXO) items. Both the Least Squares/Generalized Likelihood Ratio Test (LS/GLRT) and the Least Squares/Principal Component Analysis (LSIPCA) combinations showed significant improvement over the LS/decision tree approach. As a result, the LS/GLRT method was implemented into the portable PELAN unit. We are continuing investigation into the PCA spectral analysis method, which shows even more improvement over the LS/GLRT approach. The PCA algorithm was shown to be effective at using the entire spectrum to extract characteristics of the target for improved identification.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 2009
- Accession Number
- ADA520708
Entities
People
- Robert Sullivan
Organizations
- Leidos