科研之友
微信
新浪微博
Facebook
分享链接
摘要
Data fusion is the process of combining data gathered from two or more sensors to produce a more specific, comprehensive and unified dataset of the inspected target. On this basis, much has been said about the possible benefits resulting from the use of molecular and atomic information for the detection of explosives. The orthogonal nature of the spectral and compositional information provided by Raman spectroscopy and laser-induced breakdown spectroscopy (LIBS) makes them suitable candidates for an optimal combination of their data, thus achieving inferences that are not feasible using a single sensor. The present manuscript evaluates several architectures for the combination of spectral outputs from these two sensors in order to compare the benefits and drawbacks of data fusion for improving the overall identification performance. From the simple assembling (concatenation or addition) of Raman and LISS spectra to signals' processing on the basis of linear algebra (either the outer product or the outer sum), different identification patterns of several compounds (explosives, potential confusants and supports) have been built. The efficiency on target differentiation by using each of the architectures has been evaluated by comparing the identification yield obtained for all the inspected targets from correlation and similarity measurements. Additionally, a specific code integrated by several of these patterns to identify each compound has also been evaluated. This approach permits to obtain a better knowledge about the identity of an interrogated target, mainly in those decisive cases in which LISS or Raman cannot be effective separately to reach a decision.
- 出版日期2015-3-1
