New adaptive methods for sensing of chemical components and biological agents

It is known that leaf reflectance spectra can be used to estimate the contents of chemical components in vegetation. Recent novel applications include the detection of harmful biological agents that can originate from agricultural bioterrorism attacks. Such attacks have been identified as a major threat to the United States’ agriculture. Nevertheless, the usefulness of such approach is currently limited by distorting factors, in particular soil reflectance. The quantitative analysis of the spectral curves from the reflection of plant leaves may be the basis for the development of new methods for interpreting the data obtained by the remote measurement of plants. We consider the problem of characterizing the chemical composition from noisy spectral data using an experimental optical method. Using our experience in signal processing and optimization of complex systems we propose a new mathematical model for sensing of chemical components in vegetation. Estimates are defined as minimizers of penalized cost functionals with sequential quadratic programming (SQR) methods. A deviation measure used in risk analysis is also considered. This framework is demonstrated for different agricultural plants using adaptive filtration, principal components analysis, and optimization techniques for classification of spectral curves of chemical components. Various estimation problems will be considered to illustrate the computational aspects of the proposed method.