Optimization of wavelength sweep for time-domain spectral matching by sequential quadratic programming

This paper proposes a method of optimizing the wavelength sweep in the spectral matching imager (SMI) by sequential quadratic programming (SQP). The proposed method determines a weight function that appears in the spectral correlation integral by minimizing a cost functional composed of the squared correlation coefficient between two reference spectra, while properly incorporating equality and inequality constraints. A norm term is newly added to the cost functinoal, which increases the reliability of the solution. Experimental results confirm that the obtained wavelength sweep function, which is derived from the optimized weight function and tends to be nonlinear to time, enables the SMI to distinguish two objects with very similar spectra more easily than in the case of time-linear wavelength sweep.