The fluorescence of oil films with different thicknesses and mixing proportions was measured and classified using an improved co-training algorithm, which improves the classification accuracy and provides prospects for recognition of similar oil species.

The purpose of polarization calibration is to obtain the response matrix of an instrument such that the subsequent observation data can be corrected. The calibration precision, however, is partially restricted by the noise of the detector. We investigate the precision of the normalized response matrix in the presence of signal-independent additive noise or signal-dependent Poisson shot noise. The influences of the source intensity, type of noise, and calibration configuration on the precision are analyzed. We compare the theoretical model and the experimental measurements of the polarization calibration to show that the relative difference between the two is less than 16%. From this result, we can use the model to determine the minimum source intensity and choose the optimal configurations that provide the required precision.