Prediction model from spectral intensity to spectral polarization based on polarization spectral bidirectional reflectance distribution function

To obtain detailed multi-angle polarization spectral data in a short time is a challenging task. In this paper, a prediction model from spectral intensity to spectral polarization based on polarization spectral bidirectional reflectance distribution function (PSBRDF) is proposed. The model mainly consists of three steps: dataset preparation, eigenvalue extraction and prediction model construction, which transforms the prediction of spectral intensity into the eigenvalues. The eigenvalues predicted can recover the spectral intensity and spectral polarization based on PSBRDF. The comparison of prediction results with traditional methods shows that the method proposed slightly sacrifices accuracy when predicting spectral intensity under known detection angle conditions, but demonstrates better generalization ability and adaptability under unknown detection angle conditions. Moreover, the model can predict the spectral polarization, which traditional methods can not achieve, and control the <i>MRE</i> to be below 0.03 and <i>R²</i> to be above 0.99. The model provides a prediction-based solution for the situation where the polarization spectral data of the target is difficult to obtain.