2.5.3. Visualization
In consideration of the analytical speed in practical applications,
the developed models based on a few selected wavebands were
preferable to the models developed on full-range wavebands.
Therefore, the best PLS model performing on selected important
wavebands was employed for moisture content mapping. In the
mapping procedure, hyperspectral images at the feature
wavebands were first extracted and constructed into a multispectral
data-cube (x, y, kselected). Subsequently, the data-cube was
unfolded into a two-dimensional matrix (x y, kselected). When
applying the model, the vector (x y) was firstly multiplied by
the vector of regression coefficients extracted from the established
PLS model, and then re-folded into a color map where the moisture
content at each pixel was represented by different color scales.
Therefore, moisture content distribution can be visualized clearly
on the color map.