The invention discloses an LED-based low-cost hyperspectral detection
system, comprising: a
light source module, a light splitting module, a sampling module and a
data processing module. The light splitting module sorts each LED according to a peak
wavelength, and sequentially turns on the
LED lamp separately to complete a light splitting process of the hyperspectral detection
system; the drivingcurrent of the LED is adjusted to change the spectral distribution and the peak
wavelength of a
light beam, and current modulation is applied to the LED
light source, thereby improving the light splitting precision; the
data processing module uniformly divides a hyperspectral
image cube into sub-cubes, sequentially arranges an unfolded two-dimensional matrix of each sub-cube in a sub-cube group onthe direction of column, merges the two-dimensional matrices into a two-dimensional matrix, and achieves the low-rank constraint of the hyperspectral
image cube on the spectrum and the space throughthe rank of the two-dimensional matrix for removing all kinds of different
noise; and a spatial-spectral dictionary of sparse representation is learned by using a
dictionary learning algorithm according to the above sub-cubes, and the dictionary is used for improving the
spectral resolution of a hyperspectral data cube.