Method for identifying frostbite condition of grain seeds using spectral feature wavebands of seed embryo hyperspectral images

A method for identifying frostbite condition of a grain seed using spectral feature waveband of grain seed embryo hyperspectral images, comprising the following steps:

• 1) collecting the hyperspectral image in an embryo side of the grain seed, which composed of gray images of different wavebands;
• 2) obtaining the hyperspectral image in the embryo region of the grain seed by processing the collected hyperspectral image;
• 3) obtaining an average value of each of the gray image in an embryo region of the grain seed, forming an average spectra with different wavebands, eliminating the wavebands with noise in the average spectra, and using a successive projection algorithm to extract the feature wavebands and obtain a corresponding spectral value of the average spectra;
• 4) dividing the grain seeds with known frostbite category into multiple batches, repeating each of the above steps 1) to 2) to obtain the feature wavebands and the spectral values of the feature wavebands of the grain seeds, and then inputting the feature wavebands and the spectral values of the feature wavebands of the grain seeds into a classification model for obtaining an optimal training classification model;
• 5) repeating the above steps 1) to 2) to obtain the feature wavebands and the spectral values of the grain seeds with unknown category, and then inputting the feature wavebands and the spectral values of the feature wavebands of the grain seeds into the optimal trained classification model to obtain the frostbite category of the grain seed.