Hyperspectral Image Segmentation And Feature Extraction

In this paper, The authors have conducted a rapid non-destructive testing which used of computer image and pattern recognition technology, high spectral analysis techniques for rice seed vigor for rapid non-destructive testing.In order to study the technical feasibility of hyperspectral images of rice seed vigor for rapid non-destructive identification,the auther start this test with varying degrees of aging 22 total 5280 rice varieties of rice seeds as experimental material,and put the seeds in the box of aging to make aging process.Get three gradient of seed vigor in the environment of a temperature of 45 ℃,98% humidity and get seed aging treatment(0h, 36 h, 72h).Calculating simple vigor index after rice seed samples were artificially aged at 25℃, 20% light conditions germination test,germination rate and root length statistical,and dividing each sample into different species group dynamic gradient, using hyperspectral imaging technology,by extracting the average spectral reflectance of rice seeds, rice seeds of different vigor analyzed,Combining Savitzky-Golay(SG)smoothing algorithm, standard normal variable(SNV) and multiplicative scatter correction(MSC)spectral data 874 ~ 1 740 nm wavelength within the noise removal process,Using principal component analysis(PCA), continuous projection algorithm(SPA) for feature wavelength selection, based on full-band spectrum and the characteristic wavelengths were established based partial least squares discriminant analysis(PLS-DA) model.The results showed that after MSC pretreatment,the model which using PLS-DA model SPA selected characteristic wavelength established recognition accuracy modeling and prediction sets are set to reach more than 90 percent, as the best model. The results showed that the use of hyperspectral imaging technology used for rice seed vigor for rapid non-destructive testing is feasible.