Peach variety detection using VIS-NIR spectroscopy and deep learning

摘要: Abstract More than 1000 peach varieties with significant differences in qualities are cultivated China. Distinguishing is not only needed by dealers, but also demanded products processing enterprises and consumers. Detecting the of fruit one most important analyses. Over past decades, there many linear or nonlinear methods have been proposed, such as principal component analysis, partial least squares, support vector machine etc. However, these traditional commonly need some preprocessing steps including denoising, baseline correction, wavelength selection so on. Hence, it requires users to enough skilled knowledge before they can establish a good performance detection model. To offer information on identifying varieties, visual near-infrared (VIS-NIR) diffuse reflectance spectra between 350 820 nm were collected for five 100 samples each variety, the total samples consisted 500 peaches. Spectral has shown rapid non-destructive measurement ability research fields. In recent years, spectral was researched detecting sugar firmness fruits, examples, apple, peach, tomato Some using require application specific transforms, expertly designed constraints model parameters, limited due their maintenance costs. With development learning technology deep which plays an role different projects won eyes fields from both academy industry. order classify analyzing VIS-NIR spectra, method based principle proposed this paper. paper, realize multi-identification constructing dimensional convolution neural network establishing database containing kinds The obtained through training, then employed predict testing set data. accuracy models reached 100% validation dataset 94.4% test dataset. This study indicated that could be distinguished successfully spectroscopy learning.