Almond Origin Identification Based on Raman Spectroscopy and Portable Convolutional Neural Network
The taste and nutritional value of almonds from different producing areas are different,so the price difference is obvious.So,it is the key factor to promote the development of the almond industry how to realize almond origin identification.The Raman spectra of almonds were measured by Raman spectrometer,and the baseline of Raman spectra was corrected by LabSpec software.Then,we analyzed 7 kinds of almonds by post-processing Raman spectroscopy combined with 1D P-CNN.The model 1D P-CNN we constructed includes convolution layer conv1,pooling layer pool1,convolution layer conv2,pooling layer pool2,and full connection layers fc1 and fc2.The accuracy,recall,and F value of our model can reach 99.3%.To verify the robustness of the classifier,four different classifiers,namely,logistic regression model(LR),random forest model(RF),Bayesian network model(NB),and P-CNN,were used to identify the Raman spectra of almonds in different noise scenarios.In the test,it was found that the accuracy of the P-CNN model was higher than that of the other three models under different signal-to-noise ratios.It provides a more accurate and effective analytical method for the identification of the almond origin.
almondraman spectraP-CNNidentification of almond origin
万方数据
维普
