Research About The Nondestructive Testing Of Potato External Defect Based On Hyperspectral Imaging Technology And Multi-information Fusion

This papar makes potato external defect as research subjects,and it combines with Hyperspectral Imaging technology and Multi-imformation Fusion technology to build the nondestructive testing model for the external defect of potato. The main research contents and conclusions of this paper are as follows:(1) The interest region of different kinds of potato simple image and average spectrum data were withdrew and analyzed, which provide researchs with a theoretical basis for post-treatment and analysis.(2) Feature images were extracted from hyperspectral images of potato. The mean values of both the images on three-primary colors wavelengths and the 5 images on the neighboring wavelengths were calculated ,and then 3 images which had been calculated the mean value were superimposed, which obtained the color feature images of potato;The 5 feature bands spectra images were optimized by Principal Component Analysis(PCA),and they were used to carry out the second PCA. PC2 images of the second PCA were made as the gray feature images of potato.(3)The color images of potato were filtered by the Average Value Filter Method.Comparing De-noising effect with Recursive of Least Squares (RLS) and Wavelet,Wavelet Compressed Sensing and Recursive of Least Squares(WTCS-RLS) ,the result showed that De-noising effect with WTCS-RLS was obvious. The average values and standard deviations of components of color images under the HIS color space are extracted,they were made as the eigenvalues of color images; 6 texture features of gray images were made as the eigenvalues of gray images.BP-ANN and Bayesian classifier were applied to set up the nondestructive testing models of potato external defect based on hyperspectral image date. The result showed that the optimization method was Bayesian classifier.(4) Derivative,SNV,S-QMSC and methods of spectrum data composite pretreatment were applied to preprocess the average spectrum of potato.The result shows that he optimization method is SNV+SD.The spectral features were extracted by PCA.BP-ANN and Bayesian classifier were applied to set up the nondestructive testing models of potato external defect based on hyperspectral spectrum data. The result shows that the optimization method is Bayesian classifier.(5)The information of potato external defec are fused on the feature level. BP-ANN, Bayesian classifier,SVM and Adaboost were applied to set up the nondestructive measurement testing models of potato external defect,and chose the optimization method-SVM.