Identification of automobile transmission fluid using hyperspectral imaging technology

Abstract: An identification method combining sparse representation with principal component analysis (PCA) was proposed for discriminating varieties of transmission fluid of automobile by using hyperspectral imaging technology. Principal component analysis was applied to obtain the characteristic information in the 874-1 733 nm spectra. For each transmission fluid variety, 80 samples were randomly selected as the training set, and 20 samples as the testing set. The eigenvectors of all training samples form the matrix were used for the sparse representation, and the problem of transmission fluid types classification was transformed into one to solve a sample expressed by the overall training sample matrix through optimization under the 11 norm. The results demonstrate that the accuracy of the algorithm that was composed of sparse representation and principal component analysis (PCA) was 93%. The accuracy is higher than those of PCA-LDA (Linear Discriminant Analysis) and PCA- LS-SVM (Least Squares Support Vector Machine). Therefore, the proposed method provides a better approach for the identification of transmission fluid types.