Abstract:

In order to improve the inspection efficiency， reduce the cost of inspection and identification， and realize the rapid and nondestructive classification of sole， a Fourier?transform infrared fingerprint spectrum and a spectral derivation were adopted to analyze the soles with five different brands among the totally 50 samples， and two classification models， Bayes discriminant （BDA） and support vector machine （SVM）， were constructed. The experimental results indicated that the fusion model based on the original data， the first derivative data， and the second derivative data was superior to the single model and the intermediate fusion model in the process of sole discrimination， which resulted in an overall classification accuracy of more than 80 %. Based on the primary model， the combination of BDA with the original first derivative model led to an overall classification accuracy of 92 %. Different fusion models can be constructed by a combination of the infrared fingerprint spectrum with the first order derivation and the second order derivation to distinguish and contrast. The rapid non?destructive identification of daily leather shoes and sneakers soles can be realized by choosing the most effective fusion model， which provides a new way for the rapid detection of cases.

Key words: sole material, spectral fusion technology, Bayes discriminant analysis, support vector machine analysis