Near Infrared Spectrum Based Color Detection Modeling in Printing

Abstract

It exits nonlinear relationship between near infrared spectrum and print color. We could make use of diffuse spectrometric to get the near infrared spectra of print sample. The traditional method is using the partial least squares (PLS) to establish relation mathematical model, but the partial least squares (PLS) has the problem of low accuracy and bigger training sample size. The least squares support vector machine (LS-SVM) is presented in the paper to establish prediction model of print color. The result shows that the LS-SVM model has higher accuracy than PLS model.