Intelligent Classification of Orange Growing Areas by Using Near-Infrared Spectra

Authors

  • Xia Jiang School of Engineering, University of Guelph
  • Yifan Cai School of Engineering, University of Guelph
  • Simon Yang School of Engineering, University of Guelph
  • Gauri Mittal School of Engineering, University of Guelph

Abstract

Near-Infrared spectroscopy (NIR) is a fast and non-destructive method to identify orange growingareas. In this paper, a principal component analysis (PCA) approach was used to obtain the featuresof orange NIR spectra by reducing the divisions in the analysis. An artificial neural network (ANN)was developed to achieve enhanced classification accuracy, while a support vector machine (SVM)model was proposed for higher classification accuracy. A hybrid genetic algorithm (GA) SVMmodel was designed, with the most valuable data from the PCA selecting by GA. The simulationresults showed that the hybrid GA-SVM classifier achieved the best accuracy of 89.717%.

Downloads

Issue

Section

World Conference on Computers in Agriculture, San Jose, Costa Rica, 2014