Gaussian Mixture Model-Based Walnut Shell and Meat Classification in Hyperspectral Fluorescence Imagery

Classifying the shells and meat of walnuts is necessary when harvesting them. During the last decade, hyperspectral imaging techniques have been widely used in agriculture for quality inspection. This article demonstrates that hyperspectral fluorescence imaging is capable of analyzing the difference between walnut shells and meat, and proposes a principal component analysis and Gaussian mixture model (PCA-GMM)-based Bayesian classifier to discriminate between the shell and the meat. PCA was first used to extract features and reduce the redundancy of the input data. The optimal number of components in PCA classification was selected by a cross-validation technique. Then the PCA-GMM-based Bayesian classifier was applied to differentiate the walnut shell and meat according to the class-conditional probability and the prior estimated by the Gaussian mixture model. Furthermore, a cross-validation method was used to evaluate robustness of the proposed classification method. Finally, the PCA-GMM and GMM methods were compared. The experimental results showed the effectiveness of the proposed approach in the application of walnut shell and meat classification, and an overall 95.6% recognition rate was achieved.