Automated discrimination between walnut shell and pulp has become an important task in the walnut postharvest processing industry in the US. During the past several years, hyperspectral fluorescence imaging has been widely used to inspect agricultural products for quality and safety due to its full spectral information and ability of identifying different chemical components in the subject.
Lu Jiang of the Bio-imaging and Machine Vision Lab, at The Fischell Department of Bioengineering, University of Maryland, has studied the feasibility of analyzing the difference of black walnut shell and pulp with this technique. A Gaussian-kernel-based support vector machine (SVM) approach was used to classify the walnut shell and pulp. Experiment results showed an overall 90.3% recognition rate and that hyperspectral fluorescence imaging and the proposed classification method were effective in differentiating walnut shell and pulp.