Walnut shell and meat differentiation using fluorescence hyperspectral imagery with ICA-kNN optimal wavelength selection

In recent years, fluorescence hyperspectral imaging technology has received more attention in the food inspection field because of its ability to distinguish multiple chemical ingredients. However, one persistent problem associated with this technology is how to deal with the huge amount of data acquired by the hyperspectral imaging device, and hence improve the efficiency of the application system. To solve the aforementioned problem, this paper proposed using an Independent Component Analysis with k Nearest Neighbor classifier (ICA-kNN) approach to differentiate walnut shells from walnut meat. This project will meet two goals: (1) select optimal wavelengths to reduce data redundancy, and (2) classify walnut shells or meat according to selected wavelengths. First, the ICA ranking method was applied to select 4–10 optimal wavelengths based on their ability to distinguish the four walnut shell and meat categories: dark meat, light meat, inner shell and outer shell. Then, the kNN classifier was used to distinguish walnut shells from meat according to the computed optimal wavelengths. A total of 5,496 samples were studied, and an overall 90.6% detection rate was achieved given 10 optimal wavelengths, which constituted only 13% of the total acquired hyperspectral image data. In order to further evaluate the proposed method, the classification results of the ICA-kNN approach were also compared to the kNN classifier method alone. The experimental results showed that the ICA-kNN method with a few wavelengths had the same performance as the kNN classifier alone using information from all 79 wavelengths. This demonstrated the effectiveness of the proposed ICA-kNN method to classify walnut shells and meat.