Comparative Study of Different Pre-processing Methods for Discriminating Live Fish Based on Hyperspectral Imagery

摘要

The main aim of this study was to compare the performance of different pre-processing algorithms when coupled with Support vector machine as the classifier to differentiate live fish based on their diet received during cultivation using hyperspectral imagery system. Rainbow trout (Oncorhynchus mykiss) were fed either a fish meal-based diet or a 100% plant-based diet. Hyperspectral images were made using a push-broom hyperspectral imaging system in the spectral region of 394-1009 nm. Six spectral pre-treatment methods were used, including Savitzky-Golay, First Derivative, Second Derivative, Standard Normal Variate and Multiplicative Scatter Correction were employed to improve the robustness and performance of the classifier. According to the criteria of correct classification rate and Kappa coefficient, the support vector machine with linear kernel when coupled with Savitzky-Golay pre-processing was determined as the best method for classifying live fish due to their diet.