Performance of Supervised Classification Algorithms of Hyperspectral Imagery for Identifying Fecal and Ingesta Contaminants

Abstract

In order to select the optimum classifier, the performance of six different supervised classification algorithms including parallelepiped, minimum distance, Mahalanobis distance, maximum likelihood, spectral angle mapper, and binary coding were investigated and compared for identifying fecal and ingesta contaminants in hyperspectral poultry imagery. A pushbroom line-scan hyperspectral imager was used for hyperspectral image acquisition, with 512 narrow bands between 400 and 900 nm wavelengths. Three different feces from digestive tracts (duodenum, ceca, colon) and ingesta were considered as contaminants. These contaminants were collected from broiler carcasses that had been fed a mixture of corn, milo, or wheat with soybean. The classification accuracies varied from 62.9% to 92.3% depending on the classification method. The highest classification accuracy for identifying contaminants from corn-fed carcasses was 92.3% with a spectral angle mapper classifier. The accuracy was 82.1% with the maximum likelihood method for milo-fed carcasses, and 91.2% accuracy was obtained for wheat-fed carcasses when the same classification method was employed. The mean classification accuracy for classifying fecal and ingesta contaminants obtained in this study was 90.3%.